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 DEFINE_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 void ext4_update_super(struct super_block *sb);
69 static int ext4_commit_super(struct super_block *sb);
70 static int ext4_mark_recovery_complete(struct super_block *sb,
71 struct ext4_super_block *es);
72 static int ext4_clear_journal_err(struct super_block *sb,
73 struct ext4_super_block *es);
74 static int ext4_sync_fs(struct super_block *sb, int wait);
75 static int ext4_remount(struct super_block *sb, int *flags, char *data);
76 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
77 static int ext4_unfreeze(struct super_block *sb);
78 static int ext4_freeze(struct super_block *sb);
79 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
80 const char *dev_name, void *data);
81 static inline int ext2_feature_set_ok(struct super_block *sb);
82 static inline int ext3_feature_set_ok(struct super_block *sb);
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);
93 * mmap_lock -> sb_start_pagefault -> invalidate_lock (r) -> transaction start
94 * -> page lock -> i_data_sem (rw)
96 * buffered write path:
97 * sb_start_write -> i_mutex -> mmap_lock
98 * sb_start_write -> i_mutex -> transaction start -> page lock ->
102 * sb_start_write -> i_mutex -> invalidate_lock (w) -> i_mmap_rwsem (w) ->
104 * sb_start_write -> i_mutex -> invalidate_lock (w) -> transaction start ->
108 * sb_start_write -> i_mutex -> mmap_lock
109 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
112 * transaction start -> page lock(s) -> i_data_sem (rw)
115 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
116 static struct file_system_type ext2_fs_type = {
117 .owner = THIS_MODULE,
120 .kill_sb = kill_block_super,
121 .fs_flags = FS_REQUIRES_DEV,
123 MODULE_ALIAS_FS("ext2");
124 MODULE_ALIAS("ext2");
125 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
127 #define IS_EXT2_SB(sb) (0)
131 static struct file_system_type ext3_fs_type = {
132 .owner = THIS_MODULE,
135 .kill_sb = kill_block_super,
136 .fs_flags = FS_REQUIRES_DEV,
138 MODULE_ALIAS_FS("ext3");
139 MODULE_ALIAS("ext3");
140 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
143 static inline void __ext4_read_bh(struct buffer_head *bh, int op_flags,
147 * buffer's verified bit is no longer valid after reading from
148 * disk again due to write out error, clear it to make sure we
149 * recheck the buffer contents.
151 clear_buffer_verified(bh);
153 bh->b_end_io = end_io ? end_io : end_buffer_read_sync;
155 submit_bh(REQ_OP_READ, op_flags, bh);
158 void ext4_read_bh_nowait(struct buffer_head *bh, int op_flags,
161 BUG_ON(!buffer_locked(bh));
163 if (ext4_buffer_uptodate(bh)) {
167 __ext4_read_bh(bh, op_flags, end_io);
170 int ext4_read_bh(struct buffer_head *bh, int op_flags, bh_end_io_t *end_io)
172 BUG_ON(!buffer_locked(bh));
174 if (ext4_buffer_uptodate(bh)) {
179 __ext4_read_bh(bh, op_flags, end_io);
182 if (buffer_uptodate(bh))
187 int ext4_read_bh_lock(struct buffer_head *bh, int op_flags, bool wait)
191 ext4_read_bh_nowait(bh, op_flags, NULL);
194 return ext4_read_bh(bh, op_flags, NULL);
198 * This works like __bread_gfp() except it uses ERR_PTR for error
199 * returns. Currently with sb_bread it's impossible to distinguish
200 * between ENOMEM and EIO situations (since both result in a NULL
203 static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb,
204 sector_t block, int op_flags,
207 struct buffer_head *bh;
210 bh = sb_getblk_gfp(sb, block, gfp);
212 return ERR_PTR(-ENOMEM);
213 if (ext4_buffer_uptodate(bh))
216 ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true);
224 struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block,
227 return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE);
230 struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
233 return __ext4_sb_bread_gfp(sb, block, 0, 0);
236 void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block)
238 struct buffer_head *bh = sb_getblk_gfp(sb, block, 0);
241 if (trylock_buffer(bh))
242 ext4_read_bh_nowait(bh, REQ_RAHEAD, NULL);
247 static int ext4_verify_csum_type(struct super_block *sb,
248 struct ext4_super_block *es)
250 if (!ext4_has_feature_metadata_csum(sb))
253 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
256 static __le32 ext4_superblock_csum(struct super_block *sb,
257 struct ext4_super_block *es)
259 struct ext4_sb_info *sbi = EXT4_SB(sb);
260 int offset = offsetof(struct ext4_super_block, s_checksum);
263 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
265 return cpu_to_le32(csum);
268 static int ext4_superblock_csum_verify(struct super_block *sb,
269 struct ext4_super_block *es)
271 if (!ext4_has_metadata_csum(sb))
274 return es->s_checksum == ext4_superblock_csum(sb, es);
277 void ext4_superblock_csum_set(struct super_block *sb)
279 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
281 if (!ext4_has_metadata_csum(sb))
284 es->s_checksum = ext4_superblock_csum(sb, es);
287 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
288 struct ext4_group_desc *bg)
290 return le32_to_cpu(bg->bg_block_bitmap_lo) |
291 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
292 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
295 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
296 struct ext4_group_desc *bg)
298 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
299 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
300 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
303 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
304 struct ext4_group_desc *bg)
306 return le32_to_cpu(bg->bg_inode_table_lo) |
307 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
308 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
311 __u32 ext4_free_group_clusters(struct super_block *sb,
312 struct ext4_group_desc *bg)
314 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
315 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
316 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
319 __u32 ext4_free_inodes_count(struct super_block *sb,
320 struct ext4_group_desc *bg)
322 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
323 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
324 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
327 __u32 ext4_used_dirs_count(struct super_block *sb,
328 struct ext4_group_desc *bg)
330 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
331 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
332 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
335 __u32 ext4_itable_unused_count(struct super_block *sb,
336 struct ext4_group_desc *bg)
338 return le16_to_cpu(bg->bg_itable_unused_lo) |
339 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
340 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
343 void ext4_block_bitmap_set(struct super_block *sb,
344 struct ext4_group_desc *bg, ext4_fsblk_t blk)
346 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
347 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
348 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
351 void ext4_inode_bitmap_set(struct super_block *sb,
352 struct ext4_group_desc *bg, ext4_fsblk_t blk)
354 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
355 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
356 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
359 void ext4_inode_table_set(struct super_block *sb,
360 struct ext4_group_desc *bg, ext4_fsblk_t blk)
362 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
363 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
364 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
367 void ext4_free_group_clusters_set(struct super_block *sb,
368 struct ext4_group_desc *bg, __u32 count)
370 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
371 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
372 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
375 void ext4_free_inodes_set(struct super_block *sb,
376 struct ext4_group_desc *bg, __u32 count)
378 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
379 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
380 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
383 void ext4_used_dirs_set(struct super_block *sb,
384 struct ext4_group_desc *bg, __u32 count)
386 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
387 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
388 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
391 void ext4_itable_unused_set(struct super_block *sb,
392 struct ext4_group_desc *bg, __u32 count)
394 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
395 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
396 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
399 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now)
401 now = clamp_val(now, 0, (1ull << 40) - 1);
403 *lo = cpu_to_le32(lower_32_bits(now));
404 *hi = upper_32_bits(now);
407 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
409 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
411 #define ext4_update_tstamp(es, tstamp) \
412 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \
413 ktime_get_real_seconds())
414 #define ext4_get_tstamp(es, tstamp) \
415 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
418 * The del_gendisk() function uninitializes the disk-specific data
419 * structures, including the bdi structure, without telling anyone
420 * else. Once this happens, any attempt to call mark_buffer_dirty()
421 * (for example, by ext4_commit_super), will cause a kernel OOPS.
422 * This is a kludge to prevent these oops until we can put in a proper
423 * hook in del_gendisk() to inform the VFS and file system layers.
425 static int block_device_ejected(struct super_block *sb)
427 struct inode *bd_inode = sb->s_bdev->bd_inode;
428 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
430 return bdi->dev == NULL;
433 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
435 struct super_block *sb = journal->j_private;
436 struct ext4_sb_info *sbi = EXT4_SB(sb);
437 int error = is_journal_aborted(journal);
438 struct ext4_journal_cb_entry *jce;
440 BUG_ON(txn->t_state == T_FINISHED);
442 ext4_process_freed_data(sb, txn->t_tid);
444 spin_lock(&sbi->s_md_lock);
445 while (!list_empty(&txn->t_private_list)) {
446 jce = list_entry(txn->t_private_list.next,
447 struct ext4_journal_cb_entry, jce_list);
448 list_del_init(&jce->jce_list);
449 spin_unlock(&sbi->s_md_lock);
450 jce->jce_func(sb, jce, error);
451 spin_lock(&sbi->s_md_lock);
453 spin_unlock(&sbi->s_md_lock);
457 * This writepage callback for write_cache_pages()
458 * takes care of a few cases after page cleaning.
460 * write_cache_pages() already checks for dirty pages
461 * and calls clear_page_dirty_for_io(), which we want,
462 * to write protect the pages.
464 * However, we may have to redirty a page (see below.)
466 static int ext4_journalled_writepage_callback(struct page *page,
467 struct writeback_control *wbc,
470 transaction_t *transaction = (transaction_t *) data;
471 struct buffer_head *bh, *head;
472 struct journal_head *jh;
474 bh = head = page_buffers(page);
477 * We have to redirty a page in these cases:
478 * 1) If buffer is dirty, it means the page was dirty because it
479 * contains a buffer that needs checkpointing. So the dirty bit
480 * needs to be preserved so that checkpointing writes the buffer
482 * 2) If buffer is not part of the committing transaction
483 * (we may have just accidentally come across this buffer because
484 * inode range tracking is not exact) or if the currently running
485 * transaction already contains this buffer as well, dirty bit
486 * needs to be preserved so that the buffer gets writeprotected
487 * properly on running transaction's commit.
490 if (buffer_dirty(bh) ||
491 (jh && (jh->b_transaction != transaction ||
492 jh->b_next_transaction))) {
493 redirty_page_for_writepage(wbc, page);
496 } while ((bh = bh->b_this_page) != head);
499 return AOP_WRITEPAGE_ACTIVATE;
502 static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
504 struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
505 struct writeback_control wbc = {
506 .sync_mode = WB_SYNC_ALL,
507 .nr_to_write = LONG_MAX,
508 .range_start = jinode->i_dirty_start,
509 .range_end = jinode->i_dirty_end,
512 return write_cache_pages(mapping, &wbc,
513 ext4_journalled_writepage_callback,
514 jinode->i_transaction);
517 static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
521 if (ext4_should_journal_data(jinode->i_vfs_inode))
522 ret = ext4_journalled_submit_inode_data_buffers(jinode);
524 ret = jbd2_journal_submit_inode_data_buffers(jinode);
529 static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
533 if (!ext4_should_journal_data(jinode->i_vfs_inode))
534 ret = jbd2_journal_finish_inode_data_buffers(jinode);
539 static bool system_going_down(void)
541 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
542 || system_state == SYSTEM_RESTART;
545 struct ext4_err_translation {
550 #define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err }
552 static struct ext4_err_translation err_translation[] = {
553 EXT4_ERR_TRANSLATE(EIO),
554 EXT4_ERR_TRANSLATE(ENOMEM),
555 EXT4_ERR_TRANSLATE(EFSBADCRC),
556 EXT4_ERR_TRANSLATE(EFSCORRUPTED),
557 EXT4_ERR_TRANSLATE(ENOSPC),
558 EXT4_ERR_TRANSLATE(ENOKEY),
559 EXT4_ERR_TRANSLATE(EROFS),
560 EXT4_ERR_TRANSLATE(EFBIG),
561 EXT4_ERR_TRANSLATE(EEXIST),
562 EXT4_ERR_TRANSLATE(ERANGE),
563 EXT4_ERR_TRANSLATE(EOVERFLOW),
564 EXT4_ERR_TRANSLATE(EBUSY),
565 EXT4_ERR_TRANSLATE(ENOTDIR),
566 EXT4_ERR_TRANSLATE(ENOTEMPTY),
567 EXT4_ERR_TRANSLATE(ESHUTDOWN),
568 EXT4_ERR_TRANSLATE(EFAULT),
571 static int ext4_errno_to_code(int errno)
575 for (i = 0; i < ARRAY_SIZE(err_translation); i++)
576 if (err_translation[i].errno == errno)
577 return err_translation[i].code;
578 return EXT4_ERR_UNKNOWN;
581 static void save_error_info(struct super_block *sb, int error,
582 __u32 ino, __u64 block,
583 const char *func, unsigned int line)
585 struct ext4_sb_info *sbi = EXT4_SB(sb);
587 /* We default to EFSCORRUPTED error... */
589 error = EFSCORRUPTED;
591 spin_lock(&sbi->s_error_lock);
592 sbi->s_add_error_count++;
593 sbi->s_last_error_code = error;
594 sbi->s_last_error_line = line;
595 sbi->s_last_error_ino = ino;
596 sbi->s_last_error_block = block;
597 sbi->s_last_error_func = func;
598 sbi->s_last_error_time = ktime_get_real_seconds();
599 if (!sbi->s_first_error_time) {
600 sbi->s_first_error_code = error;
601 sbi->s_first_error_line = line;
602 sbi->s_first_error_ino = ino;
603 sbi->s_first_error_block = block;
604 sbi->s_first_error_func = func;
605 sbi->s_first_error_time = sbi->s_last_error_time;
607 spin_unlock(&sbi->s_error_lock);
610 /* Deal with the reporting of failure conditions on a filesystem such as
611 * inconsistencies detected or read IO failures.
613 * On ext2, we can store the error state of the filesystem in the
614 * superblock. That is not possible on ext4, because we may have other
615 * write ordering constraints on the superblock which prevent us from
616 * writing it out straight away; and given that the journal is about to
617 * be aborted, we can't rely on the current, or future, transactions to
618 * write out the superblock safely.
620 * We'll just use the jbd2_journal_abort() error code to record an error in
621 * the journal instead. On recovery, the journal will complain about
622 * that error until we've noted it down and cleared it.
624 * If force_ro is set, we unconditionally force the filesystem into an
625 * ABORT|READONLY state, unless the error response on the fs has been set to
626 * panic in which case we take the easy way out and panic immediately. This is
627 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM
628 * at a critical moment in log management.
630 static void ext4_handle_error(struct super_block *sb, bool force_ro, int error,
631 __u32 ino, __u64 block,
632 const char *func, unsigned int line)
634 journal_t *journal = EXT4_SB(sb)->s_journal;
635 bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT);
637 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
638 if (test_opt(sb, WARN_ON_ERROR))
641 if (!continue_fs && !sb_rdonly(sb)) {
642 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
644 jbd2_journal_abort(journal, -EIO);
647 if (!bdev_read_only(sb->s_bdev)) {
648 save_error_info(sb, error, ino, block, func, line);
650 * In case the fs should keep running, we need to writeout
651 * superblock through the journal. Due to lock ordering
652 * constraints, it may not be safe to do it right here so we
653 * defer superblock flushing to a workqueue.
655 if (continue_fs && journal)
656 schedule_work(&EXT4_SB(sb)->s_error_work);
658 ext4_commit_super(sb);
662 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
663 * could panic during 'reboot -f' as the underlying device got already
666 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
667 panic("EXT4-fs (device %s): panic forced after error\n",
671 if (sb_rdonly(sb) || continue_fs)
674 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
676 * Make sure updated value of ->s_mount_flags will be visible before
680 sb->s_flags |= SB_RDONLY;
683 static void flush_stashed_error_work(struct work_struct *work)
685 struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info,
687 journal_t *journal = sbi->s_journal;
691 * If the journal is still running, we have to write out superblock
692 * through the journal to avoid collisions of other journalled sb
695 * We use directly jbd2 functions here to avoid recursing back into
696 * ext4 error handling code during handling of previous errors.
698 if (!sb_rdonly(sbi->s_sb) && journal) {
699 struct buffer_head *sbh = sbi->s_sbh;
700 handle = jbd2_journal_start(journal, 1);
703 if (jbd2_journal_get_write_access(handle, sbh)) {
704 jbd2_journal_stop(handle);
707 ext4_update_super(sbi->s_sb);
708 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
709 ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to "
710 "superblock detected");
711 clear_buffer_write_io_error(sbh);
712 set_buffer_uptodate(sbh);
715 if (jbd2_journal_dirty_metadata(handle, sbh)) {
716 jbd2_journal_stop(handle);
719 jbd2_journal_stop(handle);
720 ext4_notify_error_sysfs(sbi);
725 * Write through journal failed. Write sb directly to get error info
726 * out and hope for the best.
728 ext4_commit_super(sbi->s_sb);
729 ext4_notify_error_sysfs(sbi);
732 #define ext4_error_ratelimit(sb) \
733 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
736 void __ext4_error(struct super_block *sb, const char *function,
737 unsigned int line, bool force_ro, int error, __u64 block,
738 const char *fmt, ...)
740 struct va_format vaf;
743 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
746 trace_ext4_error(sb, function, line);
747 if (ext4_error_ratelimit(sb)) {
752 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
753 sb->s_id, function, line, current->comm, &vaf);
756 ext4_handle_error(sb, force_ro, error, 0, block, function, line);
759 void __ext4_error_inode(struct inode *inode, const char *function,
760 unsigned int line, ext4_fsblk_t block, int error,
761 const char *fmt, ...)
764 struct va_format vaf;
766 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
769 trace_ext4_error(inode->i_sb, function, line);
770 if (ext4_error_ratelimit(inode->i_sb)) {
775 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
776 "inode #%lu: block %llu: comm %s: %pV\n",
777 inode->i_sb->s_id, function, line, inode->i_ino,
778 block, current->comm, &vaf);
780 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
781 "inode #%lu: comm %s: %pV\n",
782 inode->i_sb->s_id, function, line, inode->i_ino,
783 current->comm, &vaf);
786 ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block,
790 void __ext4_error_file(struct file *file, const char *function,
791 unsigned int line, ext4_fsblk_t block,
792 const char *fmt, ...)
795 struct va_format vaf;
796 struct inode *inode = file_inode(file);
797 char pathname[80], *path;
799 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
802 trace_ext4_error(inode->i_sb, function, line);
803 if (ext4_error_ratelimit(inode->i_sb)) {
804 path = file_path(file, pathname, sizeof(pathname));
812 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
813 "block %llu: comm %s: path %s: %pV\n",
814 inode->i_sb->s_id, function, line, inode->i_ino,
815 block, current->comm, path, &vaf);
818 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
819 "comm %s: path %s: %pV\n",
820 inode->i_sb->s_id, function, line, inode->i_ino,
821 current->comm, path, &vaf);
824 ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block,
828 const char *ext4_decode_error(struct super_block *sb, int errno,
835 errstr = "Corrupt filesystem";
838 errstr = "Filesystem failed CRC";
841 errstr = "IO failure";
844 errstr = "Out of memory";
847 if (!sb || (EXT4_SB(sb)->s_journal &&
848 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
849 errstr = "Journal has aborted";
851 errstr = "Readonly filesystem";
854 /* If the caller passed in an extra buffer for unknown
855 * errors, textualise them now. Else we just return
858 /* Check for truncated error codes... */
859 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
868 /* __ext4_std_error decodes expected errors from journaling functions
869 * automatically and invokes the appropriate error response. */
871 void __ext4_std_error(struct super_block *sb, const char *function,
872 unsigned int line, int errno)
877 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
880 /* Special case: if the error is EROFS, and we're not already
881 * inside a transaction, then there's really no point in logging
883 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
886 if (ext4_error_ratelimit(sb)) {
887 errstr = ext4_decode_error(sb, errno, nbuf);
888 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
889 sb->s_id, function, line, errstr);
892 ext4_handle_error(sb, false, -errno, 0, 0, function, line);
895 void __ext4_msg(struct super_block *sb,
896 const char *prefix, const char *fmt, ...)
898 struct va_format vaf;
901 atomic_inc(&EXT4_SB(sb)->s_msg_count);
902 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
908 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
912 static int ext4_warning_ratelimit(struct super_block *sb)
914 atomic_inc(&EXT4_SB(sb)->s_warning_count);
915 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
919 void __ext4_warning(struct super_block *sb, const char *function,
920 unsigned int line, const char *fmt, ...)
922 struct va_format vaf;
925 if (!ext4_warning_ratelimit(sb))
931 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
932 sb->s_id, function, line, &vaf);
936 void __ext4_warning_inode(const struct inode *inode, const char *function,
937 unsigned int line, const char *fmt, ...)
939 struct va_format vaf;
942 if (!ext4_warning_ratelimit(inode->i_sb))
948 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
949 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
950 function, line, inode->i_ino, current->comm, &vaf);
954 void __ext4_grp_locked_error(const char *function, unsigned int line,
955 struct super_block *sb, ext4_group_t grp,
956 unsigned long ino, ext4_fsblk_t block,
957 const char *fmt, ...)
961 struct va_format vaf;
964 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
967 trace_ext4_error(sb, function, line);
968 if (ext4_error_ratelimit(sb)) {
972 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
973 sb->s_id, function, line, grp);
975 printk(KERN_CONT "inode %lu: ", ino);
977 printk(KERN_CONT "block %llu:",
978 (unsigned long long) block);
979 printk(KERN_CONT "%pV\n", &vaf);
983 if (test_opt(sb, ERRORS_CONT)) {
984 if (test_opt(sb, WARN_ON_ERROR))
986 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
987 if (!bdev_read_only(sb->s_bdev)) {
988 save_error_info(sb, EFSCORRUPTED, ino, block, function,
990 schedule_work(&EXT4_SB(sb)->s_error_work);
994 ext4_unlock_group(sb, grp);
995 ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line);
997 * We only get here in the ERRORS_RO case; relocking the group
998 * may be dangerous, but nothing bad will happen since the
999 * filesystem will have already been marked read/only and the
1000 * journal has been aborted. We return 1 as a hint to callers
1001 * who might what to use the return value from
1002 * ext4_grp_locked_error() to distinguish between the
1003 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
1004 * aggressively from the ext4 function in question, with a
1005 * more appropriate error code.
1007 ext4_lock_group(sb, grp);
1011 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
1015 struct ext4_sb_info *sbi = EXT4_SB(sb);
1016 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1017 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
1022 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1023 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1026 percpu_counter_sub(&sbi->s_freeclusters_counter,
1030 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1031 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1036 count = ext4_free_inodes_count(sb, gdp);
1037 percpu_counter_sub(&sbi->s_freeinodes_counter,
1043 void ext4_update_dynamic_rev(struct super_block *sb)
1045 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
1047 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
1051 "updating to rev %d because of new feature flag, "
1052 "running e2fsck is recommended",
1055 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1056 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1057 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
1058 /* leave es->s_feature_*compat flags alone */
1059 /* es->s_uuid will be set by e2fsck if empty */
1062 * The rest of the superblock fields should be zero, and if not it
1063 * means they are likely already in use, so leave them alone. We
1064 * can leave it up to e2fsck to clean up any inconsistencies there.
1069 * Open the external journal device
1071 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
1073 struct block_device *bdev;
1075 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
1081 ext4_msg(sb, KERN_ERR,
1082 "failed to open journal device unknown-block(%u,%u) %ld",
1083 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
1088 * Release the journal device
1090 static void ext4_blkdev_put(struct block_device *bdev)
1092 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
1095 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
1097 struct block_device *bdev;
1098 bdev = sbi->s_journal_bdev;
1101 * Invalidate the journal device's buffers. We don't want them
1102 * floating about in memory - the physical journal device may
1103 * hotswapped, and it breaks the `ro-after' testing code.
1105 invalidate_bdev(bdev);
1106 ext4_blkdev_put(bdev);
1107 sbi->s_journal_bdev = NULL;
1111 static inline struct inode *orphan_list_entry(struct list_head *l)
1113 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1116 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1118 struct list_head *l;
1120 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1121 le32_to_cpu(sbi->s_es->s_last_orphan));
1123 printk(KERN_ERR "sb_info orphan list:\n");
1124 list_for_each(l, &sbi->s_orphan) {
1125 struct inode *inode = orphan_list_entry(l);
1127 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1128 inode->i_sb->s_id, inode->i_ino, inode,
1129 inode->i_mode, inode->i_nlink,
1130 NEXT_ORPHAN(inode));
1135 static int ext4_quota_off(struct super_block *sb, int type);
1137 static inline void ext4_quota_off_umount(struct super_block *sb)
1141 /* Use our quota_off function to clear inode flags etc. */
1142 for (type = 0; type < EXT4_MAXQUOTAS; type++)
1143 ext4_quota_off(sb, type);
1147 * This is a helper function which is used in the mount/remount
1148 * codepaths (which holds s_umount) to fetch the quota file name.
1150 static inline char *get_qf_name(struct super_block *sb,
1151 struct ext4_sb_info *sbi,
1154 return rcu_dereference_protected(sbi->s_qf_names[type],
1155 lockdep_is_held(&sb->s_umount));
1158 static inline void ext4_quota_off_umount(struct super_block *sb)
1163 static void ext4_put_super(struct super_block *sb)
1165 struct ext4_sb_info *sbi = EXT4_SB(sb);
1166 struct ext4_super_block *es = sbi->s_es;
1167 struct buffer_head **group_desc;
1168 struct flex_groups **flex_groups;
1173 * Unregister sysfs before destroying jbd2 journal.
1174 * Since we could still access attr_journal_task attribute via sysfs
1175 * path which could have sbi->s_journal->j_task as NULL
1176 * Unregister sysfs before flush sbi->s_error_work.
1177 * Since user may read /proc/fs/ext4/xx/mb_groups during umount, If
1178 * read metadata verify failed then will queue error work.
1179 * flush_stashed_error_work will call start_this_handle may trigger
1182 ext4_unregister_sysfs(sb);
1184 ext4_unregister_li_request(sb);
1185 ext4_quota_off_umount(sb);
1187 flush_work(&sbi->s_error_work);
1188 destroy_workqueue(sbi->rsv_conversion_wq);
1189 ext4_release_orphan_info(sb);
1191 if (sbi->s_journal) {
1192 aborted = is_journal_aborted(sbi->s_journal);
1193 err = jbd2_journal_destroy(sbi->s_journal);
1194 sbi->s_journal = NULL;
1195 if ((err < 0) && !aborted) {
1196 ext4_abort(sb, -err, "Couldn't clean up the journal");
1200 ext4_es_unregister_shrinker(sbi);
1201 del_timer_sync(&sbi->s_err_report);
1202 ext4_release_system_zone(sb);
1203 ext4_mb_release(sb);
1204 ext4_ext_release(sb);
1206 if (!sb_rdonly(sb) && !aborted) {
1207 ext4_clear_feature_journal_needs_recovery(sb);
1208 ext4_clear_feature_orphan_present(sb);
1209 es->s_state = cpu_to_le16(sbi->s_mount_state);
1212 ext4_commit_super(sb);
1215 group_desc = rcu_dereference(sbi->s_group_desc);
1216 for (i = 0; i < sbi->s_gdb_count; i++)
1217 brelse(group_desc[i]);
1219 flex_groups = rcu_dereference(sbi->s_flex_groups);
1221 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1222 kvfree(flex_groups[i]);
1223 kvfree(flex_groups);
1226 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1227 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1228 percpu_counter_destroy(&sbi->s_dirs_counter);
1229 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1230 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
1231 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1233 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1234 kfree(get_qf_name(sb, sbi, i));
1237 /* Debugging code just in case the in-memory inode orphan list
1238 * isn't empty. The on-disk one can be non-empty if we've
1239 * detected an error and taken the fs readonly, but the
1240 * in-memory list had better be clean by this point. */
1241 if (!list_empty(&sbi->s_orphan))
1242 dump_orphan_list(sb, sbi);
1243 ASSERT(list_empty(&sbi->s_orphan));
1245 sync_blockdev(sb->s_bdev);
1246 invalidate_bdev(sb->s_bdev);
1247 if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) {
1248 sync_blockdev(sbi->s_journal_bdev);
1249 ext4_blkdev_remove(sbi);
1252 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1253 sbi->s_ea_inode_cache = NULL;
1255 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1256 sbi->s_ea_block_cache = NULL;
1258 ext4_stop_mmpd(sbi);
1261 sb->s_fs_info = NULL;
1263 * Now that we are completely done shutting down the
1264 * superblock, we need to actually destroy the kobject.
1266 kobject_put(&sbi->s_kobj);
1267 wait_for_completion(&sbi->s_kobj_unregister);
1268 if (sbi->s_chksum_driver)
1269 crypto_free_shash(sbi->s_chksum_driver);
1270 kfree(sbi->s_blockgroup_lock);
1271 fs_put_dax(sbi->s_daxdev);
1272 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
1273 #ifdef CONFIG_UNICODE
1274 utf8_unload(sb->s_encoding);
1279 static struct kmem_cache *ext4_inode_cachep;
1282 * Called inside transaction, so use GFP_NOFS
1284 static struct inode *ext4_alloc_inode(struct super_block *sb)
1286 struct ext4_inode_info *ei;
1288 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1292 inode_set_iversion(&ei->vfs_inode, 1);
1294 spin_lock_init(&ei->i_raw_lock);
1295 INIT_LIST_HEAD(&ei->i_prealloc_list);
1296 atomic_set(&ei->i_prealloc_active, 0);
1297 spin_lock_init(&ei->i_prealloc_lock);
1298 ext4_es_init_tree(&ei->i_es_tree);
1299 rwlock_init(&ei->i_es_lock);
1300 INIT_LIST_HEAD(&ei->i_es_list);
1301 ei->i_es_all_nr = 0;
1302 ei->i_es_shk_nr = 0;
1303 ei->i_es_shrink_lblk = 0;
1304 ei->i_reserved_data_blocks = 0;
1305 spin_lock_init(&(ei->i_block_reservation_lock));
1306 ext4_init_pending_tree(&ei->i_pending_tree);
1308 ei->i_reserved_quota = 0;
1309 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1312 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1313 spin_lock_init(&ei->i_completed_io_lock);
1315 ei->i_datasync_tid = 0;
1316 atomic_set(&ei->i_unwritten, 0);
1317 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1318 ext4_fc_init_inode(&ei->vfs_inode);
1319 mutex_init(&ei->i_fc_lock);
1320 return &ei->vfs_inode;
1323 static int ext4_drop_inode(struct inode *inode)
1325 int drop = generic_drop_inode(inode);
1328 drop = fscrypt_drop_inode(inode);
1330 trace_ext4_drop_inode(inode, drop);
1334 static void ext4_free_in_core_inode(struct inode *inode)
1336 fscrypt_free_inode(inode);
1337 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1338 pr_warn("%s: inode %ld still in fc list",
1339 __func__, inode->i_ino);
1341 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1344 static void ext4_destroy_inode(struct inode *inode)
1346 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1347 ext4_msg(inode->i_sb, KERN_ERR,
1348 "Inode %lu (%p): orphan list check failed!",
1349 inode->i_ino, EXT4_I(inode));
1350 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1351 EXT4_I(inode), sizeof(struct ext4_inode_info),
1356 if (EXT4_I(inode)->i_reserved_data_blocks)
1357 ext4_msg(inode->i_sb, KERN_ERR,
1358 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!",
1359 inode->i_ino, EXT4_I(inode),
1360 EXT4_I(inode)->i_reserved_data_blocks);
1363 static void init_once(void *foo)
1365 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1367 INIT_LIST_HEAD(&ei->i_orphan);
1368 init_rwsem(&ei->xattr_sem);
1369 init_rwsem(&ei->i_data_sem);
1370 inode_init_once(&ei->vfs_inode);
1371 ext4_fc_init_inode(&ei->vfs_inode);
1374 static int __init init_inodecache(void)
1376 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1377 sizeof(struct ext4_inode_info), 0,
1378 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1380 offsetof(struct ext4_inode_info, i_data),
1381 sizeof_field(struct ext4_inode_info, i_data),
1383 if (ext4_inode_cachep == NULL)
1388 static void destroy_inodecache(void)
1391 * Make sure all delayed rcu free inodes are flushed before we
1395 kmem_cache_destroy(ext4_inode_cachep);
1398 void ext4_clear_inode(struct inode *inode)
1401 invalidate_inode_buffers(inode);
1403 ext4_discard_preallocations(inode, 0);
1404 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1406 if (EXT4_I(inode)->jinode) {
1407 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1408 EXT4_I(inode)->jinode);
1409 jbd2_free_inode(EXT4_I(inode)->jinode);
1410 EXT4_I(inode)->jinode = NULL;
1412 fscrypt_put_encryption_info(inode);
1413 fsverity_cleanup_inode(inode);
1416 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1417 u64 ino, u32 generation)
1419 struct inode *inode;
1422 * Currently we don't know the generation for parent directory, so
1423 * a generation of 0 means "accept any"
1425 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1427 return ERR_CAST(inode);
1428 if (generation && inode->i_generation != generation) {
1430 return ERR_PTR(-ESTALE);
1436 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1437 int fh_len, int fh_type)
1439 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1440 ext4_nfs_get_inode);
1443 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1444 int fh_len, int fh_type)
1446 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1447 ext4_nfs_get_inode);
1450 static int ext4_nfs_commit_metadata(struct inode *inode)
1452 struct writeback_control wbc = {
1453 .sync_mode = WB_SYNC_ALL
1456 trace_ext4_nfs_commit_metadata(inode);
1457 return ext4_write_inode(inode, &wbc);
1460 #ifdef CONFIG_FS_ENCRYPTION
1461 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1463 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1464 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1467 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1470 handle_t *handle = fs_data;
1471 int res, res2, credits, retries = 0;
1474 * Encrypting the root directory is not allowed because e2fsck expects
1475 * lost+found to exist and be unencrypted, and encrypting the root
1476 * directory would imply encrypting the lost+found directory as well as
1477 * the filename "lost+found" itself.
1479 if (inode->i_ino == EXT4_ROOT_INO)
1482 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1485 if (ext4_test_inode_flag(inode, EXT4_INODE_DAX))
1488 res = ext4_convert_inline_data(inode);
1493 * If a journal handle was specified, then the encryption context is
1494 * being set on a new inode via inheritance and is part of a larger
1495 * transaction to create the inode. Otherwise the encryption context is
1496 * being set on an existing inode in its own transaction. Only in the
1497 * latter case should the "retry on ENOSPC" logic be used.
1501 res = ext4_xattr_set_handle(handle, inode,
1502 EXT4_XATTR_INDEX_ENCRYPTION,
1503 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1506 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1507 ext4_clear_inode_state(inode,
1508 EXT4_STATE_MAY_INLINE_DATA);
1510 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1511 * S_DAX may be disabled
1513 ext4_set_inode_flags(inode, false);
1518 res = dquot_initialize(inode);
1522 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1527 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1529 return PTR_ERR(handle);
1531 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1532 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1535 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1537 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1538 * S_DAX may be disabled
1540 ext4_set_inode_flags(inode, false);
1541 res = ext4_mark_inode_dirty(handle, inode);
1543 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1545 res2 = ext4_journal_stop(handle);
1547 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1554 static const union fscrypt_policy *ext4_get_dummy_policy(struct super_block *sb)
1556 return EXT4_SB(sb)->s_dummy_enc_policy.policy;
1559 static bool ext4_has_stable_inodes(struct super_block *sb)
1561 return ext4_has_feature_stable_inodes(sb);
1564 static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1565 int *ino_bits_ret, int *lblk_bits_ret)
1567 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1568 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1571 static const struct fscrypt_operations ext4_cryptops = {
1572 .key_prefix = "ext4:",
1573 .get_context = ext4_get_context,
1574 .set_context = ext4_set_context,
1575 .get_dummy_policy = ext4_get_dummy_policy,
1576 .empty_dir = ext4_empty_dir,
1577 .max_namelen = EXT4_NAME_LEN,
1578 .has_stable_inodes = ext4_has_stable_inodes,
1579 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits,
1584 static const char * const quotatypes[] = INITQFNAMES;
1585 #define QTYPE2NAME(t) (quotatypes[t])
1587 static int ext4_write_dquot(struct dquot *dquot);
1588 static int ext4_acquire_dquot(struct dquot *dquot);
1589 static int ext4_release_dquot(struct dquot *dquot);
1590 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1591 static int ext4_write_info(struct super_block *sb, int type);
1592 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1593 const struct path *path);
1594 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1595 size_t len, loff_t off);
1596 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1597 const char *data, size_t len, loff_t off);
1598 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1599 unsigned int flags);
1601 static struct dquot **ext4_get_dquots(struct inode *inode)
1603 return EXT4_I(inode)->i_dquot;
1606 static const struct dquot_operations ext4_quota_operations = {
1607 .get_reserved_space = ext4_get_reserved_space,
1608 .write_dquot = ext4_write_dquot,
1609 .acquire_dquot = ext4_acquire_dquot,
1610 .release_dquot = ext4_release_dquot,
1611 .mark_dirty = ext4_mark_dquot_dirty,
1612 .write_info = ext4_write_info,
1613 .alloc_dquot = dquot_alloc,
1614 .destroy_dquot = dquot_destroy,
1615 .get_projid = ext4_get_projid,
1616 .get_inode_usage = ext4_get_inode_usage,
1617 .get_next_id = dquot_get_next_id,
1620 static const struct quotactl_ops ext4_qctl_operations = {
1621 .quota_on = ext4_quota_on,
1622 .quota_off = ext4_quota_off,
1623 .quota_sync = dquot_quota_sync,
1624 .get_state = dquot_get_state,
1625 .set_info = dquot_set_dqinfo,
1626 .get_dqblk = dquot_get_dqblk,
1627 .set_dqblk = dquot_set_dqblk,
1628 .get_nextdqblk = dquot_get_next_dqblk,
1632 static const struct super_operations ext4_sops = {
1633 .alloc_inode = ext4_alloc_inode,
1634 .free_inode = ext4_free_in_core_inode,
1635 .destroy_inode = ext4_destroy_inode,
1636 .write_inode = ext4_write_inode,
1637 .dirty_inode = ext4_dirty_inode,
1638 .drop_inode = ext4_drop_inode,
1639 .evict_inode = ext4_evict_inode,
1640 .put_super = ext4_put_super,
1641 .sync_fs = ext4_sync_fs,
1642 .freeze_fs = ext4_freeze,
1643 .unfreeze_fs = ext4_unfreeze,
1644 .statfs = ext4_statfs,
1645 .remount_fs = ext4_remount,
1646 .show_options = ext4_show_options,
1648 .quota_read = ext4_quota_read,
1649 .quota_write = ext4_quota_write,
1650 .get_dquots = ext4_get_dquots,
1654 static const struct export_operations ext4_export_ops = {
1655 .fh_to_dentry = ext4_fh_to_dentry,
1656 .fh_to_parent = ext4_fh_to_parent,
1657 .get_parent = ext4_get_parent,
1658 .commit_metadata = ext4_nfs_commit_metadata,
1662 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1663 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1664 Opt_nouid32, Opt_debug, Opt_removed,
1665 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1666 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1667 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1668 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1669 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1670 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1672 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1673 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1674 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1675 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version,
1676 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1677 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1678 Opt_nowarn_on_error, Opt_mblk_io_submit,
1679 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1680 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1681 Opt_inode_readahead_blks, Opt_journal_ioprio,
1682 Opt_dioread_nolock, Opt_dioread_lock,
1683 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1684 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1685 Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan,
1686 #ifdef CONFIG_EXT4_DEBUG
1687 Opt_fc_debug_max_replay, Opt_fc_debug_force
1691 static const match_table_t tokens = {
1692 {Opt_bsd_df, "bsddf"},
1693 {Opt_minix_df, "minixdf"},
1694 {Opt_grpid, "grpid"},
1695 {Opt_grpid, "bsdgroups"},
1696 {Opt_nogrpid, "nogrpid"},
1697 {Opt_nogrpid, "sysvgroups"},
1698 {Opt_resgid, "resgid=%u"},
1699 {Opt_resuid, "resuid=%u"},
1701 {Opt_err_cont, "errors=continue"},
1702 {Opt_err_panic, "errors=panic"},
1703 {Opt_err_ro, "errors=remount-ro"},
1704 {Opt_nouid32, "nouid32"},
1705 {Opt_debug, "debug"},
1706 {Opt_removed, "oldalloc"},
1707 {Opt_removed, "orlov"},
1708 {Opt_user_xattr, "user_xattr"},
1709 {Opt_nouser_xattr, "nouser_xattr"},
1711 {Opt_noacl, "noacl"},
1712 {Opt_noload, "norecovery"},
1713 {Opt_noload, "noload"},
1714 {Opt_removed, "nobh"},
1715 {Opt_removed, "bh"},
1716 {Opt_commit, "commit=%u"},
1717 {Opt_min_batch_time, "min_batch_time=%u"},
1718 {Opt_max_batch_time, "max_batch_time=%u"},
1719 {Opt_journal_dev, "journal_dev=%u"},
1720 {Opt_journal_path, "journal_path=%s"},
1721 {Opt_journal_checksum, "journal_checksum"},
1722 {Opt_nojournal_checksum, "nojournal_checksum"},
1723 {Opt_journal_async_commit, "journal_async_commit"},
1724 {Opt_abort, "abort"},
1725 {Opt_data_journal, "data=journal"},
1726 {Opt_data_ordered, "data=ordered"},
1727 {Opt_data_writeback, "data=writeback"},
1728 {Opt_data_err_abort, "data_err=abort"},
1729 {Opt_data_err_ignore, "data_err=ignore"},
1730 {Opt_offusrjquota, "usrjquota="},
1731 {Opt_usrjquota, "usrjquota=%s"},
1732 {Opt_offgrpjquota, "grpjquota="},
1733 {Opt_grpjquota, "grpjquota=%s"},
1734 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1735 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1736 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1737 {Opt_grpquota, "grpquota"},
1738 {Opt_noquota, "noquota"},
1739 {Opt_quota, "quota"},
1740 {Opt_usrquota, "usrquota"},
1741 {Opt_prjquota, "prjquota"},
1742 {Opt_barrier, "barrier=%u"},
1743 {Opt_barrier, "barrier"},
1744 {Opt_nobarrier, "nobarrier"},
1745 {Opt_i_version, "i_version"},
1747 {Opt_dax_always, "dax=always"},
1748 {Opt_dax_inode, "dax=inode"},
1749 {Opt_dax_never, "dax=never"},
1750 {Opt_stripe, "stripe=%u"},
1751 {Opt_delalloc, "delalloc"},
1752 {Opt_warn_on_error, "warn_on_error"},
1753 {Opt_nowarn_on_error, "nowarn_on_error"},
1754 {Opt_lazytime, "lazytime"},
1755 {Opt_nolazytime, "nolazytime"},
1756 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1757 {Opt_nodelalloc, "nodelalloc"},
1758 {Opt_removed, "mblk_io_submit"},
1759 {Opt_removed, "nomblk_io_submit"},
1760 {Opt_block_validity, "block_validity"},
1761 {Opt_noblock_validity, "noblock_validity"},
1762 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1763 {Opt_journal_ioprio, "journal_ioprio=%u"},
1764 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1765 {Opt_auto_da_alloc, "auto_da_alloc"},
1766 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1767 {Opt_dioread_nolock, "dioread_nolock"},
1768 {Opt_dioread_lock, "nodioread_nolock"},
1769 {Opt_dioread_lock, "dioread_lock"},
1770 {Opt_discard, "discard"},
1771 {Opt_nodiscard, "nodiscard"},
1772 {Opt_init_itable, "init_itable=%u"},
1773 {Opt_init_itable, "init_itable"},
1774 {Opt_noinit_itable, "noinit_itable"},
1775 #ifdef CONFIG_EXT4_DEBUG
1776 {Opt_fc_debug_force, "fc_debug_force"},
1777 {Opt_fc_debug_max_replay, "fc_debug_max_replay=%u"},
1779 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1780 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
1781 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1782 {Opt_inlinecrypt, "inlinecrypt"},
1783 {Opt_nombcache, "nombcache"},
1784 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1785 {Opt_removed, "prefetch_block_bitmaps"},
1786 {Opt_no_prefetch_block_bitmaps, "no_prefetch_block_bitmaps"},
1787 {Opt_mb_optimize_scan, "mb_optimize_scan=%d"},
1788 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1789 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1790 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1791 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1792 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1796 static ext4_fsblk_t get_sb_block(void **data)
1798 ext4_fsblk_t sb_block;
1799 char *options = (char *) *data;
1801 if (!options || strncmp(options, "sb=", 3) != 0)
1802 return 1; /* Default location */
1805 /* TODO: use simple_strtoll with >32bit ext4 */
1806 sb_block = simple_strtoul(options, &options, 0);
1807 if (*options && *options != ',') {
1808 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1812 if (*options == ',')
1814 *data = (void *) options;
1819 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1820 #define DEFAULT_MB_OPTIMIZE_SCAN (-1)
1822 static const char deprecated_msg[] =
1823 "Mount option \"%s\" will be removed by %s\n"
1824 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1827 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1829 struct ext4_sb_info *sbi = EXT4_SB(sb);
1830 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
1833 if (sb_any_quota_loaded(sb) && !old_qname) {
1834 ext4_msg(sb, KERN_ERR,
1835 "Cannot change journaled "
1836 "quota options when quota turned on");
1839 if (ext4_has_feature_quota(sb)) {
1840 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1841 "ignored when QUOTA feature is enabled");
1844 qname = match_strdup(args);
1846 ext4_msg(sb, KERN_ERR,
1847 "Not enough memory for storing quotafile name");
1851 if (strcmp(old_qname, qname) == 0)
1854 ext4_msg(sb, KERN_ERR,
1855 "%s quota file already specified",
1859 if (strchr(qname, '/')) {
1860 ext4_msg(sb, KERN_ERR,
1861 "quotafile must be on filesystem root");
1864 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
1872 static int clear_qf_name(struct super_block *sb, int qtype)
1875 struct ext4_sb_info *sbi = EXT4_SB(sb);
1876 char *old_qname = get_qf_name(sb, sbi, qtype);
1878 if (sb_any_quota_loaded(sb) && old_qname) {
1879 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1880 " when quota turned on");
1883 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1890 #define MOPT_SET 0x0001
1891 #define MOPT_CLEAR 0x0002
1892 #define MOPT_NOSUPPORT 0x0004
1893 #define MOPT_EXPLICIT 0x0008
1894 #define MOPT_CLEAR_ERR 0x0010
1895 #define MOPT_GTE0 0x0020
1898 #define MOPT_QFMT 0x0040
1900 #define MOPT_Q MOPT_NOSUPPORT
1901 #define MOPT_QFMT MOPT_NOSUPPORT
1903 #define MOPT_DATAJ 0x0080
1904 #define MOPT_NO_EXT2 0x0100
1905 #define MOPT_NO_EXT3 0x0200
1906 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1907 #define MOPT_STRING 0x0400
1908 #define MOPT_SKIP 0x0800
1909 #define MOPT_2 0x1000
1911 static const struct mount_opts {
1915 } ext4_mount_opts[] = {
1916 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1917 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1918 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1919 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1920 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1921 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1922 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1923 MOPT_EXT4_ONLY | MOPT_SET},
1924 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1925 MOPT_EXT4_ONLY | MOPT_CLEAR},
1926 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1927 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1928 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1929 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1930 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1931 MOPT_EXT4_ONLY | MOPT_CLEAR},
1932 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1933 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1934 {Opt_commit, 0, MOPT_NO_EXT2},
1935 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1936 MOPT_EXT4_ONLY | MOPT_CLEAR},
1937 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1938 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1939 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1940 EXT4_MOUNT_JOURNAL_CHECKSUM),
1941 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1942 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1943 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1944 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1945 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1946 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1948 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1950 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1951 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1952 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1953 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1954 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1955 {Opt_commit, 0, MOPT_GTE0},
1956 {Opt_max_batch_time, 0, MOPT_GTE0},
1957 {Opt_min_batch_time, 0, MOPT_GTE0},
1958 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1959 {Opt_init_itable, 0, MOPT_GTE0},
1960 {Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP},
1961 {Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS,
1962 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1963 {Opt_dax_inode, EXT4_MOUNT2_DAX_INODE,
1964 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1965 {Opt_dax_never, EXT4_MOUNT2_DAX_NEVER,
1966 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1967 {Opt_stripe, 0, MOPT_GTE0},
1968 {Opt_resuid, 0, MOPT_GTE0},
1969 {Opt_resgid, 0, MOPT_GTE0},
1970 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1971 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1972 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1973 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1974 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1975 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1976 MOPT_NO_EXT2 | MOPT_DATAJ},
1977 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1978 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1979 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1980 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1981 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1983 {Opt_acl, 0, MOPT_NOSUPPORT},
1984 {Opt_noacl, 0, MOPT_NOSUPPORT},
1986 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1987 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1988 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1989 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1990 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1992 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1994 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1996 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1997 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1998 MOPT_CLEAR | MOPT_Q},
1999 {Opt_usrjquota, 0, MOPT_Q | MOPT_STRING},
2000 {Opt_grpjquota, 0, MOPT_Q | MOPT_STRING},
2001 {Opt_offusrjquota, 0, MOPT_Q},
2002 {Opt_offgrpjquota, 0, MOPT_Q},
2003 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
2004 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
2005 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
2006 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
2007 {Opt_test_dummy_encryption, 0, MOPT_STRING},
2008 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
2009 {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS,
2011 {Opt_mb_optimize_scan, EXT4_MOUNT2_MB_OPTIMIZE_SCAN, MOPT_GTE0},
2012 #ifdef CONFIG_EXT4_DEBUG
2013 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
2014 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
2015 {Opt_fc_debug_max_replay, 0, MOPT_GTE0},
2020 #ifdef CONFIG_UNICODE
2021 static const struct ext4_sb_encodings {
2025 } ext4_sb_encoding_map[] = {
2026 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
2029 static int ext4_sb_read_encoding(const struct ext4_super_block *es,
2030 const struct ext4_sb_encodings **encoding,
2033 __u16 magic = le16_to_cpu(es->s_encoding);
2036 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
2037 if (magic == ext4_sb_encoding_map[i].magic)
2040 if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
2043 *encoding = &ext4_sb_encoding_map[i];
2044 *flags = le16_to_cpu(es->s_encoding_flags);
2050 static int ext4_set_test_dummy_encryption(struct super_block *sb,
2052 const substring_t *arg,
2055 #ifdef CONFIG_FS_ENCRYPTION
2056 struct ext4_sb_info *sbi = EXT4_SB(sb);
2059 if (!ext4_has_feature_encrypt(sb)) {
2060 ext4_msg(sb, KERN_WARNING,
2061 "test_dummy_encryption requires encrypt feature");
2066 * This mount option is just for testing, and it's not worthwhile to
2067 * implement the extra complexity (e.g. RCU protection) that would be
2068 * needed to allow it to be set or changed during remount. We do allow
2069 * it to be specified during remount, but only if there is no change.
2071 if (is_remount && !sbi->s_dummy_enc_policy.policy) {
2072 ext4_msg(sb, KERN_WARNING,
2073 "Can't set test_dummy_encryption on remount");
2076 err = fscrypt_set_test_dummy_encryption(sb, arg->from,
2077 &sbi->s_dummy_enc_policy);
2080 ext4_msg(sb, KERN_WARNING,
2081 "Can't change test_dummy_encryption on remount");
2082 else if (err == -EINVAL)
2083 ext4_msg(sb, KERN_WARNING,
2084 "Value of option \"%s\" is unrecognized", opt);
2086 ext4_msg(sb, KERN_WARNING,
2087 "Error processing option \"%s\" [%d]",
2091 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2094 ext4_msg(sb, KERN_WARNING,
2095 "test_dummy_encryption option not supported");
2101 struct ext4_parsed_options {
2102 unsigned long journal_devnum;
2103 unsigned int journal_ioprio;
2104 int mb_optimize_scan;
2107 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
2108 substring_t *args, struct ext4_parsed_options *parsed_opts,
2111 struct ext4_sb_info *sbi = EXT4_SB(sb);
2112 const struct mount_opts *m;
2118 if (token == Opt_usrjquota)
2119 return set_qf_name(sb, USRQUOTA, &args[0]);
2120 else if (token == Opt_grpjquota)
2121 return set_qf_name(sb, GRPQUOTA, &args[0]);
2122 else if (token == Opt_offusrjquota)
2123 return clear_qf_name(sb, USRQUOTA);
2124 else if (token == Opt_offgrpjquota)
2125 return clear_qf_name(sb, GRPQUOTA);
2129 case Opt_nouser_xattr:
2130 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
2133 return 1; /* handled by get_sb_block() */
2135 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
2138 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
2141 sb->s_flags |= SB_I_VERSION;
2144 sb->s_flags |= SB_LAZYTIME;
2146 case Opt_nolazytime:
2147 sb->s_flags &= ~SB_LAZYTIME;
2149 case Opt_inlinecrypt:
2150 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2151 sb->s_flags |= SB_INLINECRYPT;
2153 ext4_msg(sb, KERN_ERR, "inline encryption not supported");
2158 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2159 if (token == m->token)
2162 if (m->token == Opt_err) {
2163 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
2164 "or missing value", opt);
2168 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2169 ext4_msg(sb, KERN_ERR,
2170 "Mount option \"%s\" incompatible with ext2", opt);
2173 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2174 ext4_msg(sb, KERN_ERR,
2175 "Mount option \"%s\" incompatible with ext3", opt);
2179 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
2181 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
2183 if (m->flags & MOPT_EXPLICIT) {
2184 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2185 set_opt2(sb, EXPLICIT_DELALLOC);
2186 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2187 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
2191 if (m->flags & MOPT_CLEAR_ERR)
2192 clear_opt(sb, ERRORS_MASK);
2193 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
2194 ext4_msg(sb, KERN_ERR, "Cannot change quota "
2195 "options when quota turned on");
2199 if (m->flags & MOPT_NOSUPPORT) {
2200 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
2201 } else if (token == Opt_commit) {
2203 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
2204 else if (arg > INT_MAX / HZ) {
2205 ext4_msg(sb, KERN_ERR,
2206 "Invalid commit interval %d, "
2207 "must be smaller than %d",
2211 sbi->s_commit_interval = HZ * arg;
2212 } else if (token == Opt_debug_want_extra_isize) {
2215 (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
2216 ext4_msg(sb, KERN_ERR,
2217 "Invalid want_extra_isize %d", arg);
2220 sbi->s_want_extra_isize = arg;
2221 } else if (token == Opt_max_batch_time) {
2222 sbi->s_max_batch_time = arg;
2223 } else if (token == Opt_min_batch_time) {
2224 sbi->s_min_batch_time = arg;
2225 } else if (token == Opt_inode_readahead_blks) {
2226 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
2227 ext4_msg(sb, KERN_ERR,
2228 "EXT4-fs: inode_readahead_blks must be "
2229 "0 or a power of 2 smaller than 2^31");
2232 sbi->s_inode_readahead_blks = arg;
2233 } else if (token == Opt_init_itable) {
2234 set_opt(sb, INIT_INODE_TABLE);
2236 arg = EXT4_DEF_LI_WAIT_MULT;
2237 sbi->s_li_wait_mult = arg;
2238 } else if (token == Opt_max_dir_size_kb) {
2239 sbi->s_max_dir_size_kb = arg;
2240 #ifdef CONFIG_EXT4_DEBUG
2241 } else if (token == Opt_fc_debug_max_replay) {
2242 sbi->s_fc_debug_max_replay = arg;
2244 } else if (token == Opt_stripe) {
2245 sbi->s_stripe = arg;
2246 } else if (token == Opt_resuid) {
2247 uid = make_kuid(current_user_ns(), arg);
2248 if (!uid_valid(uid)) {
2249 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
2252 sbi->s_resuid = uid;
2253 } else if (token == Opt_resgid) {
2254 gid = make_kgid(current_user_ns(), arg);
2255 if (!gid_valid(gid)) {
2256 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
2259 sbi->s_resgid = gid;
2260 } else if (token == Opt_journal_dev) {
2262 ext4_msg(sb, KERN_ERR,
2263 "Cannot specify journal on remount");
2266 parsed_opts->journal_devnum = arg;
2267 } else if (token == Opt_journal_path) {
2269 struct inode *journal_inode;
2274 ext4_msg(sb, KERN_ERR,
2275 "Cannot specify journal on remount");
2278 journal_path = match_strdup(&args[0]);
2279 if (!journal_path) {
2280 ext4_msg(sb, KERN_ERR, "error: could not dup "
2281 "journal device string");
2285 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
2287 ext4_msg(sb, KERN_ERR, "error: could not find "
2288 "journal device path: error %d", error);
2289 kfree(journal_path);
2293 journal_inode = d_inode(path.dentry);
2294 if (!S_ISBLK(journal_inode->i_mode)) {
2295 ext4_msg(sb, KERN_ERR, "error: journal path %s "
2296 "is not a block device", journal_path);
2298 kfree(journal_path);
2302 parsed_opts->journal_devnum = new_encode_dev(journal_inode->i_rdev);
2304 kfree(journal_path);
2305 } else if (token == Opt_journal_ioprio) {
2307 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
2311 parsed_opts->journal_ioprio =
2312 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
2313 } else if (token == Opt_test_dummy_encryption) {
2314 return ext4_set_test_dummy_encryption(sb, opt, &args[0],
2316 } else if (m->flags & MOPT_DATAJ) {
2318 if (!sbi->s_journal)
2319 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
2320 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
2321 ext4_msg(sb, KERN_ERR,
2322 "Cannot change data mode on remount");
2326 clear_opt(sb, DATA_FLAGS);
2327 sbi->s_mount_opt |= m->mount_opt;
2330 } else if (m->flags & MOPT_QFMT) {
2331 if (sb_any_quota_loaded(sb) &&
2332 sbi->s_jquota_fmt != m->mount_opt) {
2333 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
2334 "quota options when quota turned on");
2337 if (ext4_has_feature_quota(sb)) {
2338 ext4_msg(sb, KERN_INFO,
2339 "Quota format mount options ignored "
2340 "when QUOTA feature is enabled");
2343 sbi->s_jquota_fmt = m->mount_opt;
2345 } else if (token == Opt_dax || token == Opt_dax_always ||
2346 token == Opt_dax_inode || token == Opt_dax_never) {
2347 #ifdef CONFIG_FS_DAX
2350 case Opt_dax_always:
2352 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2353 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2354 fail_dax_change_remount:
2355 ext4_msg(sb, KERN_ERR, "can't change "
2356 "dax mount option while remounting");
2360 (test_opt(sb, DATA_FLAGS) ==
2361 EXT4_MOUNT_JOURNAL_DATA)) {
2362 ext4_msg(sb, KERN_ERR, "can't mount with "
2363 "both data=journal and dax");
2366 ext4_msg(sb, KERN_WARNING,
2367 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2368 sbi->s_mount_opt |= EXT4_MOUNT_DAX_ALWAYS;
2369 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2373 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2374 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS)))
2375 goto fail_dax_change_remount;
2376 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2377 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2381 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2382 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2383 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE)))
2384 goto fail_dax_change_remount;
2385 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2386 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2387 /* Strictly for printing options */
2388 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_INODE;
2392 ext4_msg(sb, KERN_INFO, "dax option not supported");
2393 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2394 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2397 } else if (token == Opt_data_err_abort) {
2398 sbi->s_mount_opt |= m->mount_opt;
2399 } else if (token == Opt_data_err_ignore) {
2400 sbi->s_mount_opt &= ~m->mount_opt;
2401 } else if (token == Opt_mb_optimize_scan) {
2402 if (arg != 0 && arg != 1) {
2403 ext4_msg(sb, KERN_WARNING,
2404 "mb_optimize_scan should be set to 0 or 1.");
2407 parsed_opts->mb_optimize_scan = arg;
2411 if (m->flags & MOPT_CLEAR)
2413 else if (unlikely(!(m->flags & MOPT_SET))) {
2414 ext4_msg(sb, KERN_WARNING,
2415 "buggy handling of option %s", opt);
2419 if (m->flags & MOPT_2) {
2421 sbi->s_mount_opt2 |= m->mount_opt;
2423 sbi->s_mount_opt2 &= ~m->mount_opt;
2426 sbi->s_mount_opt |= m->mount_opt;
2428 sbi->s_mount_opt &= ~m->mount_opt;
2434 static int parse_options(char *options, struct super_block *sb,
2435 struct ext4_parsed_options *ret_opts,
2438 struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb);
2439 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
2440 substring_t args[MAX_OPT_ARGS];
2446 while ((p = strsep(&options, ",")) != NULL) {
2450 * Initialize args struct so we know whether arg was
2451 * found; some options take optional arguments.
2453 args[0].to = args[0].from = NULL;
2454 token = match_token(p, tokens, args);
2455 if (handle_mount_opt(sb, p, token, args, ret_opts,
2461 * We do the test below only for project quotas. 'usrquota' and
2462 * 'grpquota' mount options are allowed even without quota feature
2463 * to support legacy quotas in quota files.
2465 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2466 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2467 "Cannot enable project quota enforcement.");
2470 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2471 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2472 if (usr_qf_name || grp_qf_name) {
2473 if (test_opt(sb, USRQUOTA) && usr_qf_name)
2474 clear_opt(sb, USRQUOTA);
2476 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
2477 clear_opt(sb, GRPQUOTA);
2479 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
2480 ext4_msg(sb, KERN_ERR, "old and new quota "
2485 if (!sbi->s_jquota_fmt) {
2486 ext4_msg(sb, KERN_ERR, "journaled quota format "
2492 if (test_opt(sb, DIOREAD_NOLOCK)) {
2494 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2495 if (blocksize < PAGE_SIZE)
2496 ext4_msg(sb, KERN_WARNING, "Warning: mounting with an "
2497 "experimental mount option 'dioread_nolock' "
2498 "for blocksize < PAGE_SIZE");
2503 static inline void ext4_show_quota_options(struct seq_file *seq,
2504 struct super_block *sb)
2506 #if defined(CONFIG_QUOTA)
2507 struct ext4_sb_info *sbi = EXT4_SB(sb);
2508 char *usr_qf_name, *grp_qf_name;
2510 if (sbi->s_jquota_fmt) {
2513 switch (sbi->s_jquota_fmt) {
2524 seq_printf(seq, ",jqfmt=%s", fmtname);
2528 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2529 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2531 seq_show_option(seq, "usrjquota", usr_qf_name);
2533 seq_show_option(seq, "grpjquota", grp_qf_name);
2538 static const char *token2str(int token)
2540 const struct match_token *t;
2542 for (t = tokens; t->token != Opt_err; t++)
2543 if (t->token == token && !strchr(t->pattern, '='))
2550 * - it's set to a non-default value OR
2551 * - if the per-sb default is different from the global default
2553 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2556 struct ext4_sb_info *sbi = EXT4_SB(sb);
2557 struct ext4_super_block *es = sbi->s_es;
2558 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2559 const struct mount_opts *m;
2560 char sep = nodefs ? '\n' : ',';
2562 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2563 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2565 if (sbi->s_sb_block != 1)
2566 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2568 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2569 int want_set = m->flags & MOPT_SET;
2570 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2571 (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP)
2573 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2574 continue; /* skip if same as the default */
2576 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2577 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2578 continue; /* select Opt_noFoo vs Opt_Foo */
2579 SEQ_OPTS_PRINT("%s", token2str(m->token));
2582 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2583 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2584 SEQ_OPTS_PRINT("resuid=%u",
2585 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2586 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2587 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2588 SEQ_OPTS_PRINT("resgid=%u",
2589 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2590 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2591 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2592 SEQ_OPTS_PUTS("errors=remount-ro");
2593 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2594 SEQ_OPTS_PUTS("errors=continue");
2595 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2596 SEQ_OPTS_PUTS("errors=panic");
2597 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2598 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2599 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2600 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2601 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2602 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2603 if (sb->s_flags & SB_I_VERSION)
2604 SEQ_OPTS_PUTS("i_version");
2605 if (nodefs || sbi->s_stripe)
2606 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2607 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2608 (sbi->s_mount_opt ^ def_mount_opt)) {
2609 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2610 SEQ_OPTS_PUTS("data=journal");
2611 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2612 SEQ_OPTS_PUTS("data=ordered");
2613 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2614 SEQ_OPTS_PUTS("data=writeback");
2617 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2618 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2619 sbi->s_inode_readahead_blks);
2621 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2622 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2623 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2624 if (nodefs || sbi->s_max_dir_size_kb)
2625 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2626 if (test_opt(sb, DATA_ERR_ABORT))
2627 SEQ_OPTS_PUTS("data_err=abort");
2629 fscrypt_show_test_dummy_encryption(seq, sep, sb);
2631 if (sb->s_flags & SB_INLINECRYPT)
2632 SEQ_OPTS_PUTS("inlinecrypt");
2634 if (test_opt(sb, DAX_ALWAYS)) {
2636 SEQ_OPTS_PUTS("dax");
2638 SEQ_OPTS_PUTS("dax=always");
2639 } else if (test_opt2(sb, DAX_NEVER)) {
2640 SEQ_OPTS_PUTS("dax=never");
2641 } else if (test_opt2(sb, DAX_INODE)) {
2642 SEQ_OPTS_PUTS("dax=inode");
2644 ext4_show_quota_options(seq, sb);
2648 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2650 return _ext4_show_options(seq, root->d_sb, 0);
2653 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2655 struct super_block *sb = seq->private;
2658 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2659 rc = _ext4_show_options(seq, sb, 1);
2660 seq_puts(seq, "\n");
2664 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2667 struct ext4_sb_info *sbi = EXT4_SB(sb);
2670 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2671 ext4_msg(sb, KERN_ERR, "revision level too high, "
2672 "forcing read-only mode");
2678 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2679 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2680 "running e2fsck is recommended");
2681 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2682 ext4_msg(sb, KERN_WARNING,
2683 "warning: mounting fs with errors, "
2684 "running e2fsck is recommended");
2685 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2686 le16_to_cpu(es->s_mnt_count) >=
2687 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2688 ext4_msg(sb, KERN_WARNING,
2689 "warning: maximal mount count reached, "
2690 "running e2fsck is recommended");
2691 else if (le32_to_cpu(es->s_checkinterval) &&
2692 (ext4_get_tstamp(es, s_lastcheck) +
2693 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
2694 ext4_msg(sb, KERN_WARNING,
2695 "warning: checktime reached, "
2696 "running e2fsck is recommended");
2697 if (!sbi->s_journal)
2698 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2699 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2700 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2701 le16_add_cpu(&es->s_mnt_count, 1);
2702 ext4_update_tstamp(es, s_mtime);
2703 if (sbi->s_journal) {
2704 ext4_set_feature_journal_needs_recovery(sb);
2705 if (ext4_has_feature_orphan_file(sb))
2706 ext4_set_feature_orphan_present(sb);
2709 err = ext4_commit_super(sb);
2711 if (test_opt(sb, DEBUG))
2712 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2713 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2715 sbi->s_groups_count,
2716 EXT4_BLOCKS_PER_GROUP(sb),
2717 EXT4_INODES_PER_GROUP(sb),
2718 sbi->s_mount_opt, sbi->s_mount_opt2);
2720 cleancache_init_fs(sb);
2724 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2726 struct ext4_sb_info *sbi = EXT4_SB(sb);
2727 struct flex_groups **old_groups, **new_groups;
2730 if (!sbi->s_log_groups_per_flex)
2733 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2734 if (size <= sbi->s_flex_groups_allocated)
2737 new_groups = kvzalloc(roundup_pow_of_two(size *
2738 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
2740 ext4_msg(sb, KERN_ERR,
2741 "not enough memory for %d flex group pointers", size);
2744 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2745 new_groups[i] = kvzalloc(roundup_pow_of_two(
2746 sizeof(struct flex_groups)),
2748 if (!new_groups[i]) {
2749 for (j = sbi->s_flex_groups_allocated; j < i; j++)
2750 kvfree(new_groups[j]);
2752 ext4_msg(sb, KERN_ERR,
2753 "not enough memory for %d flex groups", size);
2758 old_groups = rcu_dereference(sbi->s_flex_groups);
2760 memcpy(new_groups, old_groups,
2761 (sbi->s_flex_groups_allocated *
2762 sizeof(struct flex_groups *)));
2764 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2765 sbi->s_flex_groups_allocated = size;
2767 ext4_kvfree_array_rcu(old_groups);
2771 static int ext4_fill_flex_info(struct super_block *sb)
2773 struct ext4_sb_info *sbi = EXT4_SB(sb);
2774 struct ext4_group_desc *gdp = NULL;
2775 struct flex_groups *fg;
2776 ext4_group_t flex_group;
2779 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2780 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2781 sbi->s_log_groups_per_flex = 0;
2785 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2789 for (i = 0; i < sbi->s_groups_count; i++) {
2790 gdp = ext4_get_group_desc(sb, i, NULL);
2792 flex_group = ext4_flex_group(sbi, i);
2793 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2794 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
2795 atomic64_add(ext4_free_group_clusters(sb, gdp),
2796 &fg->free_clusters);
2797 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
2805 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2806 struct ext4_group_desc *gdp)
2808 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2810 __le32 le_group = cpu_to_le32(block_group);
2811 struct ext4_sb_info *sbi = EXT4_SB(sb);
2813 if (ext4_has_metadata_csum(sbi->s_sb)) {
2814 /* Use new metadata_csum algorithm */
2816 __u16 dummy_csum = 0;
2818 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2820 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2821 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2822 sizeof(dummy_csum));
2823 offset += sizeof(dummy_csum);
2824 if (offset < sbi->s_desc_size)
2825 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2826 sbi->s_desc_size - offset);
2828 crc = csum32 & 0xFFFF;
2832 /* old crc16 code */
2833 if (!ext4_has_feature_gdt_csum(sb))
2836 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2837 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2838 crc = crc16(crc, (__u8 *)gdp, offset);
2839 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2840 /* for checksum of struct ext4_group_desc do the rest...*/
2841 if (ext4_has_feature_64bit(sb) && offset < sbi->s_desc_size)
2842 crc = crc16(crc, (__u8 *)gdp + offset,
2843 sbi->s_desc_size - offset);
2846 return cpu_to_le16(crc);
2849 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2850 struct ext4_group_desc *gdp)
2852 if (ext4_has_group_desc_csum(sb) &&
2853 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2859 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2860 struct ext4_group_desc *gdp)
2862 if (!ext4_has_group_desc_csum(sb))
2864 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2867 /* Called at mount-time, super-block is locked */
2868 static int ext4_check_descriptors(struct super_block *sb,
2869 ext4_fsblk_t sb_block,
2870 ext4_group_t *first_not_zeroed)
2872 struct ext4_sb_info *sbi = EXT4_SB(sb);
2873 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2874 ext4_fsblk_t last_block;
2875 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2876 ext4_fsblk_t block_bitmap;
2877 ext4_fsblk_t inode_bitmap;
2878 ext4_fsblk_t inode_table;
2879 int flexbg_flag = 0;
2880 ext4_group_t i, grp = sbi->s_groups_count;
2882 if (ext4_has_feature_flex_bg(sb))
2885 ext4_debug("Checking group descriptors");
2887 for (i = 0; i < sbi->s_groups_count; i++) {
2888 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2890 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2891 last_block = ext4_blocks_count(sbi->s_es) - 1;
2893 last_block = first_block +
2894 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2896 if ((grp == sbi->s_groups_count) &&
2897 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2900 block_bitmap = ext4_block_bitmap(sb, gdp);
2901 if (block_bitmap == sb_block) {
2902 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2903 "Block bitmap for group %u overlaps "
2908 if (block_bitmap >= sb_block + 1 &&
2909 block_bitmap <= last_bg_block) {
2910 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2911 "Block bitmap for group %u overlaps "
2912 "block group descriptors", i);
2916 if (block_bitmap < first_block || block_bitmap > last_block) {
2917 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2918 "Block bitmap for group %u not in group "
2919 "(block %llu)!", i, block_bitmap);
2922 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2923 if (inode_bitmap == sb_block) {
2924 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2925 "Inode bitmap for group %u overlaps "
2930 if (inode_bitmap >= sb_block + 1 &&
2931 inode_bitmap <= last_bg_block) {
2932 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2933 "Inode bitmap for group %u overlaps "
2934 "block group descriptors", i);
2938 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2939 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2940 "Inode bitmap for group %u not in group "
2941 "(block %llu)!", i, inode_bitmap);
2944 inode_table = ext4_inode_table(sb, gdp);
2945 if (inode_table == sb_block) {
2946 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2947 "Inode table for group %u overlaps "
2952 if (inode_table >= sb_block + 1 &&
2953 inode_table <= last_bg_block) {
2954 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2955 "Inode table for group %u overlaps "
2956 "block group descriptors", i);
2960 if (inode_table < first_block ||
2961 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2962 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2963 "Inode table for group %u not in group "
2964 "(block %llu)!", i, inode_table);
2967 ext4_lock_group(sb, i);
2968 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2969 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2970 "Checksum for group %u failed (%u!=%u)",
2971 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2972 gdp)), le16_to_cpu(gdp->bg_checksum));
2973 if (!sb_rdonly(sb)) {
2974 ext4_unlock_group(sb, i);
2978 ext4_unlock_group(sb, i);
2980 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2982 if (NULL != first_not_zeroed)
2983 *first_not_zeroed = grp;
2988 * Maximal extent format file size.
2989 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2990 * extent format containers, within a sector_t, and within i_blocks
2991 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2992 * so that won't be a limiting factor.
2994 * However there is other limiting factor. We do store extents in the form
2995 * of starting block and length, hence the resulting length of the extent
2996 * covering maximum file size must fit into on-disk format containers as
2997 * well. Given that length is always by 1 unit bigger than max unit (because
2998 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3000 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3002 static loff_t ext4_max_size(int blkbits, int has_huge_files)
3005 loff_t upper_limit = MAX_LFS_FILESIZE;
3007 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3009 if (!has_huge_files) {
3010 upper_limit = (1LL << 32) - 1;
3012 /* total blocks in file system block size */
3013 upper_limit >>= (blkbits - 9);
3014 upper_limit <<= blkbits;
3018 * 32-bit extent-start container, ee_block. We lower the maxbytes
3019 * by one fs block, so ee_len can cover the extent of maximum file
3022 res = (1LL << 32) - 1;
3025 /* Sanity check against vm- & vfs- imposed limits */
3026 if (res > upper_limit)
3033 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
3034 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3035 * We need to be 1 filesystem block less than the 2^48 sector limit.
3037 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3039 unsigned long long upper_limit, res = EXT4_NDIR_BLOCKS;
3043 * This is calculated to be the largest file size for a dense, block
3044 * mapped file such that the file's total number of 512-byte sectors,
3045 * including data and all indirect blocks, does not exceed (2^48 - 1).
3047 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3048 * number of 512-byte sectors of the file.
3050 if (!has_huge_files) {
3052 * !has_huge_files or implies that the inode i_block field
3053 * represents total file blocks in 2^32 512-byte sectors ==
3054 * size of vfs inode i_blocks * 8
3056 upper_limit = (1LL << 32) - 1;
3058 /* total blocks in file system block size */
3059 upper_limit >>= (bits - 9);
3063 * We use 48 bit ext4_inode i_blocks
3064 * With EXT4_HUGE_FILE_FL set the i_blocks
3065 * represent total number of blocks in
3066 * file system block size
3068 upper_limit = (1LL << 48) - 1;
3072 /* indirect blocks */
3074 /* double indirect blocks */
3075 meta_blocks += 1 + (1LL << (bits-2));
3076 /* tripple indirect blocks */
3077 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
3079 upper_limit -= meta_blocks;
3080 upper_limit <<= bits;
3082 res += 1LL << (bits-2);
3083 res += 1LL << (2*(bits-2));
3084 res += 1LL << (3*(bits-2));
3086 if (res > upper_limit)
3089 if (res > MAX_LFS_FILESIZE)
3090 res = MAX_LFS_FILESIZE;
3095 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3096 ext4_fsblk_t logical_sb_block, int nr)
3098 struct ext4_sb_info *sbi = EXT4_SB(sb);
3099 ext4_group_t bg, first_meta_bg;
3102 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3104 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3105 return logical_sb_block + nr + 1;
3106 bg = sbi->s_desc_per_block * nr;
3107 if (ext4_bg_has_super(sb, bg))
3111 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3112 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3113 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3116 if (sb->s_blocksize == 1024 && nr == 0 &&
3117 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3120 return (has_super + ext4_group_first_block_no(sb, bg));
3124 * ext4_get_stripe_size: Get the stripe size.
3125 * @sbi: In memory super block info
3127 * If we have specified it via mount option, then
3128 * use the mount option value. If the value specified at mount time is
3129 * greater than the blocks per group use the super block value.
3130 * If the super block value is greater than blocks per group return 0.
3131 * Allocator needs it be less than blocks per group.
3134 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3136 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3137 unsigned long stripe_width =
3138 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3141 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3142 ret = sbi->s_stripe;
3143 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3145 else if (stride && stride <= sbi->s_blocks_per_group)
3151 * If the stripe width is 1, this makes no sense and
3152 * we set it to 0 to turn off stripe handling code.
3161 * Check whether this filesystem can be mounted based on
3162 * the features present and the RDONLY/RDWR mount requested.
3163 * Returns 1 if this filesystem can be mounted as requested,
3164 * 0 if it cannot be.
3166 int ext4_feature_set_ok(struct super_block *sb, int readonly)
3168 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3169 ext4_msg(sb, KERN_ERR,
3170 "Couldn't mount because of "
3171 "unsupported optional features (%x)",
3172 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3173 ~EXT4_FEATURE_INCOMPAT_SUPP));
3177 #ifndef CONFIG_UNICODE
3178 if (ext4_has_feature_casefold(sb)) {
3179 ext4_msg(sb, KERN_ERR,
3180 "Filesystem with casefold feature cannot be "
3181 "mounted without CONFIG_UNICODE");
3189 if (ext4_has_feature_readonly(sb)) {
3190 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3191 sb->s_flags |= SB_RDONLY;
3195 /* Check that feature set is OK for a read-write mount */
3196 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3197 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3198 "unsupported optional features (%x)",
3199 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3200 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3203 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3204 ext4_msg(sb, KERN_ERR,
3205 "Can't support bigalloc feature without "
3206 "extents feature\n");
3210 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3211 if (!readonly && (ext4_has_feature_quota(sb) ||
3212 ext4_has_feature_project(sb))) {
3213 ext4_msg(sb, KERN_ERR,
3214 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3217 #endif /* CONFIG_QUOTA */
3222 * This function is called once a day if we have errors logged
3223 * on the file system
3225 static void print_daily_error_info(struct timer_list *t)
3227 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3228 struct super_block *sb = sbi->s_sb;
3229 struct ext4_super_block *es = sbi->s_es;
3231 if (es->s_error_count)
3232 /* fsck newer than v1.41.13 is needed to clean this condition. */
3233 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3234 le32_to_cpu(es->s_error_count));
3235 if (es->s_first_error_time) {
3236 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3238 ext4_get_tstamp(es, s_first_error_time),
3239 (int) sizeof(es->s_first_error_func),
3240 es->s_first_error_func,
3241 le32_to_cpu(es->s_first_error_line));
3242 if (es->s_first_error_ino)
3243 printk(KERN_CONT ": inode %u",
3244 le32_to_cpu(es->s_first_error_ino));
3245 if (es->s_first_error_block)
3246 printk(KERN_CONT ": block %llu", (unsigned long long)
3247 le64_to_cpu(es->s_first_error_block));
3248 printk(KERN_CONT "\n");
3250 if (es->s_last_error_time) {
3251 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3253 ext4_get_tstamp(es, s_last_error_time),
3254 (int) sizeof(es->s_last_error_func),
3255 es->s_last_error_func,
3256 le32_to_cpu(es->s_last_error_line));
3257 if (es->s_last_error_ino)
3258 printk(KERN_CONT ": inode %u",
3259 le32_to_cpu(es->s_last_error_ino));
3260 if (es->s_last_error_block)
3261 printk(KERN_CONT ": block %llu", (unsigned long long)
3262 le64_to_cpu(es->s_last_error_block));
3263 printk(KERN_CONT "\n");
3265 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3268 /* Find next suitable group and run ext4_init_inode_table */
3269 static int ext4_run_li_request(struct ext4_li_request *elr)
3271 struct ext4_group_desc *gdp = NULL;
3272 struct super_block *sb = elr->lr_super;
3273 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3274 ext4_group_t group = elr->lr_next_group;
3275 unsigned int prefetch_ios = 0;
3279 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3280 elr->lr_next_group = ext4_mb_prefetch(sb, group,
3281 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3283 ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3285 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3287 if (group >= elr->lr_next_group) {
3289 if (elr->lr_first_not_zeroed != ngroups &&
3290 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3291 elr->lr_next_group = elr->lr_first_not_zeroed;
3292 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3299 for (; group < ngroups; group++) {
3300 gdp = ext4_get_group_desc(sb, group, NULL);
3306 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3310 if (group >= ngroups)
3314 start_time = ktime_get_real_ns();
3315 ret = ext4_init_inode_table(sb, group,
3316 elr->lr_timeout ? 0 : 1);
3317 trace_ext4_lazy_itable_init(sb, group);
3318 if (elr->lr_timeout == 0) {
3319 elr->lr_timeout = nsecs_to_jiffies((ktime_get_real_ns() - start_time) *
3320 EXT4_SB(elr->lr_super)->s_li_wait_mult);
3322 elr->lr_next_sched = jiffies + elr->lr_timeout;
3323 elr->lr_next_group = group + 1;
3329 * Remove lr_request from the list_request and free the
3330 * request structure. Should be called with li_list_mtx held
3332 static void ext4_remove_li_request(struct ext4_li_request *elr)
3337 list_del(&elr->lr_request);
3338 EXT4_SB(elr->lr_super)->s_li_request = NULL;
3342 static void ext4_unregister_li_request(struct super_block *sb)
3344 mutex_lock(&ext4_li_mtx);
3345 if (!ext4_li_info) {
3346 mutex_unlock(&ext4_li_mtx);
3350 mutex_lock(&ext4_li_info->li_list_mtx);
3351 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3352 mutex_unlock(&ext4_li_info->li_list_mtx);
3353 mutex_unlock(&ext4_li_mtx);
3356 static struct task_struct *ext4_lazyinit_task;
3359 * This is the function where ext4lazyinit thread lives. It walks
3360 * through the request list searching for next scheduled filesystem.
3361 * When such a fs is found, run the lazy initialization request
3362 * (ext4_rn_li_request) and keep track of the time spend in this
3363 * function. Based on that time we compute next schedule time of
3364 * the request. When walking through the list is complete, compute
3365 * next waking time and put itself into sleep.
3367 static int ext4_lazyinit_thread(void *arg)
3369 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3370 struct list_head *pos, *n;
3371 struct ext4_li_request *elr;
3372 unsigned long next_wakeup, cur;
3374 BUG_ON(NULL == eli);
3379 next_wakeup = MAX_JIFFY_OFFSET;
3381 mutex_lock(&eli->li_list_mtx);
3382 if (list_empty(&eli->li_request_list)) {
3383 mutex_unlock(&eli->li_list_mtx);
3386 list_for_each_safe(pos, n, &eli->li_request_list) {
3389 elr = list_entry(pos, struct ext4_li_request,
3392 if (time_before(jiffies, elr->lr_next_sched)) {
3393 if (time_before(elr->lr_next_sched, next_wakeup))
3394 next_wakeup = elr->lr_next_sched;
3397 if (down_read_trylock(&elr->lr_super->s_umount)) {
3398 if (sb_start_write_trylock(elr->lr_super)) {
3401 * We hold sb->s_umount, sb can not
3402 * be removed from the list, it is
3403 * now safe to drop li_list_mtx
3405 mutex_unlock(&eli->li_list_mtx);
3406 err = ext4_run_li_request(elr);
3407 sb_end_write(elr->lr_super);
3408 mutex_lock(&eli->li_list_mtx);
3411 up_read((&elr->lr_super->s_umount));
3413 /* error, remove the lazy_init job */
3415 ext4_remove_li_request(elr);
3419 elr->lr_next_sched = jiffies +
3421 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3423 if (time_before(elr->lr_next_sched, next_wakeup))
3424 next_wakeup = elr->lr_next_sched;
3426 mutex_unlock(&eli->li_list_mtx);
3431 if ((time_after_eq(cur, next_wakeup)) ||
3432 (MAX_JIFFY_OFFSET == next_wakeup)) {
3437 schedule_timeout_interruptible(next_wakeup - cur);
3439 if (kthread_should_stop()) {
3440 ext4_clear_request_list();
3447 * It looks like the request list is empty, but we need
3448 * to check it under the li_list_mtx lock, to prevent any
3449 * additions into it, and of course we should lock ext4_li_mtx
3450 * to atomically free the list and ext4_li_info, because at
3451 * this point another ext4 filesystem could be registering
3454 mutex_lock(&ext4_li_mtx);
3455 mutex_lock(&eli->li_list_mtx);
3456 if (!list_empty(&eli->li_request_list)) {
3457 mutex_unlock(&eli->li_list_mtx);
3458 mutex_unlock(&ext4_li_mtx);
3461 mutex_unlock(&eli->li_list_mtx);
3462 kfree(ext4_li_info);
3463 ext4_li_info = NULL;
3464 mutex_unlock(&ext4_li_mtx);
3469 static void ext4_clear_request_list(void)
3471 struct list_head *pos, *n;
3472 struct ext4_li_request *elr;
3474 mutex_lock(&ext4_li_info->li_list_mtx);
3475 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3476 elr = list_entry(pos, struct ext4_li_request,
3478 ext4_remove_li_request(elr);
3480 mutex_unlock(&ext4_li_info->li_list_mtx);
3483 static int ext4_run_lazyinit_thread(void)
3485 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3486 ext4_li_info, "ext4lazyinit");
3487 if (IS_ERR(ext4_lazyinit_task)) {
3488 int err = PTR_ERR(ext4_lazyinit_task);
3489 ext4_clear_request_list();
3490 kfree(ext4_li_info);
3491 ext4_li_info = NULL;
3492 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3493 "initialization thread\n",
3497 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3502 * Check whether it make sense to run itable init. thread or not.
3503 * If there is at least one uninitialized inode table, return
3504 * corresponding group number, else the loop goes through all
3505 * groups and return total number of groups.
3507 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3509 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3510 struct ext4_group_desc *gdp = NULL;
3512 if (!ext4_has_group_desc_csum(sb))
3515 for (group = 0; group < ngroups; group++) {
3516 gdp = ext4_get_group_desc(sb, group, NULL);
3520 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3527 static int ext4_li_info_new(void)
3529 struct ext4_lazy_init *eli = NULL;
3531 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3535 INIT_LIST_HEAD(&eli->li_request_list);
3536 mutex_init(&eli->li_list_mtx);
3538 eli->li_state |= EXT4_LAZYINIT_QUIT;
3545 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3548 struct ext4_li_request *elr;
3550 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3555 elr->lr_first_not_zeroed = start;
3556 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) {
3557 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3558 elr->lr_next_group = start;
3560 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3564 * Randomize first schedule time of the request to
3565 * spread the inode table initialization requests
3568 elr->lr_next_sched = jiffies + (prandom_u32() %
3569 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3573 int ext4_register_li_request(struct super_block *sb,
3574 ext4_group_t first_not_zeroed)
3576 struct ext4_sb_info *sbi = EXT4_SB(sb);
3577 struct ext4_li_request *elr = NULL;
3578 ext4_group_t ngroups = sbi->s_groups_count;
3581 mutex_lock(&ext4_li_mtx);
3582 if (sbi->s_li_request != NULL) {
3584 * Reset timeout so it can be computed again, because
3585 * s_li_wait_mult might have changed.
3587 sbi->s_li_request->lr_timeout = 0;
3591 if (sb_rdonly(sb) ||
3592 (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) &&
3593 (first_not_zeroed == ngroups || !test_opt(sb, INIT_INODE_TABLE))))
3596 elr = ext4_li_request_new(sb, first_not_zeroed);
3602 if (NULL == ext4_li_info) {
3603 ret = ext4_li_info_new();
3608 mutex_lock(&ext4_li_info->li_list_mtx);
3609 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3610 mutex_unlock(&ext4_li_info->li_list_mtx);
3612 sbi->s_li_request = elr;
3614 * set elr to NULL here since it has been inserted to
3615 * the request_list and the removal and free of it is
3616 * handled by ext4_clear_request_list from now on.
3620 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3621 ret = ext4_run_lazyinit_thread();
3626 mutex_unlock(&ext4_li_mtx);
3633 * We do not need to lock anything since this is called on
3636 static void ext4_destroy_lazyinit_thread(void)
3639 * If thread exited earlier
3640 * there's nothing to be done.
3642 if (!ext4_li_info || !ext4_lazyinit_task)
3645 kthread_stop(ext4_lazyinit_task);
3648 static int set_journal_csum_feature_set(struct super_block *sb)
3651 int compat, incompat;
3652 struct ext4_sb_info *sbi = EXT4_SB(sb);
3654 if (ext4_has_metadata_csum(sb)) {
3655 /* journal checksum v3 */
3657 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3659 /* journal checksum v1 */
3660 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3664 jbd2_journal_clear_features(sbi->s_journal,
3665 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3666 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3667 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3668 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3669 ret = jbd2_journal_set_features(sbi->s_journal,
3671 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3673 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3674 ret = jbd2_journal_set_features(sbi->s_journal,
3677 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3678 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3680 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3681 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3688 * Note: calculating the overhead so we can be compatible with
3689 * historical BSD practice is quite difficult in the face of
3690 * clusters/bigalloc. This is because multiple metadata blocks from
3691 * different block group can end up in the same allocation cluster.
3692 * Calculating the exact overhead in the face of clustered allocation
3693 * requires either O(all block bitmaps) in memory or O(number of block
3694 * groups**2) in time. We will still calculate the superblock for
3695 * older file systems --- and if we come across with a bigalloc file
3696 * system with zero in s_overhead_clusters the estimate will be close to
3697 * correct especially for very large cluster sizes --- but for newer
3698 * file systems, it's better to calculate this figure once at mkfs
3699 * time, and store it in the superblock. If the superblock value is
3700 * present (even for non-bigalloc file systems), we will use it.
3702 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3705 struct ext4_sb_info *sbi = EXT4_SB(sb);
3706 struct ext4_group_desc *gdp;
3707 ext4_fsblk_t first_block, last_block, b;
3708 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3709 int s, j, count = 0;
3710 int has_super = ext4_bg_has_super(sb, grp);
3712 if (!ext4_has_feature_bigalloc(sb))
3713 return (has_super + ext4_bg_num_gdb(sb, grp) +
3714 (has_super ? le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) : 0) +
3715 sbi->s_itb_per_group + 2);
3717 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3718 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3719 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3720 for (i = 0; i < ngroups; i++) {
3721 gdp = ext4_get_group_desc(sb, i, NULL);
3722 b = ext4_block_bitmap(sb, gdp);
3723 if (b >= first_block && b <= last_block) {
3724 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3727 b = ext4_inode_bitmap(sb, gdp);
3728 if (b >= first_block && b <= last_block) {
3729 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3732 b = ext4_inode_table(sb, gdp);
3733 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3734 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3735 int c = EXT4_B2C(sbi, b - first_block);
3736 ext4_set_bit(c, buf);
3742 if (ext4_bg_has_super(sb, grp)) {
3743 ext4_set_bit(s++, buf);
3746 j = ext4_bg_num_gdb(sb, grp);
3747 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3748 ext4_error(sb, "Invalid number of block group "
3749 "descriptor blocks: %d", j);
3750 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3754 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3758 return EXT4_CLUSTERS_PER_GROUP(sb) -
3759 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3763 * Compute the overhead and stash it in sbi->s_overhead
3765 int ext4_calculate_overhead(struct super_block *sb)
3767 struct ext4_sb_info *sbi = EXT4_SB(sb);
3768 struct ext4_super_block *es = sbi->s_es;
3769 struct inode *j_inode;
3770 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3771 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3772 ext4_fsblk_t overhead = 0;
3773 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3779 * Compute the overhead (FS structures). This is constant
3780 * for a given filesystem unless the number of block groups
3781 * changes so we cache the previous value until it does.
3785 * All of the blocks before first_data_block are overhead
3787 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3790 * Add the overhead found in each block group
3792 for (i = 0; i < ngroups; i++) {
3795 blks = count_overhead(sb, i, buf);
3798 memset(buf, 0, PAGE_SIZE);
3803 * Add the internal journal blocks whether the journal has been
3806 if (sbi->s_journal && !sbi->s_journal_bdev)
3807 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
3808 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
3809 /* j_inum for internal journal is non-zero */
3810 j_inode = ext4_get_journal_inode(sb, j_inum);
3812 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3813 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3816 ext4_msg(sb, KERN_ERR, "can't get journal size");
3819 sbi->s_overhead = overhead;
3821 free_page((unsigned long) buf);
3825 static void ext4_set_resv_clusters(struct super_block *sb)
3827 ext4_fsblk_t resv_clusters;
3828 struct ext4_sb_info *sbi = EXT4_SB(sb);
3831 * There's no need to reserve anything when we aren't using extents.
3832 * The space estimates are exact, there are no unwritten extents,
3833 * hole punching doesn't need new metadata... This is needed especially
3834 * to keep ext2/3 backward compatibility.
3836 if (!ext4_has_feature_extents(sb))
3839 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3840 * This should cover the situations where we can not afford to run
3841 * out of space like for example punch hole, or converting
3842 * unwritten extents in delalloc path. In most cases such
3843 * allocation would require 1, or 2 blocks, higher numbers are
3846 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3847 sbi->s_cluster_bits);
3849 do_div(resv_clusters, 50);
3850 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3852 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3855 static const char *ext4_quota_mode(struct super_block *sb)
3858 if (!ext4_quota_capable(sb))
3861 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb))
3862 return "journalled";
3870 static void ext4_setup_csum_trigger(struct super_block *sb,
3871 enum ext4_journal_trigger_type type,
3873 struct jbd2_buffer_trigger_type *type,
3874 struct buffer_head *bh,
3878 struct ext4_sb_info *sbi = EXT4_SB(sb);
3880 sbi->s_journal_triggers[type].sb = sb;
3881 sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger;
3884 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3886 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
3887 char *orig_data = kstrdup(data, GFP_KERNEL);
3888 struct buffer_head *bh, **group_desc;
3889 struct ext4_super_block *es = NULL;
3890 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3891 struct flex_groups **flex_groups;
3893 ext4_fsblk_t sb_block = get_sb_block(&data);
3894 ext4_fsblk_t logical_sb_block;
3895 unsigned long offset = 0;
3896 unsigned long def_mount_opts;
3900 int blocksize, clustersize;
3901 unsigned int db_count;
3903 int needs_recovery, has_huge_files;
3906 ext4_group_t first_not_zeroed;
3907 struct ext4_parsed_options parsed_opts;
3909 /* Set defaults for the variables that will be set during parsing */
3910 parsed_opts.journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3911 parsed_opts.journal_devnum = 0;
3912 parsed_opts.mb_optimize_scan = DEFAULT_MB_OPTIMIZE_SCAN;
3914 if ((data && !orig_data) || !sbi)
3917 sbi->s_daxdev = dax_dev;
3918 sbi->s_blockgroup_lock =
3919 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3920 if (!sbi->s_blockgroup_lock)
3923 sb->s_fs_info = sbi;
3925 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3926 sbi->s_sb_block = sb_block;
3927 sbi->s_sectors_written_start =
3928 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]);
3930 /* Cleanup superblock name */
3931 strreplace(sb->s_id, '/', '!');
3933 /* -EINVAL is default */
3935 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3937 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3942 * The ext4 superblock will not be buffer aligned for other than 1kB
3943 * block sizes. We need to calculate the offset from buffer start.
3945 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3946 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3947 offset = do_div(logical_sb_block, blocksize);
3949 logical_sb_block = sb_block;
3952 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
3954 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3959 * Note: s_es must be initialized as soon as possible because
3960 * some ext4 macro-instructions depend on its value
3962 es = (struct ext4_super_block *) (bh->b_data + offset);
3964 sb->s_magic = le16_to_cpu(es->s_magic);
3965 if (sb->s_magic != EXT4_SUPER_MAGIC)
3967 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3969 /* Warn if metadata_csum and gdt_csum are both set. */
3970 if (ext4_has_feature_metadata_csum(sb) &&
3971 ext4_has_feature_gdt_csum(sb))
3972 ext4_warning(sb, "metadata_csum and uninit_bg are "
3973 "redundant flags; please run fsck.");
3975 /* Check for a known checksum algorithm */
3976 if (!ext4_verify_csum_type(sb, es)) {
3977 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3978 "unknown checksum algorithm.");
3982 ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE,
3983 ext4_orphan_file_block_trigger);
3985 /* Load the checksum driver */
3986 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3987 if (IS_ERR(sbi->s_chksum_driver)) {
3988 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3989 ret = PTR_ERR(sbi->s_chksum_driver);
3990 sbi->s_chksum_driver = NULL;
3994 /* Check superblock checksum */
3995 if (!ext4_superblock_csum_verify(sb, es)) {
3996 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3997 "invalid superblock checksum. Run e2fsck?");
4003 /* Precompute checksum seed for all metadata */
4004 if (ext4_has_feature_csum_seed(sb))
4005 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
4006 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
4007 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
4008 sizeof(es->s_uuid));
4010 /* Set defaults before we parse the mount options */
4011 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4012 set_opt(sb, INIT_INODE_TABLE);
4013 if (def_mount_opts & EXT4_DEFM_DEBUG)
4015 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4017 if (def_mount_opts & EXT4_DEFM_UID16)
4018 set_opt(sb, NO_UID32);
4019 /* xattr user namespace & acls are now defaulted on */
4020 set_opt(sb, XATTR_USER);
4021 #ifdef CONFIG_EXT4_FS_POSIX_ACL
4022 set_opt(sb, POSIX_ACL);
4024 if (ext4_has_feature_fast_commit(sb))
4025 set_opt2(sb, JOURNAL_FAST_COMMIT);
4026 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4027 if (ext4_has_metadata_csum(sb))
4028 set_opt(sb, JOURNAL_CHECKSUM);
4030 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4031 set_opt(sb, JOURNAL_DATA);
4032 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4033 set_opt(sb, ORDERED_DATA);
4034 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4035 set_opt(sb, WRITEBACK_DATA);
4037 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
4038 set_opt(sb, ERRORS_PANIC);
4039 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
4040 set_opt(sb, ERRORS_CONT);
4042 set_opt(sb, ERRORS_RO);
4043 /* block_validity enabled by default; disable with noblock_validity */
4044 set_opt(sb, BLOCK_VALIDITY);
4045 if (def_mount_opts & EXT4_DEFM_DISCARD)
4046 set_opt(sb, DISCARD);
4048 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
4049 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
4050 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
4051 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
4052 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
4054 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4055 set_opt(sb, BARRIER);
4058 * enable delayed allocation by default
4059 * Use -o nodelalloc to turn it off
4061 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4062 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4063 set_opt(sb, DELALLOC);
4066 * set default s_li_wait_mult for lazyinit, for the case there is
4067 * no mount option specified.
4069 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
4071 if (le32_to_cpu(es->s_log_block_size) >
4072 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4073 ext4_msg(sb, KERN_ERR,
4074 "Invalid log block size: %u",
4075 le32_to_cpu(es->s_log_block_size));
4078 if (le32_to_cpu(es->s_log_cluster_size) >
4079 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4080 ext4_msg(sb, KERN_ERR,
4081 "Invalid log cluster size: %u",
4082 le32_to_cpu(es->s_log_cluster_size));
4086 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
4088 if (blocksize == PAGE_SIZE)
4089 set_opt(sb, DIOREAD_NOLOCK);
4091 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4092 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4093 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4095 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4096 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4097 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4098 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4102 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4103 (!is_power_of_2(sbi->s_inode_size)) ||
4104 (sbi->s_inode_size > blocksize)) {
4105 ext4_msg(sb, KERN_ERR,
4106 "unsupported inode size: %d",
4108 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
4112 * i_atime_extra is the last extra field available for
4113 * [acm]times in struct ext4_inode. Checking for that
4114 * field should suffice to ensure we have extra space
4117 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4118 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4119 sb->s_time_gran = 1;
4120 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4122 sb->s_time_gran = NSEC_PER_SEC;
4123 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4125 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4127 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4128 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4129 EXT4_GOOD_OLD_INODE_SIZE;
4130 if (ext4_has_feature_extra_isize(sb)) {
4131 unsigned v, max = (sbi->s_inode_size -
4132 EXT4_GOOD_OLD_INODE_SIZE);
4134 v = le16_to_cpu(es->s_want_extra_isize);
4136 ext4_msg(sb, KERN_ERR,
4137 "bad s_want_extra_isize: %d", v);
4140 if (sbi->s_want_extra_isize < v)
4141 sbi->s_want_extra_isize = v;
4143 v = le16_to_cpu(es->s_min_extra_isize);
4145 ext4_msg(sb, KERN_ERR,
4146 "bad s_min_extra_isize: %d", v);
4149 if (sbi->s_want_extra_isize < v)
4150 sbi->s_want_extra_isize = v;
4154 if (sbi->s_es->s_mount_opts[0]) {
4155 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
4156 sizeof(sbi->s_es->s_mount_opts),
4160 if (!parse_options(s_mount_opts, sb, &parsed_opts, 0)) {
4161 ext4_msg(sb, KERN_WARNING,
4162 "failed to parse options in superblock: %s",
4165 kfree(s_mount_opts);
4167 sbi->s_def_mount_opt = sbi->s_mount_opt;
4168 if (!parse_options((char *) data, sb, &parsed_opts, 0))
4171 #ifdef CONFIG_UNICODE
4172 if (ext4_has_feature_casefold(sb) && !sb->s_encoding) {
4173 const struct ext4_sb_encodings *encoding_info;
4174 struct unicode_map *encoding;
4175 __u16 encoding_flags;
4177 if (ext4_sb_read_encoding(es, &encoding_info,
4179 ext4_msg(sb, KERN_ERR,
4180 "Encoding requested by superblock is unknown");
4184 encoding = utf8_load(encoding_info->version);
4185 if (IS_ERR(encoding)) {
4186 ext4_msg(sb, KERN_ERR,
4187 "can't mount with superblock charset: %s-%s "
4188 "not supported by the kernel. flags: 0x%x.",
4189 encoding_info->name, encoding_info->version,
4193 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4194 "%s-%s with flags 0x%hx", encoding_info->name,
4195 encoding_info->version?:"\b", encoding_flags);
4197 sb->s_encoding = encoding;
4198 sb->s_encoding_flags = encoding_flags;
4202 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4203 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, O_DIRECT and fast_commit support!\n");
4204 /* can't mount with both data=journal and dioread_nolock. */
4205 clear_opt(sb, DIOREAD_NOLOCK);
4206 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4207 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4208 ext4_msg(sb, KERN_ERR, "can't mount with "
4209 "both data=journal and delalloc");
4212 if (test_opt(sb, DAX_ALWAYS)) {
4213 ext4_msg(sb, KERN_ERR, "can't mount with "
4214 "both data=journal and dax");
4217 if (ext4_has_feature_encrypt(sb)) {
4218 ext4_msg(sb, KERN_WARNING,
4219 "encrypted files will use data=ordered "
4220 "instead of data journaling mode");
4222 if (test_opt(sb, DELALLOC))
4223 clear_opt(sb, DELALLOC);
4225 sb->s_iflags |= SB_I_CGROUPWB;
4228 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4229 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
4231 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4232 (ext4_has_compat_features(sb) ||
4233 ext4_has_ro_compat_features(sb) ||
4234 ext4_has_incompat_features(sb)))
4235 ext4_msg(sb, KERN_WARNING,
4236 "feature flags set on rev 0 fs, "
4237 "running e2fsck is recommended");
4239 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4240 set_opt2(sb, HURD_COMPAT);
4241 if (ext4_has_feature_64bit(sb)) {
4242 ext4_msg(sb, KERN_ERR,
4243 "The Hurd can't support 64-bit file systems");
4248 * ea_inode feature uses l_i_version field which is not
4249 * available in HURD_COMPAT mode.
4251 if (ext4_has_feature_ea_inode(sb)) {
4252 ext4_msg(sb, KERN_ERR,
4253 "ea_inode feature is not supported for Hurd");
4258 if (IS_EXT2_SB(sb)) {
4259 if (ext2_feature_set_ok(sb))
4260 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4261 "using the ext4 subsystem");
4264 * If we're probing be silent, if this looks like
4265 * it's actually an ext[34] filesystem.
4267 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4269 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4270 "to feature incompatibilities");
4275 if (IS_EXT3_SB(sb)) {
4276 if (ext3_feature_set_ok(sb))
4277 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4278 "using the ext4 subsystem");
4281 * If we're probing be silent, if this looks like
4282 * it's actually an ext4 filesystem.
4284 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4286 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4287 "to feature incompatibilities");
4293 * Check feature flags regardless of the revision level, since we
4294 * previously didn't change the revision level when setting the flags,
4295 * so there is a chance incompat flags are set on a rev 0 filesystem.
4297 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4300 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4301 ext4_msg(sb, KERN_ERR,
4302 "Number of reserved GDT blocks insanely large: %d",
4303 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4307 if (dax_supported(dax_dev, sb->s_bdev, blocksize, 0,
4308 bdev_nr_sectors(sb->s_bdev)))
4309 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4311 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4312 if (ext4_has_feature_inline_data(sb)) {
4313 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4314 " that may contain inline data");
4317 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4318 ext4_msg(sb, KERN_ERR,
4319 "DAX unsupported by block device.");
4324 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4325 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4326 es->s_encryption_level);
4330 if (sb->s_blocksize != blocksize) {
4332 * bh must be released before kill_bdev(), otherwise
4333 * it won't be freed and its page also. kill_bdev()
4334 * is called by sb_set_blocksize().
4337 /* Validate the filesystem blocksize */
4338 if (!sb_set_blocksize(sb, blocksize)) {
4339 ext4_msg(sb, KERN_ERR, "bad block size %d",
4345 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4346 offset = do_div(logical_sb_block, blocksize);
4347 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4349 ext4_msg(sb, KERN_ERR,
4350 "Can't read superblock on 2nd try");
4355 es = (struct ext4_super_block *)(bh->b_data + offset);
4357 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
4358 ext4_msg(sb, KERN_ERR,
4359 "Magic mismatch, very weird!");
4364 has_huge_files = ext4_has_feature_huge_file(sb);
4365 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4367 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
4369 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
4370 if (ext4_has_feature_64bit(sb)) {
4371 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
4372 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
4373 !is_power_of_2(sbi->s_desc_size)) {
4374 ext4_msg(sb, KERN_ERR,
4375 "unsupported descriptor size %lu",
4380 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
4382 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
4383 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
4385 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
4386 if (sbi->s_inodes_per_block == 0)
4388 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4389 sbi->s_inodes_per_group > blocksize * 8) {
4390 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
4391 sbi->s_inodes_per_group);
4394 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4395 sbi->s_inodes_per_block;
4396 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
4398 sbi->s_mount_state = le16_to_cpu(es->s_state) & ~EXT4_FC_REPLAY;
4399 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4400 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
4402 for (i = 0; i < 4; i++)
4403 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4404 sbi->s_def_hash_version = es->s_def_hash_version;
4405 if (ext4_has_feature_dir_index(sb)) {
4406 i = le32_to_cpu(es->s_flags);
4407 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4408 sbi->s_hash_unsigned = 3;
4409 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
4410 #ifdef __CHAR_UNSIGNED__
4413 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4414 sbi->s_hash_unsigned = 3;
4418 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
4423 /* Handle clustersize */
4424 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4425 if (ext4_has_feature_bigalloc(sb)) {
4426 if (clustersize < blocksize) {
4427 ext4_msg(sb, KERN_ERR,
4428 "cluster size (%d) smaller than "
4429 "block size (%d)", clustersize, blocksize);
4432 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4433 le32_to_cpu(es->s_log_block_size);
4434 sbi->s_clusters_per_group =
4435 le32_to_cpu(es->s_clusters_per_group);
4436 if (sbi->s_clusters_per_group > blocksize * 8) {
4437 ext4_msg(sb, KERN_ERR,
4438 "#clusters per group too big: %lu",
4439 sbi->s_clusters_per_group);
4442 if (sbi->s_blocks_per_group !=
4443 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4444 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4445 "clusters per group (%lu) inconsistent",
4446 sbi->s_blocks_per_group,
4447 sbi->s_clusters_per_group);
4451 if (clustersize != blocksize) {
4452 ext4_msg(sb, KERN_ERR,
4453 "fragment/cluster size (%d) != "
4454 "block size (%d)", clustersize, blocksize);
4457 if (sbi->s_blocks_per_group > blocksize * 8) {
4458 ext4_msg(sb, KERN_ERR,
4459 "#blocks per group too big: %lu",
4460 sbi->s_blocks_per_group);
4463 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4464 sbi->s_cluster_bits = 0;
4466 sbi->s_cluster_ratio = clustersize / blocksize;
4468 /* Do we have standard group size of clustersize * 8 blocks ? */
4469 if (sbi->s_blocks_per_group == clustersize << 3)
4470 set_opt2(sb, STD_GROUP_SIZE);
4473 * Test whether we have more sectors than will fit in sector_t,
4474 * and whether the max offset is addressable by the page cache.
4476 err = generic_check_addressable(sb->s_blocksize_bits,
4477 ext4_blocks_count(es));
4479 ext4_msg(sb, KERN_ERR, "filesystem"
4480 " too large to mount safely on this system");
4484 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4487 /* check blocks count against device size */
4488 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4489 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4490 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4491 "exceeds size of device (%llu blocks)",
4492 ext4_blocks_count(es), blocks_count);
4497 * It makes no sense for the first data block to be beyond the end
4498 * of the filesystem.
4500 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4501 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4502 "block %u is beyond end of filesystem (%llu)",
4503 le32_to_cpu(es->s_first_data_block),
4504 ext4_blocks_count(es));
4507 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4508 (sbi->s_cluster_ratio == 1)) {
4509 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4510 "block is 0 with a 1k block and cluster size");
4514 blocks_count = (ext4_blocks_count(es) -
4515 le32_to_cpu(es->s_first_data_block) +
4516 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4517 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4518 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4519 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4520 "(block count %llu, first data block %u, "
4521 "blocks per group %lu)", blocks_count,
4522 ext4_blocks_count(es),
4523 le32_to_cpu(es->s_first_data_block),
4524 EXT4_BLOCKS_PER_GROUP(sb));
4527 sbi->s_groups_count = blocks_count;
4528 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4529 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4530 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4531 le32_to_cpu(es->s_inodes_count)) {
4532 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4533 le32_to_cpu(es->s_inodes_count),
4534 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4538 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4539 EXT4_DESC_PER_BLOCK(sb);
4540 if (ext4_has_feature_meta_bg(sb)) {
4541 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4542 ext4_msg(sb, KERN_WARNING,
4543 "first meta block group too large: %u "
4544 "(group descriptor block count %u)",
4545 le32_to_cpu(es->s_first_meta_bg), db_count);
4549 rcu_assign_pointer(sbi->s_group_desc,
4550 kvmalloc_array(db_count,
4551 sizeof(struct buffer_head *),
4553 if (sbi->s_group_desc == NULL) {
4554 ext4_msg(sb, KERN_ERR, "not enough memory");
4559 bgl_lock_init(sbi->s_blockgroup_lock);
4561 /* Pre-read the descriptors into the buffer cache */
4562 for (i = 0; i < db_count; i++) {
4563 block = descriptor_loc(sb, logical_sb_block, i);
4564 ext4_sb_breadahead_unmovable(sb, block);
4567 for (i = 0; i < db_count; i++) {
4568 struct buffer_head *bh;
4570 block = descriptor_loc(sb, logical_sb_block, i);
4571 bh = ext4_sb_bread_unmovable(sb, block);
4573 ext4_msg(sb, KERN_ERR,
4574 "can't read group descriptor %d", i);
4580 rcu_dereference(sbi->s_group_desc)[i] = bh;
4583 sbi->s_gdb_count = db_count;
4584 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4585 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4586 ret = -EFSCORRUPTED;
4590 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
4591 spin_lock_init(&sbi->s_error_lock);
4592 INIT_WORK(&sbi->s_error_work, flush_stashed_error_work);
4594 /* Register extent status tree shrinker */
4595 if (ext4_es_register_shrinker(sbi))
4598 sbi->s_stripe = ext4_get_stripe_size(sbi);
4599 sbi->s_extent_max_zeroout_kb = 32;
4602 * set up enough so that it can read an inode
4604 sb->s_op = &ext4_sops;
4605 sb->s_export_op = &ext4_export_ops;
4606 sb->s_xattr = ext4_xattr_handlers;
4607 #ifdef CONFIG_FS_ENCRYPTION
4608 sb->s_cop = &ext4_cryptops;
4610 #ifdef CONFIG_FS_VERITY
4611 sb->s_vop = &ext4_verityops;
4614 sb->dq_op = &ext4_quota_operations;
4615 if (ext4_has_feature_quota(sb))
4616 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4618 sb->s_qcop = &ext4_qctl_operations;
4619 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4621 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4623 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4624 mutex_init(&sbi->s_orphan_lock);
4626 /* Initialize fast commit stuff */
4627 atomic_set(&sbi->s_fc_subtid, 0);
4628 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
4629 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
4630 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
4631 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
4632 sbi->s_fc_bytes = 0;
4633 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
4634 sbi->s_fc_ineligible_tid = 0;
4635 spin_lock_init(&sbi->s_fc_lock);
4636 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
4637 sbi->s_fc_replay_state.fc_regions = NULL;
4638 sbi->s_fc_replay_state.fc_regions_size = 0;
4639 sbi->s_fc_replay_state.fc_regions_used = 0;
4640 sbi->s_fc_replay_state.fc_regions_valid = 0;
4641 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
4642 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
4643 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
4647 needs_recovery = (es->s_last_orphan != 0 ||
4648 ext4_has_feature_orphan_present(sb) ||
4649 ext4_has_feature_journal_needs_recovery(sb));
4651 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) {
4652 err = ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block));
4654 goto failed_mount3a;
4658 * The first inode we look at is the journal inode. Don't try
4659 * root first: it may be modified in the journal!
4661 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4662 err = ext4_load_journal(sb, es, parsed_opts.journal_devnum);
4664 goto failed_mount3a;
4665 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4666 ext4_has_feature_journal_needs_recovery(sb)) {
4667 ext4_msg(sb, KERN_ERR, "required journal recovery "
4668 "suppressed and not mounted read-only");
4669 goto failed_mount3a;
4671 /* Nojournal mode, all journal mount options are illegal */
4672 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4673 ext4_msg(sb, KERN_ERR, "can't mount with "
4674 "journal_async_commit, fs mounted w/o journal");
4675 goto failed_mount3a;
4678 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4679 ext4_msg(sb, KERN_ERR, "can't mount with "
4680 "journal_checksum, fs mounted w/o journal");
4681 goto failed_mount3a;
4683 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4684 ext4_msg(sb, KERN_ERR, "can't mount with "
4685 "commit=%lu, fs mounted w/o journal",
4686 sbi->s_commit_interval / HZ);
4687 goto failed_mount3a;
4689 if (EXT4_MOUNT_DATA_FLAGS &
4690 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4691 ext4_msg(sb, KERN_ERR, "can't mount with "
4692 "data=, fs mounted w/o journal");
4693 goto failed_mount3a;
4695 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
4696 clear_opt(sb, JOURNAL_CHECKSUM);
4697 clear_opt(sb, DATA_FLAGS);
4698 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4699 sbi->s_journal = NULL;
4704 if (ext4_has_feature_64bit(sb) &&
4705 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4706 JBD2_FEATURE_INCOMPAT_64BIT)) {
4707 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4708 goto failed_mount_wq;
4711 if (!set_journal_csum_feature_set(sb)) {
4712 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4714 goto failed_mount_wq;
4717 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
4718 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4719 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
4720 ext4_msg(sb, KERN_ERR,
4721 "Failed to set fast commit journal feature");
4722 goto failed_mount_wq;
4725 /* We have now updated the journal if required, so we can
4726 * validate the data journaling mode. */
4727 switch (test_opt(sb, DATA_FLAGS)) {
4729 /* No mode set, assume a default based on the journal
4730 * capabilities: ORDERED_DATA if the journal can
4731 * cope, else JOURNAL_DATA
4733 if (jbd2_journal_check_available_features
4734 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4735 set_opt(sb, ORDERED_DATA);
4736 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4738 set_opt(sb, JOURNAL_DATA);
4739 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4743 case EXT4_MOUNT_ORDERED_DATA:
4744 case EXT4_MOUNT_WRITEBACK_DATA:
4745 if (!jbd2_journal_check_available_features
4746 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4747 ext4_msg(sb, KERN_ERR, "Journal does not support "
4748 "requested data journaling mode");
4749 goto failed_mount_wq;
4756 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4757 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4758 ext4_msg(sb, KERN_ERR, "can't mount with "
4759 "journal_async_commit in data=ordered mode");
4760 goto failed_mount_wq;
4763 set_task_ioprio(sbi->s_journal->j_task, parsed_opts.journal_ioprio);
4765 sbi->s_journal->j_submit_inode_data_buffers =
4766 ext4_journal_submit_inode_data_buffers;
4767 sbi->s_journal->j_finish_inode_data_buffers =
4768 ext4_journal_finish_inode_data_buffers;
4771 if (!test_opt(sb, NO_MBCACHE)) {
4772 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4773 if (!sbi->s_ea_block_cache) {
4774 ext4_msg(sb, KERN_ERR,
4775 "Failed to create ea_block_cache");
4776 goto failed_mount_wq;
4779 if (ext4_has_feature_ea_inode(sb)) {
4780 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4781 if (!sbi->s_ea_inode_cache) {
4782 ext4_msg(sb, KERN_ERR,
4783 "Failed to create ea_inode_cache");
4784 goto failed_mount_wq;
4789 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
4790 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
4791 goto failed_mount_wq;
4795 * Get the # of file system overhead blocks from the
4796 * superblock if present.
4798 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4799 /* ignore the precalculated value if it is ridiculous */
4800 if (sbi->s_overhead > ext4_blocks_count(es))
4801 sbi->s_overhead = 0;
4803 * If the bigalloc feature is not enabled recalculating the
4804 * overhead doesn't take long, so we might as well just redo
4805 * it to make sure we are using the correct value.
4807 if (!ext4_has_feature_bigalloc(sb))
4808 sbi->s_overhead = 0;
4809 if (sbi->s_overhead == 0) {
4810 err = ext4_calculate_overhead(sb);
4812 goto failed_mount_wq;
4816 * The maximum number of concurrent works can be high and
4817 * concurrency isn't really necessary. Limit it to 1.
4819 EXT4_SB(sb)->rsv_conversion_wq =
4820 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4821 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4822 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4828 * The jbd2_journal_load will have done any necessary log recovery,
4829 * so we can safely mount the rest of the filesystem now.
4832 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
4834 ext4_msg(sb, KERN_ERR, "get root inode failed");
4835 ret = PTR_ERR(root);
4839 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4840 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4845 sb->s_root = d_make_root(root);
4847 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4852 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4853 if (ret == -EROFS) {
4854 sb->s_flags |= SB_RDONLY;
4857 goto failed_mount4a;
4859 ext4_set_resv_clusters(sb);
4861 if (test_opt(sb, BLOCK_VALIDITY)) {
4862 err = ext4_setup_system_zone(sb);
4864 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4866 goto failed_mount4a;
4869 ext4_fc_replay_cleanup(sb);
4874 * Enable optimize_scan if number of groups is > threshold. This can be
4875 * turned off by passing "mb_optimize_scan=0". This can also be
4876 * turned on forcefully by passing "mb_optimize_scan=1".
4878 if (parsed_opts.mb_optimize_scan == 1)
4879 set_opt2(sb, MB_OPTIMIZE_SCAN);
4880 else if (parsed_opts.mb_optimize_scan == 0)
4881 clear_opt2(sb, MB_OPTIMIZE_SCAN);
4882 else if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD)
4883 set_opt2(sb, MB_OPTIMIZE_SCAN);
4885 err = ext4_mb_init(sb);
4887 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4893 * We can only set up the journal commit callback once
4894 * mballoc is initialized
4897 sbi->s_journal->j_commit_callback =
4898 ext4_journal_commit_callback;
4900 block = ext4_count_free_clusters(sb);
4901 ext4_free_blocks_count_set(sbi->s_es,
4902 EXT4_C2B(sbi, block));
4903 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4906 unsigned long freei = ext4_count_free_inodes(sb);
4907 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4908 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4912 err = percpu_counter_init(&sbi->s_dirs_counter,
4913 ext4_count_dirs(sb), GFP_KERNEL);
4915 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4918 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0,
4921 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
4924 ext4_msg(sb, KERN_ERR, "insufficient memory");
4928 if (ext4_has_feature_flex_bg(sb))
4929 if (!ext4_fill_flex_info(sb)) {
4930 ext4_msg(sb, KERN_ERR,
4931 "unable to initialize "
4932 "flex_bg meta info!");
4937 err = ext4_register_li_request(sb, first_not_zeroed);
4941 err = ext4_register_sysfs(sb);
4945 err = ext4_init_orphan_info(sb);
4949 /* Enable quota usage during mount. */
4950 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
4951 err = ext4_enable_quotas(sb);
4955 #endif /* CONFIG_QUOTA */
4958 * Save the original bdev mapping's wb_err value which could be
4959 * used to detect the metadata async write error.
4961 spin_lock_init(&sbi->s_bdev_wb_lock);
4962 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
4963 &sbi->s_bdev_wb_err);
4964 sb->s_bdev->bd_super = sb;
4965 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4966 ext4_orphan_cleanup(sb, es);
4967 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4969 * Update the checksum after updating free space/inode counters and
4970 * ext4_orphan_cleanup. Otherwise the superblock can have an incorrect
4971 * checksum in the buffer cache until it is written out and
4972 * e2fsprogs programs trying to open a file system immediately
4973 * after it is mounted can fail.
4975 ext4_superblock_csum_set(sb);
4976 if (needs_recovery) {
4977 ext4_msg(sb, KERN_INFO, "recovery complete");
4978 err = ext4_mark_recovery_complete(sb, es);
4980 goto failed_mount10;
4982 if (EXT4_SB(sb)->s_journal) {
4983 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4984 descr = " journalled data mode";
4985 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4986 descr = " ordered data mode";
4988 descr = " writeback data mode";
4990 descr = "out journal";
4992 if (test_opt(sb, DISCARD)) {
4993 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4994 if (!blk_queue_discard(q))
4995 ext4_msg(sb, KERN_WARNING,
4996 "mounting with \"discard\" option, but "
4997 "the device does not support discard");
5000 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
5001 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5002 "Opts: %.*s%s%s. Quota mode: %s.", descr,
5003 (int) sizeof(sbi->s_es->s_mount_opts),
5004 sbi->s_es->s_mount_opts,
5005 *sbi->s_es->s_mount_opts ? "; " : "", orig_data,
5006 ext4_quota_mode(sb));
5008 if (es->s_error_count)
5009 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
5011 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5012 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
5013 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
5014 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
5015 atomic_set(&sbi->s_warning_count, 0);
5016 atomic_set(&sbi->s_msg_count, 0);
5023 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5027 ext4_quota_off_umount(sb);
5028 failed_mount9: __maybe_unused
5029 ext4_release_orphan_info(sb);
5031 ext4_unregister_sysfs(sb);
5032 kobject_put(&sbi->s_kobj);
5034 ext4_unregister_li_request(sb);
5036 ext4_mb_release(sb);
5038 flex_groups = rcu_dereference(sbi->s_flex_groups);
5040 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
5041 kvfree(flex_groups[i]);
5042 kvfree(flex_groups);
5045 percpu_counter_destroy(&sbi->s_freeclusters_counter);
5046 percpu_counter_destroy(&sbi->s_freeinodes_counter);
5047 percpu_counter_destroy(&sbi->s_dirs_counter);
5048 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
5049 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
5050 percpu_free_rwsem(&sbi->s_writepages_rwsem);
5052 ext4_ext_release(sb);
5053 ext4_release_system_zone(sb);
5058 ext4_msg(sb, KERN_ERR, "mount failed");
5059 if (EXT4_SB(sb)->rsv_conversion_wq)
5060 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
5062 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5063 sbi->s_ea_inode_cache = NULL;
5065 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5066 sbi->s_ea_block_cache = NULL;
5068 if (sbi->s_journal) {
5069 /* flush s_error_work before journal destroy. */
5070 flush_work(&sbi->s_error_work);
5071 jbd2_journal_destroy(sbi->s_journal);
5072 sbi->s_journal = NULL;
5075 ext4_es_unregister_shrinker(sbi);
5077 /* flush s_error_work before sbi destroy */
5078 flush_work(&sbi->s_error_work);
5079 del_timer_sync(&sbi->s_err_report);
5080 ext4_stop_mmpd(sbi);
5083 group_desc = rcu_dereference(sbi->s_group_desc);
5084 for (i = 0; i < db_count; i++)
5085 brelse(group_desc[i]);
5089 if (sbi->s_chksum_driver)
5090 crypto_free_shash(sbi->s_chksum_driver);
5092 #ifdef CONFIG_UNICODE
5093 utf8_unload(sb->s_encoding);
5097 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5098 kfree(get_qf_name(sb, sbi, i));
5100 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
5101 /* ext4_blkdev_remove() calls kill_bdev(), release bh before it. */
5103 ext4_blkdev_remove(sbi);
5105 invalidate_bdev(sb->s_bdev);
5106 sb->s_fs_info = NULL;
5107 kfree(sbi->s_blockgroup_lock);
5111 fs_put_dax(dax_dev);
5112 return err ? err : ret;
5116 * Setup any per-fs journal parameters now. We'll do this both on
5117 * initial mount, once the journal has been initialised but before we've
5118 * done any recovery; and again on any subsequent remount.
5120 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
5122 struct ext4_sb_info *sbi = EXT4_SB(sb);
5124 journal->j_commit_interval = sbi->s_commit_interval;
5125 journal->j_min_batch_time = sbi->s_min_batch_time;
5126 journal->j_max_batch_time = sbi->s_max_batch_time;
5127 ext4_fc_init(sb, journal);
5129 write_lock(&journal->j_state_lock);
5130 if (test_opt(sb, BARRIER))
5131 journal->j_flags |= JBD2_BARRIER;
5133 journal->j_flags &= ~JBD2_BARRIER;
5134 if (test_opt(sb, DATA_ERR_ABORT))
5135 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5137 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
5138 write_unlock(&journal->j_state_lock);
5141 static struct inode *ext4_get_journal_inode(struct super_block *sb,
5142 unsigned int journal_inum)
5144 struct inode *journal_inode;
5147 * Test for the existence of a valid inode on disk. Bad things
5148 * happen if we iget() an unused inode, as the subsequent iput()
5149 * will try to delete it.
5151 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
5152 if (IS_ERR(journal_inode)) {
5153 ext4_msg(sb, KERN_ERR, "no journal found");
5156 if (!journal_inode->i_nlink) {
5157 make_bad_inode(journal_inode);
5158 iput(journal_inode);
5159 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
5163 ext4_debug("Journal inode found at %p: %lld bytes\n",
5164 journal_inode, journal_inode->i_size);
5165 if (!S_ISREG(journal_inode->i_mode) || IS_ENCRYPTED(journal_inode)) {
5166 ext4_msg(sb, KERN_ERR, "invalid journal inode");
5167 iput(journal_inode);
5170 return journal_inode;
5173 static journal_t *ext4_get_journal(struct super_block *sb,
5174 unsigned int journal_inum)
5176 struct inode *journal_inode;
5179 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5182 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5186 journal = jbd2_journal_init_inode(journal_inode);
5188 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5189 iput(journal_inode);
5192 journal->j_private = sb;
5193 ext4_init_journal_params(sb, journal);
5197 static journal_t *ext4_get_dev_journal(struct super_block *sb,
5200 struct buffer_head *bh;
5204 int hblock, blocksize;
5205 ext4_fsblk_t sb_block;
5206 unsigned long offset;
5207 struct ext4_super_block *es;
5208 struct block_device *bdev;
5210 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5213 bdev = ext4_blkdev_get(j_dev, sb);
5217 blocksize = sb->s_blocksize;
5218 hblock = bdev_logical_block_size(bdev);
5219 if (blocksize < hblock) {
5220 ext4_msg(sb, KERN_ERR,
5221 "blocksize too small for journal device");
5225 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5226 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5227 set_blocksize(bdev, blocksize);
5228 if (!(bh = __bread(bdev, sb_block, blocksize))) {
5229 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5230 "external journal");
5234 es = (struct ext4_super_block *) (bh->b_data + offset);
5235 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5236 !(le32_to_cpu(es->s_feature_incompat) &
5237 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5238 ext4_msg(sb, KERN_ERR, "external journal has "
5244 if ((le32_to_cpu(es->s_feature_ro_compat) &
5245 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5246 es->s_checksum != ext4_superblock_csum(sb, es)) {
5247 ext4_msg(sb, KERN_ERR, "external journal has "
5248 "corrupt superblock");
5253 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5254 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5259 len = ext4_blocks_count(es);
5260 start = sb_block + 1;
5261 brelse(bh); /* we're done with the superblock */
5263 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
5264 start, len, blocksize);
5266 ext4_msg(sb, KERN_ERR, "failed to create device journal");
5269 journal->j_private = sb;
5270 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) {
5271 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
5274 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5275 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5276 "user (unsupported) - %d",
5277 be32_to_cpu(journal->j_superblock->s_nr_users));
5280 EXT4_SB(sb)->s_journal_bdev = bdev;
5281 ext4_init_journal_params(sb, journal);
5285 jbd2_journal_destroy(journal);
5287 ext4_blkdev_put(bdev);
5291 static int ext4_load_journal(struct super_block *sb,
5292 struct ext4_super_block *es,
5293 unsigned long journal_devnum)
5296 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5299 int really_read_only;
5302 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5303 return -EFSCORRUPTED;
5305 if (journal_devnum &&
5306 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5307 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5308 "numbers have changed");
5309 journal_dev = new_decode_dev(journal_devnum);
5311 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5313 if (journal_inum && journal_dev) {
5314 ext4_msg(sb, KERN_ERR,
5315 "filesystem has both journal inode and journal device!");
5320 journal = ext4_get_journal(sb, journal_inum);
5324 journal = ext4_get_dev_journal(sb, journal_dev);
5329 journal_dev_ro = bdev_read_only(journal->j_dev);
5330 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5332 if (journal_dev_ro && !sb_rdonly(sb)) {
5333 ext4_msg(sb, KERN_ERR,
5334 "journal device read-only, try mounting with '-o ro'");
5340 * Are we loading a blank journal or performing recovery after a
5341 * crash? For recovery, we need to check in advance whether we
5342 * can get read-write access to the device.
5344 if (ext4_has_feature_journal_needs_recovery(sb)) {
5345 if (sb_rdonly(sb)) {
5346 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5347 "required on readonly filesystem");
5348 if (really_read_only) {
5349 ext4_msg(sb, KERN_ERR, "write access "
5350 "unavailable, cannot proceed "
5351 "(try mounting with noload)");
5355 ext4_msg(sb, KERN_INFO, "write access will "
5356 "be enabled during recovery");
5360 if (!(journal->j_flags & JBD2_BARRIER))
5361 ext4_msg(sb, KERN_INFO, "barriers disabled");
5363 if (!ext4_has_feature_journal_needs_recovery(sb))
5364 err = jbd2_journal_wipe(journal, !really_read_only);
5366 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5368 memcpy(save, ((char *) es) +
5369 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
5370 err = jbd2_journal_load(journal);
5372 memcpy(((char *) es) + EXT4_S_ERR_START,
5373 save, EXT4_S_ERR_LEN);
5378 ext4_msg(sb, KERN_ERR, "error loading journal");
5382 EXT4_SB(sb)->s_journal = journal;
5383 err = ext4_clear_journal_err(sb, es);
5385 EXT4_SB(sb)->s_journal = NULL;
5386 jbd2_journal_destroy(journal);
5390 if (!really_read_only && journal_devnum &&
5391 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5392 es->s_journal_dev = cpu_to_le32(journal_devnum);
5393 ext4_commit_super(sb);
5395 if (!really_read_only && journal_inum &&
5396 journal_inum != le32_to_cpu(es->s_journal_inum)) {
5397 es->s_journal_inum = cpu_to_le32(journal_inum);
5398 ext4_commit_super(sb);
5404 jbd2_journal_destroy(journal);
5408 /* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */
5409 static void ext4_update_super(struct super_block *sb)
5411 struct ext4_sb_info *sbi = EXT4_SB(sb);
5412 struct ext4_super_block *es = sbi->s_es;
5413 struct buffer_head *sbh = sbi->s_sbh;
5417 * If the file system is mounted read-only, don't update the
5418 * superblock write time. This avoids updating the superblock
5419 * write time when we are mounting the root file system
5420 * read/only but we need to replay the journal; at that point,
5421 * for people who are east of GMT and who make their clock
5422 * tick in localtime for Windows bug-for-bug compatibility,
5423 * the clock is set in the future, and this will cause e2fsck
5424 * to complain and force a full file system check.
5426 if (!(sb->s_flags & SB_RDONLY))
5427 ext4_update_tstamp(es, s_wtime);
5428 es->s_kbytes_written =
5429 cpu_to_le64(sbi->s_kbytes_written +
5430 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
5431 sbi->s_sectors_written_start) >> 1));
5432 if (percpu_counter_initialized(&sbi->s_freeclusters_counter))
5433 ext4_free_blocks_count_set(es,
5434 EXT4_C2B(sbi, percpu_counter_sum_positive(
5435 &sbi->s_freeclusters_counter)));
5436 if (percpu_counter_initialized(&sbi->s_freeinodes_counter))
5437 es->s_free_inodes_count =
5438 cpu_to_le32(percpu_counter_sum_positive(
5439 &sbi->s_freeinodes_counter));
5440 /* Copy error information to the on-disk superblock */
5441 spin_lock(&sbi->s_error_lock);
5442 if (sbi->s_add_error_count > 0) {
5443 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5444 if (!es->s_first_error_time && !es->s_first_error_time_hi) {
5445 __ext4_update_tstamp(&es->s_first_error_time,
5446 &es->s_first_error_time_hi,
5447 sbi->s_first_error_time);
5448 strncpy(es->s_first_error_func, sbi->s_first_error_func,
5449 sizeof(es->s_first_error_func));
5450 es->s_first_error_line =
5451 cpu_to_le32(sbi->s_first_error_line);
5452 es->s_first_error_ino =
5453 cpu_to_le32(sbi->s_first_error_ino);
5454 es->s_first_error_block =
5455 cpu_to_le64(sbi->s_first_error_block);
5456 es->s_first_error_errcode =
5457 ext4_errno_to_code(sbi->s_first_error_code);
5459 __ext4_update_tstamp(&es->s_last_error_time,
5460 &es->s_last_error_time_hi,
5461 sbi->s_last_error_time);
5462 strncpy(es->s_last_error_func, sbi->s_last_error_func,
5463 sizeof(es->s_last_error_func));
5464 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line);
5465 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino);
5466 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block);
5467 es->s_last_error_errcode =
5468 ext4_errno_to_code(sbi->s_last_error_code);
5470 * Start the daily error reporting function if it hasn't been
5473 if (!es->s_error_count)
5474 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);
5475 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count);
5476 sbi->s_add_error_count = 0;
5478 spin_unlock(&sbi->s_error_lock);
5480 ext4_superblock_csum_set(sb);
5484 static int ext4_commit_super(struct super_block *sb)
5486 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
5491 if (block_device_ejected(sb))
5494 ext4_update_super(sb);
5496 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
5498 * Oh, dear. A previous attempt to write the
5499 * superblock failed. This could happen because the
5500 * USB device was yanked out. Or it could happen to
5501 * be a transient write error and maybe the block will
5502 * be remapped. Nothing we can do but to retry the
5503 * write and hope for the best.
5505 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5506 "superblock detected");
5507 clear_buffer_write_io_error(sbh);
5508 set_buffer_uptodate(sbh);
5510 BUFFER_TRACE(sbh, "marking dirty");
5511 mark_buffer_dirty(sbh);
5512 error = __sync_dirty_buffer(sbh,
5513 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
5514 if (buffer_write_io_error(sbh)) {
5515 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5517 clear_buffer_write_io_error(sbh);
5518 set_buffer_uptodate(sbh);
5524 * Have we just finished recovery? If so, and if we are mounting (or
5525 * remounting) the filesystem readonly, then we will end up with a
5526 * consistent fs on disk. Record that fact.
5528 static int ext4_mark_recovery_complete(struct super_block *sb,
5529 struct ext4_super_block *es)
5532 journal_t *journal = EXT4_SB(sb)->s_journal;
5534 if (!ext4_has_feature_journal(sb)) {
5535 if (journal != NULL) {
5536 ext4_error(sb, "Journal got removed while the fs was "
5538 return -EFSCORRUPTED;
5542 jbd2_journal_lock_updates(journal);
5543 err = jbd2_journal_flush(journal, 0);
5547 if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) ||
5548 ext4_has_feature_orphan_present(sb))) {
5549 if (!ext4_orphan_file_empty(sb)) {
5550 ext4_error(sb, "Orphan file not empty on read-only fs.");
5551 err = -EFSCORRUPTED;
5554 ext4_clear_feature_journal_needs_recovery(sb);
5555 ext4_clear_feature_orphan_present(sb);
5556 ext4_commit_super(sb);
5559 jbd2_journal_unlock_updates(journal);
5564 * If we are mounting (or read-write remounting) a filesystem whose journal
5565 * has recorded an error from a previous lifetime, move that error to the
5566 * main filesystem now.
5568 static int ext4_clear_journal_err(struct super_block *sb,
5569 struct ext4_super_block *es)
5575 if (!ext4_has_feature_journal(sb)) {
5576 ext4_error(sb, "Journal got removed while the fs was mounted!");
5577 return -EFSCORRUPTED;
5580 journal = EXT4_SB(sb)->s_journal;
5583 * Now check for any error status which may have been recorded in the
5584 * journal by a prior ext4_error() or ext4_abort()
5587 j_errno = jbd2_journal_errno(journal);
5591 errstr = ext4_decode_error(sb, j_errno, nbuf);
5592 ext4_warning(sb, "Filesystem error recorded "
5593 "from previous mount: %s", errstr);
5594 ext4_warning(sb, "Marking fs in need of filesystem check.");
5596 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5597 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5598 ext4_commit_super(sb);
5600 jbd2_journal_clear_err(journal);
5601 jbd2_journal_update_sb_errno(journal);
5607 * Force the running and committing transactions to commit,
5608 * and wait on the commit.
5610 int ext4_force_commit(struct super_block *sb)
5617 journal = EXT4_SB(sb)->s_journal;
5618 return ext4_journal_force_commit(journal);
5621 static int ext4_sync_fs(struct super_block *sb, int wait)
5625 bool needs_barrier = false;
5626 struct ext4_sb_info *sbi = EXT4_SB(sb);
5628 if (unlikely(ext4_forced_shutdown(sbi)))
5631 trace_ext4_sync_fs(sb, wait);
5632 flush_workqueue(sbi->rsv_conversion_wq);
5634 * Writeback quota in non-journalled quota case - journalled quota has
5637 dquot_writeback_dquots(sb, -1);
5639 * Data writeback is possible w/o journal transaction, so barrier must
5640 * being sent at the end of the function. But we can skip it if
5641 * transaction_commit will do it for us.
5643 if (sbi->s_journal) {
5644 target = jbd2_get_latest_transaction(sbi->s_journal);
5645 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5646 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5647 needs_barrier = true;
5649 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5651 ret = jbd2_log_wait_commit(sbi->s_journal,
5654 } else if (wait && test_opt(sb, BARRIER))
5655 needs_barrier = true;
5656 if (needs_barrier) {
5658 err = blkdev_issue_flush(sb->s_bdev);
5667 * LVM calls this function before a (read-only) snapshot is created. This
5668 * gives us a chance to flush the journal completely and mark the fs clean.
5670 * Note that only this function cannot bring a filesystem to be in a clean
5671 * state independently. It relies on upper layer to stop all data & metadata
5674 static int ext4_freeze(struct super_block *sb)
5682 journal = EXT4_SB(sb)->s_journal;
5685 /* Now we set up the journal barrier. */
5686 jbd2_journal_lock_updates(journal);
5689 * Don't clear the needs_recovery flag if we failed to
5690 * flush the journal.
5692 error = jbd2_journal_flush(journal, 0);
5696 /* Journal blocked and flushed, clear needs_recovery flag. */
5697 ext4_clear_feature_journal_needs_recovery(sb);
5698 if (ext4_orphan_file_empty(sb))
5699 ext4_clear_feature_orphan_present(sb);
5702 error = ext4_commit_super(sb);
5705 /* we rely on upper layer to stop further updates */
5706 jbd2_journal_unlock_updates(journal);
5711 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5712 * flag here, even though the filesystem is not technically dirty yet.
5714 static int ext4_unfreeze(struct super_block *sb)
5716 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
5719 if (EXT4_SB(sb)->s_journal) {
5720 /* Reset the needs_recovery flag before the fs is unlocked. */
5721 ext4_set_feature_journal_needs_recovery(sb);
5722 if (ext4_has_feature_orphan_file(sb))
5723 ext4_set_feature_orphan_present(sb);
5726 ext4_commit_super(sb);
5731 * Structure to save mount options for ext4_remount's benefit
5733 struct ext4_mount_options {
5734 unsigned long s_mount_opt;
5735 unsigned long s_mount_opt2;
5738 unsigned long s_commit_interval;
5739 u32 s_min_batch_time, s_max_batch_time;
5742 char *s_qf_names[EXT4_MAXQUOTAS];
5746 static int ext4_remount(struct super_block *sb, int *flags, char *data)
5748 struct ext4_super_block *es;
5749 struct ext4_sb_info *sbi = EXT4_SB(sb);
5750 unsigned long old_sb_flags, vfs_flags;
5751 struct ext4_mount_options old_opts;
5755 int enable_quota = 0;
5757 char *to_free[EXT4_MAXQUOTAS];
5759 char *orig_data = kstrdup(data, GFP_KERNEL);
5760 struct ext4_parsed_options parsed_opts;
5762 parsed_opts.journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5763 parsed_opts.journal_devnum = 0;
5765 if (data && !orig_data)
5768 /* Store the original options */
5769 old_sb_flags = sb->s_flags;
5770 old_opts.s_mount_opt = sbi->s_mount_opt;
5771 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
5772 old_opts.s_resuid = sbi->s_resuid;
5773 old_opts.s_resgid = sbi->s_resgid;
5774 old_opts.s_commit_interval = sbi->s_commit_interval;
5775 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5776 old_opts.s_max_batch_time = sbi->s_max_batch_time;
5778 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
5779 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5780 if (sbi->s_qf_names[i]) {
5781 char *qf_name = get_qf_name(sb, sbi, i);
5783 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
5784 if (!old_opts.s_qf_names[i]) {
5785 for (j = 0; j < i; j++)
5786 kfree(old_opts.s_qf_names[j]);
5791 old_opts.s_qf_names[i] = NULL;
5793 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5794 parsed_opts.journal_ioprio =
5795 sbi->s_journal->j_task->io_context->ioprio;
5798 * Some options can be enabled by ext4 and/or by VFS mount flag
5799 * either way we need to make sure it matches in both *flags and
5800 * s_flags. Copy those selected flags from *flags to s_flags
5802 vfs_flags = SB_LAZYTIME | SB_I_VERSION;
5803 sb->s_flags = (sb->s_flags & ~vfs_flags) | (*flags & vfs_flags);
5805 if (!parse_options(data, sb, &parsed_opts, 1)) {
5810 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5811 test_opt(sb, JOURNAL_CHECKSUM)) {
5812 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5813 "during remount not supported; ignoring");
5814 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5817 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5818 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5819 ext4_msg(sb, KERN_ERR, "can't mount with "
5820 "both data=journal and delalloc");
5824 if (test_opt(sb, DIOREAD_NOLOCK)) {
5825 ext4_msg(sb, KERN_ERR, "can't mount with "
5826 "both data=journal and dioread_nolock");
5830 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5831 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5832 ext4_msg(sb, KERN_ERR, "can't mount with "
5833 "journal_async_commit in data=ordered mode");
5839 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5840 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5845 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED))
5846 ext4_abort(sb, EXT4_ERR_ESHUTDOWN, "Abort forced by user");
5848 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5849 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5853 if (sbi->s_journal) {
5854 ext4_init_journal_params(sb, sbi->s_journal);
5855 set_task_ioprio(sbi->s_journal->j_task, parsed_opts.journal_ioprio);
5858 /* Flush outstanding errors before changing fs state */
5859 flush_work(&sbi->s_error_work);
5861 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
5862 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) {
5867 if (*flags & SB_RDONLY) {
5868 err = sync_filesystem(sb);
5871 err = dquot_suspend(sb, -1);
5876 * First of all, the unconditional stuff we have to do
5877 * to disable replay of the journal when we next remount
5879 sb->s_flags |= SB_RDONLY;
5882 * OK, test if we are remounting a valid rw partition
5883 * readonly, and if so set the rdonly flag and then
5884 * mark the partition as valid again.
5886 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5887 (sbi->s_mount_state & EXT4_VALID_FS))
5888 es->s_state = cpu_to_le16(sbi->s_mount_state);
5890 if (sbi->s_journal) {
5892 * We let remount-ro finish even if marking fs
5893 * as clean failed...
5895 ext4_mark_recovery_complete(sb, es);
5898 /* Make sure we can mount this feature set readwrite */
5899 if (ext4_has_feature_readonly(sb) ||
5900 !ext4_feature_set_ok(sb, 0)) {
5905 * Make sure the group descriptor checksums
5906 * are sane. If they aren't, refuse to remount r/w.
5908 for (g = 0; g < sbi->s_groups_count; g++) {
5909 struct ext4_group_desc *gdp =
5910 ext4_get_group_desc(sb, g, NULL);
5912 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5913 ext4_msg(sb, KERN_ERR,
5914 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5915 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5916 le16_to_cpu(gdp->bg_checksum));
5923 * If we have an unprocessed orphan list hanging
5924 * around from a previously readonly bdev mount,
5925 * require a full umount/remount for now.
5927 if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) {
5928 ext4_msg(sb, KERN_WARNING, "Couldn't "
5929 "remount RDWR because of unprocessed "
5930 "orphan inode list. Please "
5931 "umount/remount instead");
5937 * Mounting a RDONLY partition read-write, so reread
5938 * and store the current valid flag. (It may have
5939 * been changed by e2fsck since we originally mounted
5942 if (sbi->s_journal) {
5943 err = ext4_clear_journal_err(sb, es);
5947 sbi->s_mount_state = (le16_to_cpu(es->s_state) &
5950 err = ext4_setup_super(sb, es, 0);
5954 sb->s_flags &= ~SB_RDONLY;
5955 if (ext4_has_feature_mmp(sb)) {
5956 err = ext4_multi_mount_protect(sb,
5957 le64_to_cpu(es->s_mmp_block));
5968 * Handle creation of system zone data early because it can fail.
5969 * Releasing of existing data is done when we are sure remount will
5972 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
5973 err = ext4_setup_system_zone(sb);
5978 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5979 err = ext4_commit_super(sb);
5986 if (sb_any_quota_suspended(sb))
5987 dquot_resume(sb, -1);
5988 else if (ext4_has_feature_quota(sb)) {
5989 err = ext4_enable_quotas(sb);
5994 /* Release old quota file names */
5995 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5996 kfree(old_opts.s_qf_names[i]);
5998 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
5999 ext4_release_system_zone(sb);
6002 * Reinitialize lazy itable initialization thread based on
6005 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
6006 ext4_unregister_li_request(sb);
6008 ext4_group_t first_not_zeroed;
6009 first_not_zeroed = ext4_has_uninit_itable(sb);
6010 ext4_register_li_request(sb, first_not_zeroed);
6013 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6014 ext4_stop_mmpd(sbi);
6017 * Some options can be enabled by ext4 and/or by VFS mount flag
6018 * either way we need to make sure it matches in both *flags and
6019 * s_flags. Copy those selected flags from s_flags to *flags
6021 *flags = (*flags & ~vfs_flags) | (sb->s_flags & vfs_flags);
6023 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s. Quota mode: %s.",
6024 orig_data, ext4_quota_mode(sb));
6030 * If there was a failing r/w to ro transition, we may need to
6033 if ((sb->s_flags & SB_RDONLY) && !(old_sb_flags & SB_RDONLY) &&
6034 sb_any_quota_suspended(sb))
6035 dquot_resume(sb, -1);
6036 sb->s_flags = old_sb_flags;
6037 sbi->s_mount_opt = old_opts.s_mount_opt;
6038 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
6039 sbi->s_resuid = old_opts.s_resuid;
6040 sbi->s_resgid = old_opts.s_resgid;
6041 sbi->s_commit_interval = old_opts.s_commit_interval;
6042 sbi->s_min_batch_time = old_opts.s_min_batch_time;
6043 sbi->s_max_batch_time = old_opts.s_max_batch_time;
6044 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6045 ext4_release_system_zone(sb);
6047 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
6048 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
6049 to_free[i] = get_qf_name(sb, sbi, i);
6050 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
6053 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6056 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6057 ext4_stop_mmpd(sbi);
6063 static int ext4_statfs_project(struct super_block *sb,
6064 kprojid_t projid, struct kstatfs *buf)
6067 struct dquot *dquot;
6071 qid = make_kqid_projid(projid);
6072 dquot = dqget(sb, qid);
6074 return PTR_ERR(dquot);
6075 spin_lock(&dquot->dq_dqb_lock);
6077 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6078 dquot->dq_dqb.dqb_bhardlimit);
6079 limit >>= sb->s_blocksize_bits;
6081 if (limit && buf->f_blocks > limit) {
6082 curblock = (dquot->dq_dqb.dqb_curspace +
6083 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
6084 buf->f_blocks = limit;
6085 buf->f_bfree = buf->f_bavail =
6086 (buf->f_blocks > curblock) ?
6087 (buf->f_blocks - curblock) : 0;
6090 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6091 dquot->dq_dqb.dqb_ihardlimit);
6092 if (limit && buf->f_files > limit) {
6093 buf->f_files = limit;
6095 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6096 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6099 spin_unlock(&dquot->dq_dqb_lock);
6105 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
6107 struct super_block *sb = dentry->d_sb;
6108 struct ext4_sb_info *sbi = EXT4_SB(sb);
6109 struct ext4_super_block *es = sbi->s_es;
6110 ext4_fsblk_t overhead = 0, resv_blocks;
6112 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
6114 if (!test_opt(sb, MINIX_DF))
6115 overhead = sbi->s_overhead;
6117 buf->f_type = EXT4_SUPER_MAGIC;
6118 buf->f_bsize = sb->s_blocksize;
6119 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
6120 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6121 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
6122 /* prevent underflow in case that few free space is available */
6123 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
6124 buf->f_bavail = buf->f_bfree -
6125 (ext4_r_blocks_count(es) + resv_blocks);
6126 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
6128 buf->f_files = le32_to_cpu(es->s_inodes_count);
6129 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
6130 buf->f_namelen = EXT4_NAME_LEN;
6131 buf->f_fsid = uuid_to_fsid(es->s_uuid);
6134 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6135 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6136 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6145 * Helper functions so that transaction is started before we acquire dqio_sem
6146 * to keep correct lock ordering of transaction > dqio_sem
6148 static inline struct inode *dquot_to_inode(struct dquot *dquot)
6150 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
6153 static int ext4_write_dquot(struct dquot *dquot)
6157 struct inode *inode;
6159 inode = dquot_to_inode(dquot);
6160 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
6161 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
6163 return PTR_ERR(handle);
6164 ret = dquot_commit(dquot);
6165 err = ext4_journal_stop(handle);
6171 static int ext4_acquire_dquot(struct dquot *dquot)
6176 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6177 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
6179 return PTR_ERR(handle);
6180 ret = dquot_acquire(dquot);
6181 err = ext4_journal_stop(handle);
6187 static int ext4_release_dquot(struct dquot *dquot)
6192 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6193 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
6194 if (IS_ERR(handle)) {
6195 /* Release dquot anyway to avoid endless cycle in dqput() */
6196 dquot_release(dquot);
6197 return PTR_ERR(handle);
6199 ret = dquot_release(dquot);
6200 err = ext4_journal_stop(handle);
6206 static int ext4_mark_dquot_dirty(struct dquot *dquot)
6208 struct super_block *sb = dquot->dq_sb;
6210 if (ext4_is_quota_journalled(sb)) {
6211 dquot_mark_dquot_dirty(dquot);
6212 return ext4_write_dquot(dquot);
6214 return dquot_mark_dquot_dirty(dquot);
6218 static int ext4_write_info(struct super_block *sb, int type)
6223 /* Data block + inode block */
6224 handle = ext4_journal_start_sb(sb, EXT4_HT_QUOTA, 2);
6226 return PTR_ERR(handle);
6227 ret = dquot_commit_info(sb, type);
6228 err = ext4_journal_stop(handle);
6234 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6236 struct ext4_inode_info *ei = EXT4_I(inode);
6238 /* The first argument of lockdep_set_subclass has to be
6239 * *exactly* the same as the argument to init_rwsem() --- in
6240 * this case, in init_once() --- or lockdep gets unhappy
6241 * because the name of the lock is set using the
6242 * stringification of the argument to init_rwsem().
6244 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6245 lockdep_set_subclass(&ei->i_data_sem, subclass);
6249 * Standard function to be called on quota_on
6251 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6252 const struct path *path)
6256 if (!test_opt(sb, QUOTA))
6259 /* Quotafile not on the same filesystem? */
6260 if (path->dentry->d_sb != sb)
6263 /* Quota already enabled for this file? */
6264 if (IS_NOQUOTA(d_inode(path->dentry)))
6267 /* Journaling quota? */
6268 if (EXT4_SB(sb)->s_qf_names[type]) {
6269 /* Quotafile not in fs root? */
6270 if (path->dentry->d_parent != sb->s_root)
6271 ext4_msg(sb, KERN_WARNING,
6272 "Quota file not on filesystem root. "
6273 "Journaled quota will not work");
6274 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6277 * Clear the flag just in case mount options changed since
6280 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
6284 * When we journal data on quota file, we have to flush journal to see
6285 * all updates to the file when we bypass pagecache...
6287 if (EXT4_SB(sb)->s_journal &&
6288 ext4_should_journal_data(d_inode(path->dentry))) {
6290 * We don't need to lock updates but journal_flush() could
6291 * otherwise be livelocked...
6293 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
6294 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
6295 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
6300 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6301 err = dquot_quota_on(sb, type, format_id, path);
6303 struct inode *inode = d_inode(path->dentry);
6307 * Set inode flags to prevent userspace from messing with quota
6308 * files. If this fails, we return success anyway since quotas
6309 * are already enabled and this is not a hard failure.
6312 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6315 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6316 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6317 S_NOATIME | S_IMMUTABLE);
6318 err = ext4_mark_inode_dirty(handle, inode);
6319 ext4_journal_stop(handle);
6321 inode_unlock(inode);
6323 dquot_quota_off(sb, type);
6326 lockdep_set_quota_inode(path->dentry->d_inode,
6331 static inline bool ext4_check_quota_inum(int type, unsigned long qf_inum)
6335 return qf_inum == EXT4_USR_QUOTA_INO;
6337 return qf_inum == EXT4_GRP_QUOTA_INO;
6339 return qf_inum >= EXT4_GOOD_OLD_FIRST_INO;
6345 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6349 struct inode *qf_inode;
6350 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6351 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6352 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6353 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6356 BUG_ON(!ext4_has_feature_quota(sb));
6358 if (!qf_inums[type])
6361 if (!ext4_check_quota_inum(type, qf_inums[type])) {
6362 ext4_error(sb, "Bad quota inum: %lu, type: %d",
6363 qf_inums[type], type);
6367 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
6368 if (IS_ERR(qf_inode)) {
6369 ext4_error(sb, "Bad quota inode: %lu, type: %d",
6370 qf_inums[type], type);
6371 return PTR_ERR(qf_inode);
6374 /* Don't account quota for quota files to avoid recursion */
6375 qf_inode->i_flags |= S_NOQUOTA;
6376 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
6377 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
6379 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
6385 /* Enable usage tracking for all quota types. */
6386 int ext4_enable_quotas(struct super_block *sb)
6389 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6390 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6391 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6392 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6394 bool quota_mopt[EXT4_MAXQUOTAS] = {
6395 test_opt(sb, USRQUOTA),
6396 test_opt(sb, GRPQUOTA),
6397 test_opt(sb, PRJQUOTA),
6400 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
6401 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
6402 if (qf_inums[type]) {
6403 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
6404 DQUOT_USAGE_ENABLED |
6405 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
6408 "Failed to enable quota tracking "
6409 "(type=%d, err=%d, ino=%lu). "
6410 "Please run e2fsck to fix.", type,
6411 err, qf_inums[type]);
6412 for (type--; type >= 0; type--) {
6413 struct inode *inode;
6415 inode = sb_dqopt(sb)->files[type];
6417 inode = igrab(inode);
6418 dquot_quota_off(sb, type);
6420 lockdep_set_quota_inode(inode,
6433 static int ext4_quota_off(struct super_block *sb, int type)
6435 struct inode *inode = sb_dqopt(sb)->files[type];
6439 /* Force all delayed allocation blocks to be allocated.
6440 * Caller already holds s_umount sem */
6441 if (test_opt(sb, DELALLOC))
6442 sync_filesystem(sb);
6444 if (!inode || !igrab(inode))
6447 err = dquot_quota_off(sb, type);
6448 if (err || ext4_has_feature_quota(sb))
6453 * Update modification times of quota files when userspace can
6454 * start looking at them. If we fail, we return success anyway since
6455 * this is not a hard failure and quotas are already disabled.
6457 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6458 if (IS_ERR(handle)) {
6459 err = PTR_ERR(handle);
6462 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
6463 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
6464 inode->i_mtime = inode->i_ctime = current_time(inode);
6465 err = ext4_mark_inode_dirty(handle, inode);
6466 ext4_journal_stop(handle);
6468 inode_unlock(inode);
6470 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
6474 return dquot_quota_off(sb, type);
6477 /* Read data from quotafile - avoid pagecache and such because we cannot afford
6478 * acquiring the locks... As quota files are never truncated and quota code
6479 * itself serializes the operations (and no one else should touch the files)
6480 * we don't have to be afraid of races */
6481 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
6482 size_t len, loff_t off)
6484 struct inode *inode = sb_dqopt(sb)->files[type];
6485 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6486 int offset = off & (sb->s_blocksize - 1);
6489 struct buffer_head *bh;
6490 loff_t i_size = i_size_read(inode);
6494 if (off+len > i_size)
6497 while (toread > 0) {
6498 tocopy = sb->s_blocksize - offset < toread ?
6499 sb->s_blocksize - offset : toread;
6500 bh = ext4_bread(NULL, inode, blk, 0);
6503 if (!bh) /* A hole? */
6504 memset(data, 0, tocopy);
6506 memcpy(data, bh->b_data+offset, tocopy);
6516 /* Write to quotafile (we know the transaction is already started and has
6517 * enough credits) */
6518 static ssize_t ext4_quota_write(struct super_block *sb, int type,
6519 const char *data, size_t len, loff_t off)
6521 struct inode *inode = sb_dqopt(sb)->files[type];
6522 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6523 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
6525 struct buffer_head *bh;
6526 handle_t *handle = journal_current_handle();
6529 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6530 " cancelled because transaction is not started",
6531 (unsigned long long)off, (unsigned long long)len);
6535 * Since we account only one data block in transaction credits,
6536 * then it is impossible to cross a block boundary.
6538 if (sb->s_blocksize - offset < len) {
6539 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6540 " cancelled because not block aligned",
6541 (unsigned long long)off, (unsigned long long)len);
6546 bh = ext4_bread(handle, inode, blk,
6547 EXT4_GET_BLOCKS_CREATE |
6548 EXT4_GET_BLOCKS_METADATA_NOFAIL);
6549 } while (PTR_ERR(bh) == -ENOSPC &&
6550 ext4_should_retry_alloc(inode->i_sb, &retries));
6555 BUFFER_TRACE(bh, "get write access");
6556 err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE);
6562 memcpy(bh->b_data+offset, data, len);
6563 flush_dcache_page(bh->b_page);
6565 err = ext4_handle_dirty_metadata(handle, NULL, bh);
6568 if (inode->i_size < off + len) {
6569 i_size_write(inode, off + len);
6570 EXT4_I(inode)->i_disksize = inode->i_size;
6571 err2 = ext4_mark_inode_dirty(handle, inode);
6572 if (unlikely(err2 && !err))
6575 return err ? err : len;
6579 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6580 const char *dev_name, void *data)
6582 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
6585 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
6586 static inline void register_as_ext2(void)
6588 int err = register_filesystem(&ext2_fs_type);
6591 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6594 static inline void unregister_as_ext2(void)
6596 unregister_filesystem(&ext2_fs_type);
6599 static inline int ext2_feature_set_ok(struct super_block *sb)
6601 if (ext4_has_unknown_ext2_incompat_features(sb))
6605 if (ext4_has_unknown_ext2_ro_compat_features(sb))
6610 static inline void register_as_ext2(void) { }
6611 static inline void unregister_as_ext2(void) { }
6612 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
6615 static inline void register_as_ext3(void)
6617 int err = register_filesystem(&ext3_fs_type);
6620 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6623 static inline void unregister_as_ext3(void)
6625 unregister_filesystem(&ext3_fs_type);
6628 static inline int ext3_feature_set_ok(struct super_block *sb)
6630 if (ext4_has_unknown_ext3_incompat_features(sb))
6632 if (!ext4_has_feature_journal(sb))
6636 if (ext4_has_unknown_ext3_ro_compat_features(sb))
6641 static struct file_system_type ext4_fs_type = {
6642 .owner = THIS_MODULE,
6644 .mount = ext4_mount,
6645 .kill_sb = kill_block_super,
6646 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
6648 MODULE_ALIAS_FS("ext4");
6650 /* Shared across all ext4 file systems */
6651 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
6653 static int __init ext4_init_fs(void)
6657 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
6658 ext4_li_info = NULL;
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();
6711 ext4_fc_destroy_dentry_cache();
6713 destroy_inodecache();
6715 ext4_exit_mballoc();
6719 ext4_exit_system_zone();
6723 ext4_exit_post_read_processing();
6725 ext4_exit_pending();
6732 static void __exit ext4_exit_fs(void)
6734 ext4_destroy_lazyinit_thread();
6735 unregister_as_ext2();
6736 unregister_as_ext3();
6737 unregister_filesystem(&ext4_fs_type);
6738 ext4_fc_destroy_dentry_cache();
6739 destroy_inodecache();
6740 ext4_exit_mballoc();
6742 ext4_exit_system_zone();
6744 ext4_exit_post_read_processing();
6746 ext4_exit_pending();
6749 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6750 MODULE_DESCRIPTION("Fourth Extended Filesystem");
6751 MODULE_LICENSE("GPL");
6752 MODULE_SOFTDEP("pre: crc32c");
6753 module_init(ext4_init_fs)
6754 module_exit(ext4_exit_fs)