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_policy(&sbi->s_dummy_enc_policy);
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 ext4_fc_init_inode(&ei->vfs_inode);
1335 mutex_init(&ei->i_fc_lock);
1336 return &ei->vfs_inode;
1339 static int ext4_drop_inode(struct inode *inode)
1341 int drop = generic_drop_inode(inode);
1344 drop = fscrypt_drop_inode(inode);
1346 trace_ext4_drop_inode(inode, drop);
1350 static void ext4_free_in_core_inode(struct inode *inode)
1352 fscrypt_free_inode(inode);
1353 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1354 pr_warn("%s: inode %ld still in fc list",
1355 __func__, inode->i_ino);
1357 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1360 static void ext4_destroy_inode(struct inode *inode)
1362 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1363 ext4_msg(inode->i_sb, KERN_ERR,
1364 "Inode %lu (%p): orphan list check failed!",
1365 inode->i_ino, EXT4_I(inode));
1366 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1367 EXT4_I(inode), sizeof(struct ext4_inode_info),
1373 static void init_once(void *foo)
1375 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1377 INIT_LIST_HEAD(&ei->i_orphan);
1378 init_rwsem(&ei->xattr_sem);
1379 init_rwsem(&ei->i_data_sem);
1380 init_rwsem(&ei->i_mmap_sem);
1381 inode_init_once(&ei->vfs_inode);
1382 ext4_fc_init_inode(&ei->vfs_inode);
1385 static int __init init_inodecache(void)
1387 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1388 sizeof(struct ext4_inode_info), 0,
1389 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1391 offsetof(struct ext4_inode_info, i_data),
1392 sizeof_field(struct ext4_inode_info, i_data),
1394 if (ext4_inode_cachep == NULL)
1399 static void destroy_inodecache(void)
1402 * Make sure all delayed rcu free inodes are flushed before we
1406 kmem_cache_destroy(ext4_inode_cachep);
1409 void ext4_clear_inode(struct inode *inode)
1412 invalidate_inode_buffers(inode);
1414 ext4_discard_preallocations(inode, 0);
1415 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1417 if (EXT4_I(inode)->jinode) {
1418 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1419 EXT4_I(inode)->jinode);
1420 jbd2_free_inode(EXT4_I(inode)->jinode);
1421 EXT4_I(inode)->jinode = NULL;
1423 fscrypt_put_encryption_info(inode);
1424 fsverity_cleanup_inode(inode);
1427 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1428 u64 ino, u32 generation)
1430 struct inode *inode;
1433 * Currently we don't know the generation for parent directory, so
1434 * a generation of 0 means "accept any"
1436 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1438 return ERR_CAST(inode);
1439 if (generation && inode->i_generation != generation) {
1441 return ERR_PTR(-ESTALE);
1447 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1448 int fh_len, int fh_type)
1450 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1451 ext4_nfs_get_inode);
1454 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1455 int fh_len, int fh_type)
1457 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1458 ext4_nfs_get_inode);
1461 static int ext4_nfs_commit_metadata(struct inode *inode)
1463 struct writeback_control wbc = {
1464 .sync_mode = WB_SYNC_ALL
1467 trace_ext4_nfs_commit_metadata(inode);
1468 return ext4_write_inode(inode, &wbc);
1472 * Try to release metadata pages (indirect blocks, directories) which are
1473 * mapped via the block device. Since these pages could have journal heads
1474 * which would prevent try_to_free_buffers() from freeing them, we must use
1475 * jbd2 layer's try_to_free_buffers() function to release them.
1477 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1480 journal_t *journal = EXT4_SB(sb)->s_journal;
1482 WARN_ON(PageChecked(page));
1483 if (!page_has_buffers(page))
1486 return jbd2_journal_try_to_free_buffers(journal, page);
1488 return try_to_free_buffers(page);
1491 #ifdef CONFIG_FS_ENCRYPTION
1492 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1494 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1495 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1498 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1501 handle_t *handle = fs_data;
1502 int res, res2, credits, retries = 0;
1505 * Encrypting the root directory is not allowed because e2fsck expects
1506 * lost+found to exist and be unencrypted, and encrypting the root
1507 * directory would imply encrypting the lost+found directory as well as
1508 * the filename "lost+found" itself.
1510 if (inode->i_ino == EXT4_ROOT_INO)
1513 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1516 if (ext4_test_inode_flag(inode, EXT4_INODE_DAX))
1519 res = ext4_convert_inline_data(inode);
1524 * If a journal handle was specified, then the encryption context is
1525 * being set on a new inode via inheritance and is part of a larger
1526 * transaction to create the inode. Otherwise the encryption context is
1527 * being set on an existing inode in its own transaction. Only in the
1528 * latter case should the "retry on ENOSPC" logic be used.
1532 res = ext4_xattr_set_handle(handle, inode,
1533 EXT4_XATTR_INDEX_ENCRYPTION,
1534 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1537 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1538 ext4_clear_inode_state(inode,
1539 EXT4_STATE_MAY_INLINE_DATA);
1541 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1542 * S_DAX may be disabled
1544 ext4_set_inode_flags(inode, false);
1549 res = dquot_initialize(inode);
1553 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1558 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1560 return PTR_ERR(handle);
1562 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1563 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1566 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1568 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1569 * S_DAX may be disabled
1571 ext4_set_inode_flags(inode, false);
1572 res = ext4_mark_inode_dirty(handle, inode);
1574 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1576 res2 = ext4_journal_stop(handle);
1578 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1585 static const union fscrypt_policy *ext4_get_dummy_policy(struct super_block *sb)
1587 return EXT4_SB(sb)->s_dummy_enc_policy.policy;
1590 static bool ext4_has_stable_inodes(struct super_block *sb)
1592 return ext4_has_feature_stable_inodes(sb);
1595 static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1596 int *ino_bits_ret, int *lblk_bits_ret)
1598 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1599 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1602 static const struct fscrypt_operations ext4_cryptops = {
1603 .key_prefix = "ext4:",
1604 .get_context = ext4_get_context,
1605 .set_context = ext4_set_context,
1606 .get_dummy_policy = ext4_get_dummy_policy,
1607 .empty_dir = ext4_empty_dir,
1608 .max_namelen = EXT4_NAME_LEN,
1609 .has_stable_inodes = ext4_has_stable_inodes,
1610 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits,
1615 static const char * const quotatypes[] = INITQFNAMES;
1616 #define QTYPE2NAME(t) (quotatypes[t])
1618 static int ext4_write_dquot(struct dquot *dquot);
1619 static int ext4_acquire_dquot(struct dquot *dquot);
1620 static int ext4_release_dquot(struct dquot *dquot);
1621 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1622 static int ext4_write_info(struct super_block *sb, int type);
1623 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1624 const struct path *path);
1625 static int ext4_quota_on_mount(struct super_block *sb, int type);
1626 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1627 size_t len, loff_t off);
1628 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1629 const char *data, size_t len, loff_t off);
1630 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1631 unsigned int flags);
1632 static int ext4_enable_quotas(struct super_block *sb);
1634 static struct dquot **ext4_get_dquots(struct inode *inode)
1636 return EXT4_I(inode)->i_dquot;
1639 static const struct dquot_operations ext4_quota_operations = {
1640 .get_reserved_space = ext4_get_reserved_space,
1641 .write_dquot = ext4_write_dquot,
1642 .acquire_dquot = ext4_acquire_dquot,
1643 .release_dquot = ext4_release_dquot,
1644 .mark_dirty = ext4_mark_dquot_dirty,
1645 .write_info = ext4_write_info,
1646 .alloc_dquot = dquot_alloc,
1647 .destroy_dquot = dquot_destroy,
1648 .get_projid = ext4_get_projid,
1649 .get_inode_usage = ext4_get_inode_usage,
1650 .get_next_id = dquot_get_next_id,
1653 static const struct quotactl_ops ext4_qctl_operations = {
1654 .quota_on = ext4_quota_on,
1655 .quota_off = ext4_quota_off,
1656 .quota_sync = dquot_quota_sync,
1657 .get_state = dquot_get_state,
1658 .set_info = dquot_set_dqinfo,
1659 .get_dqblk = dquot_get_dqblk,
1660 .set_dqblk = dquot_set_dqblk,
1661 .get_nextdqblk = dquot_get_next_dqblk,
1665 static const struct super_operations ext4_sops = {
1666 .alloc_inode = ext4_alloc_inode,
1667 .free_inode = ext4_free_in_core_inode,
1668 .destroy_inode = ext4_destroy_inode,
1669 .write_inode = ext4_write_inode,
1670 .dirty_inode = ext4_dirty_inode,
1671 .drop_inode = ext4_drop_inode,
1672 .evict_inode = ext4_evict_inode,
1673 .put_super = ext4_put_super,
1674 .sync_fs = ext4_sync_fs,
1675 .freeze_fs = ext4_freeze,
1676 .unfreeze_fs = ext4_unfreeze,
1677 .statfs = ext4_statfs,
1678 .remount_fs = ext4_remount,
1679 .show_options = ext4_show_options,
1681 .quota_read = ext4_quota_read,
1682 .quota_write = ext4_quota_write,
1683 .get_dquots = ext4_get_dquots,
1685 .bdev_try_to_free_page = bdev_try_to_free_page,
1688 static const struct export_operations ext4_export_ops = {
1689 .fh_to_dentry = ext4_fh_to_dentry,
1690 .fh_to_parent = ext4_fh_to_parent,
1691 .get_parent = ext4_get_parent,
1692 .commit_metadata = ext4_nfs_commit_metadata,
1696 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1697 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1698 Opt_nouid32, Opt_debug, Opt_removed,
1699 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1700 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1701 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1702 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1703 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1704 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1706 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1707 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1708 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1709 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version,
1710 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1711 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1712 Opt_nowarn_on_error, Opt_mblk_io_submit,
1713 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1714 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1715 Opt_inode_readahead_blks, Opt_journal_ioprio,
1716 Opt_dioread_nolock, Opt_dioread_lock,
1717 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1718 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1719 Opt_prefetch_block_bitmaps, Opt_no_fc,
1720 #ifdef CONFIG_EXT4_DEBUG
1721 Opt_fc_debug_max_replay,
1726 static const match_table_t tokens = {
1727 {Opt_bsd_df, "bsddf"},
1728 {Opt_minix_df, "minixdf"},
1729 {Opt_grpid, "grpid"},
1730 {Opt_grpid, "bsdgroups"},
1731 {Opt_nogrpid, "nogrpid"},
1732 {Opt_nogrpid, "sysvgroups"},
1733 {Opt_resgid, "resgid=%u"},
1734 {Opt_resuid, "resuid=%u"},
1736 {Opt_err_cont, "errors=continue"},
1737 {Opt_err_panic, "errors=panic"},
1738 {Opt_err_ro, "errors=remount-ro"},
1739 {Opt_nouid32, "nouid32"},
1740 {Opt_debug, "debug"},
1741 {Opt_removed, "oldalloc"},
1742 {Opt_removed, "orlov"},
1743 {Opt_user_xattr, "user_xattr"},
1744 {Opt_nouser_xattr, "nouser_xattr"},
1746 {Opt_noacl, "noacl"},
1747 {Opt_noload, "norecovery"},
1748 {Opt_noload, "noload"},
1749 {Opt_removed, "nobh"},
1750 {Opt_removed, "bh"},
1751 {Opt_commit, "commit=%u"},
1752 {Opt_min_batch_time, "min_batch_time=%u"},
1753 {Opt_max_batch_time, "max_batch_time=%u"},
1754 {Opt_journal_dev, "journal_dev=%u"},
1755 {Opt_journal_path, "journal_path=%s"},
1756 {Opt_journal_checksum, "journal_checksum"},
1757 {Opt_nojournal_checksum, "nojournal_checksum"},
1758 {Opt_journal_async_commit, "journal_async_commit"},
1759 {Opt_abort, "abort"},
1760 {Opt_data_journal, "data=journal"},
1761 {Opt_data_ordered, "data=ordered"},
1762 {Opt_data_writeback, "data=writeback"},
1763 {Opt_data_err_abort, "data_err=abort"},
1764 {Opt_data_err_ignore, "data_err=ignore"},
1765 {Opt_offusrjquota, "usrjquota="},
1766 {Opt_usrjquota, "usrjquota=%s"},
1767 {Opt_offgrpjquota, "grpjquota="},
1768 {Opt_grpjquota, "grpjquota=%s"},
1769 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1770 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1771 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1772 {Opt_grpquota, "grpquota"},
1773 {Opt_noquota, "noquota"},
1774 {Opt_quota, "quota"},
1775 {Opt_usrquota, "usrquota"},
1776 {Opt_prjquota, "prjquota"},
1777 {Opt_barrier, "barrier=%u"},
1778 {Opt_barrier, "barrier"},
1779 {Opt_nobarrier, "nobarrier"},
1780 {Opt_i_version, "i_version"},
1782 {Opt_dax_always, "dax=always"},
1783 {Opt_dax_inode, "dax=inode"},
1784 {Opt_dax_never, "dax=never"},
1785 {Opt_stripe, "stripe=%u"},
1786 {Opt_delalloc, "delalloc"},
1787 {Opt_warn_on_error, "warn_on_error"},
1788 {Opt_nowarn_on_error, "nowarn_on_error"},
1789 {Opt_lazytime, "lazytime"},
1790 {Opt_nolazytime, "nolazytime"},
1791 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1792 {Opt_nodelalloc, "nodelalloc"},
1793 {Opt_removed, "mblk_io_submit"},
1794 {Opt_removed, "nomblk_io_submit"},
1795 {Opt_block_validity, "block_validity"},
1796 {Opt_noblock_validity, "noblock_validity"},
1797 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1798 {Opt_journal_ioprio, "journal_ioprio=%u"},
1799 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1800 {Opt_auto_da_alloc, "auto_da_alloc"},
1801 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1802 {Opt_dioread_nolock, "dioread_nolock"},
1803 {Opt_dioread_lock, "nodioread_nolock"},
1804 {Opt_dioread_lock, "dioread_lock"},
1805 {Opt_discard, "discard"},
1806 {Opt_nodiscard, "nodiscard"},
1807 {Opt_init_itable, "init_itable=%u"},
1808 {Opt_init_itable, "init_itable"},
1809 {Opt_noinit_itable, "noinit_itable"},
1810 {Opt_no_fc, "no_fc"},
1811 {Opt_fc_debug_force, "fc_debug_force"},
1812 #ifdef CONFIG_EXT4_DEBUG
1813 {Opt_fc_debug_max_replay, "fc_debug_max_replay=%u"},
1815 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1816 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
1817 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1818 {Opt_inlinecrypt, "inlinecrypt"},
1819 {Opt_nombcache, "nombcache"},
1820 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1821 {Opt_prefetch_block_bitmaps, "prefetch_block_bitmaps"},
1822 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1823 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1824 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1825 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1826 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1830 static ext4_fsblk_t get_sb_block(void **data)
1832 ext4_fsblk_t sb_block;
1833 char *options = (char *) *data;
1835 if (!options || strncmp(options, "sb=", 3) != 0)
1836 return 1; /* Default location */
1839 /* TODO: use simple_strtoll with >32bit ext4 */
1840 sb_block = simple_strtoul(options, &options, 0);
1841 if (*options && *options != ',') {
1842 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1846 if (*options == ',')
1848 *data = (void *) options;
1853 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1854 static const char deprecated_msg[] =
1855 "Mount option \"%s\" will be removed by %s\n"
1856 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1859 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1861 struct ext4_sb_info *sbi = EXT4_SB(sb);
1862 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
1865 if (sb_any_quota_loaded(sb) && !old_qname) {
1866 ext4_msg(sb, KERN_ERR,
1867 "Cannot change journaled "
1868 "quota options when quota turned on");
1871 if (ext4_has_feature_quota(sb)) {
1872 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1873 "ignored when QUOTA feature is enabled");
1876 qname = match_strdup(args);
1878 ext4_msg(sb, KERN_ERR,
1879 "Not enough memory for storing quotafile name");
1883 if (strcmp(old_qname, qname) == 0)
1886 ext4_msg(sb, KERN_ERR,
1887 "%s quota file already specified",
1891 if (strchr(qname, '/')) {
1892 ext4_msg(sb, KERN_ERR,
1893 "quotafile must be on filesystem root");
1896 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
1904 static int clear_qf_name(struct super_block *sb, int qtype)
1907 struct ext4_sb_info *sbi = EXT4_SB(sb);
1908 char *old_qname = get_qf_name(sb, sbi, qtype);
1910 if (sb_any_quota_loaded(sb) && old_qname) {
1911 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1912 " when quota turned on");
1915 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1922 #define MOPT_SET 0x0001
1923 #define MOPT_CLEAR 0x0002
1924 #define MOPT_NOSUPPORT 0x0004
1925 #define MOPT_EXPLICIT 0x0008
1926 #define MOPT_CLEAR_ERR 0x0010
1927 #define MOPT_GTE0 0x0020
1930 #define MOPT_QFMT 0x0040
1932 #define MOPT_Q MOPT_NOSUPPORT
1933 #define MOPT_QFMT MOPT_NOSUPPORT
1935 #define MOPT_DATAJ 0x0080
1936 #define MOPT_NO_EXT2 0x0100
1937 #define MOPT_NO_EXT3 0x0200
1938 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1939 #define MOPT_STRING 0x0400
1940 #define MOPT_SKIP 0x0800
1941 #define MOPT_2 0x1000
1943 static const struct mount_opts {
1947 } ext4_mount_opts[] = {
1948 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1949 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1950 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1951 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1952 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1953 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1954 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1955 MOPT_EXT4_ONLY | MOPT_SET},
1956 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1957 MOPT_EXT4_ONLY | MOPT_CLEAR},
1958 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1959 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1960 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1961 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1962 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1963 MOPT_EXT4_ONLY | MOPT_CLEAR},
1964 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1965 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1966 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1967 MOPT_EXT4_ONLY | MOPT_CLEAR},
1968 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1969 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1970 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1971 EXT4_MOUNT_JOURNAL_CHECKSUM),
1972 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1973 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1974 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1975 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1976 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1977 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1979 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1981 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1982 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1983 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1984 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1985 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1986 {Opt_commit, 0, MOPT_GTE0},
1987 {Opt_max_batch_time, 0, MOPT_GTE0},
1988 {Opt_min_batch_time, 0, MOPT_GTE0},
1989 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1990 {Opt_init_itable, 0, MOPT_GTE0},
1991 {Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP},
1992 {Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS,
1993 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1994 {Opt_dax_inode, EXT4_MOUNT2_DAX_INODE,
1995 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1996 {Opt_dax_never, EXT4_MOUNT2_DAX_NEVER,
1997 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1998 {Opt_stripe, 0, MOPT_GTE0},
1999 {Opt_resuid, 0, MOPT_GTE0},
2000 {Opt_resgid, 0, MOPT_GTE0},
2001 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
2002 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
2003 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
2004 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
2005 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
2006 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
2007 MOPT_NO_EXT2 | MOPT_DATAJ},
2008 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
2009 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
2010 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2011 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
2012 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
2014 {Opt_acl, 0, MOPT_NOSUPPORT},
2015 {Opt_noacl, 0, MOPT_NOSUPPORT},
2017 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
2018 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
2019 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
2020 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
2021 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
2023 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
2025 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
2027 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
2028 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
2029 MOPT_CLEAR | MOPT_Q},
2030 {Opt_usrjquota, 0, MOPT_Q},
2031 {Opt_grpjquota, 0, MOPT_Q},
2032 {Opt_offusrjquota, 0, MOPT_Q},
2033 {Opt_offgrpjquota, 0, MOPT_Q},
2034 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
2035 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
2036 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
2037 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
2038 {Opt_test_dummy_encryption, 0, MOPT_STRING},
2039 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
2040 {Opt_prefetch_block_bitmaps, EXT4_MOUNT_PREFETCH_BLOCK_BITMAPS,
2042 {Opt_no_fc, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
2043 MOPT_CLEAR | MOPT_2 | MOPT_EXT4_ONLY},
2044 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
2045 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
2046 #ifdef CONFIG_EXT4_DEBUG
2047 {Opt_fc_debug_max_replay, 0, MOPT_GTE0},
2052 #ifdef CONFIG_UNICODE
2053 static const struct ext4_sb_encodings {
2057 } ext4_sb_encoding_map[] = {
2058 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
2061 static int ext4_sb_read_encoding(const struct ext4_super_block *es,
2062 const struct ext4_sb_encodings **encoding,
2065 __u16 magic = le16_to_cpu(es->s_encoding);
2068 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
2069 if (magic == ext4_sb_encoding_map[i].magic)
2072 if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
2075 *encoding = &ext4_sb_encoding_map[i];
2076 *flags = le16_to_cpu(es->s_encoding_flags);
2082 static int ext4_set_test_dummy_encryption(struct super_block *sb,
2084 const substring_t *arg,
2087 #ifdef CONFIG_FS_ENCRYPTION
2088 struct ext4_sb_info *sbi = EXT4_SB(sb);
2092 * This mount option is just for testing, and it's not worthwhile to
2093 * implement the extra complexity (e.g. RCU protection) that would be
2094 * needed to allow it to be set or changed during remount. We do allow
2095 * it to be specified during remount, but only if there is no change.
2097 if (is_remount && !sbi->s_dummy_enc_policy.policy) {
2098 ext4_msg(sb, KERN_WARNING,
2099 "Can't set test_dummy_encryption on remount");
2102 err = fscrypt_set_test_dummy_encryption(sb, arg->from,
2103 &sbi->s_dummy_enc_policy);
2106 ext4_msg(sb, KERN_WARNING,
2107 "Can't change test_dummy_encryption on remount");
2108 else if (err == -EINVAL)
2109 ext4_msg(sb, KERN_WARNING,
2110 "Value of option \"%s\" is unrecognized", opt);
2112 ext4_msg(sb, KERN_WARNING,
2113 "Error processing option \"%s\" [%d]",
2117 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2119 ext4_msg(sb, KERN_WARNING,
2120 "Test dummy encryption mount option ignored");
2125 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
2126 substring_t *args, unsigned long *journal_devnum,
2127 unsigned int *journal_ioprio, int is_remount)
2129 struct ext4_sb_info *sbi = EXT4_SB(sb);
2130 const struct mount_opts *m;
2136 if (token == Opt_usrjquota)
2137 return set_qf_name(sb, USRQUOTA, &args[0]);
2138 else if (token == Opt_grpjquota)
2139 return set_qf_name(sb, GRPQUOTA, &args[0]);
2140 else if (token == Opt_offusrjquota)
2141 return clear_qf_name(sb, USRQUOTA);
2142 else if (token == Opt_offgrpjquota)
2143 return clear_qf_name(sb, GRPQUOTA);
2147 case Opt_nouser_xattr:
2148 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
2151 return 1; /* handled by get_sb_block() */
2153 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
2156 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
2159 sb->s_flags |= SB_I_VERSION;
2162 sb->s_flags |= SB_LAZYTIME;
2164 case Opt_nolazytime:
2165 sb->s_flags &= ~SB_LAZYTIME;
2167 case Opt_inlinecrypt:
2168 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2169 sb->s_flags |= SB_INLINECRYPT;
2171 ext4_msg(sb, KERN_ERR, "inline encryption not supported");
2176 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2177 if (token == m->token)
2180 if (m->token == Opt_err) {
2181 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
2182 "or missing value", opt);
2186 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2187 ext4_msg(sb, KERN_ERR,
2188 "Mount option \"%s\" incompatible with ext2", opt);
2191 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2192 ext4_msg(sb, KERN_ERR,
2193 "Mount option \"%s\" incompatible with ext3", opt);
2197 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
2199 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
2201 if (m->flags & MOPT_EXPLICIT) {
2202 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2203 set_opt2(sb, EXPLICIT_DELALLOC);
2204 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2205 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
2209 if (m->flags & MOPT_CLEAR_ERR)
2210 clear_opt(sb, ERRORS_MASK);
2211 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
2212 ext4_msg(sb, KERN_ERR, "Cannot change quota "
2213 "options when quota turned on");
2217 if (m->flags & MOPT_NOSUPPORT) {
2218 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
2219 } else if (token == Opt_commit) {
2221 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
2222 else if (arg > INT_MAX / HZ) {
2223 ext4_msg(sb, KERN_ERR,
2224 "Invalid commit interval %d, "
2225 "must be smaller than %d",
2229 sbi->s_commit_interval = HZ * arg;
2230 } else if (token == Opt_debug_want_extra_isize) {
2233 (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
2234 ext4_msg(sb, KERN_ERR,
2235 "Invalid want_extra_isize %d", arg);
2238 sbi->s_want_extra_isize = arg;
2239 } else if (token == Opt_max_batch_time) {
2240 sbi->s_max_batch_time = arg;
2241 } else if (token == Opt_min_batch_time) {
2242 sbi->s_min_batch_time = arg;
2243 } else if (token == Opt_inode_readahead_blks) {
2244 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
2245 ext4_msg(sb, KERN_ERR,
2246 "EXT4-fs: inode_readahead_blks must be "
2247 "0 or a power of 2 smaller than 2^31");
2250 sbi->s_inode_readahead_blks = arg;
2251 } else if (token == Opt_init_itable) {
2252 set_opt(sb, INIT_INODE_TABLE);
2254 arg = EXT4_DEF_LI_WAIT_MULT;
2255 sbi->s_li_wait_mult = arg;
2256 } else if (token == Opt_max_dir_size_kb) {
2257 sbi->s_max_dir_size_kb = arg;
2258 #ifdef CONFIG_EXT4_DEBUG
2259 } else if (token == Opt_fc_debug_max_replay) {
2260 sbi->s_fc_debug_max_replay = arg;
2262 } else if (token == Opt_stripe) {
2263 sbi->s_stripe = arg;
2264 } else if (token == Opt_resuid) {
2265 uid = make_kuid(current_user_ns(), arg);
2266 if (!uid_valid(uid)) {
2267 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
2270 sbi->s_resuid = uid;
2271 } else if (token == Opt_resgid) {
2272 gid = make_kgid(current_user_ns(), arg);
2273 if (!gid_valid(gid)) {
2274 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
2277 sbi->s_resgid = gid;
2278 } else if (token == Opt_journal_dev) {
2280 ext4_msg(sb, KERN_ERR,
2281 "Cannot specify journal on remount");
2284 *journal_devnum = arg;
2285 } else if (token == Opt_journal_path) {
2287 struct inode *journal_inode;
2292 ext4_msg(sb, KERN_ERR,
2293 "Cannot specify journal on remount");
2296 journal_path = match_strdup(&args[0]);
2297 if (!journal_path) {
2298 ext4_msg(sb, KERN_ERR, "error: could not dup "
2299 "journal device string");
2303 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
2305 ext4_msg(sb, KERN_ERR, "error: could not find "
2306 "journal device path: error %d", error);
2307 kfree(journal_path);
2311 journal_inode = d_inode(path.dentry);
2312 if (!S_ISBLK(journal_inode->i_mode)) {
2313 ext4_msg(sb, KERN_ERR, "error: journal path %s "
2314 "is not a block device", journal_path);
2316 kfree(journal_path);
2320 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
2322 kfree(journal_path);
2323 } else if (token == Opt_journal_ioprio) {
2325 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
2330 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
2331 } else if (token == Opt_test_dummy_encryption) {
2332 return ext4_set_test_dummy_encryption(sb, opt, &args[0],
2334 } else if (m->flags & MOPT_DATAJ) {
2336 if (!sbi->s_journal)
2337 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
2338 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
2339 ext4_msg(sb, KERN_ERR,
2340 "Cannot change data mode on remount");
2344 clear_opt(sb, DATA_FLAGS);
2345 sbi->s_mount_opt |= m->mount_opt;
2348 } else if (m->flags & MOPT_QFMT) {
2349 if (sb_any_quota_loaded(sb) &&
2350 sbi->s_jquota_fmt != m->mount_opt) {
2351 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
2352 "quota options when quota turned on");
2355 if (ext4_has_feature_quota(sb)) {
2356 ext4_msg(sb, KERN_INFO,
2357 "Quota format mount options ignored "
2358 "when QUOTA feature is enabled");
2361 sbi->s_jquota_fmt = m->mount_opt;
2363 } else if (token == Opt_dax || token == Opt_dax_always ||
2364 token == Opt_dax_inode || token == Opt_dax_never) {
2365 #ifdef CONFIG_FS_DAX
2368 case Opt_dax_always:
2370 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2371 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2372 fail_dax_change_remount:
2373 ext4_msg(sb, KERN_ERR, "can't change "
2374 "dax mount option while remounting");
2378 (test_opt(sb, DATA_FLAGS) ==
2379 EXT4_MOUNT_JOURNAL_DATA)) {
2380 ext4_msg(sb, KERN_ERR, "can't mount with "
2381 "both data=journal and dax");
2384 ext4_msg(sb, KERN_WARNING,
2385 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2386 sbi->s_mount_opt |= EXT4_MOUNT_DAX_ALWAYS;
2387 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2391 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2392 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS)))
2393 goto fail_dax_change_remount;
2394 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2395 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2399 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2400 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2401 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE)))
2402 goto fail_dax_change_remount;
2403 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2404 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2405 /* Strictly for printing options */
2406 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_INODE;
2410 ext4_msg(sb, KERN_INFO, "dax option not supported");
2411 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2412 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2415 } else if (token == Opt_data_err_abort) {
2416 sbi->s_mount_opt |= m->mount_opt;
2417 } else if (token == Opt_data_err_ignore) {
2418 sbi->s_mount_opt &= ~m->mount_opt;
2422 if (m->flags & MOPT_CLEAR)
2424 else if (unlikely(!(m->flags & MOPT_SET))) {
2425 ext4_msg(sb, KERN_WARNING,
2426 "buggy handling of option %s", opt);
2430 if (m->flags & MOPT_2) {
2432 sbi->s_mount_opt2 |= m->mount_opt;
2434 sbi->s_mount_opt2 &= ~m->mount_opt;
2437 sbi->s_mount_opt |= m->mount_opt;
2439 sbi->s_mount_opt &= ~m->mount_opt;
2445 static int parse_options(char *options, struct super_block *sb,
2446 unsigned long *journal_devnum,
2447 unsigned int *journal_ioprio,
2450 struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb);
2451 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
2452 substring_t args[MAX_OPT_ARGS];
2458 while ((p = strsep(&options, ",")) != NULL) {
2462 * Initialize args struct so we know whether arg was
2463 * found; some options take optional arguments.
2465 args[0].to = args[0].from = NULL;
2466 token = match_token(p, tokens, args);
2467 if (handle_mount_opt(sb, p, token, args, journal_devnum,
2468 journal_ioprio, is_remount) < 0)
2473 * We do the test below only for project quotas. 'usrquota' and
2474 * 'grpquota' mount options are allowed even without quota feature
2475 * to support legacy quotas in quota files.
2477 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2478 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2479 "Cannot enable project quota enforcement.");
2482 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2483 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2484 if (usr_qf_name || grp_qf_name) {
2485 if (test_opt(sb, USRQUOTA) && usr_qf_name)
2486 clear_opt(sb, USRQUOTA);
2488 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
2489 clear_opt(sb, GRPQUOTA);
2491 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
2492 ext4_msg(sb, KERN_ERR, "old and new quota "
2497 if (!sbi->s_jquota_fmt) {
2498 ext4_msg(sb, KERN_ERR, "journaled quota format "
2504 if (test_opt(sb, DIOREAD_NOLOCK)) {
2506 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2507 if (blocksize < PAGE_SIZE)
2508 ext4_msg(sb, KERN_WARNING, "Warning: mounting with an "
2509 "experimental mount option 'dioread_nolock' "
2510 "for blocksize < PAGE_SIZE");
2515 static inline void ext4_show_quota_options(struct seq_file *seq,
2516 struct super_block *sb)
2518 #if defined(CONFIG_QUOTA)
2519 struct ext4_sb_info *sbi = EXT4_SB(sb);
2520 char *usr_qf_name, *grp_qf_name;
2522 if (sbi->s_jquota_fmt) {
2525 switch (sbi->s_jquota_fmt) {
2536 seq_printf(seq, ",jqfmt=%s", fmtname);
2540 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2541 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2543 seq_show_option(seq, "usrjquota", usr_qf_name);
2545 seq_show_option(seq, "grpjquota", grp_qf_name);
2550 static const char *token2str(int token)
2552 const struct match_token *t;
2554 for (t = tokens; t->token != Opt_err; t++)
2555 if (t->token == token && !strchr(t->pattern, '='))
2562 * - it's set to a non-default value OR
2563 * - if the per-sb default is different from the global default
2565 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2568 struct ext4_sb_info *sbi = EXT4_SB(sb);
2569 struct ext4_super_block *es = sbi->s_es;
2570 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2571 const struct mount_opts *m;
2572 char sep = nodefs ? '\n' : ',';
2574 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2575 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2577 if (sbi->s_sb_block != 1)
2578 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2580 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2581 int want_set = m->flags & MOPT_SET;
2582 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2583 (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP)
2585 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2586 continue; /* skip if same as the default */
2588 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2589 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2590 continue; /* select Opt_noFoo vs Opt_Foo */
2591 SEQ_OPTS_PRINT("%s", token2str(m->token));
2594 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2595 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2596 SEQ_OPTS_PRINT("resuid=%u",
2597 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2598 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2599 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2600 SEQ_OPTS_PRINT("resgid=%u",
2601 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2602 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2603 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2604 SEQ_OPTS_PUTS("errors=remount-ro");
2605 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2606 SEQ_OPTS_PUTS("errors=continue");
2607 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2608 SEQ_OPTS_PUTS("errors=panic");
2609 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2610 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2611 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2612 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2613 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2614 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2615 if (sb->s_flags & SB_I_VERSION)
2616 SEQ_OPTS_PUTS("i_version");
2617 if (nodefs || sbi->s_stripe)
2618 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2619 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2620 (sbi->s_mount_opt ^ def_mount_opt)) {
2621 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2622 SEQ_OPTS_PUTS("data=journal");
2623 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2624 SEQ_OPTS_PUTS("data=ordered");
2625 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2626 SEQ_OPTS_PUTS("data=writeback");
2629 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2630 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2631 sbi->s_inode_readahead_blks);
2633 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2634 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2635 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2636 if (nodefs || sbi->s_max_dir_size_kb)
2637 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2638 if (test_opt(sb, DATA_ERR_ABORT))
2639 SEQ_OPTS_PUTS("data_err=abort");
2641 fscrypt_show_test_dummy_encryption(seq, sep, sb);
2643 if (sb->s_flags & SB_INLINECRYPT)
2644 SEQ_OPTS_PUTS("inlinecrypt");
2646 if (test_opt(sb, DAX_ALWAYS)) {
2648 SEQ_OPTS_PUTS("dax");
2650 SEQ_OPTS_PUTS("dax=always");
2651 } else if (test_opt2(sb, DAX_NEVER)) {
2652 SEQ_OPTS_PUTS("dax=never");
2653 } else if (test_opt2(sb, DAX_INODE)) {
2654 SEQ_OPTS_PUTS("dax=inode");
2657 if (test_opt2(sb, JOURNAL_FAST_COMMIT))
2658 SEQ_OPTS_PUTS("fast_commit");
2660 ext4_show_quota_options(seq, sb);
2664 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2666 return _ext4_show_options(seq, root->d_sb, 0);
2669 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2671 struct super_block *sb = seq->private;
2674 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2675 rc = _ext4_show_options(seq, sb, 1);
2676 seq_puts(seq, "\n");
2680 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2683 struct ext4_sb_info *sbi = EXT4_SB(sb);
2686 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2687 ext4_msg(sb, KERN_ERR, "revision level too high, "
2688 "forcing read-only mode");
2694 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2695 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2696 "running e2fsck is recommended");
2697 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2698 ext4_msg(sb, KERN_WARNING,
2699 "warning: mounting fs with errors, "
2700 "running e2fsck is recommended");
2701 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2702 le16_to_cpu(es->s_mnt_count) >=
2703 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2704 ext4_msg(sb, KERN_WARNING,
2705 "warning: maximal mount count reached, "
2706 "running e2fsck is recommended");
2707 else if (le32_to_cpu(es->s_checkinterval) &&
2708 (ext4_get_tstamp(es, s_lastcheck) +
2709 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
2710 ext4_msg(sb, KERN_WARNING,
2711 "warning: checktime reached, "
2712 "running e2fsck is recommended");
2713 if (!sbi->s_journal)
2714 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2715 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2716 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2717 le16_add_cpu(&es->s_mnt_count, 1);
2718 ext4_update_tstamp(es, s_mtime);
2720 ext4_set_feature_journal_needs_recovery(sb);
2722 err = ext4_commit_super(sb, 1);
2724 if (test_opt(sb, DEBUG))
2725 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2726 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2728 sbi->s_groups_count,
2729 EXT4_BLOCKS_PER_GROUP(sb),
2730 EXT4_INODES_PER_GROUP(sb),
2731 sbi->s_mount_opt, sbi->s_mount_opt2);
2733 cleancache_init_fs(sb);
2737 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2739 struct ext4_sb_info *sbi = EXT4_SB(sb);
2740 struct flex_groups **old_groups, **new_groups;
2743 if (!sbi->s_log_groups_per_flex)
2746 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2747 if (size <= sbi->s_flex_groups_allocated)
2750 new_groups = kvzalloc(roundup_pow_of_two(size *
2751 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
2753 ext4_msg(sb, KERN_ERR,
2754 "not enough memory for %d flex group pointers", size);
2757 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2758 new_groups[i] = kvzalloc(roundup_pow_of_two(
2759 sizeof(struct flex_groups)),
2761 if (!new_groups[i]) {
2762 for (j = sbi->s_flex_groups_allocated; j < i; j++)
2763 kvfree(new_groups[j]);
2765 ext4_msg(sb, KERN_ERR,
2766 "not enough memory for %d flex groups", size);
2771 old_groups = rcu_dereference(sbi->s_flex_groups);
2773 memcpy(new_groups, old_groups,
2774 (sbi->s_flex_groups_allocated *
2775 sizeof(struct flex_groups *)));
2777 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2778 sbi->s_flex_groups_allocated = size;
2780 ext4_kvfree_array_rcu(old_groups);
2784 static int ext4_fill_flex_info(struct super_block *sb)
2786 struct ext4_sb_info *sbi = EXT4_SB(sb);
2787 struct ext4_group_desc *gdp = NULL;
2788 struct flex_groups *fg;
2789 ext4_group_t flex_group;
2792 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2793 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2794 sbi->s_log_groups_per_flex = 0;
2798 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2802 for (i = 0; i < sbi->s_groups_count; i++) {
2803 gdp = ext4_get_group_desc(sb, i, NULL);
2805 flex_group = ext4_flex_group(sbi, i);
2806 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2807 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
2808 atomic64_add(ext4_free_group_clusters(sb, gdp),
2809 &fg->free_clusters);
2810 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
2818 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2819 struct ext4_group_desc *gdp)
2821 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2823 __le32 le_group = cpu_to_le32(block_group);
2824 struct ext4_sb_info *sbi = EXT4_SB(sb);
2826 if (ext4_has_metadata_csum(sbi->s_sb)) {
2827 /* Use new metadata_csum algorithm */
2829 __u16 dummy_csum = 0;
2831 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2833 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2834 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2835 sizeof(dummy_csum));
2836 offset += sizeof(dummy_csum);
2837 if (offset < sbi->s_desc_size)
2838 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2839 sbi->s_desc_size - offset);
2841 crc = csum32 & 0xFFFF;
2845 /* old crc16 code */
2846 if (!ext4_has_feature_gdt_csum(sb))
2849 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2850 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2851 crc = crc16(crc, (__u8 *)gdp, offset);
2852 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2853 /* for checksum of struct ext4_group_desc do the rest...*/
2854 if (ext4_has_feature_64bit(sb) &&
2855 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2856 crc = crc16(crc, (__u8 *)gdp + offset,
2857 le16_to_cpu(sbi->s_es->s_desc_size) -
2861 return cpu_to_le16(crc);
2864 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2865 struct ext4_group_desc *gdp)
2867 if (ext4_has_group_desc_csum(sb) &&
2868 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2874 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2875 struct ext4_group_desc *gdp)
2877 if (!ext4_has_group_desc_csum(sb))
2879 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2882 /* Called at mount-time, super-block is locked */
2883 static int ext4_check_descriptors(struct super_block *sb,
2884 ext4_fsblk_t sb_block,
2885 ext4_group_t *first_not_zeroed)
2887 struct ext4_sb_info *sbi = EXT4_SB(sb);
2888 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2889 ext4_fsblk_t last_block;
2890 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2891 ext4_fsblk_t block_bitmap;
2892 ext4_fsblk_t inode_bitmap;
2893 ext4_fsblk_t inode_table;
2894 int flexbg_flag = 0;
2895 ext4_group_t i, grp = sbi->s_groups_count;
2897 if (ext4_has_feature_flex_bg(sb))
2900 ext4_debug("Checking group descriptors");
2902 for (i = 0; i < sbi->s_groups_count; i++) {
2903 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2905 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2906 last_block = ext4_blocks_count(sbi->s_es) - 1;
2908 last_block = first_block +
2909 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2911 if ((grp == sbi->s_groups_count) &&
2912 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2915 block_bitmap = ext4_block_bitmap(sb, gdp);
2916 if (block_bitmap == sb_block) {
2917 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2918 "Block bitmap for group %u overlaps "
2923 if (block_bitmap >= sb_block + 1 &&
2924 block_bitmap <= last_bg_block) {
2925 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2926 "Block bitmap for group %u overlaps "
2927 "block group descriptors", i);
2931 if (block_bitmap < first_block || block_bitmap > last_block) {
2932 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2933 "Block bitmap for group %u not in group "
2934 "(block %llu)!", i, block_bitmap);
2937 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2938 if (inode_bitmap == sb_block) {
2939 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2940 "Inode bitmap for group %u overlaps "
2945 if (inode_bitmap >= sb_block + 1 &&
2946 inode_bitmap <= last_bg_block) {
2947 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2948 "Inode bitmap for group %u overlaps "
2949 "block group descriptors", i);
2953 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2954 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2955 "Inode bitmap for group %u not in group "
2956 "(block %llu)!", i, inode_bitmap);
2959 inode_table = ext4_inode_table(sb, gdp);
2960 if (inode_table == sb_block) {
2961 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2962 "Inode table for group %u overlaps "
2967 if (inode_table >= sb_block + 1 &&
2968 inode_table <= last_bg_block) {
2969 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2970 "Inode table for group %u overlaps "
2971 "block group descriptors", i);
2975 if (inode_table < first_block ||
2976 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2977 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2978 "Inode table for group %u not in group "
2979 "(block %llu)!", i, inode_table);
2982 ext4_lock_group(sb, i);
2983 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2984 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2985 "Checksum for group %u failed (%u!=%u)",
2986 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2987 gdp)), le16_to_cpu(gdp->bg_checksum));
2988 if (!sb_rdonly(sb)) {
2989 ext4_unlock_group(sb, i);
2993 ext4_unlock_group(sb, i);
2995 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2997 if (NULL != first_not_zeroed)
2998 *first_not_zeroed = grp;
3002 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
3003 * the superblock) which were deleted from all directories, but held open by
3004 * a process at the time of a crash. We walk the list and try to delete these
3005 * inodes at recovery time (only with a read-write filesystem).
3007 * In order to keep the orphan inode chain consistent during traversal (in
3008 * case of crash during recovery), we link each inode into the superblock
3009 * orphan list_head and handle it the same way as an inode deletion during
3010 * normal operation (which journals the operations for us).
3012 * We only do an iget() and an iput() on each inode, which is very safe if we
3013 * accidentally point at an in-use or already deleted inode. The worst that
3014 * can happen in this case is that we get a "bit already cleared" message from
3015 * ext4_free_inode(). The only reason we would point at a wrong inode is if
3016 * e2fsck was run on this filesystem, and it must have already done the orphan
3017 * inode cleanup for us, so we can safely abort without any further action.
3019 static void ext4_orphan_cleanup(struct super_block *sb,
3020 struct ext4_super_block *es)
3022 unsigned int s_flags = sb->s_flags;
3023 int ret, nr_orphans = 0, nr_truncates = 0;
3025 int quota_update = 0;
3028 if (!es->s_last_orphan) {
3029 jbd_debug(4, "no orphan inodes to clean up\n");
3033 if (bdev_read_only(sb->s_bdev)) {
3034 ext4_msg(sb, KERN_ERR, "write access "
3035 "unavailable, skipping orphan cleanup");
3039 /* Check if feature set would not allow a r/w mount */
3040 if (!ext4_feature_set_ok(sb, 0)) {
3041 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
3042 "unknown ROCOMPAT features");
3046 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
3047 /* don't clear list on RO mount w/ errors */
3048 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
3049 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
3050 "clearing orphan list.\n");
3051 es->s_last_orphan = 0;
3053 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
3057 if (s_flags & SB_RDONLY) {
3058 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
3059 sb->s_flags &= ~SB_RDONLY;
3062 /* Needed for iput() to work correctly and not trash data */
3063 sb->s_flags |= SB_ACTIVE;
3066 * Turn on quotas which were not enabled for read-only mounts if
3067 * filesystem has quota feature, so that they are updated correctly.
3069 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
3070 int ret = ext4_enable_quotas(sb);
3075 ext4_msg(sb, KERN_ERR,
3076 "Cannot turn on quotas: error %d", ret);
3079 /* Turn on journaled quotas used for old sytle */
3080 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
3081 if (EXT4_SB(sb)->s_qf_names[i]) {
3082 int ret = ext4_quota_on_mount(sb, i);
3087 ext4_msg(sb, KERN_ERR,
3088 "Cannot turn on journaled "
3089 "quota: type %d: error %d", i, ret);
3094 while (es->s_last_orphan) {
3095 struct inode *inode;
3098 * We may have encountered an error during cleanup; if
3099 * so, skip the rest.
3101 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
3102 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
3103 es->s_last_orphan = 0;
3107 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
3108 if (IS_ERR(inode)) {
3109 es->s_last_orphan = 0;
3113 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
3114 dquot_initialize(inode);
3115 if (inode->i_nlink) {
3116 if (test_opt(sb, DEBUG))
3117 ext4_msg(sb, KERN_DEBUG,
3118 "%s: truncating inode %lu to %lld bytes",
3119 __func__, inode->i_ino, inode->i_size);
3120 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
3121 inode->i_ino, inode->i_size);
3123 truncate_inode_pages(inode->i_mapping, inode->i_size);
3124 ret = ext4_truncate(inode);
3126 ext4_std_error(inode->i_sb, ret);
3127 inode_unlock(inode);
3130 if (test_opt(sb, DEBUG))
3131 ext4_msg(sb, KERN_DEBUG,
3132 "%s: deleting unreferenced inode %lu",
3133 __func__, inode->i_ino);
3134 jbd_debug(2, "deleting unreferenced inode %lu\n",
3138 iput(inode); /* The delete magic happens here! */
3141 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
3144 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
3145 PLURAL(nr_orphans));
3147 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
3148 PLURAL(nr_truncates));
3150 /* Turn off quotas if they were enabled for orphan cleanup */
3152 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
3153 if (sb_dqopt(sb)->files[i])
3154 dquot_quota_off(sb, i);
3158 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
3162 * Maximal extent format file size.
3163 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3164 * extent format containers, within a sector_t, and within i_blocks
3165 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3166 * so that won't be a limiting factor.
3168 * However there is other limiting factor. We do store extents in the form
3169 * of starting block and length, hence the resulting length of the extent
3170 * covering maximum file size must fit into on-disk format containers as
3171 * well. Given that length is always by 1 unit bigger than max unit (because
3172 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3174 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3176 static loff_t ext4_max_size(int blkbits, int has_huge_files)
3179 loff_t upper_limit = MAX_LFS_FILESIZE;
3181 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3183 if (!has_huge_files) {
3184 upper_limit = (1LL << 32) - 1;
3186 /* total blocks in file system block size */
3187 upper_limit >>= (blkbits - 9);
3188 upper_limit <<= blkbits;
3192 * 32-bit extent-start container, ee_block. We lower the maxbytes
3193 * by one fs block, so ee_len can cover the extent of maximum file
3196 res = (1LL << 32) - 1;
3199 /* Sanity check against vm- & vfs- imposed limits */
3200 if (res > upper_limit)
3207 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
3208 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3209 * We need to be 1 filesystem block less than the 2^48 sector limit.
3211 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3213 loff_t res = EXT4_NDIR_BLOCKS;
3216 /* This is calculated to be the largest file size for a dense, block
3217 * mapped file such that the file's total number of 512-byte sectors,
3218 * including data and all indirect blocks, does not exceed (2^48 - 1).
3220 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3221 * number of 512-byte sectors of the file.
3224 if (!has_huge_files) {
3226 * !has_huge_files or implies that the inode i_block field
3227 * represents total file blocks in 2^32 512-byte sectors ==
3228 * size of vfs inode i_blocks * 8
3230 upper_limit = (1LL << 32) - 1;
3232 /* total blocks in file system block size */
3233 upper_limit >>= (bits - 9);
3237 * We use 48 bit ext4_inode i_blocks
3238 * With EXT4_HUGE_FILE_FL set the i_blocks
3239 * represent total number of blocks in
3240 * file system block size
3242 upper_limit = (1LL << 48) - 1;
3246 /* indirect blocks */
3248 /* double indirect blocks */
3249 meta_blocks += 1 + (1LL << (bits-2));
3250 /* tripple indirect blocks */
3251 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
3253 upper_limit -= meta_blocks;
3254 upper_limit <<= bits;
3256 res += 1LL << (bits-2);
3257 res += 1LL << (2*(bits-2));
3258 res += 1LL << (3*(bits-2));
3260 if (res > upper_limit)
3263 if (res > MAX_LFS_FILESIZE)
3264 res = MAX_LFS_FILESIZE;
3269 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3270 ext4_fsblk_t logical_sb_block, int nr)
3272 struct ext4_sb_info *sbi = EXT4_SB(sb);
3273 ext4_group_t bg, first_meta_bg;
3276 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3278 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3279 return logical_sb_block + nr + 1;
3280 bg = sbi->s_desc_per_block * nr;
3281 if (ext4_bg_has_super(sb, bg))
3285 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3286 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3287 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3290 if (sb->s_blocksize == 1024 && nr == 0 &&
3291 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3294 return (has_super + ext4_group_first_block_no(sb, bg));
3298 * ext4_get_stripe_size: Get the stripe size.
3299 * @sbi: In memory super block info
3301 * If we have specified it via mount option, then
3302 * use the mount option value. If the value specified at mount time is
3303 * greater than the blocks per group use the super block value.
3304 * If the super block value is greater than blocks per group return 0.
3305 * Allocator needs it be less than blocks per group.
3308 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3310 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3311 unsigned long stripe_width =
3312 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3315 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3316 ret = sbi->s_stripe;
3317 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3319 else if (stride && stride <= sbi->s_blocks_per_group)
3325 * If the stripe width is 1, this makes no sense and
3326 * we set it to 0 to turn off stripe handling code.
3335 * Check whether this filesystem can be mounted based on
3336 * the features present and the RDONLY/RDWR mount requested.
3337 * Returns 1 if this filesystem can be mounted as requested,
3338 * 0 if it cannot be.
3340 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
3342 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3343 ext4_msg(sb, KERN_ERR,
3344 "Couldn't mount because of "
3345 "unsupported optional features (%x)",
3346 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3347 ~EXT4_FEATURE_INCOMPAT_SUPP));
3351 #ifndef CONFIG_UNICODE
3352 if (ext4_has_feature_casefold(sb)) {
3353 ext4_msg(sb, KERN_ERR,
3354 "Filesystem with casefold feature cannot be "
3355 "mounted without CONFIG_UNICODE");
3363 if (ext4_has_feature_readonly(sb)) {
3364 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3365 sb->s_flags |= SB_RDONLY;
3369 /* Check that feature set is OK for a read-write mount */
3370 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3371 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3372 "unsupported optional features (%x)",
3373 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3374 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3377 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3378 ext4_msg(sb, KERN_ERR,
3379 "Can't support bigalloc feature without "
3380 "extents feature\n");
3384 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3385 if (!readonly && (ext4_has_feature_quota(sb) ||
3386 ext4_has_feature_project(sb))) {
3387 ext4_msg(sb, KERN_ERR,
3388 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3391 #endif /* CONFIG_QUOTA */
3396 * This function is called once a day if we have errors logged
3397 * on the file system
3399 static void print_daily_error_info(struct timer_list *t)
3401 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3402 struct super_block *sb = sbi->s_sb;
3403 struct ext4_super_block *es = sbi->s_es;
3405 if (es->s_error_count)
3406 /* fsck newer than v1.41.13 is needed to clean this condition. */
3407 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3408 le32_to_cpu(es->s_error_count));
3409 if (es->s_first_error_time) {
3410 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3412 ext4_get_tstamp(es, s_first_error_time),
3413 (int) sizeof(es->s_first_error_func),
3414 es->s_first_error_func,
3415 le32_to_cpu(es->s_first_error_line));
3416 if (es->s_first_error_ino)
3417 printk(KERN_CONT ": inode %u",
3418 le32_to_cpu(es->s_first_error_ino));
3419 if (es->s_first_error_block)
3420 printk(KERN_CONT ": block %llu", (unsigned long long)
3421 le64_to_cpu(es->s_first_error_block));
3422 printk(KERN_CONT "\n");
3424 if (es->s_last_error_time) {
3425 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3427 ext4_get_tstamp(es, s_last_error_time),
3428 (int) sizeof(es->s_last_error_func),
3429 es->s_last_error_func,
3430 le32_to_cpu(es->s_last_error_line));
3431 if (es->s_last_error_ino)
3432 printk(KERN_CONT ": inode %u",
3433 le32_to_cpu(es->s_last_error_ino));
3434 if (es->s_last_error_block)
3435 printk(KERN_CONT ": block %llu", (unsigned long long)
3436 le64_to_cpu(es->s_last_error_block));
3437 printk(KERN_CONT "\n");
3439 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3442 /* Find next suitable group and run ext4_init_inode_table */
3443 static int ext4_run_li_request(struct ext4_li_request *elr)
3445 struct ext4_group_desc *gdp = NULL;
3446 struct super_block *sb = elr->lr_super;
3447 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3448 ext4_group_t group = elr->lr_next_group;
3449 unsigned long timeout = 0;
3450 unsigned int prefetch_ios = 0;
3453 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3454 elr->lr_next_group = ext4_mb_prefetch(sb, group,
3455 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3457 ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3459 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3461 if (group >= elr->lr_next_group) {
3463 if (elr->lr_first_not_zeroed != ngroups &&
3464 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3465 elr->lr_next_group = elr->lr_first_not_zeroed;
3466 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3473 for (; group < ngroups; group++) {
3474 gdp = ext4_get_group_desc(sb, group, NULL);
3480 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3484 if (group >= ngroups)
3489 ret = ext4_init_inode_table(sb, group,
3490 elr->lr_timeout ? 0 : 1);
3491 trace_ext4_lazy_itable_init(sb, group);
3492 if (elr->lr_timeout == 0) {
3493 timeout = (jiffies - timeout) *
3494 EXT4_SB(elr->lr_super)->s_li_wait_mult;
3495 elr->lr_timeout = timeout;
3497 elr->lr_next_sched = jiffies + elr->lr_timeout;
3498 elr->lr_next_group = group + 1;
3504 * Remove lr_request from the list_request and free the
3505 * request structure. Should be called with li_list_mtx held
3507 static void ext4_remove_li_request(struct ext4_li_request *elr)
3512 list_del(&elr->lr_request);
3513 EXT4_SB(elr->lr_super)->s_li_request = NULL;
3517 static void ext4_unregister_li_request(struct super_block *sb)
3519 mutex_lock(&ext4_li_mtx);
3520 if (!ext4_li_info) {
3521 mutex_unlock(&ext4_li_mtx);
3525 mutex_lock(&ext4_li_info->li_list_mtx);
3526 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3527 mutex_unlock(&ext4_li_info->li_list_mtx);
3528 mutex_unlock(&ext4_li_mtx);
3531 static struct task_struct *ext4_lazyinit_task;
3534 * This is the function where ext4lazyinit thread lives. It walks
3535 * through the request list searching for next scheduled filesystem.
3536 * When such a fs is found, run the lazy initialization request
3537 * (ext4_rn_li_request) and keep track of the time spend in this
3538 * function. Based on that time we compute next schedule time of
3539 * the request. When walking through the list is complete, compute
3540 * next waking time and put itself into sleep.
3542 static int ext4_lazyinit_thread(void *arg)
3544 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3545 struct list_head *pos, *n;
3546 struct ext4_li_request *elr;
3547 unsigned long next_wakeup, cur;
3549 BUG_ON(NULL == eli);
3553 next_wakeup = MAX_JIFFY_OFFSET;
3555 mutex_lock(&eli->li_list_mtx);
3556 if (list_empty(&eli->li_request_list)) {
3557 mutex_unlock(&eli->li_list_mtx);
3560 list_for_each_safe(pos, n, &eli->li_request_list) {
3563 elr = list_entry(pos, struct ext4_li_request,
3566 if (time_before(jiffies, elr->lr_next_sched)) {
3567 if (time_before(elr->lr_next_sched, next_wakeup))
3568 next_wakeup = elr->lr_next_sched;
3571 if (down_read_trylock(&elr->lr_super->s_umount)) {
3572 if (sb_start_write_trylock(elr->lr_super)) {
3575 * We hold sb->s_umount, sb can not
3576 * be removed from the list, it is
3577 * now safe to drop li_list_mtx
3579 mutex_unlock(&eli->li_list_mtx);
3580 err = ext4_run_li_request(elr);
3581 sb_end_write(elr->lr_super);
3582 mutex_lock(&eli->li_list_mtx);
3585 up_read((&elr->lr_super->s_umount));
3587 /* error, remove the lazy_init job */
3589 ext4_remove_li_request(elr);
3593 elr->lr_next_sched = jiffies +
3595 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3597 if (time_before(elr->lr_next_sched, next_wakeup))
3598 next_wakeup = elr->lr_next_sched;
3600 mutex_unlock(&eli->li_list_mtx);
3605 if ((time_after_eq(cur, next_wakeup)) ||
3606 (MAX_JIFFY_OFFSET == next_wakeup)) {
3611 schedule_timeout_interruptible(next_wakeup - cur);
3613 if (kthread_should_stop()) {
3614 ext4_clear_request_list();
3621 * It looks like the request list is empty, but we need
3622 * to check it under the li_list_mtx lock, to prevent any
3623 * additions into it, and of course we should lock ext4_li_mtx
3624 * to atomically free the list and ext4_li_info, because at
3625 * this point another ext4 filesystem could be registering
3628 mutex_lock(&ext4_li_mtx);
3629 mutex_lock(&eli->li_list_mtx);
3630 if (!list_empty(&eli->li_request_list)) {
3631 mutex_unlock(&eli->li_list_mtx);
3632 mutex_unlock(&ext4_li_mtx);
3635 mutex_unlock(&eli->li_list_mtx);
3636 kfree(ext4_li_info);
3637 ext4_li_info = NULL;
3638 mutex_unlock(&ext4_li_mtx);
3643 static void ext4_clear_request_list(void)
3645 struct list_head *pos, *n;
3646 struct ext4_li_request *elr;
3648 mutex_lock(&ext4_li_info->li_list_mtx);
3649 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3650 elr = list_entry(pos, struct ext4_li_request,
3652 ext4_remove_li_request(elr);
3654 mutex_unlock(&ext4_li_info->li_list_mtx);
3657 static int ext4_run_lazyinit_thread(void)
3659 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3660 ext4_li_info, "ext4lazyinit");
3661 if (IS_ERR(ext4_lazyinit_task)) {
3662 int err = PTR_ERR(ext4_lazyinit_task);
3663 ext4_clear_request_list();
3664 kfree(ext4_li_info);
3665 ext4_li_info = NULL;
3666 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3667 "initialization thread\n",
3671 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3676 * Check whether it make sense to run itable init. thread or not.
3677 * If there is at least one uninitialized inode table, return
3678 * corresponding group number, else the loop goes through all
3679 * groups and return total number of groups.
3681 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3683 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3684 struct ext4_group_desc *gdp = NULL;
3686 if (!ext4_has_group_desc_csum(sb))
3689 for (group = 0; group < ngroups; group++) {
3690 gdp = ext4_get_group_desc(sb, group, NULL);
3694 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3701 static int ext4_li_info_new(void)
3703 struct ext4_lazy_init *eli = NULL;
3705 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3709 INIT_LIST_HEAD(&eli->li_request_list);
3710 mutex_init(&eli->li_list_mtx);
3712 eli->li_state |= EXT4_LAZYINIT_QUIT;
3719 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3722 struct ext4_li_request *elr;
3724 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3729 elr->lr_first_not_zeroed = start;
3730 if (test_opt(sb, PREFETCH_BLOCK_BITMAPS))
3731 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3733 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3734 elr->lr_next_group = start;
3738 * Randomize first schedule time of the request to
3739 * spread the inode table initialization requests
3742 elr->lr_next_sched = jiffies + (prandom_u32() %
3743 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3747 int ext4_register_li_request(struct super_block *sb,
3748 ext4_group_t first_not_zeroed)
3750 struct ext4_sb_info *sbi = EXT4_SB(sb);
3751 struct ext4_li_request *elr = NULL;
3752 ext4_group_t ngroups = sbi->s_groups_count;
3755 mutex_lock(&ext4_li_mtx);
3756 if (sbi->s_li_request != NULL) {
3758 * Reset timeout so it can be computed again, because
3759 * s_li_wait_mult might have changed.
3761 sbi->s_li_request->lr_timeout = 0;
3765 if (!test_opt(sb, PREFETCH_BLOCK_BITMAPS) &&
3766 (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3767 !test_opt(sb, INIT_INODE_TABLE)))
3770 elr = ext4_li_request_new(sb, first_not_zeroed);
3776 if (NULL == ext4_li_info) {
3777 ret = ext4_li_info_new();
3782 mutex_lock(&ext4_li_info->li_list_mtx);
3783 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3784 mutex_unlock(&ext4_li_info->li_list_mtx);
3786 sbi->s_li_request = elr;
3788 * set elr to NULL here since it has been inserted to
3789 * the request_list and the removal and free of it is
3790 * handled by ext4_clear_request_list from now on.
3794 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3795 ret = ext4_run_lazyinit_thread();
3800 mutex_unlock(&ext4_li_mtx);
3807 * We do not need to lock anything since this is called on
3810 static void ext4_destroy_lazyinit_thread(void)
3813 * If thread exited earlier
3814 * there's nothing to be done.
3816 if (!ext4_li_info || !ext4_lazyinit_task)
3819 kthread_stop(ext4_lazyinit_task);
3822 static int set_journal_csum_feature_set(struct super_block *sb)
3825 int compat, incompat;
3826 struct ext4_sb_info *sbi = EXT4_SB(sb);
3828 if (ext4_has_metadata_csum(sb)) {
3829 /* journal checksum v3 */
3831 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3833 /* journal checksum v1 */
3834 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3838 jbd2_journal_clear_features(sbi->s_journal,
3839 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3840 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3841 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3842 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3843 ret = jbd2_journal_set_features(sbi->s_journal,
3845 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3847 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3848 ret = jbd2_journal_set_features(sbi->s_journal,
3851 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3852 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3854 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3855 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3862 * Note: calculating the overhead so we can be compatible with
3863 * historical BSD practice is quite difficult in the face of
3864 * clusters/bigalloc. This is because multiple metadata blocks from
3865 * different block group can end up in the same allocation cluster.
3866 * Calculating the exact overhead in the face of clustered allocation
3867 * requires either O(all block bitmaps) in memory or O(number of block
3868 * groups**2) in time. We will still calculate the superblock for
3869 * older file systems --- and if we come across with a bigalloc file
3870 * system with zero in s_overhead_clusters the estimate will be close to
3871 * correct especially for very large cluster sizes --- but for newer
3872 * file systems, it's better to calculate this figure once at mkfs
3873 * time, and store it in the superblock. If the superblock value is
3874 * present (even for non-bigalloc file systems), we will use it.
3876 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3879 struct ext4_sb_info *sbi = EXT4_SB(sb);
3880 struct ext4_group_desc *gdp;
3881 ext4_fsblk_t first_block, last_block, b;
3882 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3883 int s, j, count = 0;
3885 if (!ext4_has_feature_bigalloc(sb))
3886 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3887 sbi->s_itb_per_group + 2);
3889 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3890 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3891 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3892 for (i = 0; i < ngroups; i++) {
3893 gdp = ext4_get_group_desc(sb, i, NULL);
3894 b = ext4_block_bitmap(sb, gdp);
3895 if (b >= first_block && b <= last_block) {
3896 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3899 b = ext4_inode_bitmap(sb, gdp);
3900 if (b >= first_block && b <= last_block) {
3901 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3904 b = ext4_inode_table(sb, gdp);
3905 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3906 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3907 int c = EXT4_B2C(sbi, b - first_block);
3908 ext4_set_bit(c, buf);
3914 if (ext4_bg_has_super(sb, grp)) {
3915 ext4_set_bit(s++, buf);
3918 j = ext4_bg_num_gdb(sb, grp);
3919 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3920 ext4_error(sb, "Invalid number of block group "
3921 "descriptor blocks: %d", j);
3922 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3926 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3930 return EXT4_CLUSTERS_PER_GROUP(sb) -
3931 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3935 * Compute the overhead and stash it in sbi->s_overhead
3937 int ext4_calculate_overhead(struct super_block *sb)
3939 struct ext4_sb_info *sbi = EXT4_SB(sb);
3940 struct ext4_super_block *es = sbi->s_es;
3941 struct inode *j_inode;
3942 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3943 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3944 ext4_fsblk_t overhead = 0;
3945 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3951 * Compute the overhead (FS structures). This is constant
3952 * for a given filesystem unless the number of block groups
3953 * changes so we cache the previous value until it does.
3957 * All of the blocks before first_data_block are overhead
3959 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3962 * Add the overhead found in each block group
3964 for (i = 0; i < ngroups; i++) {
3967 blks = count_overhead(sb, i, buf);
3970 memset(buf, 0, PAGE_SIZE);
3975 * Add the internal journal blocks whether the journal has been
3978 if (sbi->s_journal && !sbi->s_journal_bdev)
3979 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3980 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
3981 /* j_inum for internal journal is non-zero */
3982 j_inode = ext4_get_journal_inode(sb, j_inum);
3984 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3985 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3988 ext4_msg(sb, KERN_ERR, "can't get journal size");
3991 sbi->s_overhead = overhead;
3993 free_page((unsigned long) buf);
3997 static void ext4_set_resv_clusters(struct super_block *sb)
3999 ext4_fsblk_t resv_clusters;
4000 struct ext4_sb_info *sbi = EXT4_SB(sb);
4003 * There's no need to reserve anything when we aren't using extents.
4004 * The space estimates are exact, there are no unwritten extents,
4005 * hole punching doesn't need new metadata... This is needed especially
4006 * to keep ext2/3 backward compatibility.
4008 if (!ext4_has_feature_extents(sb))
4011 * By default we reserve 2% or 4096 clusters, whichever is smaller.
4012 * This should cover the situations where we can not afford to run
4013 * out of space like for example punch hole, or converting
4014 * unwritten extents in delalloc path. In most cases such
4015 * allocation would require 1, or 2 blocks, higher numbers are
4018 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
4019 sbi->s_cluster_bits);
4021 do_div(resv_clusters, 50);
4022 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
4024 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
4027 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
4029 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
4030 char *orig_data = kstrdup(data, GFP_KERNEL);
4031 struct buffer_head *bh, **group_desc;
4032 struct ext4_super_block *es = NULL;
4033 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
4034 struct flex_groups **flex_groups;
4036 ext4_fsblk_t sb_block = get_sb_block(&data);
4037 ext4_fsblk_t logical_sb_block;
4038 unsigned long offset = 0;
4039 unsigned long journal_devnum = 0;
4040 unsigned long def_mount_opts;
4044 int blocksize, clustersize;
4045 unsigned int db_count;
4047 int needs_recovery, has_huge_files;
4050 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4051 ext4_group_t first_not_zeroed;
4053 if ((data && !orig_data) || !sbi)
4056 sbi->s_daxdev = dax_dev;
4057 sbi->s_blockgroup_lock =
4058 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
4059 if (!sbi->s_blockgroup_lock)
4062 sb->s_fs_info = sbi;
4064 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
4065 sbi->s_sb_block = sb_block;
4066 if (sb->s_bdev->bd_part)
4067 sbi->s_sectors_written_start =
4068 part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]);
4070 /* Cleanup superblock name */
4071 strreplace(sb->s_id, '/', '!');
4073 /* -EINVAL is default */
4075 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
4077 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
4082 * The ext4 superblock will not be buffer aligned for other than 1kB
4083 * block sizes. We need to calculate the offset from buffer start.
4085 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
4086 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4087 offset = do_div(logical_sb_block, blocksize);
4089 logical_sb_block = sb_block;
4092 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4094 ext4_msg(sb, KERN_ERR, "unable to read superblock");
4100 * Note: s_es must be initialized as soon as possible because
4101 * some ext4 macro-instructions depend on its value
4103 es = (struct ext4_super_block *) (bh->b_data + offset);
4105 sb->s_magic = le16_to_cpu(es->s_magic);
4106 if (sb->s_magic != EXT4_SUPER_MAGIC)
4108 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
4110 /* Warn if metadata_csum and gdt_csum are both set. */
4111 if (ext4_has_feature_metadata_csum(sb) &&
4112 ext4_has_feature_gdt_csum(sb))
4113 ext4_warning(sb, "metadata_csum and uninit_bg are "
4114 "redundant flags; please run fsck.");
4116 /* Check for a known checksum algorithm */
4117 if (!ext4_verify_csum_type(sb, es)) {
4118 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4119 "unknown checksum algorithm.");
4124 /* Load the checksum driver */
4125 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
4126 if (IS_ERR(sbi->s_chksum_driver)) {
4127 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
4128 ret = PTR_ERR(sbi->s_chksum_driver);
4129 sbi->s_chksum_driver = NULL;
4133 /* Check superblock checksum */
4134 if (!ext4_superblock_csum_verify(sb, es)) {
4135 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4136 "invalid superblock checksum. Run e2fsck?");
4142 /* Precompute checksum seed for all metadata */
4143 if (ext4_has_feature_csum_seed(sb))
4144 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
4145 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
4146 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
4147 sizeof(es->s_uuid));
4149 /* Set defaults before we parse the mount options */
4150 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4151 set_opt(sb, INIT_INODE_TABLE);
4152 if (def_mount_opts & EXT4_DEFM_DEBUG)
4154 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4156 if (def_mount_opts & EXT4_DEFM_UID16)
4157 set_opt(sb, NO_UID32);
4158 /* xattr user namespace & acls are now defaulted on */
4159 set_opt(sb, XATTR_USER);
4160 #ifdef CONFIG_EXT4_FS_POSIX_ACL
4161 set_opt(sb, POSIX_ACL);
4163 if (ext4_has_feature_fast_commit(sb))
4164 set_opt2(sb, JOURNAL_FAST_COMMIT);
4165 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4166 if (ext4_has_metadata_csum(sb))
4167 set_opt(sb, JOURNAL_CHECKSUM);
4169 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4170 set_opt(sb, JOURNAL_DATA);
4171 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4172 set_opt(sb, ORDERED_DATA);
4173 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4174 set_opt(sb, WRITEBACK_DATA);
4176 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
4177 set_opt(sb, ERRORS_PANIC);
4178 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
4179 set_opt(sb, ERRORS_CONT);
4181 set_opt(sb, ERRORS_RO);
4182 /* block_validity enabled by default; disable with noblock_validity */
4183 set_opt(sb, BLOCK_VALIDITY);
4184 if (def_mount_opts & EXT4_DEFM_DISCARD)
4185 set_opt(sb, DISCARD);
4187 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
4188 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
4189 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
4190 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
4191 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
4193 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4194 set_opt(sb, BARRIER);
4197 * enable delayed allocation by default
4198 * Use -o nodelalloc to turn it off
4200 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4201 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4202 set_opt(sb, DELALLOC);
4205 * set default s_li_wait_mult for lazyinit, for the case there is
4206 * no mount option specified.
4208 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
4210 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
4212 if (blocksize == PAGE_SIZE)
4213 set_opt(sb, DIOREAD_NOLOCK);
4215 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
4216 blocksize > EXT4_MAX_BLOCK_SIZE) {
4217 ext4_msg(sb, KERN_ERR,
4218 "Unsupported filesystem blocksize %d (%d log_block_size)",
4219 blocksize, le32_to_cpu(es->s_log_block_size));
4223 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4224 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4225 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4227 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4228 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4229 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4230 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4234 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4235 (!is_power_of_2(sbi->s_inode_size)) ||
4236 (sbi->s_inode_size > blocksize)) {
4237 ext4_msg(sb, KERN_ERR,
4238 "unsupported inode size: %d",
4240 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
4244 * i_atime_extra is the last extra field available for
4245 * [acm]times in struct ext4_inode. Checking for that
4246 * field should suffice to ensure we have extra space
4249 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4250 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4251 sb->s_time_gran = 1;
4252 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4254 sb->s_time_gran = NSEC_PER_SEC;
4255 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4257 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4259 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4260 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4261 EXT4_GOOD_OLD_INODE_SIZE;
4262 if (ext4_has_feature_extra_isize(sb)) {
4263 unsigned v, max = (sbi->s_inode_size -
4264 EXT4_GOOD_OLD_INODE_SIZE);
4266 v = le16_to_cpu(es->s_want_extra_isize);
4268 ext4_msg(sb, KERN_ERR,
4269 "bad s_want_extra_isize: %d", v);
4272 if (sbi->s_want_extra_isize < v)
4273 sbi->s_want_extra_isize = v;
4275 v = le16_to_cpu(es->s_min_extra_isize);
4277 ext4_msg(sb, KERN_ERR,
4278 "bad s_min_extra_isize: %d", v);
4281 if (sbi->s_want_extra_isize < v)
4282 sbi->s_want_extra_isize = v;
4286 if (sbi->s_es->s_mount_opts[0]) {
4287 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
4288 sizeof(sbi->s_es->s_mount_opts),
4292 if (!parse_options(s_mount_opts, sb, &journal_devnum,
4293 &journal_ioprio, 0)) {
4294 ext4_msg(sb, KERN_WARNING,
4295 "failed to parse options in superblock: %s",
4298 kfree(s_mount_opts);
4300 sbi->s_def_mount_opt = sbi->s_mount_opt;
4301 if (!parse_options((char *) data, sb, &journal_devnum,
4302 &journal_ioprio, 0))
4305 #ifdef CONFIG_UNICODE
4306 if (ext4_has_feature_casefold(sb) && !sbi->s_encoding) {
4307 const struct ext4_sb_encodings *encoding_info;
4308 struct unicode_map *encoding;
4309 __u16 encoding_flags;
4311 if (ext4_has_feature_encrypt(sb)) {
4312 ext4_msg(sb, KERN_ERR,
4313 "Can't mount with encoding and encryption");
4317 if (ext4_sb_read_encoding(es, &encoding_info,
4319 ext4_msg(sb, KERN_ERR,
4320 "Encoding requested by superblock is unknown");
4324 encoding = utf8_load(encoding_info->version);
4325 if (IS_ERR(encoding)) {
4326 ext4_msg(sb, KERN_ERR,
4327 "can't mount with superblock charset: %s-%s "
4328 "not supported by the kernel. flags: 0x%x.",
4329 encoding_info->name, encoding_info->version,
4333 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4334 "%s-%s with flags 0x%hx", encoding_info->name,
4335 encoding_info->version?:"\b", encoding_flags);
4337 sbi->s_encoding = encoding;
4338 sbi->s_encoding_flags = encoding_flags;
4342 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4343 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, and O_DIRECT support!\n");
4344 /* can't mount with both data=journal and dioread_nolock. */
4345 clear_opt(sb, DIOREAD_NOLOCK);
4346 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4347 ext4_msg(sb, KERN_ERR, "can't mount with "
4348 "both data=journal and delalloc");
4351 if (test_opt(sb, DAX_ALWAYS)) {
4352 ext4_msg(sb, KERN_ERR, "can't mount with "
4353 "both data=journal and dax");
4356 if (ext4_has_feature_encrypt(sb)) {
4357 ext4_msg(sb, KERN_WARNING,
4358 "encrypted files will use data=ordered "
4359 "instead of data journaling mode");
4361 if (test_opt(sb, DELALLOC))
4362 clear_opt(sb, DELALLOC);
4364 sb->s_iflags |= SB_I_CGROUPWB;
4367 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4368 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
4370 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4371 (ext4_has_compat_features(sb) ||
4372 ext4_has_ro_compat_features(sb) ||
4373 ext4_has_incompat_features(sb)))
4374 ext4_msg(sb, KERN_WARNING,
4375 "feature flags set on rev 0 fs, "
4376 "running e2fsck is recommended");
4378 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4379 set_opt2(sb, HURD_COMPAT);
4380 if (ext4_has_feature_64bit(sb)) {
4381 ext4_msg(sb, KERN_ERR,
4382 "The Hurd can't support 64-bit file systems");
4387 * ea_inode feature uses l_i_version field which is not
4388 * available in HURD_COMPAT mode.
4390 if (ext4_has_feature_ea_inode(sb)) {
4391 ext4_msg(sb, KERN_ERR,
4392 "ea_inode feature is not supported for Hurd");
4397 if (IS_EXT2_SB(sb)) {
4398 if (ext2_feature_set_ok(sb))
4399 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4400 "using the ext4 subsystem");
4403 * If we're probing be silent, if this looks like
4404 * it's actually an ext[34] filesystem.
4406 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4408 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4409 "to feature incompatibilities");
4414 if (IS_EXT3_SB(sb)) {
4415 if (ext3_feature_set_ok(sb))
4416 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4417 "using the ext4 subsystem");
4420 * If we're probing be silent, if this looks like
4421 * it's actually an ext4 filesystem.
4423 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4425 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4426 "to feature incompatibilities");
4432 * Check feature flags regardless of the revision level, since we
4433 * previously didn't change the revision level when setting the flags,
4434 * so there is a chance incompat flags are set on a rev 0 filesystem.
4436 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4439 if (le32_to_cpu(es->s_log_block_size) >
4440 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4441 ext4_msg(sb, KERN_ERR,
4442 "Invalid log block size: %u",
4443 le32_to_cpu(es->s_log_block_size));
4446 if (le32_to_cpu(es->s_log_cluster_size) >
4447 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4448 ext4_msg(sb, KERN_ERR,
4449 "Invalid log cluster size: %u",
4450 le32_to_cpu(es->s_log_cluster_size));
4454 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4455 ext4_msg(sb, KERN_ERR,
4456 "Number of reserved GDT blocks insanely large: %d",
4457 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4461 if (bdev_dax_supported(sb->s_bdev, blocksize))
4462 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4464 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4465 if (ext4_has_feature_inline_data(sb)) {
4466 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4467 " that may contain inline data");
4470 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4471 ext4_msg(sb, KERN_ERR,
4472 "DAX unsupported by block device.");
4477 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4478 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4479 es->s_encryption_level);
4483 if (sb->s_blocksize != blocksize) {
4484 /* Validate the filesystem blocksize */
4485 if (!sb_set_blocksize(sb, blocksize)) {
4486 ext4_msg(sb, KERN_ERR, "bad block size %d",
4492 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4493 offset = do_div(logical_sb_block, blocksize);
4494 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4496 ext4_msg(sb, KERN_ERR,
4497 "Can't read superblock on 2nd try");
4502 es = (struct ext4_super_block *)(bh->b_data + offset);
4504 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
4505 ext4_msg(sb, KERN_ERR,
4506 "Magic mismatch, very weird!");
4511 has_huge_files = ext4_has_feature_huge_file(sb);
4512 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4514 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
4516 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
4517 if (ext4_has_feature_64bit(sb)) {
4518 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
4519 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
4520 !is_power_of_2(sbi->s_desc_size)) {
4521 ext4_msg(sb, KERN_ERR,
4522 "unsupported descriptor size %lu",
4527 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
4529 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
4530 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
4532 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
4533 if (sbi->s_inodes_per_block == 0)
4535 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4536 sbi->s_inodes_per_group > blocksize * 8) {
4537 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
4538 sbi->s_inodes_per_group);
4541 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4542 sbi->s_inodes_per_block;
4543 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
4545 sbi->s_mount_state = le16_to_cpu(es->s_state);
4546 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4547 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
4549 for (i = 0; i < 4; i++)
4550 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4551 sbi->s_def_hash_version = es->s_def_hash_version;
4552 if (ext4_has_feature_dir_index(sb)) {
4553 i = le32_to_cpu(es->s_flags);
4554 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4555 sbi->s_hash_unsigned = 3;
4556 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
4557 #ifdef __CHAR_UNSIGNED__
4560 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4561 sbi->s_hash_unsigned = 3;
4565 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
4570 /* Handle clustersize */
4571 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4572 if (ext4_has_feature_bigalloc(sb)) {
4573 if (clustersize < blocksize) {
4574 ext4_msg(sb, KERN_ERR,
4575 "cluster size (%d) smaller than "
4576 "block size (%d)", clustersize, blocksize);
4579 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4580 le32_to_cpu(es->s_log_block_size);
4581 sbi->s_clusters_per_group =
4582 le32_to_cpu(es->s_clusters_per_group);
4583 if (sbi->s_clusters_per_group > blocksize * 8) {
4584 ext4_msg(sb, KERN_ERR,
4585 "#clusters per group too big: %lu",
4586 sbi->s_clusters_per_group);
4589 if (sbi->s_blocks_per_group !=
4590 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4591 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4592 "clusters per group (%lu) inconsistent",
4593 sbi->s_blocks_per_group,
4594 sbi->s_clusters_per_group);
4598 if (clustersize != blocksize) {
4599 ext4_msg(sb, KERN_ERR,
4600 "fragment/cluster size (%d) != "
4601 "block size (%d)", clustersize, blocksize);
4604 if (sbi->s_blocks_per_group > blocksize * 8) {
4605 ext4_msg(sb, KERN_ERR,
4606 "#blocks per group too big: %lu",
4607 sbi->s_blocks_per_group);
4610 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4611 sbi->s_cluster_bits = 0;
4613 sbi->s_cluster_ratio = clustersize / blocksize;
4615 /* Do we have standard group size of clustersize * 8 blocks ? */
4616 if (sbi->s_blocks_per_group == clustersize << 3)
4617 set_opt2(sb, STD_GROUP_SIZE);
4620 * Test whether we have more sectors than will fit in sector_t,
4621 * and whether the max offset is addressable by the page cache.
4623 err = generic_check_addressable(sb->s_blocksize_bits,
4624 ext4_blocks_count(es));
4626 ext4_msg(sb, KERN_ERR, "filesystem"
4627 " too large to mount safely on this system");
4631 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4634 /* check blocks count against device size */
4635 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4636 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4637 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4638 "exceeds size of device (%llu blocks)",
4639 ext4_blocks_count(es), blocks_count);
4644 * It makes no sense for the first data block to be beyond the end
4645 * of the filesystem.
4647 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4648 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4649 "block %u is beyond end of filesystem (%llu)",
4650 le32_to_cpu(es->s_first_data_block),
4651 ext4_blocks_count(es));
4654 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4655 (sbi->s_cluster_ratio == 1)) {
4656 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4657 "block is 0 with a 1k block and cluster size");
4661 blocks_count = (ext4_blocks_count(es) -
4662 le32_to_cpu(es->s_first_data_block) +
4663 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4664 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4665 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4666 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4667 "(block count %llu, first data block %u, "
4668 "blocks per group %lu)", blocks_count,
4669 ext4_blocks_count(es),
4670 le32_to_cpu(es->s_first_data_block),
4671 EXT4_BLOCKS_PER_GROUP(sb));
4674 sbi->s_groups_count = blocks_count;
4675 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4676 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4677 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4678 le32_to_cpu(es->s_inodes_count)) {
4679 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4680 le32_to_cpu(es->s_inodes_count),
4681 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4685 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4686 EXT4_DESC_PER_BLOCK(sb);
4687 if (ext4_has_feature_meta_bg(sb)) {
4688 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4689 ext4_msg(sb, KERN_WARNING,
4690 "first meta block group too large: %u "
4691 "(group descriptor block count %u)",
4692 le32_to_cpu(es->s_first_meta_bg), db_count);
4696 rcu_assign_pointer(sbi->s_group_desc,
4697 kvmalloc_array(db_count,
4698 sizeof(struct buffer_head *),
4700 if (sbi->s_group_desc == NULL) {
4701 ext4_msg(sb, KERN_ERR, "not enough memory");
4706 bgl_lock_init(sbi->s_blockgroup_lock);
4708 /* Pre-read the descriptors into the buffer cache */
4709 for (i = 0; i < db_count; i++) {
4710 block = descriptor_loc(sb, logical_sb_block, i);
4711 ext4_sb_breadahead_unmovable(sb, block);
4714 for (i = 0; i < db_count; i++) {
4715 struct buffer_head *bh;
4717 block = descriptor_loc(sb, logical_sb_block, i);
4718 bh = ext4_sb_bread_unmovable(sb, block);
4720 ext4_msg(sb, KERN_ERR,
4721 "can't read group descriptor %d", i);
4728 rcu_dereference(sbi->s_group_desc)[i] = bh;
4731 sbi->s_gdb_count = db_count;
4732 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4733 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4734 ret = -EFSCORRUPTED;
4738 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
4740 /* Register extent status tree shrinker */
4741 if (ext4_es_register_shrinker(sbi))
4744 sbi->s_stripe = ext4_get_stripe_size(sbi);
4745 sbi->s_extent_max_zeroout_kb = 32;
4748 * set up enough so that it can read an inode
4750 sb->s_op = &ext4_sops;
4751 sb->s_export_op = &ext4_export_ops;
4752 sb->s_xattr = ext4_xattr_handlers;
4753 #ifdef CONFIG_FS_ENCRYPTION
4754 sb->s_cop = &ext4_cryptops;
4756 #ifdef CONFIG_FS_VERITY
4757 sb->s_vop = &ext4_verityops;
4760 sb->dq_op = &ext4_quota_operations;
4761 if (ext4_has_feature_quota(sb))
4762 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4764 sb->s_qcop = &ext4_qctl_operations;
4765 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4767 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4769 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4770 mutex_init(&sbi->s_orphan_lock);
4772 /* Initialize fast commit stuff */
4773 atomic_set(&sbi->s_fc_subtid, 0);
4774 atomic_set(&sbi->s_fc_ineligible_updates, 0);
4775 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
4776 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
4777 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
4778 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
4779 sbi->s_fc_bytes = 0;
4780 sbi->s_mount_state &= ~EXT4_FC_INELIGIBLE;
4781 sbi->s_mount_state &= ~EXT4_FC_COMMITTING;
4782 spin_lock_init(&sbi->s_fc_lock);
4783 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
4784 sbi->s_fc_replay_state.fc_regions = NULL;
4785 sbi->s_fc_replay_state.fc_regions_size = 0;
4786 sbi->s_fc_replay_state.fc_regions_used = 0;
4787 sbi->s_fc_replay_state.fc_regions_valid = 0;
4788 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
4789 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
4790 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
4794 needs_recovery = (es->s_last_orphan != 0 ||
4795 ext4_has_feature_journal_needs_recovery(sb));
4797 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
4798 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4799 goto failed_mount3a;
4802 * The first inode we look at is the journal inode. Don't try
4803 * root first: it may be modified in the journal!
4805 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4806 err = ext4_load_journal(sb, es, journal_devnum);
4808 goto failed_mount3a;
4809 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4810 ext4_has_feature_journal_needs_recovery(sb)) {
4811 ext4_msg(sb, KERN_ERR, "required journal recovery "
4812 "suppressed and not mounted read-only");
4813 goto failed_mount_wq;
4815 /* Nojournal mode, all journal mount options are illegal */
4816 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4817 ext4_msg(sb, KERN_ERR, "can't mount with "
4818 "journal_checksum, fs mounted w/o journal");
4819 goto failed_mount_wq;
4821 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4822 ext4_msg(sb, KERN_ERR, "can't mount with "
4823 "journal_async_commit, fs mounted w/o journal");
4824 goto failed_mount_wq;
4826 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4827 ext4_msg(sb, KERN_ERR, "can't mount with "
4828 "commit=%lu, fs mounted w/o journal",
4829 sbi->s_commit_interval / HZ);
4830 goto failed_mount_wq;
4832 if (EXT4_MOUNT_DATA_FLAGS &
4833 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4834 ext4_msg(sb, KERN_ERR, "can't mount with "
4835 "data=, fs mounted w/o journal");
4836 goto failed_mount_wq;
4838 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
4839 clear_opt(sb, JOURNAL_CHECKSUM);
4840 clear_opt(sb, DATA_FLAGS);
4841 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4842 sbi->s_journal = NULL;
4847 if (ext4_has_feature_64bit(sb) &&
4848 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4849 JBD2_FEATURE_INCOMPAT_64BIT)) {
4850 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4851 goto failed_mount_wq;
4854 if (!set_journal_csum_feature_set(sb)) {
4855 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4857 goto failed_mount_wq;
4860 /* We have now updated the journal if required, so we can
4861 * validate the data journaling mode. */
4862 switch (test_opt(sb, DATA_FLAGS)) {
4864 /* No mode set, assume a default based on the journal
4865 * capabilities: ORDERED_DATA if the journal can
4866 * cope, else JOURNAL_DATA
4868 if (jbd2_journal_check_available_features
4869 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4870 set_opt(sb, ORDERED_DATA);
4871 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4873 set_opt(sb, JOURNAL_DATA);
4874 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4878 case EXT4_MOUNT_ORDERED_DATA:
4879 case EXT4_MOUNT_WRITEBACK_DATA:
4880 if (!jbd2_journal_check_available_features
4881 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4882 ext4_msg(sb, KERN_ERR, "Journal does not support "
4883 "requested data journaling mode");
4884 goto failed_mount_wq;
4890 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4891 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4892 ext4_msg(sb, KERN_ERR, "can't mount with "
4893 "journal_async_commit in data=ordered mode");
4894 goto failed_mount_wq;
4897 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4899 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4900 sbi->s_journal->j_submit_inode_data_buffers =
4901 ext4_journal_submit_inode_data_buffers;
4902 sbi->s_journal->j_finish_inode_data_buffers =
4903 ext4_journal_finish_inode_data_buffers;
4906 if (!test_opt(sb, NO_MBCACHE)) {
4907 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4908 if (!sbi->s_ea_block_cache) {
4909 ext4_msg(sb, KERN_ERR,
4910 "Failed to create ea_block_cache");
4911 goto failed_mount_wq;
4914 if (ext4_has_feature_ea_inode(sb)) {
4915 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4916 if (!sbi->s_ea_inode_cache) {
4917 ext4_msg(sb, KERN_ERR,
4918 "Failed to create ea_inode_cache");
4919 goto failed_mount_wq;
4924 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
4925 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
4926 goto failed_mount_wq;
4929 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
4930 !ext4_has_feature_encrypt(sb)) {
4931 ext4_set_feature_encrypt(sb);
4932 ext4_commit_super(sb, 1);
4936 * Get the # of file system overhead blocks from the
4937 * superblock if present.
4939 if (es->s_overhead_clusters)
4940 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4942 err = ext4_calculate_overhead(sb);
4944 goto failed_mount_wq;
4948 * The maximum number of concurrent works can be high and
4949 * concurrency isn't really necessary. Limit it to 1.
4951 EXT4_SB(sb)->rsv_conversion_wq =
4952 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4953 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4954 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4960 * The jbd2_journal_load will have done any necessary log recovery,
4961 * so we can safely mount the rest of the filesystem now.
4964 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
4966 ext4_msg(sb, KERN_ERR, "get root inode failed");
4967 ret = PTR_ERR(root);
4971 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4972 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4977 #ifdef CONFIG_UNICODE
4978 if (sbi->s_encoding)
4979 sb->s_d_op = &ext4_dentry_ops;
4982 sb->s_root = d_make_root(root);
4984 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4989 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4990 if (ret == -EROFS) {
4991 sb->s_flags |= SB_RDONLY;
4994 goto failed_mount4a;
4996 ext4_set_resv_clusters(sb);
4998 if (test_opt(sb, BLOCK_VALIDITY)) {
4999 err = ext4_setup_system_zone(sb);
5001 ext4_msg(sb, KERN_ERR, "failed to initialize system "
5003 goto failed_mount4a;
5006 ext4_fc_replay_cleanup(sb);
5009 err = ext4_mb_init(sb);
5011 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
5016 block = ext4_count_free_clusters(sb);
5017 ext4_free_blocks_count_set(sbi->s_es,
5018 EXT4_C2B(sbi, block));
5019 ext4_superblock_csum_set(sb);
5020 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
5023 unsigned long freei = ext4_count_free_inodes(sb);
5024 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
5025 ext4_superblock_csum_set(sb);
5026 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
5030 err = percpu_counter_init(&sbi->s_dirs_counter,
5031 ext4_count_dirs(sb), GFP_KERNEL);
5033 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
5036 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
5039 ext4_msg(sb, KERN_ERR, "insufficient memory");
5043 if (ext4_has_feature_flex_bg(sb))
5044 if (!ext4_fill_flex_info(sb)) {
5045 ext4_msg(sb, KERN_ERR,
5046 "unable to initialize "
5047 "flex_bg meta info!");
5051 err = ext4_register_li_request(sb, first_not_zeroed);
5055 err = ext4_register_sysfs(sb);
5060 /* Enable quota usage during mount. */
5061 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
5062 err = ext4_enable_quotas(sb);
5066 #endif /* CONFIG_QUOTA */
5069 * Save the original bdev mapping's wb_err value which could be
5070 * used to detect the metadata async write error.
5072 spin_lock_init(&sbi->s_bdev_wb_lock);
5073 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
5074 &sbi->s_bdev_wb_err);
5075 sb->s_bdev->bd_super = sb;
5076 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
5077 ext4_orphan_cleanup(sb, es);
5078 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
5079 if (needs_recovery) {
5080 ext4_msg(sb, KERN_INFO, "recovery complete");
5081 err = ext4_mark_recovery_complete(sb, es);
5085 if (EXT4_SB(sb)->s_journal) {
5086 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
5087 descr = " journalled data mode";
5088 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
5089 descr = " ordered data mode";
5091 descr = " writeback data mode";
5093 descr = "out journal";
5095 if (test_opt(sb, DISCARD)) {
5096 struct request_queue *q = bdev_get_queue(sb->s_bdev);
5097 if (!blk_queue_discard(q))
5098 ext4_msg(sb, KERN_WARNING,
5099 "mounting with \"discard\" option, but "
5100 "the device does not support discard");
5103 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
5104 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5105 "Opts: %.*s%s%s", descr,
5106 (int) sizeof(sbi->s_es->s_mount_opts),
5107 sbi->s_es->s_mount_opts,
5108 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
5110 if (es->s_error_count)
5111 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
5113 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5114 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
5115 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
5116 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
5117 atomic_set(&sbi->s_warning_count, 0);
5118 atomic_set(&sbi->s_msg_count, 0);
5125 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5129 ext4_unregister_sysfs(sb);
5130 kobject_put(&sbi->s_kobj);
5132 ext4_unregister_li_request(sb);
5134 ext4_mb_release(sb);
5136 flex_groups = rcu_dereference(sbi->s_flex_groups);
5138 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
5139 kvfree(flex_groups[i]);
5140 kvfree(flex_groups);
5143 percpu_counter_destroy(&sbi->s_freeclusters_counter);
5144 percpu_counter_destroy(&sbi->s_freeinodes_counter);
5145 percpu_counter_destroy(&sbi->s_dirs_counter);
5146 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
5147 percpu_free_rwsem(&sbi->s_writepages_rwsem);
5149 ext4_ext_release(sb);
5150 ext4_release_system_zone(sb);
5155 ext4_msg(sb, KERN_ERR, "mount failed");
5156 if (EXT4_SB(sb)->rsv_conversion_wq)
5157 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
5159 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5160 sbi->s_ea_inode_cache = NULL;
5162 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5163 sbi->s_ea_block_cache = NULL;
5165 if (sbi->s_journal) {
5166 jbd2_journal_destroy(sbi->s_journal);
5167 sbi->s_journal = NULL;
5170 ext4_es_unregister_shrinker(sbi);
5172 del_timer_sync(&sbi->s_err_report);
5174 kthread_stop(sbi->s_mmp_tsk);
5177 group_desc = rcu_dereference(sbi->s_group_desc);
5178 for (i = 0; i < db_count; i++)
5179 brelse(group_desc[i]);
5183 if (sbi->s_chksum_driver)
5184 crypto_free_shash(sbi->s_chksum_driver);
5186 #ifdef CONFIG_UNICODE
5187 utf8_unload(sbi->s_encoding);
5191 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5192 kfree(get_qf_name(sb, sbi, i));
5194 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
5195 ext4_blkdev_remove(sbi);
5198 sb->s_fs_info = NULL;
5199 kfree(sbi->s_blockgroup_lock);
5203 fs_put_dax(dax_dev);
5204 return err ? err : ret;
5208 * Setup any per-fs journal parameters now. We'll do this both on
5209 * initial mount, once the journal has been initialised but before we've
5210 * done any recovery; and again on any subsequent remount.
5212 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
5214 struct ext4_sb_info *sbi = EXT4_SB(sb);
5216 journal->j_commit_interval = sbi->s_commit_interval;
5217 journal->j_min_batch_time = sbi->s_min_batch_time;
5218 journal->j_max_batch_time = sbi->s_max_batch_time;
5219 ext4_fc_init(sb, journal);
5221 write_lock(&journal->j_state_lock);
5222 if (test_opt(sb, BARRIER))
5223 journal->j_flags |= JBD2_BARRIER;
5225 journal->j_flags &= ~JBD2_BARRIER;
5226 if (test_opt(sb, DATA_ERR_ABORT))
5227 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5229 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
5230 write_unlock(&journal->j_state_lock);
5233 static struct inode *ext4_get_journal_inode(struct super_block *sb,
5234 unsigned int journal_inum)
5236 struct inode *journal_inode;
5239 * Test for the existence of a valid inode on disk. Bad things
5240 * happen if we iget() an unused inode, as the subsequent iput()
5241 * will try to delete it.
5243 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
5244 if (IS_ERR(journal_inode)) {
5245 ext4_msg(sb, KERN_ERR, "no journal found");
5248 if (!journal_inode->i_nlink) {
5249 make_bad_inode(journal_inode);
5250 iput(journal_inode);
5251 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
5255 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
5256 journal_inode, journal_inode->i_size);
5257 if (!S_ISREG(journal_inode->i_mode)) {
5258 ext4_msg(sb, KERN_ERR, "invalid journal inode");
5259 iput(journal_inode);
5262 return journal_inode;
5265 static journal_t *ext4_get_journal(struct super_block *sb,
5266 unsigned int journal_inum)
5268 struct inode *journal_inode;
5271 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5274 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5278 journal = jbd2_journal_init_inode(journal_inode);
5280 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5281 iput(journal_inode);
5284 journal->j_private = sb;
5285 ext4_init_journal_params(sb, journal);
5289 static journal_t *ext4_get_dev_journal(struct super_block *sb,
5292 struct buffer_head *bh;
5296 int hblock, blocksize;
5297 ext4_fsblk_t sb_block;
5298 unsigned long offset;
5299 struct ext4_super_block *es;
5300 struct block_device *bdev;
5302 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5305 bdev = ext4_blkdev_get(j_dev, sb);
5309 blocksize = sb->s_blocksize;
5310 hblock = bdev_logical_block_size(bdev);
5311 if (blocksize < hblock) {
5312 ext4_msg(sb, KERN_ERR,
5313 "blocksize too small for journal device");
5317 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5318 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5319 set_blocksize(bdev, blocksize);
5320 if (!(bh = __bread(bdev, sb_block, blocksize))) {
5321 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5322 "external journal");
5326 es = (struct ext4_super_block *) (bh->b_data + offset);
5327 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5328 !(le32_to_cpu(es->s_feature_incompat) &
5329 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5330 ext4_msg(sb, KERN_ERR, "external journal has "
5336 if ((le32_to_cpu(es->s_feature_ro_compat) &
5337 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5338 es->s_checksum != ext4_superblock_csum(sb, es)) {
5339 ext4_msg(sb, KERN_ERR, "external journal has "
5340 "corrupt superblock");
5345 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5346 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5351 len = ext4_blocks_count(es);
5352 start = sb_block + 1;
5353 brelse(bh); /* we're done with the superblock */
5355 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
5356 start, len, blocksize);
5358 ext4_msg(sb, KERN_ERR, "failed to create device journal");
5361 journal->j_private = sb;
5362 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) {
5363 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
5366 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5367 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5368 "user (unsupported) - %d",
5369 be32_to_cpu(journal->j_superblock->s_nr_users));
5372 EXT4_SB(sb)->s_journal_bdev = bdev;
5373 ext4_init_journal_params(sb, journal);
5377 jbd2_journal_destroy(journal);
5379 ext4_blkdev_put(bdev);
5383 static int ext4_load_journal(struct super_block *sb,
5384 struct ext4_super_block *es,
5385 unsigned long journal_devnum)
5388 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5391 int really_read_only;
5394 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5395 return -EFSCORRUPTED;
5397 if (journal_devnum &&
5398 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5399 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5400 "numbers have changed");
5401 journal_dev = new_decode_dev(journal_devnum);
5403 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5405 if (journal_inum && journal_dev) {
5406 ext4_msg(sb, KERN_ERR,
5407 "filesystem has both journal inode and journal device!");
5412 journal = ext4_get_journal(sb, journal_inum);
5416 journal = ext4_get_dev_journal(sb, journal_dev);
5421 journal_dev_ro = bdev_read_only(journal->j_dev);
5422 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5424 if (journal_dev_ro && !sb_rdonly(sb)) {
5425 ext4_msg(sb, KERN_ERR,
5426 "journal device read-only, try mounting with '-o ro'");
5432 * Are we loading a blank journal or performing recovery after a
5433 * crash? For recovery, we need to check in advance whether we
5434 * can get read-write access to the device.
5436 if (ext4_has_feature_journal_needs_recovery(sb)) {
5437 if (sb_rdonly(sb)) {
5438 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5439 "required on readonly filesystem");
5440 if (really_read_only) {
5441 ext4_msg(sb, KERN_ERR, "write access "
5442 "unavailable, cannot proceed "
5443 "(try mounting with noload)");
5447 ext4_msg(sb, KERN_INFO, "write access will "
5448 "be enabled during recovery");
5452 if (!(journal->j_flags & JBD2_BARRIER))
5453 ext4_msg(sb, KERN_INFO, "barriers disabled");
5455 if (!ext4_has_feature_journal_needs_recovery(sb))
5456 err = jbd2_journal_wipe(journal, !really_read_only);
5458 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5460 memcpy(save, ((char *) es) +
5461 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
5462 err = jbd2_journal_load(journal);
5464 memcpy(((char *) es) + EXT4_S_ERR_START,
5465 save, EXT4_S_ERR_LEN);
5470 ext4_msg(sb, KERN_ERR, "error loading journal");
5474 EXT4_SB(sb)->s_journal = journal;
5475 err = ext4_clear_journal_err(sb, es);
5477 EXT4_SB(sb)->s_journal = NULL;
5478 jbd2_journal_destroy(journal);
5482 if (!really_read_only && journal_devnum &&
5483 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5484 es->s_journal_dev = cpu_to_le32(journal_devnum);
5486 /* Make sure we flush the recovery flag to disk. */
5487 ext4_commit_super(sb, 1);
5493 jbd2_journal_destroy(journal);
5497 static int ext4_commit_super(struct super_block *sb, int sync)
5499 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
5500 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
5503 if (!sbh || block_device_ejected(sb))
5507 * If the file system is mounted read-only, don't update the
5508 * superblock write time. This avoids updating the superblock
5509 * write time when we are mounting the root file system
5510 * read/only but we need to replay the journal; at that point,
5511 * for people who are east of GMT and who make their clock
5512 * tick in localtime for Windows bug-for-bug compatibility,
5513 * the clock is set in the future, and this will cause e2fsck
5514 * to complain and force a full file system check.
5516 if (!(sb->s_flags & SB_RDONLY))
5517 ext4_update_tstamp(es, s_wtime);
5518 if (sb->s_bdev->bd_part)
5519 es->s_kbytes_written =
5520 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
5521 ((part_stat_read(sb->s_bdev->bd_part,
5522 sectors[STAT_WRITE]) -
5523 EXT4_SB(sb)->s_sectors_written_start) >> 1));
5525 es->s_kbytes_written =
5526 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
5527 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
5528 ext4_free_blocks_count_set(es,
5529 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
5530 &EXT4_SB(sb)->s_freeclusters_counter)));
5531 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
5532 es->s_free_inodes_count =
5533 cpu_to_le32(percpu_counter_sum_positive(
5534 &EXT4_SB(sb)->s_freeinodes_counter));
5535 BUFFER_TRACE(sbh, "marking dirty");
5536 ext4_superblock_csum_set(sb);
5539 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
5541 * Oh, dear. A previous attempt to write the
5542 * superblock failed. This could happen because the
5543 * USB device was yanked out. Or it could happen to
5544 * be a transient write error and maybe the block will
5545 * be remapped. Nothing we can do but to retry the
5546 * write and hope for the best.
5548 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5549 "superblock detected");
5550 clear_buffer_write_io_error(sbh);
5551 set_buffer_uptodate(sbh);
5553 mark_buffer_dirty(sbh);
5556 error = __sync_dirty_buffer(sbh,
5557 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
5558 if (buffer_write_io_error(sbh)) {
5559 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5561 clear_buffer_write_io_error(sbh);
5562 set_buffer_uptodate(sbh);
5569 * Have we just finished recovery? If so, and if we are mounting (or
5570 * remounting) the filesystem readonly, then we will end up with a
5571 * consistent fs on disk. Record that fact.
5573 static int ext4_mark_recovery_complete(struct super_block *sb,
5574 struct ext4_super_block *es)
5577 journal_t *journal = EXT4_SB(sb)->s_journal;
5579 if (!ext4_has_feature_journal(sb)) {
5580 if (journal != NULL) {
5581 ext4_error(sb, "Journal got removed while the fs was "
5583 return -EFSCORRUPTED;
5587 jbd2_journal_lock_updates(journal);
5588 err = jbd2_journal_flush(journal);
5592 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
5593 ext4_clear_feature_journal_needs_recovery(sb);
5594 ext4_commit_super(sb, 1);
5597 jbd2_journal_unlock_updates(journal);
5602 * If we are mounting (or read-write remounting) a filesystem whose journal
5603 * has recorded an error from a previous lifetime, move that error to the
5604 * main filesystem now.
5606 static int ext4_clear_journal_err(struct super_block *sb,
5607 struct ext4_super_block *es)
5613 if (!ext4_has_feature_journal(sb)) {
5614 ext4_error(sb, "Journal got removed while the fs was mounted!");
5615 return -EFSCORRUPTED;
5618 journal = EXT4_SB(sb)->s_journal;
5621 * Now check for any error status which may have been recorded in the
5622 * journal by a prior ext4_error() or ext4_abort()
5625 j_errno = jbd2_journal_errno(journal);
5629 errstr = ext4_decode_error(sb, j_errno, nbuf);
5630 ext4_warning(sb, "Filesystem error recorded "
5631 "from previous mount: %s", errstr);
5632 ext4_warning(sb, "Marking fs in need of filesystem check.");
5634 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5635 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5636 ext4_commit_super(sb, 1);
5638 jbd2_journal_clear_err(journal);
5639 jbd2_journal_update_sb_errno(journal);
5645 * Force the running and committing transactions to commit,
5646 * and wait on the commit.
5648 int ext4_force_commit(struct super_block *sb)
5655 journal = EXT4_SB(sb)->s_journal;
5656 return ext4_journal_force_commit(journal);
5659 static int ext4_sync_fs(struct super_block *sb, int wait)
5663 bool needs_barrier = false;
5664 struct ext4_sb_info *sbi = EXT4_SB(sb);
5666 if (unlikely(ext4_forced_shutdown(sbi)))
5669 trace_ext4_sync_fs(sb, wait);
5670 flush_workqueue(sbi->rsv_conversion_wq);
5672 * Writeback quota in non-journalled quota case - journalled quota has
5675 dquot_writeback_dquots(sb, -1);
5677 * Data writeback is possible w/o journal transaction, so barrier must
5678 * being sent at the end of the function. But we can skip it if
5679 * transaction_commit will do it for us.
5681 if (sbi->s_journal) {
5682 target = jbd2_get_latest_transaction(sbi->s_journal);
5683 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5684 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5685 needs_barrier = true;
5687 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5689 ret = jbd2_log_wait_commit(sbi->s_journal,
5692 } else if (wait && test_opt(sb, BARRIER))
5693 needs_barrier = true;
5694 if (needs_barrier) {
5696 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL);
5705 * LVM calls this function before a (read-only) snapshot is created. This
5706 * gives us a chance to flush the journal completely and mark the fs clean.
5708 * Note that only this function cannot bring a filesystem to be in a clean
5709 * state independently. It relies on upper layer to stop all data & metadata
5712 static int ext4_freeze(struct super_block *sb)
5720 journal = EXT4_SB(sb)->s_journal;
5723 /* Now we set up the journal barrier. */
5724 jbd2_journal_lock_updates(journal);
5727 * Don't clear the needs_recovery flag if we failed to
5728 * flush the journal.
5730 error = jbd2_journal_flush(journal);
5734 /* Journal blocked and flushed, clear needs_recovery flag. */
5735 ext4_clear_feature_journal_needs_recovery(sb);
5738 error = ext4_commit_super(sb, 1);
5741 /* we rely on upper layer to stop further updates */
5742 jbd2_journal_unlock_updates(journal);
5747 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5748 * flag here, even though the filesystem is not technically dirty yet.
5750 static int ext4_unfreeze(struct super_block *sb)
5752 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
5755 if (EXT4_SB(sb)->s_journal) {
5756 /* Reset the needs_recovery flag before the fs is unlocked. */
5757 ext4_set_feature_journal_needs_recovery(sb);
5760 ext4_commit_super(sb, 1);
5765 * Structure to save mount options for ext4_remount's benefit
5767 struct ext4_mount_options {
5768 unsigned long s_mount_opt;
5769 unsigned long s_mount_opt2;
5772 unsigned long s_commit_interval;
5773 u32 s_min_batch_time, s_max_batch_time;
5776 char *s_qf_names[EXT4_MAXQUOTAS];
5780 static int ext4_remount(struct super_block *sb, int *flags, char *data)
5782 struct ext4_super_block *es;
5783 struct ext4_sb_info *sbi = EXT4_SB(sb);
5784 unsigned long old_sb_flags, vfs_flags;
5785 struct ext4_mount_options old_opts;
5786 int enable_quota = 0;
5788 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5792 char *to_free[EXT4_MAXQUOTAS];
5794 char *orig_data = kstrdup(data, GFP_KERNEL);
5796 if (data && !orig_data)
5799 /* Store the original options */
5800 old_sb_flags = sb->s_flags;
5801 old_opts.s_mount_opt = sbi->s_mount_opt;
5802 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
5803 old_opts.s_resuid = sbi->s_resuid;
5804 old_opts.s_resgid = sbi->s_resgid;
5805 old_opts.s_commit_interval = sbi->s_commit_interval;
5806 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5807 old_opts.s_max_batch_time = sbi->s_max_batch_time;
5809 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
5810 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5811 if (sbi->s_qf_names[i]) {
5812 char *qf_name = get_qf_name(sb, sbi, i);
5814 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
5815 if (!old_opts.s_qf_names[i]) {
5816 for (j = 0; j < i; j++)
5817 kfree(old_opts.s_qf_names[j]);
5822 old_opts.s_qf_names[i] = NULL;
5824 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5825 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
5828 * Some options can be enabled by ext4 and/or by VFS mount flag
5829 * either way we need to make sure it matches in both *flags and
5830 * s_flags. Copy those selected flags from *flags to s_flags
5832 vfs_flags = SB_LAZYTIME | SB_I_VERSION;
5833 sb->s_flags = (sb->s_flags & ~vfs_flags) | (*flags & vfs_flags);
5835 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
5840 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5841 test_opt(sb, JOURNAL_CHECKSUM)) {
5842 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5843 "during remount not supported; ignoring");
5844 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5847 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5848 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5849 ext4_msg(sb, KERN_ERR, "can't mount with "
5850 "both data=journal and delalloc");
5854 if (test_opt(sb, DIOREAD_NOLOCK)) {
5855 ext4_msg(sb, KERN_ERR, "can't mount with "
5856 "both data=journal and dioread_nolock");
5860 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5861 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5862 ext4_msg(sb, KERN_ERR, "can't mount with "
5863 "journal_async_commit in data=ordered mode");
5869 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5870 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5875 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
5876 ext4_abort(sb, EXT4_ERR_ESHUTDOWN, "Abort forced by user");
5878 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5879 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5883 if (sbi->s_journal) {
5884 ext4_init_journal_params(sb, sbi->s_journal);
5885 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5888 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
5889 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
5894 if (*flags & SB_RDONLY) {
5895 err = sync_filesystem(sb);
5898 err = dquot_suspend(sb, -1);
5903 * First of all, the unconditional stuff we have to do
5904 * to disable replay of the journal when we next remount
5906 sb->s_flags |= SB_RDONLY;
5909 * OK, test if we are remounting a valid rw partition
5910 * readonly, and if so set the rdonly flag and then
5911 * mark the partition as valid again.
5913 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5914 (sbi->s_mount_state & EXT4_VALID_FS))
5915 es->s_state = cpu_to_le16(sbi->s_mount_state);
5917 if (sbi->s_journal) {
5919 * We let remount-ro finish even if marking fs
5920 * as clean failed...
5922 ext4_mark_recovery_complete(sb, es);
5925 kthread_stop(sbi->s_mmp_tsk);
5927 /* Make sure we can mount this feature set readwrite */
5928 if (ext4_has_feature_readonly(sb) ||
5929 !ext4_feature_set_ok(sb, 0)) {
5934 * Make sure the group descriptor checksums
5935 * are sane. If they aren't, refuse to remount r/w.
5937 for (g = 0; g < sbi->s_groups_count; g++) {
5938 struct ext4_group_desc *gdp =
5939 ext4_get_group_desc(sb, g, NULL);
5941 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5942 ext4_msg(sb, KERN_ERR,
5943 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5944 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5945 le16_to_cpu(gdp->bg_checksum));
5952 * If we have an unprocessed orphan list hanging
5953 * around from a previously readonly bdev mount,
5954 * require a full umount/remount for now.
5956 if (es->s_last_orphan) {
5957 ext4_msg(sb, KERN_WARNING, "Couldn't "
5958 "remount RDWR because of unprocessed "
5959 "orphan inode list. Please "
5960 "umount/remount instead");
5966 * Mounting a RDONLY partition read-write, so reread
5967 * and store the current valid flag. (It may have
5968 * been changed by e2fsck since we originally mounted
5971 if (sbi->s_journal) {
5972 err = ext4_clear_journal_err(sb, es);
5976 sbi->s_mount_state = le16_to_cpu(es->s_state);
5978 err = ext4_setup_super(sb, es, 0);
5982 sb->s_flags &= ~SB_RDONLY;
5983 if (ext4_has_feature_mmp(sb))
5984 if (ext4_multi_mount_protect(sb,
5985 le64_to_cpu(es->s_mmp_block))) {
5994 * Reinitialize lazy itable initialization thread based on
5997 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
5998 ext4_unregister_li_request(sb);
6000 ext4_group_t first_not_zeroed;
6001 first_not_zeroed = ext4_has_uninit_itable(sb);
6002 ext4_register_li_request(sb, first_not_zeroed);
6006 * Handle creation of system zone data early because it can fail.
6007 * Releasing of existing data is done when we are sure remount will
6010 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
6011 err = ext4_setup_system_zone(sb);
6016 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
6017 err = ext4_commit_super(sb, 1);
6023 /* Release old quota file names */
6024 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6025 kfree(old_opts.s_qf_names[i]);
6027 if (sb_any_quota_suspended(sb))
6028 dquot_resume(sb, -1);
6029 else if (ext4_has_feature_quota(sb)) {
6030 err = ext4_enable_quotas(sb);
6036 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6037 ext4_release_system_zone(sb);
6040 * Some options can be enabled by ext4 and/or by VFS mount flag
6041 * either way we need to make sure it matches in both *flags and
6042 * s_flags. Copy those selected flags from s_flags to *flags
6044 *flags = (*flags & ~vfs_flags) | (sb->s_flags & vfs_flags);
6046 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
6051 sb->s_flags = old_sb_flags;
6052 sbi->s_mount_opt = old_opts.s_mount_opt;
6053 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
6054 sbi->s_resuid = old_opts.s_resuid;
6055 sbi->s_resgid = old_opts.s_resgid;
6056 sbi->s_commit_interval = old_opts.s_commit_interval;
6057 sbi->s_min_batch_time = old_opts.s_min_batch_time;
6058 sbi->s_max_batch_time = old_opts.s_max_batch_time;
6059 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6060 ext4_release_system_zone(sb);
6062 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
6063 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
6064 to_free[i] = get_qf_name(sb, sbi, i);
6065 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
6068 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6076 static int ext4_statfs_project(struct super_block *sb,
6077 kprojid_t projid, struct kstatfs *buf)
6080 struct dquot *dquot;
6084 qid = make_kqid_projid(projid);
6085 dquot = dqget(sb, qid);
6087 return PTR_ERR(dquot);
6088 spin_lock(&dquot->dq_dqb_lock);
6090 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6091 dquot->dq_dqb.dqb_bhardlimit);
6092 limit >>= sb->s_blocksize_bits;
6094 if (limit && buf->f_blocks > limit) {
6095 curblock = (dquot->dq_dqb.dqb_curspace +
6096 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
6097 buf->f_blocks = limit;
6098 buf->f_bfree = buf->f_bavail =
6099 (buf->f_blocks > curblock) ?
6100 (buf->f_blocks - curblock) : 0;
6103 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6104 dquot->dq_dqb.dqb_ihardlimit);
6105 if (limit && buf->f_files > limit) {
6106 buf->f_files = limit;
6108 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6109 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6112 spin_unlock(&dquot->dq_dqb_lock);
6118 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
6120 struct super_block *sb = dentry->d_sb;
6121 struct ext4_sb_info *sbi = EXT4_SB(sb);
6122 struct ext4_super_block *es = sbi->s_es;
6123 ext4_fsblk_t overhead = 0, resv_blocks;
6126 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
6128 if (!test_opt(sb, MINIX_DF))
6129 overhead = sbi->s_overhead;
6131 buf->f_type = EXT4_SUPER_MAGIC;
6132 buf->f_bsize = sb->s_blocksize;
6133 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
6134 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6135 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
6136 /* prevent underflow in case that few free space is available */
6137 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
6138 buf->f_bavail = buf->f_bfree -
6139 (ext4_r_blocks_count(es) + resv_blocks);
6140 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
6142 buf->f_files = le32_to_cpu(es->s_inodes_count);
6143 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
6144 buf->f_namelen = EXT4_NAME_LEN;
6145 fsid = le64_to_cpup((void *)es->s_uuid) ^
6146 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
6147 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
6148 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
6151 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6152 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6153 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6162 * Helper functions so that transaction is started before we acquire dqio_sem
6163 * to keep correct lock ordering of transaction > dqio_sem
6165 static inline struct inode *dquot_to_inode(struct dquot *dquot)
6167 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
6170 static int ext4_write_dquot(struct dquot *dquot)
6174 struct inode *inode;
6176 inode = dquot_to_inode(dquot);
6177 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
6178 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
6180 return PTR_ERR(handle);
6181 ret = dquot_commit(dquot);
6182 err = ext4_journal_stop(handle);
6188 static int ext4_acquire_dquot(struct dquot *dquot)
6193 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6194 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
6196 return PTR_ERR(handle);
6197 ret = dquot_acquire(dquot);
6198 err = ext4_journal_stop(handle);
6204 static int ext4_release_dquot(struct dquot *dquot)
6209 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6210 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
6211 if (IS_ERR(handle)) {
6212 /* Release dquot anyway to avoid endless cycle in dqput() */
6213 dquot_release(dquot);
6214 return PTR_ERR(handle);
6216 ret = dquot_release(dquot);
6217 err = ext4_journal_stop(handle);
6223 static int ext4_mark_dquot_dirty(struct dquot *dquot)
6225 struct super_block *sb = dquot->dq_sb;
6226 struct ext4_sb_info *sbi = EXT4_SB(sb);
6228 /* Are we journaling quotas? */
6229 if (ext4_has_feature_quota(sb) ||
6230 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
6231 dquot_mark_dquot_dirty(dquot);
6232 return ext4_write_dquot(dquot);
6234 return dquot_mark_dquot_dirty(dquot);
6238 static int ext4_write_info(struct super_block *sb, int type)
6243 /* Data block + inode block */
6244 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
6246 return PTR_ERR(handle);
6247 ret = dquot_commit_info(sb, type);
6248 err = ext4_journal_stop(handle);
6255 * Turn on quotas during mount time - we need to find
6256 * the quota file and such...
6258 static int ext4_quota_on_mount(struct super_block *sb, int type)
6260 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
6261 EXT4_SB(sb)->s_jquota_fmt, type);
6264 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6266 struct ext4_inode_info *ei = EXT4_I(inode);
6268 /* The first argument of lockdep_set_subclass has to be
6269 * *exactly* the same as the argument to init_rwsem() --- in
6270 * this case, in init_once() --- or lockdep gets unhappy
6271 * because the name of the lock is set using the
6272 * stringification of the argument to init_rwsem().
6274 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6275 lockdep_set_subclass(&ei->i_data_sem, subclass);
6279 * Standard function to be called on quota_on
6281 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6282 const struct path *path)
6286 if (!test_opt(sb, QUOTA))
6289 /* Quotafile not on the same filesystem? */
6290 if (path->dentry->d_sb != sb)
6293 /* Quota already enabled for this file? */
6294 if (IS_NOQUOTA(d_inode(path->dentry)))
6297 /* Journaling quota? */
6298 if (EXT4_SB(sb)->s_qf_names[type]) {
6299 /* Quotafile not in fs root? */
6300 if (path->dentry->d_parent != sb->s_root)
6301 ext4_msg(sb, KERN_WARNING,
6302 "Quota file not on filesystem root. "
6303 "Journaled quota will not work");
6304 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6307 * Clear the flag just in case mount options changed since
6310 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
6314 * When we journal data on quota file, we have to flush journal to see
6315 * all updates to the file when we bypass pagecache...
6317 if (EXT4_SB(sb)->s_journal &&
6318 ext4_should_journal_data(d_inode(path->dentry))) {
6320 * We don't need to lock updates but journal_flush() could
6321 * otherwise be livelocked...
6323 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
6324 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
6325 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
6330 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6331 err = dquot_quota_on(sb, type, format_id, path);
6333 lockdep_set_quota_inode(path->dentry->d_inode,
6336 struct inode *inode = d_inode(path->dentry);
6340 * Set inode flags to prevent userspace from messing with quota
6341 * files. If this fails, we return success anyway since quotas
6342 * are already enabled and this is not a hard failure.
6345 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6348 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6349 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6350 S_NOATIME | S_IMMUTABLE);
6351 err = ext4_mark_inode_dirty(handle, inode);
6352 ext4_journal_stop(handle);
6354 inode_unlock(inode);
6359 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6363 struct inode *qf_inode;
6364 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6365 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6366 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6367 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6370 BUG_ON(!ext4_has_feature_quota(sb));
6372 if (!qf_inums[type])
6375 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
6376 if (IS_ERR(qf_inode)) {
6377 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
6378 return PTR_ERR(qf_inode);
6381 /* Don't account quota for quota files to avoid recursion */
6382 qf_inode->i_flags |= S_NOQUOTA;
6383 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
6384 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
6386 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
6392 /* Enable usage tracking for all quota types. */
6393 static int ext4_enable_quotas(struct super_block *sb)
6396 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6397 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6398 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6399 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6401 bool quota_mopt[EXT4_MAXQUOTAS] = {
6402 test_opt(sb, USRQUOTA),
6403 test_opt(sb, GRPQUOTA),
6404 test_opt(sb, PRJQUOTA),
6407 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
6408 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
6409 if (qf_inums[type]) {
6410 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
6411 DQUOT_USAGE_ENABLED |
6412 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
6415 "Failed to enable quota tracking "
6416 "(type=%d, err=%d). Please run "
6417 "e2fsck to fix.", type, err);
6418 for (type--; type >= 0; type--)
6419 dquot_quota_off(sb, type);
6428 static int ext4_quota_off(struct super_block *sb, int type)
6430 struct inode *inode = sb_dqopt(sb)->files[type];
6434 /* Force all delayed allocation blocks to be allocated.
6435 * Caller already holds s_umount sem */
6436 if (test_opt(sb, DELALLOC))
6437 sync_filesystem(sb);
6439 if (!inode || !igrab(inode))
6442 err = dquot_quota_off(sb, type);
6443 if (err || ext4_has_feature_quota(sb))
6448 * Update modification times of quota files when userspace can
6449 * start looking at them. If we fail, we return success anyway since
6450 * this is not a hard failure and quotas are already disabled.
6452 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6453 if (IS_ERR(handle)) {
6454 err = PTR_ERR(handle);
6457 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
6458 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
6459 inode->i_mtime = inode->i_ctime = current_time(inode);
6460 err = ext4_mark_inode_dirty(handle, inode);
6461 ext4_journal_stop(handle);
6463 inode_unlock(inode);
6465 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
6469 return dquot_quota_off(sb, type);
6472 /* Read data from quotafile - avoid pagecache and such because we cannot afford
6473 * acquiring the locks... As quota files are never truncated and quota code
6474 * itself serializes the operations (and no one else should touch the files)
6475 * we don't have to be afraid of races */
6476 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
6477 size_t len, loff_t off)
6479 struct inode *inode = sb_dqopt(sb)->files[type];
6480 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6481 int offset = off & (sb->s_blocksize - 1);
6484 struct buffer_head *bh;
6485 loff_t i_size = i_size_read(inode);
6489 if (off+len > i_size)
6492 while (toread > 0) {
6493 tocopy = sb->s_blocksize - offset < toread ?
6494 sb->s_blocksize - offset : toread;
6495 bh = ext4_bread(NULL, inode, blk, 0);
6498 if (!bh) /* A hole? */
6499 memset(data, 0, tocopy);
6501 memcpy(data, bh->b_data+offset, tocopy);
6511 /* Write to quotafile (we know the transaction is already started and has
6512 * enough credits) */
6513 static ssize_t ext4_quota_write(struct super_block *sb, int type,
6514 const char *data, size_t len, loff_t off)
6516 struct inode *inode = sb_dqopt(sb)->files[type];
6517 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6518 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
6520 struct buffer_head *bh;
6521 handle_t *handle = journal_current_handle();
6523 if (EXT4_SB(sb)->s_journal && !handle) {
6524 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6525 " cancelled because transaction is not started",
6526 (unsigned long long)off, (unsigned long long)len);
6530 * Since we account only one data block in transaction credits,
6531 * then it is impossible to cross a block boundary.
6533 if (sb->s_blocksize - offset < len) {
6534 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6535 " cancelled because not block aligned",
6536 (unsigned long long)off, (unsigned long long)len);
6541 bh = ext4_bread(handle, inode, blk,
6542 EXT4_GET_BLOCKS_CREATE |
6543 EXT4_GET_BLOCKS_METADATA_NOFAIL);
6544 } while (PTR_ERR(bh) == -ENOSPC &&
6545 ext4_should_retry_alloc(inode->i_sb, &retries));
6550 BUFFER_TRACE(bh, "get write access");
6551 err = ext4_journal_get_write_access(handle, bh);
6557 memcpy(bh->b_data+offset, data, len);
6558 flush_dcache_page(bh->b_page);
6560 err = ext4_handle_dirty_metadata(handle, NULL, bh);
6563 if (inode->i_size < off + len) {
6564 ext4_fc_track_range(inode,
6565 (inode->i_size > 0 ? inode->i_size - 1 : 0)
6566 >> inode->i_sb->s_blocksize_bits,
6567 (off + len) >> inode->i_sb->s_blocksize_bits);
6568 i_size_write(inode, off + len);
6569 EXT4_I(inode)->i_disksize = inode->i_size;
6570 err2 = ext4_mark_inode_dirty(handle, inode);
6571 if (unlikely(err2 && !err))
6574 return err ? err : len;
6578 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6579 const char *dev_name, void *data)
6581 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
6584 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
6585 static inline void register_as_ext2(void)
6587 int err = register_filesystem(&ext2_fs_type);
6590 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6593 static inline void unregister_as_ext2(void)
6595 unregister_filesystem(&ext2_fs_type);
6598 static inline int ext2_feature_set_ok(struct super_block *sb)
6600 if (ext4_has_unknown_ext2_incompat_features(sb))
6604 if (ext4_has_unknown_ext2_ro_compat_features(sb))
6609 static inline void register_as_ext2(void) { }
6610 static inline void unregister_as_ext2(void) { }
6611 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
6614 static inline void register_as_ext3(void)
6616 int err = register_filesystem(&ext3_fs_type);
6619 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6622 static inline void unregister_as_ext3(void)
6624 unregister_filesystem(&ext3_fs_type);
6627 static inline int ext3_feature_set_ok(struct super_block *sb)
6629 if (ext4_has_unknown_ext3_incompat_features(sb))
6631 if (!ext4_has_feature_journal(sb))
6635 if (ext4_has_unknown_ext3_ro_compat_features(sb))
6640 static struct file_system_type ext4_fs_type = {
6641 .owner = THIS_MODULE,
6643 .mount = ext4_mount,
6644 .kill_sb = kill_block_super,
6645 .fs_flags = FS_REQUIRES_DEV,
6647 MODULE_ALIAS_FS("ext4");
6649 /* Shared across all ext4 file systems */
6650 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
6652 static int __init ext4_init_fs(void)
6656 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
6657 ext4_li_info = NULL;
6658 mutex_init(&ext4_li_mtx);
6660 /* Build-time check for flags consistency */
6661 ext4_check_flag_values();
6663 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
6664 init_waitqueue_head(&ext4__ioend_wq[i]);
6666 err = ext4_init_es();
6670 err = ext4_init_pending();
6674 err = ext4_init_post_read_processing();
6678 err = ext4_init_pageio();
6682 err = ext4_init_system_zone();
6686 err = ext4_init_sysfs();
6690 err = ext4_init_mballoc();
6693 err = init_inodecache();
6697 err = ext4_fc_init_dentry_cache();
6703 err = register_filesystem(&ext4_fs_type);
6709 unregister_as_ext2();
6710 unregister_as_ext3();
6712 destroy_inodecache();
6714 ext4_exit_mballoc();
6718 ext4_exit_system_zone();
6722 ext4_exit_post_read_processing();
6724 ext4_exit_pending();
6731 static void __exit ext4_exit_fs(void)
6733 ext4_destroy_lazyinit_thread();
6734 unregister_as_ext2();
6735 unregister_as_ext3();
6736 unregister_filesystem(&ext4_fs_type);
6737 destroy_inodecache();
6738 ext4_exit_mballoc();
6740 ext4_exit_system_zone();
6742 ext4_exit_post_read_processing();
6744 ext4_exit_pending();
6747 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6748 MODULE_DESCRIPTION("Fourth Extended Filesystem");
6749 MODULE_LICENSE("GPL");
6750 MODULE_SOFTDEP("pre: crc32c");
6751 module_init(ext4_init_fs)
6752 module_exit(ext4_exit_fs)