1 // SPDX-License-Identifier: GPL-2.0
3 * linux/fs/ext4/super.c
5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI)
12 * linux/fs/minix/inode.c
14 * Copyright (C) 1991, 1992 Linus Torvalds
16 * Big-endian to little-endian byte-swapping/bitmaps by
17 * David S. Miller (davem@caip.rutgers.edu), 1995
20 #include <linux/module.h>
21 #include <linux/string.h>
23 #include <linux/time.h>
24 #include <linux/vmalloc.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/backing-dev.h>
29 #include <linux/parser.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/dax.h>
42 #include <linux/cleancache.h>
43 #include <linux/uaccess.h>
44 #include <linux/iversion.h>
45 #include <linux/unicode.h>
46 #include <linux/part_stat.h>
47 #include <linux/kthread.h>
48 #include <linux/freezer.h>
51 #include "ext4_extents.h" /* Needed for trace points definition */
52 #include "ext4_jbd2.h"
58 #define CREATE_TRACE_POINTS
59 #include <trace/events/ext4.h>
61 static struct ext4_lazy_init *ext4_li_info;
62 static struct mutex ext4_li_mtx;
63 static struct ratelimit_state ext4_mount_msg_ratelimit;
65 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
66 unsigned long journal_devnum);
67 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
68 static int ext4_commit_super(struct super_block *sb, int sync);
69 static int ext4_mark_recovery_complete(struct super_block *sb,
70 struct ext4_super_block *es);
71 static int ext4_clear_journal_err(struct super_block *sb,
72 struct ext4_super_block *es);
73 static int ext4_sync_fs(struct super_block *sb, int wait);
74 static int ext4_remount(struct super_block *sb, int *flags, char *data);
75 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
76 static int ext4_unfreeze(struct super_block *sb);
77 static int ext4_freeze(struct super_block *sb);
78 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
80 static inline int ext2_feature_set_ok(struct super_block *sb);
81 static inline int ext3_feature_set_ok(struct super_block *sb);
82 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
83 static void ext4_destroy_lazyinit_thread(void);
84 static void ext4_unregister_li_request(struct super_block *sb);
85 static void ext4_clear_request_list(void);
86 static struct inode *ext4_get_journal_inode(struct super_block *sb,
87 unsigned int journal_inum);
92 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
93 * i_mmap_rwsem (inode->i_mmap_rwsem)!
96 * mmap_lock -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
97 * page lock -> i_data_sem (rw)
99 * buffered write path:
100 * sb_start_write -> i_mutex -> mmap_lock
101 * sb_start_write -> i_mutex -> transaction start -> page lock ->
105 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock
106 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start ->
110 * sb_start_write -> i_mutex -> mmap_lock
111 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
114 * transaction start -> page lock(s) -> i_data_sem (rw)
117 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
118 static struct file_system_type ext2_fs_type = {
119 .owner = THIS_MODULE,
122 .kill_sb = kill_block_super,
123 .fs_flags = FS_REQUIRES_DEV,
125 MODULE_ALIAS_FS("ext2");
126 MODULE_ALIAS("ext2");
127 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
129 #define IS_EXT2_SB(sb) (0)
133 static struct file_system_type ext3_fs_type = {
134 .owner = THIS_MODULE,
137 .kill_sb = kill_block_super,
138 .fs_flags = FS_REQUIRES_DEV,
140 MODULE_ALIAS_FS("ext3");
141 MODULE_ALIAS("ext3");
142 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
145 static inline void __ext4_read_bh(struct buffer_head *bh, int op_flags,
149 * buffer's verified bit is no longer valid after reading from
150 * disk again due to write out error, clear it to make sure we
151 * recheck the buffer contents.
153 clear_buffer_verified(bh);
155 bh->b_end_io = end_io ? end_io : end_buffer_read_sync;
157 submit_bh(REQ_OP_READ, op_flags, bh);
160 void ext4_read_bh_nowait(struct buffer_head *bh, int op_flags,
163 BUG_ON(!buffer_locked(bh));
165 if (ext4_buffer_uptodate(bh)) {
169 __ext4_read_bh(bh, op_flags, end_io);
172 int ext4_read_bh(struct buffer_head *bh, int op_flags, bh_end_io_t *end_io)
174 BUG_ON(!buffer_locked(bh));
176 if (ext4_buffer_uptodate(bh)) {
181 __ext4_read_bh(bh, op_flags, end_io);
184 if (buffer_uptodate(bh))
189 int ext4_read_bh_lock(struct buffer_head *bh, int op_flags, bool wait)
191 if (trylock_buffer(bh)) {
193 return ext4_read_bh(bh, op_flags, NULL);
194 ext4_read_bh_nowait(bh, op_flags, NULL);
199 if (buffer_uptodate(bh))
207 * This works like __bread_gfp() except it uses ERR_PTR for error
208 * returns. Currently with sb_bread it's impossible to distinguish
209 * between ENOMEM and EIO situations (since both result in a NULL
212 static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb,
213 sector_t block, int op_flags,
216 struct buffer_head *bh;
219 bh = sb_getblk_gfp(sb, block, gfp);
221 return ERR_PTR(-ENOMEM);
222 if (ext4_buffer_uptodate(bh))
225 ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true);
233 struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block,
236 return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE);
239 struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
242 return __ext4_sb_bread_gfp(sb, block, 0, 0);
245 void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block)
247 struct buffer_head *bh = sb_getblk_gfp(sb, block, 0);
250 ext4_read_bh_lock(bh, REQ_RAHEAD, false);
255 static int ext4_verify_csum_type(struct super_block *sb,
256 struct ext4_super_block *es)
258 if (!ext4_has_feature_metadata_csum(sb))
261 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
264 static __le32 ext4_superblock_csum(struct super_block *sb,
265 struct ext4_super_block *es)
267 struct ext4_sb_info *sbi = EXT4_SB(sb);
268 int offset = offsetof(struct ext4_super_block, s_checksum);
271 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
273 return cpu_to_le32(csum);
276 static int ext4_superblock_csum_verify(struct super_block *sb,
277 struct ext4_super_block *es)
279 if (!ext4_has_metadata_csum(sb))
282 return es->s_checksum == ext4_superblock_csum(sb, es);
285 void ext4_superblock_csum_set(struct super_block *sb)
287 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
289 if (!ext4_has_metadata_csum(sb))
293 * Locking the superblock prevents the scenario
295 * 1) a first thread pauses during checksum calculation.
296 * 2) a second thread updates the superblock, recalculates
297 * the checksum, and updates s_checksum
298 * 3) the first thread resumes and finishes its checksum calculation
299 * and updates s_checksum with a potentially stale or torn value.
301 lock_buffer(EXT4_SB(sb)->s_sbh);
302 es->s_checksum = ext4_superblock_csum(sb, es);
303 unlock_buffer(EXT4_SB(sb)->s_sbh);
306 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
307 struct ext4_group_desc *bg)
309 return le32_to_cpu(bg->bg_block_bitmap_lo) |
310 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
311 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
314 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
315 struct ext4_group_desc *bg)
317 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
318 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
319 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
322 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
323 struct ext4_group_desc *bg)
325 return le32_to_cpu(bg->bg_inode_table_lo) |
326 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
327 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
330 __u32 ext4_free_group_clusters(struct super_block *sb,
331 struct ext4_group_desc *bg)
333 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
334 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
335 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
338 __u32 ext4_free_inodes_count(struct super_block *sb,
339 struct ext4_group_desc *bg)
341 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
342 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
343 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
346 __u32 ext4_used_dirs_count(struct super_block *sb,
347 struct ext4_group_desc *bg)
349 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
350 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
351 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
354 __u32 ext4_itable_unused_count(struct super_block *sb,
355 struct ext4_group_desc *bg)
357 return le16_to_cpu(bg->bg_itable_unused_lo) |
358 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
359 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
362 void ext4_block_bitmap_set(struct super_block *sb,
363 struct ext4_group_desc *bg, ext4_fsblk_t blk)
365 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
366 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
367 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
370 void ext4_inode_bitmap_set(struct super_block *sb,
371 struct ext4_group_desc *bg, ext4_fsblk_t blk)
373 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
374 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
375 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
378 void ext4_inode_table_set(struct super_block *sb,
379 struct ext4_group_desc *bg, ext4_fsblk_t blk)
381 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
382 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
383 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
386 void ext4_free_group_clusters_set(struct super_block *sb,
387 struct ext4_group_desc *bg, __u32 count)
389 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
390 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
391 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
394 void ext4_free_inodes_set(struct super_block *sb,
395 struct ext4_group_desc *bg, __u32 count)
397 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
398 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
399 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
402 void ext4_used_dirs_set(struct super_block *sb,
403 struct ext4_group_desc *bg, __u32 count)
405 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
406 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
407 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
410 void ext4_itable_unused_set(struct super_block *sb,
411 struct ext4_group_desc *bg, __u32 count)
413 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
414 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
415 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
418 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi)
420 time64_t now = ktime_get_real_seconds();
422 now = clamp_val(now, 0, (1ull << 40) - 1);
424 *lo = cpu_to_le32(lower_32_bits(now));
425 *hi = upper_32_bits(now);
428 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
430 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
432 #define ext4_update_tstamp(es, tstamp) \
433 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
434 #define ext4_get_tstamp(es, tstamp) \
435 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
437 static void __save_error_info(struct super_block *sb, int error,
438 __u32 ino, __u64 block,
439 const char *func, unsigned int line)
441 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
444 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
445 if (bdev_read_only(sb->s_bdev))
447 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
448 ext4_update_tstamp(es, s_last_error_time);
449 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
450 es->s_last_error_line = cpu_to_le32(line);
451 es->s_last_error_ino = cpu_to_le32(ino);
452 es->s_last_error_block = cpu_to_le64(block);
458 err = EXT4_ERR_ENOMEM;
461 err = EXT4_ERR_EFSBADCRC;
465 err = EXT4_ERR_EFSCORRUPTED;
468 err = EXT4_ERR_ENOSPC;
471 err = EXT4_ERR_ENOKEY;
474 err = EXT4_ERR_EROFS;
477 err = EXT4_ERR_EFBIG;
480 err = EXT4_ERR_EEXIST;
483 err = EXT4_ERR_ERANGE;
486 err = EXT4_ERR_EOVERFLOW;
489 err = EXT4_ERR_EBUSY;
492 err = EXT4_ERR_ENOTDIR;
495 err = EXT4_ERR_ENOTEMPTY;
498 err = EXT4_ERR_ESHUTDOWN;
501 err = EXT4_ERR_EFAULT;
504 err = EXT4_ERR_UNKNOWN;
506 es->s_last_error_errcode = err;
507 if (!es->s_first_error_time) {
508 es->s_first_error_time = es->s_last_error_time;
509 es->s_first_error_time_hi = es->s_last_error_time_hi;
510 strncpy(es->s_first_error_func, func,
511 sizeof(es->s_first_error_func));
512 es->s_first_error_line = cpu_to_le32(line);
513 es->s_first_error_ino = es->s_last_error_ino;
514 es->s_first_error_block = es->s_last_error_block;
515 es->s_first_error_errcode = es->s_last_error_errcode;
518 * Start the daily error reporting function if it hasn't been
521 if (!es->s_error_count)
522 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
523 le32_add_cpu(&es->s_error_count, 1);
526 static void save_error_info(struct super_block *sb, int error,
527 __u32 ino, __u64 block,
528 const char *func, unsigned int line)
530 __save_error_info(sb, error, ino, block, func, line);
531 if (!bdev_read_only(sb->s_bdev))
532 ext4_commit_super(sb, 1);
536 * The del_gendisk() function uninitializes the disk-specific data
537 * structures, including the bdi structure, without telling anyone
538 * else. Once this happens, any attempt to call mark_buffer_dirty()
539 * (for example, by ext4_commit_super), will cause a kernel OOPS.
540 * This is a kludge to prevent these oops until we can put in a proper
541 * hook in del_gendisk() to inform the VFS and file system layers.
543 static int block_device_ejected(struct super_block *sb)
545 struct inode *bd_inode = sb->s_bdev->bd_inode;
546 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
548 return bdi->dev == NULL;
551 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
553 struct super_block *sb = journal->j_private;
554 struct ext4_sb_info *sbi = EXT4_SB(sb);
555 int error = is_journal_aborted(journal);
556 struct ext4_journal_cb_entry *jce;
558 BUG_ON(txn->t_state == T_FINISHED);
560 ext4_process_freed_data(sb, txn->t_tid);
562 spin_lock(&sbi->s_md_lock);
563 while (!list_empty(&txn->t_private_list)) {
564 jce = list_entry(txn->t_private_list.next,
565 struct ext4_journal_cb_entry, jce_list);
566 list_del_init(&jce->jce_list);
567 spin_unlock(&sbi->s_md_lock);
568 jce->jce_func(sb, jce, error);
569 spin_lock(&sbi->s_md_lock);
571 spin_unlock(&sbi->s_md_lock);
575 * This writepage callback for write_cache_pages()
576 * takes care of a few cases after page cleaning.
578 * write_cache_pages() already checks for dirty pages
579 * and calls clear_page_dirty_for_io(), which we want,
580 * to write protect the pages.
582 * However, we may have to redirty a page (see below.)
584 static int ext4_journalled_writepage_callback(struct page *page,
585 struct writeback_control *wbc,
588 transaction_t *transaction = (transaction_t *) data;
589 struct buffer_head *bh, *head;
590 struct journal_head *jh;
592 bh = head = page_buffers(page);
595 * We have to redirty a page in these cases:
596 * 1) If buffer is dirty, it means the page was dirty because it
597 * contains a buffer that needs checkpointing. So the dirty bit
598 * needs to be preserved so that checkpointing writes the buffer
600 * 2) If buffer is not part of the committing transaction
601 * (we may have just accidentally come across this buffer because
602 * inode range tracking is not exact) or if the currently running
603 * transaction already contains this buffer as well, dirty bit
604 * needs to be preserved so that the buffer gets writeprotected
605 * properly on running transaction's commit.
608 if (buffer_dirty(bh) ||
609 (jh && (jh->b_transaction != transaction ||
610 jh->b_next_transaction))) {
611 redirty_page_for_writepage(wbc, page);
614 } while ((bh = bh->b_this_page) != head);
617 return AOP_WRITEPAGE_ACTIVATE;
620 static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
622 struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
623 struct writeback_control wbc = {
624 .sync_mode = WB_SYNC_ALL,
625 .nr_to_write = LONG_MAX,
626 .range_start = jinode->i_dirty_start,
627 .range_end = jinode->i_dirty_end,
630 return write_cache_pages(mapping, &wbc,
631 ext4_journalled_writepage_callback,
632 jinode->i_transaction);
635 static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
639 if (ext4_should_journal_data(jinode->i_vfs_inode))
640 ret = ext4_journalled_submit_inode_data_buffers(jinode);
642 ret = jbd2_journal_submit_inode_data_buffers(jinode);
647 static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
651 if (!ext4_should_journal_data(jinode->i_vfs_inode))
652 ret = jbd2_journal_finish_inode_data_buffers(jinode);
657 static bool system_going_down(void)
659 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
660 || system_state == SYSTEM_RESTART;
663 /* Deal with the reporting of failure conditions on a filesystem such as
664 * inconsistencies detected or read IO failures.
666 * On ext2, we can store the error state of the filesystem in the
667 * superblock. That is not possible on ext4, because we may have other
668 * write ordering constraints on the superblock which prevent us from
669 * writing it out straight away; and given that the journal is about to
670 * be aborted, we can't rely on the current, or future, transactions to
671 * write out the superblock safely.
673 * We'll just use the jbd2_journal_abort() error code to record an error in
674 * the journal instead. On recovery, the journal will complain about
675 * that error until we've noted it down and cleared it.
678 static void ext4_handle_error(struct super_block *sb)
680 if (test_opt(sb, WARN_ON_ERROR))
686 if (!test_opt(sb, ERRORS_CONT)) {
687 journal_t *journal = EXT4_SB(sb)->s_journal;
689 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
691 jbd2_journal_abort(journal, -EIO);
694 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
695 * could panic during 'reboot -f' as the underlying device got already
698 if (test_opt(sb, ERRORS_RO) || system_going_down()) {
699 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
701 * Make sure updated value of ->s_mount_flags will be visible
702 * before ->s_flags update
705 sb->s_flags |= SB_RDONLY;
706 } else if (test_opt(sb, ERRORS_PANIC)) {
707 panic("EXT4-fs (device %s): panic forced after error\n",
712 #define ext4_error_ratelimit(sb) \
713 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
716 void __ext4_error(struct super_block *sb, const char *function,
717 unsigned int line, int error, __u64 block,
718 const char *fmt, ...)
720 struct va_format vaf;
723 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
726 trace_ext4_error(sb, function, line);
727 if (ext4_error_ratelimit(sb)) {
732 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
733 sb->s_id, function, line, current->comm, &vaf);
736 save_error_info(sb, error, 0, block, function, line);
737 ext4_handle_error(sb);
740 void __ext4_error_inode(struct inode *inode, const char *function,
741 unsigned int line, ext4_fsblk_t block, int error,
742 const char *fmt, ...)
745 struct va_format vaf;
747 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
750 trace_ext4_error(inode->i_sb, function, line);
751 if (ext4_error_ratelimit(inode->i_sb)) {
756 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
757 "inode #%lu: block %llu: comm %s: %pV\n",
758 inode->i_sb->s_id, function, line, inode->i_ino,
759 block, current->comm, &vaf);
761 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
762 "inode #%lu: comm %s: %pV\n",
763 inode->i_sb->s_id, function, line, inode->i_ino,
764 current->comm, &vaf);
767 save_error_info(inode->i_sb, error, inode->i_ino, block,
769 ext4_handle_error(inode->i_sb);
772 void __ext4_error_file(struct file *file, const char *function,
773 unsigned int line, ext4_fsblk_t block,
774 const char *fmt, ...)
777 struct va_format vaf;
778 struct inode *inode = file_inode(file);
779 char pathname[80], *path;
781 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
784 trace_ext4_error(inode->i_sb, function, line);
785 if (ext4_error_ratelimit(inode->i_sb)) {
786 path = file_path(file, pathname, sizeof(pathname));
794 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
795 "block %llu: comm %s: path %s: %pV\n",
796 inode->i_sb->s_id, function, line, inode->i_ino,
797 block, current->comm, path, &vaf);
800 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
801 "comm %s: path %s: %pV\n",
802 inode->i_sb->s_id, function, line, inode->i_ino,
803 current->comm, path, &vaf);
806 save_error_info(inode->i_sb, EFSCORRUPTED, inode->i_ino, block,
808 ext4_handle_error(inode->i_sb);
811 const char *ext4_decode_error(struct super_block *sb, int errno,
818 errstr = "Corrupt filesystem";
821 errstr = "Filesystem failed CRC";
824 errstr = "IO failure";
827 errstr = "Out of memory";
830 if (!sb || (EXT4_SB(sb)->s_journal &&
831 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
832 errstr = "Journal has aborted";
834 errstr = "Readonly filesystem";
837 /* If the caller passed in an extra buffer for unknown
838 * errors, textualise them now. Else we just return
841 /* Check for truncated error codes... */
842 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
851 /* __ext4_std_error decodes expected errors from journaling functions
852 * automatically and invokes the appropriate error response. */
854 void __ext4_std_error(struct super_block *sb, const char *function,
855 unsigned int line, int errno)
860 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
863 /* Special case: if the error is EROFS, and we're not already
864 * inside a transaction, then there's really no point in logging
866 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
869 if (ext4_error_ratelimit(sb)) {
870 errstr = ext4_decode_error(sb, errno, nbuf);
871 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
872 sb->s_id, function, line, errstr);
875 save_error_info(sb, -errno, 0, 0, function, line);
876 ext4_handle_error(sb);
880 * ext4_abort is a much stronger failure handler than ext4_error. The
881 * abort function may be used to deal with unrecoverable failures such
882 * as journal IO errors or ENOMEM at a critical moment in log management.
884 * We unconditionally force the filesystem into an ABORT|READONLY state,
885 * unless the error response on the fs has been set to panic in which
886 * case we take the easy way out and panic immediately.
889 void __ext4_abort(struct super_block *sb, const char *function,
890 unsigned int line, int error, const char *fmt, ...)
892 struct va_format vaf;
895 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
898 save_error_info(sb, error, 0, 0, function, line);
902 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
903 sb->s_id, function, line, &vaf);
906 if (sb_rdonly(sb) == 0) {
907 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
908 if (EXT4_SB(sb)->s_journal)
909 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
911 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
913 * Make sure updated value of ->s_mount_flags will be visible
914 * before ->s_flags update
917 sb->s_flags |= SB_RDONLY;
919 if (test_opt(sb, ERRORS_PANIC) && !system_going_down())
920 panic("EXT4-fs panic from previous error\n");
923 void __ext4_msg(struct super_block *sb,
924 const char *prefix, const char *fmt, ...)
926 struct va_format vaf;
929 atomic_inc(&EXT4_SB(sb)->s_msg_count);
930 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
936 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
940 static int ext4_warning_ratelimit(struct super_block *sb)
942 atomic_inc(&EXT4_SB(sb)->s_warning_count);
943 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
947 void __ext4_warning(struct super_block *sb, const char *function,
948 unsigned int line, const char *fmt, ...)
950 struct va_format vaf;
953 if (!ext4_warning_ratelimit(sb))
959 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
960 sb->s_id, function, line, &vaf);
964 void __ext4_warning_inode(const struct inode *inode, const char *function,
965 unsigned int line, const char *fmt, ...)
967 struct va_format vaf;
970 if (!ext4_warning_ratelimit(inode->i_sb))
976 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
977 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
978 function, line, inode->i_ino, current->comm, &vaf);
982 void __ext4_grp_locked_error(const char *function, unsigned int line,
983 struct super_block *sb, ext4_group_t grp,
984 unsigned long ino, ext4_fsblk_t block,
985 const char *fmt, ...)
989 struct va_format vaf;
992 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
995 trace_ext4_error(sb, function, line);
996 __save_error_info(sb, EFSCORRUPTED, ino, block, function, line);
998 if (ext4_error_ratelimit(sb)) {
1002 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
1003 sb->s_id, function, line, grp);
1005 printk(KERN_CONT "inode %lu: ", ino);
1007 printk(KERN_CONT "block %llu:",
1008 (unsigned long long) block);
1009 printk(KERN_CONT "%pV\n", &vaf);
1013 if (test_opt(sb, WARN_ON_ERROR))
1016 if (test_opt(sb, ERRORS_CONT)) {
1017 ext4_commit_super(sb, 0);
1021 ext4_unlock_group(sb, grp);
1022 ext4_commit_super(sb, 1);
1023 ext4_handle_error(sb);
1025 * We only get here in the ERRORS_RO case; relocking the group
1026 * may be dangerous, but nothing bad will happen since the
1027 * filesystem will have already been marked read/only and the
1028 * journal has been aborted. We return 1 as a hint to callers
1029 * who might what to use the return value from
1030 * ext4_grp_locked_error() to distinguish between the
1031 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
1032 * aggressively from the ext4 function in question, with a
1033 * more appropriate error code.
1035 ext4_lock_group(sb, grp);
1039 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
1043 struct ext4_sb_info *sbi = EXT4_SB(sb);
1044 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1045 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
1048 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1049 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1052 percpu_counter_sub(&sbi->s_freeclusters_counter,
1056 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1057 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1062 count = ext4_free_inodes_count(sb, gdp);
1063 percpu_counter_sub(&sbi->s_freeinodes_counter,
1069 void ext4_update_dynamic_rev(struct super_block *sb)
1071 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
1073 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
1077 "updating to rev %d because of new feature flag, "
1078 "running e2fsck is recommended",
1081 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1082 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1083 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
1084 /* leave es->s_feature_*compat flags alone */
1085 /* es->s_uuid will be set by e2fsck if empty */
1088 * The rest of the superblock fields should be zero, and if not it
1089 * means they are likely already in use, so leave them alone. We
1090 * can leave it up to e2fsck to clean up any inconsistencies there.
1095 * Open the external journal device
1097 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
1099 struct block_device *bdev;
1101 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
1107 ext4_msg(sb, KERN_ERR,
1108 "failed to open journal device unknown-block(%u,%u) %ld",
1109 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
1114 * Release the journal device
1116 static void ext4_blkdev_put(struct block_device *bdev)
1118 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
1121 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
1123 struct block_device *bdev;
1124 bdev = sbi->s_journal_bdev;
1126 ext4_blkdev_put(bdev);
1127 sbi->s_journal_bdev = NULL;
1131 static inline struct inode *orphan_list_entry(struct list_head *l)
1133 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1136 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1138 struct list_head *l;
1140 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1141 le32_to_cpu(sbi->s_es->s_last_orphan));
1143 printk(KERN_ERR "sb_info orphan list:\n");
1144 list_for_each(l, &sbi->s_orphan) {
1145 struct inode *inode = orphan_list_entry(l);
1147 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1148 inode->i_sb->s_id, inode->i_ino, inode,
1149 inode->i_mode, inode->i_nlink,
1150 NEXT_ORPHAN(inode));
1155 static int ext4_quota_off(struct super_block *sb, int type);
1157 static inline void ext4_quota_off_umount(struct super_block *sb)
1161 /* Use our quota_off function to clear inode flags etc. */
1162 for (type = 0; type < EXT4_MAXQUOTAS; type++)
1163 ext4_quota_off(sb, type);
1167 * This is a helper function which is used in the mount/remount
1168 * codepaths (which holds s_umount) to fetch the quota file name.
1170 static inline char *get_qf_name(struct super_block *sb,
1171 struct ext4_sb_info *sbi,
1174 return rcu_dereference_protected(sbi->s_qf_names[type],
1175 lockdep_is_held(&sb->s_umount));
1178 static inline void ext4_quota_off_umount(struct super_block *sb)
1183 static void ext4_put_super(struct super_block *sb)
1185 struct ext4_sb_info *sbi = EXT4_SB(sb);
1186 struct ext4_super_block *es = sbi->s_es;
1187 struct buffer_head **group_desc;
1188 struct flex_groups **flex_groups;
1192 ext4_unregister_li_request(sb);
1193 ext4_quota_off_umount(sb);
1195 destroy_workqueue(sbi->rsv_conversion_wq);
1198 * Unregister sysfs before destroying jbd2 journal.
1199 * Since we could still access attr_journal_task attribute via sysfs
1200 * path which could have sbi->s_journal->j_task as NULL
1202 ext4_unregister_sysfs(sb);
1204 if (sbi->s_journal) {
1205 aborted = is_journal_aborted(sbi->s_journal);
1206 err = jbd2_journal_destroy(sbi->s_journal);
1207 sbi->s_journal = NULL;
1208 if ((err < 0) && !aborted) {
1209 ext4_abort(sb, -err, "Couldn't clean up the journal");
1213 ext4_es_unregister_shrinker(sbi);
1214 del_timer_sync(&sbi->s_err_report);
1215 ext4_release_system_zone(sb);
1216 ext4_mb_release(sb);
1217 ext4_ext_release(sb);
1219 if (!sb_rdonly(sb) && !aborted) {
1220 ext4_clear_feature_journal_needs_recovery(sb);
1221 es->s_state = cpu_to_le16(sbi->s_mount_state);
1224 ext4_commit_super(sb, 1);
1227 group_desc = rcu_dereference(sbi->s_group_desc);
1228 for (i = 0; i < sbi->s_gdb_count; i++)
1229 brelse(group_desc[i]);
1231 flex_groups = rcu_dereference(sbi->s_flex_groups);
1233 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1234 kvfree(flex_groups[i]);
1235 kvfree(flex_groups);
1238 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1239 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1240 percpu_counter_destroy(&sbi->s_dirs_counter);
1241 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1242 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1244 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1245 kfree(get_qf_name(sb, sbi, i));
1248 /* Debugging code just in case the in-memory inode orphan list
1249 * isn't empty. The on-disk one can be non-empty if we've
1250 * detected an error and taken the fs readonly, but the
1251 * in-memory list had better be clean by this point. */
1252 if (!list_empty(&sbi->s_orphan))
1253 dump_orphan_list(sb, sbi);
1254 J_ASSERT(list_empty(&sbi->s_orphan));
1256 sync_blockdev(sb->s_bdev);
1257 invalidate_bdev(sb->s_bdev);
1258 if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) {
1260 * Invalidate the journal device's buffers. We don't want them
1261 * floating about in memory - the physical journal device may
1262 * hotswapped, and it breaks the `ro-after' testing code.
1264 sync_blockdev(sbi->s_journal_bdev);
1265 invalidate_bdev(sbi->s_journal_bdev);
1266 ext4_blkdev_remove(sbi);
1269 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1270 sbi->s_ea_inode_cache = NULL;
1272 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1273 sbi->s_ea_block_cache = NULL;
1276 kthread_stop(sbi->s_mmp_tsk);
1278 sb->s_fs_info = NULL;
1280 * Now that we are completely done shutting down the
1281 * superblock, we need to actually destroy the kobject.
1283 kobject_put(&sbi->s_kobj);
1284 wait_for_completion(&sbi->s_kobj_unregister);
1285 if (sbi->s_chksum_driver)
1286 crypto_free_shash(sbi->s_chksum_driver);
1287 kfree(sbi->s_blockgroup_lock);
1288 fs_put_dax(sbi->s_daxdev);
1289 fscrypt_free_dummy_context(&sbi->s_dummy_enc_ctx);
1290 #ifdef CONFIG_UNICODE
1291 utf8_unload(sbi->s_encoding);
1296 static struct kmem_cache *ext4_inode_cachep;
1299 * Called inside transaction, so use GFP_NOFS
1301 static struct inode *ext4_alloc_inode(struct super_block *sb)
1303 struct ext4_inode_info *ei;
1305 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1309 inode_set_iversion(&ei->vfs_inode, 1);
1310 spin_lock_init(&ei->i_raw_lock);
1311 INIT_LIST_HEAD(&ei->i_prealloc_list);
1312 atomic_set(&ei->i_prealloc_active, 0);
1313 spin_lock_init(&ei->i_prealloc_lock);
1314 ext4_es_init_tree(&ei->i_es_tree);
1315 rwlock_init(&ei->i_es_lock);
1316 INIT_LIST_HEAD(&ei->i_es_list);
1317 ei->i_es_all_nr = 0;
1318 ei->i_es_shk_nr = 0;
1319 ei->i_es_shrink_lblk = 0;
1320 ei->i_reserved_data_blocks = 0;
1321 spin_lock_init(&(ei->i_block_reservation_lock));
1322 ext4_init_pending_tree(&ei->i_pending_tree);
1324 ei->i_reserved_quota = 0;
1325 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1328 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1329 spin_lock_init(&ei->i_completed_io_lock);
1331 ei->i_datasync_tid = 0;
1332 atomic_set(&ei->i_unwritten, 0);
1333 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1334 return &ei->vfs_inode;
1337 static int ext4_drop_inode(struct inode *inode)
1339 int drop = generic_drop_inode(inode);
1342 drop = fscrypt_drop_inode(inode);
1344 trace_ext4_drop_inode(inode, drop);
1348 static void ext4_free_in_core_inode(struct inode *inode)
1350 fscrypt_free_inode(inode);
1351 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1354 static void ext4_destroy_inode(struct inode *inode)
1356 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1357 ext4_msg(inode->i_sb, KERN_ERR,
1358 "Inode %lu (%p): orphan list check failed!",
1359 inode->i_ino, EXT4_I(inode));
1360 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1361 EXT4_I(inode), sizeof(struct ext4_inode_info),
1367 static void init_once(void *foo)
1369 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1371 INIT_LIST_HEAD(&ei->i_orphan);
1372 init_rwsem(&ei->xattr_sem);
1373 init_rwsem(&ei->i_data_sem);
1374 init_rwsem(&ei->i_mmap_sem);
1375 inode_init_once(&ei->vfs_inode);
1378 static int __init init_inodecache(void)
1380 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1381 sizeof(struct ext4_inode_info), 0,
1382 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1384 offsetof(struct ext4_inode_info, i_data),
1385 sizeof_field(struct ext4_inode_info, i_data),
1387 if (ext4_inode_cachep == NULL)
1392 static void destroy_inodecache(void)
1395 * Make sure all delayed rcu free inodes are flushed before we
1399 kmem_cache_destroy(ext4_inode_cachep);
1402 void ext4_clear_inode(struct inode *inode)
1404 invalidate_inode_buffers(inode);
1406 ext4_discard_preallocations(inode, 0);
1407 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1409 if (EXT4_I(inode)->jinode) {
1410 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1411 EXT4_I(inode)->jinode);
1412 jbd2_free_inode(EXT4_I(inode)->jinode);
1413 EXT4_I(inode)->jinode = NULL;
1415 fscrypt_put_encryption_info(inode);
1416 fsverity_cleanup_inode(inode);
1419 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1420 u64 ino, u32 generation)
1422 struct inode *inode;
1425 * Currently we don't know the generation for parent directory, so
1426 * a generation of 0 means "accept any"
1428 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1430 return ERR_CAST(inode);
1431 if (generation && inode->i_generation != generation) {
1433 return ERR_PTR(-ESTALE);
1439 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1440 int fh_len, int fh_type)
1442 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1443 ext4_nfs_get_inode);
1446 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1447 int fh_len, int fh_type)
1449 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1450 ext4_nfs_get_inode);
1453 static int ext4_nfs_commit_metadata(struct inode *inode)
1455 struct writeback_control wbc = {
1456 .sync_mode = WB_SYNC_ALL
1459 trace_ext4_nfs_commit_metadata(inode);
1460 return ext4_write_inode(inode, &wbc);
1464 * Try to release metadata pages (indirect blocks, directories) which are
1465 * mapped via the block device. Since these pages could have journal heads
1466 * which would prevent try_to_free_buffers() from freeing them, we must use
1467 * jbd2 layer's try_to_free_buffers() function to release them.
1469 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1472 journal_t *journal = EXT4_SB(sb)->s_journal;
1474 WARN_ON(PageChecked(page));
1475 if (!page_has_buffers(page))
1478 return jbd2_journal_try_to_free_buffers(journal, page);
1480 return try_to_free_buffers(page);
1483 #ifdef CONFIG_FS_ENCRYPTION
1484 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1486 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1487 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1490 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1493 handle_t *handle = fs_data;
1494 int res, res2, credits, retries = 0;
1497 * Encrypting the root directory is not allowed because e2fsck expects
1498 * lost+found to exist and be unencrypted, and encrypting the root
1499 * directory would imply encrypting the lost+found directory as well as
1500 * the filename "lost+found" itself.
1502 if (inode->i_ino == EXT4_ROOT_INO)
1505 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1508 if (ext4_test_inode_flag(inode, EXT4_INODE_DAX))
1511 res = ext4_convert_inline_data(inode);
1516 * If a journal handle was specified, then the encryption context is
1517 * being set on a new inode via inheritance and is part of a larger
1518 * transaction to create the inode. Otherwise the encryption context is
1519 * being set on an existing inode in its own transaction. Only in the
1520 * latter case should the "retry on ENOSPC" logic be used.
1524 res = ext4_xattr_set_handle(handle, inode,
1525 EXT4_XATTR_INDEX_ENCRYPTION,
1526 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1529 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1530 ext4_clear_inode_state(inode,
1531 EXT4_STATE_MAY_INLINE_DATA);
1533 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1534 * S_DAX may be disabled
1536 ext4_set_inode_flags(inode, false);
1541 res = dquot_initialize(inode);
1545 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1550 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1552 return PTR_ERR(handle);
1554 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1555 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1558 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1560 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1561 * S_DAX may be disabled
1563 ext4_set_inode_flags(inode, false);
1564 res = ext4_mark_inode_dirty(handle, inode);
1566 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1568 res2 = ext4_journal_stop(handle);
1570 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1577 static const union fscrypt_context *
1578 ext4_get_dummy_context(struct super_block *sb)
1580 return EXT4_SB(sb)->s_dummy_enc_ctx.ctx;
1583 static bool ext4_has_stable_inodes(struct super_block *sb)
1585 return ext4_has_feature_stable_inodes(sb);
1588 static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1589 int *ino_bits_ret, int *lblk_bits_ret)
1591 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1592 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1595 static const struct fscrypt_operations ext4_cryptops = {
1596 .key_prefix = "ext4:",
1597 .get_context = ext4_get_context,
1598 .set_context = ext4_set_context,
1599 .get_dummy_context = ext4_get_dummy_context,
1600 .empty_dir = ext4_empty_dir,
1601 .max_namelen = EXT4_NAME_LEN,
1602 .has_stable_inodes = ext4_has_stable_inodes,
1603 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits,
1608 static const char * const quotatypes[] = INITQFNAMES;
1609 #define QTYPE2NAME(t) (quotatypes[t])
1611 static int ext4_write_dquot(struct dquot *dquot);
1612 static int ext4_acquire_dquot(struct dquot *dquot);
1613 static int ext4_release_dquot(struct dquot *dquot);
1614 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1615 static int ext4_write_info(struct super_block *sb, int type);
1616 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1617 const struct path *path);
1618 static int ext4_quota_on_mount(struct super_block *sb, int type);
1619 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1620 size_t len, loff_t off);
1621 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1622 const char *data, size_t len, loff_t off);
1623 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1624 unsigned int flags);
1625 static int ext4_enable_quotas(struct super_block *sb);
1627 static struct dquot **ext4_get_dquots(struct inode *inode)
1629 return EXT4_I(inode)->i_dquot;
1632 static const struct dquot_operations ext4_quota_operations = {
1633 .get_reserved_space = ext4_get_reserved_space,
1634 .write_dquot = ext4_write_dquot,
1635 .acquire_dquot = ext4_acquire_dquot,
1636 .release_dquot = ext4_release_dquot,
1637 .mark_dirty = ext4_mark_dquot_dirty,
1638 .write_info = ext4_write_info,
1639 .alloc_dquot = dquot_alloc,
1640 .destroy_dquot = dquot_destroy,
1641 .get_projid = ext4_get_projid,
1642 .get_inode_usage = ext4_get_inode_usage,
1643 .get_next_id = dquot_get_next_id,
1646 static const struct quotactl_ops ext4_qctl_operations = {
1647 .quota_on = ext4_quota_on,
1648 .quota_off = ext4_quota_off,
1649 .quota_sync = dquot_quota_sync,
1650 .get_state = dquot_get_state,
1651 .set_info = dquot_set_dqinfo,
1652 .get_dqblk = dquot_get_dqblk,
1653 .set_dqblk = dquot_set_dqblk,
1654 .get_nextdqblk = dquot_get_next_dqblk,
1658 static const struct super_operations ext4_sops = {
1659 .alloc_inode = ext4_alloc_inode,
1660 .free_inode = ext4_free_in_core_inode,
1661 .destroy_inode = ext4_destroy_inode,
1662 .write_inode = ext4_write_inode,
1663 .dirty_inode = ext4_dirty_inode,
1664 .drop_inode = ext4_drop_inode,
1665 .evict_inode = ext4_evict_inode,
1666 .put_super = ext4_put_super,
1667 .sync_fs = ext4_sync_fs,
1668 .freeze_fs = ext4_freeze,
1669 .unfreeze_fs = ext4_unfreeze,
1670 .statfs = ext4_statfs,
1671 .remount_fs = ext4_remount,
1672 .show_options = ext4_show_options,
1674 .quota_read = ext4_quota_read,
1675 .quota_write = ext4_quota_write,
1676 .get_dquots = ext4_get_dquots,
1678 .bdev_try_to_free_page = bdev_try_to_free_page,
1681 static const struct export_operations ext4_export_ops = {
1682 .fh_to_dentry = ext4_fh_to_dentry,
1683 .fh_to_parent = ext4_fh_to_parent,
1684 .get_parent = ext4_get_parent,
1685 .commit_metadata = ext4_nfs_commit_metadata,
1689 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1690 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1691 Opt_nouid32, Opt_debug, Opt_removed,
1692 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1693 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1694 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1695 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1696 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1697 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1699 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1700 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1701 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1702 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version,
1703 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1704 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1705 Opt_nowarn_on_error, Opt_mblk_io_submit,
1706 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1707 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1708 Opt_inode_readahead_blks, Opt_journal_ioprio,
1709 Opt_dioread_nolock, Opt_dioread_lock,
1710 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1711 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1712 Opt_prefetch_block_bitmaps, Opt_no_fc,
1715 static const match_table_t tokens = {
1716 {Opt_bsd_df, "bsddf"},
1717 {Opt_minix_df, "minixdf"},
1718 {Opt_grpid, "grpid"},
1719 {Opt_grpid, "bsdgroups"},
1720 {Opt_nogrpid, "nogrpid"},
1721 {Opt_nogrpid, "sysvgroups"},
1722 {Opt_resgid, "resgid=%u"},
1723 {Opt_resuid, "resuid=%u"},
1725 {Opt_err_cont, "errors=continue"},
1726 {Opt_err_panic, "errors=panic"},
1727 {Opt_err_ro, "errors=remount-ro"},
1728 {Opt_nouid32, "nouid32"},
1729 {Opt_debug, "debug"},
1730 {Opt_removed, "oldalloc"},
1731 {Opt_removed, "orlov"},
1732 {Opt_user_xattr, "user_xattr"},
1733 {Opt_nouser_xattr, "nouser_xattr"},
1735 {Opt_noacl, "noacl"},
1736 {Opt_noload, "norecovery"},
1737 {Opt_noload, "noload"},
1738 {Opt_removed, "nobh"},
1739 {Opt_removed, "bh"},
1740 {Opt_commit, "commit=%u"},
1741 {Opt_min_batch_time, "min_batch_time=%u"},
1742 {Opt_max_batch_time, "max_batch_time=%u"},
1743 {Opt_journal_dev, "journal_dev=%u"},
1744 {Opt_journal_path, "journal_path=%s"},
1745 {Opt_journal_checksum, "journal_checksum"},
1746 {Opt_nojournal_checksum, "nojournal_checksum"},
1747 {Opt_journal_async_commit, "journal_async_commit"},
1748 {Opt_abort, "abort"},
1749 {Opt_data_journal, "data=journal"},
1750 {Opt_data_ordered, "data=ordered"},
1751 {Opt_data_writeback, "data=writeback"},
1752 {Opt_data_err_abort, "data_err=abort"},
1753 {Opt_data_err_ignore, "data_err=ignore"},
1754 {Opt_offusrjquota, "usrjquota="},
1755 {Opt_usrjquota, "usrjquota=%s"},
1756 {Opt_offgrpjquota, "grpjquota="},
1757 {Opt_grpjquota, "grpjquota=%s"},
1758 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1759 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1760 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1761 {Opt_grpquota, "grpquota"},
1762 {Opt_noquota, "noquota"},
1763 {Opt_quota, "quota"},
1764 {Opt_usrquota, "usrquota"},
1765 {Opt_prjquota, "prjquota"},
1766 {Opt_barrier, "barrier=%u"},
1767 {Opt_barrier, "barrier"},
1768 {Opt_nobarrier, "nobarrier"},
1769 {Opt_i_version, "i_version"},
1771 {Opt_dax_always, "dax=always"},
1772 {Opt_dax_inode, "dax=inode"},
1773 {Opt_dax_never, "dax=never"},
1774 {Opt_stripe, "stripe=%u"},
1775 {Opt_delalloc, "delalloc"},
1776 {Opt_warn_on_error, "warn_on_error"},
1777 {Opt_nowarn_on_error, "nowarn_on_error"},
1778 {Opt_lazytime, "lazytime"},
1779 {Opt_nolazytime, "nolazytime"},
1780 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1781 {Opt_nodelalloc, "nodelalloc"},
1782 {Opt_removed, "mblk_io_submit"},
1783 {Opt_removed, "nomblk_io_submit"},
1784 {Opt_block_validity, "block_validity"},
1785 {Opt_noblock_validity, "noblock_validity"},
1786 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1787 {Opt_journal_ioprio, "journal_ioprio=%u"},
1788 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1789 {Opt_auto_da_alloc, "auto_da_alloc"},
1790 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1791 {Opt_dioread_nolock, "dioread_nolock"},
1792 {Opt_dioread_lock, "nodioread_nolock"},
1793 {Opt_dioread_lock, "dioread_lock"},
1794 {Opt_discard, "discard"},
1795 {Opt_nodiscard, "nodiscard"},
1796 {Opt_init_itable, "init_itable=%u"},
1797 {Opt_init_itable, "init_itable"},
1798 {Opt_noinit_itable, "noinit_itable"},
1799 {Opt_no_fc, "no_fc"},
1800 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1801 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
1802 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1803 {Opt_inlinecrypt, "inlinecrypt"},
1804 {Opt_nombcache, "nombcache"},
1805 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1806 {Opt_prefetch_block_bitmaps, "prefetch_block_bitmaps"},
1807 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1808 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1809 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1810 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1811 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1815 static ext4_fsblk_t get_sb_block(void **data)
1817 ext4_fsblk_t sb_block;
1818 char *options = (char *) *data;
1820 if (!options || strncmp(options, "sb=", 3) != 0)
1821 return 1; /* Default location */
1824 /* TODO: use simple_strtoll with >32bit ext4 */
1825 sb_block = simple_strtoul(options, &options, 0);
1826 if (*options && *options != ',') {
1827 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1831 if (*options == ',')
1833 *data = (void *) options;
1838 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1839 static const char deprecated_msg[] =
1840 "Mount option \"%s\" will be removed by %s\n"
1841 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1844 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1846 struct ext4_sb_info *sbi = EXT4_SB(sb);
1847 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
1850 if (sb_any_quota_loaded(sb) && !old_qname) {
1851 ext4_msg(sb, KERN_ERR,
1852 "Cannot change journaled "
1853 "quota options when quota turned on");
1856 if (ext4_has_feature_quota(sb)) {
1857 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1858 "ignored when QUOTA feature is enabled");
1861 qname = match_strdup(args);
1863 ext4_msg(sb, KERN_ERR,
1864 "Not enough memory for storing quotafile name");
1868 if (strcmp(old_qname, qname) == 0)
1871 ext4_msg(sb, KERN_ERR,
1872 "%s quota file already specified",
1876 if (strchr(qname, '/')) {
1877 ext4_msg(sb, KERN_ERR,
1878 "quotafile must be on filesystem root");
1881 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
1889 static int clear_qf_name(struct super_block *sb, int qtype)
1892 struct ext4_sb_info *sbi = EXT4_SB(sb);
1893 char *old_qname = get_qf_name(sb, sbi, qtype);
1895 if (sb_any_quota_loaded(sb) && old_qname) {
1896 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1897 " when quota turned on");
1900 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1907 #define MOPT_SET 0x0001
1908 #define MOPT_CLEAR 0x0002
1909 #define MOPT_NOSUPPORT 0x0004
1910 #define MOPT_EXPLICIT 0x0008
1911 #define MOPT_CLEAR_ERR 0x0010
1912 #define MOPT_GTE0 0x0020
1915 #define MOPT_QFMT 0x0040
1917 #define MOPT_Q MOPT_NOSUPPORT
1918 #define MOPT_QFMT MOPT_NOSUPPORT
1920 #define MOPT_DATAJ 0x0080
1921 #define MOPT_NO_EXT2 0x0100
1922 #define MOPT_NO_EXT3 0x0200
1923 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1924 #define MOPT_STRING 0x0400
1925 #define MOPT_SKIP 0x0800
1926 #define MOPT_2 0x1000
1928 static const struct mount_opts {
1932 } ext4_mount_opts[] = {
1933 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1934 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1935 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1936 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1937 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1938 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1939 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1940 MOPT_EXT4_ONLY | MOPT_SET},
1941 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1942 MOPT_EXT4_ONLY | MOPT_CLEAR},
1943 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1944 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1945 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1946 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1947 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1948 MOPT_EXT4_ONLY | MOPT_CLEAR},
1949 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1950 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1951 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1952 MOPT_EXT4_ONLY | MOPT_CLEAR},
1953 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1954 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1955 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1956 EXT4_MOUNT_JOURNAL_CHECKSUM),
1957 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1958 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1959 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1960 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1961 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1962 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1964 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1966 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1967 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1968 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1969 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1970 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1971 {Opt_commit, 0, MOPT_GTE0},
1972 {Opt_max_batch_time, 0, MOPT_GTE0},
1973 {Opt_min_batch_time, 0, MOPT_GTE0},
1974 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1975 {Opt_init_itable, 0, MOPT_GTE0},
1976 {Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP},
1977 {Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS,
1978 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1979 {Opt_dax_inode, EXT4_MOUNT2_DAX_INODE,
1980 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1981 {Opt_dax_never, EXT4_MOUNT2_DAX_NEVER,
1982 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1983 {Opt_stripe, 0, MOPT_GTE0},
1984 {Opt_resuid, 0, MOPT_GTE0},
1985 {Opt_resgid, 0, MOPT_GTE0},
1986 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1987 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1988 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1989 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1990 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1991 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1992 MOPT_NO_EXT2 | MOPT_DATAJ},
1993 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1994 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1995 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1996 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1997 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1999 {Opt_acl, 0, MOPT_NOSUPPORT},
2000 {Opt_noacl, 0, MOPT_NOSUPPORT},
2002 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
2003 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
2004 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
2005 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
2006 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
2008 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
2010 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
2012 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
2013 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
2014 MOPT_CLEAR | MOPT_Q},
2015 {Opt_usrjquota, 0, MOPT_Q},
2016 {Opt_grpjquota, 0, MOPT_Q},
2017 {Opt_offusrjquota, 0, MOPT_Q},
2018 {Opt_offgrpjquota, 0, MOPT_Q},
2019 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
2020 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
2021 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
2022 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
2023 {Opt_test_dummy_encryption, 0, MOPT_STRING},
2024 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
2025 {Opt_prefetch_block_bitmaps, EXT4_MOUNT_PREFETCH_BLOCK_BITMAPS,
2027 {Opt_no_fc, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
2028 MOPT_CLEAR | MOPT_2 | MOPT_EXT4_ONLY},
2032 #ifdef CONFIG_UNICODE
2033 static const struct ext4_sb_encodings {
2037 } ext4_sb_encoding_map[] = {
2038 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
2041 static int ext4_sb_read_encoding(const struct ext4_super_block *es,
2042 const struct ext4_sb_encodings **encoding,
2045 __u16 magic = le16_to_cpu(es->s_encoding);
2048 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
2049 if (magic == ext4_sb_encoding_map[i].magic)
2052 if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
2055 *encoding = &ext4_sb_encoding_map[i];
2056 *flags = le16_to_cpu(es->s_encoding_flags);
2062 static int ext4_set_test_dummy_encryption(struct super_block *sb,
2064 const substring_t *arg,
2067 #ifdef CONFIG_FS_ENCRYPTION
2068 struct ext4_sb_info *sbi = EXT4_SB(sb);
2072 * This mount option is just for testing, and it's not worthwhile to
2073 * implement the extra complexity (e.g. RCU protection) that would be
2074 * needed to allow it to be set or changed during remount. We do allow
2075 * it to be specified during remount, but only if there is no change.
2077 if (is_remount && !sbi->s_dummy_enc_ctx.ctx) {
2078 ext4_msg(sb, KERN_WARNING,
2079 "Can't set test_dummy_encryption on remount");
2082 err = fscrypt_set_test_dummy_encryption(sb, arg, &sbi->s_dummy_enc_ctx);
2085 ext4_msg(sb, KERN_WARNING,
2086 "Can't change test_dummy_encryption on remount");
2087 else if (err == -EINVAL)
2088 ext4_msg(sb, KERN_WARNING,
2089 "Value of option \"%s\" is unrecognized", opt);
2091 ext4_msg(sb, KERN_WARNING,
2092 "Error processing option \"%s\" [%d]",
2096 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2098 ext4_msg(sb, KERN_WARNING,
2099 "Test dummy encryption mount option ignored");
2104 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
2105 substring_t *args, unsigned long *journal_devnum,
2106 unsigned int *journal_ioprio, int is_remount)
2108 struct ext4_sb_info *sbi = EXT4_SB(sb);
2109 const struct mount_opts *m;
2115 if (token == Opt_usrjquota)
2116 return set_qf_name(sb, USRQUOTA, &args[0]);
2117 else if (token == Opt_grpjquota)
2118 return set_qf_name(sb, GRPQUOTA, &args[0]);
2119 else if (token == Opt_offusrjquota)
2120 return clear_qf_name(sb, USRQUOTA);
2121 else if (token == Opt_offgrpjquota)
2122 return clear_qf_name(sb, GRPQUOTA);
2126 case Opt_nouser_xattr:
2127 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
2130 return 1; /* handled by get_sb_block() */
2132 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
2135 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
2138 sb->s_flags |= SB_I_VERSION;
2141 sb->s_flags |= SB_LAZYTIME;
2143 case Opt_nolazytime:
2144 sb->s_flags &= ~SB_LAZYTIME;
2146 case Opt_inlinecrypt:
2147 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2148 sb->s_flags |= SB_INLINECRYPT;
2150 ext4_msg(sb, KERN_ERR, "inline encryption not supported");
2155 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2156 if (token == m->token)
2159 if (m->token == Opt_err) {
2160 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
2161 "or missing value", opt);
2165 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2166 ext4_msg(sb, KERN_ERR,
2167 "Mount option \"%s\" incompatible with ext2", opt);
2170 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2171 ext4_msg(sb, KERN_ERR,
2172 "Mount option \"%s\" incompatible with ext3", opt);
2176 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
2178 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
2180 if (m->flags & MOPT_EXPLICIT) {
2181 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2182 set_opt2(sb, EXPLICIT_DELALLOC);
2183 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2184 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
2188 if (m->flags & MOPT_CLEAR_ERR)
2189 clear_opt(sb, ERRORS_MASK);
2190 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
2191 ext4_msg(sb, KERN_ERR, "Cannot change quota "
2192 "options when quota turned on");
2196 if (m->flags & MOPT_NOSUPPORT) {
2197 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
2198 } else if (token == Opt_commit) {
2200 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
2201 else if (arg > INT_MAX / HZ) {
2202 ext4_msg(sb, KERN_ERR,
2203 "Invalid commit interval %d, "
2204 "must be smaller than %d",
2208 sbi->s_commit_interval = HZ * arg;
2209 } else if (token == Opt_debug_want_extra_isize) {
2212 (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
2213 ext4_msg(sb, KERN_ERR,
2214 "Invalid want_extra_isize %d", arg);
2217 sbi->s_want_extra_isize = arg;
2218 } else if (token == Opt_max_batch_time) {
2219 sbi->s_max_batch_time = arg;
2220 } else if (token == Opt_min_batch_time) {
2221 sbi->s_min_batch_time = arg;
2222 } else if (token == Opt_inode_readahead_blks) {
2223 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
2224 ext4_msg(sb, KERN_ERR,
2225 "EXT4-fs: inode_readahead_blks must be "
2226 "0 or a power of 2 smaller than 2^31");
2229 sbi->s_inode_readahead_blks = arg;
2230 } else if (token == Opt_init_itable) {
2231 set_opt(sb, INIT_INODE_TABLE);
2233 arg = EXT4_DEF_LI_WAIT_MULT;
2234 sbi->s_li_wait_mult = arg;
2235 } else if (token == Opt_max_dir_size_kb) {
2236 sbi->s_max_dir_size_kb = arg;
2237 } else if (token == Opt_stripe) {
2238 sbi->s_stripe = arg;
2239 } else if (token == Opt_resuid) {
2240 uid = make_kuid(current_user_ns(), arg);
2241 if (!uid_valid(uid)) {
2242 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
2245 sbi->s_resuid = uid;
2246 } else if (token == Opt_resgid) {
2247 gid = make_kgid(current_user_ns(), arg);
2248 if (!gid_valid(gid)) {
2249 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
2252 sbi->s_resgid = gid;
2253 } else if (token == Opt_journal_dev) {
2255 ext4_msg(sb, KERN_ERR,
2256 "Cannot specify journal on remount");
2259 *journal_devnum = arg;
2260 } else if (token == Opt_journal_path) {
2262 struct inode *journal_inode;
2267 ext4_msg(sb, KERN_ERR,
2268 "Cannot specify journal on remount");
2271 journal_path = match_strdup(&args[0]);
2272 if (!journal_path) {
2273 ext4_msg(sb, KERN_ERR, "error: could not dup "
2274 "journal device string");
2278 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
2280 ext4_msg(sb, KERN_ERR, "error: could not find "
2281 "journal device path: error %d", error);
2282 kfree(journal_path);
2286 journal_inode = d_inode(path.dentry);
2287 if (!S_ISBLK(journal_inode->i_mode)) {
2288 ext4_msg(sb, KERN_ERR, "error: journal path %s "
2289 "is not a block device", journal_path);
2291 kfree(journal_path);
2295 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
2297 kfree(journal_path);
2298 } else if (token == Opt_journal_ioprio) {
2300 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
2305 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
2306 } else if (token == Opt_test_dummy_encryption) {
2307 return ext4_set_test_dummy_encryption(sb, opt, &args[0],
2309 } else if (m->flags & MOPT_DATAJ) {
2311 if (!sbi->s_journal)
2312 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
2313 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
2314 ext4_msg(sb, KERN_ERR,
2315 "Cannot change data mode on remount");
2319 clear_opt(sb, DATA_FLAGS);
2320 sbi->s_mount_opt |= m->mount_opt;
2323 } else if (m->flags & MOPT_QFMT) {
2324 if (sb_any_quota_loaded(sb) &&
2325 sbi->s_jquota_fmt != m->mount_opt) {
2326 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
2327 "quota options when quota turned on");
2330 if (ext4_has_feature_quota(sb)) {
2331 ext4_msg(sb, KERN_INFO,
2332 "Quota format mount options ignored "
2333 "when QUOTA feature is enabled");
2336 sbi->s_jquota_fmt = m->mount_opt;
2338 } else if (token == Opt_dax || token == Opt_dax_always ||
2339 token == Opt_dax_inode || token == Opt_dax_never) {
2340 #ifdef CONFIG_FS_DAX
2343 case Opt_dax_always:
2345 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2346 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2347 fail_dax_change_remount:
2348 ext4_msg(sb, KERN_ERR, "can't change "
2349 "dax mount option while remounting");
2353 (test_opt(sb, DATA_FLAGS) ==
2354 EXT4_MOUNT_JOURNAL_DATA)) {
2355 ext4_msg(sb, KERN_ERR, "can't mount with "
2356 "both data=journal and dax");
2359 ext4_msg(sb, KERN_WARNING,
2360 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2361 sbi->s_mount_opt |= EXT4_MOUNT_DAX_ALWAYS;
2362 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2366 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2367 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS)))
2368 goto fail_dax_change_remount;
2369 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2370 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2374 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2375 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2376 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE)))
2377 goto fail_dax_change_remount;
2378 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2379 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2380 /* Strictly for printing options */
2381 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_INODE;
2385 ext4_msg(sb, KERN_INFO, "dax option not supported");
2386 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2387 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2390 } else if (token == Opt_data_err_abort) {
2391 sbi->s_mount_opt |= m->mount_opt;
2392 } else if (token == Opt_data_err_ignore) {
2393 sbi->s_mount_opt &= ~m->mount_opt;
2397 if (m->flags & MOPT_CLEAR)
2399 else if (unlikely(!(m->flags & MOPT_SET))) {
2400 ext4_msg(sb, KERN_WARNING,
2401 "buggy handling of option %s", opt);
2405 if (m->flags & MOPT_2) {
2407 sbi->s_mount_opt2 |= m->mount_opt;
2409 sbi->s_mount_opt2 &= ~m->mount_opt;
2412 sbi->s_mount_opt |= m->mount_opt;
2414 sbi->s_mount_opt &= ~m->mount_opt;
2420 static int parse_options(char *options, struct super_block *sb,
2421 unsigned long *journal_devnum,
2422 unsigned int *journal_ioprio,
2425 struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb);
2426 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
2427 substring_t args[MAX_OPT_ARGS];
2433 while ((p = strsep(&options, ",")) != NULL) {
2437 * Initialize args struct so we know whether arg was
2438 * found; some options take optional arguments.
2440 args[0].to = args[0].from = NULL;
2441 token = match_token(p, tokens, args);
2442 if (handle_mount_opt(sb, p, token, args, journal_devnum,
2443 journal_ioprio, is_remount) < 0)
2448 * We do the test below only for project quotas. 'usrquota' and
2449 * 'grpquota' mount options are allowed even without quota feature
2450 * to support legacy quotas in quota files.
2452 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2453 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2454 "Cannot enable project quota enforcement.");
2457 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2458 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2459 if (usr_qf_name || grp_qf_name) {
2460 if (test_opt(sb, USRQUOTA) && usr_qf_name)
2461 clear_opt(sb, USRQUOTA);
2463 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
2464 clear_opt(sb, GRPQUOTA);
2466 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
2467 ext4_msg(sb, KERN_ERR, "old and new quota "
2472 if (!sbi->s_jquota_fmt) {
2473 ext4_msg(sb, KERN_ERR, "journaled quota format "
2479 if (test_opt(sb, DIOREAD_NOLOCK)) {
2481 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2482 if (blocksize < PAGE_SIZE)
2483 ext4_msg(sb, KERN_WARNING, "Warning: mounting with an "
2484 "experimental mount option 'dioread_nolock' "
2485 "for blocksize < PAGE_SIZE");
2490 static inline void ext4_show_quota_options(struct seq_file *seq,
2491 struct super_block *sb)
2493 #if defined(CONFIG_QUOTA)
2494 struct ext4_sb_info *sbi = EXT4_SB(sb);
2495 char *usr_qf_name, *grp_qf_name;
2497 if (sbi->s_jquota_fmt) {
2500 switch (sbi->s_jquota_fmt) {
2511 seq_printf(seq, ",jqfmt=%s", fmtname);
2515 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2516 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2518 seq_show_option(seq, "usrjquota", usr_qf_name);
2520 seq_show_option(seq, "grpjquota", grp_qf_name);
2525 static const char *token2str(int token)
2527 const struct match_token *t;
2529 for (t = tokens; t->token != Opt_err; t++)
2530 if (t->token == token && !strchr(t->pattern, '='))
2537 * - it's set to a non-default value OR
2538 * - if the per-sb default is different from the global default
2540 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2543 struct ext4_sb_info *sbi = EXT4_SB(sb);
2544 struct ext4_super_block *es = sbi->s_es;
2545 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2546 const struct mount_opts *m;
2547 char sep = nodefs ? '\n' : ',';
2549 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2550 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2552 if (sbi->s_sb_block != 1)
2553 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2555 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2556 int want_set = m->flags & MOPT_SET;
2557 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2558 (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP)
2560 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2561 continue; /* skip if same as the default */
2563 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2564 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2565 continue; /* select Opt_noFoo vs Opt_Foo */
2566 SEQ_OPTS_PRINT("%s", token2str(m->token));
2569 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2570 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2571 SEQ_OPTS_PRINT("resuid=%u",
2572 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2573 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2574 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2575 SEQ_OPTS_PRINT("resgid=%u",
2576 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2577 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2578 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2579 SEQ_OPTS_PUTS("errors=remount-ro");
2580 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2581 SEQ_OPTS_PUTS("errors=continue");
2582 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2583 SEQ_OPTS_PUTS("errors=panic");
2584 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2585 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2586 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2587 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2588 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2589 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2590 if (sb->s_flags & SB_I_VERSION)
2591 SEQ_OPTS_PUTS("i_version");
2592 if (nodefs || sbi->s_stripe)
2593 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2594 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2595 (sbi->s_mount_opt ^ def_mount_opt)) {
2596 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2597 SEQ_OPTS_PUTS("data=journal");
2598 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2599 SEQ_OPTS_PUTS("data=ordered");
2600 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2601 SEQ_OPTS_PUTS("data=writeback");
2604 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2605 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2606 sbi->s_inode_readahead_blks);
2608 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2609 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2610 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2611 if (nodefs || sbi->s_max_dir_size_kb)
2612 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2613 if (test_opt(sb, DATA_ERR_ABORT))
2614 SEQ_OPTS_PUTS("data_err=abort");
2616 fscrypt_show_test_dummy_encryption(seq, sep, sb);
2618 if (sb->s_flags & SB_INLINECRYPT)
2619 SEQ_OPTS_PUTS("inlinecrypt");
2621 if (test_opt(sb, DAX_ALWAYS)) {
2623 SEQ_OPTS_PUTS("dax");
2625 SEQ_OPTS_PUTS("dax=always");
2626 } else if (test_opt2(sb, DAX_NEVER)) {
2627 SEQ_OPTS_PUTS("dax=never");
2628 } else if (test_opt2(sb, DAX_INODE)) {
2629 SEQ_OPTS_PUTS("dax=inode");
2632 if (test_opt2(sb, JOURNAL_FAST_COMMIT))
2633 SEQ_OPTS_PUTS("fast_commit");
2635 ext4_show_quota_options(seq, sb);
2639 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2641 return _ext4_show_options(seq, root->d_sb, 0);
2644 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2646 struct super_block *sb = seq->private;
2649 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2650 rc = _ext4_show_options(seq, sb, 1);
2651 seq_puts(seq, "\n");
2655 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2658 struct ext4_sb_info *sbi = EXT4_SB(sb);
2661 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2662 ext4_msg(sb, KERN_ERR, "revision level too high, "
2663 "forcing read-only mode");
2669 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2670 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2671 "running e2fsck is recommended");
2672 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2673 ext4_msg(sb, KERN_WARNING,
2674 "warning: mounting fs with errors, "
2675 "running e2fsck is recommended");
2676 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2677 le16_to_cpu(es->s_mnt_count) >=
2678 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2679 ext4_msg(sb, KERN_WARNING,
2680 "warning: maximal mount count reached, "
2681 "running e2fsck is recommended");
2682 else if (le32_to_cpu(es->s_checkinterval) &&
2683 (ext4_get_tstamp(es, s_lastcheck) +
2684 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
2685 ext4_msg(sb, KERN_WARNING,
2686 "warning: checktime reached, "
2687 "running e2fsck is recommended");
2688 if (!sbi->s_journal)
2689 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2690 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2691 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2692 le16_add_cpu(&es->s_mnt_count, 1);
2693 ext4_update_tstamp(es, s_mtime);
2695 ext4_set_feature_journal_needs_recovery(sb);
2697 err = ext4_commit_super(sb, 1);
2699 if (test_opt(sb, DEBUG))
2700 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2701 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2703 sbi->s_groups_count,
2704 EXT4_BLOCKS_PER_GROUP(sb),
2705 EXT4_INODES_PER_GROUP(sb),
2706 sbi->s_mount_opt, sbi->s_mount_opt2);
2708 cleancache_init_fs(sb);
2712 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2714 struct ext4_sb_info *sbi = EXT4_SB(sb);
2715 struct flex_groups **old_groups, **new_groups;
2718 if (!sbi->s_log_groups_per_flex)
2721 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2722 if (size <= sbi->s_flex_groups_allocated)
2725 new_groups = kvzalloc(roundup_pow_of_two(size *
2726 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
2728 ext4_msg(sb, KERN_ERR,
2729 "not enough memory for %d flex group pointers", size);
2732 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2733 new_groups[i] = kvzalloc(roundup_pow_of_two(
2734 sizeof(struct flex_groups)),
2736 if (!new_groups[i]) {
2737 for (j = sbi->s_flex_groups_allocated; j < i; j++)
2738 kvfree(new_groups[j]);
2740 ext4_msg(sb, KERN_ERR,
2741 "not enough memory for %d flex groups", size);
2746 old_groups = rcu_dereference(sbi->s_flex_groups);
2748 memcpy(new_groups, old_groups,
2749 (sbi->s_flex_groups_allocated *
2750 sizeof(struct flex_groups *)));
2752 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2753 sbi->s_flex_groups_allocated = size;
2755 ext4_kvfree_array_rcu(old_groups);
2759 static int ext4_fill_flex_info(struct super_block *sb)
2761 struct ext4_sb_info *sbi = EXT4_SB(sb);
2762 struct ext4_group_desc *gdp = NULL;
2763 struct flex_groups *fg;
2764 ext4_group_t flex_group;
2767 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2768 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2769 sbi->s_log_groups_per_flex = 0;
2773 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2777 for (i = 0; i < sbi->s_groups_count; i++) {
2778 gdp = ext4_get_group_desc(sb, i, NULL);
2780 flex_group = ext4_flex_group(sbi, i);
2781 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2782 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
2783 atomic64_add(ext4_free_group_clusters(sb, gdp),
2784 &fg->free_clusters);
2785 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
2793 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2794 struct ext4_group_desc *gdp)
2796 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2798 __le32 le_group = cpu_to_le32(block_group);
2799 struct ext4_sb_info *sbi = EXT4_SB(sb);
2801 if (ext4_has_metadata_csum(sbi->s_sb)) {
2802 /* Use new metadata_csum algorithm */
2804 __u16 dummy_csum = 0;
2806 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2808 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2809 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2810 sizeof(dummy_csum));
2811 offset += sizeof(dummy_csum);
2812 if (offset < sbi->s_desc_size)
2813 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2814 sbi->s_desc_size - offset);
2816 crc = csum32 & 0xFFFF;
2820 /* old crc16 code */
2821 if (!ext4_has_feature_gdt_csum(sb))
2824 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2825 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2826 crc = crc16(crc, (__u8 *)gdp, offset);
2827 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2828 /* for checksum of struct ext4_group_desc do the rest...*/
2829 if (ext4_has_feature_64bit(sb) &&
2830 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2831 crc = crc16(crc, (__u8 *)gdp + offset,
2832 le16_to_cpu(sbi->s_es->s_desc_size) -
2836 return cpu_to_le16(crc);
2839 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2840 struct ext4_group_desc *gdp)
2842 if (ext4_has_group_desc_csum(sb) &&
2843 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2849 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2850 struct ext4_group_desc *gdp)
2852 if (!ext4_has_group_desc_csum(sb))
2854 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2857 /* Called at mount-time, super-block is locked */
2858 static int ext4_check_descriptors(struct super_block *sb,
2859 ext4_fsblk_t sb_block,
2860 ext4_group_t *first_not_zeroed)
2862 struct ext4_sb_info *sbi = EXT4_SB(sb);
2863 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2864 ext4_fsblk_t last_block;
2865 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2866 ext4_fsblk_t block_bitmap;
2867 ext4_fsblk_t inode_bitmap;
2868 ext4_fsblk_t inode_table;
2869 int flexbg_flag = 0;
2870 ext4_group_t i, grp = sbi->s_groups_count;
2872 if (ext4_has_feature_flex_bg(sb))
2875 ext4_debug("Checking group descriptors");
2877 for (i = 0; i < sbi->s_groups_count; i++) {
2878 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2880 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2881 last_block = ext4_blocks_count(sbi->s_es) - 1;
2883 last_block = first_block +
2884 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2886 if ((grp == sbi->s_groups_count) &&
2887 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2890 block_bitmap = ext4_block_bitmap(sb, gdp);
2891 if (block_bitmap == sb_block) {
2892 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2893 "Block bitmap for group %u overlaps "
2898 if (block_bitmap >= sb_block + 1 &&
2899 block_bitmap <= last_bg_block) {
2900 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2901 "Block bitmap for group %u overlaps "
2902 "block group descriptors", i);
2906 if (block_bitmap < first_block || block_bitmap > last_block) {
2907 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2908 "Block bitmap for group %u not in group "
2909 "(block %llu)!", i, block_bitmap);
2912 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2913 if (inode_bitmap == sb_block) {
2914 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2915 "Inode bitmap for group %u overlaps "
2920 if (inode_bitmap >= sb_block + 1 &&
2921 inode_bitmap <= last_bg_block) {
2922 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2923 "Inode bitmap for group %u overlaps "
2924 "block group descriptors", i);
2928 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2929 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2930 "Inode bitmap for group %u not in group "
2931 "(block %llu)!", i, inode_bitmap);
2934 inode_table = ext4_inode_table(sb, gdp);
2935 if (inode_table == sb_block) {
2936 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2937 "Inode table for group %u overlaps "
2942 if (inode_table >= sb_block + 1 &&
2943 inode_table <= last_bg_block) {
2944 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2945 "Inode table for group %u overlaps "
2946 "block group descriptors", i);
2950 if (inode_table < first_block ||
2951 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2952 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2953 "Inode table for group %u not in group "
2954 "(block %llu)!", i, inode_table);
2957 ext4_lock_group(sb, i);
2958 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2959 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2960 "Checksum for group %u failed (%u!=%u)",
2961 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2962 gdp)), le16_to_cpu(gdp->bg_checksum));
2963 if (!sb_rdonly(sb)) {
2964 ext4_unlock_group(sb, i);
2968 ext4_unlock_group(sb, i);
2970 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2972 if (NULL != first_not_zeroed)
2973 *first_not_zeroed = grp;
2977 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2978 * the superblock) which were deleted from all directories, but held open by
2979 * a process at the time of a crash. We walk the list and try to delete these
2980 * inodes at recovery time (only with a read-write filesystem).
2982 * In order to keep the orphan inode chain consistent during traversal (in
2983 * case of crash during recovery), we link each inode into the superblock
2984 * orphan list_head and handle it the same way as an inode deletion during
2985 * normal operation (which journals the operations for us).
2987 * We only do an iget() and an iput() on each inode, which is very safe if we
2988 * accidentally point at an in-use or already deleted inode. The worst that
2989 * can happen in this case is that we get a "bit already cleared" message from
2990 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2991 * e2fsck was run on this filesystem, and it must have already done the orphan
2992 * inode cleanup for us, so we can safely abort without any further action.
2994 static void ext4_orphan_cleanup(struct super_block *sb,
2995 struct ext4_super_block *es)
2997 unsigned int s_flags = sb->s_flags;
2998 int ret, nr_orphans = 0, nr_truncates = 0;
3000 int quota_update = 0;
3003 if (!es->s_last_orphan) {
3004 jbd_debug(4, "no orphan inodes to clean up\n");
3008 if (bdev_read_only(sb->s_bdev)) {
3009 ext4_msg(sb, KERN_ERR, "write access "
3010 "unavailable, skipping orphan cleanup");
3014 /* Check if feature set would not allow a r/w mount */
3015 if (!ext4_feature_set_ok(sb, 0)) {
3016 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
3017 "unknown ROCOMPAT features");
3021 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
3022 /* don't clear list on RO mount w/ errors */
3023 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
3024 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
3025 "clearing orphan list.\n");
3026 es->s_last_orphan = 0;
3028 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
3032 if (s_flags & SB_RDONLY) {
3033 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
3034 sb->s_flags &= ~SB_RDONLY;
3037 /* Needed for iput() to work correctly and not trash data */
3038 sb->s_flags |= SB_ACTIVE;
3041 * Turn on quotas which were not enabled for read-only mounts if
3042 * filesystem has quota feature, so that they are updated correctly.
3044 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
3045 int ret = ext4_enable_quotas(sb);
3050 ext4_msg(sb, KERN_ERR,
3051 "Cannot turn on quotas: error %d", ret);
3054 /* Turn on journaled quotas used for old sytle */
3055 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
3056 if (EXT4_SB(sb)->s_qf_names[i]) {
3057 int ret = ext4_quota_on_mount(sb, i);
3062 ext4_msg(sb, KERN_ERR,
3063 "Cannot turn on journaled "
3064 "quota: type %d: error %d", i, ret);
3069 while (es->s_last_orphan) {
3070 struct inode *inode;
3073 * We may have encountered an error during cleanup; if
3074 * so, skip the rest.
3076 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
3077 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
3078 es->s_last_orphan = 0;
3082 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
3083 if (IS_ERR(inode)) {
3084 es->s_last_orphan = 0;
3088 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
3089 dquot_initialize(inode);
3090 if (inode->i_nlink) {
3091 if (test_opt(sb, DEBUG))
3092 ext4_msg(sb, KERN_DEBUG,
3093 "%s: truncating inode %lu to %lld bytes",
3094 __func__, inode->i_ino, inode->i_size);
3095 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
3096 inode->i_ino, inode->i_size);
3098 truncate_inode_pages(inode->i_mapping, inode->i_size);
3099 ret = ext4_truncate(inode);
3101 ext4_std_error(inode->i_sb, ret);
3102 inode_unlock(inode);
3105 if (test_opt(sb, DEBUG))
3106 ext4_msg(sb, KERN_DEBUG,
3107 "%s: deleting unreferenced inode %lu",
3108 __func__, inode->i_ino);
3109 jbd_debug(2, "deleting unreferenced inode %lu\n",
3113 iput(inode); /* The delete magic happens here! */
3116 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
3119 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
3120 PLURAL(nr_orphans));
3122 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
3123 PLURAL(nr_truncates));
3125 /* Turn off quotas if they were enabled for orphan cleanup */
3127 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
3128 if (sb_dqopt(sb)->files[i])
3129 dquot_quota_off(sb, i);
3133 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
3137 * Maximal extent format file size.
3138 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3139 * extent format containers, within a sector_t, and within i_blocks
3140 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3141 * so that won't be a limiting factor.
3143 * However there is other limiting factor. We do store extents in the form
3144 * of starting block and length, hence the resulting length of the extent
3145 * covering maximum file size must fit into on-disk format containers as
3146 * well. Given that length is always by 1 unit bigger than max unit (because
3147 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3149 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3151 static loff_t ext4_max_size(int blkbits, int has_huge_files)
3154 loff_t upper_limit = MAX_LFS_FILESIZE;
3156 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3158 if (!has_huge_files) {
3159 upper_limit = (1LL << 32) - 1;
3161 /* total blocks in file system block size */
3162 upper_limit >>= (blkbits - 9);
3163 upper_limit <<= blkbits;
3167 * 32-bit extent-start container, ee_block. We lower the maxbytes
3168 * by one fs block, so ee_len can cover the extent of maximum file
3171 res = (1LL << 32) - 1;
3174 /* Sanity check against vm- & vfs- imposed limits */
3175 if (res > upper_limit)
3182 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
3183 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3184 * We need to be 1 filesystem block less than the 2^48 sector limit.
3186 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3188 loff_t res = EXT4_NDIR_BLOCKS;
3191 /* This is calculated to be the largest file size for a dense, block
3192 * mapped file such that the file's total number of 512-byte sectors,
3193 * including data and all indirect blocks, does not exceed (2^48 - 1).
3195 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3196 * number of 512-byte sectors of the file.
3199 if (!has_huge_files) {
3201 * !has_huge_files or implies that the inode i_block field
3202 * represents total file blocks in 2^32 512-byte sectors ==
3203 * size of vfs inode i_blocks * 8
3205 upper_limit = (1LL << 32) - 1;
3207 /* total blocks in file system block size */
3208 upper_limit >>= (bits - 9);
3212 * We use 48 bit ext4_inode i_blocks
3213 * With EXT4_HUGE_FILE_FL set the i_blocks
3214 * represent total number of blocks in
3215 * file system block size
3217 upper_limit = (1LL << 48) - 1;
3221 /* indirect blocks */
3223 /* double indirect blocks */
3224 meta_blocks += 1 + (1LL << (bits-2));
3225 /* tripple indirect blocks */
3226 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
3228 upper_limit -= meta_blocks;
3229 upper_limit <<= bits;
3231 res += 1LL << (bits-2);
3232 res += 1LL << (2*(bits-2));
3233 res += 1LL << (3*(bits-2));
3235 if (res > upper_limit)
3238 if (res > MAX_LFS_FILESIZE)
3239 res = MAX_LFS_FILESIZE;
3244 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3245 ext4_fsblk_t logical_sb_block, int nr)
3247 struct ext4_sb_info *sbi = EXT4_SB(sb);
3248 ext4_group_t bg, first_meta_bg;
3251 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3253 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3254 return logical_sb_block + nr + 1;
3255 bg = sbi->s_desc_per_block * nr;
3256 if (ext4_bg_has_super(sb, bg))
3260 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3261 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3262 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3265 if (sb->s_blocksize == 1024 && nr == 0 &&
3266 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3269 return (has_super + ext4_group_first_block_no(sb, bg));
3273 * ext4_get_stripe_size: Get the stripe size.
3274 * @sbi: In memory super block info
3276 * If we have specified it via mount option, then
3277 * use the mount option value. If the value specified at mount time is
3278 * greater than the blocks per group use the super block value.
3279 * If the super block value is greater than blocks per group return 0.
3280 * Allocator needs it be less than blocks per group.
3283 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3285 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3286 unsigned long stripe_width =
3287 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3290 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3291 ret = sbi->s_stripe;
3292 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3294 else if (stride && stride <= sbi->s_blocks_per_group)
3300 * If the stripe width is 1, this makes no sense and
3301 * we set it to 0 to turn off stripe handling code.
3310 * Check whether this filesystem can be mounted based on
3311 * the features present and the RDONLY/RDWR mount requested.
3312 * Returns 1 if this filesystem can be mounted as requested,
3313 * 0 if it cannot be.
3315 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
3317 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3318 ext4_msg(sb, KERN_ERR,
3319 "Couldn't mount because of "
3320 "unsupported optional features (%x)",
3321 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3322 ~EXT4_FEATURE_INCOMPAT_SUPP));
3326 #ifndef CONFIG_UNICODE
3327 if (ext4_has_feature_casefold(sb)) {
3328 ext4_msg(sb, KERN_ERR,
3329 "Filesystem with casefold feature cannot be "
3330 "mounted without CONFIG_UNICODE");
3338 if (ext4_has_feature_readonly(sb)) {
3339 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3340 sb->s_flags |= SB_RDONLY;
3344 /* Check that feature set is OK for a read-write mount */
3345 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3346 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3347 "unsupported optional features (%x)",
3348 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3349 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3352 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3353 ext4_msg(sb, KERN_ERR,
3354 "Can't support bigalloc feature without "
3355 "extents feature\n");
3359 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3360 if (!readonly && (ext4_has_feature_quota(sb) ||
3361 ext4_has_feature_project(sb))) {
3362 ext4_msg(sb, KERN_ERR,
3363 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3366 #endif /* CONFIG_QUOTA */
3371 * This function is called once a day if we have errors logged
3372 * on the file system
3374 static void print_daily_error_info(struct timer_list *t)
3376 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3377 struct super_block *sb = sbi->s_sb;
3378 struct ext4_super_block *es = sbi->s_es;
3380 if (es->s_error_count)
3381 /* fsck newer than v1.41.13 is needed to clean this condition. */
3382 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3383 le32_to_cpu(es->s_error_count));
3384 if (es->s_first_error_time) {
3385 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3387 ext4_get_tstamp(es, s_first_error_time),
3388 (int) sizeof(es->s_first_error_func),
3389 es->s_first_error_func,
3390 le32_to_cpu(es->s_first_error_line));
3391 if (es->s_first_error_ino)
3392 printk(KERN_CONT ": inode %u",
3393 le32_to_cpu(es->s_first_error_ino));
3394 if (es->s_first_error_block)
3395 printk(KERN_CONT ": block %llu", (unsigned long long)
3396 le64_to_cpu(es->s_first_error_block));
3397 printk(KERN_CONT "\n");
3399 if (es->s_last_error_time) {
3400 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3402 ext4_get_tstamp(es, s_last_error_time),
3403 (int) sizeof(es->s_last_error_func),
3404 es->s_last_error_func,
3405 le32_to_cpu(es->s_last_error_line));
3406 if (es->s_last_error_ino)
3407 printk(KERN_CONT ": inode %u",
3408 le32_to_cpu(es->s_last_error_ino));
3409 if (es->s_last_error_block)
3410 printk(KERN_CONT ": block %llu", (unsigned long long)
3411 le64_to_cpu(es->s_last_error_block));
3412 printk(KERN_CONT "\n");
3414 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3417 /* Find next suitable group and run ext4_init_inode_table */
3418 static int ext4_run_li_request(struct ext4_li_request *elr)
3420 struct ext4_group_desc *gdp = NULL;
3421 struct super_block *sb = elr->lr_super;
3422 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3423 ext4_group_t group = elr->lr_next_group;
3424 unsigned long timeout = 0;
3425 unsigned int prefetch_ios = 0;
3428 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3429 elr->lr_next_group = ext4_mb_prefetch(sb, group,
3430 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3432 ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3434 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3436 if (group >= elr->lr_next_group) {
3438 if (elr->lr_first_not_zeroed != ngroups &&
3439 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3440 elr->lr_next_group = elr->lr_first_not_zeroed;
3441 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3448 for (; group < ngroups; group++) {
3449 gdp = ext4_get_group_desc(sb, group, NULL);
3455 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3459 if (group >= ngroups)
3464 ret = ext4_init_inode_table(sb, group,
3465 elr->lr_timeout ? 0 : 1);
3466 trace_ext4_lazy_itable_init(sb, group);
3467 if (elr->lr_timeout == 0) {
3468 timeout = (jiffies - timeout) *
3469 EXT4_SB(elr->lr_super)->s_li_wait_mult;
3470 elr->lr_timeout = timeout;
3472 elr->lr_next_sched = jiffies + elr->lr_timeout;
3473 elr->lr_next_group = group + 1;
3479 * Remove lr_request from the list_request and free the
3480 * request structure. Should be called with li_list_mtx held
3482 static void ext4_remove_li_request(struct ext4_li_request *elr)
3487 list_del(&elr->lr_request);
3488 EXT4_SB(elr->lr_super)->s_li_request = NULL;
3492 static void ext4_unregister_li_request(struct super_block *sb)
3494 mutex_lock(&ext4_li_mtx);
3495 if (!ext4_li_info) {
3496 mutex_unlock(&ext4_li_mtx);
3500 mutex_lock(&ext4_li_info->li_list_mtx);
3501 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3502 mutex_unlock(&ext4_li_info->li_list_mtx);
3503 mutex_unlock(&ext4_li_mtx);
3506 static struct task_struct *ext4_lazyinit_task;
3509 * This is the function where ext4lazyinit thread lives. It walks
3510 * through the request list searching for next scheduled filesystem.
3511 * When such a fs is found, run the lazy initialization request
3512 * (ext4_rn_li_request) and keep track of the time spend in this
3513 * function. Based on that time we compute next schedule time of
3514 * the request. When walking through the list is complete, compute
3515 * next waking time and put itself into sleep.
3517 static int ext4_lazyinit_thread(void *arg)
3519 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3520 struct list_head *pos, *n;
3521 struct ext4_li_request *elr;
3522 unsigned long next_wakeup, cur;
3524 BUG_ON(NULL == eli);
3528 next_wakeup = MAX_JIFFY_OFFSET;
3530 mutex_lock(&eli->li_list_mtx);
3531 if (list_empty(&eli->li_request_list)) {
3532 mutex_unlock(&eli->li_list_mtx);
3535 list_for_each_safe(pos, n, &eli->li_request_list) {
3538 elr = list_entry(pos, struct ext4_li_request,
3541 if (time_before(jiffies, elr->lr_next_sched)) {
3542 if (time_before(elr->lr_next_sched, next_wakeup))
3543 next_wakeup = elr->lr_next_sched;
3546 if (down_read_trylock(&elr->lr_super->s_umount)) {
3547 if (sb_start_write_trylock(elr->lr_super)) {
3550 * We hold sb->s_umount, sb can not
3551 * be removed from the list, it is
3552 * now safe to drop li_list_mtx
3554 mutex_unlock(&eli->li_list_mtx);
3555 err = ext4_run_li_request(elr);
3556 sb_end_write(elr->lr_super);
3557 mutex_lock(&eli->li_list_mtx);
3560 up_read((&elr->lr_super->s_umount));
3562 /* error, remove the lazy_init job */
3564 ext4_remove_li_request(elr);
3568 elr->lr_next_sched = jiffies +
3570 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3572 if (time_before(elr->lr_next_sched, next_wakeup))
3573 next_wakeup = elr->lr_next_sched;
3575 mutex_unlock(&eli->li_list_mtx);
3580 if ((time_after_eq(cur, next_wakeup)) ||
3581 (MAX_JIFFY_OFFSET == next_wakeup)) {
3586 schedule_timeout_interruptible(next_wakeup - cur);
3588 if (kthread_should_stop()) {
3589 ext4_clear_request_list();
3596 * It looks like the request list is empty, but we need
3597 * to check it under the li_list_mtx lock, to prevent any
3598 * additions into it, and of course we should lock ext4_li_mtx
3599 * to atomically free the list and ext4_li_info, because at
3600 * this point another ext4 filesystem could be registering
3603 mutex_lock(&ext4_li_mtx);
3604 mutex_lock(&eli->li_list_mtx);
3605 if (!list_empty(&eli->li_request_list)) {
3606 mutex_unlock(&eli->li_list_mtx);
3607 mutex_unlock(&ext4_li_mtx);
3610 mutex_unlock(&eli->li_list_mtx);
3611 kfree(ext4_li_info);
3612 ext4_li_info = NULL;
3613 mutex_unlock(&ext4_li_mtx);
3618 static void ext4_clear_request_list(void)
3620 struct list_head *pos, *n;
3621 struct ext4_li_request *elr;
3623 mutex_lock(&ext4_li_info->li_list_mtx);
3624 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3625 elr = list_entry(pos, struct ext4_li_request,
3627 ext4_remove_li_request(elr);
3629 mutex_unlock(&ext4_li_info->li_list_mtx);
3632 static int ext4_run_lazyinit_thread(void)
3634 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3635 ext4_li_info, "ext4lazyinit");
3636 if (IS_ERR(ext4_lazyinit_task)) {
3637 int err = PTR_ERR(ext4_lazyinit_task);
3638 ext4_clear_request_list();
3639 kfree(ext4_li_info);
3640 ext4_li_info = NULL;
3641 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3642 "initialization thread\n",
3646 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3651 * Check whether it make sense to run itable init. thread or not.
3652 * If there is at least one uninitialized inode table, return
3653 * corresponding group number, else the loop goes through all
3654 * groups and return total number of groups.
3656 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3658 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3659 struct ext4_group_desc *gdp = NULL;
3661 if (!ext4_has_group_desc_csum(sb))
3664 for (group = 0; group < ngroups; group++) {
3665 gdp = ext4_get_group_desc(sb, group, NULL);
3669 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3676 static int ext4_li_info_new(void)
3678 struct ext4_lazy_init *eli = NULL;
3680 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3684 INIT_LIST_HEAD(&eli->li_request_list);
3685 mutex_init(&eli->li_list_mtx);
3687 eli->li_state |= EXT4_LAZYINIT_QUIT;
3694 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3697 struct ext4_li_request *elr;
3699 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3704 elr->lr_first_not_zeroed = start;
3705 if (test_opt(sb, PREFETCH_BLOCK_BITMAPS))
3706 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3708 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3709 elr->lr_next_group = start;
3713 * Randomize first schedule time of the request to
3714 * spread the inode table initialization requests
3717 elr->lr_next_sched = jiffies + (prandom_u32() %
3718 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3722 int ext4_register_li_request(struct super_block *sb,
3723 ext4_group_t first_not_zeroed)
3725 struct ext4_sb_info *sbi = EXT4_SB(sb);
3726 struct ext4_li_request *elr = NULL;
3727 ext4_group_t ngroups = sbi->s_groups_count;
3730 mutex_lock(&ext4_li_mtx);
3731 if (sbi->s_li_request != NULL) {
3733 * Reset timeout so it can be computed again, because
3734 * s_li_wait_mult might have changed.
3736 sbi->s_li_request->lr_timeout = 0;
3740 if (!test_opt(sb, PREFETCH_BLOCK_BITMAPS) &&
3741 (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3742 !test_opt(sb, INIT_INODE_TABLE)))
3745 elr = ext4_li_request_new(sb, first_not_zeroed);
3751 if (NULL == ext4_li_info) {
3752 ret = ext4_li_info_new();
3757 mutex_lock(&ext4_li_info->li_list_mtx);
3758 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3759 mutex_unlock(&ext4_li_info->li_list_mtx);
3761 sbi->s_li_request = elr;
3763 * set elr to NULL here since it has been inserted to
3764 * the request_list and the removal and free of it is
3765 * handled by ext4_clear_request_list from now on.
3769 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3770 ret = ext4_run_lazyinit_thread();
3775 mutex_unlock(&ext4_li_mtx);
3782 * We do not need to lock anything since this is called on
3785 static void ext4_destroy_lazyinit_thread(void)
3788 * If thread exited earlier
3789 * there's nothing to be done.
3791 if (!ext4_li_info || !ext4_lazyinit_task)
3794 kthread_stop(ext4_lazyinit_task);
3797 static int set_journal_csum_feature_set(struct super_block *sb)
3800 int compat, incompat;
3801 struct ext4_sb_info *sbi = EXT4_SB(sb);
3803 if (ext4_has_metadata_csum(sb)) {
3804 /* journal checksum v3 */
3806 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3808 /* journal checksum v1 */
3809 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3813 jbd2_journal_clear_features(sbi->s_journal,
3814 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3815 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3816 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3817 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3818 ret = jbd2_journal_set_features(sbi->s_journal,
3820 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3822 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3823 ret = jbd2_journal_set_features(sbi->s_journal,
3826 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3827 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3829 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3830 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3837 * Note: calculating the overhead so we can be compatible with
3838 * historical BSD practice is quite difficult in the face of
3839 * clusters/bigalloc. This is because multiple metadata blocks from
3840 * different block group can end up in the same allocation cluster.
3841 * Calculating the exact overhead in the face of clustered allocation
3842 * requires either O(all block bitmaps) in memory or O(number of block
3843 * groups**2) in time. We will still calculate the superblock for
3844 * older file systems --- and if we come across with a bigalloc file
3845 * system with zero in s_overhead_clusters the estimate will be close to
3846 * correct especially for very large cluster sizes --- but for newer
3847 * file systems, it's better to calculate this figure once at mkfs
3848 * time, and store it in the superblock. If the superblock value is
3849 * present (even for non-bigalloc file systems), we will use it.
3851 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3854 struct ext4_sb_info *sbi = EXT4_SB(sb);
3855 struct ext4_group_desc *gdp;
3856 ext4_fsblk_t first_block, last_block, b;
3857 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3858 int s, j, count = 0;
3860 if (!ext4_has_feature_bigalloc(sb))
3861 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3862 sbi->s_itb_per_group + 2);
3864 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3865 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3866 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3867 for (i = 0; i < ngroups; i++) {
3868 gdp = ext4_get_group_desc(sb, i, NULL);
3869 b = ext4_block_bitmap(sb, gdp);
3870 if (b >= first_block && b <= last_block) {
3871 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3874 b = ext4_inode_bitmap(sb, gdp);
3875 if (b >= first_block && b <= last_block) {
3876 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3879 b = ext4_inode_table(sb, gdp);
3880 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3881 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3882 int c = EXT4_B2C(sbi, b - first_block);
3883 ext4_set_bit(c, buf);
3889 if (ext4_bg_has_super(sb, grp)) {
3890 ext4_set_bit(s++, buf);
3893 j = ext4_bg_num_gdb(sb, grp);
3894 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3895 ext4_error(sb, "Invalid number of block group "
3896 "descriptor blocks: %d", j);
3897 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3901 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3905 return EXT4_CLUSTERS_PER_GROUP(sb) -
3906 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3910 * Compute the overhead and stash it in sbi->s_overhead
3912 int ext4_calculate_overhead(struct super_block *sb)
3914 struct ext4_sb_info *sbi = EXT4_SB(sb);
3915 struct ext4_super_block *es = sbi->s_es;
3916 struct inode *j_inode;
3917 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3918 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3919 ext4_fsblk_t overhead = 0;
3920 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3926 * Compute the overhead (FS structures). This is constant
3927 * for a given filesystem unless the number of block groups
3928 * changes so we cache the previous value until it does.
3932 * All of the blocks before first_data_block are overhead
3934 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3937 * Add the overhead found in each block group
3939 for (i = 0; i < ngroups; i++) {
3942 blks = count_overhead(sb, i, buf);
3945 memset(buf, 0, PAGE_SIZE);
3950 * Add the internal journal blocks whether the journal has been
3953 if (sbi->s_journal && !sbi->s_journal_bdev)
3954 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3955 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
3956 /* j_inum for internal journal is non-zero */
3957 j_inode = ext4_get_journal_inode(sb, j_inum);
3959 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3960 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3963 ext4_msg(sb, KERN_ERR, "can't get journal size");
3966 sbi->s_overhead = overhead;
3968 free_page((unsigned long) buf);
3972 static void ext4_set_resv_clusters(struct super_block *sb)
3974 ext4_fsblk_t resv_clusters;
3975 struct ext4_sb_info *sbi = EXT4_SB(sb);
3978 * There's no need to reserve anything when we aren't using extents.
3979 * The space estimates are exact, there are no unwritten extents,
3980 * hole punching doesn't need new metadata... This is needed especially
3981 * to keep ext2/3 backward compatibility.
3983 if (!ext4_has_feature_extents(sb))
3986 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3987 * This should cover the situations where we can not afford to run
3988 * out of space like for example punch hole, or converting
3989 * unwritten extents in delalloc path. In most cases such
3990 * allocation would require 1, or 2 blocks, higher numbers are
3993 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3994 sbi->s_cluster_bits);
3996 do_div(resv_clusters, 50);
3997 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3999 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
4002 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
4004 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
4005 char *orig_data = kstrdup(data, GFP_KERNEL);
4006 struct buffer_head *bh, **group_desc;
4007 struct ext4_super_block *es = NULL;
4008 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
4009 struct flex_groups **flex_groups;
4011 ext4_fsblk_t sb_block = get_sb_block(&data);
4012 ext4_fsblk_t logical_sb_block;
4013 unsigned long offset = 0;
4014 unsigned long journal_devnum = 0;
4015 unsigned long def_mount_opts;
4019 int blocksize, clustersize;
4020 unsigned int db_count;
4022 int needs_recovery, has_huge_files;
4025 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4026 ext4_group_t first_not_zeroed;
4028 if ((data && !orig_data) || !sbi)
4031 sbi->s_daxdev = dax_dev;
4032 sbi->s_blockgroup_lock =
4033 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
4034 if (!sbi->s_blockgroup_lock)
4037 sb->s_fs_info = sbi;
4039 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
4040 sbi->s_sb_block = sb_block;
4041 if (sb->s_bdev->bd_part)
4042 sbi->s_sectors_written_start =
4043 part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]);
4045 /* Cleanup superblock name */
4046 strreplace(sb->s_id, '/', '!');
4048 /* -EINVAL is default */
4050 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
4052 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
4057 * The ext4 superblock will not be buffer aligned for other than 1kB
4058 * block sizes. We need to calculate the offset from buffer start.
4060 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
4061 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4062 offset = do_div(logical_sb_block, blocksize);
4064 logical_sb_block = sb_block;
4067 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4069 ext4_msg(sb, KERN_ERR, "unable to read superblock");
4075 * Note: s_es must be initialized as soon as possible because
4076 * some ext4 macro-instructions depend on its value
4078 es = (struct ext4_super_block *) (bh->b_data + offset);
4080 sb->s_magic = le16_to_cpu(es->s_magic);
4081 if (sb->s_magic != EXT4_SUPER_MAGIC)
4083 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
4085 /* Warn if metadata_csum and gdt_csum are both set. */
4086 if (ext4_has_feature_metadata_csum(sb) &&
4087 ext4_has_feature_gdt_csum(sb))
4088 ext4_warning(sb, "metadata_csum and uninit_bg are "
4089 "redundant flags; please run fsck.");
4091 /* Check for a known checksum algorithm */
4092 if (!ext4_verify_csum_type(sb, es)) {
4093 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4094 "unknown checksum algorithm.");
4099 /* Load the checksum driver */
4100 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
4101 if (IS_ERR(sbi->s_chksum_driver)) {
4102 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
4103 ret = PTR_ERR(sbi->s_chksum_driver);
4104 sbi->s_chksum_driver = NULL;
4108 /* Check superblock checksum */
4109 if (!ext4_superblock_csum_verify(sb, es)) {
4110 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4111 "invalid superblock checksum. Run e2fsck?");
4117 /* Precompute checksum seed for all metadata */
4118 if (ext4_has_feature_csum_seed(sb))
4119 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
4120 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
4121 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
4122 sizeof(es->s_uuid));
4124 /* Set defaults before we parse the mount options */
4125 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4126 set_opt(sb, INIT_INODE_TABLE);
4127 if (def_mount_opts & EXT4_DEFM_DEBUG)
4129 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4131 if (def_mount_opts & EXT4_DEFM_UID16)
4132 set_opt(sb, NO_UID32);
4133 /* xattr user namespace & acls are now defaulted on */
4134 set_opt(sb, XATTR_USER);
4135 #ifdef CONFIG_EXT4_FS_POSIX_ACL
4136 set_opt(sb, POSIX_ACL);
4138 if (ext4_has_feature_fast_commit(sb))
4139 set_opt2(sb, JOURNAL_FAST_COMMIT);
4140 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4141 if (ext4_has_metadata_csum(sb))
4142 set_opt(sb, JOURNAL_CHECKSUM);
4144 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4145 set_opt(sb, JOURNAL_DATA);
4146 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4147 set_opt(sb, ORDERED_DATA);
4148 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4149 set_opt(sb, WRITEBACK_DATA);
4151 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
4152 set_opt(sb, ERRORS_PANIC);
4153 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
4154 set_opt(sb, ERRORS_CONT);
4156 set_opt(sb, ERRORS_RO);
4157 /* block_validity enabled by default; disable with noblock_validity */
4158 set_opt(sb, BLOCK_VALIDITY);
4159 if (def_mount_opts & EXT4_DEFM_DISCARD)
4160 set_opt(sb, DISCARD);
4162 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
4163 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
4164 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
4165 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
4166 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
4168 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4169 set_opt(sb, BARRIER);
4172 * enable delayed allocation by default
4173 * Use -o nodelalloc to turn it off
4175 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4176 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4177 set_opt(sb, DELALLOC);
4180 * set default s_li_wait_mult for lazyinit, for the case there is
4181 * no mount option specified.
4183 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
4185 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
4187 if (blocksize == PAGE_SIZE)
4188 set_opt(sb, DIOREAD_NOLOCK);
4190 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
4191 blocksize > EXT4_MAX_BLOCK_SIZE) {
4192 ext4_msg(sb, KERN_ERR,
4193 "Unsupported filesystem blocksize %d (%d log_block_size)",
4194 blocksize, le32_to_cpu(es->s_log_block_size));
4198 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4199 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4200 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4202 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4203 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4204 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4205 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4209 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4210 (!is_power_of_2(sbi->s_inode_size)) ||
4211 (sbi->s_inode_size > blocksize)) {
4212 ext4_msg(sb, KERN_ERR,
4213 "unsupported inode size: %d",
4215 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
4219 * i_atime_extra is the last extra field available for
4220 * [acm]times in struct ext4_inode. Checking for that
4221 * field should suffice to ensure we have extra space
4224 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4225 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4226 sb->s_time_gran = 1;
4227 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4229 sb->s_time_gran = NSEC_PER_SEC;
4230 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4232 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4234 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4235 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4236 EXT4_GOOD_OLD_INODE_SIZE;
4237 if (ext4_has_feature_extra_isize(sb)) {
4238 unsigned v, max = (sbi->s_inode_size -
4239 EXT4_GOOD_OLD_INODE_SIZE);
4241 v = le16_to_cpu(es->s_want_extra_isize);
4243 ext4_msg(sb, KERN_ERR,
4244 "bad s_want_extra_isize: %d", v);
4247 if (sbi->s_want_extra_isize < v)
4248 sbi->s_want_extra_isize = v;
4250 v = le16_to_cpu(es->s_min_extra_isize);
4252 ext4_msg(sb, KERN_ERR,
4253 "bad s_min_extra_isize: %d", v);
4256 if (sbi->s_want_extra_isize < v)
4257 sbi->s_want_extra_isize = v;
4261 if (sbi->s_es->s_mount_opts[0]) {
4262 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
4263 sizeof(sbi->s_es->s_mount_opts),
4267 if (!parse_options(s_mount_opts, sb, &journal_devnum,
4268 &journal_ioprio, 0)) {
4269 ext4_msg(sb, KERN_WARNING,
4270 "failed to parse options in superblock: %s",
4273 kfree(s_mount_opts);
4275 sbi->s_def_mount_opt = sbi->s_mount_opt;
4276 if (!parse_options((char *) data, sb, &journal_devnum,
4277 &journal_ioprio, 0))
4280 #ifdef CONFIG_UNICODE
4281 if (ext4_has_feature_casefold(sb) && !sbi->s_encoding) {
4282 const struct ext4_sb_encodings *encoding_info;
4283 struct unicode_map *encoding;
4284 __u16 encoding_flags;
4286 if (ext4_has_feature_encrypt(sb)) {
4287 ext4_msg(sb, KERN_ERR,
4288 "Can't mount with encoding and encryption");
4292 if (ext4_sb_read_encoding(es, &encoding_info,
4294 ext4_msg(sb, KERN_ERR,
4295 "Encoding requested by superblock is unknown");
4299 encoding = utf8_load(encoding_info->version);
4300 if (IS_ERR(encoding)) {
4301 ext4_msg(sb, KERN_ERR,
4302 "can't mount with superblock charset: %s-%s "
4303 "not supported by the kernel. flags: 0x%x.",
4304 encoding_info->name, encoding_info->version,
4308 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4309 "%s-%s with flags 0x%hx", encoding_info->name,
4310 encoding_info->version?:"\b", encoding_flags);
4312 sbi->s_encoding = encoding;
4313 sbi->s_encoding_flags = encoding_flags;
4317 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4318 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, and O_DIRECT support!\n");
4319 /* can't mount with both data=journal and dioread_nolock. */
4320 clear_opt(sb, DIOREAD_NOLOCK);
4321 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4322 ext4_msg(sb, KERN_ERR, "can't mount with "
4323 "both data=journal and delalloc");
4326 if (test_opt(sb, DAX_ALWAYS)) {
4327 ext4_msg(sb, KERN_ERR, "can't mount with "
4328 "both data=journal and dax");
4331 if (ext4_has_feature_encrypt(sb)) {
4332 ext4_msg(sb, KERN_WARNING,
4333 "encrypted files will use data=ordered "
4334 "instead of data journaling mode");
4336 if (test_opt(sb, DELALLOC))
4337 clear_opt(sb, DELALLOC);
4339 sb->s_iflags |= SB_I_CGROUPWB;
4342 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4343 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
4345 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4346 (ext4_has_compat_features(sb) ||
4347 ext4_has_ro_compat_features(sb) ||
4348 ext4_has_incompat_features(sb)))
4349 ext4_msg(sb, KERN_WARNING,
4350 "feature flags set on rev 0 fs, "
4351 "running e2fsck is recommended");
4353 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4354 set_opt2(sb, HURD_COMPAT);
4355 if (ext4_has_feature_64bit(sb)) {
4356 ext4_msg(sb, KERN_ERR,
4357 "The Hurd can't support 64-bit file systems");
4362 * ea_inode feature uses l_i_version field which is not
4363 * available in HURD_COMPAT mode.
4365 if (ext4_has_feature_ea_inode(sb)) {
4366 ext4_msg(sb, KERN_ERR,
4367 "ea_inode feature is not supported for Hurd");
4372 if (IS_EXT2_SB(sb)) {
4373 if (ext2_feature_set_ok(sb))
4374 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4375 "using the ext4 subsystem");
4378 * If we're probing be silent, if this looks like
4379 * it's actually an ext[34] filesystem.
4381 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4383 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4384 "to feature incompatibilities");
4389 if (IS_EXT3_SB(sb)) {
4390 if (ext3_feature_set_ok(sb))
4391 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4392 "using the ext4 subsystem");
4395 * If we're probing be silent, if this looks like
4396 * it's actually an ext4 filesystem.
4398 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4400 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4401 "to feature incompatibilities");
4407 * Check feature flags regardless of the revision level, since we
4408 * previously didn't change the revision level when setting the flags,
4409 * so there is a chance incompat flags are set on a rev 0 filesystem.
4411 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4414 if (le32_to_cpu(es->s_log_block_size) >
4415 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4416 ext4_msg(sb, KERN_ERR,
4417 "Invalid log block size: %u",
4418 le32_to_cpu(es->s_log_block_size));
4421 if (le32_to_cpu(es->s_log_cluster_size) >
4422 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4423 ext4_msg(sb, KERN_ERR,
4424 "Invalid log cluster size: %u",
4425 le32_to_cpu(es->s_log_cluster_size));
4429 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4430 ext4_msg(sb, KERN_ERR,
4431 "Number of reserved GDT blocks insanely large: %d",
4432 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4436 if (bdev_dax_supported(sb->s_bdev, blocksize))
4437 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4439 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4440 if (ext4_has_feature_inline_data(sb)) {
4441 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4442 " that may contain inline data");
4445 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4446 ext4_msg(sb, KERN_ERR,
4447 "DAX unsupported by block device.");
4452 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4453 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4454 es->s_encryption_level);
4458 if (sb->s_blocksize != blocksize) {
4459 /* Validate the filesystem blocksize */
4460 if (!sb_set_blocksize(sb, blocksize)) {
4461 ext4_msg(sb, KERN_ERR, "bad block size %d",
4467 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4468 offset = do_div(logical_sb_block, blocksize);
4469 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4471 ext4_msg(sb, KERN_ERR,
4472 "Can't read superblock on 2nd try");
4477 es = (struct ext4_super_block *)(bh->b_data + offset);
4479 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
4480 ext4_msg(sb, KERN_ERR,
4481 "Magic mismatch, very weird!");
4486 has_huge_files = ext4_has_feature_huge_file(sb);
4487 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4489 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
4491 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
4492 if (ext4_has_feature_64bit(sb)) {
4493 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
4494 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
4495 !is_power_of_2(sbi->s_desc_size)) {
4496 ext4_msg(sb, KERN_ERR,
4497 "unsupported descriptor size %lu",
4502 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
4504 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
4505 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
4507 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
4508 if (sbi->s_inodes_per_block == 0)
4510 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4511 sbi->s_inodes_per_group > blocksize * 8) {
4512 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
4513 sbi->s_inodes_per_group);
4516 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4517 sbi->s_inodes_per_block;
4518 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
4520 sbi->s_mount_state = le16_to_cpu(es->s_state);
4521 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4522 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
4524 for (i = 0; i < 4; i++)
4525 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4526 sbi->s_def_hash_version = es->s_def_hash_version;
4527 if (ext4_has_feature_dir_index(sb)) {
4528 i = le32_to_cpu(es->s_flags);
4529 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4530 sbi->s_hash_unsigned = 3;
4531 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
4532 #ifdef __CHAR_UNSIGNED__
4535 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4536 sbi->s_hash_unsigned = 3;
4540 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
4545 /* Handle clustersize */
4546 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4547 if (ext4_has_feature_bigalloc(sb)) {
4548 if (clustersize < blocksize) {
4549 ext4_msg(sb, KERN_ERR,
4550 "cluster size (%d) smaller than "
4551 "block size (%d)", clustersize, blocksize);
4554 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4555 le32_to_cpu(es->s_log_block_size);
4556 sbi->s_clusters_per_group =
4557 le32_to_cpu(es->s_clusters_per_group);
4558 if (sbi->s_clusters_per_group > blocksize * 8) {
4559 ext4_msg(sb, KERN_ERR,
4560 "#clusters per group too big: %lu",
4561 sbi->s_clusters_per_group);
4564 if (sbi->s_blocks_per_group !=
4565 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4566 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4567 "clusters per group (%lu) inconsistent",
4568 sbi->s_blocks_per_group,
4569 sbi->s_clusters_per_group);
4573 if (clustersize != blocksize) {
4574 ext4_msg(sb, KERN_ERR,
4575 "fragment/cluster size (%d) != "
4576 "block size (%d)", clustersize, blocksize);
4579 if (sbi->s_blocks_per_group > blocksize * 8) {
4580 ext4_msg(sb, KERN_ERR,
4581 "#blocks per group too big: %lu",
4582 sbi->s_blocks_per_group);
4585 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4586 sbi->s_cluster_bits = 0;
4588 sbi->s_cluster_ratio = clustersize / blocksize;
4590 /* Do we have standard group size of clustersize * 8 blocks ? */
4591 if (sbi->s_blocks_per_group == clustersize << 3)
4592 set_opt2(sb, STD_GROUP_SIZE);
4595 * Test whether we have more sectors than will fit in sector_t,
4596 * and whether the max offset is addressable by the page cache.
4598 err = generic_check_addressable(sb->s_blocksize_bits,
4599 ext4_blocks_count(es));
4601 ext4_msg(sb, KERN_ERR, "filesystem"
4602 " too large to mount safely on this system");
4606 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4609 /* check blocks count against device size */
4610 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4611 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4612 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4613 "exceeds size of device (%llu blocks)",
4614 ext4_blocks_count(es), blocks_count);
4619 * It makes no sense for the first data block to be beyond the end
4620 * of the filesystem.
4622 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4623 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4624 "block %u is beyond end of filesystem (%llu)",
4625 le32_to_cpu(es->s_first_data_block),
4626 ext4_blocks_count(es));
4629 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4630 (sbi->s_cluster_ratio == 1)) {
4631 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4632 "block is 0 with a 1k block and cluster size");
4636 blocks_count = (ext4_blocks_count(es) -
4637 le32_to_cpu(es->s_first_data_block) +
4638 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4639 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4640 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4641 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4642 "(block count %llu, first data block %u, "
4643 "blocks per group %lu)", blocks_count,
4644 ext4_blocks_count(es),
4645 le32_to_cpu(es->s_first_data_block),
4646 EXT4_BLOCKS_PER_GROUP(sb));
4649 sbi->s_groups_count = blocks_count;
4650 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4651 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4652 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4653 le32_to_cpu(es->s_inodes_count)) {
4654 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4655 le32_to_cpu(es->s_inodes_count),
4656 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4660 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4661 EXT4_DESC_PER_BLOCK(sb);
4662 if (ext4_has_feature_meta_bg(sb)) {
4663 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4664 ext4_msg(sb, KERN_WARNING,
4665 "first meta block group too large: %u "
4666 "(group descriptor block count %u)",
4667 le32_to_cpu(es->s_first_meta_bg), db_count);
4671 rcu_assign_pointer(sbi->s_group_desc,
4672 kvmalloc_array(db_count,
4673 sizeof(struct buffer_head *),
4675 if (sbi->s_group_desc == NULL) {
4676 ext4_msg(sb, KERN_ERR, "not enough memory");
4681 bgl_lock_init(sbi->s_blockgroup_lock);
4683 /* Pre-read the descriptors into the buffer cache */
4684 for (i = 0; i < db_count; i++) {
4685 block = descriptor_loc(sb, logical_sb_block, i);
4686 ext4_sb_breadahead_unmovable(sb, block);
4689 for (i = 0; i < db_count; i++) {
4690 struct buffer_head *bh;
4692 block = descriptor_loc(sb, logical_sb_block, i);
4693 bh = ext4_sb_bread_unmovable(sb, block);
4695 ext4_msg(sb, KERN_ERR,
4696 "can't read group descriptor %d", i);
4703 rcu_dereference(sbi->s_group_desc)[i] = bh;
4706 sbi->s_gdb_count = db_count;
4707 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4708 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4709 ret = -EFSCORRUPTED;
4713 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
4715 /* Register extent status tree shrinker */
4716 if (ext4_es_register_shrinker(sbi))
4719 sbi->s_stripe = ext4_get_stripe_size(sbi);
4720 sbi->s_extent_max_zeroout_kb = 32;
4723 * set up enough so that it can read an inode
4725 sb->s_op = &ext4_sops;
4726 sb->s_export_op = &ext4_export_ops;
4727 sb->s_xattr = ext4_xattr_handlers;
4728 #ifdef CONFIG_FS_ENCRYPTION
4729 sb->s_cop = &ext4_cryptops;
4731 #ifdef CONFIG_FS_VERITY
4732 sb->s_vop = &ext4_verityops;
4735 sb->dq_op = &ext4_quota_operations;
4736 if (ext4_has_feature_quota(sb))
4737 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4739 sb->s_qcop = &ext4_qctl_operations;
4740 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4742 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4744 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4745 mutex_init(&sbi->s_orphan_lock);
4749 needs_recovery = (es->s_last_orphan != 0 ||
4750 ext4_has_feature_journal_needs_recovery(sb));
4752 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
4753 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4754 goto failed_mount3a;
4757 * The first inode we look at is the journal inode. Don't try
4758 * root first: it may be modified in the journal!
4760 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4761 err = ext4_load_journal(sb, es, journal_devnum);
4763 goto failed_mount3a;
4764 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4765 ext4_has_feature_journal_needs_recovery(sb)) {
4766 ext4_msg(sb, KERN_ERR, "required journal recovery "
4767 "suppressed and not mounted read-only");
4768 goto failed_mount_wq;
4770 /* Nojournal mode, all journal mount options are illegal */
4771 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4772 ext4_msg(sb, KERN_ERR, "can't mount with "
4773 "journal_checksum, fs mounted w/o journal");
4774 goto failed_mount_wq;
4776 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4777 ext4_msg(sb, KERN_ERR, "can't mount with "
4778 "journal_async_commit, fs mounted w/o journal");
4779 goto failed_mount_wq;
4781 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4782 ext4_msg(sb, KERN_ERR, "can't mount with "
4783 "commit=%lu, fs mounted w/o journal",
4784 sbi->s_commit_interval / HZ);
4785 goto failed_mount_wq;
4787 if (EXT4_MOUNT_DATA_FLAGS &
4788 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4789 ext4_msg(sb, KERN_ERR, "can't mount with "
4790 "data=, fs mounted w/o journal");
4791 goto failed_mount_wq;
4793 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
4794 clear_opt(sb, JOURNAL_CHECKSUM);
4795 clear_opt(sb, DATA_FLAGS);
4796 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4797 sbi->s_journal = NULL;
4802 if (ext4_has_feature_64bit(sb) &&
4803 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4804 JBD2_FEATURE_INCOMPAT_64BIT)) {
4805 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4806 goto failed_mount_wq;
4809 if (!set_journal_csum_feature_set(sb)) {
4810 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4812 goto failed_mount_wq;
4815 /* We have now updated the journal if required, so we can
4816 * validate the data journaling mode. */
4817 switch (test_opt(sb, DATA_FLAGS)) {
4819 /* No mode set, assume a default based on the journal
4820 * capabilities: ORDERED_DATA if the journal can
4821 * cope, else JOURNAL_DATA
4823 if (jbd2_journal_check_available_features
4824 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4825 set_opt(sb, ORDERED_DATA);
4826 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4828 set_opt(sb, JOURNAL_DATA);
4829 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4833 case EXT4_MOUNT_ORDERED_DATA:
4834 case EXT4_MOUNT_WRITEBACK_DATA:
4835 if (!jbd2_journal_check_available_features
4836 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4837 ext4_msg(sb, KERN_ERR, "Journal does not support "
4838 "requested data journaling mode");
4839 goto failed_mount_wq;
4845 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4846 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4847 ext4_msg(sb, KERN_ERR, "can't mount with "
4848 "journal_async_commit in data=ordered mode");
4849 goto failed_mount_wq;
4852 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4854 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4855 sbi->s_journal->j_submit_inode_data_buffers =
4856 ext4_journal_submit_inode_data_buffers;
4857 sbi->s_journal->j_finish_inode_data_buffers =
4858 ext4_journal_finish_inode_data_buffers;
4861 if (!test_opt(sb, NO_MBCACHE)) {
4862 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4863 if (!sbi->s_ea_block_cache) {
4864 ext4_msg(sb, KERN_ERR,
4865 "Failed to create ea_block_cache");
4866 goto failed_mount_wq;
4869 if (ext4_has_feature_ea_inode(sb)) {
4870 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4871 if (!sbi->s_ea_inode_cache) {
4872 ext4_msg(sb, KERN_ERR,
4873 "Failed to create ea_inode_cache");
4874 goto failed_mount_wq;
4879 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
4880 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
4881 goto failed_mount_wq;
4884 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
4885 !ext4_has_feature_encrypt(sb)) {
4886 ext4_set_feature_encrypt(sb);
4887 ext4_commit_super(sb, 1);
4891 * Get the # of file system overhead blocks from the
4892 * superblock if present.
4894 if (es->s_overhead_clusters)
4895 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4897 err = ext4_calculate_overhead(sb);
4899 goto failed_mount_wq;
4903 * The maximum number of concurrent works can be high and
4904 * concurrency isn't really necessary. Limit it to 1.
4906 EXT4_SB(sb)->rsv_conversion_wq =
4907 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4908 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4909 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4915 * The jbd2_journal_load will have done any necessary log recovery,
4916 * so we can safely mount the rest of the filesystem now.
4919 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
4921 ext4_msg(sb, KERN_ERR, "get root inode failed");
4922 ret = PTR_ERR(root);
4926 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4927 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4932 #ifdef CONFIG_UNICODE
4933 if (sbi->s_encoding)
4934 sb->s_d_op = &ext4_dentry_ops;
4937 sb->s_root = d_make_root(root);
4939 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4944 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4945 if (ret == -EROFS) {
4946 sb->s_flags |= SB_RDONLY;
4949 goto failed_mount4a;
4951 ext4_set_resv_clusters(sb);
4953 if (test_opt(sb, BLOCK_VALIDITY)) {
4954 err = ext4_setup_system_zone(sb);
4956 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4958 goto failed_mount4a;
4963 err = ext4_mb_init(sb);
4965 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4970 block = ext4_count_free_clusters(sb);
4971 ext4_free_blocks_count_set(sbi->s_es,
4972 EXT4_C2B(sbi, block));
4973 ext4_superblock_csum_set(sb);
4974 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4977 unsigned long freei = ext4_count_free_inodes(sb);
4978 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4979 ext4_superblock_csum_set(sb);
4980 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4984 err = percpu_counter_init(&sbi->s_dirs_counter,
4985 ext4_count_dirs(sb), GFP_KERNEL);
4987 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4990 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
4993 ext4_msg(sb, KERN_ERR, "insufficient memory");
4997 if (ext4_has_feature_flex_bg(sb))
4998 if (!ext4_fill_flex_info(sb)) {
4999 ext4_msg(sb, KERN_ERR,
5000 "unable to initialize "
5001 "flex_bg meta info!");
5005 err = ext4_register_li_request(sb, first_not_zeroed);
5009 err = ext4_register_sysfs(sb);
5014 /* Enable quota usage during mount. */
5015 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
5016 err = ext4_enable_quotas(sb);
5020 #endif /* CONFIG_QUOTA */
5023 * Save the original bdev mapping's wb_err value which could be
5024 * used to detect the metadata async write error.
5026 spin_lock_init(&sbi->s_bdev_wb_lock);
5027 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
5028 &sbi->s_bdev_wb_err);
5029 sb->s_bdev->bd_super = sb;
5030 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
5031 ext4_orphan_cleanup(sb, es);
5032 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
5033 if (needs_recovery) {
5034 ext4_msg(sb, KERN_INFO, "recovery complete");
5035 err = ext4_mark_recovery_complete(sb, es);
5039 if (EXT4_SB(sb)->s_journal) {
5040 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
5041 descr = " journalled data mode";
5042 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
5043 descr = " ordered data mode";
5045 descr = " writeback data mode";
5047 descr = "out journal";
5049 if (test_opt(sb, DISCARD)) {
5050 struct request_queue *q = bdev_get_queue(sb->s_bdev);
5051 if (!blk_queue_discard(q))
5052 ext4_msg(sb, KERN_WARNING,
5053 "mounting with \"discard\" option, but "
5054 "the device does not support discard");
5057 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
5058 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5059 "Opts: %.*s%s%s", descr,
5060 (int) sizeof(sbi->s_es->s_mount_opts),
5061 sbi->s_es->s_mount_opts,
5062 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
5064 if (es->s_error_count)
5065 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
5067 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5068 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
5069 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
5070 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
5071 atomic_set(&sbi->s_warning_count, 0);
5072 atomic_set(&sbi->s_msg_count, 0);
5079 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5083 ext4_unregister_sysfs(sb);
5084 kobject_put(&sbi->s_kobj);
5086 ext4_unregister_li_request(sb);
5088 ext4_mb_release(sb);
5090 flex_groups = rcu_dereference(sbi->s_flex_groups);
5092 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
5093 kvfree(flex_groups[i]);
5094 kvfree(flex_groups);
5097 percpu_counter_destroy(&sbi->s_freeclusters_counter);
5098 percpu_counter_destroy(&sbi->s_freeinodes_counter);
5099 percpu_counter_destroy(&sbi->s_dirs_counter);
5100 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
5101 percpu_free_rwsem(&sbi->s_writepages_rwsem);
5103 ext4_ext_release(sb);
5104 ext4_release_system_zone(sb);
5109 ext4_msg(sb, KERN_ERR, "mount failed");
5110 if (EXT4_SB(sb)->rsv_conversion_wq)
5111 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
5113 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5114 sbi->s_ea_inode_cache = NULL;
5116 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5117 sbi->s_ea_block_cache = NULL;
5119 if (sbi->s_journal) {
5120 jbd2_journal_destroy(sbi->s_journal);
5121 sbi->s_journal = NULL;
5124 ext4_es_unregister_shrinker(sbi);
5126 del_timer_sync(&sbi->s_err_report);
5128 kthread_stop(sbi->s_mmp_tsk);
5131 group_desc = rcu_dereference(sbi->s_group_desc);
5132 for (i = 0; i < db_count; i++)
5133 brelse(group_desc[i]);
5137 if (sbi->s_chksum_driver)
5138 crypto_free_shash(sbi->s_chksum_driver);
5140 #ifdef CONFIG_UNICODE
5141 utf8_unload(sbi->s_encoding);
5145 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5146 kfree(get_qf_name(sb, sbi, i));
5148 fscrypt_free_dummy_context(&sbi->s_dummy_enc_ctx);
5149 ext4_blkdev_remove(sbi);
5152 sb->s_fs_info = NULL;
5153 kfree(sbi->s_blockgroup_lock);
5157 fs_put_dax(dax_dev);
5158 return err ? err : ret;
5162 * Setup any per-fs journal parameters now. We'll do this both on
5163 * initial mount, once the journal has been initialised but before we've
5164 * done any recovery; and again on any subsequent remount.
5166 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
5168 struct ext4_sb_info *sbi = EXT4_SB(sb);
5170 journal->j_commit_interval = sbi->s_commit_interval;
5171 journal->j_min_batch_time = sbi->s_min_batch_time;
5172 journal->j_max_batch_time = sbi->s_max_batch_time;
5174 write_lock(&journal->j_state_lock);
5175 if (test_opt(sb, BARRIER))
5176 journal->j_flags |= JBD2_BARRIER;
5178 journal->j_flags &= ~JBD2_BARRIER;
5179 if (test_opt(sb, DATA_ERR_ABORT))
5180 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5182 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
5183 write_unlock(&journal->j_state_lock);
5186 static struct inode *ext4_get_journal_inode(struct super_block *sb,
5187 unsigned int journal_inum)
5189 struct inode *journal_inode;
5192 * Test for the existence of a valid inode on disk. Bad things
5193 * happen if we iget() an unused inode, as the subsequent iput()
5194 * will try to delete it.
5196 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
5197 if (IS_ERR(journal_inode)) {
5198 ext4_msg(sb, KERN_ERR, "no journal found");
5201 if (!journal_inode->i_nlink) {
5202 make_bad_inode(journal_inode);
5203 iput(journal_inode);
5204 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
5208 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
5209 journal_inode, journal_inode->i_size);
5210 if (!S_ISREG(journal_inode->i_mode)) {
5211 ext4_msg(sb, KERN_ERR, "invalid journal inode");
5212 iput(journal_inode);
5215 return journal_inode;
5218 static journal_t *ext4_get_journal(struct super_block *sb,
5219 unsigned int journal_inum)
5221 struct inode *journal_inode;
5224 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5227 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5231 journal = jbd2_journal_init_inode(journal_inode);
5233 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5234 iput(journal_inode);
5237 journal->j_private = sb;
5238 ext4_init_journal_params(sb, journal);
5242 static journal_t *ext4_get_dev_journal(struct super_block *sb,
5245 struct buffer_head *bh;
5249 int hblock, blocksize;
5250 ext4_fsblk_t sb_block;
5251 unsigned long offset;
5252 struct ext4_super_block *es;
5253 struct block_device *bdev;
5255 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5258 bdev = ext4_blkdev_get(j_dev, sb);
5262 blocksize = sb->s_blocksize;
5263 hblock = bdev_logical_block_size(bdev);
5264 if (blocksize < hblock) {
5265 ext4_msg(sb, KERN_ERR,
5266 "blocksize too small for journal device");
5270 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5271 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5272 set_blocksize(bdev, blocksize);
5273 if (!(bh = __bread(bdev, sb_block, blocksize))) {
5274 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5275 "external journal");
5279 es = (struct ext4_super_block *) (bh->b_data + offset);
5280 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5281 !(le32_to_cpu(es->s_feature_incompat) &
5282 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5283 ext4_msg(sb, KERN_ERR, "external journal has "
5289 if ((le32_to_cpu(es->s_feature_ro_compat) &
5290 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5291 es->s_checksum != ext4_superblock_csum(sb, es)) {
5292 ext4_msg(sb, KERN_ERR, "external journal has "
5293 "corrupt superblock");
5298 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5299 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5304 len = ext4_blocks_count(es);
5305 start = sb_block + 1;
5306 brelse(bh); /* we're done with the superblock */
5308 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
5309 start, len, blocksize);
5311 ext4_msg(sb, KERN_ERR, "failed to create device journal");
5314 journal->j_private = sb;
5315 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) {
5316 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
5319 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5320 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5321 "user (unsupported) - %d",
5322 be32_to_cpu(journal->j_superblock->s_nr_users));
5325 EXT4_SB(sb)->s_journal_bdev = bdev;
5326 ext4_init_journal_params(sb, journal);
5330 jbd2_journal_destroy(journal);
5332 ext4_blkdev_put(bdev);
5336 static int ext4_load_journal(struct super_block *sb,
5337 struct ext4_super_block *es,
5338 unsigned long journal_devnum)
5341 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5344 int really_read_only;
5347 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5348 return -EFSCORRUPTED;
5350 if (journal_devnum &&
5351 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5352 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5353 "numbers have changed");
5354 journal_dev = new_decode_dev(journal_devnum);
5356 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5358 if (journal_inum && journal_dev) {
5359 ext4_msg(sb, KERN_ERR,
5360 "filesystem has both journal inode and journal device!");
5365 journal = ext4_get_journal(sb, journal_inum);
5369 journal = ext4_get_dev_journal(sb, journal_dev);
5374 journal_dev_ro = bdev_read_only(journal->j_dev);
5375 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5377 if (journal_dev_ro && !sb_rdonly(sb)) {
5378 ext4_msg(sb, KERN_ERR,
5379 "journal device read-only, try mounting with '-o ro'");
5385 * Are we loading a blank journal or performing recovery after a
5386 * crash? For recovery, we need to check in advance whether we
5387 * can get read-write access to the device.
5389 if (ext4_has_feature_journal_needs_recovery(sb)) {
5390 if (sb_rdonly(sb)) {
5391 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5392 "required on readonly filesystem");
5393 if (really_read_only) {
5394 ext4_msg(sb, KERN_ERR, "write access "
5395 "unavailable, cannot proceed "
5396 "(try mounting with noload)");
5400 ext4_msg(sb, KERN_INFO, "write access will "
5401 "be enabled during recovery");
5405 if (!(journal->j_flags & JBD2_BARRIER))
5406 ext4_msg(sb, KERN_INFO, "barriers disabled");
5408 if (!ext4_has_feature_journal_needs_recovery(sb))
5409 err = jbd2_journal_wipe(journal, !really_read_only);
5411 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5413 memcpy(save, ((char *) es) +
5414 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
5415 err = jbd2_journal_load(journal);
5417 memcpy(((char *) es) + EXT4_S_ERR_START,
5418 save, EXT4_S_ERR_LEN);
5423 ext4_msg(sb, KERN_ERR, "error loading journal");
5427 EXT4_SB(sb)->s_journal = journal;
5428 err = ext4_clear_journal_err(sb, es);
5430 EXT4_SB(sb)->s_journal = NULL;
5431 jbd2_journal_destroy(journal);
5435 if (!really_read_only && journal_devnum &&
5436 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5437 es->s_journal_dev = cpu_to_le32(journal_devnum);
5439 /* Make sure we flush the recovery flag to disk. */
5440 ext4_commit_super(sb, 1);
5446 jbd2_journal_destroy(journal);
5450 static int ext4_commit_super(struct super_block *sb, int sync)
5452 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
5453 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
5456 if (!sbh || block_device_ejected(sb))
5460 * If the file system is mounted read-only, don't update the
5461 * superblock write time. This avoids updating the superblock
5462 * write time when we are mounting the root file system
5463 * read/only but we need to replay the journal; at that point,
5464 * for people who are east of GMT and who make their clock
5465 * tick in localtime for Windows bug-for-bug compatibility,
5466 * the clock is set in the future, and this will cause e2fsck
5467 * to complain and force a full file system check.
5469 if (!(sb->s_flags & SB_RDONLY))
5470 ext4_update_tstamp(es, s_wtime);
5471 if (sb->s_bdev->bd_part)
5472 es->s_kbytes_written =
5473 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
5474 ((part_stat_read(sb->s_bdev->bd_part,
5475 sectors[STAT_WRITE]) -
5476 EXT4_SB(sb)->s_sectors_written_start) >> 1));
5478 es->s_kbytes_written =
5479 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
5480 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
5481 ext4_free_blocks_count_set(es,
5482 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
5483 &EXT4_SB(sb)->s_freeclusters_counter)));
5484 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
5485 es->s_free_inodes_count =
5486 cpu_to_le32(percpu_counter_sum_positive(
5487 &EXT4_SB(sb)->s_freeinodes_counter));
5488 BUFFER_TRACE(sbh, "marking dirty");
5489 ext4_superblock_csum_set(sb);
5492 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
5494 * Oh, dear. A previous attempt to write the
5495 * superblock failed. This could happen because the
5496 * USB device was yanked out. Or it could happen to
5497 * be a transient write error and maybe the block will
5498 * be remapped. Nothing we can do but to retry the
5499 * write and hope for the best.
5501 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5502 "superblock detected");
5503 clear_buffer_write_io_error(sbh);
5504 set_buffer_uptodate(sbh);
5506 mark_buffer_dirty(sbh);
5509 error = __sync_dirty_buffer(sbh,
5510 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
5511 if (buffer_write_io_error(sbh)) {
5512 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5514 clear_buffer_write_io_error(sbh);
5515 set_buffer_uptodate(sbh);
5522 * Have we just finished recovery? If so, and if we are mounting (or
5523 * remounting) the filesystem readonly, then we will end up with a
5524 * consistent fs on disk. Record that fact.
5526 static int ext4_mark_recovery_complete(struct super_block *sb,
5527 struct ext4_super_block *es)
5530 journal_t *journal = EXT4_SB(sb)->s_journal;
5532 if (!ext4_has_feature_journal(sb)) {
5533 if (journal != NULL) {
5534 ext4_error(sb, "Journal got removed while the fs was "
5536 return -EFSCORRUPTED;
5540 jbd2_journal_lock_updates(journal);
5541 err = jbd2_journal_flush(journal);
5545 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
5546 ext4_clear_feature_journal_needs_recovery(sb);
5547 ext4_commit_super(sb, 1);
5550 jbd2_journal_unlock_updates(journal);
5555 * If we are mounting (or read-write remounting) a filesystem whose journal
5556 * has recorded an error from a previous lifetime, move that error to the
5557 * main filesystem now.
5559 static int ext4_clear_journal_err(struct super_block *sb,
5560 struct ext4_super_block *es)
5566 if (!ext4_has_feature_journal(sb)) {
5567 ext4_error(sb, "Journal got removed while the fs was mounted!");
5568 return -EFSCORRUPTED;
5571 journal = EXT4_SB(sb)->s_journal;
5574 * Now check for any error status which may have been recorded in the
5575 * journal by a prior ext4_error() or ext4_abort()
5578 j_errno = jbd2_journal_errno(journal);
5582 errstr = ext4_decode_error(sb, j_errno, nbuf);
5583 ext4_warning(sb, "Filesystem error recorded "
5584 "from previous mount: %s", errstr);
5585 ext4_warning(sb, "Marking fs in need of filesystem check.");
5587 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5588 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5589 ext4_commit_super(sb, 1);
5591 jbd2_journal_clear_err(journal);
5592 jbd2_journal_update_sb_errno(journal);
5598 * Force the running and committing transactions to commit,
5599 * and wait on the commit.
5601 int ext4_force_commit(struct super_block *sb)
5608 journal = EXT4_SB(sb)->s_journal;
5609 return ext4_journal_force_commit(journal);
5612 static int ext4_sync_fs(struct super_block *sb, int wait)
5616 bool needs_barrier = false;
5617 struct ext4_sb_info *sbi = EXT4_SB(sb);
5619 if (unlikely(ext4_forced_shutdown(sbi)))
5622 trace_ext4_sync_fs(sb, wait);
5623 flush_workqueue(sbi->rsv_conversion_wq);
5625 * Writeback quota in non-journalled quota case - journalled quota has
5628 dquot_writeback_dquots(sb, -1);
5630 * Data writeback is possible w/o journal transaction, so barrier must
5631 * being sent at the end of the function. But we can skip it if
5632 * transaction_commit will do it for us.
5634 if (sbi->s_journal) {
5635 target = jbd2_get_latest_transaction(sbi->s_journal);
5636 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5637 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5638 needs_barrier = true;
5640 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5642 ret = jbd2_log_wait_commit(sbi->s_journal,
5645 } else if (wait && test_opt(sb, BARRIER))
5646 needs_barrier = true;
5647 if (needs_barrier) {
5649 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL);
5658 * LVM calls this function before a (read-only) snapshot is created. This
5659 * gives us a chance to flush the journal completely and mark the fs clean.
5661 * Note that only this function cannot bring a filesystem to be in a clean
5662 * state independently. It relies on upper layer to stop all data & metadata
5665 static int ext4_freeze(struct super_block *sb)
5673 journal = EXT4_SB(sb)->s_journal;
5676 /* Now we set up the journal barrier. */
5677 jbd2_journal_lock_updates(journal);
5680 * Don't clear the needs_recovery flag if we failed to
5681 * flush the journal.
5683 error = jbd2_journal_flush(journal);
5687 /* Journal blocked and flushed, clear needs_recovery flag. */
5688 ext4_clear_feature_journal_needs_recovery(sb);
5691 error = ext4_commit_super(sb, 1);
5694 /* we rely on upper layer to stop further updates */
5695 jbd2_journal_unlock_updates(journal);
5700 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5701 * flag here, even though the filesystem is not technically dirty yet.
5703 static int ext4_unfreeze(struct super_block *sb)
5705 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
5708 if (EXT4_SB(sb)->s_journal) {
5709 /* Reset the needs_recovery flag before the fs is unlocked. */
5710 ext4_set_feature_journal_needs_recovery(sb);
5713 ext4_commit_super(sb, 1);
5718 * Structure to save mount options for ext4_remount's benefit
5720 struct ext4_mount_options {
5721 unsigned long s_mount_opt;
5722 unsigned long s_mount_opt2;
5725 unsigned long s_commit_interval;
5726 u32 s_min_batch_time, s_max_batch_time;
5729 char *s_qf_names[EXT4_MAXQUOTAS];
5733 static int ext4_remount(struct super_block *sb, int *flags, char *data)
5735 struct ext4_super_block *es;
5736 struct ext4_sb_info *sbi = EXT4_SB(sb);
5737 unsigned long old_sb_flags, vfs_flags;
5738 struct ext4_mount_options old_opts;
5739 int enable_quota = 0;
5741 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5745 char *to_free[EXT4_MAXQUOTAS];
5747 char *orig_data = kstrdup(data, GFP_KERNEL);
5749 if (data && !orig_data)
5752 /* Store the original options */
5753 old_sb_flags = sb->s_flags;
5754 old_opts.s_mount_opt = sbi->s_mount_opt;
5755 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
5756 old_opts.s_resuid = sbi->s_resuid;
5757 old_opts.s_resgid = sbi->s_resgid;
5758 old_opts.s_commit_interval = sbi->s_commit_interval;
5759 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5760 old_opts.s_max_batch_time = sbi->s_max_batch_time;
5762 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
5763 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5764 if (sbi->s_qf_names[i]) {
5765 char *qf_name = get_qf_name(sb, sbi, i);
5767 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
5768 if (!old_opts.s_qf_names[i]) {
5769 for (j = 0; j < i; j++)
5770 kfree(old_opts.s_qf_names[j]);
5775 old_opts.s_qf_names[i] = NULL;
5777 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5778 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
5781 * Some options can be enabled by ext4 and/or by VFS mount flag
5782 * either way we need to make sure it matches in both *flags and
5783 * s_flags. Copy those selected flags from *flags to s_flags
5785 vfs_flags = SB_LAZYTIME | SB_I_VERSION;
5786 sb->s_flags = (sb->s_flags & ~vfs_flags) | (*flags & vfs_flags);
5788 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
5793 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5794 test_opt(sb, JOURNAL_CHECKSUM)) {
5795 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5796 "during remount not supported; ignoring");
5797 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5800 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5801 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5802 ext4_msg(sb, KERN_ERR, "can't mount with "
5803 "both data=journal and delalloc");
5807 if (test_opt(sb, DIOREAD_NOLOCK)) {
5808 ext4_msg(sb, KERN_ERR, "can't mount with "
5809 "both data=journal and dioread_nolock");
5813 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5814 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5815 ext4_msg(sb, KERN_ERR, "can't mount with "
5816 "journal_async_commit in data=ordered mode");
5822 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5823 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5828 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
5829 ext4_abort(sb, EXT4_ERR_ESHUTDOWN, "Abort forced by user");
5831 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5832 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5836 if (sbi->s_journal) {
5837 ext4_init_journal_params(sb, sbi->s_journal);
5838 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5841 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
5842 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
5847 if (*flags & SB_RDONLY) {
5848 err = sync_filesystem(sb);
5851 err = dquot_suspend(sb, -1);
5856 * First of all, the unconditional stuff we have to do
5857 * to disable replay of the journal when we next remount
5859 sb->s_flags |= SB_RDONLY;
5862 * OK, test if we are remounting a valid rw partition
5863 * readonly, and if so set the rdonly flag and then
5864 * mark the partition as valid again.
5866 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5867 (sbi->s_mount_state & EXT4_VALID_FS))
5868 es->s_state = cpu_to_le16(sbi->s_mount_state);
5870 if (sbi->s_journal) {
5872 * We let remount-ro finish even if marking fs
5873 * as clean failed...
5875 ext4_mark_recovery_complete(sb, es);
5878 kthread_stop(sbi->s_mmp_tsk);
5880 /* Make sure we can mount this feature set readwrite */
5881 if (ext4_has_feature_readonly(sb) ||
5882 !ext4_feature_set_ok(sb, 0)) {
5887 * Make sure the group descriptor checksums
5888 * are sane. If they aren't, refuse to remount r/w.
5890 for (g = 0; g < sbi->s_groups_count; g++) {
5891 struct ext4_group_desc *gdp =
5892 ext4_get_group_desc(sb, g, NULL);
5894 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5895 ext4_msg(sb, KERN_ERR,
5896 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5897 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5898 le16_to_cpu(gdp->bg_checksum));
5905 * If we have an unprocessed orphan list hanging
5906 * around from a previously readonly bdev mount,
5907 * require a full umount/remount for now.
5909 if (es->s_last_orphan) {
5910 ext4_msg(sb, KERN_WARNING, "Couldn't "
5911 "remount RDWR because of unprocessed "
5912 "orphan inode list. Please "
5913 "umount/remount instead");
5919 * Mounting a RDONLY partition read-write, so reread
5920 * and store the current valid flag. (It may have
5921 * been changed by e2fsck since we originally mounted
5924 if (sbi->s_journal) {
5925 err = ext4_clear_journal_err(sb, es);
5929 sbi->s_mount_state = le16_to_cpu(es->s_state);
5931 err = ext4_setup_super(sb, es, 0);
5935 sb->s_flags &= ~SB_RDONLY;
5936 if (ext4_has_feature_mmp(sb))
5937 if (ext4_multi_mount_protect(sb,
5938 le64_to_cpu(es->s_mmp_block))) {
5947 * Reinitialize lazy itable initialization thread based on
5950 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
5951 ext4_unregister_li_request(sb);
5953 ext4_group_t first_not_zeroed;
5954 first_not_zeroed = ext4_has_uninit_itable(sb);
5955 ext4_register_li_request(sb, first_not_zeroed);
5959 * Handle creation of system zone data early because it can fail.
5960 * Releasing of existing data is done when we are sure remount will
5963 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
5964 err = ext4_setup_system_zone(sb);
5969 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5970 err = ext4_commit_super(sb, 1);
5976 /* Release old quota file names */
5977 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5978 kfree(old_opts.s_qf_names[i]);
5980 if (sb_any_quota_suspended(sb))
5981 dquot_resume(sb, -1);
5982 else if (ext4_has_feature_quota(sb)) {
5983 err = ext4_enable_quotas(sb);
5989 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
5990 ext4_release_system_zone(sb);
5993 * Some options can be enabled by ext4 and/or by VFS mount flag
5994 * either way we need to make sure it matches in both *flags and
5995 * s_flags. Copy those selected flags from s_flags to *flags
5997 *flags = (*flags & ~vfs_flags) | (sb->s_flags & vfs_flags);
5999 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
6004 sb->s_flags = old_sb_flags;
6005 sbi->s_mount_opt = old_opts.s_mount_opt;
6006 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
6007 sbi->s_resuid = old_opts.s_resuid;
6008 sbi->s_resgid = old_opts.s_resgid;
6009 sbi->s_commit_interval = old_opts.s_commit_interval;
6010 sbi->s_min_batch_time = old_opts.s_min_batch_time;
6011 sbi->s_max_batch_time = old_opts.s_max_batch_time;
6012 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6013 ext4_release_system_zone(sb);
6015 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
6016 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
6017 to_free[i] = get_qf_name(sb, sbi, i);
6018 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
6021 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6029 static int ext4_statfs_project(struct super_block *sb,
6030 kprojid_t projid, struct kstatfs *buf)
6033 struct dquot *dquot;
6037 qid = make_kqid_projid(projid);
6038 dquot = dqget(sb, qid);
6040 return PTR_ERR(dquot);
6041 spin_lock(&dquot->dq_dqb_lock);
6043 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6044 dquot->dq_dqb.dqb_bhardlimit);
6045 limit >>= sb->s_blocksize_bits;
6047 if (limit && buf->f_blocks > limit) {
6048 curblock = (dquot->dq_dqb.dqb_curspace +
6049 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
6050 buf->f_blocks = limit;
6051 buf->f_bfree = buf->f_bavail =
6052 (buf->f_blocks > curblock) ?
6053 (buf->f_blocks - curblock) : 0;
6056 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6057 dquot->dq_dqb.dqb_ihardlimit);
6058 if (limit && buf->f_files > limit) {
6059 buf->f_files = limit;
6061 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6062 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6065 spin_unlock(&dquot->dq_dqb_lock);
6071 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
6073 struct super_block *sb = dentry->d_sb;
6074 struct ext4_sb_info *sbi = EXT4_SB(sb);
6075 struct ext4_super_block *es = sbi->s_es;
6076 ext4_fsblk_t overhead = 0, resv_blocks;
6079 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
6081 if (!test_opt(sb, MINIX_DF))
6082 overhead = sbi->s_overhead;
6084 buf->f_type = EXT4_SUPER_MAGIC;
6085 buf->f_bsize = sb->s_blocksize;
6086 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
6087 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6088 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
6089 /* prevent underflow in case that few free space is available */
6090 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
6091 buf->f_bavail = buf->f_bfree -
6092 (ext4_r_blocks_count(es) + resv_blocks);
6093 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
6095 buf->f_files = le32_to_cpu(es->s_inodes_count);
6096 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
6097 buf->f_namelen = EXT4_NAME_LEN;
6098 fsid = le64_to_cpup((void *)es->s_uuid) ^
6099 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
6100 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
6101 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
6104 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6105 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6106 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6115 * Helper functions so that transaction is started before we acquire dqio_sem
6116 * to keep correct lock ordering of transaction > dqio_sem
6118 static inline struct inode *dquot_to_inode(struct dquot *dquot)
6120 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
6123 static int ext4_write_dquot(struct dquot *dquot)
6127 struct inode *inode;
6129 inode = dquot_to_inode(dquot);
6130 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
6131 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
6133 return PTR_ERR(handle);
6134 ret = dquot_commit(dquot);
6135 err = ext4_journal_stop(handle);
6141 static int ext4_acquire_dquot(struct dquot *dquot)
6146 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6147 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
6149 return PTR_ERR(handle);
6150 ret = dquot_acquire(dquot);
6151 err = ext4_journal_stop(handle);
6157 static int ext4_release_dquot(struct dquot *dquot)
6162 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6163 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
6164 if (IS_ERR(handle)) {
6165 /* Release dquot anyway to avoid endless cycle in dqput() */
6166 dquot_release(dquot);
6167 return PTR_ERR(handle);
6169 ret = dquot_release(dquot);
6170 err = ext4_journal_stop(handle);
6176 static int ext4_mark_dquot_dirty(struct dquot *dquot)
6178 struct super_block *sb = dquot->dq_sb;
6179 struct ext4_sb_info *sbi = EXT4_SB(sb);
6181 /* Are we journaling quotas? */
6182 if (ext4_has_feature_quota(sb) ||
6183 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
6184 dquot_mark_dquot_dirty(dquot);
6185 return ext4_write_dquot(dquot);
6187 return dquot_mark_dquot_dirty(dquot);
6191 static int ext4_write_info(struct super_block *sb, int type)
6196 /* Data block + inode block */
6197 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
6199 return PTR_ERR(handle);
6200 ret = dquot_commit_info(sb, type);
6201 err = ext4_journal_stop(handle);
6208 * Turn on quotas during mount time - we need to find
6209 * the quota file and such...
6211 static int ext4_quota_on_mount(struct super_block *sb, int type)
6213 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
6214 EXT4_SB(sb)->s_jquota_fmt, type);
6217 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6219 struct ext4_inode_info *ei = EXT4_I(inode);
6221 /* The first argument of lockdep_set_subclass has to be
6222 * *exactly* the same as the argument to init_rwsem() --- in
6223 * this case, in init_once() --- or lockdep gets unhappy
6224 * because the name of the lock is set using the
6225 * stringification of the argument to init_rwsem().
6227 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6228 lockdep_set_subclass(&ei->i_data_sem, subclass);
6232 * Standard function to be called on quota_on
6234 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6235 const struct path *path)
6239 if (!test_opt(sb, QUOTA))
6242 /* Quotafile not on the same filesystem? */
6243 if (path->dentry->d_sb != sb)
6246 /* Quota already enabled for this file? */
6247 if (IS_NOQUOTA(d_inode(path->dentry)))
6250 /* Journaling quota? */
6251 if (EXT4_SB(sb)->s_qf_names[type]) {
6252 /* Quotafile not in fs root? */
6253 if (path->dentry->d_parent != sb->s_root)
6254 ext4_msg(sb, KERN_WARNING,
6255 "Quota file not on filesystem root. "
6256 "Journaled quota will not work");
6257 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6260 * Clear the flag just in case mount options changed since
6263 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
6267 * When we journal data on quota file, we have to flush journal to see
6268 * all updates to the file when we bypass pagecache...
6270 if (EXT4_SB(sb)->s_journal &&
6271 ext4_should_journal_data(d_inode(path->dentry))) {
6273 * We don't need to lock updates but journal_flush() could
6274 * otherwise be livelocked...
6276 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
6277 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
6278 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
6283 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6284 err = dquot_quota_on(sb, type, format_id, path);
6286 lockdep_set_quota_inode(path->dentry->d_inode,
6289 struct inode *inode = d_inode(path->dentry);
6293 * Set inode flags to prevent userspace from messing with quota
6294 * files. If this fails, we return success anyway since quotas
6295 * are already enabled and this is not a hard failure.
6298 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6301 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6302 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6303 S_NOATIME | S_IMMUTABLE);
6304 err = ext4_mark_inode_dirty(handle, inode);
6305 ext4_journal_stop(handle);
6307 inode_unlock(inode);
6312 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6316 struct inode *qf_inode;
6317 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6318 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6319 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6320 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6323 BUG_ON(!ext4_has_feature_quota(sb));
6325 if (!qf_inums[type])
6328 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
6329 if (IS_ERR(qf_inode)) {
6330 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
6331 return PTR_ERR(qf_inode);
6334 /* Don't account quota for quota files to avoid recursion */
6335 qf_inode->i_flags |= S_NOQUOTA;
6336 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
6337 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
6339 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
6345 /* Enable usage tracking for all quota types. */
6346 static int ext4_enable_quotas(struct super_block *sb)
6349 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6350 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6351 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6352 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6354 bool quota_mopt[EXT4_MAXQUOTAS] = {
6355 test_opt(sb, USRQUOTA),
6356 test_opt(sb, GRPQUOTA),
6357 test_opt(sb, PRJQUOTA),
6360 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
6361 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
6362 if (qf_inums[type]) {
6363 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
6364 DQUOT_USAGE_ENABLED |
6365 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
6368 "Failed to enable quota tracking "
6369 "(type=%d, err=%d). Please run "
6370 "e2fsck to fix.", type, err);
6371 for (type--; type >= 0; type--)
6372 dquot_quota_off(sb, type);
6381 static int ext4_quota_off(struct super_block *sb, int type)
6383 struct inode *inode = sb_dqopt(sb)->files[type];
6387 /* Force all delayed allocation blocks to be allocated.
6388 * Caller already holds s_umount sem */
6389 if (test_opt(sb, DELALLOC))
6390 sync_filesystem(sb);
6392 if (!inode || !igrab(inode))
6395 err = dquot_quota_off(sb, type);
6396 if (err || ext4_has_feature_quota(sb))
6401 * Update modification times of quota files when userspace can
6402 * start looking at them. If we fail, we return success anyway since
6403 * this is not a hard failure and quotas are already disabled.
6405 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6406 if (IS_ERR(handle)) {
6407 err = PTR_ERR(handle);
6410 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
6411 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
6412 inode->i_mtime = inode->i_ctime = current_time(inode);
6413 err = ext4_mark_inode_dirty(handle, inode);
6414 ext4_journal_stop(handle);
6416 inode_unlock(inode);
6418 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
6422 return dquot_quota_off(sb, type);
6425 /* Read data from quotafile - avoid pagecache and such because we cannot afford
6426 * acquiring the locks... As quota files are never truncated and quota code
6427 * itself serializes the operations (and no one else should touch the files)
6428 * we don't have to be afraid of races */
6429 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
6430 size_t len, loff_t off)
6432 struct inode *inode = sb_dqopt(sb)->files[type];
6433 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6434 int offset = off & (sb->s_blocksize - 1);
6437 struct buffer_head *bh;
6438 loff_t i_size = i_size_read(inode);
6442 if (off+len > i_size)
6445 while (toread > 0) {
6446 tocopy = sb->s_blocksize - offset < toread ?
6447 sb->s_blocksize - offset : toread;
6448 bh = ext4_bread(NULL, inode, blk, 0);
6451 if (!bh) /* A hole? */
6452 memset(data, 0, tocopy);
6454 memcpy(data, bh->b_data+offset, tocopy);
6464 /* Write to quotafile (we know the transaction is already started and has
6465 * enough credits) */
6466 static ssize_t ext4_quota_write(struct super_block *sb, int type,
6467 const char *data, size_t len, loff_t off)
6469 struct inode *inode = sb_dqopt(sb)->files[type];
6470 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6471 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
6473 struct buffer_head *bh;
6474 handle_t *handle = journal_current_handle();
6476 if (EXT4_SB(sb)->s_journal && !handle) {
6477 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6478 " cancelled because transaction is not started",
6479 (unsigned long long)off, (unsigned long long)len);
6483 * Since we account only one data block in transaction credits,
6484 * then it is impossible to cross a block boundary.
6486 if (sb->s_blocksize - offset < len) {
6487 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6488 " cancelled because not block aligned",
6489 (unsigned long long)off, (unsigned long long)len);
6494 bh = ext4_bread(handle, inode, blk,
6495 EXT4_GET_BLOCKS_CREATE |
6496 EXT4_GET_BLOCKS_METADATA_NOFAIL);
6497 } while (PTR_ERR(bh) == -ENOSPC &&
6498 ext4_should_retry_alloc(inode->i_sb, &retries));
6503 BUFFER_TRACE(bh, "get write access");
6504 err = ext4_journal_get_write_access(handle, bh);
6510 memcpy(bh->b_data+offset, data, len);
6511 flush_dcache_page(bh->b_page);
6513 err = ext4_handle_dirty_metadata(handle, NULL, bh);
6516 if (inode->i_size < off + len) {
6517 i_size_write(inode, off + len);
6518 EXT4_I(inode)->i_disksize = inode->i_size;
6519 err2 = ext4_mark_inode_dirty(handle, inode);
6520 if (unlikely(err2 && !err))
6523 return err ? err : len;
6527 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6528 const char *dev_name, void *data)
6530 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
6533 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
6534 static inline void register_as_ext2(void)
6536 int err = register_filesystem(&ext2_fs_type);
6539 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6542 static inline void unregister_as_ext2(void)
6544 unregister_filesystem(&ext2_fs_type);
6547 static inline int ext2_feature_set_ok(struct super_block *sb)
6549 if (ext4_has_unknown_ext2_incompat_features(sb))
6553 if (ext4_has_unknown_ext2_ro_compat_features(sb))
6558 static inline void register_as_ext2(void) { }
6559 static inline void unregister_as_ext2(void) { }
6560 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
6563 static inline void register_as_ext3(void)
6565 int err = register_filesystem(&ext3_fs_type);
6568 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6571 static inline void unregister_as_ext3(void)
6573 unregister_filesystem(&ext3_fs_type);
6576 static inline int ext3_feature_set_ok(struct super_block *sb)
6578 if (ext4_has_unknown_ext3_incompat_features(sb))
6580 if (!ext4_has_feature_journal(sb))
6584 if (ext4_has_unknown_ext3_ro_compat_features(sb))
6589 static struct file_system_type ext4_fs_type = {
6590 .owner = THIS_MODULE,
6592 .mount = ext4_mount,
6593 .kill_sb = kill_block_super,
6594 .fs_flags = FS_REQUIRES_DEV,
6596 MODULE_ALIAS_FS("ext4");
6598 /* Shared across all ext4 file systems */
6599 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
6601 static int __init ext4_init_fs(void)
6605 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
6606 ext4_li_info = NULL;
6607 mutex_init(&ext4_li_mtx);
6609 /* Build-time check for flags consistency */
6610 ext4_check_flag_values();
6612 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
6613 init_waitqueue_head(&ext4__ioend_wq[i]);
6615 err = ext4_init_es();
6619 err = ext4_init_pending();
6623 err = ext4_init_post_read_processing();
6627 err = ext4_init_pageio();
6631 err = ext4_init_system_zone();
6635 err = ext4_init_sysfs();
6639 err = ext4_init_mballoc();
6642 err = init_inodecache();
6647 err = register_filesystem(&ext4_fs_type);
6653 unregister_as_ext2();
6654 unregister_as_ext3();
6655 destroy_inodecache();
6657 ext4_exit_mballoc();
6661 ext4_exit_system_zone();
6665 ext4_exit_post_read_processing();
6667 ext4_exit_pending();
6674 static void __exit ext4_exit_fs(void)
6676 ext4_destroy_lazyinit_thread();
6677 unregister_as_ext2();
6678 unregister_as_ext3();
6679 unregister_filesystem(&ext4_fs_type);
6680 destroy_inodecache();
6681 ext4_exit_mballoc();
6683 ext4_exit_system_zone();
6685 ext4_exit_post_read_processing();
6687 ext4_exit_pending();
6690 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6691 MODULE_DESCRIPTION("Fourth Extended Filesystem");
6692 MODULE_LICENSE("GPL");
6693 MODULE_SOFTDEP("pre: crc32c");
6694 module_init(ext4_init_fs)
6695 module_exit(ext4_exit_fs)