1 // SPDX-License-Identifier: GPL-2.0
3 * linux/fs/ext4/super.c
5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI)
12 * linux/fs/minix/inode.c
14 * Copyright (C) 1991, 1992 Linus Torvalds
16 * Big-endian to little-endian byte-swapping/bitmaps by
17 * David S. Miller (davem@caip.rutgers.edu), 1995
20 #include <linux/module.h>
21 #include <linux/string.h>
23 #include <linux/time.h>
24 #include <linux/vmalloc.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/backing-dev.h>
29 #include <linux/parser.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/dax.h>
42 #include <linux/cleancache.h>
43 #include <linux/uaccess.h>
44 #include <linux/iversion.h>
45 #include <linux/unicode.h>
46 #include <linux/part_stat.h>
47 #include <linux/kthread.h>
48 #include <linux/freezer.h>
51 #include "ext4_extents.h" /* Needed for trace points definition */
52 #include "ext4_jbd2.h"
58 #define CREATE_TRACE_POINTS
59 #include <trace/events/ext4.h>
61 static struct ext4_lazy_init *ext4_li_info;
62 static struct mutex ext4_li_mtx;
63 static struct ratelimit_state ext4_mount_msg_ratelimit;
65 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
66 unsigned long journal_devnum);
67 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
68 static int ext4_commit_super(struct super_block *sb, int sync);
69 static int ext4_mark_recovery_complete(struct super_block *sb,
70 struct ext4_super_block *es);
71 static int ext4_clear_journal_err(struct super_block *sb,
72 struct ext4_super_block *es);
73 static int ext4_sync_fs(struct super_block *sb, int wait);
74 static int ext4_remount(struct super_block *sb, int *flags, char *data);
75 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
76 static int ext4_unfreeze(struct super_block *sb);
77 static int ext4_freeze(struct super_block *sb);
78 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
80 static inline int ext2_feature_set_ok(struct super_block *sb);
81 static inline int ext3_feature_set_ok(struct super_block *sb);
82 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
83 static void ext4_destroy_lazyinit_thread(void);
84 static void ext4_unregister_li_request(struct super_block *sb);
85 static void ext4_clear_request_list(void);
86 static struct inode *ext4_get_journal_inode(struct super_block *sb,
87 unsigned int journal_inum);
92 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
93 * i_mmap_rwsem (inode->i_mmap_rwsem)!
96 * mmap_lock -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
97 * page lock -> i_data_sem (rw)
99 * buffered write path:
100 * sb_start_write -> i_mutex -> mmap_lock
101 * sb_start_write -> i_mutex -> transaction start -> page lock ->
105 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock
106 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start ->
110 * sb_start_write -> i_mutex -> mmap_lock
111 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
114 * transaction start -> page lock(s) -> i_data_sem (rw)
117 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
118 static struct file_system_type ext2_fs_type = {
119 .owner = THIS_MODULE,
122 .kill_sb = kill_block_super,
123 .fs_flags = FS_REQUIRES_DEV,
125 MODULE_ALIAS_FS("ext2");
126 MODULE_ALIAS("ext2");
127 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
129 #define IS_EXT2_SB(sb) (0)
133 static struct file_system_type ext3_fs_type = {
134 .owner = THIS_MODULE,
137 .kill_sb = kill_block_super,
138 .fs_flags = FS_REQUIRES_DEV,
140 MODULE_ALIAS_FS("ext3");
141 MODULE_ALIAS("ext3");
142 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
145 static inline void __ext4_read_bh(struct buffer_head *bh, int op_flags,
149 * buffer's verified bit is no longer valid after reading from
150 * disk again due to write out error, clear it to make sure we
151 * recheck the buffer contents.
153 clear_buffer_verified(bh);
155 bh->b_end_io = end_io ? end_io : end_buffer_read_sync;
157 submit_bh(REQ_OP_READ, op_flags, bh);
160 void ext4_read_bh_nowait(struct buffer_head *bh, int op_flags,
163 BUG_ON(!buffer_locked(bh));
165 if (ext4_buffer_uptodate(bh)) {
169 __ext4_read_bh(bh, op_flags, end_io);
172 int ext4_read_bh(struct buffer_head *bh, int op_flags, bh_end_io_t *end_io)
174 BUG_ON(!buffer_locked(bh));
176 if (ext4_buffer_uptodate(bh)) {
181 __ext4_read_bh(bh, op_flags, end_io);
184 if (buffer_uptodate(bh))
189 int ext4_read_bh_lock(struct buffer_head *bh, int op_flags, bool wait)
191 if (trylock_buffer(bh)) {
193 return ext4_read_bh(bh, op_flags, NULL);
194 ext4_read_bh_nowait(bh, op_flags, NULL);
199 if (buffer_uptodate(bh))
207 * This works like sb_bread() except it uses ERR_PTR for error
208 * returns. Currently with sb_bread it's impossible to distinguish
209 * between ENOMEM and EIO situations (since both result in a NULL
213 ext4_sb_bread(struct super_block *sb, sector_t block, int op_flags)
215 struct buffer_head *bh = sb_getblk(sb, block);
218 return ERR_PTR(-ENOMEM);
219 if (ext4_buffer_uptodate(bh))
221 clear_buffer_verified(bh);
222 ll_rw_block(REQ_OP_READ, REQ_META | op_flags, 1, &bh);
224 if (buffer_uptodate(bh))
227 return ERR_PTR(-EIO);
230 static int ext4_verify_csum_type(struct super_block *sb,
231 struct ext4_super_block *es)
233 if (!ext4_has_feature_metadata_csum(sb))
236 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
239 static __le32 ext4_superblock_csum(struct super_block *sb,
240 struct ext4_super_block *es)
242 struct ext4_sb_info *sbi = EXT4_SB(sb);
243 int offset = offsetof(struct ext4_super_block, s_checksum);
246 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
248 return cpu_to_le32(csum);
251 static int ext4_superblock_csum_verify(struct super_block *sb,
252 struct ext4_super_block *es)
254 if (!ext4_has_metadata_csum(sb))
257 return es->s_checksum == ext4_superblock_csum(sb, es);
260 void ext4_superblock_csum_set(struct super_block *sb)
262 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
264 if (!ext4_has_metadata_csum(sb))
268 * Locking the superblock prevents the scenario
270 * 1) a first thread pauses during checksum calculation.
271 * 2) a second thread updates the superblock, recalculates
272 * the checksum, and updates s_checksum
273 * 3) the first thread resumes and finishes its checksum calculation
274 * and updates s_checksum with a potentially stale or torn value.
276 lock_buffer(EXT4_SB(sb)->s_sbh);
277 es->s_checksum = ext4_superblock_csum(sb, es);
278 unlock_buffer(EXT4_SB(sb)->s_sbh);
281 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
282 struct ext4_group_desc *bg)
284 return le32_to_cpu(bg->bg_block_bitmap_lo) |
285 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
286 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
289 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
290 struct ext4_group_desc *bg)
292 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
293 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
294 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
297 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
298 struct ext4_group_desc *bg)
300 return le32_to_cpu(bg->bg_inode_table_lo) |
301 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
302 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
305 __u32 ext4_free_group_clusters(struct super_block *sb,
306 struct ext4_group_desc *bg)
308 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
309 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
310 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
313 __u32 ext4_free_inodes_count(struct super_block *sb,
314 struct ext4_group_desc *bg)
316 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
317 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
318 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
321 __u32 ext4_used_dirs_count(struct super_block *sb,
322 struct ext4_group_desc *bg)
324 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
325 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
326 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
329 __u32 ext4_itable_unused_count(struct super_block *sb,
330 struct ext4_group_desc *bg)
332 return le16_to_cpu(bg->bg_itable_unused_lo) |
333 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
334 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
337 void ext4_block_bitmap_set(struct super_block *sb,
338 struct ext4_group_desc *bg, ext4_fsblk_t blk)
340 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
341 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
342 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
345 void ext4_inode_bitmap_set(struct super_block *sb,
346 struct ext4_group_desc *bg, ext4_fsblk_t blk)
348 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
349 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
350 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
353 void ext4_inode_table_set(struct super_block *sb,
354 struct ext4_group_desc *bg, ext4_fsblk_t blk)
356 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
357 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
358 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
361 void ext4_free_group_clusters_set(struct super_block *sb,
362 struct ext4_group_desc *bg, __u32 count)
364 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
365 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
366 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
369 void ext4_free_inodes_set(struct super_block *sb,
370 struct ext4_group_desc *bg, __u32 count)
372 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
373 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
374 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
377 void ext4_used_dirs_set(struct super_block *sb,
378 struct ext4_group_desc *bg, __u32 count)
380 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
381 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
382 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
385 void ext4_itable_unused_set(struct super_block *sb,
386 struct ext4_group_desc *bg, __u32 count)
388 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
389 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
390 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
393 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi)
395 time64_t now = ktime_get_real_seconds();
397 now = clamp_val(now, 0, (1ull << 40) - 1);
399 *lo = cpu_to_le32(lower_32_bits(now));
400 *hi = upper_32_bits(now);
403 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
405 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
407 #define ext4_update_tstamp(es, tstamp) \
408 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
409 #define ext4_get_tstamp(es, tstamp) \
410 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
412 static void __save_error_info(struct super_block *sb, int error,
413 __u32 ino, __u64 block,
414 const char *func, unsigned int line)
416 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
419 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
420 if (bdev_read_only(sb->s_bdev))
422 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
423 ext4_update_tstamp(es, s_last_error_time);
424 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
425 es->s_last_error_line = cpu_to_le32(line);
426 es->s_last_error_ino = cpu_to_le32(ino);
427 es->s_last_error_block = cpu_to_le64(block);
433 err = EXT4_ERR_ENOMEM;
436 err = EXT4_ERR_EFSBADCRC;
440 err = EXT4_ERR_EFSCORRUPTED;
443 err = EXT4_ERR_ENOSPC;
446 err = EXT4_ERR_ENOKEY;
449 err = EXT4_ERR_EROFS;
452 err = EXT4_ERR_EFBIG;
455 err = EXT4_ERR_EEXIST;
458 err = EXT4_ERR_ERANGE;
461 err = EXT4_ERR_EOVERFLOW;
464 err = EXT4_ERR_EBUSY;
467 err = EXT4_ERR_ENOTDIR;
470 err = EXT4_ERR_ENOTEMPTY;
473 err = EXT4_ERR_ESHUTDOWN;
476 err = EXT4_ERR_EFAULT;
479 err = EXT4_ERR_UNKNOWN;
481 es->s_last_error_errcode = err;
482 if (!es->s_first_error_time) {
483 es->s_first_error_time = es->s_last_error_time;
484 es->s_first_error_time_hi = es->s_last_error_time_hi;
485 strncpy(es->s_first_error_func, func,
486 sizeof(es->s_first_error_func));
487 es->s_first_error_line = cpu_to_le32(line);
488 es->s_first_error_ino = es->s_last_error_ino;
489 es->s_first_error_block = es->s_last_error_block;
490 es->s_first_error_errcode = es->s_last_error_errcode;
493 * Start the daily error reporting function if it hasn't been
496 if (!es->s_error_count)
497 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
498 le32_add_cpu(&es->s_error_count, 1);
501 static void save_error_info(struct super_block *sb, int error,
502 __u32 ino, __u64 block,
503 const char *func, unsigned int line)
505 __save_error_info(sb, error, ino, block, func, line);
506 if (!bdev_read_only(sb->s_bdev))
507 ext4_commit_super(sb, 1);
511 * The del_gendisk() function uninitializes the disk-specific data
512 * structures, including the bdi structure, without telling anyone
513 * else. Once this happens, any attempt to call mark_buffer_dirty()
514 * (for example, by ext4_commit_super), will cause a kernel OOPS.
515 * This is a kludge to prevent these oops until we can put in a proper
516 * hook in del_gendisk() to inform the VFS and file system layers.
518 static int block_device_ejected(struct super_block *sb)
520 struct inode *bd_inode = sb->s_bdev->bd_inode;
521 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
523 return bdi->dev == NULL;
526 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
528 struct super_block *sb = journal->j_private;
529 struct ext4_sb_info *sbi = EXT4_SB(sb);
530 int error = is_journal_aborted(journal);
531 struct ext4_journal_cb_entry *jce;
533 BUG_ON(txn->t_state == T_FINISHED);
535 ext4_process_freed_data(sb, txn->t_tid);
537 spin_lock(&sbi->s_md_lock);
538 while (!list_empty(&txn->t_private_list)) {
539 jce = list_entry(txn->t_private_list.next,
540 struct ext4_journal_cb_entry, jce_list);
541 list_del_init(&jce->jce_list);
542 spin_unlock(&sbi->s_md_lock);
543 jce->jce_func(sb, jce, error);
544 spin_lock(&sbi->s_md_lock);
546 spin_unlock(&sbi->s_md_lock);
549 static bool system_going_down(void)
551 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
552 || system_state == SYSTEM_RESTART;
555 /* Deal with the reporting of failure conditions on a filesystem such as
556 * inconsistencies detected or read IO failures.
558 * On ext2, we can store the error state of the filesystem in the
559 * superblock. That is not possible on ext4, because we may have other
560 * write ordering constraints on the superblock which prevent us from
561 * writing it out straight away; and given that the journal is about to
562 * be aborted, we can't rely on the current, or future, transactions to
563 * write out the superblock safely.
565 * We'll just use the jbd2_journal_abort() error code to record an error in
566 * the journal instead. On recovery, the journal will complain about
567 * that error until we've noted it down and cleared it.
570 static void ext4_handle_error(struct super_block *sb)
572 if (test_opt(sb, WARN_ON_ERROR))
578 if (!test_opt(sb, ERRORS_CONT)) {
579 journal_t *journal = EXT4_SB(sb)->s_journal;
581 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
583 jbd2_journal_abort(journal, -EIO);
586 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
587 * could panic during 'reboot -f' as the underlying device got already
590 if (test_opt(sb, ERRORS_RO) || system_going_down()) {
591 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
593 * Make sure updated value of ->s_mount_flags will be visible
594 * before ->s_flags update
597 sb->s_flags |= SB_RDONLY;
598 } else if (test_opt(sb, ERRORS_PANIC)) {
599 panic("EXT4-fs (device %s): panic forced after error\n",
604 #define ext4_error_ratelimit(sb) \
605 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
608 void __ext4_error(struct super_block *sb, const char *function,
609 unsigned int line, int error, __u64 block,
610 const char *fmt, ...)
612 struct va_format vaf;
615 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
618 trace_ext4_error(sb, function, line);
619 if (ext4_error_ratelimit(sb)) {
624 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
625 sb->s_id, function, line, current->comm, &vaf);
628 save_error_info(sb, error, 0, block, function, line);
629 ext4_handle_error(sb);
632 void __ext4_error_inode(struct inode *inode, const char *function,
633 unsigned int line, ext4_fsblk_t block, int error,
634 const char *fmt, ...)
637 struct va_format vaf;
639 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
642 trace_ext4_error(inode->i_sb, function, line);
643 if (ext4_error_ratelimit(inode->i_sb)) {
648 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
649 "inode #%lu: block %llu: comm %s: %pV\n",
650 inode->i_sb->s_id, function, line, inode->i_ino,
651 block, current->comm, &vaf);
653 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
654 "inode #%lu: comm %s: %pV\n",
655 inode->i_sb->s_id, function, line, inode->i_ino,
656 current->comm, &vaf);
659 save_error_info(inode->i_sb, error, inode->i_ino, block,
661 ext4_handle_error(inode->i_sb);
664 void __ext4_error_file(struct file *file, const char *function,
665 unsigned int line, ext4_fsblk_t block,
666 const char *fmt, ...)
669 struct va_format vaf;
670 struct inode *inode = file_inode(file);
671 char pathname[80], *path;
673 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
676 trace_ext4_error(inode->i_sb, function, line);
677 if (ext4_error_ratelimit(inode->i_sb)) {
678 path = file_path(file, pathname, sizeof(pathname));
686 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
687 "block %llu: comm %s: path %s: %pV\n",
688 inode->i_sb->s_id, function, line, inode->i_ino,
689 block, current->comm, path, &vaf);
692 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
693 "comm %s: path %s: %pV\n",
694 inode->i_sb->s_id, function, line, inode->i_ino,
695 current->comm, path, &vaf);
698 save_error_info(inode->i_sb, EFSCORRUPTED, inode->i_ino, block,
700 ext4_handle_error(inode->i_sb);
703 const char *ext4_decode_error(struct super_block *sb, int errno,
710 errstr = "Corrupt filesystem";
713 errstr = "Filesystem failed CRC";
716 errstr = "IO failure";
719 errstr = "Out of memory";
722 if (!sb || (EXT4_SB(sb)->s_journal &&
723 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
724 errstr = "Journal has aborted";
726 errstr = "Readonly filesystem";
729 /* If the caller passed in an extra buffer for unknown
730 * errors, textualise them now. Else we just return
733 /* Check for truncated error codes... */
734 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
743 /* __ext4_std_error decodes expected errors from journaling functions
744 * automatically and invokes the appropriate error response. */
746 void __ext4_std_error(struct super_block *sb, const char *function,
747 unsigned int line, int errno)
752 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
755 /* Special case: if the error is EROFS, and we're not already
756 * inside a transaction, then there's really no point in logging
758 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
761 if (ext4_error_ratelimit(sb)) {
762 errstr = ext4_decode_error(sb, errno, nbuf);
763 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
764 sb->s_id, function, line, errstr);
767 save_error_info(sb, -errno, 0, 0, function, line);
768 ext4_handle_error(sb);
772 * ext4_abort is a much stronger failure handler than ext4_error. The
773 * abort function may be used to deal with unrecoverable failures such
774 * as journal IO errors or ENOMEM at a critical moment in log management.
776 * We unconditionally force the filesystem into an ABORT|READONLY state,
777 * unless the error response on the fs has been set to panic in which
778 * case we take the easy way out and panic immediately.
781 void __ext4_abort(struct super_block *sb, const char *function,
782 unsigned int line, int error, const char *fmt, ...)
784 struct va_format vaf;
787 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
790 save_error_info(sb, error, 0, 0, function, line);
794 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
795 sb->s_id, function, line, &vaf);
798 if (sb_rdonly(sb) == 0) {
799 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
800 if (EXT4_SB(sb)->s_journal)
801 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
803 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
805 * Make sure updated value of ->s_mount_flags will be visible
806 * before ->s_flags update
809 sb->s_flags |= SB_RDONLY;
811 if (test_opt(sb, ERRORS_PANIC) && !system_going_down())
812 panic("EXT4-fs panic from previous error\n");
815 void __ext4_msg(struct super_block *sb,
816 const char *prefix, const char *fmt, ...)
818 struct va_format vaf;
821 atomic_inc(&EXT4_SB(sb)->s_msg_count);
822 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
828 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
832 static int ext4_warning_ratelimit(struct super_block *sb)
834 atomic_inc(&EXT4_SB(sb)->s_warning_count);
835 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
839 void __ext4_warning(struct super_block *sb, const char *function,
840 unsigned int line, const char *fmt, ...)
842 struct va_format vaf;
845 if (!ext4_warning_ratelimit(sb))
851 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
852 sb->s_id, function, line, &vaf);
856 void __ext4_warning_inode(const struct inode *inode, const char *function,
857 unsigned int line, const char *fmt, ...)
859 struct va_format vaf;
862 if (!ext4_warning_ratelimit(inode->i_sb))
868 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
869 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
870 function, line, inode->i_ino, current->comm, &vaf);
874 void __ext4_grp_locked_error(const char *function, unsigned int line,
875 struct super_block *sb, ext4_group_t grp,
876 unsigned long ino, ext4_fsblk_t block,
877 const char *fmt, ...)
881 struct va_format vaf;
884 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
887 trace_ext4_error(sb, function, line);
888 __save_error_info(sb, EFSCORRUPTED, ino, block, function, line);
890 if (ext4_error_ratelimit(sb)) {
894 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
895 sb->s_id, function, line, grp);
897 printk(KERN_CONT "inode %lu: ", ino);
899 printk(KERN_CONT "block %llu:",
900 (unsigned long long) block);
901 printk(KERN_CONT "%pV\n", &vaf);
905 if (test_opt(sb, WARN_ON_ERROR))
908 if (test_opt(sb, ERRORS_CONT)) {
909 ext4_commit_super(sb, 0);
913 ext4_unlock_group(sb, grp);
914 ext4_commit_super(sb, 1);
915 ext4_handle_error(sb);
917 * We only get here in the ERRORS_RO case; relocking the group
918 * may be dangerous, but nothing bad will happen since the
919 * filesystem will have already been marked read/only and the
920 * journal has been aborted. We return 1 as a hint to callers
921 * who might what to use the return value from
922 * ext4_grp_locked_error() to distinguish between the
923 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
924 * aggressively from the ext4 function in question, with a
925 * more appropriate error code.
927 ext4_lock_group(sb, grp);
931 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
935 struct ext4_sb_info *sbi = EXT4_SB(sb);
936 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
937 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
940 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
941 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
944 percpu_counter_sub(&sbi->s_freeclusters_counter,
948 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
949 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
954 count = ext4_free_inodes_count(sb, gdp);
955 percpu_counter_sub(&sbi->s_freeinodes_counter,
961 void ext4_update_dynamic_rev(struct super_block *sb)
963 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
965 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
969 "updating to rev %d because of new feature flag, "
970 "running e2fsck is recommended",
973 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
974 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
975 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
976 /* leave es->s_feature_*compat flags alone */
977 /* es->s_uuid will be set by e2fsck if empty */
980 * The rest of the superblock fields should be zero, and if not it
981 * means they are likely already in use, so leave them alone. We
982 * can leave it up to e2fsck to clean up any inconsistencies there.
987 * Open the external journal device
989 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
991 struct block_device *bdev;
993 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
999 ext4_msg(sb, KERN_ERR,
1000 "failed to open journal device unknown-block(%u,%u) %ld",
1001 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
1006 * Release the journal device
1008 static void ext4_blkdev_put(struct block_device *bdev)
1010 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
1013 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
1015 struct block_device *bdev;
1016 bdev = sbi->s_journal_bdev;
1018 ext4_blkdev_put(bdev);
1019 sbi->s_journal_bdev = NULL;
1023 static inline struct inode *orphan_list_entry(struct list_head *l)
1025 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1028 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1030 struct list_head *l;
1032 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1033 le32_to_cpu(sbi->s_es->s_last_orphan));
1035 printk(KERN_ERR "sb_info orphan list:\n");
1036 list_for_each(l, &sbi->s_orphan) {
1037 struct inode *inode = orphan_list_entry(l);
1039 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1040 inode->i_sb->s_id, inode->i_ino, inode,
1041 inode->i_mode, inode->i_nlink,
1042 NEXT_ORPHAN(inode));
1047 static int ext4_quota_off(struct super_block *sb, int type);
1049 static inline void ext4_quota_off_umount(struct super_block *sb)
1053 /* Use our quota_off function to clear inode flags etc. */
1054 for (type = 0; type < EXT4_MAXQUOTAS; type++)
1055 ext4_quota_off(sb, type);
1059 * This is a helper function which is used in the mount/remount
1060 * codepaths (which holds s_umount) to fetch the quota file name.
1062 static inline char *get_qf_name(struct super_block *sb,
1063 struct ext4_sb_info *sbi,
1066 return rcu_dereference_protected(sbi->s_qf_names[type],
1067 lockdep_is_held(&sb->s_umount));
1070 static inline void ext4_quota_off_umount(struct super_block *sb)
1075 static void ext4_put_super(struct super_block *sb)
1077 struct ext4_sb_info *sbi = EXT4_SB(sb);
1078 struct ext4_super_block *es = sbi->s_es;
1079 struct buffer_head **group_desc;
1080 struct flex_groups **flex_groups;
1084 ext4_unregister_li_request(sb);
1085 ext4_quota_off_umount(sb);
1087 destroy_workqueue(sbi->rsv_conversion_wq);
1090 * Unregister sysfs before destroying jbd2 journal.
1091 * Since we could still access attr_journal_task attribute via sysfs
1092 * path which could have sbi->s_journal->j_task as NULL
1094 ext4_unregister_sysfs(sb);
1096 if (sbi->s_journal) {
1097 aborted = is_journal_aborted(sbi->s_journal);
1098 err = jbd2_journal_destroy(sbi->s_journal);
1099 sbi->s_journal = NULL;
1100 if ((err < 0) && !aborted) {
1101 ext4_abort(sb, -err, "Couldn't clean up the journal");
1105 ext4_es_unregister_shrinker(sbi);
1106 del_timer_sync(&sbi->s_err_report);
1107 ext4_release_system_zone(sb);
1108 ext4_mb_release(sb);
1109 ext4_ext_release(sb);
1111 if (!sb_rdonly(sb) && !aborted) {
1112 ext4_clear_feature_journal_needs_recovery(sb);
1113 es->s_state = cpu_to_le16(sbi->s_mount_state);
1116 ext4_commit_super(sb, 1);
1119 group_desc = rcu_dereference(sbi->s_group_desc);
1120 for (i = 0; i < sbi->s_gdb_count; i++)
1121 brelse(group_desc[i]);
1123 flex_groups = rcu_dereference(sbi->s_flex_groups);
1125 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1126 kvfree(flex_groups[i]);
1127 kvfree(flex_groups);
1130 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1131 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1132 percpu_counter_destroy(&sbi->s_dirs_counter);
1133 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1134 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1136 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1137 kfree(get_qf_name(sb, sbi, i));
1140 /* Debugging code just in case the in-memory inode orphan list
1141 * isn't empty. The on-disk one can be non-empty if we've
1142 * detected an error and taken the fs readonly, but the
1143 * in-memory list had better be clean by this point. */
1144 if (!list_empty(&sbi->s_orphan))
1145 dump_orphan_list(sb, sbi);
1146 J_ASSERT(list_empty(&sbi->s_orphan));
1148 sync_blockdev(sb->s_bdev);
1149 invalidate_bdev(sb->s_bdev);
1150 if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) {
1152 * Invalidate the journal device's buffers. We don't want them
1153 * floating about in memory - the physical journal device may
1154 * hotswapped, and it breaks the `ro-after' testing code.
1156 sync_blockdev(sbi->s_journal_bdev);
1157 invalidate_bdev(sbi->s_journal_bdev);
1158 ext4_blkdev_remove(sbi);
1161 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1162 sbi->s_ea_inode_cache = NULL;
1164 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1165 sbi->s_ea_block_cache = NULL;
1168 kthread_stop(sbi->s_mmp_tsk);
1170 sb->s_fs_info = NULL;
1172 * Now that we are completely done shutting down the
1173 * superblock, we need to actually destroy the kobject.
1175 kobject_put(&sbi->s_kobj);
1176 wait_for_completion(&sbi->s_kobj_unregister);
1177 if (sbi->s_chksum_driver)
1178 crypto_free_shash(sbi->s_chksum_driver);
1179 kfree(sbi->s_blockgroup_lock);
1180 fs_put_dax(sbi->s_daxdev);
1181 fscrypt_free_dummy_context(&sbi->s_dummy_enc_ctx);
1182 #ifdef CONFIG_UNICODE
1183 utf8_unload(sbi->s_encoding);
1188 static struct kmem_cache *ext4_inode_cachep;
1191 * Called inside transaction, so use GFP_NOFS
1193 static struct inode *ext4_alloc_inode(struct super_block *sb)
1195 struct ext4_inode_info *ei;
1197 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1201 inode_set_iversion(&ei->vfs_inode, 1);
1202 spin_lock_init(&ei->i_raw_lock);
1203 INIT_LIST_HEAD(&ei->i_prealloc_list);
1204 atomic_set(&ei->i_prealloc_active, 0);
1205 spin_lock_init(&ei->i_prealloc_lock);
1206 ext4_es_init_tree(&ei->i_es_tree);
1207 rwlock_init(&ei->i_es_lock);
1208 INIT_LIST_HEAD(&ei->i_es_list);
1209 ei->i_es_all_nr = 0;
1210 ei->i_es_shk_nr = 0;
1211 ei->i_es_shrink_lblk = 0;
1212 ei->i_reserved_data_blocks = 0;
1213 spin_lock_init(&(ei->i_block_reservation_lock));
1214 ext4_init_pending_tree(&ei->i_pending_tree);
1216 ei->i_reserved_quota = 0;
1217 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1220 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1221 spin_lock_init(&ei->i_completed_io_lock);
1223 ei->i_datasync_tid = 0;
1224 atomic_set(&ei->i_unwritten, 0);
1225 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1226 return &ei->vfs_inode;
1229 static int ext4_drop_inode(struct inode *inode)
1231 int drop = generic_drop_inode(inode);
1234 drop = fscrypt_drop_inode(inode);
1236 trace_ext4_drop_inode(inode, drop);
1240 static void ext4_free_in_core_inode(struct inode *inode)
1242 fscrypt_free_inode(inode);
1243 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1246 static void ext4_destroy_inode(struct inode *inode)
1248 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1249 ext4_msg(inode->i_sb, KERN_ERR,
1250 "Inode %lu (%p): orphan list check failed!",
1251 inode->i_ino, EXT4_I(inode));
1252 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1253 EXT4_I(inode), sizeof(struct ext4_inode_info),
1259 static void init_once(void *foo)
1261 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1263 INIT_LIST_HEAD(&ei->i_orphan);
1264 init_rwsem(&ei->xattr_sem);
1265 init_rwsem(&ei->i_data_sem);
1266 init_rwsem(&ei->i_mmap_sem);
1267 inode_init_once(&ei->vfs_inode);
1270 static int __init init_inodecache(void)
1272 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1273 sizeof(struct ext4_inode_info), 0,
1274 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1276 offsetof(struct ext4_inode_info, i_data),
1277 sizeof_field(struct ext4_inode_info, i_data),
1279 if (ext4_inode_cachep == NULL)
1284 static void destroy_inodecache(void)
1287 * Make sure all delayed rcu free inodes are flushed before we
1291 kmem_cache_destroy(ext4_inode_cachep);
1294 void ext4_clear_inode(struct inode *inode)
1296 invalidate_inode_buffers(inode);
1298 ext4_discard_preallocations(inode, 0);
1299 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1301 if (EXT4_I(inode)->jinode) {
1302 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1303 EXT4_I(inode)->jinode);
1304 jbd2_free_inode(EXT4_I(inode)->jinode);
1305 EXT4_I(inode)->jinode = NULL;
1307 fscrypt_put_encryption_info(inode);
1308 fsverity_cleanup_inode(inode);
1311 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1312 u64 ino, u32 generation)
1314 struct inode *inode;
1317 * Currently we don't know the generation for parent directory, so
1318 * a generation of 0 means "accept any"
1320 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1322 return ERR_CAST(inode);
1323 if (generation && inode->i_generation != generation) {
1325 return ERR_PTR(-ESTALE);
1331 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1332 int fh_len, int fh_type)
1334 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1335 ext4_nfs_get_inode);
1338 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1339 int fh_len, int fh_type)
1341 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1342 ext4_nfs_get_inode);
1345 static int ext4_nfs_commit_metadata(struct inode *inode)
1347 struct writeback_control wbc = {
1348 .sync_mode = WB_SYNC_ALL
1351 trace_ext4_nfs_commit_metadata(inode);
1352 return ext4_write_inode(inode, &wbc);
1356 * Try to release metadata pages (indirect blocks, directories) which are
1357 * mapped via the block device. Since these pages could have journal heads
1358 * which would prevent try_to_free_buffers() from freeing them, we must use
1359 * jbd2 layer's try_to_free_buffers() function to release them.
1361 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1364 journal_t *journal = EXT4_SB(sb)->s_journal;
1366 WARN_ON(PageChecked(page));
1367 if (!page_has_buffers(page))
1370 return jbd2_journal_try_to_free_buffers(journal, page);
1372 return try_to_free_buffers(page);
1375 #ifdef CONFIG_FS_ENCRYPTION
1376 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1378 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1379 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1382 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1385 handle_t *handle = fs_data;
1386 int res, res2, credits, retries = 0;
1389 * Encrypting the root directory is not allowed because e2fsck expects
1390 * lost+found to exist and be unencrypted, and encrypting the root
1391 * directory would imply encrypting the lost+found directory as well as
1392 * the filename "lost+found" itself.
1394 if (inode->i_ino == EXT4_ROOT_INO)
1397 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1400 if (ext4_test_inode_flag(inode, EXT4_INODE_DAX))
1403 res = ext4_convert_inline_data(inode);
1408 * If a journal handle was specified, then the encryption context is
1409 * being set on a new inode via inheritance and is part of a larger
1410 * transaction to create the inode. Otherwise the encryption context is
1411 * being set on an existing inode in its own transaction. Only in the
1412 * latter case should the "retry on ENOSPC" logic be used.
1416 res = ext4_xattr_set_handle(handle, inode,
1417 EXT4_XATTR_INDEX_ENCRYPTION,
1418 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1421 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1422 ext4_clear_inode_state(inode,
1423 EXT4_STATE_MAY_INLINE_DATA);
1425 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1426 * S_DAX may be disabled
1428 ext4_set_inode_flags(inode, false);
1433 res = dquot_initialize(inode);
1437 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1442 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1444 return PTR_ERR(handle);
1446 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1447 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1450 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1452 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1453 * S_DAX may be disabled
1455 ext4_set_inode_flags(inode, false);
1456 res = ext4_mark_inode_dirty(handle, inode);
1458 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1460 res2 = ext4_journal_stop(handle);
1462 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1469 static const union fscrypt_context *
1470 ext4_get_dummy_context(struct super_block *sb)
1472 return EXT4_SB(sb)->s_dummy_enc_ctx.ctx;
1475 static bool ext4_has_stable_inodes(struct super_block *sb)
1477 return ext4_has_feature_stable_inodes(sb);
1480 static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1481 int *ino_bits_ret, int *lblk_bits_ret)
1483 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1484 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1487 static const struct fscrypt_operations ext4_cryptops = {
1488 .key_prefix = "ext4:",
1489 .get_context = ext4_get_context,
1490 .set_context = ext4_set_context,
1491 .get_dummy_context = ext4_get_dummy_context,
1492 .empty_dir = ext4_empty_dir,
1493 .max_namelen = EXT4_NAME_LEN,
1494 .has_stable_inodes = ext4_has_stable_inodes,
1495 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits,
1500 static const char * const quotatypes[] = INITQFNAMES;
1501 #define QTYPE2NAME(t) (quotatypes[t])
1503 static int ext4_write_dquot(struct dquot *dquot);
1504 static int ext4_acquire_dquot(struct dquot *dquot);
1505 static int ext4_release_dquot(struct dquot *dquot);
1506 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1507 static int ext4_write_info(struct super_block *sb, int type);
1508 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1509 const struct path *path);
1510 static int ext4_quota_on_mount(struct super_block *sb, int type);
1511 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1512 size_t len, loff_t off);
1513 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1514 const char *data, size_t len, loff_t off);
1515 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1516 unsigned int flags);
1517 static int ext4_enable_quotas(struct super_block *sb);
1519 static struct dquot **ext4_get_dquots(struct inode *inode)
1521 return EXT4_I(inode)->i_dquot;
1524 static const struct dquot_operations ext4_quota_operations = {
1525 .get_reserved_space = ext4_get_reserved_space,
1526 .write_dquot = ext4_write_dquot,
1527 .acquire_dquot = ext4_acquire_dquot,
1528 .release_dquot = ext4_release_dquot,
1529 .mark_dirty = ext4_mark_dquot_dirty,
1530 .write_info = ext4_write_info,
1531 .alloc_dquot = dquot_alloc,
1532 .destroy_dquot = dquot_destroy,
1533 .get_projid = ext4_get_projid,
1534 .get_inode_usage = ext4_get_inode_usage,
1535 .get_next_id = dquot_get_next_id,
1538 static const struct quotactl_ops ext4_qctl_operations = {
1539 .quota_on = ext4_quota_on,
1540 .quota_off = ext4_quota_off,
1541 .quota_sync = dquot_quota_sync,
1542 .get_state = dquot_get_state,
1543 .set_info = dquot_set_dqinfo,
1544 .get_dqblk = dquot_get_dqblk,
1545 .set_dqblk = dquot_set_dqblk,
1546 .get_nextdqblk = dquot_get_next_dqblk,
1550 static const struct super_operations ext4_sops = {
1551 .alloc_inode = ext4_alloc_inode,
1552 .free_inode = ext4_free_in_core_inode,
1553 .destroy_inode = ext4_destroy_inode,
1554 .write_inode = ext4_write_inode,
1555 .dirty_inode = ext4_dirty_inode,
1556 .drop_inode = ext4_drop_inode,
1557 .evict_inode = ext4_evict_inode,
1558 .put_super = ext4_put_super,
1559 .sync_fs = ext4_sync_fs,
1560 .freeze_fs = ext4_freeze,
1561 .unfreeze_fs = ext4_unfreeze,
1562 .statfs = ext4_statfs,
1563 .remount_fs = ext4_remount,
1564 .show_options = ext4_show_options,
1566 .quota_read = ext4_quota_read,
1567 .quota_write = ext4_quota_write,
1568 .get_dquots = ext4_get_dquots,
1570 .bdev_try_to_free_page = bdev_try_to_free_page,
1573 static const struct export_operations ext4_export_ops = {
1574 .fh_to_dentry = ext4_fh_to_dentry,
1575 .fh_to_parent = ext4_fh_to_parent,
1576 .get_parent = ext4_get_parent,
1577 .commit_metadata = ext4_nfs_commit_metadata,
1581 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1582 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1583 Opt_nouid32, Opt_debug, Opt_removed,
1584 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1585 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1586 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1587 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1588 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1589 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1591 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1592 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1593 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1594 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version,
1595 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1596 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1597 Opt_nowarn_on_error, Opt_mblk_io_submit,
1598 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1599 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1600 Opt_inode_readahead_blks, Opt_journal_ioprio,
1601 Opt_dioread_nolock, Opt_dioread_lock,
1602 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1603 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1604 Opt_prefetch_block_bitmaps,
1607 static const match_table_t tokens = {
1608 {Opt_bsd_df, "bsddf"},
1609 {Opt_minix_df, "minixdf"},
1610 {Opt_grpid, "grpid"},
1611 {Opt_grpid, "bsdgroups"},
1612 {Opt_nogrpid, "nogrpid"},
1613 {Opt_nogrpid, "sysvgroups"},
1614 {Opt_resgid, "resgid=%u"},
1615 {Opt_resuid, "resuid=%u"},
1617 {Opt_err_cont, "errors=continue"},
1618 {Opt_err_panic, "errors=panic"},
1619 {Opt_err_ro, "errors=remount-ro"},
1620 {Opt_nouid32, "nouid32"},
1621 {Opt_debug, "debug"},
1622 {Opt_removed, "oldalloc"},
1623 {Opt_removed, "orlov"},
1624 {Opt_user_xattr, "user_xattr"},
1625 {Opt_nouser_xattr, "nouser_xattr"},
1627 {Opt_noacl, "noacl"},
1628 {Opt_noload, "norecovery"},
1629 {Opt_noload, "noload"},
1630 {Opt_removed, "nobh"},
1631 {Opt_removed, "bh"},
1632 {Opt_commit, "commit=%u"},
1633 {Opt_min_batch_time, "min_batch_time=%u"},
1634 {Opt_max_batch_time, "max_batch_time=%u"},
1635 {Opt_journal_dev, "journal_dev=%u"},
1636 {Opt_journal_path, "journal_path=%s"},
1637 {Opt_journal_checksum, "journal_checksum"},
1638 {Opt_nojournal_checksum, "nojournal_checksum"},
1639 {Opt_journal_async_commit, "journal_async_commit"},
1640 {Opt_abort, "abort"},
1641 {Opt_data_journal, "data=journal"},
1642 {Opt_data_ordered, "data=ordered"},
1643 {Opt_data_writeback, "data=writeback"},
1644 {Opt_data_err_abort, "data_err=abort"},
1645 {Opt_data_err_ignore, "data_err=ignore"},
1646 {Opt_offusrjquota, "usrjquota="},
1647 {Opt_usrjquota, "usrjquota=%s"},
1648 {Opt_offgrpjquota, "grpjquota="},
1649 {Opt_grpjquota, "grpjquota=%s"},
1650 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1651 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1652 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1653 {Opt_grpquota, "grpquota"},
1654 {Opt_noquota, "noquota"},
1655 {Opt_quota, "quota"},
1656 {Opt_usrquota, "usrquota"},
1657 {Opt_prjquota, "prjquota"},
1658 {Opt_barrier, "barrier=%u"},
1659 {Opt_barrier, "barrier"},
1660 {Opt_nobarrier, "nobarrier"},
1661 {Opt_i_version, "i_version"},
1663 {Opt_dax_always, "dax=always"},
1664 {Opt_dax_inode, "dax=inode"},
1665 {Opt_dax_never, "dax=never"},
1666 {Opt_stripe, "stripe=%u"},
1667 {Opt_delalloc, "delalloc"},
1668 {Opt_warn_on_error, "warn_on_error"},
1669 {Opt_nowarn_on_error, "nowarn_on_error"},
1670 {Opt_lazytime, "lazytime"},
1671 {Opt_nolazytime, "nolazytime"},
1672 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1673 {Opt_nodelalloc, "nodelalloc"},
1674 {Opt_removed, "mblk_io_submit"},
1675 {Opt_removed, "nomblk_io_submit"},
1676 {Opt_block_validity, "block_validity"},
1677 {Opt_noblock_validity, "noblock_validity"},
1678 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1679 {Opt_journal_ioprio, "journal_ioprio=%u"},
1680 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1681 {Opt_auto_da_alloc, "auto_da_alloc"},
1682 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1683 {Opt_dioread_nolock, "dioread_nolock"},
1684 {Opt_dioread_lock, "nodioread_nolock"},
1685 {Opt_dioread_lock, "dioread_lock"},
1686 {Opt_discard, "discard"},
1687 {Opt_nodiscard, "nodiscard"},
1688 {Opt_init_itable, "init_itable=%u"},
1689 {Opt_init_itable, "init_itable"},
1690 {Opt_noinit_itable, "noinit_itable"},
1691 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1692 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
1693 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1694 {Opt_inlinecrypt, "inlinecrypt"},
1695 {Opt_nombcache, "nombcache"},
1696 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1697 {Opt_prefetch_block_bitmaps, "prefetch_block_bitmaps"},
1698 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1699 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1700 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1701 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1702 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1706 static ext4_fsblk_t get_sb_block(void **data)
1708 ext4_fsblk_t sb_block;
1709 char *options = (char *) *data;
1711 if (!options || strncmp(options, "sb=", 3) != 0)
1712 return 1; /* Default location */
1715 /* TODO: use simple_strtoll with >32bit ext4 */
1716 sb_block = simple_strtoul(options, &options, 0);
1717 if (*options && *options != ',') {
1718 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1722 if (*options == ',')
1724 *data = (void *) options;
1729 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1730 static const char deprecated_msg[] =
1731 "Mount option \"%s\" will be removed by %s\n"
1732 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1735 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1737 struct ext4_sb_info *sbi = EXT4_SB(sb);
1738 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
1741 if (sb_any_quota_loaded(sb) && !old_qname) {
1742 ext4_msg(sb, KERN_ERR,
1743 "Cannot change journaled "
1744 "quota options when quota turned on");
1747 if (ext4_has_feature_quota(sb)) {
1748 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1749 "ignored when QUOTA feature is enabled");
1752 qname = match_strdup(args);
1754 ext4_msg(sb, KERN_ERR,
1755 "Not enough memory for storing quotafile name");
1759 if (strcmp(old_qname, qname) == 0)
1762 ext4_msg(sb, KERN_ERR,
1763 "%s quota file already specified",
1767 if (strchr(qname, '/')) {
1768 ext4_msg(sb, KERN_ERR,
1769 "quotafile must be on filesystem root");
1772 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
1780 static int clear_qf_name(struct super_block *sb, int qtype)
1783 struct ext4_sb_info *sbi = EXT4_SB(sb);
1784 char *old_qname = get_qf_name(sb, sbi, qtype);
1786 if (sb_any_quota_loaded(sb) && old_qname) {
1787 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1788 " when quota turned on");
1791 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1798 #define MOPT_SET 0x0001
1799 #define MOPT_CLEAR 0x0002
1800 #define MOPT_NOSUPPORT 0x0004
1801 #define MOPT_EXPLICIT 0x0008
1802 #define MOPT_CLEAR_ERR 0x0010
1803 #define MOPT_GTE0 0x0020
1806 #define MOPT_QFMT 0x0040
1808 #define MOPT_Q MOPT_NOSUPPORT
1809 #define MOPT_QFMT MOPT_NOSUPPORT
1811 #define MOPT_DATAJ 0x0080
1812 #define MOPT_NO_EXT2 0x0100
1813 #define MOPT_NO_EXT3 0x0200
1814 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1815 #define MOPT_STRING 0x0400
1816 #define MOPT_SKIP 0x0800
1818 static const struct mount_opts {
1822 } ext4_mount_opts[] = {
1823 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1824 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1825 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1826 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1827 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1828 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1829 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1830 MOPT_EXT4_ONLY | MOPT_SET},
1831 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1832 MOPT_EXT4_ONLY | MOPT_CLEAR},
1833 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1834 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1835 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1836 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1837 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1838 MOPT_EXT4_ONLY | MOPT_CLEAR},
1839 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1840 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1841 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1842 MOPT_EXT4_ONLY | MOPT_CLEAR},
1843 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1844 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1845 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1846 EXT4_MOUNT_JOURNAL_CHECKSUM),
1847 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1848 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1849 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1850 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1851 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1852 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1854 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1856 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1857 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1858 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1859 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1860 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1861 {Opt_commit, 0, MOPT_GTE0},
1862 {Opt_max_batch_time, 0, MOPT_GTE0},
1863 {Opt_min_batch_time, 0, MOPT_GTE0},
1864 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1865 {Opt_init_itable, 0, MOPT_GTE0},
1866 {Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP},
1867 {Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS,
1868 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1869 {Opt_dax_inode, EXT4_MOUNT2_DAX_INODE,
1870 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1871 {Opt_dax_never, EXT4_MOUNT2_DAX_NEVER,
1872 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1873 {Opt_stripe, 0, MOPT_GTE0},
1874 {Opt_resuid, 0, MOPT_GTE0},
1875 {Opt_resgid, 0, MOPT_GTE0},
1876 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1877 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1878 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1879 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1880 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1881 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1882 MOPT_NO_EXT2 | MOPT_DATAJ},
1883 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1884 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1885 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1886 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1887 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1889 {Opt_acl, 0, MOPT_NOSUPPORT},
1890 {Opt_noacl, 0, MOPT_NOSUPPORT},
1892 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1893 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1894 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1895 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1896 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1898 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1900 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1902 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1903 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1904 MOPT_CLEAR | MOPT_Q},
1905 {Opt_usrjquota, 0, MOPT_Q},
1906 {Opt_grpjquota, 0, MOPT_Q},
1907 {Opt_offusrjquota, 0, MOPT_Q},
1908 {Opt_offgrpjquota, 0, MOPT_Q},
1909 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1910 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1911 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1912 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1913 {Opt_test_dummy_encryption, 0, MOPT_STRING},
1914 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1915 {Opt_prefetch_block_bitmaps, EXT4_MOUNT_PREFETCH_BLOCK_BITMAPS,
1920 #ifdef CONFIG_UNICODE
1921 static const struct ext4_sb_encodings {
1925 } ext4_sb_encoding_map[] = {
1926 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
1929 static int ext4_sb_read_encoding(const struct ext4_super_block *es,
1930 const struct ext4_sb_encodings **encoding,
1933 __u16 magic = le16_to_cpu(es->s_encoding);
1936 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1937 if (magic == ext4_sb_encoding_map[i].magic)
1940 if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
1943 *encoding = &ext4_sb_encoding_map[i];
1944 *flags = le16_to_cpu(es->s_encoding_flags);
1950 static int ext4_set_test_dummy_encryption(struct super_block *sb,
1952 const substring_t *arg,
1955 #ifdef CONFIG_FS_ENCRYPTION
1956 struct ext4_sb_info *sbi = EXT4_SB(sb);
1960 * This mount option is just for testing, and it's not worthwhile to
1961 * implement the extra complexity (e.g. RCU protection) that would be
1962 * needed to allow it to be set or changed during remount. We do allow
1963 * it to be specified during remount, but only if there is no change.
1965 if (is_remount && !sbi->s_dummy_enc_ctx.ctx) {
1966 ext4_msg(sb, KERN_WARNING,
1967 "Can't set test_dummy_encryption on remount");
1970 err = fscrypt_set_test_dummy_encryption(sb, arg, &sbi->s_dummy_enc_ctx);
1973 ext4_msg(sb, KERN_WARNING,
1974 "Can't change test_dummy_encryption on remount");
1975 else if (err == -EINVAL)
1976 ext4_msg(sb, KERN_WARNING,
1977 "Value of option \"%s\" is unrecognized", opt);
1979 ext4_msg(sb, KERN_WARNING,
1980 "Error processing option \"%s\" [%d]",
1984 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
1986 ext4_msg(sb, KERN_WARNING,
1987 "Test dummy encryption mount option ignored");
1992 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1993 substring_t *args, unsigned long *journal_devnum,
1994 unsigned int *journal_ioprio, int is_remount)
1996 struct ext4_sb_info *sbi = EXT4_SB(sb);
1997 const struct mount_opts *m;
2003 if (token == Opt_usrjquota)
2004 return set_qf_name(sb, USRQUOTA, &args[0]);
2005 else if (token == Opt_grpjquota)
2006 return set_qf_name(sb, GRPQUOTA, &args[0]);
2007 else if (token == Opt_offusrjquota)
2008 return clear_qf_name(sb, USRQUOTA);
2009 else if (token == Opt_offgrpjquota)
2010 return clear_qf_name(sb, GRPQUOTA);
2014 case Opt_nouser_xattr:
2015 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
2018 return 1; /* handled by get_sb_block() */
2020 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
2023 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
2026 sb->s_flags |= SB_I_VERSION;
2029 sb->s_flags |= SB_LAZYTIME;
2031 case Opt_nolazytime:
2032 sb->s_flags &= ~SB_LAZYTIME;
2034 case Opt_inlinecrypt:
2035 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2036 sb->s_flags |= SB_INLINECRYPT;
2038 ext4_msg(sb, KERN_ERR, "inline encryption not supported");
2043 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2044 if (token == m->token)
2047 if (m->token == Opt_err) {
2048 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
2049 "or missing value", opt);
2053 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2054 ext4_msg(sb, KERN_ERR,
2055 "Mount option \"%s\" incompatible with ext2", opt);
2058 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2059 ext4_msg(sb, KERN_ERR,
2060 "Mount option \"%s\" incompatible with ext3", opt);
2064 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
2066 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
2068 if (m->flags & MOPT_EXPLICIT) {
2069 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2070 set_opt2(sb, EXPLICIT_DELALLOC);
2071 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2072 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
2076 if (m->flags & MOPT_CLEAR_ERR)
2077 clear_opt(sb, ERRORS_MASK);
2078 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
2079 ext4_msg(sb, KERN_ERR, "Cannot change quota "
2080 "options when quota turned on");
2084 if (m->flags & MOPT_NOSUPPORT) {
2085 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
2086 } else if (token == Opt_commit) {
2088 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
2089 else if (arg > INT_MAX / HZ) {
2090 ext4_msg(sb, KERN_ERR,
2091 "Invalid commit interval %d, "
2092 "must be smaller than %d",
2096 sbi->s_commit_interval = HZ * arg;
2097 } else if (token == Opt_debug_want_extra_isize) {
2100 (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
2101 ext4_msg(sb, KERN_ERR,
2102 "Invalid want_extra_isize %d", arg);
2105 sbi->s_want_extra_isize = arg;
2106 } else if (token == Opt_max_batch_time) {
2107 sbi->s_max_batch_time = arg;
2108 } else if (token == Opt_min_batch_time) {
2109 sbi->s_min_batch_time = arg;
2110 } else if (token == Opt_inode_readahead_blks) {
2111 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
2112 ext4_msg(sb, KERN_ERR,
2113 "EXT4-fs: inode_readahead_blks must be "
2114 "0 or a power of 2 smaller than 2^31");
2117 sbi->s_inode_readahead_blks = arg;
2118 } else if (token == Opt_init_itable) {
2119 set_opt(sb, INIT_INODE_TABLE);
2121 arg = EXT4_DEF_LI_WAIT_MULT;
2122 sbi->s_li_wait_mult = arg;
2123 } else if (token == Opt_max_dir_size_kb) {
2124 sbi->s_max_dir_size_kb = arg;
2125 } else if (token == Opt_stripe) {
2126 sbi->s_stripe = arg;
2127 } else if (token == Opt_resuid) {
2128 uid = make_kuid(current_user_ns(), arg);
2129 if (!uid_valid(uid)) {
2130 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
2133 sbi->s_resuid = uid;
2134 } else if (token == Opt_resgid) {
2135 gid = make_kgid(current_user_ns(), arg);
2136 if (!gid_valid(gid)) {
2137 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
2140 sbi->s_resgid = gid;
2141 } else if (token == Opt_journal_dev) {
2143 ext4_msg(sb, KERN_ERR,
2144 "Cannot specify journal on remount");
2147 *journal_devnum = arg;
2148 } else if (token == Opt_journal_path) {
2150 struct inode *journal_inode;
2155 ext4_msg(sb, KERN_ERR,
2156 "Cannot specify journal on remount");
2159 journal_path = match_strdup(&args[0]);
2160 if (!journal_path) {
2161 ext4_msg(sb, KERN_ERR, "error: could not dup "
2162 "journal device string");
2166 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
2168 ext4_msg(sb, KERN_ERR, "error: could not find "
2169 "journal device path: error %d", error);
2170 kfree(journal_path);
2174 journal_inode = d_inode(path.dentry);
2175 if (!S_ISBLK(journal_inode->i_mode)) {
2176 ext4_msg(sb, KERN_ERR, "error: journal path %s "
2177 "is not a block device", journal_path);
2179 kfree(journal_path);
2183 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
2185 kfree(journal_path);
2186 } else if (token == Opt_journal_ioprio) {
2188 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
2193 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
2194 } else if (token == Opt_test_dummy_encryption) {
2195 return ext4_set_test_dummy_encryption(sb, opt, &args[0],
2197 } else if (m->flags & MOPT_DATAJ) {
2199 if (!sbi->s_journal)
2200 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
2201 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
2202 ext4_msg(sb, KERN_ERR,
2203 "Cannot change data mode on remount");
2207 clear_opt(sb, DATA_FLAGS);
2208 sbi->s_mount_opt |= m->mount_opt;
2211 } else if (m->flags & MOPT_QFMT) {
2212 if (sb_any_quota_loaded(sb) &&
2213 sbi->s_jquota_fmt != m->mount_opt) {
2214 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
2215 "quota options when quota turned on");
2218 if (ext4_has_feature_quota(sb)) {
2219 ext4_msg(sb, KERN_INFO,
2220 "Quota format mount options ignored "
2221 "when QUOTA feature is enabled");
2224 sbi->s_jquota_fmt = m->mount_opt;
2226 } else if (token == Opt_dax || token == Opt_dax_always ||
2227 token == Opt_dax_inode || token == Opt_dax_never) {
2228 #ifdef CONFIG_FS_DAX
2231 case Opt_dax_always:
2233 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2234 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2235 fail_dax_change_remount:
2236 ext4_msg(sb, KERN_ERR, "can't change "
2237 "dax mount option while remounting");
2241 (test_opt(sb, DATA_FLAGS) ==
2242 EXT4_MOUNT_JOURNAL_DATA)) {
2243 ext4_msg(sb, KERN_ERR, "can't mount with "
2244 "both data=journal and dax");
2247 ext4_msg(sb, KERN_WARNING,
2248 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2249 sbi->s_mount_opt |= EXT4_MOUNT_DAX_ALWAYS;
2250 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2254 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2255 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS)))
2256 goto fail_dax_change_remount;
2257 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2258 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2262 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2263 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2264 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE)))
2265 goto fail_dax_change_remount;
2266 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2267 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2268 /* Strictly for printing options */
2269 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_INODE;
2273 ext4_msg(sb, KERN_INFO, "dax option not supported");
2274 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2275 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2278 } else if (token == Opt_data_err_abort) {
2279 sbi->s_mount_opt |= m->mount_opt;
2280 } else if (token == Opt_data_err_ignore) {
2281 sbi->s_mount_opt &= ~m->mount_opt;
2285 if (m->flags & MOPT_CLEAR)
2287 else if (unlikely(!(m->flags & MOPT_SET))) {
2288 ext4_msg(sb, KERN_WARNING,
2289 "buggy handling of option %s", opt);
2294 sbi->s_mount_opt |= m->mount_opt;
2296 sbi->s_mount_opt &= ~m->mount_opt;
2301 static int parse_options(char *options, struct super_block *sb,
2302 unsigned long *journal_devnum,
2303 unsigned int *journal_ioprio,
2306 struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb);
2307 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
2308 substring_t args[MAX_OPT_ARGS];
2314 while ((p = strsep(&options, ",")) != NULL) {
2318 * Initialize args struct so we know whether arg was
2319 * found; some options take optional arguments.
2321 args[0].to = args[0].from = NULL;
2322 token = match_token(p, tokens, args);
2323 if (handle_mount_opt(sb, p, token, args, journal_devnum,
2324 journal_ioprio, is_remount) < 0)
2329 * We do the test below only for project quotas. 'usrquota' and
2330 * 'grpquota' mount options are allowed even without quota feature
2331 * to support legacy quotas in quota files.
2333 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2334 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2335 "Cannot enable project quota enforcement.");
2338 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2339 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2340 if (usr_qf_name || grp_qf_name) {
2341 if (test_opt(sb, USRQUOTA) && usr_qf_name)
2342 clear_opt(sb, USRQUOTA);
2344 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
2345 clear_opt(sb, GRPQUOTA);
2347 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
2348 ext4_msg(sb, KERN_ERR, "old and new quota "
2353 if (!sbi->s_jquota_fmt) {
2354 ext4_msg(sb, KERN_ERR, "journaled quota format "
2360 if (test_opt(sb, DIOREAD_NOLOCK)) {
2362 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2363 if (blocksize < PAGE_SIZE)
2364 ext4_msg(sb, KERN_WARNING, "Warning: mounting with an "
2365 "experimental mount option 'dioread_nolock' "
2366 "for blocksize < PAGE_SIZE");
2371 static inline void ext4_show_quota_options(struct seq_file *seq,
2372 struct super_block *sb)
2374 #if defined(CONFIG_QUOTA)
2375 struct ext4_sb_info *sbi = EXT4_SB(sb);
2376 char *usr_qf_name, *grp_qf_name;
2378 if (sbi->s_jquota_fmt) {
2381 switch (sbi->s_jquota_fmt) {
2392 seq_printf(seq, ",jqfmt=%s", fmtname);
2396 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2397 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2399 seq_show_option(seq, "usrjquota", usr_qf_name);
2401 seq_show_option(seq, "grpjquota", grp_qf_name);
2406 static const char *token2str(int token)
2408 const struct match_token *t;
2410 for (t = tokens; t->token != Opt_err; t++)
2411 if (t->token == token && !strchr(t->pattern, '='))
2418 * - it's set to a non-default value OR
2419 * - if the per-sb default is different from the global default
2421 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2424 struct ext4_sb_info *sbi = EXT4_SB(sb);
2425 struct ext4_super_block *es = sbi->s_es;
2426 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2427 const struct mount_opts *m;
2428 char sep = nodefs ? '\n' : ',';
2430 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2431 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2433 if (sbi->s_sb_block != 1)
2434 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2436 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2437 int want_set = m->flags & MOPT_SET;
2438 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2439 (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP)
2441 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2442 continue; /* skip if same as the default */
2444 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2445 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2446 continue; /* select Opt_noFoo vs Opt_Foo */
2447 SEQ_OPTS_PRINT("%s", token2str(m->token));
2450 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2451 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2452 SEQ_OPTS_PRINT("resuid=%u",
2453 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2454 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2455 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2456 SEQ_OPTS_PRINT("resgid=%u",
2457 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2458 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2459 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2460 SEQ_OPTS_PUTS("errors=remount-ro");
2461 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2462 SEQ_OPTS_PUTS("errors=continue");
2463 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2464 SEQ_OPTS_PUTS("errors=panic");
2465 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2466 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2467 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2468 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2469 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2470 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2471 if (sb->s_flags & SB_I_VERSION)
2472 SEQ_OPTS_PUTS("i_version");
2473 if (nodefs || sbi->s_stripe)
2474 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2475 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2476 (sbi->s_mount_opt ^ def_mount_opt)) {
2477 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2478 SEQ_OPTS_PUTS("data=journal");
2479 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2480 SEQ_OPTS_PUTS("data=ordered");
2481 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2482 SEQ_OPTS_PUTS("data=writeback");
2485 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2486 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2487 sbi->s_inode_readahead_blks);
2489 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2490 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2491 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2492 if (nodefs || sbi->s_max_dir_size_kb)
2493 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2494 if (test_opt(sb, DATA_ERR_ABORT))
2495 SEQ_OPTS_PUTS("data_err=abort");
2497 fscrypt_show_test_dummy_encryption(seq, sep, sb);
2499 if (sb->s_flags & SB_INLINECRYPT)
2500 SEQ_OPTS_PUTS("inlinecrypt");
2502 if (test_opt(sb, DAX_ALWAYS)) {
2504 SEQ_OPTS_PUTS("dax");
2506 SEQ_OPTS_PUTS("dax=always");
2507 } else if (test_opt2(sb, DAX_NEVER)) {
2508 SEQ_OPTS_PUTS("dax=never");
2509 } else if (test_opt2(sb, DAX_INODE)) {
2510 SEQ_OPTS_PUTS("dax=inode");
2513 ext4_show_quota_options(seq, sb);
2517 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2519 return _ext4_show_options(seq, root->d_sb, 0);
2522 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2524 struct super_block *sb = seq->private;
2527 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2528 rc = _ext4_show_options(seq, sb, 1);
2529 seq_puts(seq, "\n");
2533 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2536 struct ext4_sb_info *sbi = EXT4_SB(sb);
2539 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2540 ext4_msg(sb, KERN_ERR, "revision level too high, "
2541 "forcing read-only mode");
2547 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2548 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2549 "running e2fsck is recommended");
2550 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2551 ext4_msg(sb, KERN_WARNING,
2552 "warning: mounting fs with errors, "
2553 "running e2fsck is recommended");
2554 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2555 le16_to_cpu(es->s_mnt_count) >=
2556 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2557 ext4_msg(sb, KERN_WARNING,
2558 "warning: maximal mount count reached, "
2559 "running e2fsck is recommended");
2560 else if (le32_to_cpu(es->s_checkinterval) &&
2561 (ext4_get_tstamp(es, s_lastcheck) +
2562 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
2563 ext4_msg(sb, KERN_WARNING,
2564 "warning: checktime reached, "
2565 "running e2fsck is recommended");
2566 if (!sbi->s_journal)
2567 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2568 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2569 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2570 le16_add_cpu(&es->s_mnt_count, 1);
2571 ext4_update_tstamp(es, s_mtime);
2573 ext4_set_feature_journal_needs_recovery(sb);
2575 err = ext4_commit_super(sb, 1);
2577 if (test_opt(sb, DEBUG))
2578 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2579 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2581 sbi->s_groups_count,
2582 EXT4_BLOCKS_PER_GROUP(sb),
2583 EXT4_INODES_PER_GROUP(sb),
2584 sbi->s_mount_opt, sbi->s_mount_opt2);
2586 cleancache_init_fs(sb);
2590 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2592 struct ext4_sb_info *sbi = EXT4_SB(sb);
2593 struct flex_groups **old_groups, **new_groups;
2596 if (!sbi->s_log_groups_per_flex)
2599 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2600 if (size <= sbi->s_flex_groups_allocated)
2603 new_groups = kvzalloc(roundup_pow_of_two(size *
2604 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
2606 ext4_msg(sb, KERN_ERR,
2607 "not enough memory for %d flex group pointers", size);
2610 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2611 new_groups[i] = kvzalloc(roundup_pow_of_two(
2612 sizeof(struct flex_groups)),
2614 if (!new_groups[i]) {
2615 for (j = sbi->s_flex_groups_allocated; j < i; j++)
2616 kvfree(new_groups[j]);
2618 ext4_msg(sb, KERN_ERR,
2619 "not enough memory for %d flex groups", size);
2624 old_groups = rcu_dereference(sbi->s_flex_groups);
2626 memcpy(new_groups, old_groups,
2627 (sbi->s_flex_groups_allocated *
2628 sizeof(struct flex_groups *)));
2630 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2631 sbi->s_flex_groups_allocated = size;
2633 ext4_kvfree_array_rcu(old_groups);
2637 static int ext4_fill_flex_info(struct super_block *sb)
2639 struct ext4_sb_info *sbi = EXT4_SB(sb);
2640 struct ext4_group_desc *gdp = NULL;
2641 struct flex_groups *fg;
2642 ext4_group_t flex_group;
2645 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2646 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2647 sbi->s_log_groups_per_flex = 0;
2651 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2655 for (i = 0; i < sbi->s_groups_count; i++) {
2656 gdp = ext4_get_group_desc(sb, i, NULL);
2658 flex_group = ext4_flex_group(sbi, i);
2659 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2660 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
2661 atomic64_add(ext4_free_group_clusters(sb, gdp),
2662 &fg->free_clusters);
2663 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
2671 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2672 struct ext4_group_desc *gdp)
2674 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2676 __le32 le_group = cpu_to_le32(block_group);
2677 struct ext4_sb_info *sbi = EXT4_SB(sb);
2679 if (ext4_has_metadata_csum(sbi->s_sb)) {
2680 /* Use new metadata_csum algorithm */
2682 __u16 dummy_csum = 0;
2684 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2686 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2687 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2688 sizeof(dummy_csum));
2689 offset += sizeof(dummy_csum);
2690 if (offset < sbi->s_desc_size)
2691 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2692 sbi->s_desc_size - offset);
2694 crc = csum32 & 0xFFFF;
2698 /* old crc16 code */
2699 if (!ext4_has_feature_gdt_csum(sb))
2702 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2703 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2704 crc = crc16(crc, (__u8 *)gdp, offset);
2705 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2706 /* for checksum of struct ext4_group_desc do the rest...*/
2707 if (ext4_has_feature_64bit(sb) &&
2708 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2709 crc = crc16(crc, (__u8 *)gdp + offset,
2710 le16_to_cpu(sbi->s_es->s_desc_size) -
2714 return cpu_to_le16(crc);
2717 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2718 struct ext4_group_desc *gdp)
2720 if (ext4_has_group_desc_csum(sb) &&
2721 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2727 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2728 struct ext4_group_desc *gdp)
2730 if (!ext4_has_group_desc_csum(sb))
2732 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2735 /* Called at mount-time, super-block is locked */
2736 static int ext4_check_descriptors(struct super_block *sb,
2737 ext4_fsblk_t sb_block,
2738 ext4_group_t *first_not_zeroed)
2740 struct ext4_sb_info *sbi = EXT4_SB(sb);
2741 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2742 ext4_fsblk_t last_block;
2743 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2744 ext4_fsblk_t block_bitmap;
2745 ext4_fsblk_t inode_bitmap;
2746 ext4_fsblk_t inode_table;
2747 int flexbg_flag = 0;
2748 ext4_group_t i, grp = sbi->s_groups_count;
2750 if (ext4_has_feature_flex_bg(sb))
2753 ext4_debug("Checking group descriptors");
2755 for (i = 0; i < sbi->s_groups_count; i++) {
2756 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2758 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2759 last_block = ext4_blocks_count(sbi->s_es) - 1;
2761 last_block = first_block +
2762 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2764 if ((grp == sbi->s_groups_count) &&
2765 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2768 block_bitmap = ext4_block_bitmap(sb, gdp);
2769 if (block_bitmap == sb_block) {
2770 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2771 "Block bitmap for group %u overlaps "
2776 if (block_bitmap >= sb_block + 1 &&
2777 block_bitmap <= last_bg_block) {
2778 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2779 "Block bitmap for group %u overlaps "
2780 "block group descriptors", i);
2784 if (block_bitmap < first_block || block_bitmap > last_block) {
2785 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2786 "Block bitmap for group %u not in group "
2787 "(block %llu)!", i, block_bitmap);
2790 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2791 if (inode_bitmap == sb_block) {
2792 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2793 "Inode bitmap for group %u overlaps "
2798 if (inode_bitmap >= sb_block + 1 &&
2799 inode_bitmap <= last_bg_block) {
2800 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2801 "Inode bitmap for group %u overlaps "
2802 "block group descriptors", i);
2806 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2807 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2808 "Inode bitmap for group %u not in group "
2809 "(block %llu)!", i, inode_bitmap);
2812 inode_table = ext4_inode_table(sb, gdp);
2813 if (inode_table == sb_block) {
2814 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2815 "Inode table for group %u overlaps "
2820 if (inode_table >= sb_block + 1 &&
2821 inode_table <= last_bg_block) {
2822 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2823 "Inode table for group %u overlaps "
2824 "block group descriptors", i);
2828 if (inode_table < first_block ||
2829 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2830 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2831 "Inode table for group %u not in group "
2832 "(block %llu)!", i, inode_table);
2835 ext4_lock_group(sb, i);
2836 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2837 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2838 "Checksum for group %u failed (%u!=%u)",
2839 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2840 gdp)), le16_to_cpu(gdp->bg_checksum));
2841 if (!sb_rdonly(sb)) {
2842 ext4_unlock_group(sb, i);
2846 ext4_unlock_group(sb, i);
2848 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2850 if (NULL != first_not_zeroed)
2851 *first_not_zeroed = grp;
2855 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2856 * the superblock) which were deleted from all directories, but held open by
2857 * a process at the time of a crash. We walk the list and try to delete these
2858 * inodes at recovery time (only with a read-write filesystem).
2860 * In order to keep the orphan inode chain consistent during traversal (in
2861 * case of crash during recovery), we link each inode into the superblock
2862 * orphan list_head and handle it the same way as an inode deletion during
2863 * normal operation (which journals the operations for us).
2865 * We only do an iget() and an iput() on each inode, which is very safe if we
2866 * accidentally point at an in-use or already deleted inode. The worst that
2867 * can happen in this case is that we get a "bit already cleared" message from
2868 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2869 * e2fsck was run on this filesystem, and it must have already done the orphan
2870 * inode cleanup for us, so we can safely abort without any further action.
2872 static void ext4_orphan_cleanup(struct super_block *sb,
2873 struct ext4_super_block *es)
2875 unsigned int s_flags = sb->s_flags;
2876 int ret, nr_orphans = 0, nr_truncates = 0;
2878 int quota_update = 0;
2881 if (!es->s_last_orphan) {
2882 jbd_debug(4, "no orphan inodes to clean up\n");
2886 if (bdev_read_only(sb->s_bdev)) {
2887 ext4_msg(sb, KERN_ERR, "write access "
2888 "unavailable, skipping orphan cleanup");
2892 /* Check if feature set would not allow a r/w mount */
2893 if (!ext4_feature_set_ok(sb, 0)) {
2894 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2895 "unknown ROCOMPAT features");
2899 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2900 /* don't clear list on RO mount w/ errors */
2901 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
2902 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2903 "clearing orphan list.\n");
2904 es->s_last_orphan = 0;
2906 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2910 if (s_flags & SB_RDONLY) {
2911 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2912 sb->s_flags &= ~SB_RDONLY;
2915 /* Needed for iput() to work correctly and not trash data */
2916 sb->s_flags |= SB_ACTIVE;
2919 * Turn on quotas which were not enabled for read-only mounts if
2920 * filesystem has quota feature, so that they are updated correctly.
2922 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
2923 int ret = ext4_enable_quotas(sb);
2928 ext4_msg(sb, KERN_ERR,
2929 "Cannot turn on quotas: error %d", ret);
2932 /* Turn on journaled quotas used for old sytle */
2933 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2934 if (EXT4_SB(sb)->s_qf_names[i]) {
2935 int ret = ext4_quota_on_mount(sb, i);
2940 ext4_msg(sb, KERN_ERR,
2941 "Cannot turn on journaled "
2942 "quota: type %d: error %d", i, ret);
2947 while (es->s_last_orphan) {
2948 struct inode *inode;
2951 * We may have encountered an error during cleanup; if
2952 * so, skip the rest.
2954 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2955 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2956 es->s_last_orphan = 0;
2960 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2961 if (IS_ERR(inode)) {
2962 es->s_last_orphan = 0;
2966 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2967 dquot_initialize(inode);
2968 if (inode->i_nlink) {
2969 if (test_opt(sb, DEBUG))
2970 ext4_msg(sb, KERN_DEBUG,
2971 "%s: truncating inode %lu to %lld bytes",
2972 __func__, inode->i_ino, inode->i_size);
2973 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2974 inode->i_ino, inode->i_size);
2976 truncate_inode_pages(inode->i_mapping, inode->i_size);
2977 ret = ext4_truncate(inode);
2979 ext4_std_error(inode->i_sb, ret);
2980 inode_unlock(inode);
2983 if (test_opt(sb, DEBUG))
2984 ext4_msg(sb, KERN_DEBUG,
2985 "%s: deleting unreferenced inode %lu",
2986 __func__, inode->i_ino);
2987 jbd_debug(2, "deleting unreferenced inode %lu\n",
2991 iput(inode); /* The delete magic happens here! */
2994 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2997 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2998 PLURAL(nr_orphans));
3000 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
3001 PLURAL(nr_truncates));
3003 /* Turn off quotas if they were enabled for orphan cleanup */
3005 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
3006 if (sb_dqopt(sb)->files[i])
3007 dquot_quota_off(sb, i);
3011 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
3015 * Maximal extent format file size.
3016 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3017 * extent format containers, within a sector_t, and within i_blocks
3018 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3019 * so that won't be a limiting factor.
3021 * However there is other limiting factor. We do store extents in the form
3022 * of starting block and length, hence the resulting length of the extent
3023 * covering maximum file size must fit into on-disk format containers as
3024 * well. Given that length is always by 1 unit bigger than max unit (because
3025 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3027 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3029 static loff_t ext4_max_size(int blkbits, int has_huge_files)
3032 loff_t upper_limit = MAX_LFS_FILESIZE;
3034 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3036 if (!has_huge_files) {
3037 upper_limit = (1LL << 32) - 1;
3039 /* total blocks in file system block size */
3040 upper_limit >>= (blkbits - 9);
3041 upper_limit <<= blkbits;
3045 * 32-bit extent-start container, ee_block. We lower the maxbytes
3046 * by one fs block, so ee_len can cover the extent of maximum file
3049 res = (1LL << 32) - 1;
3052 /* Sanity check against vm- & vfs- imposed limits */
3053 if (res > upper_limit)
3060 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
3061 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3062 * We need to be 1 filesystem block less than the 2^48 sector limit.
3064 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3066 loff_t res = EXT4_NDIR_BLOCKS;
3069 /* This is calculated to be the largest file size for a dense, block
3070 * mapped file such that the file's total number of 512-byte sectors,
3071 * including data and all indirect blocks, does not exceed (2^48 - 1).
3073 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3074 * number of 512-byte sectors of the file.
3077 if (!has_huge_files) {
3079 * !has_huge_files or implies that the inode i_block field
3080 * represents total file blocks in 2^32 512-byte sectors ==
3081 * size of vfs inode i_blocks * 8
3083 upper_limit = (1LL << 32) - 1;
3085 /* total blocks in file system block size */
3086 upper_limit >>= (bits - 9);
3090 * We use 48 bit ext4_inode i_blocks
3091 * With EXT4_HUGE_FILE_FL set the i_blocks
3092 * represent total number of blocks in
3093 * file system block size
3095 upper_limit = (1LL << 48) - 1;
3099 /* indirect blocks */
3101 /* double indirect blocks */
3102 meta_blocks += 1 + (1LL << (bits-2));
3103 /* tripple indirect blocks */
3104 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
3106 upper_limit -= meta_blocks;
3107 upper_limit <<= bits;
3109 res += 1LL << (bits-2);
3110 res += 1LL << (2*(bits-2));
3111 res += 1LL << (3*(bits-2));
3113 if (res > upper_limit)
3116 if (res > MAX_LFS_FILESIZE)
3117 res = MAX_LFS_FILESIZE;
3122 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3123 ext4_fsblk_t logical_sb_block, int nr)
3125 struct ext4_sb_info *sbi = EXT4_SB(sb);
3126 ext4_group_t bg, first_meta_bg;
3129 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3131 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3132 return logical_sb_block + nr + 1;
3133 bg = sbi->s_desc_per_block * nr;
3134 if (ext4_bg_has_super(sb, bg))
3138 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3139 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3140 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3143 if (sb->s_blocksize == 1024 && nr == 0 &&
3144 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3147 return (has_super + ext4_group_first_block_no(sb, bg));
3151 * ext4_get_stripe_size: Get the stripe size.
3152 * @sbi: In memory super block info
3154 * If we have specified it via mount option, then
3155 * use the mount option value. If the value specified at mount time is
3156 * greater than the blocks per group use the super block value.
3157 * If the super block value is greater than blocks per group return 0.
3158 * Allocator needs it be less than blocks per group.
3161 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3163 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3164 unsigned long stripe_width =
3165 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3168 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3169 ret = sbi->s_stripe;
3170 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3172 else if (stride && stride <= sbi->s_blocks_per_group)
3178 * If the stripe width is 1, this makes no sense and
3179 * we set it to 0 to turn off stripe handling code.
3188 * Check whether this filesystem can be mounted based on
3189 * the features present and the RDONLY/RDWR mount requested.
3190 * Returns 1 if this filesystem can be mounted as requested,
3191 * 0 if it cannot be.
3193 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
3195 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3196 ext4_msg(sb, KERN_ERR,
3197 "Couldn't mount because of "
3198 "unsupported optional features (%x)",
3199 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3200 ~EXT4_FEATURE_INCOMPAT_SUPP));
3204 #ifndef CONFIG_UNICODE
3205 if (ext4_has_feature_casefold(sb)) {
3206 ext4_msg(sb, KERN_ERR,
3207 "Filesystem with casefold feature cannot be "
3208 "mounted without CONFIG_UNICODE");
3216 if (ext4_has_feature_readonly(sb)) {
3217 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3218 sb->s_flags |= SB_RDONLY;
3222 /* Check that feature set is OK for a read-write mount */
3223 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3224 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3225 "unsupported optional features (%x)",
3226 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3227 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3230 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3231 ext4_msg(sb, KERN_ERR,
3232 "Can't support bigalloc feature without "
3233 "extents feature\n");
3237 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3238 if (!readonly && (ext4_has_feature_quota(sb) ||
3239 ext4_has_feature_project(sb))) {
3240 ext4_msg(sb, KERN_ERR,
3241 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3244 #endif /* CONFIG_QUOTA */
3249 * This function is called once a day if we have errors logged
3250 * on the file system
3252 static void print_daily_error_info(struct timer_list *t)
3254 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3255 struct super_block *sb = sbi->s_sb;
3256 struct ext4_super_block *es = sbi->s_es;
3258 if (es->s_error_count)
3259 /* fsck newer than v1.41.13 is needed to clean this condition. */
3260 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3261 le32_to_cpu(es->s_error_count));
3262 if (es->s_first_error_time) {
3263 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3265 ext4_get_tstamp(es, s_first_error_time),
3266 (int) sizeof(es->s_first_error_func),
3267 es->s_first_error_func,
3268 le32_to_cpu(es->s_first_error_line));
3269 if (es->s_first_error_ino)
3270 printk(KERN_CONT ": inode %u",
3271 le32_to_cpu(es->s_first_error_ino));
3272 if (es->s_first_error_block)
3273 printk(KERN_CONT ": block %llu", (unsigned long long)
3274 le64_to_cpu(es->s_first_error_block));
3275 printk(KERN_CONT "\n");
3277 if (es->s_last_error_time) {
3278 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3280 ext4_get_tstamp(es, s_last_error_time),
3281 (int) sizeof(es->s_last_error_func),
3282 es->s_last_error_func,
3283 le32_to_cpu(es->s_last_error_line));
3284 if (es->s_last_error_ino)
3285 printk(KERN_CONT ": inode %u",
3286 le32_to_cpu(es->s_last_error_ino));
3287 if (es->s_last_error_block)
3288 printk(KERN_CONT ": block %llu", (unsigned long long)
3289 le64_to_cpu(es->s_last_error_block));
3290 printk(KERN_CONT "\n");
3292 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3295 /* Find next suitable group and run ext4_init_inode_table */
3296 static int ext4_run_li_request(struct ext4_li_request *elr)
3298 struct ext4_group_desc *gdp = NULL;
3299 struct super_block *sb = elr->lr_super;
3300 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3301 ext4_group_t group = elr->lr_next_group;
3302 unsigned long timeout = 0;
3303 unsigned int prefetch_ios = 0;
3306 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3307 elr->lr_next_group = ext4_mb_prefetch(sb, group,
3308 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3310 ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3312 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3314 if (group >= elr->lr_next_group) {
3316 if (elr->lr_first_not_zeroed != ngroups &&
3317 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3318 elr->lr_next_group = elr->lr_first_not_zeroed;
3319 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3326 for (; group < ngroups; group++) {
3327 gdp = ext4_get_group_desc(sb, group, NULL);
3333 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3337 if (group >= ngroups)
3342 ret = ext4_init_inode_table(sb, group,
3343 elr->lr_timeout ? 0 : 1);
3344 trace_ext4_lazy_itable_init(sb, group);
3345 if (elr->lr_timeout == 0) {
3346 timeout = (jiffies - timeout) *
3347 EXT4_SB(elr->lr_super)->s_li_wait_mult;
3348 elr->lr_timeout = timeout;
3350 elr->lr_next_sched = jiffies + elr->lr_timeout;
3351 elr->lr_next_group = group + 1;
3357 * Remove lr_request from the list_request and free the
3358 * request structure. Should be called with li_list_mtx held
3360 static void ext4_remove_li_request(struct ext4_li_request *elr)
3365 list_del(&elr->lr_request);
3366 EXT4_SB(elr->lr_super)->s_li_request = NULL;
3370 static void ext4_unregister_li_request(struct super_block *sb)
3372 mutex_lock(&ext4_li_mtx);
3373 if (!ext4_li_info) {
3374 mutex_unlock(&ext4_li_mtx);
3378 mutex_lock(&ext4_li_info->li_list_mtx);
3379 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3380 mutex_unlock(&ext4_li_info->li_list_mtx);
3381 mutex_unlock(&ext4_li_mtx);
3384 static struct task_struct *ext4_lazyinit_task;
3387 * This is the function where ext4lazyinit thread lives. It walks
3388 * through the request list searching for next scheduled filesystem.
3389 * When such a fs is found, run the lazy initialization request
3390 * (ext4_rn_li_request) and keep track of the time spend in this
3391 * function. Based on that time we compute next schedule time of
3392 * the request. When walking through the list is complete, compute
3393 * next waking time and put itself into sleep.
3395 static int ext4_lazyinit_thread(void *arg)
3397 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3398 struct list_head *pos, *n;
3399 struct ext4_li_request *elr;
3400 unsigned long next_wakeup, cur;
3402 BUG_ON(NULL == eli);
3406 next_wakeup = MAX_JIFFY_OFFSET;
3408 mutex_lock(&eli->li_list_mtx);
3409 if (list_empty(&eli->li_request_list)) {
3410 mutex_unlock(&eli->li_list_mtx);
3413 list_for_each_safe(pos, n, &eli->li_request_list) {
3416 elr = list_entry(pos, struct ext4_li_request,
3419 if (time_before(jiffies, elr->lr_next_sched)) {
3420 if (time_before(elr->lr_next_sched, next_wakeup))
3421 next_wakeup = elr->lr_next_sched;
3424 if (down_read_trylock(&elr->lr_super->s_umount)) {
3425 if (sb_start_write_trylock(elr->lr_super)) {
3428 * We hold sb->s_umount, sb can not
3429 * be removed from the list, it is
3430 * now safe to drop li_list_mtx
3432 mutex_unlock(&eli->li_list_mtx);
3433 err = ext4_run_li_request(elr);
3434 sb_end_write(elr->lr_super);
3435 mutex_lock(&eli->li_list_mtx);
3438 up_read((&elr->lr_super->s_umount));
3440 /* error, remove the lazy_init job */
3442 ext4_remove_li_request(elr);
3446 elr->lr_next_sched = jiffies +
3448 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3450 if (time_before(elr->lr_next_sched, next_wakeup))
3451 next_wakeup = elr->lr_next_sched;
3453 mutex_unlock(&eli->li_list_mtx);
3458 if ((time_after_eq(cur, next_wakeup)) ||
3459 (MAX_JIFFY_OFFSET == next_wakeup)) {
3464 schedule_timeout_interruptible(next_wakeup - cur);
3466 if (kthread_should_stop()) {
3467 ext4_clear_request_list();
3474 * It looks like the request list is empty, but we need
3475 * to check it under the li_list_mtx lock, to prevent any
3476 * additions into it, and of course we should lock ext4_li_mtx
3477 * to atomically free the list and ext4_li_info, because at
3478 * this point another ext4 filesystem could be registering
3481 mutex_lock(&ext4_li_mtx);
3482 mutex_lock(&eli->li_list_mtx);
3483 if (!list_empty(&eli->li_request_list)) {
3484 mutex_unlock(&eli->li_list_mtx);
3485 mutex_unlock(&ext4_li_mtx);
3488 mutex_unlock(&eli->li_list_mtx);
3489 kfree(ext4_li_info);
3490 ext4_li_info = NULL;
3491 mutex_unlock(&ext4_li_mtx);
3496 static void ext4_clear_request_list(void)
3498 struct list_head *pos, *n;
3499 struct ext4_li_request *elr;
3501 mutex_lock(&ext4_li_info->li_list_mtx);
3502 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3503 elr = list_entry(pos, struct ext4_li_request,
3505 ext4_remove_li_request(elr);
3507 mutex_unlock(&ext4_li_info->li_list_mtx);
3510 static int ext4_run_lazyinit_thread(void)
3512 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3513 ext4_li_info, "ext4lazyinit");
3514 if (IS_ERR(ext4_lazyinit_task)) {
3515 int err = PTR_ERR(ext4_lazyinit_task);
3516 ext4_clear_request_list();
3517 kfree(ext4_li_info);
3518 ext4_li_info = NULL;
3519 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3520 "initialization thread\n",
3524 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3529 * Check whether it make sense to run itable init. thread or not.
3530 * If there is at least one uninitialized inode table, return
3531 * corresponding group number, else the loop goes through all
3532 * groups and return total number of groups.
3534 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3536 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3537 struct ext4_group_desc *gdp = NULL;
3539 if (!ext4_has_group_desc_csum(sb))
3542 for (group = 0; group < ngroups; group++) {
3543 gdp = ext4_get_group_desc(sb, group, NULL);
3547 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3554 static int ext4_li_info_new(void)
3556 struct ext4_lazy_init *eli = NULL;
3558 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3562 INIT_LIST_HEAD(&eli->li_request_list);
3563 mutex_init(&eli->li_list_mtx);
3565 eli->li_state |= EXT4_LAZYINIT_QUIT;
3572 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3575 struct ext4_li_request *elr;
3577 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3582 elr->lr_first_not_zeroed = start;
3583 if (test_opt(sb, PREFETCH_BLOCK_BITMAPS))
3584 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3586 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3587 elr->lr_next_group = start;
3591 * Randomize first schedule time of the request to
3592 * spread the inode table initialization requests
3595 elr->lr_next_sched = jiffies + (prandom_u32() %
3596 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3600 int ext4_register_li_request(struct super_block *sb,
3601 ext4_group_t first_not_zeroed)
3603 struct ext4_sb_info *sbi = EXT4_SB(sb);
3604 struct ext4_li_request *elr = NULL;
3605 ext4_group_t ngroups = sbi->s_groups_count;
3608 mutex_lock(&ext4_li_mtx);
3609 if (sbi->s_li_request != NULL) {
3611 * Reset timeout so it can be computed again, because
3612 * s_li_wait_mult might have changed.
3614 sbi->s_li_request->lr_timeout = 0;
3618 if (!test_opt(sb, PREFETCH_BLOCK_BITMAPS) &&
3619 (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3620 !test_opt(sb, INIT_INODE_TABLE)))
3623 elr = ext4_li_request_new(sb, first_not_zeroed);
3629 if (NULL == ext4_li_info) {
3630 ret = ext4_li_info_new();
3635 mutex_lock(&ext4_li_info->li_list_mtx);
3636 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3637 mutex_unlock(&ext4_li_info->li_list_mtx);
3639 sbi->s_li_request = elr;
3641 * set elr to NULL here since it has been inserted to
3642 * the request_list and the removal and free of it is
3643 * handled by ext4_clear_request_list from now on.
3647 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3648 ret = ext4_run_lazyinit_thread();
3653 mutex_unlock(&ext4_li_mtx);
3660 * We do not need to lock anything since this is called on
3663 static void ext4_destroy_lazyinit_thread(void)
3666 * If thread exited earlier
3667 * there's nothing to be done.
3669 if (!ext4_li_info || !ext4_lazyinit_task)
3672 kthread_stop(ext4_lazyinit_task);
3675 static int set_journal_csum_feature_set(struct super_block *sb)
3678 int compat, incompat;
3679 struct ext4_sb_info *sbi = EXT4_SB(sb);
3681 if (ext4_has_metadata_csum(sb)) {
3682 /* journal checksum v3 */
3684 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3686 /* journal checksum v1 */
3687 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3691 jbd2_journal_clear_features(sbi->s_journal,
3692 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3693 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3694 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3695 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3696 ret = jbd2_journal_set_features(sbi->s_journal,
3698 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3700 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3701 ret = jbd2_journal_set_features(sbi->s_journal,
3704 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3705 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3707 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3708 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3715 * Note: calculating the overhead so we can be compatible with
3716 * historical BSD practice is quite difficult in the face of
3717 * clusters/bigalloc. This is because multiple metadata blocks from
3718 * different block group can end up in the same allocation cluster.
3719 * Calculating the exact overhead in the face of clustered allocation
3720 * requires either O(all block bitmaps) in memory or O(number of block
3721 * groups**2) in time. We will still calculate the superblock for
3722 * older file systems --- and if we come across with a bigalloc file
3723 * system with zero in s_overhead_clusters the estimate will be close to
3724 * correct especially for very large cluster sizes --- but for newer
3725 * file systems, it's better to calculate this figure once at mkfs
3726 * time, and store it in the superblock. If the superblock value is
3727 * present (even for non-bigalloc file systems), we will use it.
3729 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3732 struct ext4_sb_info *sbi = EXT4_SB(sb);
3733 struct ext4_group_desc *gdp;
3734 ext4_fsblk_t first_block, last_block, b;
3735 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3736 int s, j, count = 0;
3738 if (!ext4_has_feature_bigalloc(sb))
3739 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3740 sbi->s_itb_per_group + 2);
3742 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3743 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3744 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3745 for (i = 0; i < ngroups; i++) {
3746 gdp = ext4_get_group_desc(sb, i, NULL);
3747 b = ext4_block_bitmap(sb, gdp);
3748 if (b >= first_block && b <= last_block) {
3749 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3752 b = ext4_inode_bitmap(sb, gdp);
3753 if (b >= first_block && b <= last_block) {
3754 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3757 b = ext4_inode_table(sb, gdp);
3758 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3759 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3760 int c = EXT4_B2C(sbi, b - first_block);
3761 ext4_set_bit(c, buf);
3767 if (ext4_bg_has_super(sb, grp)) {
3768 ext4_set_bit(s++, buf);
3771 j = ext4_bg_num_gdb(sb, grp);
3772 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3773 ext4_error(sb, "Invalid number of block group "
3774 "descriptor blocks: %d", j);
3775 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3779 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3783 return EXT4_CLUSTERS_PER_GROUP(sb) -
3784 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3788 * Compute the overhead and stash it in sbi->s_overhead
3790 int ext4_calculate_overhead(struct super_block *sb)
3792 struct ext4_sb_info *sbi = EXT4_SB(sb);
3793 struct ext4_super_block *es = sbi->s_es;
3794 struct inode *j_inode;
3795 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3796 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3797 ext4_fsblk_t overhead = 0;
3798 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3804 * Compute the overhead (FS structures). This is constant
3805 * for a given filesystem unless the number of block groups
3806 * changes so we cache the previous value until it does.
3810 * All of the blocks before first_data_block are overhead
3812 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3815 * Add the overhead found in each block group
3817 for (i = 0; i < ngroups; i++) {
3820 blks = count_overhead(sb, i, buf);
3823 memset(buf, 0, PAGE_SIZE);
3828 * Add the internal journal blocks whether the journal has been
3831 if (sbi->s_journal && !sbi->s_journal_bdev)
3832 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3833 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
3834 /* j_inum for internal journal is non-zero */
3835 j_inode = ext4_get_journal_inode(sb, j_inum);
3837 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3838 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3841 ext4_msg(sb, KERN_ERR, "can't get journal size");
3844 sbi->s_overhead = overhead;
3846 free_page((unsigned long) buf);
3850 static void ext4_set_resv_clusters(struct super_block *sb)
3852 ext4_fsblk_t resv_clusters;
3853 struct ext4_sb_info *sbi = EXT4_SB(sb);
3856 * There's no need to reserve anything when we aren't using extents.
3857 * The space estimates are exact, there are no unwritten extents,
3858 * hole punching doesn't need new metadata... This is needed especially
3859 * to keep ext2/3 backward compatibility.
3861 if (!ext4_has_feature_extents(sb))
3864 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3865 * This should cover the situations where we can not afford to run
3866 * out of space like for example punch hole, or converting
3867 * unwritten extents in delalloc path. In most cases such
3868 * allocation would require 1, or 2 blocks, higher numbers are
3871 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3872 sbi->s_cluster_bits);
3874 do_div(resv_clusters, 50);
3875 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3877 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3880 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3882 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
3883 char *orig_data = kstrdup(data, GFP_KERNEL);
3884 struct buffer_head *bh, **group_desc;
3885 struct ext4_super_block *es = NULL;
3886 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3887 struct flex_groups **flex_groups;
3889 ext4_fsblk_t sb_block = get_sb_block(&data);
3890 ext4_fsblk_t logical_sb_block;
3891 unsigned long offset = 0;
3892 unsigned long journal_devnum = 0;
3893 unsigned long def_mount_opts;
3897 int blocksize, clustersize;
3898 unsigned int db_count;
3900 int needs_recovery, has_huge_files;
3903 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3904 ext4_group_t first_not_zeroed;
3906 if ((data && !orig_data) || !sbi)
3909 sbi->s_daxdev = dax_dev;
3910 sbi->s_blockgroup_lock =
3911 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3912 if (!sbi->s_blockgroup_lock)
3915 sb->s_fs_info = sbi;
3917 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3918 sbi->s_sb_block = sb_block;
3919 if (sb->s_bdev->bd_part)
3920 sbi->s_sectors_written_start =
3921 part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]);
3923 /* Cleanup superblock name */
3924 strreplace(sb->s_id, '/', '!');
3926 /* -EINVAL is default */
3928 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3930 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3935 * The ext4 superblock will not be buffer aligned for other than 1kB
3936 * block sizes. We need to calculate the offset from buffer start.
3938 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3939 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3940 offset = do_div(logical_sb_block, blocksize);
3942 logical_sb_block = sb_block;
3945 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3946 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3950 * Note: s_es must be initialized as soon as possible because
3951 * some ext4 macro-instructions depend on its value
3953 es = (struct ext4_super_block *) (bh->b_data + offset);
3955 sb->s_magic = le16_to_cpu(es->s_magic);
3956 if (sb->s_magic != EXT4_SUPER_MAGIC)
3958 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3960 /* Warn if metadata_csum and gdt_csum are both set. */
3961 if (ext4_has_feature_metadata_csum(sb) &&
3962 ext4_has_feature_gdt_csum(sb))
3963 ext4_warning(sb, "metadata_csum and uninit_bg are "
3964 "redundant flags; please run fsck.");
3966 /* Check for a known checksum algorithm */
3967 if (!ext4_verify_csum_type(sb, es)) {
3968 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3969 "unknown checksum algorithm.");
3974 /* Load the checksum driver */
3975 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3976 if (IS_ERR(sbi->s_chksum_driver)) {
3977 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3978 ret = PTR_ERR(sbi->s_chksum_driver);
3979 sbi->s_chksum_driver = NULL;
3983 /* Check superblock checksum */
3984 if (!ext4_superblock_csum_verify(sb, es)) {
3985 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3986 "invalid superblock checksum. Run e2fsck?");
3992 /* Precompute checksum seed for all metadata */
3993 if (ext4_has_feature_csum_seed(sb))
3994 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3995 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
3996 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3997 sizeof(es->s_uuid));
3999 /* Set defaults before we parse the mount options */
4000 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4001 set_opt(sb, INIT_INODE_TABLE);
4002 if (def_mount_opts & EXT4_DEFM_DEBUG)
4004 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4006 if (def_mount_opts & EXT4_DEFM_UID16)
4007 set_opt(sb, NO_UID32);
4008 /* xattr user namespace & acls are now defaulted on */
4009 set_opt(sb, XATTR_USER);
4010 #ifdef CONFIG_EXT4_FS_POSIX_ACL
4011 set_opt(sb, POSIX_ACL);
4013 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4014 if (ext4_has_metadata_csum(sb))
4015 set_opt(sb, JOURNAL_CHECKSUM);
4017 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4018 set_opt(sb, JOURNAL_DATA);
4019 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4020 set_opt(sb, ORDERED_DATA);
4021 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4022 set_opt(sb, WRITEBACK_DATA);
4024 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
4025 set_opt(sb, ERRORS_PANIC);
4026 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
4027 set_opt(sb, ERRORS_CONT);
4029 set_opt(sb, ERRORS_RO);
4030 /* block_validity enabled by default; disable with noblock_validity */
4031 set_opt(sb, BLOCK_VALIDITY);
4032 if (def_mount_opts & EXT4_DEFM_DISCARD)
4033 set_opt(sb, DISCARD);
4035 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
4036 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
4037 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
4038 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
4039 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
4041 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4042 set_opt(sb, BARRIER);
4045 * enable delayed allocation by default
4046 * Use -o nodelalloc to turn it off
4048 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4049 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4050 set_opt(sb, DELALLOC);
4053 * set default s_li_wait_mult for lazyinit, for the case there is
4054 * no mount option specified.
4056 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
4058 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
4060 if (blocksize == PAGE_SIZE)
4061 set_opt(sb, DIOREAD_NOLOCK);
4063 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
4064 blocksize > EXT4_MAX_BLOCK_SIZE) {
4065 ext4_msg(sb, KERN_ERR,
4066 "Unsupported filesystem blocksize %d (%d log_block_size)",
4067 blocksize, le32_to_cpu(es->s_log_block_size));
4071 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4072 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4073 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4075 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4076 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4077 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4078 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4082 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4083 (!is_power_of_2(sbi->s_inode_size)) ||
4084 (sbi->s_inode_size > blocksize)) {
4085 ext4_msg(sb, KERN_ERR,
4086 "unsupported inode size: %d",
4088 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
4092 * i_atime_extra is the last extra field available for
4093 * [acm]times in struct ext4_inode. Checking for that
4094 * field should suffice to ensure we have extra space
4097 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4098 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4099 sb->s_time_gran = 1;
4100 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4102 sb->s_time_gran = NSEC_PER_SEC;
4103 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4105 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4107 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4108 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4109 EXT4_GOOD_OLD_INODE_SIZE;
4110 if (ext4_has_feature_extra_isize(sb)) {
4111 unsigned v, max = (sbi->s_inode_size -
4112 EXT4_GOOD_OLD_INODE_SIZE);
4114 v = le16_to_cpu(es->s_want_extra_isize);
4116 ext4_msg(sb, KERN_ERR,
4117 "bad s_want_extra_isize: %d", v);
4120 if (sbi->s_want_extra_isize < v)
4121 sbi->s_want_extra_isize = v;
4123 v = le16_to_cpu(es->s_min_extra_isize);
4125 ext4_msg(sb, KERN_ERR,
4126 "bad s_min_extra_isize: %d", v);
4129 if (sbi->s_want_extra_isize < v)
4130 sbi->s_want_extra_isize = v;
4134 if (sbi->s_es->s_mount_opts[0]) {
4135 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
4136 sizeof(sbi->s_es->s_mount_opts),
4140 if (!parse_options(s_mount_opts, sb, &journal_devnum,
4141 &journal_ioprio, 0)) {
4142 ext4_msg(sb, KERN_WARNING,
4143 "failed to parse options in superblock: %s",
4146 kfree(s_mount_opts);
4148 sbi->s_def_mount_opt = sbi->s_mount_opt;
4149 if (!parse_options((char *) data, sb, &journal_devnum,
4150 &journal_ioprio, 0))
4153 #ifdef CONFIG_UNICODE
4154 if (ext4_has_feature_casefold(sb) && !sbi->s_encoding) {
4155 const struct ext4_sb_encodings *encoding_info;
4156 struct unicode_map *encoding;
4157 __u16 encoding_flags;
4159 if (ext4_has_feature_encrypt(sb)) {
4160 ext4_msg(sb, KERN_ERR,
4161 "Can't mount with encoding and encryption");
4165 if (ext4_sb_read_encoding(es, &encoding_info,
4167 ext4_msg(sb, KERN_ERR,
4168 "Encoding requested by superblock is unknown");
4172 encoding = utf8_load(encoding_info->version);
4173 if (IS_ERR(encoding)) {
4174 ext4_msg(sb, KERN_ERR,
4175 "can't mount with superblock charset: %s-%s "
4176 "not supported by the kernel. flags: 0x%x.",
4177 encoding_info->name, encoding_info->version,
4181 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4182 "%s-%s with flags 0x%hx", encoding_info->name,
4183 encoding_info->version?:"\b", encoding_flags);
4185 sbi->s_encoding = encoding;
4186 sbi->s_encoding_flags = encoding_flags;
4190 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4191 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, and O_DIRECT support!\n");
4192 /* can't mount with both data=journal and dioread_nolock. */
4193 clear_opt(sb, DIOREAD_NOLOCK);
4194 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4195 ext4_msg(sb, KERN_ERR, "can't mount with "
4196 "both data=journal and delalloc");
4199 if (test_opt(sb, DAX_ALWAYS)) {
4200 ext4_msg(sb, KERN_ERR, "can't mount with "
4201 "both data=journal and dax");
4204 if (ext4_has_feature_encrypt(sb)) {
4205 ext4_msg(sb, KERN_WARNING,
4206 "encrypted files will use data=ordered "
4207 "instead of data journaling mode");
4209 if (test_opt(sb, DELALLOC))
4210 clear_opt(sb, DELALLOC);
4212 sb->s_iflags |= SB_I_CGROUPWB;
4215 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4216 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
4218 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4219 (ext4_has_compat_features(sb) ||
4220 ext4_has_ro_compat_features(sb) ||
4221 ext4_has_incompat_features(sb)))
4222 ext4_msg(sb, KERN_WARNING,
4223 "feature flags set on rev 0 fs, "
4224 "running e2fsck is recommended");
4226 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4227 set_opt2(sb, HURD_COMPAT);
4228 if (ext4_has_feature_64bit(sb)) {
4229 ext4_msg(sb, KERN_ERR,
4230 "The Hurd can't support 64-bit file systems");
4235 * ea_inode feature uses l_i_version field which is not
4236 * available in HURD_COMPAT mode.
4238 if (ext4_has_feature_ea_inode(sb)) {
4239 ext4_msg(sb, KERN_ERR,
4240 "ea_inode feature is not supported for Hurd");
4245 if (IS_EXT2_SB(sb)) {
4246 if (ext2_feature_set_ok(sb))
4247 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4248 "using the ext4 subsystem");
4251 * If we're probing be silent, if this looks like
4252 * it's actually an ext[34] filesystem.
4254 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4256 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4257 "to feature incompatibilities");
4262 if (IS_EXT3_SB(sb)) {
4263 if (ext3_feature_set_ok(sb))
4264 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4265 "using the ext4 subsystem");
4268 * If we're probing be silent, if this looks like
4269 * it's actually an ext4 filesystem.
4271 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4273 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4274 "to feature incompatibilities");
4280 * Check feature flags regardless of the revision level, since we
4281 * previously didn't change the revision level when setting the flags,
4282 * so there is a chance incompat flags are set on a rev 0 filesystem.
4284 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4287 if (le32_to_cpu(es->s_log_block_size) >
4288 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4289 ext4_msg(sb, KERN_ERR,
4290 "Invalid log block size: %u",
4291 le32_to_cpu(es->s_log_block_size));
4294 if (le32_to_cpu(es->s_log_cluster_size) >
4295 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4296 ext4_msg(sb, KERN_ERR,
4297 "Invalid log cluster size: %u",
4298 le32_to_cpu(es->s_log_cluster_size));
4302 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4303 ext4_msg(sb, KERN_ERR,
4304 "Number of reserved GDT blocks insanely large: %d",
4305 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4309 if (bdev_dax_supported(sb->s_bdev, blocksize))
4310 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4312 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4313 if (ext4_has_feature_inline_data(sb)) {
4314 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4315 " that may contain inline data");
4318 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4319 ext4_msg(sb, KERN_ERR,
4320 "DAX unsupported by block device.");
4325 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4326 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4327 es->s_encryption_level);
4331 if (sb->s_blocksize != blocksize) {
4332 /* Validate the filesystem blocksize */
4333 if (!sb_set_blocksize(sb, blocksize)) {
4334 ext4_msg(sb, KERN_ERR, "bad block size %d",
4340 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4341 offset = do_div(logical_sb_block, blocksize);
4342 bh = sb_bread_unmovable(sb, logical_sb_block);
4344 ext4_msg(sb, KERN_ERR,
4345 "Can't read superblock on 2nd try");
4348 es = (struct ext4_super_block *)(bh->b_data + offset);
4350 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
4351 ext4_msg(sb, KERN_ERR,
4352 "Magic mismatch, very weird!");
4357 has_huge_files = ext4_has_feature_huge_file(sb);
4358 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4360 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
4362 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
4363 if (ext4_has_feature_64bit(sb)) {
4364 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
4365 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
4366 !is_power_of_2(sbi->s_desc_size)) {
4367 ext4_msg(sb, KERN_ERR,
4368 "unsupported descriptor size %lu",
4373 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
4375 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
4376 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
4378 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
4379 if (sbi->s_inodes_per_block == 0)
4381 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4382 sbi->s_inodes_per_group > blocksize * 8) {
4383 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
4384 sbi->s_inodes_per_group);
4387 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4388 sbi->s_inodes_per_block;
4389 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
4391 sbi->s_mount_state = le16_to_cpu(es->s_state);
4392 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4393 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
4395 for (i = 0; i < 4; i++)
4396 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4397 sbi->s_def_hash_version = es->s_def_hash_version;
4398 if (ext4_has_feature_dir_index(sb)) {
4399 i = le32_to_cpu(es->s_flags);
4400 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4401 sbi->s_hash_unsigned = 3;
4402 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
4403 #ifdef __CHAR_UNSIGNED__
4406 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4407 sbi->s_hash_unsigned = 3;
4411 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
4416 /* Handle clustersize */
4417 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4418 if (ext4_has_feature_bigalloc(sb)) {
4419 if (clustersize < blocksize) {
4420 ext4_msg(sb, KERN_ERR,
4421 "cluster size (%d) smaller than "
4422 "block size (%d)", clustersize, blocksize);
4425 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4426 le32_to_cpu(es->s_log_block_size);
4427 sbi->s_clusters_per_group =
4428 le32_to_cpu(es->s_clusters_per_group);
4429 if (sbi->s_clusters_per_group > blocksize * 8) {
4430 ext4_msg(sb, KERN_ERR,
4431 "#clusters per group too big: %lu",
4432 sbi->s_clusters_per_group);
4435 if (sbi->s_blocks_per_group !=
4436 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4437 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4438 "clusters per group (%lu) inconsistent",
4439 sbi->s_blocks_per_group,
4440 sbi->s_clusters_per_group);
4444 if (clustersize != blocksize) {
4445 ext4_msg(sb, KERN_ERR,
4446 "fragment/cluster size (%d) != "
4447 "block size (%d)", clustersize, blocksize);
4450 if (sbi->s_blocks_per_group > blocksize * 8) {
4451 ext4_msg(sb, KERN_ERR,
4452 "#blocks per group too big: %lu",
4453 sbi->s_blocks_per_group);
4456 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4457 sbi->s_cluster_bits = 0;
4459 sbi->s_cluster_ratio = clustersize / blocksize;
4461 /* Do we have standard group size of clustersize * 8 blocks ? */
4462 if (sbi->s_blocks_per_group == clustersize << 3)
4463 set_opt2(sb, STD_GROUP_SIZE);
4466 * Test whether we have more sectors than will fit in sector_t,
4467 * and whether the max offset is addressable by the page cache.
4469 err = generic_check_addressable(sb->s_blocksize_bits,
4470 ext4_blocks_count(es));
4472 ext4_msg(sb, KERN_ERR, "filesystem"
4473 " too large to mount safely on this system");
4477 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4480 /* check blocks count against device size */
4481 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4482 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4483 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4484 "exceeds size of device (%llu blocks)",
4485 ext4_blocks_count(es), blocks_count);
4490 * It makes no sense for the first data block to be beyond the end
4491 * of the filesystem.
4493 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4494 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4495 "block %u is beyond end of filesystem (%llu)",
4496 le32_to_cpu(es->s_first_data_block),
4497 ext4_blocks_count(es));
4500 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4501 (sbi->s_cluster_ratio == 1)) {
4502 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4503 "block is 0 with a 1k block and cluster size");
4507 blocks_count = (ext4_blocks_count(es) -
4508 le32_to_cpu(es->s_first_data_block) +
4509 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4510 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4511 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4512 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4513 "(block count %llu, first data block %u, "
4514 "blocks per group %lu)", blocks_count,
4515 ext4_blocks_count(es),
4516 le32_to_cpu(es->s_first_data_block),
4517 EXT4_BLOCKS_PER_GROUP(sb));
4520 sbi->s_groups_count = blocks_count;
4521 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4522 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4523 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4524 le32_to_cpu(es->s_inodes_count)) {
4525 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4526 le32_to_cpu(es->s_inodes_count),
4527 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4531 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4532 EXT4_DESC_PER_BLOCK(sb);
4533 if (ext4_has_feature_meta_bg(sb)) {
4534 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4535 ext4_msg(sb, KERN_WARNING,
4536 "first meta block group too large: %u "
4537 "(group descriptor block count %u)",
4538 le32_to_cpu(es->s_first_meta_bg), db_count);
4542 rcu_assign_pointer(sbi->s_group_desc,
4543 kvmalloc_array(db_count,
4544 sizeof(struct buffer_head *),
4546 if (sbi->s_group_desc == NULL) {
4547 ext4_msg(sb, KERN_ERR, "not enough memory");
4552 bgl_lock_init(sbi->s_blockgroup_lock);
4554 /* Pre-read the descriptors into the buffer cache */
4555 for (i = 0; i < db_count; i++) {
4556 block = descriptor_loc(sb, logical_sb_block, i);
4557 sb_breadahead_unmovable(sb, block);
4560 for (i = 0; i < db_count; i++) {
4561 struct buffer_head *bh;
4563 block = descriptor_loc(sb, logical_sb_block, i);
4564 bh = sb_bread_unmovable(sb, block);
4566 ext4_msg(sb, KERN_ERR,
4567 "can't read group descriptor %d", i);
4572 rcu_dereference(sbi->s_group_desc)[i] = bh;
4575 sbi->s_gdb_count = db_count;
4576 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4577 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4578 ret = -EFSCORRUPTED;
4582 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
4584 /* Register extent status tree shrinker */
4585 if (ext4_es_register_shrinker(sbi))
4588 sbi->s_stripe = ext4_get_stripe_size(sbi);
4589 sbi->s_extent_max_zeroout_kb = 32;
4592 * set up enough so that it can read an inode
4594 sb->s_op = &ext4_sops;
4595 sb->s_export_op = &ext4_export_ops;
4596 sb->s_xattr = ext4_xattr_handlers;
4597 #ifdef CONFIG_FS_ENCRYPTION
4598 sb->s_cop = &ext4_cryptops;
4600 #ifdef CONFIG_FS_VERITY
4601 sb->s_vop = &ext4_verityops;
4604 sb->dq_op = &ext4_quota_operations;
4605 if (ext4_has_feature_quota(sb))
4606 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4608 sb->s_qcop = &ext4_qctl_operations;
4609 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4611 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4613 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4614 mutex_init(&sbi->s_orphan_lock);
4618 needs_recovery = (es->s_last_orphan != 0 ||
4619 ext4_has_feature_journal_needs_recovery(sb));
4621 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
4622 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4623 goto failed_mount3a;
4626 * The first inode we look at is the journal inode. Don't try
4627 * root first: it may be modified in the journal!
4629 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4630 err = ext4_load_journal(sb, es, journal_devnum);
4632 goto failed_mount3a;
4633 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4634 ext4_has_feature_journal_needs_recovery(sb)) {
4635 ext4_msg(sb, KERN_ERR, "required journal recovery "
4636 "suppressed and not mounted read-only");
4637 goto failed_mount_wq;
4639 /* Nojournal mode, all journal mount options are illegal */
4640 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4641 ext4_msg(sb, KERN_ERR, "can't mount with "
4642 "journal_checksum, fs mounted w/o journal");
4643 goto failed_mount_wq;
4645 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4646 ext4_msg(sb, KERN_ERR, "can't mount with "
4647 "journal_async_commit, fs mounted w/o journal");
4648 goto failed_mount_wq;
4650 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4651 ext4_msg(sb, KERN_ERR, "can't mount with "
4652 "commit=%lu, fs mounted w/o journal",
4653 sbi->s_commit_interval / HZ);
4654 goto failed_mount_wq;
4656 if (EXT4_MOUNT_DATA_FLAGS &
4657 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4658 ext4_msg(sb, KERN_ERR, "can't mount with "
4659 "data=, fs mounted w/o journal");
4660 goto failed_mount_wq;
4662 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
4663 clear_opt(sb, JOURNAL_CHECKSUM);
4664 clear_opt(sb, DATA_FLAGS);
4665 sbi->s_journal = NULL;
4670 if (ext4_has_feature_64bit(sb) &&
4671 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4672 JBD2_FEATURE_INCOMPAT_64BIT)) {
4673 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4674 goto failed_mount_wq;
4677 if (!set_journal_csum_feature_set(sb)) {
4678 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4680 goto failed_mount_wq;
4683 /* We have now updated the journal if required, so we can
4684 * validate the data journaling mode. */
4685 switch (test_opt(sb, DATA_FLAGS)) {
4687 /* No mode set, assume a default based on the journal
4688 * capabilities: ORDERED_DATA if the journal can
4689 * cope, else JOURNAL_DATA
4691 if (jbd2_journal_check_available_features
4692 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4693 set_opt(sb, ORDERED_DATA);
4694 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4696 set_opt(sb, JOURNAL_DATA);
4697 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4701 case EXT4_MOUNT_ORDERED_DATA:
4702 case EXT4_MOUNT_WRITEBACK_DATA:
4703 if (!jbd2_journal_check_available_features
4704 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4705 ext4_msg(sb, KERN_ERR, "Journal does not support "
4706 "requested data journaling mode");
4707 goto failed_mount_wq;
4713 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4714 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4715 ext4_msg(sb, KERN_ERR, "can't mount with "
4716 "journal_async_commit in data=ordered mode");
4717 goto failed_mount_wq;
4720 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4722 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4725 if (!test_opt(sb, NO_MBCACHE)) {
4726 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4727 if (!sbi->s_ea_block_cache) {
4728 ext4_msg(sb, KERN_ERR,
4729 "Failed to create ea_block_cache");
4730 goto failed_mount_wq;
4733 if (ext4_has_feature_ea_inode(sb)) {
4734 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4735 if (!sbi->s_ea_inode_cache) {
4736 ext4_msg(sb, KERN_ERR,
4737 "Failed to create ea_inode_cache");
4738 goto failed_mount_wq;
4743 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
4744 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
4745 goto failed_mount_wq;
4748 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
4749 !ext4_has_feature_encrypt(sb)) {
4750 ext4_set_feature_encrypt(sb);
4751 ext4_commit_super(sb, 1);
4755 * Get the # of file system overhead blocks from the
4756 * superblock if present.
4758 if (es->s_overhead_clusters)
4759 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4761 err = ext4_calculate_overhead(sb);
4763 goto failed_mount_wq;
4767 * The maximum number of concurrent works can be high and
4768 * concurrency isn't really necessary. Limit it to 1.
4770 EXT4_SB(sb)->rsv_conversion_wq =
4771 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4772 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4773 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4779 * The jbd2_journal_load will have done any necessary log recovery,
4780 * so we can safely mount the rest of the filesystem now.
4783 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
4785 ext4_msg(sb, KERN_ERR, "get root inode failed");
4786 ret = PTR_ERR(root);
4790 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4791 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4796 #ifdef CONFIG_UNICODE
4797 if (sbi->s_encoding)
4798 sb->s_d_op = &ext4_dentry_ops;
4801 sb->s_root = d_make_root(root);
4803 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4808 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4809 if (ret == -EROFS) {
4810 sb->s_flags |= SB_RDONLY;
4813 goto failed_mount4a;
4815 ext4_set_resv_clusters(sb);
4817 if (test_opt(sb, BLOCK_VALIDITY)) {
4818 err = ext4_setup_system_zone(sb);
4820 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4822 goto failed_mount4a;
4827 err = ext4_mb_init(sb);
4829 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4834 block = ext4_count_free_clusters(sb);
4835 ext4_free_blocks_count_set(sbi->s_es,
4836 EXT4_C2B(sbi, block));
4837 ext4_superblock_csum_set(sb);
4838 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4841 unsigned long freei = ext4_count_free_inodes(sb);
4842 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4843 ext4_superblock_csum_set(sb);
4844 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4848 err = percpu_counter_init(&sbi->s_dirs_counter,
4849 ext4_count_dirs(sb), GFP_KERNEL);
4851 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4854 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
4857 ext4_msg(sb, KERN_ERR, "insufficient memory");
4861 if (ext4_has_feature_flex_bg(sb))
4862 if (!ext4_fill_flex_info(sb)) {
4863 ext4_msg(sb, KERN_ERR,
4864 "unable to initialize "
4865 "flex_bg meta info!");
4869 err = ext4_register_li_request(sb, first_not_zeroed);
4873 err = ext4_register_sysfs(sb);
4878 /* Enable quota usage during mount. */
4879 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
4880 err = ext4_enable_quotas(sb);
4884 #endif /* CONFIG_QUOTA */
4887 * Save the original bdev mapping's wb_err value which could be
4888 * used to detect the metadata async write error.
4890 spin_lock_init(&sbi->s_bdev_wb_lock);
4891 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
4892 &sbi->s_bdev_wb_err);
4893 sb->s_bdev->bd_super = sb;
4894 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4895 ext4_orphan_cleanup(sb, es);
4896 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4897 if (needs_recovery) {
4898 ext4_msg(sb, KERN_INFO, "recovery complete");
4899 err = ext4_mark_recovery_complete(sb, es);
4903 if (EXT4_SB(sb)->s_journal) {
4904 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4905 descr = " journalled data mode";
4906 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4907 descr = " ordered data mode";
4909 descr = " writeback data mode";
4911 descr = "out journal";
4913 if (test_opt(sb, DISCARD)) {
4914 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4915 if (!blk_queue_discard(q))
4916 ext4_msg(sb, KERN_WARNING,
4917 "mounting with \"discard\" option, but "
4918 "the device does not support discard");
4921 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4922 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4923 "Opts: %.*s%s%s", descr,
4924 (int) sizeof(sbi->s_es->s_mount_opts),
4925 sbi->s_es->s_mount_opts,
4926 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4928 if (es->s_error_count)
4929 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4931 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4932 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4933 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4934 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4935 atomic_set(&sbi->s_warning_count, 0);
4936 atomic_set(&sbi->s_msg_count, 0);
4943 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4947 ext4_unregister_sysfs(sb);
4948 kobject_put(&sbi->s_kobj);
4950 ext4_unregister_li_request(sb);
4952 ext4_mb_release(sb);
4954 flex_groups = rcu_dereference(sbi->s_flex_groups);
4956 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
4957 kvfree(flex_groups[i]);
4958 kvfree(flex_groups);
4961 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4962 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4963 percpu_counter_destroy(&sbi->s_dirs_counter);
4964 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4965 percpu_free_rwsem(&sbi->s_writepages_rwsem);
4967 ext4_ext_release(sb);
4968 ext4_release_system_zone(sb);
4973 ext4_msg(sb, KERN_ERR, "mount failed");
4974 if (EXT4_SB(sb)->rsv_conversion_wq)
4975 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4977 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4978 sbi->s_ea_inode_cache = NULL;
4980 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4981 sbi->s_ea_block_cache = NULL;
4983 if (sbi->s_journal) {
4984 jbd2_journal_destroy(sbi->s_journal);
4985 sbi->s_journal = NULL;
4988 ext4_es_unregister_shrinker(sbi);
4990 del_timer_sync(&sbi->s_err_report);
4992 kthread_stop(sbi->s_mmp_tsk);
4995 group_desc = rcu_dereference(sbi->s_group_desc);
4996 for (i = 0; i < db_count; i++)
4997 brelse(group_desc[i]);
5001 if (sbi->s_chksum_driver)
5002 crypto_free_shash(sbi->s_chksum_driver);
5004 #ifdef CONFIG_UNICODE
5005 utf8_unload(sbi->s_encoding);
5009 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5010 kfree(get_qf_name(sb, sbi, i));
5012 fscrypt_free_dummy_context(&sbi->s_dummy_enc_ctx);
5013 ext4_blkdev_remove(sbi);
5016 sb->s_fs_info = NULL;
5017 kfree(sbi->s_blockgroup_lock);
5021 fs_put_dax(dax_dev);
5022 return err ? err : ret;
5026 * Setup any per-fs journal parameters now. We'll do this both on
5027 * initial mount, once the journal has been initialised but before we've
5028 * done any recovery; and again on any subsequent remount.
5030 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
5032 struct ext4_sb_info *sbi = EXT4_SB(sb);
5034 journal->j_commit_interval = sbi->s_commit_interval;
5035 journal->j_min_batch_time = sbi->s_min_batch_time;
5036 journal->j_max_batch_time = sbi->s_max_batch_time;
5038 write_lock(&journal->j_state_lock);
5039 if (test_opt(sb, BARRIER))
5040 journal->j_flags |= JBD2_BARRIER;
5042 journal->j_flags &= ~JBD2_BARRIER;
5043 if (test_opt(sb, DATA_ERR_ABORT))
5044 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5046 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
5047 write_unlock(&journal->j_state_lock);
5050 static struct inode *ext4_get_journal_inode(struct super_block *sb,
5051 unsigned int journal_inum)
5053 struct inode *journal_inode;
5056 * Test for the existence of a valid inode on disk. Bad things
5057 * happen if we iget() an unused inode, as the subsequent iput()
5058 * will try to delete it.
5060 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
5061 if (IS_ERR(journal_inode)) {
5062 ext4_msg(sb, KERN_ERR, "no journal found");
5065 if (!journal_inode->i_nlink) {
5066 make_bad_inode(journal_inode);
5067 iput(journal_inode);
5068 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
5072 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
5073 journal_inode, journal_inode->i_size);
5074 if (!S_ISREG(journal_inode->i_mode)) {
5075 ext4_msg(sb, KERN_ERR, "invalid journal inode");
5076 iput(journal_inode);
5079 return journal_inode;
5082 static journal_t *ext4_get_journal(struct super_block *sb,
5083 unsigned int journal_inum)
5085 struct inode *journal_inode;
5088 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5091 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5095 journal = jbd2_journal_init_inode(journal_inode);
5097 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5098 iput(journal_inode);
5101 journal->j_private = sb;
5102 ext4_init_journal_params(sb, journal);
5106 static journal_t *ext4_get_dev_journal(struct super_block *sb,
5109 struct buffer_head *bh;
5113 int hblock, blocksize;
5114 ext4_fsblk_t sb_block;
5115 unsigned long offset;
5116 struct ext4_super_block *es;
5117 struct block_device *bdev;
5119 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5122 bdev = ext4_blkdev_get(j_dev, sb);
5126 blocksize = sb->s_blocksize;
5127 hblock = bdev_logical_block_size(bdev);
5128 if (blocksize < hblock) {
5129 ext4_msg(sb, KERN_ERR,
5130 "blocksize too small for journal device");
5134 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5135 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5136 set_blocksize(bdev, blocksize);
5137 if (!(bh = __bread(bdev, sb_block, blocksize))) {
5138 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5139 "external journal");
5143 es = (struct ext4_super_block *) (bh->b_data + offset);
5144 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5145 !(le32_to_cpu(es->s_feature_incompat) &
5146 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5147 ext4_msg(sb, KERN_ERR, "external journal has "
5153 if ((le32_to_cpu(es->s_feature_ro_compat) &
5154 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5155 es->s_checksum != ext4_superblock_csum(sb, es)) {
5156 ext4_msg(sb, KERN_ERR, "external journal has "
5157 "corrupt superblock");
5162 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5163 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5168 len = ext4_blocks_count(es);
5169 start = sb_block + 1;
5170 brelse(bh); /* we're done with the superblock */
5172 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
5173 start, len, blocksize);
5175 ext4_msg(sb, KERN_ERR, "failed to create device journal");
5178 journal->j_private = sb;
5179 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
5180 wait_on_buffer(journal->j_sb_buffer);
5181 if (!buffer_uptodate(journal->j_sb_buffer)) {
5182 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
5185 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5186 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5187 "user (unsupported) - %d",
5188 be32_to_cpu(journal->j_superblock->s_nr_users));
5191 EXT4_SB(sb)->s_journal_bdev = bdev;
5192 ext4_init_journal_params(sb, journal);
5196 jbd2_journal_destroy(journal);
5198 ext4_blkdev_put(bdev);
5202 static int ext4_load_journal(struct super_block *sb,
5203 struct ext4_super_block *es,
5204 unsigned long journal_devnum)
5207 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5210 int really_read_only;
5213 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5214 return -EFSCORRUPTED;
5216 if (journal_devnum &&
5217 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5218 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5219 "numbers have changed");
5220 journal_dev = new_decode_dev(journal_devnum);
5222 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5224 if (journal_inum && journal_dev) {
5225 ext4_msg(sb, KERN_ERR,
5226 "filesystem has both journal inode and journal device!");
5231 journal = ext4_get_journal(sb, journal_inum);
5235 journal = ext4_get_dev_journal(sb, journal_dev);
5240 journal_dev_ro = bdev_read_only(journal->j_dev);
5241 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5243 if (journal_dev_ro && !sb_rdonly(sb)) {
5244 ext4_msg(sb, KERN_ERR,
5245 "journal device read-only, try mounting with '-o ro'");
5251 * Are we loading a blank journal or performing recovery after a
5252 * crash? For recovery, we need to check in advance whether we
5253 * can get read-write access to the device.
5255 if (ext4_has_feature_journal_needs_recovery(sb)) {
5256 if (sb_rdonly(sb)) {
5257 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5258 "required on readonly filesystem");
5259 if (really_read_only) {
5260 ext4_msg(sb, KERN_ERR, "write access "
5261 "unavailable, cannot proceed "
5262 "(try mounting with noload)");
5266 ext4_msg(sb, KERN_INFO, "write access will "
5267 "be enabled during recovery");
5271 if (!(journal->j_flags & JBD2_BARRIER))
5272 ext4_msg(sb, KERN_INFO, "barriers disabled");
5274 if (!ext4_has_feature_journal_needs_recovery(sb))
5275 err = jbd2_journal_wipe(journal, !really_read_only);
5277 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5279 memcpy(save, ((char *) es) +
5280 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
5281 err = jbd2_journal_load(journal);
5283 memcpy(((char *) es) + EXT4_S_ERR_START,
5284 save, EXT4_S_ERR_LEN);
5289 ext4_msg(sb, KERN_ERR, "error loading journal");
5293 EXT4_SB(sb)->s_journal = journal;
5294 err = ext4_clear_journal_err(sb, es);
5296 EXT4_SB(sb)->s_journal = NULL;
5297 jbd2_journal_destroy(journal);
5301 if (!really_read_only && journal_devnum &&
5302 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5303 es->s_journal_dev = cpu_to_le32(journal_devnum);
5305 /* Make sure we flush the recovery flag to disk. */
5306 ext4_commit_super(sb, 1);
5312 jbd2_journal_destroy(journal);
5316 static int ext4_commit_super(struct super_block *sb, int sync)
5318 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
5319 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
5322 if (!sbh || block_device_ejected(sb))
5326 * If the file system is mounted read-only, don't update the
5327 * superblock write time. This avoids updating the superblock
5328 * write time when we are mounting the root file system
5329 * read/only but we need to replay the journal; at that point,
5330 * for people who are east of GMT and who make their clock
5331 * tick in localtime for Windows bug-for-bug compatibility,
5332 * the clock is set in the future, and this will cause e2fsck
5333 * to complain and force a full file system check.
5335 if (!(sb->s_flags & SB_RDONLY))
5336 ext4_update_tstamp(es, s_wtime);
5337 if (sb->s_bdev->bd_part)
5338 es->s_kbytes_written =
5339 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
5340 ((part_stat_read(sb->s_bdev->bd_part,
5341 sectors[STAT_WRITE]) -
5342 EXT4_SB(sb)->s_sectors_written_start) >> 1));
5344 es->s_kbytes_written =
5345 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
5346 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
5347 ext4_free_blocks_count_set(es,
5348 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
5349 &EXT4_SB(sb)->s_freeclusters_counter)));
5350 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
5351 es->s_free_inodes_count =
5352 cpu_to_le32(percpu_counter_sum_positive(
5353 &EXT4_SB(sb)->s_freeinodes_counter));
5354 BUFFER_TRACE(sbh, "marking dirty");
5355 ext4_superblock_csum_set(sb);
5358 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
5360 * Oh, dear. A previous attempt to write the
5361 * superblock failed. This could happen because the
5362 * USB device was yanked out. Or it could happen to
5363 * be a transient write error and maybe the block will
5364 * be remapped. Nothing we can do but to retry the
5365 * write and hope for the best.
5367 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5368 "superblock detected");
5369 clear_buffer_write_io_error(sbh);
5370 set_buffer_uptodate(sbh);
5372 mark_buffer_dirty(sbh);
5375 error = __sync_dirty_buffer(sbh,
5376 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
5377 if (buffer_write_io_error(sbh)) {
5378 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5380 clear_buffer_write_io_error(sbh);
5381 set_buffer_uptodate(sbh);
5388 * Have we just finished recovery? If so, and if we are mounting (or
5389 * remounting) the filesystem readonly, then we will end up with a
5390 * consistent fs on disk. Record that fact.
5392 static int ext4_mark_recovery_complete(struct super_block *sb,
5393 struct ext4_super_block *es)
5396 journal_t *journal = EXT4_SB(sb)->s_journal;
5398 if (!ext4_has_feature_journal(sb)) {
5399 if (journal != NULL) {
5400 ext4_error(sb, "Journal got removed while the fs was "
5402 return -EFSCORRUPTED;
5406 jbd2_journal_lock_updates(journal);
5407 err = jbd2_journal_flush(journal);
5411 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
5412 ext4_clear_feature_journal_needs_recovery(sb);
5413 ext4_commit_super(sb, 1);
5416 jbd2_journal_unlock_updates(journal);
5421 * If we are mounting (or read-write remounting) a filesystem whose journal
5422 * has recorded an error from a previous lifetime, move that error to the
5423 * main filesystem now.
5425 static int ext4_clear_journal_err(struct super_block *sb,
5426 struct ext4_super_block *es)
5432 if (!ext4_has_feature_journal(sb)) {
5433 ext4_error(sb, "Journal got removed while the fs was mounted!");
5434 return -EFSCORRUPTED;
5437 journal = EXT4_SB(sb)->s_journal;
5440 * Now check for any error status which may have been recorded in the
5441 * journal by a prior ext4_error() or ext4_abort()
5444 j_errno = jbd2_journal_errno(journal);
5448 errstr = ext4_decode_error(sb, j_errno, nbuf);
5449 ext4_warning(sb, "Filesystem error recorded "
5450 "from previous mount: %s", errstr);
5451 ext4_warning(sb, "Marking fs in need of filesystem check.");
5453 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5454 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5455 ext4_commit_super(sb, 1);
5457 jbd2_journal_clear_err(journal);
5458 jbd2_journal_update_sb_errno(journal);
5464 * Force the running and committing transactions to commit,
5465 * and wait on the commit.
5467 int ext4_force_commit(struct super_block *sb)
5474 journal = EXT4_SB(sb)->s_journal;
5475 return ext4_journal_force_commit(journal);
5478 static int ext4_sync_fs(struct super_block *sb, int wait)
5482 bool needs_barrier = false;
5483 struct ext4_sb_info *sbi = EXT4_SB(sb);
5485 if (unlikely(ext4_forced_shutdown(sbi)))
5488 trace_ext4_sync_fs(sb, wait);
5489 flush_workqueue(sbi->rsv_conversion_wq);
5491 * Writeback quota in non-journalled quota case - journalled quota has
5494 dquot_writeback_dquots(sb, -1);
5496 * Data writeback is possible w/o journal transaction, so barrier must
5497 * being sent at the end of the function. But we can skip it if
5498 * transaction_commit will do it for us.
5500 if (sbi->s_journal) {
5501 target = jbd2_get_latest_transaction(sbi->s_journal);
5502 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5503 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5504 needs_barrier = true;
5506 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5508 ret = jbd2_log_wait_commit(sbi->s_journal,
5511 } else if (wait && test_opt(sb, BARRIER))
5512 needs_barrier = true;
5513 if (needs_barrier) {
5515 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL);
5524 * LVM calls this function before a (read-only) snapshot is created. This
5525 * gives us a chance to flush the journal completely and mark the fs clean.
5527 * Note that only this function cannot bring a filesystem to be in a clean
5528 * state independently. It relies on upper layer to stop all data & metadata
5531 static int ext4_freeze(struct super_block *sb)
5539 journal = EXT4_SB(sb)->s_journal;
5542 /* Now we set up the journal barrier. */
5543 jbd2_journal_lock_updates(journal);
5546 * Don't clear the needs_recovery flag if we failed to
5547 * flush the journal.
5549 error = jbd2_journal_flush(journal);
5553 /* Journal blocked and flushed, clear needs_recovery flag. */
5554 ext4_clear_feature_journal_needs_recovery(sb);
5557 error = ext4_commit_super(sb, 1);
5560 /* we rely on upper layer to stop further updates */
5561 jbd2_journal_unlock_updates(journal);
5566 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5567 * flag here, even though the filesystem is not technically dirty yet.
5569 static int ext4_unfreeze(struct super_block *sb)
5571 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
5574 if (EXT4_SB(sb)->s_journal) {
5575 /* Reset the needs_recovery flag before the fs is unlocked. */
5576 ext4_set_feature_journal_needs_recovery(sb);
5579 ext4_commit_super(sb, 1);
5584 * Structure to save mount options for ext4_remount's benefit
5586 struct ext4_mount_options {
5587 unsigned long s_mount_opt;
5588 unsigned long s_mount_opt2;
5591 unsigned long s_commit_interval;
5592 u32 s_min_batch_time, s_max_batch_time;
5595 char *s_qf_names[EXT4_MAXQUOTAS];
5599 static int ext4_remount(struct super_block *sb, int *flags, char *data)
5601 struct ext4_super_block *es;
5602 struct ext4_sb_info *sbi = EXT4_SB(sb);
5603 unsigned long old_sb_flags, vfs_flags;
5604 struct ext4_mount_options old_opts;
5605 int enable_quota = 0;
5607 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5611 char *to_free[EXT4_MAXQUOTAS];
5613 char *orig_data = kstrdup(data, GFP_KERNEL);
5615 if (data && !orig_data)
5618 /* Store the original options */
5619 old_sb_flags = sb->s_flags;
5620 old_opts.s_mount_opt = sbi->s_mount_opt;
5621 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
5622 old_opts.s_resuid = sbi->s_resuid;
5623 old_opts.s_resgid = sbi->s_resgid;
5624 old_opts.s_commit_interval = sbi->s_commit_interval;
5625 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5626 old_opts.s_max_batch_time = sbi->s_max_batch_time;
5628 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
5629 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5630 if (sbi->s_qf_names[i]) {
5631 char *qf_name = get_qf_name(sb, sbi, i);
5633 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
5634 if (!old_opts.s_qf_names[i]) {
5635 for (j = 0; j < i; j++)
5636 kfree(old_opts.s_qf_names[j]);
5641 old_opts.s_qf_names[i] = NULL;
5643 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5644 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
5647 * Some options can be enabled by ext4 and/or by VFS mount flag
5648 * either way we need to make sure it matches in both *flags and
5649 * s_flags. Copy those selected flags from *flags to s_flags
5651 vfs_flags = SB_LAZYTIME | SB_I_VERSION;
5652 sb->s_flags = (sb->s_flags & ~vfs_flags) | (*flags & vfs_flags);
5654 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
5659 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5660 test_opt(sb, JOURNAL_CHECKSUM)) {
5661 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5662 "during remount not supported; ignoring");
5663 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5666 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5667 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5668 ext4_msg(sb, KERN_ERR, "can't mount with "
5669 "both data=journal and delalloc");
5673 if (test_opt(sb, DIOREAD_NOLOCK)) {
5674 ext4_msg(sb, KERN_ERR, "can't mount with "
5675 "both data=journal and dioread_nolock");
5679 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5680 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5681 ext4_msg(sb, KERN_ERR, "can't mount with "
5682 "journal_async_commit in data=ordered mode");
5688 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5689 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5694 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
5695 ext4_abort(sb, EXT4_ERR_ESHUTDOWN, "Abort forced by user");
5697 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5698 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5702 if (sbi->s_journal) {
5703 ext4_init_journal_params(sb, sbi->s_journal);
5704 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5707 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
5708 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
5713 if (*flags & SB_RDONLY) {
5714 err = sync_filesystem(sb);
5717 err = dquot_suspend(sb, -1);
5722 * First of all, the unconditional stuff we have to do
5723 * to disable replay of the journal when we next remount
5725 sb->s_flags |= SB_RDONLY;
5728 * OK, test if we are remounting a valid rw partition
5729 * readonly, and if so set the rdonly flag and then
5730 * mark the partition as valid again.
5732 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5733 (sbi->s_mount_state & EXT4_VALID_FS))
5734 es->s_state = cpu_to_le16(sbi->s_mount_state);
5736 if (sbi->s_journal) {
5738 * We let remount-ro finish even if marking fs
5739 * as clean failed...
5741 ext4_mark_recovery_complete(sb, es);
5744 kthread_stop(sbi->s_mmp_tsk);
5746 /* Make sure we can mount this feature set readwrite */
5747 if (ext4_has_feature_readonly(sb) ||
5748 !ext4_feature_set_ok(sb, 0)) {
5753 * Make sure the group descriptor checksums
5754 * are sane. If they aren't, refuse to remount r/w.
5756 for (g = 0; g < sbi->s_groups_count; g++) {
5757 struct ext4_group_desc *gdp =
5758 ext4_get_group_desc(sb, g, NULL);
5760 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5761 ext4_msg(sb, KERN_ERR,
5762 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5763 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5764 le16_to_cpu(gdp->bg_checksum));
5771 * If we have an unprocessed orphan list hanging
5772 * around from a previously readonly bdev mount,
5773 * require a full umount/remount for now.
5775 if (es->s_last_orphan) {
5776 ext4_msg(sb, KERN_WARNING, "Couldn't "
5777 "remount RDWR because of unprocessed "
5778 "orphan inode list. Please "
5779 "umount/remount instead");
5785 * Mounting a RDONLY partition read-write, so reread
5786 * and store the current valid flag. (It may have
5787 * been changed by e2fsck since we originally mounted
5790 if (sbi->s_journal) {
5791 err = ext4_clear_journal_err(sb, es);
5795 sbi->s_mount_state = le16_to_cpu(es->s_state);
5797 err = ext4_setup_super(sb, es, 0);
5801 sb->s_flags &= ~SB_RDONLY;
5802 if (ext4_has_feature_mmp(sb))
5803 if (ext4_multi_mount_protect(sb,
5804 le64_to_cpu(es->s_mmp_block))) {
5813 * Reinitialize lazy itable initialization thread based on
5816 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
5817 ext4_unregister_li_request(sb);
5819 ext4_group_t first_not_zeroed;
5820 first_not_zeroed = ext4_has_uninit_itable(sb);
5821 ext4_register_li_request(sb, first_not_zeroed);
5825 * Handle creation of system zone data early because it can fail.
5826 * Releasing of existing data is done when we are sure remount will
5829 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
5830 err = ext4_setup_system_zone(sb);
5835 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5836 err = ext4_commit_super(sb, 1);
5842 /* Release old quota file names */
5843 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5844 kfree(old_opts.s_qf_names[i]);
5846 if (sb_any_quota_suspended(sb))
5847 dquot_resume(sb, -1);
5848 else if (ext4_has_feature_quota(sb)) {
5849 err = ext4_enable_quotas(sb);
5855 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
5856 ext4_release_system_zone(sb);
5859 * Some options can be enabled by ext4 and/or by VFS mount flag
5860 * either way we need to make sure it matches in both *flags and
5861 * s_flags. Copy those selected flags from s_flags to *flags
5863 *flags = (*flags & ~vfs_flags) | (sb->s_flags & vfs_flags);
5865 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5870 sb->s_flags = old_sb_flags;
5871 sbi->s_mount_opt = old_opts.s_mount_opt;
5872 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5873 sbi->s_resuid = old_opts.s_resuid;
5874 sbi->s_resgid = old_opts.s_resgid;
5875 sbi->s_commit_interval = old_opts.s_commit_interval;
5876 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5877 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5878 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
5879 ext4_release_system_zone(sb);
5881 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5882 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5883 to_free[i] = get_qf_name(sb, sbi, i);
5884 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
5887 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5895 static int ext4_statfs_project(struct super_block *sb,
5896 kprojid_t projid, struct kstatfs *buf)
5899 struct dquot *dquot;
5903 qid = make_kqid_projid(projid);
5904 dquot = dqget(sb, qid);
5906 return PTR_ERR(dquot);
5907 spin_lock(&dquot->dq_dqb_lock);
5909 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
5910 dquot->dq_dqb.dqb_bhardlimit);
5911 limit >>= sb->s_blocksize_bits;
5913 if (limit && buf->f_blocks > limit) {
5914 curblock = (dquot->dq_dqb.dqb_curspace +
5915 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
5916 buf->f_blocks = limit;
5917 buf->f_bfree = buf->f_bavail =
5918 (buf->f_blocks > curblock) ?
5919 (buf->f_blocks - curblock) : 0;
5922 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
5923 dquot->dq_dqb.dqb_ihardlimit);
5924 if (limit && buf->f_files > limit) {
5925 buf->f_files = limit;
5927 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5928 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5931 spin_unlock(&dquot->dq_dqb_lock);
5937 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5939 struct super_block *sb = dentry->d_sb;
5940 struct ext4_sb_info *sbi = EXT4_SB(sb);
5941 struct ext4_super_block *es = sbi->s_es;
5942 ext4_fsblk_t overhead = 0, resv_blocks;
5945 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5947 if (!test_opt(sb, MINIX_DF))
5948 overhead = sbi->s_overhead;
5950 buf->f_type = EXT4_SUPER_MAGIC;
5951 buf->f_bsize = sb->s_blocksize;
5952 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5953 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5954 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5955 /* prevent underflow in case that few free space is available */
5956 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5957 buf->f_bavail = buf->f_bfree -
5958 (ext4_r_blocks_count(es) + resv_blocks);
5959 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5961 buf->f_files = le32_to_cpu(es->s_inodes_count);
5962 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5963 buf->f_namelen = EXT4_NAME_LEN;
5964 fsid = le64_to_cpup((void *)es->s_uuid) ^
5965 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5966 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5967 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5970 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5971 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5972 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5981 * Helper functions so that transaction is started before we acquire dqio_sem
5982 * to keep correct lock ordering of transaction > dqio_sem
5984 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5986 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5989 static int ext4_write_dquot(struct dquot *dquot)
5993 struct inode *inode;
5995 inode = dquot_to_inode(dquot);
5996 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5997 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5999 return PTR_ERR(handle);
6000 ret = dquot_commit(dquot);
6001 err = ext4_journal_stop(handle);
6007 static int ext4_acquire_dquot(struct dquot *dquot)
6012 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6013 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
6015 return PTR_ERR(handle);
6016 ret = dquot_acquire(dquot);
6017 err = ext4_journal_stop(handle);
6023 static int ext4_release_dquot(struct dquot *dquot)
6028 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6029 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
6030 if (IS_ERR(handle)) {
6031 /* Release dquot anyway to avoid endless cycle in dqput() */
6032 dquot_release(dquot);
6033 return PTR_ERR(handle);
6035 ret = dquot_release(dquot);
6036 err = ext4_journal_stop(handle);
6042 static int ext4_mark_dquot_dirty(struct dquot *dquot)
6044 struct super_block *sb = dquot->dq_sb;
6045 struct ext4_sb_info *sbi = EXT4_SB(sb);
6047 /* Are we journaling quotas? */
6048 if (ext4_has_feature_quota(sb) ||
6049 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
6050 dquot_mark_dquot_dirty(dquot);
6051 return ext4_write_dquot(dquot);
6053 return dquot_mark_dquot_dirty(dquot);
6057 static int ext4_write_info(struct super_block *sb, int type)
6062 /* Data block + inode block */
6063 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
6065 return PTR_ERR(handle);
6066 ret = dquot_commit_info(sb, type);
6067 err = ext4_journal_stop(handle);
6074 * Turn on quotas during mount time - we need to find
6075 * the quota file and such...
6077 static int ext4_quota_on_mount(struct super_block *sb, int type)
6079 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
6080 EXT4_SB(sb)->s_jquota_fmt, type);
6083 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6085 struct ext4_inode_info *ei = EXT4_I(inode);
6087 /* The first argument of lockdep_set_subclass has to be
6088 * *exactly* the same as the argument to init_rwsem() --- in
6089 * this case, in init_once() --- or lockdep gets unhappy
6090 * because the name of the lock is set using the
6091 * stringification of the argument to init_rwsem().
6093 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6094 lockdep_set_subclass(&ei->i_data_sem, subclass);
6098 * Standard function to be called on quota_on
6100 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6101 const struct path *path)
6105 if (!test_opt(sb, QUOTA))
6108 /* Quotafile not on the same filesystem? */
6109 if (path->dentry->d_sb != sb)
6111 /* Journaling quota? */
6112 if (EXT4_SB(sb)->s_qf_names[type]) {
6113 /* Quotafile not in fs root? */
6114 if (path->dentry->d_parent != sb->s_root)
6115 ext4_msg(sb, KERN_WARNING,
6116 "Quota file not on filesystem root. "
6117 "Journaled quota will not work");
6118 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6121 * Clear the flag just in case mount options changed since
6124 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
6128 * When we journal data on quota file, we have to flush journal to see
6129 * all updates to the file when we bypass pagecache...
6131 if (EXT4_SB(sb)->s_journal &&
6132 ext4_should_journal_data(d_inode(path->dentry))) {
6134 * We don't need to lock updates but journal_flush() could
6135 * otherwise be livelocked...
6137 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
6138 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
6139 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
6144 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6145 err = dquot_quota_on(sb, type, format_id, path);
6147 lockdep_set_quota_inode(path->dentry->d_inode,
6150 struct inode *inode = d_inode(path->dentry);
6154 * Set inode flags to prevent userspace from messing with quota
6155 * files. If this fails, we return success anyway since quotas
6156 * are already enabled and this is not a hard failure.
6159 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6162 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6163 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6164 S_NOATIME | S_IMMUTABLE);
6165 err = ext4_mark_inode_dirty(handle, inode);
6166 ext4_journal_stop(handle);
6168 inode_unlock(inode);
6173 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6177 struct inode *qf_inode;
6178 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6179 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6180 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6181 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6184 BUG_ON(!ext4_has_feature_quota(sb));
6186 if (!qf_inums[type])
6189 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
6190 if (IS_ERR(qf_inode)) {
6191 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
6192 return PTR_ERR(qf_inode);
6195 /* Don't account quota for quota files to avoid recursion */
6196 qf_inode->i_flags |= S_NOQUOTA;
6197 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
6198 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
6200 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
6206 /* Enable usage tracking for all quota types. */
6207 static int ext4_enable_quotas(struct super_block *sb)
6210 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6211 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6212 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6213 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6215 bool quota_mopt[EXT4_MAXQUOTAS] = {
6216 test_opt(sb, USRQUOTA),
6217 test_opt(sb, GRPQUOTA),
6218 test_opt(sb, PRJQUOTA),
6221 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
6222 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
6223 if (qf_inums[type]) {
6224 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
6225 DQUOT_USAGE_ENABLED |
6226 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
6229 "Failed to enable quota tracking "
6230 "(type=%d, err=%d). Please run "
6231 "e2fsck to fix.", type, err);
6232 for (type--; type >= 0; type--)
6233 dquot_quota_off(sb, type);
6242 static int ext4_quota_off(struct super_block *sb, int type)
6244 struct inode *inode = sb_dqopt(sb)->files[type];
6248 /* Force all delayed allocation blocks to be allocated.
6249 * Caller already holds s_umount sem */
6250 if (test_opt(sb, DELALLOC))
6251 sync_filesystem(sb);
6253 if (!inode || !igrab(inode))
6256 err = dquot_quota_off(sb, type);
6257 if (err || ext4_has_feature_quota(sb))
6262 * Update modification times of quota files when userspace can
6263 * start looking at them. If we fail, we return success anyway since
6264 * this is not a hard failure and quotas are already disabled.
6266 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6267 if (IS_ERR(handle)) {
6268 err = PTR_ERR(handle);
6271 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
6272 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
6273 inode->i_mtime = inode->i_ctime = current_time(inode);
6274 err = ext4_mark_inode_dirty(handle, inode);
6275 ext4_journal_stop(handle);
6277 inode_unlock(inode);
6279 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
6283 return dquot_quota_off(sb, type);
6286 /* Read data from quotafile - avoid pagecache and such because we cannot afford
6287 * acquiring the locks... As quota files are never truncated and quota code
6288 * itself serializes the operations (and no one else should touch the files)
6289 * we don't have to be afraid of races */
6290 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
6291 size_t len, loff_t off)
6293 struct inode *inode = sb_dqopt(sb)->files[type];
6294 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6295 int offset = off & (sb->s_blocksize - 1);
6298 struct buffer_head *bh;
6299 loff_t i_size = i_size_read(inode);
6303 if (off+len > i_size)
6306 while (toread > 0) {
6307 tocopy = sb->s_blocksize - offset < toread ?
6308 sb->s_blocksize - offset : toread;
6309 bh = ext4_bread(NULL, inode, blk, 0);
6312 if (!bh) /* A hole? */
6313 memset(data, 0, tocopy);
6315 memcpy(data, bh->b_data+offset, tocopy);
6325 /* Write to quotafile (we know the transaction is already started and has
6326 * enough credits) */
6327 static ssize_t ext4_quota_write(struct super_block *sb, int type,
6328 const char *data, size_t len, loff_t off)
6330 struct inode *inode = sb_dqopt(sb)->files[type];
6331 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6332 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
6334 struct buffer_head *bh;
6335 handle_t *handle = journal_current_handle();
6337 if (EXT4_SB(sb)->s_journal && !handle) {
6338 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6339 " cancelled because transaction is not started",
6340 (unsigned long long)off, (unsigned long long)len);
6344 * Since we account only one data block in transaction credits,
6345 * then it is impossible to cross a block boundary.
6347 if (sb->s_blocksize - offset < len) {
6348 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6349 " cancelled because not block aligned",
6350 (unsigned long long)off, (unsigned long long)len);
6355 bh = ext4_bread(handle, inode, blk,
6356 EXT4_GET_BLOCKS_CREATE |
6357 EXT4_GET_BLOCKS_METADATA_NOFAIL);
6358 } while (PTR_ERR(bh) == -ENOSPC &&
6359 ext4_should_retry_alloc(inode->i_sb, &retries));
6364 BUFFER_TRACE(bh, "get write access");
6365 err = ext4_journal_get_write_access(handle, bh);
6371 memcpy(bh->b_data+offset, data, len);
6372 flush_dcache_page(bh->b_page);
6374 err = ext4_handle_dirty_metadata(handle, NULL, bh);
6377 if (inode->i_size < off + len) {
6378 i_size_write(inode, off + len);
6379 EXT4_I(inode)->i_disksize = inode->i_size;
6380 err2 = ext4_mark_inode_dirty(handle, inode);
6381 if (unlikely(err2 && !err))
6384 return err ? err : len;
6388 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6389 const char *dev_name, void *data)
6391 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
6394 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
6395 static inline void register_as_ext2(void)
6397 int err = register_filesystem(&ext2_fs_type);
6400 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6403 static inline void unregister_as_ext2(void)
6405 unregister_filesystem(&ext2_fs_type);
6408 static inline int ext2_feature_set_ok(struct super_block *sb)
6410 if (ext4_has_unknown_ext2_incompat_features(sb))
6414 if (ext4_has_unknown_ext2_ro_compat_features(sb))
6419 static inline void register_as_ext2(void) { }
6420 static inline void unregister_as_ext2(void) { }
6421 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
6424 static inline void register_as_ext3(void)
6426 int err = register_filesystem(&ext3_fs_type);
6429 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6432 static inline void unregister_as_ext3(void)
6434 unregister_filesystem(&ext3_fs_type);
6437 static inline int ext3_feature_set_ok(struct super_block *sb)
6439 if (ext4_has_unknown_ext3_incompat_features(sb))
6441 if (!ext4_has_feature_journal(sb))
6445 if (ext4_has_unknown_ext3_ro_compat_features(sb))
6450 static struct file_system_type ext4_fs_type = {
6451 .owner = THIS_MODULE,
6453 .mount = ext4_mount,
6454 .kill_sb = kill_block_super,
6455 .fs_flags = FS_REQUIRES_DEV,
6457 MODULE_ALIAS_FS("ext4");
6459 /* Shared across all ext4 file systems */
6460 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
6462 static int __init ext4_init_fs(void)
6466 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
6467 ext4_li_info = NULL;
6468 mutex_init(&ext4_li_mtx);
6470 /* Build-time check for flags consistency */
6471 ext4_check_flag_values();
6473 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
6474 init_waitqueue_head(&ext4__ioend_wq[i]);
6476 err = ext4_init_es();
6480 err = ext4_init_pending();
6484 err = ext4_init_post_read_processing();
6488 err = ext4_init_pageio();
6492 err = ext4_init_system_zone();
6496 err = ext4_init_sysfs();
6500 err = ext4_init_mballoc();
6503 err = init_inodecache();
6508 err = register_filesystem(&ext4_fs_type);
6514 unregister_as_ext2();
6515 unregister_as_ext3();
6516 destroy_inodecache();
6518 ext4_exit_mballoc();
6522 ext4_exit_system_zone();
6526 ext4_exit_post_read_processing();
6528 ext4_exit_pending();
6535 static void __exit ext4_exit_fs(void)
6537 ext4_destroy_lazyinit_thread();
6538 unregister_as_ext2();
6539 unregister_as_ext3();
6540 unregister_filesystem(&ext4_fs_type);
6541 destroy_inodecache();
6542 ext4_exit_mballoc();
6544 ext4_exit_system_zone();
6546 ext4_exit_post_read_processing();
6548 ext4_exit_pending();
6551 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6552 MODULE_DESCRIPTION("Fourth Extended Filesystem");
6553 MODULE_LICENSE("GPL");
6554 MODULE_SOFTDEP("pre: crc32c");
6555 module_init(ext4_init_fs)
6556 module_exit(ext4_exit_fs)