1 // SPDX-License-Identifier: GPL-2.0
3 * linux/fs/ext4/super.c
5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI)
12 * linux/fs/minix/inode.c
14 * Copyright (C) 1991, 1992 Linus Torvalds
16 * Big-endian to little-endian byte-swapping/bitmaps by
17 * David S. Miller (davem@caip.rutgers.edu), 1995
20 #include <linux/module.h>
21 #include <linux/string.h>
23 #include <linux/time.h>
24 #include <linux/vmalloc.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/backing-dev.h>
29 #include <linux/parser.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/dax.h>
42 #include <linux/cleancache.h>
43 #include <linux/uaccess.h>
44 #include <linux/iversion.h>
45 #include <linux/unicode.h>
46 #include <linux/part_stat.h>
47 #include <linux/kthread.h>
48 #include <linux/freezer.h>
51 #include "ext4_extents.h" /* Needed for trace points definition */
52 #include "ext4_jbd2.h"
58 #define CREATE_TRACE_POINTS
59 #include <trace/events/ext4.h>
61 static struct ext4_lazy_init *ext4_li_info;
62 static struct mutex ext4_li_mtx;
63 static struct ratelimit_state ext4_mount_msg_ratelimit;
65 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
66 unsigned long journal_devnum);
67 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
68 static int ext4_commit_super(struct super_block *sb, int sync);
69 static int ext4_mark_recovery_complete(struct super_block *sb,
70 struct ext4_super_block *es);
71 static int ext4_clear_journal_err(struct super_block *sb,
72 struct ext4_super_block *es);
73 static int ext4_sync_fs(struct super_block *sb, int wait);
74 static int ext4_remount(struct super_block *sb, int *flags, char *data);
75 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
76 static int ext4_unfreeze(struct super_block *sb);
77 static int ext4_freeze(struct super_block *sb);
78 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
80 static inline int ext2_feature_set_ok(struct super_block *sb);
81 static inline int ext3_feature_set_ok(struct super_block *sb);
82 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
83 static void ext4_destroy_lazyinit_thread(void);
84 static void ext4_unregister_li_request(struct super_block *sb);
85 static void ext4_clear_request_list(void);
86 static struct inode *ext4_get_journal_inode(struct super_block *sb,
87 unsigned int journal_inum);
92 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
93 * i_mmap_rwsem (inode->i_mmap_rwsem)!
96 * mmap_lock -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
97 * page lock -> i_data_sem (rw)
99 * buffered write path:
100 * sb_start_write -> i_mutex -> mmap_lock
101 * sb_start_write -> i_mutex -> transaction start -> page lock ->
105 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock
106 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start ->
110 * sb_start_write -> i_mutex -> mmap_lock
111 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
114 * transaction start -> page lock(s) -> i_data_sem (rw)
117 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
118 static struct file_system_type ext2_fs_type = {
119 .owner = THIS_MODULE,
122 .kill_sb = kill_block_super,
123 .fs_flags = FS_REQUIRES_DEV,
125 MODULE_ALIAS_FS("ext2");
126 MODULE_ALIAS("ext2");
127 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
129 #define IS_EXT2_SB(sb) (0)
133 static struct file_system_type ext3_fs_type = {
134 .owner = THIS_MODULE,
137 .kill_sb = kill_block_super,
138 .fs_flags = FS_REQUIRES_DEV,
140 MODULE_ALIAS_FS("ext3");
141 MODULE_ALIAS("ext3");
142 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
145 static inline void __ext4_read_bh(struct buffer_head *bh, int op_flags,
149 * buffer's verified bit is no longer valid after reading from
150 * disk again due to write out error, clear it to make sure we
151 * recheck the buffer contents.
153 clear_buffer_verified(bh);
155 bh->b_end_io = end_io ? end_io : end_buffer_read_sync;
157 submit_bh(REQ_OP_READ, op_flags, bh);
160 void ext4_read_bh_nowait(struct buffer_head *bh, int op_flags,
163 BUG_ON(!buffer_locked(bh));
165 if (ext4_buffer_uptodate(bh)) {
169 __ext4_read_bh(bh, op_flags, end_io);
172 int ext4_read_bh(struct buffer_head *bh, int op_flags, bh_end_io_t *end_io)
174 BUG_ON(!buffer_locked(bh));
176 if (ext4_buffer_uptodate(bh)) {
181 __ext4_read_bh(bh, op_flags, end_io);
184 if (buffer_uptodate(bh))
189 int ext4_read_bh_lock(struct buffer_head *bh, int op_flags, bool wait)
193 ext4_read_bh_nowait(bh, op_flags, NULL);
196 return ext4_read_bh(bh, op_flags, NULL);
200 * This works like __bread_gfp() except it uses ERR_PTR for error
201 * returns. Currently with sb_bread it's impossible to distinguish
202 * between ENOMEM and EIO situations (since both result in a NULL
205 static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb,
206 sector_t block, int op_flags,
209 struct buffer_head *bh;
212 bh = sb_getblk_gfp(sb, block, gfp);
214 return ERR_PTR(-ENOMEM);
215 if (ext4_buffer_uptodate(bh))
218 ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true);
226 struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block,
229 return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE);
232 struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
235 return __ext4_sb_bread_gfp(sb, block, 0, 0);
238 void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block)
240 struct buffer_head *bh = sb_getblk_gfp(sb, block, 0);
243 if (trylock_buffer(bh))
244 ext4_read_bh_nowait(bh, REQ_RAHEAD, NULL);
249 static int ext4_verify_csum_type(struct super_block *sb,
250 struct ext4_super_block *es)
252 if (!ext4_has_feature_metadata_csum(sb))
255 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
258 static __le32 ext4_superblock_csum(struct super_block *sb,
259 struct ext4_super_block *es)
261 struct ext4_sb_info *sbi = EXT4_SB(sb);
262 int offset = offsetof(struct ext4_super_block, s_checksum);
265 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
267 return cpu_to_le32(csum);
270 static int ext4_superblock_csum_verify(struct super_block *sb,
271 struct ext4_super_block *es)
273 if (!ext4_has_metadata_csum(sb))
276 return es->s_checksum == ext4_superblock_csum(sb, es);
279 void ext4_superblock_csum_set(struct super_block *sb)
281 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
283 if (!ext4_has_metadata_csum(sb))
286 es->s_checksum = ext4_superblock_csum(sb, es);
289 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
290 struct ext4_group_desc *bg)
292 return le32_to_cpu(bg->bg_block_bitmap_lo) |
293 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
294 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
297 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
298 struct ext4_group_desc *bg)
300 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
301 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
302 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
305 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
306 struct ext4_group_desc *bg)
308 return le32_to_cpu(bg->bg_inode_table_lo) |
309 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
310 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
313 __u32 ext4_free_group_clusters(struct super_block *sb,
314 struct ext4_group_desc *bg)
316 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
317 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
318 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
321 __u32 ext4_free_inodes_count(struct super_block *sb,
322 struct ext4_group_desc *bg)
324 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
325 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
326 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
329 __u32 ext4_used_dirs_count(struct super_block *sb,
330 struct ext4_group_desc *bg)
332 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
333 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
334 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
337 __u32 ext4_itable_unused_count(struct super_block *sb,
338 struct ext4_group_desc *bg)
340 return le16_to_cpu(bg->bg_itable_unused_lo) |
341 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
342 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
345 void ext4_block_bitmap_set(struct super_block *sb,
346 struct ext4_group_desc *bg, ext4_fsblk_t blk)
348 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
349 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
350 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
353 void ext4_inode_bitmap_set(struct super_block *sb,
354 struct ext4_group_desc *bg, ext4_fsblk_t blk)
356 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
357 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
358 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
361 void ext4_inode_table_set(struct super_block *sb,
362 struct ext4_group_desc *bg, ext4_fsblk_t blk)
364 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
365 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
366 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
369 void ext4_free_group_clusters_set(struct super_block *sb,
370 struct ext4_group_desc *bg, __u32 count)
372 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
373 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
374 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
377 void ext4_free_inodes_set(struct super_block *sb,
378 struct ext4_group_desc *bg, __u32 count)
380 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
381 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
382 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
385 void ext4_used_dirs_set(struct super_block *sb,
386 struct ext4_group_desc *bg, __u32 count)
388 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
389 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
390 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
393 void ext4_itable_unused_set(struct super_block *sb,
394 struct ext4_group_desc *bg, __u32 count)
396 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
397 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
398 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
401 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi)
403 time64_t now = ktime_get_real_seconds();
405 now = clamp_val(now, 0, (1ull << 40) - 1);
407 *lo = cpu_to_le32(lower_32_bits(now));
408 *hi = upper_32_bits(now);
411 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
413 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
415 #define ext4_update_tstamp(es, tstamp) \
416 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
417 #define ext4_get_tstamp(es, tstamp) \
418 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
421 * The del_gendisk() function uninitializes the disk-specific data
422 * structures, including the bdi structure, without telling anyone
423 * else. Once this happens, any attempt to call mark_buffer_dirty()
424 * (for example, by ext4_commit_super), will cause a kernel OOPS.
425 * This is a kludge to prevent these oops until we can put in a proper
426 * hook in del_gendisk() to inform the VFS and file system layers.
428 static int block_device_ejected(struct super_block *sb)
430 struct inode *bd_inode = sb->s_bdev->bd_inode;
431 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
433 return bdi->dev == NULL;
436 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
438 struct super_block *sb = journal->j_private;
439 struct ext4_sb_info *sbi = EXT4_SB(sb);
440 int error = is_journal_aborted(journal);
441 struct ext4_journal_cb_entry *jce;
443 BUG_ON(txn->t_state == T_FINISHED);
445 ext4_process_freed_data(sb, txn->t_tid);
447 spin_lock(&sbi->s_md_lock);
448 while (!list_empty(&txn->t_private_list)) {
449 jce = list_entry(txn->t_private_list.next,
450 struct ext4_journal_cb_entry, jce_list);
451 list_del_init(&jce->jce_list);
452 spin_unlock(&sbi->s_md_lock);
453 jce->jce_func(sb, jce, error);
454 spin_lock(&sbi->s_md_lock);
456 spin_unlock(&sbi->s_md_lock);
460 * This writepage callback for write_cache_pages()
461 * takes care of a few cases after page cleaning.
463 * write_cache_pages() already checks for dirty pages
464 * and calls clear_page_dirty_for_io(), which we want,
465 * to write protect the pages.
467 * However, we may have to redirty a page (see below.)
469 static int ext4_journalled_writepage_callback(struct page *page,
470 struct writeback_control *wbc,
473 transaction_t *transaction = (transaction_t *) data;
474 struct buffer_head *bh, *head;
475 struct journal_head *jh;
477 bh = head = page_buffers(page);
480 * We have to redirty a page in these cases:
481 * 1) If buffer is dirty, it means the page was dirty because it
482 * contains a buffer that needs checkpointing. So the dirty bit
483 * needs to be preserved so that checkpointing writes the buffer
485 * 2) If buffer is not part of the committing transaction
486 * (we may have just accidentally come across this buffer because
487 * inode range tracking is not exact) or if the currently running
488 * transaction already contains this buffer as well, dirty bit
489 * needs to be preserved so that the buffer gets writeprotected
490 * properly on running transaction's commit.
493 if (buffer_dirty(bh) ||
494 (jh && (jh->b_transaction != transaction ||
495 jh->b_next_transaction))) {
496 redirty_page_for_writepage(wbc, page);
499 } while ((bh = bh->b_this_page) != head);
502 return AOP_WRITEPAGE_ACTIVATE;
505 static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
507 struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
508 struct writeback_control wbc = {
509 .sync_mode = WB_SYNC_ALL,
510 .nr_to_write = LONG_MAX,
511 .range_start = jinode->i_dirty_start,
512 .range_end = jinode->i_dirty_end,
515 return write_cache_pages(mapping, &wbc,
516 ext4_journalled_writepage_callback,
517 jinode->i_transaction);
520 static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
524 if (ext4_should_journal_data(jinode->i_vfs_inode))
525 ret = ext4_journalled_submit_inode_data_buffers(jinode);
527 ret = jbd2_journal_submit_inode_data_buffers(jinode);
532 static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
536 if (!ext4_should_journal_data(jinode->i_vfs_inode))
537 ret = jbd2_journal_finish_inode_data_buffers(jinode);
542 static bool system_going_down(void)
544 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
545 || system_state == SYSTEM_RESTART;
548 struct ext4_err_translation {
553 #define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err }
555 static struct ext4_err_translation err_translation[] = {
556 EXT4_ERR_TRANSLATE(EIO),
557 EXT4_ERR_TRANSLATE(ENOMEM),
558 EXT4_ERR_TRANSLATE(EFSBADCRC),
559 EXT4_ERR_TRANSLATE(EFSCORRUPTED),
560 EXT4_ERR_TRANSLATE(ENOSPC),
561 EXT4_ERR_TRANSLATE(ENOKEY),
562 EXT4_ERR_TRANSLATE(EROFS),
563 EXT4_ERR_TRANSLATE(EFBIG),
564 EXT4_ERR_TRANSLATE(EEXIST),
565 EXT4_ERR_TRANSLATE(ERANGE),
566 EXT4_ERR_TRANSLATE(EOVERFLOW),
567 EXT4_ERR_TRANSLATE(EBUSY),
568 EXT4_ERR_TRANSLATE(ENOTDIR),
569 EXT4_ERR_TRANSLATE(ENOTEMPTY),
570 EXT4_ERR_TRANSLATE(ESHUTDOWN),
571 EXT4_ERR_TRANSLATE(EFAULT),
574 static int ext4_errno_to_code(int errno)
578 for (i = 0; i < ARRAY_SIZE(err_translation); i++)
579 if (err_translation[i].errno == errno)
580 return err_translation[i].code;
581 return EXT4_ERR_UNKNOWN;
584 static void __save_error_info(struct super_block *sb, int error,
585 __u32 ino, __u64 block,
586 const char *func, unsigned int line)
588 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
590 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
591 if (bdev_read_only(sb->s_bdev))
593 /* We default to EFSCORRUPTED error... */
595 error = EFSCORRUPTED;
596 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
597 ext4_update_tstamp(es, s_last_error_time);
598 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
599 es->s_last_error_line = cpu_to_le32(line);
600 es->s_last_error_ino = cpu_to_le32(ino);
601 es->s_last_error_block = cpu_to_le64(block);
602 es->s_last_error_errcode = ext4_errno_to_code(error);
603 if (!es->s_first_error_time) {
604 es->s_first_error_time = es->s_last_error_time;
605 es->s_first_error_time_hi = es->s_last_error_time_hi;
606 strncpy(es->s_first_error_func, func,
607 sizeof(es->s_first_error_func));
608 es->s_first_error_line = cpu_to_le32(line);
609 es->s_first_error_ino = es->s_last_error_ino;
610 es->s_first_error_block = es->s_last_error_block;
611 es->s_first_error_errcode = es->s_last_error_errcode;
614 * Start the daily error reporting function if it hasn't been
617 if (!es->s_error_count)
618 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
619 le32_add_cpu(&es->s_error_count, 1);
622 static void save_error_info(struct super_block *sb, int error,
623 __u32 ino, __u64 block,
624 const char *func, unsigned int line)
626 __save_error_info(sb, error, ino, block, func, line);
627 if (!bdev_read_only(sb->s_bdev))
628 ext4_commit_super(sb, 1);
631 /* Deal with the reporting of failure conditions on a filesystem such as
632 * inconsistencies detected or read IO failures.
634 * On ext2, we can store the error state of the filesystem in the
635 * superblock. That is not possible on ext4, because we may have other
636 * write ordering constraints on the superblock which prevent us from
637 * writing it out straight away; and given that the journal is about to
638 * be aborted, we can't rely on the current, or future, transactions to
639 * write out the superblock safely.
641 * We'll just use the jbd2_journal_abort() error code to record an error in
642 * the journal instead. On recovery, the journal will complain about
643 * that error until we've noted it down and cleared it.
646 static void ext4_handle_error(struct super_block *sb)
648 journal_t *journal = EXT4_SB(sb)->s_journal;
650 if (test_opt(sb, WARN_ON_ERROR))
653 if (sb_rdonly(sb) || test_opt(sb, ERRORS_CONT))
656 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
658 jbd2_journal_abort(journal, -EIO);
660 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
661 * could panic during 'reboot -f' as the underlying device got already
664 if (test_opt(sb, ERRORS_RO) || system_going_down()) {
665 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
667 * Make sure updated value of ->s_mount_flags will be visible
668 * before ->s_flags update
671 sb->s_flags |= SB_RDONLY;
672 } else if (test_opt(sb, ERRORS_PANIC)) {
673 panic("EXT4-fs (device %s): panic forced after error\n",
678 #define ext4_error_ratelimit(sb) \
679 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
682 void __ext4_error(struct super_block *sb, const char *function,
683 unsigned int line, int error, __u64 block,
684 const char *fmt, ...)
686 struct va_format vaf;
689 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
692 trace_ext4_error(sb, function, line);
693 if (ext4_error_ratelimit(sb)) {
698 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
699 sb->s_id, function, line, current->comm, &vaf);
702 save_error_info(sb, error, 0, block, function, line);
703 ext4_handle_error(sb);
706 void __ext4_error_inode(struct inode *inode, const char *function,
707 unsigned int line, ext4_fsblk_t block, int error,
708 const char *fmt, ...)
711 struct va_format vaf;
713 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
716 trace_ext4_error(inode->i_sb, function, line);
717 if (ext4_error_ratelimit(inode->i_sb)) {
722 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
723 "inode #%lu: block %llu: comm %s: %pV\n",
724 inode->i_sb->s_id, function, line, inode->i_ino,
725 block, current->comm, &vaf);
727 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
728 "inode #%lu: comm %s: %pV\n",
729 inode->i_sb->s_id, function, line, inode->i_ino,
730 current->comm, &vaf);
733 save_error_info(inode->i_sb, error, inode->i_ino, block,
735 ext4_handle_error(inode->i_sb);
738 void __ext4_error_file(struct file *file, const char *function,
739 unsigned int line, ext4_fsblk_t block,
740 const char *fmt, ...)
743 struct va_format vaf;
744 struct inode *inode = file_inode(file);
745 char pathname[80], *path;
747 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
750 trace_ext4_error(inode->i_sb, function, line);
751 if (ext4_error_ratelimit(inode->i_sb)) {
752 path = file_path(file, pathname, sizeof(pathname));
760 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
761 "block %llu: comm %s: path %s: %pV\n",
762 inode->i_sb->s_id, function, line, inode->i_ino,
763 block, current->comm, path, &vaf);
766 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
767 "comm %s: path %s: %pV\n",
768 inode->i_sb->s_id, function, line, inode->i_ino,
769 current->comm, path, &vaf);
772 save_error_info(inode->i_sb, EFSCORRUPTED, inode->i_ino, block,
774 ext4_handle_error(inode->i_sb);
777 const char *ext4_decode_error(struct super_block *sb, int errno,
784 errstr = "Corrupt filesystem";
787 errstr = "Filesystem failed CRC";
790 errstr = "IO failure";
793 errstr = "Out of memory";
796 if (!sb || (EXT4_SB(sb)->s_journal &&
797 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
798 errstr = "Journal has aborted";
800 errstr = "Readonly filesystem";
803 /* If the caller passed in an extra buffer for unknown
804 * errors, textualise them now. Else we just return
807 /* Check for truncated error codes... */
808 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
817 /* __ext4_std_error decodes expected errors from journaling functions
818 * automatically and invokes the appropriate error response. */
820 void __ext4_std_error(struct super_block *sb, const char *function,
821 unsigned int line, int errno)
826 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
829 /* Special case: if the error is EROFS, and we're not already
830 * inside a transaction, then there's really no point in logging
832 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
835 if (ext4_error_ratelimit(sb)) {
836 errstr = ext4_decode_error(sb, errno, nbuf);
837 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
838 sb->s_id, function, line, errstr);
841 save_error_info(sb, -errno, 0, 0, function, line);
842 ext4_handle_error(sb);
846 * ext4_abort is a much stronger failure handler than ext4_error. The
847 * abort function may be used to deal with unrecoverable failures such
848 * as journal IO errors or ENOMEM at a critical moment in log management.
850 * We unconditionally force the filesystem into an ABORT|READONLY state,
851 * unless the error response on the fs has been set to panic in which
852 * case we take the easy way out and panic immediately.
855 void __ext4_abort(struct super_block *sb, const char *function,
856 unsigned int line, int error, const char *fmt, ...)
858 struct va_format vaf;
861 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
864 save_error_info(sb, error, 0, 0, function, line);
868 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
869 sb->s_id, function, line, &vaf);
872 if (sb_rdonly(sb) == 0) {
873 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
874 if (EXT4_SB(sb)->s_journal)
875 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
877 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
879 * Make sure updated value of ->s_mount_flags will be visible
880 * before ->s_flags update
883 sb->s_flags |= SB_RDONLY;
885 if (test_opt(sb, ERRORS_PANIC) && !system_going_down())
886 panic("EXT4-fs panic from previous error\n");
889 void __ext4_msg(struct super_block *sb,
890 const char *prefix, const char *fmt, ...)
892 struct va_format vaf;
895 atomic_inc(&EXT4_SB(sb)->s_msg_count);
896 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
902 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
906 static int ext4_warning_ratelimit(struct super_block *sb)
908 atomic_inc(&EXT4_SB(sb)->s_warning_count);
909 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
913 void __ext4_warning(struct super_block *sb, const char *function,
914 unsigned int line, const char *fmt, ...)
916 struct va_format vaf;
919 if (!ext4_warning_ratelimit(sb))
925 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
926 sb->s_id, function, line, &vaf);
930 void __ext4_warning_inode(const struct inode *inode, const char *function,
931 unsigned int line, const char *fmt, ...)
933 struct va_format vaf;
936 if (!ext4_warning_ratelimit(inode->i_sb))
942 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
943 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
944 function, line, inode->i_ino, current->comm, &vaf);
948 void __ext4_grp_locked_error(const char *function, unsigned int line,
949 struct super_block *sb, ext4_group_t grp,
950 unsigned long ino, ext4_fsblk_t block,
951 const char *fmt, ...)
955 struct va_format vaf;
958 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
961 trace_ext4_error(sb, function, line);
962 __save_error_info(sb, EFSCORRUPTED, ino, block, function, line);
964 if (ext4_error_ratelimit(sb)) {
968 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
969 sb->s_id, function, line, grp);
971 printk(KERN_CONT "inode %lu: ", ino);
973 printk(KERN_CONT "block %llu:",
974 (unsigned long long) block);
975 printk(KERN_CONT "%pV\n", &vaf);
979 if (test_opt(sb, WARN_ON_ERROR))
982 if (test_opt(sb, ERRORS_CONT)) {
983 ext4_commit_super(sb, 0);
987 ext4_unlock_group(sb, grp);
988 ext4_commit_super(sb, 1);
989 ext4_handle_error(sb);
991 * We only get here in the ERRORS_RO case; relocking the group
992 * may be dangerous, but nothing bad will happen since the
993 * filesystem will have already been marked read/only and the
994 * journal has been aborted. We return 1 as a hint to callers
995 * who might what to use the return value from
996 * ext4_grp_locked_error() to distinguish between the
997 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
998 * aggressively from the ext4 function in question, with a
999 * more appropriate error code.
1001 ext4_lock_group(sb, grp);
1005 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
1009 struct ext4_sb_info *sbi = EXT4_SB(sb);
1010 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1011 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
1016 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1017 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1020 percpu_counter_sub(&sbi->s_freeclusters_counter,
1024 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1025 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1030 count = ext4_free_inodes_count(sb, gdp);
1031 percpu_counter_sub(&sbi->s_freeinodes_counter,
1037 void ext4_update_dynamic_rev(struct super_block *sb)
1039 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
1041 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
1045 "updating to rev %d because of new feature flag, "
1046 "running e2fsck is recommended",
1049 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1050 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1051 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
1052 /* leave es->s_feature_*compat flags alone */
1053 /* es->s_uuid will be set by e2fsck if empty */
1056 * The rest of the superblock fields should be zero, and if not it
1057 * means they are likely already in use, so leave them alone. We
1058 * can leave it up to e2fsck to clean up any inconsistencies there.
1063 * Open the external journal device
1065 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
1067 struct block_device *bdev;
1069 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
1075 ext4_msg(sb, KERN_ERR,
1076 "failed to open journal device unknown-block(%u,%u) %ld",
1077 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
1082 * Release the journal device
1084 static void ext4_blkdev_put(struct block_device *bdev)
1086 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
1089 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
1091 struct block_device *bdev;
1092 bdev = sbi->s_journal_bdev;
1095 * Invalidate the journal device's buffers. We don't want them
1096 * floating about in memory - the physical journal device may
1097 * hotswapped, and it breaks the `ro-after' testing code.
1099 invalidate_bdev(bdev);
1100 ext4_blkdev_put(bdev);
1101 sbi->s_journal_bdev = NULL;
1105 static inline struct inode *orphan_list_entry(struct list_head *l)
1107 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1110 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1112 struct list_head *l;
1114 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1115 le32_to_cpu(sbi->s_es->s_last_orphan));
1117 printk(KERN_ERR "sb_info orphan list:\n");
1118 list_for_each(l, &sbi->s_orphan) {
1119 struct inode *inode = orphan_list_entry(l);
1121 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1122 inode->i_sb->s_id, inode->i_ino, inode,
1123 inode->i_mode, inode->i_nlink,
1124 NEXT_ORPHAN(inode));
1129 static int ext4_quota_off(struct super_block *sb, int type);
1131 static inline void ext4_quota_off_umount(struct super_block *sb)
1135 /* Use our quota_off function to clear inode flags etc. */
1136 for (type = 0; type < EXT4_MAXQUOTAS; type++)
1137 ext4_quota_off(sb, type);
1141 * This is a helper function which is used in the mount/remount
1142 * codepaths (which holds s_umount) to fetch the quota file name.
1144 static inline char *get_qf_name(struct super_block *sb,
1145 struct ext4_sb_info *sbi,
1148 return rcu_dereference_protected(sbi->s_qf_names[type],
1149 lockdep_is_held(&sb->s_umount));
1152 static inline void ext4_quota_off_umount(struct super_block *sb)
1157 static void ext4_put_super(struct super_block *sb)
1159 struct ext4_sb_info *sbi = EXT4_SB(sb);
1160 struct ext4_super_block *es = sbi->s_es;
1161 struct buffer_head **group_desc;
1162 struct flex_groups **flex_groups;
1167 * Unregister sysfs before destroying jbd2 journal.
1168 * Since we could still access attr_journal_task attribute via sysfs
1169 * path which could have sbi->s_journal->j_task as NULL
1170 * Unregister sysfs before flush sbi->s_error_work.
1171 * Since user may read /proc/fs/ext4/xx/mb_groups during umount, If
1172 * read metadata verify failed then will queue error work.
1173 * flush_stashed_error_work will call start_this_handle may trigger
1176 ext4_unregister_sysfs(sb);
1178 ext4_unregister_li_request(sb);
1179 ext4_quota_off_umount(sb);
1181 destroy_workqueue(sbi->rsv_conversion_wq);
1183 if (sbi->s_journal) {
1184 aborted = is_journal_aborted(sbi->s_journal);
1185 err = jbd2_journal_destroy(sbi->s_journal);
1186 sbi->s_journal = NULL;
1187 if ((err < 0) && !aborted) {
1188 ext4_abort(sb, -err, "Couldn't clean up the journal");
1192 ext4_es_unregister_shrinker(sbi);
1193 del_timer_sync(&sbi->s_err_report);
1194 ext4_release_system_zone(sb);
1195 ext4_mb_release(sb);
1196 ext4_ext_release(sb);
1198 if (!sb_rdonly(sb) && !aborted) {
1199 ext4_clear_feature_journal_needs_recovery(sb);
1200 es->s_state = cpu_to_le16(sbi->s_mount_state);
1203 ext4_commit_super(sb, 1);
1206 group_desc = rcu_dereference(sbi->s_group_desc);
1207 for (i = 0; i < sbi->s_gdb_count; i++)
1208 brelse(group_desc[i]);
1210 flex_groups = rcu_dereference(sbi->s_flex_groups);
1212 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1213 kvfree(flex_groups[i]);
1214 kvfree(flex_groups);
1217 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1218 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1219 percpu_counter_destroy(&sbi->s_dirs_counter);
1220 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1221 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
1222 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1224 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1225 kfree(get_qf_name(sb, sbi, i));
1228 /* Debugging code just in case the in-memory inode orphan list
1229 * isn't empty. The on-disk one can be non-empty if we've
1230 * detected an error and taken the fs readonly, but the
1231 * in-memory list had better be clean by this point. */
1232 if (!list_empty(&sbi->s_orphan))
1233 dump_orphan_list(sb, sbi);
1234 J_ASSERT(list_empty(&sbi->s_orphan));
1236 sync_blockdev(sb->s_bdev);
1237 invalidate_bdev(sb->s_bdev);
1238 if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) {
1239 sync_blockdev(sbi->s_journal_bdev);
1240 ext4_blkdev_remove(sbi);
1243 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1244 sbi->s_ea_inode_cache = NULL;
1246 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1247 sbi->s_ea_block_cache = NULL;
1249 ext4_stop_mmpd(sbi);
1252 sb->s_fs_info = NULL;
1254 * Now that we are completely done shutting down the
1255 * superblock, we need to actually destroy the kobject.
1257 kobject_put(&sbi->s_kobj);
1258 wait_for_completion(&sbi->s_kobj_unregister);
1259 if (sbi->s_chksum_driver)
1260 crypto_free_shash(sbi->s_chksum_driver);
1261 kfree(sbi->s_blockgroup_lock);
1262 fs_put_dax(sbi->s_daxdev);
1263 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
1264 #ifdef CONFIG_UNICODE
1265 utf8_unload(sb->s_encoding);
1270 static struct kmem_cache *ext4_inode_cachep;
1273 * Called inside transaction, so use GFP_NOFS
1275 static struct inode *ext4_alloc_inode(struct super_block *sb)
1277 struct ext4_inode_info *ei;
1279 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1283 inode_set_iversion(&ei->vfs_inode, 1);
1285 spin_lock_init(&ei->i_raw_lock);
1286 INIT_LIST_HEAD(&ei->i_prealloc_list);
1287 atomic_set(&ei->i_prealloc_active, 0);
1288 spin_lock_init(&ei->i_prealloc_lock);
1289 ext4_es_init_tree(&ei->i_es_tree);
1290 rwlock_init(&ei->i_es_lock);
1291 INIT_LIST_HEAD(&ei->i_es_list);
1292 ei->i_es_all_nr = 0;
1293 ei->i_es_shk_nr = 0;
1294 ei->i_es_shrink_lblk = 0;
1295 ei->i_reserved_data_blocks = 0;
1296 spin_lock_init(&(ei->i_block_reservation_lock));
1297 ext4_init_pending_tree(&ei->i_pending_tree);
1299 ei->i_reserved_quota = 0;
1300 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1303 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1304 spin_lock_init(&ei->i_completed_io_lock);
1306 ei->i_datasync_tid = 0;
1307 atomic_set(&ei->i_unwritten, 0);
1308 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1309 ext4_fc_init_inode(&ei->vfs_inode);
1310 mutex_init(&ei->i_fc_lock);
1311 return &ei->vfs_inode;
1314 static int ext4_drop_inode(struct inode *inode)
1316 int drop = generic_drop_inode(inode);
1319 drop = fscrypt_drop_inode(inode);
1321 trace_ext4_drop_inode(inode, drop);
1325 static void ext4_free_in_core_inode(struct inode *inode)
1327 fscrypt_free_inode(inode);
1328 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1329 pr_warn("%s: inode %ld still in fc list",
1330 __func__, inode->i_ino);
1332 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1335 static void ext4_destroy_inode(struct inode *inode)
1337 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1338 ext4_msg(inode->i_sb, KERN_ERR,
1339 "Inode %lu (%p): orphan list check failed!",
1340 inode->i_ino, EXT4_I(inode));
1341 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1342 EXT4_I(inode), sizeof(struct ext4_inode_info),
1347 if (EXT4_I(inode)->i_reserved_data_blocks)
1348 ext4_msg(inode->i_sb, KERN_ERR,
1349 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!",
1350 inode->i_ino, EXT4_I(inode),
1351 EXT4_I(inode)->i_reserved_data_blocks);
1354 static void init_once(void *foo)
1356 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1358 INIT_LIST_HEAD(&ei->i_orphan);
1359 init_rwsem(&ei->xattr_sem);
1360 init_rwsem(&ei->i_data_sem);
1361 init_rwsem(&ei->i_mmap_sem);
1362 inode_init_once(&ei->vfs_inode);
1363 ext4_fc_init_inode(&ei->vfs_inode);
1366 static int __init init_inodecache(void)
1368 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1369 sizeof(struct ext4_inode_info), 0,
1370 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1372 offsetof(struct ext4_inode_info, i_data),
1373 sizeof_field(struct ext4_inode_info, i_data),
1375 if (ext4_inode_cachep == NULL)
1380 static void destroy_inodecache(void)
1383 * Make sure all delayed rcu free inodes are flushed before we
1387 kmem_cache_destroy(ext4_inode_cachep);
1390 void ext4_clear_inode(struct inode *inode)
1393 invalidate_inode_buffers(inode);
1395 ext4_discard_preallocations(inode, 0);
1396 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1398 if (EXT4_I(inode)->jinode) {
1399 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1400 EXT4_I(inode)->jinode);
1401 jbd2_free_inode(EXT4_I(inode)->jinode);
1402 EXT4_I(inode)->jinode = NULL;
1404 fscrypt_put_encryption_info(inode);
1405 fsverity_cleanup_inode(inode);
1408 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1409 u64 ino, u32 generation)
1411 struct inode *inode;
1414 * Currently we don't know the generation for parent directory, so
1415 * a generation of 0 means "accept any"
1417 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1419 return ERR_CAST(inode);
1420 if (generation && inode->i_generation != generation) {
1422 return ERR_PTR(-ESTALE);
1428 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1429 int fh_len, int fh_type)
1431 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1432 ext4_nfs_get_inode);
1435 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1436 int fh_len, int fh_type)
1438 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1439 ext4_nfs_get_inode);
1442 static int ext4_nfs_commit_metadata(struct inode *inode)
1444 struct writeback_control wbc = {
1445 .sync_mode = WB_SYNC_ALL
1448 trace_ext4_nfs_commit_metadata(inode);
1449 return ext4_write_inode(inode, &wbc);
1453 * Try to release metadata pages (indirect blocks, directories) which are
1454 * mapped via the block device. Since these pages could have journal heads
1455 * which would prevent try_to_free_buffers() from freeing them, we must use
1456 * jbd2 layer's try_to_free_buffers() function to release them.
1458 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1461 journal_t *journal = EXT4_SB(sb)->s_journal;
1463 WARN_ON(PageChecked(page));
1464 if (!page_has_buffers(page))
1467 return jbd2_journal_try_to_free_buffers(journal, page);
1469 return try_to_free_buffers(page);
1472 #ifdef CONFIG_FS_ENCRYPTION
1473 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1475 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1476 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1479 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1482 handle_t *handle = fs_data;
1483 int res, res2, credits, retries = 0;
1486 * Encrypting the root directory is not allowed because e2fsck expects
1487 * lost+found to exist and be unencrypted, and encrypting the root
1488 * directory would imply encrypting the lost+found directory as well as
1489 * the filename "lost+found" itself.
1491 if (inode->i_ino == EXT4_ROOT_INO)
1494 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1497 if (ext4_test_inode_flag(inode, EXT4_INODE_DAX))
1500 res = ext4_convert_inline_data(inode);
1505 * If a journal handle was specified, then the encryption context is
1506 * being set on a new inode via inheritance and is part of a larger
1507 * transaction to create the inode. Otherwise the encryption context is
1508 * being set on an existing inode in its own transaction. Only in the
1509 * latter case should the "retry on ENOSPC" logic be used.
1513 res = ext4_xattr_set_handle(handle, inode,
1514 EXT4_XATTR_INDEX_ENCRYPTION,
1515 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1518 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1519 ext4_clear_inode_state(inode,
1520 EXT4_STATE_MAY_INLINE_DATA);
1522 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1523 * S_DAX may be disabled
1525 ext4_set_inode_flags(inode, false);
1530 res = dquot_initialize(inode);
1534 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1539 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1541 return PTR_ERR(handle);
1543 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1544 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1547 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1549 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1550 * S_DAX may be disabled
1552 ext4_set_inode_flags(inode, false);
1553 res = ext4_mark_inode_dirty(handle, inode);
1555 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1557 res2 = ext4_journal_stop(handle);
1559 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1566 static const union fscrypt_policy *ext4_get_dummy_policy(struct super_block *sb)
1568 return EXT4_SB(sb)->s_dummy_enc_policy.policy;
1571 static bool ext4_has_stable_inodes(struct super_block *sb)
1573 return ext4_has_feature_stable_inodes(sb);
1576 static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1577 int *ino_bits_ret, int *lblk_bits_ret)
1579 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1580 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1583 static const struct fscrypt_operations ext4_cryptops = {
1584 .key_prefix = "ext4:",
1585 .get_context = ext4_get_context,
1586 .set_context = ext4_set_context,
1587 .get_dummy_policy = ext4_get_dummy_policy,
1588 .empty_dir = ext4_empty_dir,
1589 .max_namelen = EXT4_NAME_LEN,
1590 .has_stable_inodes = ext4_has_stable_inodes,
1591 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits,
1596 static const char * const quotatypes[] = INITQFNAMES;
1597 #define QTYPE2NAME(t) (quotatypes[t])
1599 static int ext4_write_dquot(struct dquot *dquot);
1600 static int ext4_acquire_dquot(struct dquot *dquot);
1601 static int ext4_release_dquot(struct dquot *dquot);
1602 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1603 static int ext4_write_info(struct super_block *sb, int type);
1604 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1605 const struct path *path);
1606 static int ext4_quota_on_mount(struct super_block *sb, int type);
1607 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1608 size_t len, loff_t off);
1609 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1610 const char *data, size_t len, loff_t off);
1611 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1612 unsigned int flags);
1613 static int ext4_enable_quotas(struct super_block *sb);
1615 static struct dquot **ext4_get_dquots(struct inode *inode)
1617 return EXT4_I(inode)->i_dquot;
1620 static const struct dquot_operations ext4_quota_operations = {
1621 .get_reserved_space = ext4_get_reserved_space,
1622 .write_dquot = ext4_write_dquot,
1623 .acquire_dquot = ext4_acquire_dquot,
1624 .release_dquot = ext4_release_dquot,
1625 .mark_dirty = ext4_mark_dquot_dirty,
1626 .write_info = ext4_write_info,
1627 .alloc_dquot = dquot_alloc,
1628 .destroy_dquot = dquot_destroy,
1629 .get_projid = ext4_get_projid,
1630 .get_inode_usage = ext4_get_inode_usage,
1631 .get_next_id = dquot_get_next_id,
1634 static const struct quotactl_ops ext4_qctl_operations = {
1635 .quota_on = ext4_quota_on,
1636 .quota_off = ext4_quota_off,
1637 .quota_sync = dquot_quota_sync,
1638 .get_state = dquot_get_state,
1639 .set_info = dquot_set_dqinfo,
1640 .get_dqblk = dquot_get_dqblk,
1641 .set_dqblk = dquot_set_dqblk,
1642 .get_nextdqblk = dquot_get_next_dqblk,
1646 static const struct super_operations ext4_sops = {
1647 .alloc_inode = ext4_alloc_inode,
1648 .free_inode = ext4_free_in_core_inode,
1649 .destroy_inode = ext4_destroy_inode,
1650 .write_inode = ext4_write_inode,
1651 .dirty_inode = ext4_dirty_inode,
1652 .drop_inode = ext4_drop_inode,
1653 .evict_inode = ext4_evict_inode,
1654 .put_super = ext4_put_super,
1655 .sync_fs = ext4_sync_fs,
1656 .freeze_fs = ext4_freeze,
1657 .unfreeze_fs = ext4_unfreeze,
1658 .statfs = ext4_statfs,
1659 .remount_fs = ext4_remount,
1660 .show_options = ext4_show_options,
1662 .quota_read = ext4_quota_read,
1663 .quota_write = ext4_quota_write,
1664 .get_dquots = ext4_get_dquots,
1666 .bdev_try_to_free_page = bdev_try_to_free_page,
1669 static const struct export_operations ext4_export_ops = {
1670 .fh_to_dentry = ext4_fh_to_dentry,
1671 .fh_to_parent = ext4_fh_to_parent,
1672 .get_parent = ext4_get_parent,
1673 .commit_metadata = ext4_nfs_commit_metadata,
1677 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1678 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1679 Opt_nouid32, Opt_debug, Opt_removed,
1680 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1681 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1682 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1683 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1684 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1685 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1687 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1688 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1689 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1690 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version,
1691 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1692 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1693 Opt_nowarn_on_error, Opt_mblk_io_submit,
1694 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1695 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1696 Opt_inode_readahead_blks, Opt_journal_ioprio,
1697 Opt_dioread_nolock, Opt_dioread_lock,
1698 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1699 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1700 Opt_prefetch_block_bitmaps,
1701 #ifdef CONFIG_EXT4_DEBUG
1702 Opt_fc_debug_max_replay, Opt_fc_debug_force
1706 static const match_table_t tokens = {
1707 {Opt_bsd_df, "bsddf"},
1708 {Opt_minix_df, "minixdf"},
1709 {Opt_grpid, "grpid"},
1710 {Opt_grpid, "bsdgroups"},
1711 {Opt_nogrpid, "nogrpid"},
1712 {Opt_nogrpid, "sysvgroups"},
1713 {Opt_resgid, "resgid=%u"},
1714 {Opt_resuid, "resuid=%u"},
1716 {Opt_err_cont, "errors=continue"},
1717 {Opt_err_panic, "errors=panic"},
1718 {Opt_err_ro, "errors=remount-ro"},
1719 {Opt_nouid32, "nouid32"},
1720 {Opt_debug, "debug"},
1721 {Opt_removed, "oldalloc"},
1722 {Opt_removed, "orlov"},
1723 {Opt_user_xattr, "user_xattr"},
1724 {Opt_nouser_xattr, "nouser_xattr"},
1726 {Opt_noacl, "noacl"},
1727 {Opt_noload, "norecovery"},
1728 {Opt_noload, "noload"},
1729 {Opt_removed, "nobh"},
1730 {Opt_removed, "bh"},
1731 {Opt_commit, "commit=%u"},
1732 {Opt_min_batch_time, "min_batch_time=%u"},
1733 {Opt_max_batch_time, "max_batch_time=%u"},
1734 {Opt_journal_dev, "journal_dev=%u"},
1735 {Opt_journal_path, "journal_path=%s"},
1736 {Opt_journal_checksum, "journal_checksum"},
1737 {Opt_nojournal_checksum, "nojournal_checksum"},
1738 {Opt_journal_async_commit, "journal_async_commit"},
1739 {Opt_abort, "abort"},
1740 {Opt_data_journal, "data=journal"},
1741 {Opt_data_ordered, "data=ordered"},
1742 {Opt_data_writeback, "data=writeback"},
1743 {Opt_data_err_abort, "data_err=abort"},
1744 {Opt_data_err_ignore, "data_err=ignore"},
1745 {Opt_offusrjquota, "usrjquota="},
1746 {Opt_usrjquota, "usrjquota=%s"},
1747 {Opt_offgrpjquota, "grpjquota="},
1748 {Opt_grpjquota, "grpjquota=%s"},
1749 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1750 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1751 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1752 {Opt_grpquota, "grpquota"},
1753 {Opt_noquota, "noquota"},
1754 {Opt_quota, "quota"},
1755 {Opt_usrquota, "usrquota"},
1756 {Opt_prjquota, "prjquota"},
1757 {Opt_barrier, "barrier=%u"},
1758 {Opt_barrier, "barrier"},
1759 {Opt_nobarrier, "nobarrier"},
1760 {Opt_i_version, "i_version"},
1762 {Opt_dax_always, "dax=always"},
1763 {Opt_dax_inode, "dax=inode"},
1764 {Opt_dax_never, "dax=never"},
1765 {Opt_stripe, "stripe=%u"},
1766 {Opt_delalloc, "delalloc"},
1767 {Opt_warn_on_error, "warn_on_error"},
1768 {Opt_nowarn_on_error, "nowarn_on_error"},
1769 {Opt_lazytime, "lazytime"},
1770 {Opt_nolazytime, "nolazytime"},
1771 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1772 {Opt_nodelalloc, "nodelalloc"},
1773 {Opt_removed, "mblk_io_submit"},
1774 {Opt_removed, "nomblk_io_submit"},
1775 {Opt_block_validity, "block_validity"},
1776 {Opt_noblock_validity, "noblock_validity"},
1777 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1778 {Opt_journal_ioprio, "journal_ioprio=%u"},
1779 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1780 {Opt_auto_da_alloc, "auto_da_alloc"},
1781 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1782 {Opt_dioread_nolock, "dioread_nolock"},
1783 {Opt_dioread_lock, "nodioread_nolock"},
1784 {Opt_dioread_lock, "dioread_lock"},
1785 {Opt_discard, "discard"},
1786 {Opt_nodiscard, "nodiscard"},
1787 {Opt_init_itable, "init_itable=%u"},
1788 {Opt_init_itable, "init_itable"},
1789 {Opt_noinit_itable, "noinit_itable"},
1790 #ifdef CONFIG_EXT4_DEBUG
1791 {Opt_fc_debug_force, "fc_debug_force"},
1792 {Opt_fc_debug_max_replay, "fc_debug_max_replay=%u"},
1794 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1795 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
1796 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1797 {Opt_inlinecrypt, "inlinecrypt"},
1798 {Opt_nombcache, "nombcache"},
1799 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1800 {Opt_prefetch_block_bitmaps, "prefetch_block_bitmaps"},
1801 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1802 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1803 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1804 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1805 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1809 static ext4_fsblk_t get_sb_block(void **data)
1811 ext4_fsblk_t sb_block;
1812 char *options = (char *) *data;
1814 if (!options || strncmp(options, "sb=", 3) != 0)
1815 return 1; /* Default location */
1818 /* TODO: use simple_strtoll with >32bit ext4 */
1819 sb_block = simple_strtoul(options, &options, 0);
1820 if (*options && *options != ',') {
1821 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1825 if (*options == ',')
1827 *data = (void *) options;
1832 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1833 static const char deprecated_msg[] =
1834 "Mount option \"%s\" will be removed by %s\n"
1835 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1838 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1840 struct ext4_sb_info *sbi = EXT4_SB(sb);
1841 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
1844 if (sb_any_quota_loaded(sb) && !old_qname) {
1845 ext4_msg(sb, KERN_ERR,
1846 "Cannot change journaled "
1847 "quota options when quota turned on");
1850 if (ext4_has_feature_quota(sb)) {
1851 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1852 "ignored when QUOTA feature is enabled");
1855 qname = match_strdup(args);
1857 ext4_msg(sb, KERN_ERR,
1858 "Not enough memory for storing quotafile name");
1862 if (strcmp(old_qname, qname) == 0)
1865 ext4_msg(sb, KERN_ERR,
1866 "%s quota file already specified",
1870 if (strchr(qname, '/')) {
1871 ext4_msg(sb, KERN_ERR,
1872 "quotafile must be on filesystem root");
1875 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
1883 static int clear_qf_name(struct super_block *sb, int qtype)
1886 struct ext4_sb_info *sbi = EXT4_SB(sb);
1887 char *old_qname = get_qf_name(sb, sbi, qtype);
1889 if (sb_any_quota_loaded(sb) && old_qname) {
1890 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1891 " when quota turned on");
1894 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1901 #define MOPT_SET 0x0001
1902 #define MOPT_CLEAR 0x0002
1903 #define MOPT_NOSUPPORT 0x0004
1904 #define MOPT_EXPLICIT 0x0008
1905 #define MOPT_CLEAR_ERR 0x0010
1906 #define MOPT_GTE0 0x0020
1909 #define MOPT_QFMT 0x0040
1911 #define MOPT_Q MOPT_NOSUPPORT
1912 #define MOPT_QFMT MOPT_NOSUPPORT
1914 #define MOPT_DATAJ 0x0080
1915 #define MOPT_NO_EXT2 0x0100
1916 #define MOPT_NO_EXT3 0x0200
1917 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1918 #define MOPT_STRING 0x0400
1919 #define MOPT_SKIP 0x0800
1920 #define MOPT_2 0x1000
1922 static const struct mount_opts {
1926 } ext4_mount_opts[] = {
1927 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1928 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1929 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1930 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1931 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1932 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1933 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1934 MOPT_EXT4_ONLY | MOPT_SET},
1935 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1936 MOPT_EXT4_ONLY | MOPT_CLEAR},
1937 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1938 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1939 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1940 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1941 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1942 MOPT_EXT4_ONLY | MOPT_CLEAR},
1943 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1944 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1945 {Opt_commit, 0, MOPT_NO_EXT2},
1946 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1947 MOPT_EXT4_ONLY | MOPT_CLEAR},
1948 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1949 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1950 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1951 EXT4_MOUNT_JOURNAL_CHECKSUM),
1952 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1953 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1954 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1955 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1956 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1957 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1959 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1961 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1962 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1963 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1964 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1965 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1966 {Opt_commit, 0, MOPT_GTE0},
1967 {Opt_max_batch_time, 0, MOPT_GTE0},
1968 {Opt_min_batch_time, 0, MOPT_GTE0},
1969 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1970 {Opt_init_itable, 0, MOPT_GTE0},
1971 {Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP},
1972 {Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS,
1973 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1974 {Opt_dax_inode, EXT4_MOUNT2_DAX_INODE,
1975 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1976 {Opt_dax_never, EXT4_MOUNT2_DAX_NEVER,
1977 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1978 {Opt_stripe, 0, MOPT_GTE0},
1979 {Opt_resuid, 0, MOPT_GTE0},
1980 {Opt_resgid, 0, MOPT_GTE0},
1981 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1982 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1983 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1984 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1985 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1986 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1987 MOPT_NO_EXT2 | MOPT_DATAJ},
1988 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1989 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1990 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1991 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1992 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1994 {Opt_acl, 0, MOPT_NOSUPPORT},
1995 {Opt_noacl, 0, MOPT_NOSUPPORT},
1997 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1998 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1999 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
2000 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
2001 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
2003 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
2005 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
2007 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
2008 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
2009 MOPT_CLEAR | MOPT_Q},
2010 {Opt_usrjquota, 0, MOPT_Q | MOPT_STRING},
2011 {Opt_grpjquota, 0, MOPT_Q | MOPT_STRING},
2012 {Opt_offusrjquota, 0, MOPT_Q},
2013 {Opt_offgrpjquota, 0, MOPT_Q},
2014 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
2015 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
2016 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
2017 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
2018 {Opt_test_dummy_encryption, 0, MOPT_STRING},
2019 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
2020 {Opt_prefetch_block_bitmaps, EXT4_MOUNT_PREFETCH_BLOCK_BITMAPS,
2022 #ifdef CONFIG_EXT4_DEBUG
2023 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
2024 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
2025 {Opt_fc_debug_max_replay, 0, MOPT_GTE0},
2030 #ifdef CONFIG_UNICODE
2031 static const struct ext4_sb_encodings {
2035 } ext4_sb_encoding_map[] = {
2036 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
2039 static int ext4_sb_read_encoding(const struct ext4_super_block *es,
2040 const struct ext4_sb_encodings **encoding,
2043 __u16 magic = le16_to_cpu(es->s_encoding);
2046 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
2047 if (magic == ext4_sb_encoding_map[i].magic)
2050 if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
2053 *encoding = &ext4_sb_encoding_map[i];
2054 *flags = le16_to_cpu(es->s_encoding_flags);
2060 static int ext4_set_test_dummy_encryption(struct super_block *sb,
2062 const substring_t *arg,
2065 #ifdef CONFIG_FS_ENCRYPTION
2066 struct ext4_sb_info *sbi = EXT4_SB(sb);
2069 if (!ext4_has_feature_encrypt(sb)) {
2070 ext4_msg(sb, KERN_WARNING,
2071 "test_dummy_encryption requires encrypt feature");
2076 * This mount option is just for testing, and it's not worthwhile to
2077 * implement the extra complexity (e.g. RCU protection) that would be
2078 * needed to allow it to be set or changed during remount. We do allow
2079 * it to be specified during remount, but only if there is no change.
2081 if (is_remount && !sbi->s_dummy_enc_policy.policy) {
2082 ext4_msg(sb, KERN_WARNING,
2083 "Can't set test_dummy_encryption on remount");
2086 err = fscrypt_set_test_dummy_encryption(sb, arg->from,
2087 &sbi->s_dummy_enc_policy);
2090 ext4_msg(sb, KERN_WARNING,
2091 "Can't change test_dummy_encryption on remount");
2092 else if (err == -EINVAL)
2093 ext4_msg(sb, KERN_WARNING,
2094 "Value of option \"%s\" is unrecognized", opt);
2096 ext4_msg(sb, KERN_WARNING,
2097 "Error processing option \"%s\" [%d]",
2101 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2104 ext4_msg(sb, KERN_WARNING,
2105 "test_dummy_encryption option not supported");
2111 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
2112 substring_t *args, unsigned long *journal_devnum,
2113 unsigned int *journal_ioprio, int is_remount)
2115 struct ext4_sb_info *sbi = EXT4_SB(sb);
2116 const struct mount_opts *m;
2122 if (token == Opt_usrjquota)
2123 return set_qf_name(sb, USRQUOTA, &args[0]);
2124 else if (token == Opt_grpjquota)
2125 return set_qf_name(sb, GRPQUOTA, &args[0]);
2126 else if (token == Opt_offusrjquota)
2127 return clear_qf_name(sb, USRQUOTA);
2128 else if (token == Opt_offgrpjquota)
2129 return clear_qf_name(sb, GRPQUOTA);
2133 case Opt_nouser_xattr:
2134 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
2137 return 1; /* handled by get_sb_block() */
2139 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
2142 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
2145 sb->s_flags |= SB_I_VERSION;
2148 sb->s_flags |= SB_LAZYTIME;
2150 case Opt_nolazytime:
2151 sb->s_flags &= ~SB_LAZYTIME;
2153 case Opt_inlinecrypt:
2154 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2155 sb->s_flags |= SB_INLINECRYPT;
2157 ext4_msg(sb, KERN_ERR, "inline encryption not supported");
2162 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2163 if (token == m->token)
2166 if (m->token == Opt_err) {
2167 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
2168 "or missing value", opt);
2172 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2173 ext4_msg(sb, KERN_ERR,
2174 "Mount option \"%s\" incompatible with ext2", opt);
2177 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2178 ext4_msg(sb, KERN_ERR,
2179 "Mount option \"%s\" incompatible with ext3", opt);
2183 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
2185 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
2187 if (m->flags & MOPT_EXPLICIT) {
2188 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2189 set_opt2(sb, EXPLICIT_DELALLOC);
2190 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2191 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
2195 if (m->flags & MOPT_CLEAR_ERR)
2196 clear_opt(sb, ERRORS_MASK);
2197 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
2198 ext4_msg(sb, KERN_ERR, "Cannot change quota "
2199 "options when quota turned on");
2203 if (m->flags & MOPT_NOSUPPORT) {
2204 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
2205 } else if (token == Opt_commit) {
2207 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
2208 else if (arg > INT_MAX / HZ) {
2209 ext4_msg(sb, KERN_ERR,
2210 "Invalid commit interval %d, "
2211 "must be smaller than %d",
2215 sbi->s_commit_interval = HZ * arg;
2216 } else if (token == Opt_debug_want_extra_isize) {
2219 (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
2220 ext4_msg(sb, KERN_ERR,
2221 "Invalid want_extra_isize %d", arg);
2224 sbi->s_want_extra_isize = arg;
2225 } else if (token == Opt_max_batch_time) {
2226 sbi->s_max_batch_time = arg;
2227 } else if (token == Opt_min_batch_time) {
2228 sbi->s_min_batch_time = arg;
2229 } else if (token == Opt_inode_readahead_blks) {
2230 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
2231 ext4_msg(sb, KERN_ERR,
2232 "EXT4-fs: inode_readahead_blks must be "
2233 "0 or a power of 2 smaller than 2^31");
2236 sbi->s_inode_readahead_blks = arg;
2237 } else if (token == Opt_init_itable) {
2238 set_opt(sb, INIT_INODE_TABLE);
2240 arg = EXT4_DEF_LI_WAIT_MULT;
2241 sbi->s_li_wait_mult = arg;
2242 } else if (token == Opt_max_dir_size_kb) {
2243 sbi->s_max_dir_size_kb = arg;
2244 #ifdef CONFIG_EXT4_DEBUG
2245 } else if (token == Opt_fc_debug_max_replay) {
2246 sbi->s_fc_debug_max_replay = arg;
2248 } else if (token == Opt_stripe) {
2249 sbi->s_stripe = arg;
2250 } else if (token == Opt_resuid) {
2251 uid = make_kuid(current_user_ns(), arg);
2252 if (!uid_valid(uid)) {
2253 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
2256 sbi->s_resuid = uid;
2257 } else if (token == Opt_resgid) {
2258 gid = make_kgid(current_user_ns(), arg);
2259 if (!gid_valid(gid)) {
2260 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
2263 sbi->s_resgid = gid;
2264 } else if (token == Opt_journal_dev) {
2266 ext4_msg(sb, KERN_ERR,
2267 "Cannot specify journal on remount");
2270 *journal_devnum = arg;
2271 } else if (token == Opt_journal_path) {
2273 struct inode *journal_inode;
2278 ext4_msg(sb, KERN_ERR,
2279 "Cannot specify journal on remount");
2282 journal_path = match_strdup(&args[0]);
2283 if (!journal_path) {
2284 ext4_msg(sb, KERN_ERR, "error: could not dup "
2285 "journal device string");
2289 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
2291 ext4_msg(sb, KERN_ERR, "error: could not find "
2292 "journal device path: error %d", error);
2293 kfree(journal_path);
2297 journal_inode = d_inode(path.dentry);
2298 if (!S_ISBLK(journal_inode->i_mode)) {
2299 ext4_msg(sb, KERN_ERR, "error: journal path %s "
2300 "is not a block device", journal_path);
2302 kfree(journal_path);
2306 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
2308 kfree(journal_path);
2309 } else if (token == Opt_journal_ioprio) {
2311 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
2316 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
2317 } else if (token == Opt_test_dummy_encryption) {
2318 return ext4_set_test_dummy_encryption(sb, opt, &args[0],
2320 } else if (m->flags & MOPT_DATAJ) {
2322 if (!sbi->s_journal)
2323 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
2324 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
2325 ext4_msg(sb, KERN_ERR,
2326 "Cannot change data mode on remount");
2330 clear_opt(sb, DATA_FLAGS);
2331 sbi->s_mount_opt |= m->mount_opt;
2334 } else if (m->flags & MOPT_QFMT) {
2335 if (sb_any_quota_loaded(sb) &&
2336 sbi->s_jquota_fmt != m->mount_opt) {
2337 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
2338 "quota options when quota turned on");
2341 if (ext4_has_feature_quota(sb)) {
2342 ext4_msg(sb, KERN_INFO,
2343 "Quota format mount options ignored "
2344 "when QUOTA feature is enabled");
2347 sbi->s_jquota_fmt = m->mount_opt;
2349 } else if (token == Opt_dax || token == Opt_dax_always ||
2350 token == Opt_dax_inode || token == Opt_dax_never) {
2351 #ifdef CONFIG_FS_DAX
2354 case Opt_dax_always:
2356 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2357 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2358 fail_dax_change_remount:
2359 ext4_msg(sb, KERN_ERR, "can't change "
2360 "dax mount option while remounting");
2364 (test_opt(sb, DATA_FLAGS) ==
2365 EXT4_MOUNT_JOURNAL_DATA)) {
2366 ext4_msg(sb, KERN_ERR, "can't mount with "
2367 "both data=journal and dax");
2370 ext4_msg(sb, KERN_WARNING,
2371 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2372 sbi->s_mount_opt |= EXT4_MOUNT_DAX_ALWAYS;
2373 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2377 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2378 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS)))
2379 goto fail_dax_change_remount;
2380 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2381 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2385 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2386 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2387 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE)))
2388 goto fail_dax_change_remount;
2389 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2390 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2391 /* Strictly for printing options */
2392 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_INODE;
2396 ext4_msg(sb, KERN_INFO, "dax option not supported");
2397 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2398 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2401 } else if (token == Opt_data_err_abort) {
2402 sbi->s_mount_opt |= m->mount_opt;
2403 } else if (token == Opt_data_err_ignore) {
2404 sbi->s_mount_opt &= ~m->mount_opt;
2408 if (m->flags & MOPT_CLEAR)
2410 else if (unlikely(!(m->flags & MOPT_SET))) {
2411 ext4_msg(sb, KERN_WARNING,
2412 "buggy handling of option %s", opt);
2416 if (m->flags & MOPT_2) {
2418 sbi->s_mount_opt2 |= m->mount_opt;
2420 sbi->s_mount_opt2 &= ~m->mount_opt;
2423 sbi->s_mount_opt |= m->mount_opt;
2425 sbi->s_mount_opt &= ~m->mount_opt;
2431 static int parse_options(char *options, struct super_block *sb,
2432 unsigned long *journal_devnum,
2433 unsigned int *journal_ioprio,
2436 struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb);
2437 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
2438 substring_t args[MAX_OPT_ARGS];
2444 while ((p = strsep(&options, ",")) != NULL) {
2448 * Initialize args struct so we know whether arg was
2449 * found; some options take optional arguments.
2451 args[0].to = args[0].from = NULL;
2452 token = match_token(p, tokens, args);
2453 if (handle_mount_opt(sb, p, token, args, journal_devnum,
2454 journal_ioprio, is_remount) < 0)
2459 * We do the test below only for project quotas. 'usrquota' and
2460 * 'grpquota' mount options are allowed even without quota feature
2461 * to support legacy quotas in quota files.
2463 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2464 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2465 "Cannot enable project quota enforcement.");
2468 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2469 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2470 if (usr_qf_name || grp_qf_name) {
2471 if (test_opt(sb, USRQUOTA) && usr_qf_name)
2472 clear_opt(sb, USRQUOTA);
2474 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
2475 clear_opt(sb, GRPQUOTA);
2477 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
2478 ext4_msg(sb, KERN_ERR, "old and new quota "
2483 if (!sbi->s_jquota_fmt) {
2484 ext4_msg(sb, KERN_ERR, "journaled quota format "
2490 if (test_opt(sb, DIOREAD_NOLOCK)) {
2492 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2493 if (blocksize < PAGE_SIZE)
2494 ext4_msg(sb, KERN_WARNING, "Warning: mounting with an "
2495 "experimental mount option 'dioread_nolock' "
2496 "for blocksize < PAGE_SIZE");
2501 static inline void ext4_show_quota_options(struct seq_file *seq,
2502 struct super_block *sb)
2504 #if defined(CONFIG_QUOTA)
2505 struct ext4_sb_info *sbi = EXT4_SB(sb);
2506 char *usr_qf_name, *grp_qf_name;
2508 if (sbi->s_jquota_fmt) {
2511 switch (sbi->s_jquota_fmt) {
2522 seq_printf(seq, ",jqfmt=%s", fmtname);
2526 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2527 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2529 seq_show_option(seq, "usrjquota", usr_qf_name);
2531 seq_show_option(seq, "grpjquota", grp_qf_name);
2536 static const char *token2str(int token)
2538 const struct match_token *t;
2540 for (t = tokens; t->token != Opt_err; t++)
2541 if (t->token == token && !strchr(t->pattern, '='))
2548 * - it's set to a non-default value OR
2549 * - if the per-sb default is different from the global default
2551 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2554 struct ext4_sb_info *sbi = EXT4_SB(sb);
2555 struct ext4_super_block *es = sbi->s_es;
2556 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2557 const struct mount_opts *m;
2558 char sep = nodefs ? '\n' : ',';
2560 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2561 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2563 if (sbi->s_sb_block != 1)
2564 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2566 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2567 int want_set = m->flags & MOPT_SET;
2568 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2569 (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP)
2571 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2572 continue; /* skip if same as the default */
2574 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2575 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2576 continue; /* select Opt_noFoo vs Opt_Foo */
2577 SEQ_OPTS_PRINT("%s", token2str(m->token));
2580 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2581 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2582 SEQ_OPTS_PRINT("resuid=%u",
2583 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2584 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2585 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2586 SEQ_OPTS_PRINT("resgid=%u",
2587 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2588 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2589 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2590 SEQ_OPTS_PUTS("errors=remount-ro");
2591 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2592 SEQ_OPTS_PUTS("errors=continue");
2593 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2594 SEQ_OPTS_PUTS("errors=panic");
2595 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2596 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2597 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2598 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2599 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2600 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2601 if (sb->s_flags & SB_I_VERSION)
2602 SEQ_OPTS_PUTS("i_version");
2603 if (nodefs || sbi->s_stripe)
2604 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2605 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2606 (sbi->s_mount_opt ^ def_mount_opt)) {
2607 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2608 SEQ_OPTS_PUTS("data=journal");
2609 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2610 SEQ_OPTS_PUTS("data=ordered");
2611 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2612 SEQ_OPTS_PUTS("data=writeback");
2615 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2616 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2617 sbi->s_inode_readahead_blks);
2619 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2620 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2621 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2622 if (nodefs || sbi->s_max_dir_size_kb)
2623 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2624 if (test_opt(sb, DATA_ERR_ABORT))
2625 SEQ_OPTS_PUTS("data_err=abort");
2627 fscrypt_show_test_dummy_encryption(seq, sep, sb);
2629 if (sb->s_flags & SB_INLINECRYPT)
2630 SEQ_OPTS_PUTS("inlinecrypt");
2632 if (test_opt(sb, DAX_ALWAYS)) {
2634 SEQ_OPTS_PUTS("dax");
2636 SEQ_OPTS_PUTS("dax=always");
2637 } else if (test_opt2(sb, DAX_NEVER)) {
2638 SEQ_OPTS_PUTS("dax=never");
2639 } else if (test_opt2(sb, DAX_INODE)) {
2640 SEQ_OPTS_PUTS("dax=inode");
2642 ext4_show_quota_options(seq, sb);
2646 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2648 return _ext4_show_options(seq, root->d_sb, 0);
2651 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2653 struct super_block *sb = seq->private;
2656 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2657 rc = _ext4_show_options(seq, sb, 1);
2658 seq_puts(seq, "\n");
2662 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2665 struct ext4_sb_info *sbi = EXT4_SB(sb);
2668 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2669 ext4_msg(sb, KERN_ERR, "revision level too high, "
2670 "forcing read-only mode");
2676 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2677 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2678 "running e2fsck is recommended");
2679 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2680 ext4_msg(sb, KERN_WARNING,
2681 "warning: mounting fs with errors, "
2682 "running e2fsck is recommended");
2683 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2684 le16_to_cpu(es->s_mnt_count) >=
2685 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2686 ext4_msg(sb, KERN_WARNING,
2687 "warning: maximal mount count reached, "
2688 "running e2fsck is recommended");
2689 else if (le32_to_cpu(es->s_checkinterval) &&
2690 (ext4_get_tstamp(es, s_lastcheck) +
2691 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
2692 ext4_msg(sb, KERN_WARNING,
2693 "warning: checktime reached, "
2694 "running e2fsck is recommended");
2695 if (!sbi->s_journal)
2696 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2697 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2698 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2699 le16_add_cpu(&es->s_mnt_count, 1);
2700 ext4_update_tstamp(es, s_mtime);
2702 ext4_set_feature_journal_needs_recovery(sb);
2704 err = ext4_commit_super(sb, 1);
2706 if (test_opt(sb, DEBUG))
2707 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2708 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2710 sbi->s_groups_count,
2711 EXT4_BLOCKS_PER_GROUP(sb),
2712 EXT4_INODES_PER_GROUP(sb),
2713 sbi->s_mount_opt, sbi->s_mount_opt2);
2715 cleancache_init_fs(sb);
2719 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2721 struct ext4_sb_info *sbi = EXT4_SB(sb);
2722 struct flex_groups **old_groups, **new_groups;
2725 if (!sbi->s_log_groups_per_flex)
2728 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2729 if (size <= sbi->s_flex_groups_allocated)
2732 new_groups = kvzalloc(roundup_pow_of_two(size *
2733 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
2735 ext4_msg(sb, KERN_ERR,
2736 "not enough memory for %d flex group pointers", size);
2739 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2740 new_groups[i] = kvzalloc(roundup_pow_of_two(
2741 sizeof(struct flex_groups)),
2743 if (!new_groups[i]) {
2744 for (j = sbi->s_flex_groups_allocated; j < i; j++)
2745 kvfree(new_groups[j]);
2747 ext4_msg(sb, KERN_ERR,
2748 "not enough memory for %d flex groups", size);
2753 old_groups = rcu_dereference(sbi->s_flex_groups);
2755 memcpy(new_groups, old_groups,
2756 (sbi->s_flex_groups_allocated *
2757 sizeof(struct flex_groups *)));
2759 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2760 sbi->s_flex_groups_allocated = size;
2762 ext4_kvfree_array_rcu(old_groups);
2766 static int ext4_fill_flex_info(struct super_block *sb)
2768 struct ext4_sb_info *sbi = EXT4_SB(sb);
2769 struct ext4_group_desc *gdp = NULL;
2770 struct flex_groups *fg;
2771 ext4_group_t flex_group;
2774 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2775 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2776 sbi->s_log_groups_per_flex = 0;
2780 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2784 for (i = 0; i < sbi->s_groups_count; i++) {
2785 gdp = ext4_get_group_desc(sb, i, NULL);
2787 flex_group = ext4_flex_group(sbi, i);
2788 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2789 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
2790 atomic64_add(ext4_free_group_clusters(sb, gdp),
2791 &fg->free_clusters);
2792 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
2800 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2801 struct ext4_group_desc *gdp)
2803 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2805 __le32 le_group = cpu_to_le32(block_group);
2806 struct ext4_sb_info *sbi = EXT4_SB(sb);
2808 if (ext4_has_metadata_csum(sbi->s_sb)) {
2809 /* Use new metadata_csum algorithm */
2811 __u16 dummy_csum = 0;
2813 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2815 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2816 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2817 sizeof(dummy_csum));
2818 offset += sizeof(dummy_csum);
2819 if (offset < sbi->s_desc_size)
2820 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2821 sbi->s_desc_size - offset);
2823 crc = csum32 & 0xFFFF;
2827 /* old crc16 code */
2828 if (!ext4_has_feature_gdt_csum(sb))
2831 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2832 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2833 crc = crc16(crc, (__u8 *)gdp, offset);
2834 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2835 /* for checksum of struct ext4_group_desc do the rest...*/
2836 if (ext4_has_feature_64bit(sb) && offset < sbi->s_desc_size)
2837 crc = crc16(crc, (__u8 *)gdp + offset,
2838 sbi->s_desc_size - offset);
2841 return cpu_to_le16(crc);
2844 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2845 struct ext4_group_desc *gdp)
2847 if (ext4_has_group_desc_csum(sb) &&
2848 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2854 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2855 struct ext4_group_desc *gdp)
2857 if (!ext4_has_group_desc_csum(sb))
2859 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2862 /* Called at mount-time, super-block is locked */
2863 static int ext4_check_descriptors(struct super_block *sb,
2864 ext4_fsblk_t sb_block,
2865 ext4_group_t *first_not_zeroed)
2867 struct ext4_sb_info *sbi = EXT4_SB(sb);
2868 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2869 ext4_fsblk_t last_block;
2870 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2871 ext4_fsblk_t block_bitmap;
2872 ext4_fsblk_t inode_bitmap;
2873 ext4_fsblk_t inode_table;
2874 int flexbg_flag = 0;
2875 ext4_group_t i, grp = sbi->s_groups_count;
2877 if (ext4_has_feature_flex_bg(sb))
2880 ext4_debug("Checking group descriptors");
2882 for (i = 0; i < sbi->s_groups_count; i++) {
2883 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2885 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2886 last_block = ext4_blocks_count(sbi->s_es) - 1;
2888 last_block = first_block +
2889 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2891 if ((grp == sbi->s_groups_count) &&
2892 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2895 block_bitmap = ext4_block_bitmap(sb, gdp);
2896 if (block_bitmap == sb_block) {
2897 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2898 "Block bitmap for group %u overlaps "
2903 if (block_bitmap >= sb_block + 1 &&
2904 block_bitmap <= last_bg_block) {
2905 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2906 "Block bitmap for group %u overlaps "
2907 "block group descriptors", i);
2911 if (block_bitmap < first_block || block_bitmap > last_block) {
2912 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2913 "Block bitmap for group %u not in group "
2914 "(block %llu)!", i, block_bitmap);
2917 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2918 if (inode_bitmap == sb_block) {
2919 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2920 "Inode bitmap for group %u overlaps "
2925 if (inode_bitmap >= sb_block + 1 &&
2926 inode_bitmap <= last_bg_block) {
2927 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2928 "Inode bitmap for group %u overlaps "
2929 "block group descriptors", i);
2933 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2934 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2935 "Inode bitmap for group %u not in group "
2936 "(block %llu)!", i, inode_bitmap);
2939 inode_table = ext4_inode_table(sb, gdp);
2940 if (inode_table == sb_block) {
2941 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2942 "Inode table for group %u overlaps "
2947 if (inode_table >= sb_block + 1 &&
2948 inode_table <= last_bg_block) {
2949 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2950 "Inode table for group %u overlaps "
2951 "block group descriptors", i);
2955 if (inode_table < first_block ||
2956 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2957 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2958 "Inode table for group %u not in group "
2959 "(block %llu)!", i, inode_table);
2962 ext4_lock_group(sb, i);
2963 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2964 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2965 "Checksum for group %u failed (%u!=%u)",
2966 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2967 gdp)), le16_to_cpu(gdp->bg_checksum));
2968 if (!sb_rdonly(sb)) {
2969 ext4_unlock_group(sb, i);
2973 ext4_unlock_group(sb, i);
2975 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2977 if (NULL != first_not_zeroed)
2978 *first_not_zeroed = grp;
2982 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2983 * the superblock) which were deleted from all directories, but held open by
2984 * a process at the time of a crash. We walk the list and try to delete these
2985 * inodes at recovery time (only with a read-write filesystem).
2987 * In order to keep the orphan inode chain consistent during traversal (in
2988 * case of crash during recovery), we link each inode into the superblock
2989 * orphan list_head and handle it the same way as an inode deletion during
2990 * normal operation (which journals the operations for us).
2992 * We only do an iget() and an iput() on each inode, which is very safe if we
2993 * accidentally point at an in-use or already deleted inode. The worst that
2994 * can happen in this case is that we get a "bit already cleared" message from
2995 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2996 * e2fsck was run on this filesystem, and it must have already done the orphan
2997 * inode cleanup for us, so we can safely abort without any further action.
2999 static void ext4_orphan_cleanup(struct super_block *sb,
3000 struct ext4_super_block *es)
3002 unsigned int s_flags = sb->s_flags;
3003 int ret, nr_orphans = 0, nr_truncates = 0;
3005 int quota_update = 0;
3008 if (!es->s_last_orphan) {
3009 jbd_debug(4, "no orphan inodes to clean up\n");
3013 if (bdev_read_only(sb->s_bdev)) {
3014 ext4_msg(sb, KERN_ERR, "write access "
3015 "unavailable, skipping orphan cleanup");
3019 /* Check if feature set would not allow a r/w mount */
3020 if (!ext4_feature_set_ok(sb, 0)) {
3021 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
3022 "unknown ROCOMPAT features");
3026 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
3027 /* don't clear list on RO mount w/ errors */
3028 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
3029 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
3030 "clearing orphan list.\n");
3031 es->s_last_orphan = 0;
3033 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
3037 if (s_flags & SB_RDONLY) {
3038 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
3039 sb->s_flags &= ~SB_RDONLY;
3043 * Turn on quotas which were not enabled for read-only mounts if
3044 * filesystem has quota feature, so that they are updated correctly.
3046 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
3047 int ret = ext4_enable_quotas(sb);
3052 ext4_msg(sb, KERN_ERR,
3053 "Cannot turn on quotas: error %d", ret);
3056 /* Turn on journaled quotas used for old sytle */
3057 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
3058 if (EXT4_SB(sb)->s_qf_names[i]) {
3059 int ret = ext4_quota_on_mount(sb, i);
3064 ext4_msg(sb, KERN_ERR,
3065 "Cannot turn on journaled "
3066 "quota: type %d: error %d", i, ret);
3071 while (es->s_last_orphan) {
3072 struct inode *inode;
3075 * We may have encountered an error during cleanup; if
3076 * so, skip the rest.
3078 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
3079 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
3080 es->s_last_orphan = 0;
3084 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
3085 if (IS_ERR(inode)) {
3086 es->s_last_orphan = 0;
3090 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
3091 dquot_initialize(inode);
3092 if (inode->i_nlink) {
3093 if (test_opt(sb, DEBUG))
3094 ext4_msg(sb, KERN_DEBUG,
3095 "%s: truncating inode %lu to %lld bytes",
3096 __func__, inode->i_ino, inode->i_size);
3097 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
3098 inode->i_ino, inode->i_size);
3100 truncate_inode_pages(inode->i_mapping, inode->i_size);
3101 ret = ext4_truncate(inode);
3104 * We need to clean up the in-core orphan list
3105 * manually if ext4_truncate() failed to get a
3106 * transaction handle.
3108 ext4_orphan_del(NULL, inode);
3109 ext4_std_error(inode->i_sb, ret);
3111 inode_unlock(inode);
3114 if (test_opt(sb, DEBUG))
3115 ext4_msg(sb, KERN_DEBUG,
3116 "%s: deleting unreferenced inode %lu",
3117 __func__, inode->i_ino);
3118 jbd_debug(2, "deleting unreferenced inode %lu\n",
3122 iput(inode); /* The delete magic happens here! */
3125 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
3128 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
3129 PLURAL(nr_orphans));
3131 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
3132 PLURAL(nr_truncates));
3134 /* Turn off quotas if they were enabled for orphan cleanup */
3136 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
3137 if (sb_dqopt(sb)->files[i])
3138 dquot_quota_off(sb, i);
3142 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
3146 * Maximal extent format file size.
3147 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3148 * extent format containers, within a sector_t, and within i_blocks
3149 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3150 * so that won't be a limiting factor.
3152 * However there is other limiting factor. We do store extents in the form
3153 * of starting block and length, hence the resulting length of the extent
3154 * covering maximum file size must fit into on-disk format containers as
3155 * well. Given that length is always by 1 unit bigger than max unit (because
3156 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3158 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3160 static loff_t ext4_max_size(int blkbits, int has_huge_files)
3163 loff_t upper_limit = MAX_LFS_FILESIZE;
3165 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3167 if (!has_huge_files) {
3168 upper_limit = (1LL << 32) - 1;
3170 /* total blocks in file system block size */
3171 upper_limit >>= (blkbits - 9);
3172 upper_limit <<= blkbits;
3176 * 32-bit extent-start container, ee_block. We lower the maxbytes
3177 * by one fs block, so ee_len can cover the extent of maximum file
3180 res = (1LL << 32) - 1;
3183 /* Sanity check against vm- & vfs- imposed limits */
3184 if (res > upper_limit)
3191 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
3192 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3193 * We need to be 1 filesystem block less than the 2^48 sector limit.
3195 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3197 unsigned long long upper_limit, res = EXT4_NDIR_BLOCKS;
3201 * This is calculated to be the largest file size for a dense, block
3202 * mapped file such that the file's total number of 512-byte sectors,
3203 * including data and all indirect blocks, does not exceed (2^48 - 1).
3205 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3206 * number of 512-byte sectors of the file.
3208 if (!has_huge_files) {
3210 * !has_huge_files or implies that the inode i_block field
3211 * represents total file blocks in 2^32 512-byte sectors ==
3212 * size of vfs inode i_blocks * 8
3214 upper_limit = (1LL << 32) - 1;
3216 /* total blocks in file system block size */
3217 upper_limit >>= (bits - 9);
3221 * We use 48 bit ext4_inode i_blocks
3222 * With EXT4_HUGE_FILE_FL set the i_blocks
3223 * represent total number of blocks in
3224 * file system block size
3226 upper_limit = (1LL << 48) - 1;
3230 /* indirect blocks */
3232 /* double indirect blocks */
3233 meta_blocks += 1 + (1LL << (bits-2));
3234 /* tripple indirect blocks */
3235 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
3237 upper_limit -= meta_blocks;
3238 upper_limit <<= bits;
3240 res += 1LL << (bits-2);
3241 res += 1LL << (2*(bits-2));
3242 res += 1LL << (3*(bits-2));
3244 if (res > upper_limit)
3247 if (res > MAX_LFS_FILESIZE)
3248 res = MAX_LFS_FILESIZE;
3253 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3254 ext4_fsblk_t logical_sb_block, int nr)
3256 struct ext4_sb_info *sbi = EXT4_SB(sb);
3257 ext4_group_t bg, first_meta_bg;
3260 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3262 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3263 return logical_sb_block + nr + 1;
3264 bg = sbi->s_desc_per_block * nr;
3265 if (ext4_bg_has_super(sb, bg))
3269 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3270 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3271 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3274 if (sb->s_blocksize == 1024 && nr == 0 &&
3275 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3278 return (has_super + ext4_group_first_block_no(sb, bg));
3282 * ext4_get_stripe_size: Get the stripe size.
3283 * @sbi: In memory super block info
3285 * If we have specified it via mount option, then
3286 * use the mount option value. If the value specified at mount time is
3287 * greater than the blocks per group use the super block value.
3288 * If the super block value is greater than blocks per group return 0.
3289 * Allocator needs it be less than blocks per group.
3292 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3294 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3295 unsigned long stripe_width =
3296 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3299 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3300 ret = sbi->s_stripe;
3301 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3303 else if (stride && stride <= sbi->s_blocks_per_group)
3309 * If the stripe width is 1, this makes no sense and
3310 * we set it to 0 to turn off stripe handling code.
3319 * Check whether this filesystem can be mounted based on
3320 * the features present and the RDONLY/RDWR mount requested.
3321 * Returns 1 if this filesystem can be mounted as requested,
3322 * 0 if it cannot be.
3324 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
3326 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3327 ext4_msg(sb, KERN_ERR,
3328 "Couldn't mount because of "
3329 "unsupported optional features (%x)",
3330 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3331 ~EXT4_FEATURE_INCOMPAT_SUPP));
3335 #ifndef CONFIG_UNICODE
3336 if (ext4_has_feature_casefold(sb)) {
3337 ext4_msg(sb, KERN_ERR,
3338 "Filesystem with casefold feature cannot be "
3339 "mounted without CONFIG_UNICODE");
3347 if (ext4_has_feature_readonly(sb)) {
3348 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3349 sb->s_flags |= SB_RDONLY;
3353 /* Check that feature set is OK for a read-write mount */
3354 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3355 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3356 "unsupported optional features (%x)",
3357 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3358 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3361 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3362 ext4_msg(sb, KERN_ERR,
3363 "Can't support bigalloc feature without "
3364 "extents feature\n");
3368 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3369 if (!readonly && (ext4_has_feature_quota(sb) ||
3370 ext4_has_feature_project(sb))) {
3371 ext4_msg(sb, KERN_ERR,
3372 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3375 #endif /* CONFIG_QUOTA */
3380 * This function is called once a day if we have errors logged
3381 * on the file system
3383 static void print_daily_error_info(struct timer_list *t)
3385 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3386 struct super_block *sb = sbi->s_sb;
3387 struct ext4_super_block *es = sbi->s_es;
3389 if (es->s_error_count)
3390 /* fsck newer than v1.41.13 is needed to clean this condition. */
3391 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3392 le32_to_cpu(es->s_error_count));
3393 if (es->s_first_error_time) {
3394 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3396 ext4_get_tstamp(es, s_first_error_time),
3397 (int) sizeof(es->s_first_error_func),
3398 es->s_first_error_func,
3399 le32_to_cpu(es->s_first_error_line));
3400 if (es->s_first_error_ino)
3401 printk(KERN_CONT ": inode %u",
3402 le32_to_cpu(es->s_first_error_ino));
3403 if (es->s_first_error_block)
3404 printk(KERN_CONT ": block %llu", (unsigned long long)
3405 le64_to_cpu(es->s_first_error_block));
3406 printk(KERN_CONT "\n");
3408 if (es->s_last_error_time) {
3409 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3411 ext4_get_tstamp(es, s_last_error_time),
3412 (int) sizeof(es->s_last_error_func),
3413 es->s_last_error_func,
3414 le32_to_cpu(es->s_last_error_line));
3415 if (es->s_last_error_ino)
3416 printk(KERN_CONT ": inode %u",
3417 le32_to_cpu(es->s_last_error_ino));
3418 if (es->s_last_error_block)
3419 printk(KERN_CONT ": block %llu", (unsigned long long)
3420 le64_to_cpu(es->s_last_error_block));
3421 printk(KERN_CONT "\n");
3423 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3426 /* Find next suitable group and run ext4_init_inode_table */
3427 static int ext4_run_li_request(struct ext4_li_request *elr)
3429 struct ext4_group_desc *gdp = NULL;
3430 struct super_block *sb = elr->lr_super;
3431 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3432 ext4_group_t group = elr->lr_next_group;
3433 unsigned int prefetch_ios = 0;
3437 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3438 elr->lr_next_group = ext4_mb_prefetch(sb, group,
3439 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3441 ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3443 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3445 if (group >= elr->lr_next_group) {
3447 if (elr->lr_first_not_zeroed != ngroups &&
3448 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3449 elr->lr_next_group = elr->lr_first_not_zeroed;
3450 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3457 for (; group < ngroups; group++) {
3458 gdp = ext4_get_group_desc(sb, group, NULL);
3464 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3468 if (group >= ngroups)
3472 start_time = ktime_get_real_ns();
3473 ret = ext4_init_inode_table(sb, group,
3474 elr->lr_timeout ? 0 : 1);
3475 trace_ext4_lazy_itable_init(sb, group);
3476 if (elr->lr_timeout == 0) {
3477 elr->lr_timeout = nsecs_to_jiffies((ktime_get_real_ns() - start_time) *
3478 EXT4_SB(elr->lr_super)->s_li_wait_mult);
3480 elr->lr_next_sched = jiffies + elr->lr_timeout;
3481 elr->lr_next_group = group + 1;
3487 * Remove lr_request from the list_request and free the
3488 * request structure. Should be called with li_list_mtx held
3490 static void ext4_remove_li_request(struct ext4_li_request *elr)
3495 list_del(&elr->lr_request);
3496 EXT4_SB(elr->lr_super)->s_li_request = NULL;
3500 static void ext4_unregister_li_request(struct super_block *sb)
3502 mutex_lock(&ext4_li_mtx);
3503 if (!ext4_li_info) {
3504 mutex_unlock(&ext4_li_mtx);
3508 mutex_lock(&ext4_li_info->li_list_mtx);
3509 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3510 mutex_unlock(&ext4_li_info->li_list_mtx);
3511 mutex_unlock(&ext4_li_mtx);
3514 static struct task_struct *ext4_lazyinit_task;
3517 * This is the function where ext4lazyinit thread lives. It walks
3518 * through the request list searching for next scheduled filesystem.
3519 * When such a fs is found, run the lazy initialization request
3520 * (ext4_rn_li_request) and keep track of the time spend in this
3521 * function. Based on that time we compute next schedule time of
3522 * the request. When walking through the list is complete, compute
3523 * next waking time and put itself into sleep.
3525 static int ext4_lazyinit_thread(void *arg)
3527 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3528 struct list_head *pos, *n;
3529 struct ext4_li_request *elr;
3530 unsigned long next_wakeup, cur;
3532 BUG_ON(NULL == eli);
3537 next_wakeup = MAX_JIFFY_OFFSET;
3539 mutex_lock(&eli->li_list_mtx);
3540 if (list_empty(&eli->li_request_list)) {
3541 mutex_unlock(&eli->li_list_mtx);
3544 list_for_each_safe(pos, n, &eli->li_request_list) {
3547 elr = list_entry(pos, struct ext4_li_request,
3550 if (time_before(jiffies, elr->lr_next_sched)) {
3551 if (time_before(elr->lr_next_sched, next_wakeup))
3552 next_wakeup = elr->lr_next_sched;
3555 if (down_read_trylock(&elr->lr_super->s_umount)) {
3556 if (sb_start_write_trylock(elr->lr_super)) {
3559 * We hold sb->s_umount, sb can not
3560 * be removed from the list, it is
3561 * now safe to drop li_list_mtx
3563 mutex_unlock(&eli->li_list_mtx);
3564 err = ext4_run_li_request(elr);
3565 sb_end_write(elr->lr_super);
3566 mutex_lock(&eli->li_list_mtx);
3569 up_read((&elr->lr_super->s_umount));
3571 /* error, remove the lazy_init job */
3573 ext4_remove_li_request(elr);
3577 elr->lr_next_sched = jiffies +
3579 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3581 if (time_before(elr->lr_next_sched, next_wakeup))
3582 next_wakeup = elr->lr_next_sched;
3584 mutex_unlock(&eli->li_list_mtx);
3589 if ((time_after_eq(cur, next_wakeup)) ||
3590 (MAX_JIFFY_OFFSET == next_wakeup)) {
3595 schedule_timeout_interruptible(next_wakeup - cur);
3597 if (kthread_should_stop()) {
3598 ext4_clear_request_list();
3605 * It looks like the request list is empty, but we need
3606 * to check it under the li_list_mtx lock, to prevent any
3607 * additions into it, and of course we should lock ext4_li_mtx
3608 * to atomically free the list and ext4_li_info, because at
3609 * this point another ext4 filesystem could be registering
3612 mutex_lock(&ext4_li_mtx);
3613 mutex_lock(&eli->li_list_mtx);
3614 if (!list_empty(&eli->li_request_list)) {
3615 mutex_unlock(&eli->li_list_mtx);
3616 mutex_unlock(&ext4_li_mtx);
3619 mutex_unlock(&eli->li_list_mtx);
3620 kfree(ext4_li_info);
3621 ext4_li_info = NULL;
3622 mutex_unlock(&ext4_li_mtx);
3627 static void ext4_clear_request_list(void)
3629 struct list_head *pos, *n;
3630 struct ext4_li_request *elr;
3632 mutex_lock(&ext4_li_info->li_list_mtx);
3633 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3634 elr = list_entry(pos, struct ext4_li_request,
3636 ext4_remove_li_request(elr);
3638 mutex_unlock(&ext4_li_info->li_list_mtx);
3641 static int ext4_run_lazyinit_thread(void)
3643 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3644 ext4_li_info, "ext4lazyinit");
3645 if (IS_ERR(ext4_lazyinit_task)) {
3646 int err = PTR_ERR(ext4_lazyinit_task);
3647 ext4_clear_request_list();
3648 kfree(ext4_li_info);
3649 ext4_li_info = NULL;
3650 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3651 "initialization thread\n",
3655 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3660 * Check whether it make sense to run itable init. thread or not.
3661 * If there is at least one uninitialized inode table, return
3662 * corresponding group number, else the loop goes through all
3663 * groups and return total number of groups.
3665 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3667 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3668 struct ext4_group_desc *gdp = NULL;
3670 if (!ext4_has_group_desc_csum(sb))
3673 for (group = 0; group < ngroups; group++) {
3674 gdp = ext4_get_group_desc(sb, group, NULL);
3678 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3685 static int ext4_li_info_new(void)
3687 struct ext4_lazy_init *eli = NULL;
3689 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3693 INIT_LIST_HEAD(&eli->li_request_list);
3694 mutex_init(&eli->li_list_mtx);
3696 eli->li_state |= EXT4_LAZYINIT_QUIT;
3703 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3706 struct ext4_li_request *elr;
3708 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3713 elr->lr_first_not_zeroed = start;
3714 if (test_opt(sb, PREFETCH_BLOCK_BITMAPS))
3715 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3717 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3718 elr->lr_next_group = start;
3722 * Randomize first schedule time of the request to
3723 * spread the inode table initialization requests
3726 elr->lr_next_sched = jiffies + (prandom_u32() %
3727 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3731 int ext4_register_li_request(struct super_block *sb,
3732 ext4_group_t first_not_zeroed)
3734 struct ext4_sb_info *sbi = EXT4_SB(sb);
3735 struct ext4_li_request *elr = NULL;
3736 ext4_group_t ngroups = sbi->s_groups_count;
3739 mutex_lock(&ext4_li_mtx);
3740 if (sbi->s_li_request != NULL) {
3742 * Reset timeout so it can be computed again, because
3743 * s_li_wait_mult might have changed.
3745 sbi->s_li_request->lr_timeout = 0;
3749 if (!test_opt(sb, PREFETCH_BLOCK_BITMAPS) &&
3750 (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3751 !test_opt(sb, INIT_INODE_TABLE)))
3754 elr = ext4_li_request_new(sb, first_not_zeroed);
3760 if (NULL == ext4_li_info) {
3761 ret = ext4_li_info_new();
3766 mutex_lock(&ext4_li_info->li_list_mtx);
3767 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3768 mutex_unlock(&ext4_li_info->li_list_mtx);
3770 sbi->s_li_request = elr;
3772 * set elr to NULL here since it has been inserted to
3773 * the request_list and the removal and free of it is
3774 * handled by ext4_clear_request_list from now on.
3778 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3779 ret = ext4_run_lazyinit_thread();
3784 mutex_unlock(&ext4_li_mtx);
3791 * We do not need to lock anything since this is called on
3794 static void ext4_destroy_lazyinit_thread(void)
3797 * If thread exited earlier
3798 * there's nothing to be done.
3800 if (!ext4_li_info || !ext4_lazyinit_task)
3803 kthread_stop(ext4_lazyinit_task);
3806 static int set_journal_csum_feature_set(struct super_block *sb)
3809 int compat, incompat;
3810 struct ext4_sb_info *sbi = EXT4_SB(sb);
3812 if (ext4_has_metadata_csum(sb)) {
3813 /* journal checksum v3 */
3815 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3817 /* journal checksum v1 */
3818 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3822 jbd2_journal_clear_features(sbi->s_journal,
3823 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3824 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3825 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3826 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3827 ret = jbd2_journal_set_features(sbi->s_journal,
3829 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3831 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3832 ret = jbd2_journal_set_features(sbi->s_journal,
3835 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3836 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3838 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3839 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3846 * Note: calculating the overhead so we can be compatible with
3847 * historical BSD practice is quite difficult in the face of
3848 * clusters/bigalloc. This is because multiple metadata blocks from
3849 * different block group can end up in the same allocation cluster.
3850 * Calculating the exact overhead in the face of clustered allocation
3851 * requires either O(all block bitmaps) in memory or O(number of block
3852 * groups**2) in time. We will still calculate the superblock for
3853 * older file systems --- and if we come across with a bigalloc file
3854 * system with zero in s_overhead_clusters the estimate will be close to
3855 * correct especially for very large cluster sizes --- but for newer
3856 * file systems, it's better to calculate this figure once at mkfs
3857 * time, and store it in the superblock. If the superblock value is
3858 * present (even for non-bigalloc file systems), we will use it.
3860 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3863 struct ext4_sb_info *sbi = EXT4_SB(sb);
3864 struct ext4_group_desc *gdp;
3865 ext4_fsblk_t first_block, last_block, b;
3866 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3867 int s, j, count = 0;
3868 int has_super = ext4_bg_has_super(sb, grp);
3870 if (!ext4_has_feature_bigalloc(sb))
3871 return (has_super + ext4_bg_num_gdb(sb, grp) +
3872 (has_super ? le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) : 0) +
3873 sbi->s_itb_per_group + 2);
3875 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3876 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3877 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3878 for (i = 0; i < ngroups; i++) {
3879 gdp = ext4_get_group_desc(sb, i, NULL);
3880 b = ext4_block_bitmap(sb, gdp);
3881 if (b >= first_block && b <= last_block) {
3882 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3885 b = ext4_inode_bitmap(sb, gdp);
3886 if (b >= first_block && b <= last_block) {
3887 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3890 b = ext4_inode_table(sb, gdp);
3891 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3892 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3893 int c = EXT4_B2C(sbi, b - first_block);
3894 ext4_set_bit(c, buf);
3900 if (ext4_bg_has_super(sb, grp)) {
3901 ext4_set_bit(s++, buf);
3904 j = ext4_bg_num_gdb(sb, grp);
3905 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3906 ext4_error(sb, "Invalid number of block group "
3907 "descriptor blocks: %d", j);
3908 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3912 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3916 return EXT4_CLUSTERS_PER_GROUP(sb) -
3917 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3921 * Compute the overhead and stash it in sbi->s_overhead
3923 int ext4_calculate_overhead(struct super_block *sb)
3925 struct ext4_sb_info *sbi = EXT4_SB(sb);
3926 struct ext4_super_block *es = sbi->s_es;
3927 struct inode *j_inode;
3928 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3929 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3930 ext4_fsblk_t overhead = 0;
3931 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3937 * Compute the overhead (FS structures). This is constant
3938 * for a given filesystem unless the number of block groups
3939 * changes so we cache the previous value until it does.
3943 * All of the blocks before first_data_block are overhead
3945 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3948 * Add the overhead found in each block group
3950 for (i = 0; i < ngroups; i++) {
3953 blks = count_overhead(sb, i, buf);
3956 memset(buf, 0, PAGE_SIZE);
3961 * Add the internal journal blocks whether the journal has been
3964 if (sbi->s_journal && !sbi->s_journal_bdev)
3965 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
3966 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
3967 /* j_inum for internal journal is non-zero */
3968 j_inode = ext4_get_journal_inode(sb, j_inum);
3970 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3971 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3974 ext4_msg(sb, KERN_ERR, "can't get journal size");
3977 sbi->s_overhead = overhead;
3979 free_page((unsigned long) buf);
3983 static void ext4_set_resv_clusters(struct super_block *sb)
3985 ext4_fsblk_t resv_clusters;
3986 struct ext4_sb_info *sbi = EXT4_SB(sb);
3989 * There's no need to reserve anything when we aren't using extents.
3990 * The space estimates are exact, there are no unwritten extents,
3991 * hole punching doesn't need new metadata... This is needed especially
3992 * to keep ext2/3 backward compatibility.
3994 if (!ext4_has_feature_extents(sb))
3997 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3998 * This should cover the situations where we can not afford to run
3999 * out of space like for example punch hole, or converting
4000 * unwritten extents in delalloc path. In most cases such
4001 * allocation would require 1, or 2 blocks, higher numbers are
4004 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
4005 sbi->s_cluster_bits);
4007 do_div(resv_clusters, 50);
4008 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
4010 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
4013 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
4015 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
4016 char *orig_data = kstrdup(data, GFP_KERNEL);
4017 struct buffer_head *bh, **group_desc;
4018 struct ext4_super_block *es = NULL;
4019 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
4020 struct flex_groups **flex_groups;
4022 ext4_fsblk_t sb_block = get_sb_block(&data);
4023 ext4_fsblk_t logical_sb_block;
4024 unsigned long offset = 0;
4025 unsigned long journal_devnum = 0;
4026 unsigned long def_mount_opts;
4030 int blocksize, clustersize;
4031 unsigned int db_count;
4033 int needs_recovery, has_huge_files;
4036 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4037 ext4_group_t first_not_zeroed;
4039 if ((data && !orig_data) || !sbi)
4042 sbi->s_daxdev = dax_dev;
4043 sbi->s_blockgroup_lock =
4044 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
4045 if (!sbi->s_blockgroup_lock)
4048 sb->s_fs_info = sbi;
4050 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
4051 sbi->s_sb_block = sb_block;
4052 if (sb->s_bdev->bd_part)
4053 sbi->s_sectors_written_start =
4054 part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]);
4056 /* Cleanup superblock name */
4057 strreplace(sb->s_id, '/', '!');
4059 /* -EINVAL is default */
4061 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
4063 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
4068 * The ext4 superblock will not be buffer aligned for other than 1kB
4069 * block sizes. We need to calculate the offset from buffer start.
4071 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
4072 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4073 offset = do_div(logical_sb_block, blocksize);
4075 logical_sb_block = sb_block;
4078 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4080 ext4_msg(sb, KERN_ERR, "unable to read superblock");
4086 * Note: s_es must be initialized as soon as possible because
4087 * some ext4 macro-instructions depend on its value
4089 es = (struct ext4_super_block *) (bh->b_data + offset);
4091 sb->s_magic = le16_to_cpu(es->s_magic);
4092 if (sb->s_magic != EXT4_SUPER_MAGIC)
4094 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
4096 /* Warn if metadata_csum and gdt_csum are both set. */
4097 if (ext4_has_feature_metadata_csum(sb) &&
4098 ext4_has_feature_gdt_csum(sb))
4099 ext4_warning(sb, "metadata_csum and uninit_bg are "
4100 "redundant flags; please run fsck.");
4102 /* Check for a known checksum algorithm */
4103 if (!ext4_verify_csum_type(sb, es)) {
4104 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4105 "unknown checksum algorithm.");
4110 /* Load the checksum driver */
4111 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
4112 if (IS_ERR(sbi->s_chksum_driver)) {
4113 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
4114 ret = PTR_ERR(sbi->s_chksum_driver);
4115 sbi->s_chksum_driver = NULL;
4119 /* Check superblock checksum */
4120 if (!ext4_superblock_csum_verify(sb, es)) {
4121 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4122 "invalid superblock checksum. Run e2fsck?");
4128 /* Precompute checksum seed for all metadata */
4129 if (ext4_has_feature_csum_seed(sb))
4130 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
4131 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
4132 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
4133 sizeof(es->s_uuid));
4135 /* Set defaults before we parse the mount options */
4136 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4137 set_opt(sb, INIT_INODE_TABLE);
4138 if (def_mount_opts & EXT4_DEFM_DEBUG)
4140 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4142 if (def_mount_opts & EXT4_DEFM_UID16)
4143 set_opt(sb, NO_UID32);
4144 /* xattr user namespace & acls are now defaulted on */
4145 set_opt(sb, XATTR_USER);
4146 #ifdef CONFIG_EXT4_FS_POSIX_ACL
4147 set_opt(sb, POSIX_ACL);
4149 if (ext4_has_feature_fast_commit(sb))
4150 set_opt2(sb, JOURNAL_FAST_COMMIT);
4151 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4152 if (ext4_has_metadata_csum(sb))
4153 set_opt(sb, JOURNAL_CHECKSUM);
4155 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4156 set_opt(sb, JOURNAL_DATA);
4157 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4158 set_opt(sb, ORDERED_DATA);
4159 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4160 set_opt(sb, WRITEBACK_DATA);
4162 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
4163 set_opt(sb, ERRORS_PANIC);
4164 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
4165 set_opt(sb, ERRORS_CONT);
4167 set_opt(sb, ERRORS_RO);
4168 /* block_validity enabled by default; disable with noblock_validity */
4169 set_opt(sb, BLOCK_VALIDITY);
4170 if (def_mount_opts & EXT4_DEFM_DISCARD)
4171 set_opt(sb, DISCARD);
4173 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
4174 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
4175 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
4176 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
4177 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
4179 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4180 set_opt(sb, BARRIER);
4183 * enable delayed allocation by default
4184 * Use -o nodelalloc to turn it off
4186 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4187 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4188 set_opt(sb, DELALLOC);
4191 * set default s_li_wait_mult for lazyinit, for the case there is
4192 * no mount option specified.
4194 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
4196 if (le32_to_cpu(es->s_log_block_size) >
4197 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4198 ext4_msg(sb, KERN_ERR,
4199 "Invalid log block size: %u",
4200 le32_to_cpu(es->s_log_block_size));
4203 if (le32_to_cpu(es->s_log_cluster_size) >
4204 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4205 ext4_msg(sb, KERN_ERR,
4206 "Invalid log cluster size: %u",
4207 le32_to_cpu(es->s_log_cluster_size));
4211 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
4213 if (blocksize == PAGE_SIZE)
4214 set_opt(sb, DIOREAD_NOLOCK);
4216 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4217 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4218 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4220 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4221 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4222 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4223 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4227 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4228 (!is_power_of_2(sbi->s_inode_size)) ||
4229 (sbi->s_inode_size > blocksize)) {
4230 ext4_msg(sb, KERN_ERR,
4231 "unsupported inode size: %d",
4233 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
4237 * i_atime_extra is the last extra field available for
4238 * [acm]times in struct ext4_inode. Checking for that
4239 * field should suffice to ensure we have extra space
4242 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4243 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4244 sb->s_time_gran = 1;
4245 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4247 sb->s_time_gran = NSEC_PER_SEC;
4248 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4250 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4252 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4253 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4254 EXT4_GOOD_OLD_INODE_SIZE;
4255 if (ext4_has_feature_extra_isize(sb)) {
4256 unsigned v, max = (sbi->s_inode_size -
4257 EXT4_GOOD_OLD_INODE_SIZE);
4259 v = le16_to_cpu(es->s_want_extra_isize);
4261 ext4_msg(sb, KERN_ERR,
4262 "bad s_want_extra_isize: %d", v);
4265 if (sbi->s_want_extra_isize < v)
4266 sbi->s_want_extra_isize = v;
4268 v = le16_to_cpu(es->s_min_extra_isize);
4270 ext4_msg(sb, KERN_ERR,
4271 "bad s_min_extra_isize: %d", v);
4274 if (sbi->s_want_extra_isize < v)
4275 sbi->s_want_extra_isize = v;
4279 if (sbi->s_es->s_mount_opts[0]) {
4280 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
4281 sizeof(sbi->s_es->s_mount_opts),
4285 if (!parse_options(s_mount_opts, sb, &journal_devnum,
4286 &journal_ioprio, 0)) {
4287 ext4_msg(sb, KERN_WARNING,
4288 "failed to parse options in superblock: %s",
4291 kfree(s_mount_opts);
4293 sbi->s_def_mount_opt = sbi->s_mount_opt;
4294 if (!parse_options((char *) data, sb, &journal_devnum,
4295 &journal_ioprio, 0))
4298 #ifdef CONFIG_UNICODE
4299 if (ext4_has_feature_casefold(sb) && !sb->s_encoding) {
4300 const struct ext4_sb_encodings *encoding_info;
4301 struct unicode_map *encoding;
4302 __u16 encoding_flags;
4304 if (ext4_has_feature_encrypt(sb)) {
4305 ext4_msg(sb, KERN_ERR,
4306 "Can't mount with encoding and encryption");
4310 if (ext4_sb_read_encoding(es, &encoding_info,
4312 ext4_msg(sb, KERN_ERR,
4313 "Encoding requested by superblock is unknown");
4317 encoding = utf8_load(encoding_info->version);
4318 if (IS_ERR(encoding)) {
4319 ext4_msg(sb, KERN_ERR,
4320 "can't mount with superblock charset: %s-%s "
4321 "not supported by the kernel. flags: 0x%x.",
4322 encoding_info->name, encoding_info->version,
4326 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4327 "%s-%s with flags 0x%hx", encoding_info->name,
4328 encoding_info->version?:"\b", encoding_flags);
4330 sb->s_encoding = encoding;
4331 sb->s_encoding_flags = encoding_flags;
4335 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4336 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, O_DIRECT and fast_commit support!\n");
4337 /* can't mount with both data=journal and dioread_nolock. */
4338 clear_opt(sb, DIOREAD_NOLOCK);
4339 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4340 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4341 ext4_msg(sb, KERN_ERR, "can't mount with "
4342 "both data=journal and delalloc");
4345 if (test_opt(sb, DAX_ALWAYS)) {
4346 ext4_msg(sb, KERN_ERR, "can't mount with "
4347 "both data=journal and dax");
4350 if (ext4_has_feature_encrypt(sb)) {
4351 ext4_msg(sb, KERN_WARNING,
4352 "encrypted files will use data=ordered "
4353 "instead of data journaling mode");
4355 if (test_opt(sb, DELALLOC))
4356 clear_opt(sb, DELALLOC);
4358 sb->s_iflags |= SB_I_CGROUPWB;
4361 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4362 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
4364 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4365 (ext4_has_compat_features(sb) ||
4366 ext4_has_ro_compat_features(sb) ||
4367 ext4_has_incompat_features(sb)))
4368 ext4_msg(sb, KERN_WARNING,
4369 "feature flags set on rev 0 fs, "
4370 "running e2fsck is recommended");
4372 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4373 set_opt2(sb, HURD_COMPAT);
4374 if (ext4_has_feature_64bit(sb)) {
4375 ext4_msg(sb, KERN_ERR,
4376 "The Hurd can't support 64-bit file systems");
4381 * ea_inode feature uses l_i_version field which is not
4382 * available in HURD_COMPAT mode.
4384 if (ext4_has_feature_ea_inode(sb)) {
4385 ext4_msg(sb, KERN_ERR,
4386 "ea_inode feature is not supported for Hurd");
4391 if (IS_EXT2_SB(sb)) {
4392 if (ext2_feature_set_ok(sb))
4393 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4394 "using the ext4 subsystem");
4397 * If we're probing be silent, if this looks like
4398 * it's actually an ext[34] filesystem.
4400 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4402 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4403 "to feature incompatibilities");
4408 if (IS_EXT3_SB(sb)) {
4409 if (ext3_feature_set_ok(sb))
4410 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4411 "using the ext4 subsystem");
4414 * If we're probing be silent, if this looks like
4415 * it's actually an ext4 filesystem.
4417 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4419 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4420 "to feature incompatibilities");
4426 * Check feature flags regardless of the revision level, since we
4427 * previously didn't change the revision level when setting the flags,
4428 * so there is a chance incompat flags are set on a rev 0 filesystem.
4430 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4433 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4434 ext4_msg(sb, KERN_ERR,
4435 "Number of reserved GDT blocks insanely large: %d",
4436 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4440 if (bdev_dax_supported(sb->s_bdev, blocksize))
4441 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4443 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4444 if (ext4_has_feature_inline_data(sb)) {
4445 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4446 " that may contain inline data");
4449 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4450 ext4_msg(sb, KERN_ERR,
4451 "DAX unsupported by block device.");
4456 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4457 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4458 es->s_encryption_level);
4462 if (sb->s_blocksize != blocksize) {
4464 * bh must be released before kill_bdev(), otherwise
4465 * it won't be freed and its page also. kill_bdev()
4466 * is called by sb_set_blocksize().
4469 /* Validate the filesystem blocksize */
4470 if (!sb_set_blocksize(sb, blocksize)) {
4471 ext4_msg(sb, KERN_ERR, "bad block size %d",
4477 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4478 offset = do_div(logical_sb_block, blocksize);
4479 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4481 ext4_msg(sb, KERN_ERR,
4482 "Can't read superblock on 2nd try");
4487 es = (struct ext4_super_block *)(bh->b_data + offset);
4489 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
4490 ext4_msg(sb, KERN_ERR,
4491 "Magic mismatch, very weird!");
4496 has_huge_files = ext4_has_feature_huge_file(sb);
4497 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4499 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
4501 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
4502 if (ext4_has_feature_64bit(sb)) {
4503 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
4504 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
4505 !is_power_of_2(sbi->s_desc_size)) {
4506 ext4_msg(sb, KERN_ERR,
4507 "unsupported descriptor size %lu",
4512 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
4514 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
4515 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
4517 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
4518 if (sbi->s_inodes_per_block == 0)
4520 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4521 sbi->s_inodes_per_group > blocksize * 8) {
4522 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
4523 sbi->s_inodes_per_group);
4526 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4527 sbi->s_inodes_per_block;
4528 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
4530 sbi->s_mount_state = le16_to_cpu(es->s_state) & ~EXT4_FC_REPLAY;
4531 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4532 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
4534 for (i = 0; i < 4; i++)
4535 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4536 sbi->s_def_hash_version = es->s_def_hash_version;
4537 if (ext4_has_feature_dir_index(sb)) {
4538 i = le32_to_cpu(es->s_flags);
4539 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4540 sbi->s_hash_unsigned = 3;
4541 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
4542 #ifdef __CHAR_UNSIGNED__
4545 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4546 sbi->s_hash_unsigned = 3;
4550 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
4555 /* Handle clustersize */
4556 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4557 if (ext4_has_feature_bigalloc(sb)) {
4558 if (clustersize < blocksize) {
4559 ext4_msg(sb, KERN_ERR,
4560 "cluster size (%d) smaller than "
4561 "block size (%d)", clustersize, blocksize);
4564 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4565 le32_to_cpu(es->s_log_block_size);
4566 sbi->s_clusters_per_group =
4567 le32_to_cpu(es->s_clusters_per_group);
4568 if (sbi->s_clusters_per_group > blocksize * 8) {
4569 ext4_msg(sb, KERN_ERR,
4570 "#clusters per group too big: %lu",
4571 sbi->s_clusters_per_group);
4574 if (sbi->s_blocks_per_group !=
4575 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4576 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4577 "clusters per group (%lu) inconsistent",
4578 sbi->s_blocks_per_group,
4579 sbi->s_clusters_per_group);
4583 if (clustersize != blocksize) {
4584 ext4_msg(sb, KERN_ERR,
4585 "fragment/cluster size (%d) != "
4586 "block size (%d)", clustersize, blocksize);
4589 if (sbi->s_blocks_per_group > blocksize * 8) {
4590 ext4_msg(sb, KERN_ERR,
4591 "#blocks per group too big: %lu",
4592 sbi->s_blocks_per_group);
4595 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4596 sbi->s_cluster_bits = 0;
4598 sbi->s_cluster_ratio = clustersize / blocksize;
4600 /* Do we have standard group size of clustersize * 8 blocks ? */
4601 if (sbi->s_blocks_per_group == clustersize << 3)
4602 set_opt2(sb, STD_GROUP_SIZE);
4605 * Test whether we have more sectors than will fit in sector_t,
4606 * and whether the max offset is addressable by the page cache.
4608 err = generic_check_addressable(sb->s_blocksize_bits,
4609 ext4_blocks_count(es));
4611 ext4_msg(sb, KERN_ERR, "filesystem"
4612 " too large to mount safely on this system");
4616 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4619 /* check blocks count against device size */
4620 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4621 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4622 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4623 "exceeds size of device (%llu blocks)",
4624 ext4_blocks_count(es), blocks_count);
4629 * It makes no sense for the first data block to be beyond the end
4630 * of the filesystem.
4632 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4633 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4634 "block %u is beyond end of filesystem (%llu)",
4635 le32_to_cpu(es->s_first_data_block),
4636 ext4_blocks_count(es));
4639 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4640 (sbi->s_cluster_ratio == 1)) {
4641 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4642 "block is 0 with a 1k block and cluster size");
4646 blocks_count = (ext4_blocks_count(es) -
4647 le32_to_cpu(es->s_first_data_block) +
4648 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4649 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4650 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4651 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4652 "(block count %llu, first data block %u, "
4653 "blocks per group %lu)", blocks_count,
4654 ext4_blocks_count(es),
4655 le32_to_cpu(es->s_first_data_block),
4656 EXT4_BLOCKS_PER_GROUP(sb));
4659 sbi->s_groups_count = blocks_count;
4660 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4661 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4662 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4663 le32_to_cpu(es->s_inodes_count)) {
4664 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4665 le32_to_cpu(es->s_inodes_count),
4666 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4670 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4671 EXT4_DESC_PER_BLOCK(sb);
4672 if (ext4_has_feature_meta_bg(sb)) {
4673 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4674 ext4_msg(sb, KERN_WARNING,
4675 "first meta block group too large: %u "
4676 "(group descriptor block count %u)",
4677 le32_to_cpu(es->s_first_meta_bg), db_count);
4681 rcu_assign_pointer(sbi->s_group_desc,
4682 kvmalloc_array(db_count,
4683 sizeof(struct buffer_head *),
4685 if (sbi->s_group_desc == NULL) {
4686 ext4_msg(sb, KERN_ERR, "not enough memory");
4691 bgl_lock_init(sbi->s_blockgroup_lock);
4693 /* Pre-read the descriptors into the buffer cache */
4694 for (i = 0; i < db_count; i++) {
4695 block = descriptor_loc(sb, logical_sb_block, i);
4696 ext4_sb_breadahead_unmovable(sb, block);
4699 for (i = 0; i < db_count; i++) {
4700 struct buffer_head *bh;
4702 block = descriptor_loc(sb, logical_sb_block, i);
4703 bh = ext4_sb_bread_unmovable(sb, block);
4705 ext4_msg(sb, KERN_ERR,
4706 "can't read group descriptor %d", i);
4713 rcu_dereference(sbi->s_group_desc)[i] = bh;
4716 sbi->s_gdb_count = db_count;
4717 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4718 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4719 ret = -EFSCORRUPTED;
4723 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
4725 /* Register extent status tree shrinker */
4726 if (ext4_es_register_shrinker(sbi))
4729 sbi->s_stripe = ext4_get_stripe_size(sbi);
4730 sbi->s_extent_max_zeroout_kb = 32;
4733 * set up enough so that it can read an inode
4735 sb->s_op = &ext4_sops;
4736 sb->s_export_op = &ext4_export_ops;
4737 sb->s_xattr = ext4_xattr_handlers;
4738 #ifdef CONFIG_FS_ENCRYPTION
4739 sb->s_cop = &ext4_cryptops;
4741 #ifdef CONFIG_FS_VERITY
4742 sb->s_vop = &ext4_verityops;
4745 sb->dq_op = &ext4_quota_operations;
4746 if (ext4_has_feature_quota(sb))
4747 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4749 sb->s_qcop = &ext4_qctl_operations;
4750 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4752 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4754 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4755 mutex_init(&sbi->s_orphan_lock);
4757 /* Initialize fast commit stuff */
4758 atomic_set(&sbi->s_fc_subtid, 0);
4759 atomic_set(&sbi->s_fc_ineligible_updates, 0);
4760 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
4761 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
4762 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
4763 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
4764 sbi->s_fc_bytes = 0;
4765 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
4766 ext4_clear_mount_flag(sb, EXT4_MF_FC_COMMITTING);
4767 spin_lock_init(&sbi->s_fc_lock);
4768 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
4769 sbi->s_fc_replay_state.fc_regions = NULL;
4770 sbi->s_fc_replay_state.fc_regions_size = 0;
4771 sbi->s_fc_replay_state.fc_regions_used = 0;
4772 sbi->s_fc_replay_state.fc_regions_valid = 0;
4773 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
4774 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
4775 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
4779 needs_recovery = (es->s_last_orphan != 0 ||
4780 ext4_has_feature_journal_needs_recovery(sb));
4782 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) {
4783 err = ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block));
4785 goto failed_mount3a;
4789 * The first inode we look at is the journal inode. Don't try
4790 * root first: it may be modified in the journal!
4792 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4793 err = ext4_load_journal(sb, es, journal_devnum);
4795 goto failed_mount3a;
4796 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4797 ext4_has_feature_journal_needs_recovery(sb)) {
4798 ext4_msg(sb, KERN_ERR, "required journal recovery "
4799 "suppressed and not mounted read-only");
4800 goto failed_mount3a;
4802 /* Nojournal mode, all journal mount options are illegal */
4803 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4804 ext4_msg(sb, KERN_ERR, "can't mount with "
4805 "journal_async_commit, fs mounted w/o journal");
4806 goto failed_mount3a;
4809 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4810 ext4_msg(sb, KERN_ERR, "can't mount with "
4811 "journal_checksum, fs mounted w/o journal");
4812 goto failed_mount3a;
4814 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4815 ext4_msg(sb, KERN_ERR, "can't mount with "
4816 "commit=%lu, fs mounted w/o journal",
4817 sbi->s_commit_interval / HZ);
4818 goto failed_mount3a;
4820 if (EXT4_MOUNT_DATA_FLAGS &
4821 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4822 ext4_msg(sb, KERN_ERR, "can't mount with "
4823 "data=, fs mounted w/o journal");
4824 goto failed_mount3a;
4826 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
4827 clear_opt(sb, JOURNAL_CHECKSUM);
4828 clear_opt(sb, DATA_FLAGS);
4829 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4830 sbi->s_journal = NULL;
4835 if (ext4_has_feature_64bit(sb) &&
4836 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4837 JBD2_FEATURE_INCOMPAT_64BIT)) {
4838 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4839 goto failed_mount_wq;
4842 if (!set_journal_csum_feature_set(sb)) {
4843 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4845 goto failed_mount_wq;
4848 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
4849 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4850 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
4851 ext4_msg(sb, KERN_ERR,
4852 "Failed to set fast commit journal feature");
4853 goto failed_mount_wq;
4856 /* We have now updated the journal if required, so we can
4857 * validate the data journaling mode. */
4858 switch (test_opt(sb, DATA_FLAGS)) {
4860 /* No mode set, assume a default based on the journal
4861 * capabilities: ORDERED_DATA if the journal can
4862 * cope, else JOURNAL_DATA
4864 if (jbd2_journal_check_available_features
4865 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4866 set_opt(sb, ORDERED_DATA);
4867 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4869 set_opt(sb, JOURNAL_DATA);
4870 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4874 case EXT4_MOUNT_ORDERED_DATA:
4875 case EXT4_MOUNT_WRITEBACK_DATA:
4876 if (!jbd2_journal_check_available_features
4877 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4878 ext4_msg(sb, KERN_ERR, "Journal does not support "
4879 "requested data journaling mode");
4880 goto failed_mount_wq;
4886 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4887 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4888 ext4_msg(sb, KERN_ERR, "can't mount with "
4889 "journal_async_commit in data=ordered mode");
4890 goto failed_mount_wq;
4893 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4895 sbi->s_journal->j_submit_inode_data_buffers =
4896 ext4_journal_submit_inode_data_buffers;
4897 sbi->s_journal->j_finish_inode_data_buffers =
4898 ext4_journal_finish_inode_data_buffers;
4901 if (!test_opt(sb, NO_MBCACHE)) {
4902 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4903 if (!sbi->s_ea_block_cache) {
4904 ext4_msg(sb, KERN_ERR,
4905 "Failed to create ea_block_cache");
4906 goto failed_mount_wq;
4909 if (ext4_has_feature_ea_inode(sb)) {
4910 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4911 if (!sbi->s_ea_inode_cache) {
4912 ext4_msg(sb, KERN_ERR,
4913 "Failed to create ea_inode_cache");
4914 goto failed_mount_wq;
4919 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
4920 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
4921 goto failed_mount_wq;
4925 * Get the # of file system overhead blocks from the
4926 * superblock if present.
4928 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4929 /* ignore the precalculated value if it is ridiculous */
4930 if (sbi->s_overhead > ext4_blocks_count(es))
4931 sbi->s_overhead = 0;
4933 * If the bigalloc feature is not enabled recalculating the
4934 * overhead doesn't take long, so we might as well just redo
4935 * it to make sure we are using the correct value.
4937 if (!ext4_has_feature_bigalloc(sb))
4938 sbi->s_overhead = 0;
4939 if (sbi->s_overhead == 0) {
4940 err = ext4_calculate_overhead(sb);
4942 goto failed_mount_wq;
4946 * The maximum number of concurrent works can be high and
4947 * concurrency isn't really necessary. Limit it to 1.
4949 EXT4_SB(sb)->rsv_conversion_wq =
4950 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4951 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4952 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4958 * The jbd2_journal_load will have done any necessary log recovery,
4959 * so we can safely mount the rest of the filesystem now.
4962 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
4964 ext4_msg(sb, KERN_ERR, "get root inode failed");
4965 ret = PTR_ERR(root);
4969 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4970 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4975 #ifdef CONFIG_UNICODE
4977 sb->s_d_op = &ext4_dentry_ops;
4980 sb->s_root = d_make_root(root);
4982 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4987 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4988 if (ret == -EROFS) {
4989 sb->s_flags |= SB_RDONLY;
4992 goto failed_mount4a;
4994 ext4_set_resv_clusters(sb);
4996 if (test_opt(sb, BLOCK_VALIDITY)) {
4997 err = ext4_setup_system_zone(sb);
4999 ext4_msg(sb, KERN_ERR, "failed to initialize system "
5001 goto failed_mount4a;
5004 ext4_fc_replay_cleanup(sb);
5007 err = ext4_mb_init(sb);
5009 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
5015 * We can only set up the journal commit callback once
5016 * mballoc is initialized
5019 sbi->s_journal->j_commit_callback =
5020 ext4_journal_commit_callback;
5022 block = ext4_count_free_clusters(sb);
5023 ext4_free_blocks_count_set(sbi->s_es,
5024 EXT4_C2B(sbi, block));
5025 ext4_superblock_csum_set(sb);
5026 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
5029 unsigned long freei = ext4_count_free_inodes(sb);
5030 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
5031 ext4_superblock_csum_set(sb);
5032 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
5036 err = percpu_counter_init(&sbi->s_dirs_counter,
5037 ext4_count_dirs(sb), GFP_KERNEL);
5039 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
5042 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0,
5045 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
5048 ext4_msg(sb, KERN_ERR, "insufficient memory");
5052 if (ext4_has_feature_flex_bg(sb))
5053 if (!ext4_fill_flex_info(sb)) {
5054 ext4_msg(sb, KERN_ERR,
5055 "unable to initialize "
5056 "flex_bg meta info!");
5061 err = ext4_register_li_request(sb, first_not_zeroed);
5065 err = ext4_register_sysfs(sb);
5070 /* Enable quota usage during mount. */
5071 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
5072 err = ext4_enable_quotas(sb);
5076 #endif /* CONFIG_QUOTA */
5079 * Save the original bdev mapping's wb_err value which could be
5080 * used to detect the metadata async write error.
5082 spin_lock_init(&sbi->s_bdev_wb_lock);
5083 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
5084 &sbi->s_bdev_wb_err);
5085 sb->s_bdev->bd_super = sb;
5086 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
5087 ext4_orphan_cleanup(sb, es);
5088 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
5089 if (needs_recovery) {
5090 ext4_msg(sb, KERN_INFO, "recovery complete");
5091 err = ext4_mark_recovery_complete(sb, es);
5095 if (EXT4_SB(sb)->s_journal) {
5096 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
5097 descr = " journalled data mode";
5098 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
5099 descr = " ordered data mode";
5101 descr = " writeback data mode";
5103 descr = "out journal";
5105 if (test_opt(sb, DISCARD)) {
5106 struct request_queue *q = bdev_get_queue(sb->s_bdev);
5107 if (!blk_queue_discard(q))
5108 ext4_msg(sb, KERN_WARNING,
5109 "mounting with \"discard\" option, but "
5110 "the device does not support discard");
5113 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
5114 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5115 "Opts: %.*s%s%s", descr,
5116 (int) sizeof(sbi->s_es->s_mount_opts),
5117 sbi->s_es->s_mount_opts,
5118 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
5120 if (es->s_error_count)
5121 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
5123 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5124 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
5125 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
5126 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
5127 atomic_set(&sbi->s_warning_count, 0);
5128 atomic_set(&sbi->s_msg_count, 0);
5135 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5139 ext4_unregister_sysfs(sb);
5140 kobject_put(&sbi->s_kobj);
5142 ext4_unregister_li_request(sb);
5144 ext4_mb_release(sb);
5146 flex_groups = rcu_dereference(sbi->s_flex_groups);
5148 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
5149 kvfree(flex_groups[i]);
5150 kvfree(flex_groups);
5153 percpu_counter_destroy(&sbi->s_freeclusters_counter);
5154 percpu_counter_destroy(&sbi->s_freeinodes_counter);
5155 percpu_counter_destroy(&sbi->s_dirs_counter);
5156 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
5157 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
5158 percpu_free_rwsem(&sbi->s_writepages_rwsem);
5160 ext4_ext_release(sb);
5161 ext4_release_system_zone(sb);
5166 ext4_msg(sb, KERN_ERR, "mount failed");
5167 if (EXT4_SB(sb)->rsv_conversion_wq)
5168 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
5170 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5171 sbi->s_ea_inode_cache = NULL;
5173 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5174 sbi->s_ea_block_cache = NULL;
5176 if (sbi->s_journal) {
5177 jbd2_journal_destroy(sbi->s_journal);
5178 sbi->s_journal = NULL;
5181 ext4_es_unregister_shrinker(sbi);
5183 del_timer_sync(&sbi->s_err_report);
5184 ext4_stop_mmpd(sbi);
5187 group_desc = rcu_dereference(sbi->s_group_desc);
5188 for (i = 0; i < db_count; i++)
5189 brelse(group_desc[i]);
5193 if (sbi->s_chksum_driver)
5194 crypto_free_shash(sbi->s_chksum_driver);
5196 #ifdef CONFIG_UNICODE
5197 utf8_unload(sb->s_encoding);
5201 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5202 kfree(get_qf_name(sb, sbi, i));
5204 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
5205 /* ext4_blkdev_remove() calls kill_bdev(), release bh before it. */
5207 ext4_blkdev_remove(sbi);
5209 invalidate_bdev(sb->s_bdev);
5210 sb->s_fs_info = NULL;
5211 kfree(sbi->s_blockgroup_lock);
5215 fs_put_dax(dax_dev);
5216 return err ? err : ret;
5220 * Setup any per-fs journal parameters now. We'll do this both on
5221 * initial mount, once the journal has been initialised but before we've
5222 * done any recovery; and again on any subsequent remount.
5224 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
5226 struct ext4_sb_info *sbi = EXT4_SB(sb);
5228 journal->j_commit_interval = sbi->s_commit_interval;
5229 journal->j_min_batch_time = sbi->s_min_batch_time;
5230 journal->j_max_batch_time = sbi->s_max_batch_time;
5231 ext4_fc_init(sb, journal);
5233 write_lock(&journal->j_state_lock);
5234 if (test_opt(sb, BARRIER))
5235 journal->j_flags |= JBD2_BARRIER;
5237 journal->j_flags &= ~JBD2_BARRIER;
5238 if (test_opt(sb, DATA_ERR_ABORT))
5239 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5241 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
5242 write_unlock(&journal->j_state_lock);
5245 static struct inode *ext4_get_journal_inode(struct super_block *sb,
5246 unsigned int journal_inum)
5248 struct inode *journal_inode;
5251 * Test for the existence of a valid inode on disk. Bad things
5252 * happen if we iget() an unused inode, as the subsequent iput()
5253 * will try to delete it.
5255 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
5256 if (IS_ERR(journal_inode)) {
5257 ext4_msg(sb, KERN_ERR, "no journal found");
5260 if (!journal_inode->i_nlink) {
5261 make_bad_inode(journal_inode);
5262 iput(journal_inode);
5263 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
5267 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
5268 journal_inode, journal_inode->i_size);
5269 if (!S_ISREG(journal_inode->i_mode) || IS_ENCRYPTED(journal_inode)) {
5270 ext4_msg(sb, KERN_ERR, "invalid journal inode");
5271 iput(journal_inode);
5274 return journal_inode;
5277 static journal_t *ext4_get_journal(struct super_block *sb,
5278 unsigned int journal_inum)
5280 struct inode *journal_inode;
5283 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5286 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5290 journal = jbd2_journal_init_inode(journal_inode);
5292 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5293 iput(journal_inode);
5296 journal->j_private = sb;
5297 ext4_init_journal_params(sb, journal);
5301 static journal_t *ext4_get_dev_journal(struct super_block *sb,
5304 struct buffer_head *bh;
5308 int hblock, blocksize;
5309 ext4_fsblk_t sb_block;
5310 unsigned long offset;
5311 struct ext4_super_block *es;
5312 struct block_device *bdev;
5314 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5317 bdev = ext4_blkdev_get(j_dev, sb);
5321 blocksize = sb->s_blocksize;
5322 hblock = bdev_logical_block_size(bdev);
5323 if (blocksize < hblock) {
5324 ext4_msg(sb, KERN_ERR,
5325 "blocksize too small for journal device");
5329 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5330 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5331 set_blocksize(bdev, blocksize);
5332 if (!(bh = __bread(bdev, sb_block, blocksize))) {
5333 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5334 "external journal");
5338 es = (struct ext4_super_block *) (bh->b_data + offset);
5339 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5340 !(le32_to_cpu(es->s_feature_incompat) &
5341 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5342 ext4_msg(sb, KERN_ERR, "external journal has "
5348 if ((le32_to_cpu(es->s_feature_ro_compat) &
5349 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5350 es->s_checksum != ext4_superblock_csum(sb, es)) {
5351 ext4_msg(sb, KERN_ERR, "external journal has "
5352 "corrupt superblock");
5357 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5358 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5363 len = ext4_blocks_count(es);
5364 start = sb_block + 1;
5365 brelse(bh); /* we're done with the superblock */
5367 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
5368 start, len, blocksize);
5370 ext4_msg(sb, KERN_ERR, "failed to create device journal");
5373 journal->j_private = sb;
5374 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) {
5375 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
5378 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5379 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5380 "user (unsupported) - %d",
5381 be32_to_cpu(journal->j_superblock->s_nr_users));
5384 EXT4_SB(sb)->s_journal_bdev = bdev;
5385 ext4_init_journal_params(sb, journal);
5389 jbd2_journal_destroy(journal);
5391 ext4_blkdev_put(bdev);
5395 static int ext4_load_journal(struct super_block *sb,
5396 struct ext4_super_block *es,
5397 unsigned long journal_devnum)
5400 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5403 int really_read_only;
5406 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5407 return -EFSCORRUPTED;
5409 if (journal_devnum &&
5410 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5411 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5412 "numbers have changed");
5413 journal_dev = new_decode_dev(journal_devnum);
5415 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5417 if (journal_inum && journal_dev) {
5418 ext4_msg(sb, KERN_ERR,
5419 "filesystem has both journal inode and journal device!");
5424 journal = ext4_get_journal(sb, journal_inum);
5428 journal = ext4_get_dev_journal(sb, journal_dev);
5433 journal_dev_ro = bdev_read_only(journal->j_dev);
5434 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5436 if (journal_dev_ro && !sb_rdonly(sb)) {
5437 ext4_msg(sb, KERN_ERR,
5438 "journal device read-only, try mounting with '-o ro'");
5444 * Are we loading a blank journal or performing recovery after a
5445 * crash? For recovery, we need to check in advance whether we
5446 * can get read-write access to the device.
5448 if (ext4_has_feature_journal_needs_recovery(sb)) {
5449 if (sb_rdonly(sb)) {
5450 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5451 "required on readonly filesystem");
5452 if (really_read_only) {
5453 ext4_msg(sb, KERN_ERR, "write access "
5454 "unavailable, cannot proceed "
5455 "(try mounting with noload)");
5459 ext4_msg(sb, KERN_INFO, "write access will "
5460 "be enabled during recovery");
5464 if (!(journal->j_flags & JBD2_BARRIER))
5465 ext4_msg(sb, KERN_INFO, "barriers disabled");
5467 if (!ext4_has_feature_journal_needs_recovery(sb))
5468 err = jbd2_journal_wipe(journal, !really_read_only);
5470 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5472 memcpy(save, ((char *) es) +
5473 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
5474 err = jbd2_journal_load(journal);
5476 memcpy(((char *) es) + EXT4_S_ERR_START,
5477 save, EXT4_S_ERR_LEN);
5482 ext4_msg(sb, KERN_ERR, "error loading journal");
5486 EXT4_SB(sb)->s_journal = journal;
5487 err = ext4_clear_journal_err(sb, es);
5489 EXT4_SB(sb)->s_journal = NULL;
5490 jbd2_journal_destroy(journal);
5494 if (!really_read_only && journal_devnum &&
5495 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5496 es->s_journal_dev = cpu_to_le32(journal_devnum);
5498 /* Make sure we flush the recovery flag to disk. */
5499 ext4_commit_super(sb, 1);
5505 jbd2_journal_destroy(journal);
5509 static int ext4_commit_super(struct super_block *sb, int sync)
5511 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
5512 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
5517 if (block_device_ejected(sb))
5521 * If the file system is mounted read-only, don't update the
5522 * superblock write time. This avoids updating the superblock
5523 * write time when we are mounting the root file system
5524 * read/only but we need to replay the journal; at that point,
5525 * for people who are east of GMT and who make their clock
5526 * tick in localtime for Windows bug-for-bug compatibility,
5527 * the clock is set in the future, and this will cause e2fsck
5528 * to complain and force a full file system check.
5530 if (!(sb->s_flags & SB_RDONLY))
5531 ext4_update_tstamp(es, s_wtime);
5532 if (sb->s_bdev->bd_part)
5533 es->s_kbytes_written =
5534 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
5535 ((part_stat_read(sb->s_bdev->bd_part,
5536 sectors[STAT_WRITE]) -
5537 EXT4_SB(sb)->s_sectors_written_start) >> 1));
5539 es->s_kbytes_written =
5540 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
5541 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
5542 ext4_free_blocks_count_set(es,
5543 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
5544 &EXT4_SB(sb)->s_freeclusters_counter)));
5545 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
5546 es->s_free_inodes_count =
5547 cpu_to_le32(percpu_counter_sum_positive(
5548 &EXT4_SB(sb)->s_freeinodes_counter));
5549 BUFFER_TRACE(sbh, "marking dirty");
5550 ext4_superblock_csum_set(sb);
5553 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
5555 * Oh, dear. A previous attempt to write the
5556 * superblock failed. This could happen because the
5557 * USB device was yanked out. Or it could happen to
5558 * be a transient write error and maybe the block will
5559 * be remapped. Nothing we can do but to retry the
5560 * write and hope for the best.
5562 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5563 "superblock detected");
5564 clear_buffer_write_io_error(sbh);
5565 set_buffer_uptodate(sbh);
5567 mark_buffer_dirty(sbh);
5570 error = __sync_dirty_buffer(sbh,
5571 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
5572 if (buffer_write_io_error(sbh)) {
5573 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5575 clear_buffer_write_io_error(sbh);
5576 set_buffer_uptodate(sbh);
5583 * Have we just finished recovery? If so, and if we are mounting (or
5584 * remounting) the filesystem readonly, then we will end up with a
5585 * consistent fs on disk. Record that fact.
5587 static int ext4_mark_recovery_complete(struct super_block *sb,
5588 struct ext4_super_block *es)
5591 journal_t *journal = EXT4_SB(sb)->s_journal;
5593 if (!ext4_has_feature_journal(sb)) {
5594 if (journal != NULL) {
5595 ext4_error(sb, "Journal got removed while the fs was "
5597 return -EFSCORRUPTED;
5601 jbd2_journal_lock_updates(journal);
5602 err = jbd2_journal_flush(journal);
5606 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
5607 ext4_clear_feature_journal_needs_recovery(sb);
5608 ext4_commit_super(sb, 1);
5611 jbd2_journal_unlock_updates(journal);
5616 * If we are mounting (or read-write remounting) a filesystem whose journal
5617 * has recorded an error from a previous lifetime, move that error to the
5618 * main filesystem now.
5620 static int ext4_clear_journal_err(struct super_block *sb,
5621 struct ext4_super_block *es)
5627 if (!ext4_has_feature_journal(sb)) {
5628 ext4_error(sb, "Journal got removed while the fs was mounted!");
5629 return -EFSCORRUPTED;
5632 journal = EXT4_SB(sb)->s_journal;
5635 * Now check for any error status which may have been recorded in the
5636 * journal by a prior ext4_error() or ext4_abort()
5639 j_errno = jbd2_journal_errno(journal);
5643 errstr = ext4_decode_error(sb, j_errno, nbuf);
5644 ext4_warning(sb, "Filesystem error recorded "
5645 "from previous mount: %s", errstr);
5646 ext4_warning(sb, "Marking fs in need of filesystem check.");
5648 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5649 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5650 ext4_commit_super(sb, 1);
5652 jbd2_journal_clear_err(journal);
5653 jbd2_journal_update_sb_errno(journal);
5659 * Force the running and committing transactions to commit,
5660 * and wait on the commit.
5662 int ext4_force_commit(struct super_block *sb)
5669 journal = EXT4_SB(sb)->s_journal;
5670 return ext4_journal_force_commit(journal);
5673 static int ext4_sync_fs(struct super_block *sb, int wait)
5677 bool needs_barrier = false;
5678 struct ext4_sb_info *sbi = EXT4_SB(sb);
5680 if (unlikely(ext4_forced_shutdown(sbi)))
5683 trace_ext4_sync_fs(sb, wait);
5684 flush_workqueue(sbi->rsv_conversion_wq);
5686 * Writeback quota in non-journalled quota case - journalled quota has
5689 dquot_writeback_dquots(sb, -1);
5691 * Data writeback is possible w/o journal transaction, so barrier must
5692 * being sent at the end of the function. But we can skip it if
5693 * transaction_commit will do it for us.
5695 if (sbi->s_journal) {
5696 target = jbd2_get_latest_transaction(sbi->s_journal);
5697 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5698 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5699 needs_barrier = true;
5701 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5703 ret = jbd2_log_wait_commit(sbi->s_journal,
5706 } else if (wait && test_opt(sb, BARRIER))
5707 needs_barrier = true;
5708 if (needs_barrier) {
5710 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL);
5719 * LVM calls this function before a (read-only) snapshot is created. This
5720 * gives us a chance to flush the journal completely and mark the fs clean.
5722 * Note that only this function cannot bring a filesystem to be in a clean
5723 * state independently. It relies on upper layer to stop all data & metadata
5726 static int ext4_freeze(struct super_block *sb)
5734 journal = EXT4_SB(sb)->s_journal;
5737 /* Now we set up the journal barrier. */
5738 jbd2_journal_lock_updates(journal);
5741 * Don't clear the needs_recovery flag if we failed to
5742 * flush the journal.
5744 error = jbd2_journal_flush(journal);
5748 /* Journal blocked and flushed, clear needs_recovery flag. */
5749 ext4_clear_feature_journal_needs_recovery(sb);
5752 error = ext4_commit_super(sb, 1);
5755 /* we rely on upper layer to stop further updates */
5756 jbd2_journal_unlock_updates(journal);
5761 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5762 * flag here, even though the filesystem is not technically dirty yet.
5764 static int ext4_unfreeze(struct super_block *sb)
5766 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
5769 if (EXT4_SB(sb)->s_journal) {
5770 /* Reset the needs_recovery flag before the fs is unlocked. */
5771 ext4_set_feature_journal_needs_recovery(sb);
5774 ext4_commit_super(sb, 1);
5779 * Structure to save mount options for ext4_remount's benefit
5781 struct ext4_mount_options {
5782 unsigned long s_mount_opt;
5783 unsigned long s_mount_opt2;
5786 unsigned long s_commit_interval;
5787 u32 s_min_batch_time, s_max_batch_time;
5790 char *s_qf_names[EXT4_MAXQUOTAS];
5794 static int ext4_remount(struct super_block *sb, int *flags, char *data)
5796 struct ext4_super_block *es;
5797 struct ext4_sb_info *sbi = EXT4_SB(sb);
5798 unsigned long old_sb_flags, vfs_flags;
5799 struct ext4_mount_options old_opts;
5801 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5804 int enable_quota = 0;
5806 char *to_free[EXT4_MAXQUOTAS];
5808 char *orig_data = kstrdup(data, GFP_KERNEL);
5810 if (data && !orig_data)
5813 /* Store the original options */
5814 old_sb_flags = sb->s_flags;
5815 old_opts.s_mount_opt = sbi->s_mount_opt;
5816 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
5817 old_opts.s_resuid = sbi->s_resuid;
5818 old_opts.s_resgid = sbi->s_resgid;
5819 old_opts.s_commit_interval = sbi->s_commit_interval;
5820 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5821 old_opts.s_max_batch_time = sbi->s_max_batch_time;
5823 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
5824 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5825 if (sbi->s_qf_names[i]) {
5826 char *qf_name = get_qf_name(sb, sbi, i);
5828 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
5829 if (!old_opts.s_qf_names[i]) {
5830 for (j = 0; j < i; j++)
5831 kfree(old_opts.s_qf_names[j]);
5836 old_opts.s_qf_names[i] = NULL;
5838 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5839 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
5842 * Some options can be enabled by ext4 and/or by VFS mount flag
5843 * either way we need to make sure it matches in both *flags and
5844 * s_flags. Copy those selected flags from *flags to s_flags
5846 vfs_flags = SB_LAZYTIME | SB_I_VERSION;
5847 sb->s_flags = (sb->s_flags & ~vfs_flags) | (*flags & vfs_flags);
5849 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
5854 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5855 test_opt(sb, JOURNAL_CHECKSUM)) {
5856 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5857 "during remount not supported; ignoring");
5858 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5861 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5862 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5863 ext4_msg(sb, KERN_ERR, "can't mount with "
5864 "both data=journal and delalloc");
5868 if (test_opt(sb, DIOREAD_NOLOCK)) {
5869 ext4_msg(sb, KERN_ERR, "can't mount with "
5870 "both data=journal and dioread_nolock");
5874 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5875 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5876 ext4_msg(sb, KERN_ERR, "can't mount with "
5877 "journal_async_commit in data=ordered mode");
5883 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5884 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5889 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED))
5890 ext4_abort(sb, EXT4_ERR_ESHUTDOWN, "Abort forced by user");
5892 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5893 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5897 if (sbi->s_journal) {
5898 ext4_init_journal_params(sb, sbi->s_journal);
5899 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5902 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
5903 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) {
5908 if (*flags & SB_RDONLY) {
5909 err = sync_filesystem(sb);
5912 err = dquot_suspend(sb, -1);
5917 * First of all, the unconditional stuff we have to do
5918 * to disable replay of the journal when we next remount
5920 sb->s_flags |= SB_RDONLY;
5923 * OK, test if we are remounting a valid rw partition
5924 * readonly, and if so set the rdonly flag and then
5925 * mark the partition as valid again.
5927 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5928 (sbi->s_mount_state & EXT4_VALID_FS))
5929 es->s_state = cpu_to_le16(sbi->s_mount_state);
5931 if (sbi->s_journal) {
5933 * We let remount-ro finish even if marking fs
5934 * as clean failed...
5936 ext4_mark_recovery_complete(sb, es);
5939 /* Make sure we can mount this feature set readwrite */
5940 if (ext4_has_feature_readonly(sb) ||
5941 !ext4_feature_set_ok(sb, 0)) {
5946 * Make sure the group descriptor checksums
5947 * are sane. If they aren't, refuse to remount r/w.
5949 for (g = 0; g < sbi->s_groups_count; g++) {
5950 struct ext4_group_desc *gdp =
5951 ext4_get_group_desc(sb, g, NULL);
5953 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5954 ext4_msg(sb, KERN_ERR,
5955 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5956 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5957 le16_to_cpu(gdp->bg_checksum));
5964 * If we have an unprocessed orphan list hanging
5965 * around from a previously readonly bdev mount,
5966 * require a full umount/remount for now.
5968 if (es->s_last_orphan) {
5969 ext4_msg(sb, KERN_WARNING, "Couldn't "
5970 "remount RDWR because of unprocessed "
5971 "orphan inode list. Please "
5972 "umount/remount instead");
5978 * Mounting a RDONLY partition read-write, so reread
5979 * and store the current valid flag. (It may have
5980 * been changed by e2fsck since we originally mounted
5983 if (sbi->s_journal) {
5984 err = ext4_clear_journal_err(sb, es);
5988 sbi->s_mount_state = (le16_to_cpu(es->s_state) &
5991 err = ext4_setup_super(sb, es, 0);
5995 sb->s_flags &= ~SB_RDONLY;
5996 if (ext4_has_feature_mmp(sb)) {
5997 err = ext4_multi_mount_protect(sb,
5998 le64_to_cpu(es->s_mmp_block));
6009 * Handle creation of system zone data early because it can fail.
6010 * Releasing of existing data is done when we are sure remount will
6013 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
6014 err = ext4_setup_system_zone(sb);
6019 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
6020 err = ext4_commit_super(sb, 1);
6027 if (sb_any_quota_suspended(sb))
6028 dquot_resume(sb, -1);
6029 else if (ext4_has_feature_quota(sb)) {
6030 err = ext4_enable_quotas(sb);
6035 /* Release old quota file names */
6036 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6037 kfree(old_opts.s_qf_names[i]);
6039 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6040 ext4_release_system_zone(sb);
6043 * Reinitialize lazy itable initialization thread based on
6046 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
6047 ext4_unregister_li_request(sb);
6049 ext4_group_t first_not_zeroed;
6050 first_not_zeroed = ext4_has_uninit_itable(sb);
6051 ext4_register_li_request(sb, first_not_zeroed);
6054 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6055 ext4_stop_mmpd(sbi);
6058 * Some options can be enabled by ext4 and/or by VFS mount flag
6059 * either way we need to make sure it matches in both *flags and
6060 * s_flags. Copy those selected flags from s_flags to *flags
6062 *flags = (*flags & ~vfs_flags) | (sb->s_flags & vfs_flags);
6064 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
6070 * If there was a failing r/w to ro transition, we may need to
6073 if ((sb->s_flags & SB_RDONLY) && !(old_sb_flags & SB_RDONLY) &&
6074 sb_any_quota_suspended(sb))
6075 dquot_resume(sb, -1);
6076 sb->s_flags = old_sb_flags;
6077 sbi->s_mount_opt = old_opts.s_mount_opt;
6078 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
6079 sbi->s_resuid = old_opts.s_resuid;
6080 sbi->s_resgid = old_opts.s_resgid;
6081 sbi->s_commit_interval = old_opts.s_commit_interval;
6082 sbi->s_min_batch_time = old_opts.s_min_batch_time;
6083 sbi->s_max_batch_time = old_opts.s_max_batch_time;
6084 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6085 ext4_release_system_zone(sb);
6087 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
6088 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
6089 to_free[i] = get_qf_name(sb, sbi, i);
6090 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
6093 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6096 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6097 ext4_stop_mmpd(sbi);
6103 static int ext4_statfs_project(struct super_block *sb,
6104 kprojid_t projid, struct kstatfs *buf)
6107 struct dquot *dquot;
6111 qid = make_kqid_projid(projid);
6112 dquot = dqget(sb, qid);
6114 return PTR_ERR(dquot);
6115 spin_lock(&dquot->dq_dqb_lock);
6117 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6118 dquot->dq_dqb.dqb_bhardlimit);
6119 limit >>= sb->s_blocksize_bits;
6121 if (limit && buf->f_blocks > limit) {
6122 curblock = (dquot->dq_dqb.dqb_curspace +
6123 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
6124 buf->f_blocks = limit;
6125 buf->f_bfree = buf->f_bavail =
6126 (buf->f_blocks > curblock) ?
6127 (buf->f_blocks - curblock) : 0;
6130 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6131 dquot->dq_dqb.dqb_ihardlimit);
6132 if (limit && buf->f_files > limit) {
6133 buf->f_files = limit;
6135 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6136 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6139 spin_unlock(&dquot->dq_dqb_lock);
6145 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
6147 struct super_block *sb = dentry->d_sb;
6148 struct ext4_sb_info *sbi = EXT4_SB(sb);
6149 struct ext4_super_block *es = sbi->s_es;
6150 ext4_fsblk_t overhead = 0, resv_blocks;
6153 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
6155 if (!test_opt(sb, MINIX_DF))
6156 overhead = sbi->s_overhead;
6158 buf->f_type = EXT4_SUPER_MAGIC;
6159 buf->f_bsize = sb->s_blocksize;
6160 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
6161 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6162 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
6163 /* prevent underflow in case that few free space is available */
6164 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
6165 buf->f_bavail = buf->f_bfree -
6166 (ext4_r_blocks_count(es) + resv_blocks);
6167 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
6169 buf->f_files = le32_to_cpu(es->s_inodes_count);
6170 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
6171 buf->f_namelen = EXT4_NAME_LEN;
6172 fsid = le64_to_cpup((void *)es->s_uuid) ^
6173 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
6174 buf->f_fsid = u64_to_fsid(fsid);
6177 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6178 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6179 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6188 * Helper functions so that transaction is started before we acquire dqio_sem
6189 * to keep correct lock ordering of transaction > dqio_sem
6191 static inline struct inode *dquot_to_inode(struct dquot *dquot)
6193 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
6196 static int ext4_write_dquot(struct dquot *dquot)
6200 struct inode *inode;
6202 inode = dquot_to_inode(dquot);
6203 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
6204 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
6206 return PTR_ERR(handle);
6207 ret = dquot_commit(dquot);
6209 ext4_error_err(dquot->dq_sb, -ret,
6210 "Failed to commit dquot type %d",
6212 err = ext4_journal_stop(handle);
6218 static int ext4_acquire_dquot(struct dquot *dquot)
6223 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6224 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
6226 return PTR_ERR(handle);
6227 ret = dquot_acquire(dquot);
6229 ext4_error_err(dquot->dq_sb, -ret,
6230 "Failed to acquire dquot type %d",
6232 err = ext4_journal_stop(handle);
6238 static int ext4_release_dquot(struct dquot *dquot)
6243 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6244 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
6245 if (IS_ERR(handle)) {
6246 /* Release dquot anyway to avoid endless cycle in dqput() */
6247 dquot_release(dquot);
6248 return PTR_ERR(handle);
6250 ret = dquot_release(dquot);
6252 ext4_error_err(dquot->dq_sb, -ret,
6253 "Failed to release dquot type %d",
6255 err = ext4_journal_stop(handle);
6261 static int ext4_mark_dquot_dirty(struct dquot *dquot)
6263 struct super_block *sb = dquot->dq_sb;
6264 struct ext4_sb_info *sbi = EXT4_SB(sb);
6266 /* Are we journaling quotas? */
6267 if (ext4_has_feature_quota(sb) ||
6268 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
6269 dquot_mark_dquot_dirty(dquot);
6270 return ext4_write_dquot(dquot);
6272 return dquot_mark_dquot_dirty(dquot);
6276 static int ext4_write_info(struct super_block *sb, int type)
6281 /* Data block + inode block */
6282 handle = ext4_journal_start_sb(sb, EXT4_HT_QUOTA, 2);
6284 return PTR_ERR(handle);
6285 ret = dquot_commit_info(sb, type);
6286 err = ext4_journal_stop(handle);
6293 * Turn on quotas during mount time - we need to find
6294 * the quota file and such...
6296 static int ext4_quota_on_mount(struct super_block *sb, int type)
6298 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
6299 EXT4_SB(sb)->s_jquota_fmt, type);
6302 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6304 struct ext4_inode_info *ei = EXT4_I(inode);
6306 /* The first argument of lockdep_set_subclass has to be
6307 * *exactly* the same as the argument to init_rwsem() --- in
6308 * this case, in init_once() --- or lockdep gets unhappy
6309 * because the name of the lock is set using the
6310 * stringification of the argument to init_rwsem().
6312 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6313 lockdep_set_subclass(&ei->i_data_sem, subclass);
6317 * Standard function to be called on quota_on
6319 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6320 const struct path *path)
6324 if (!test_opt(sb, QUOTA))
6327 /* Quotafile not on the same filesystem? */
6328 if (path->dentry->d_sb != sb)
6331 /* Quota already enabled for this file? */
6332 if (IS_NOQUOTA(d_inode(path->dentry)))
6335 /* Journaling quota? */
6336 if (EXT4_SB(sb)->s_qf_names[type]) {
6337 /* Quotafile not in fs root? */
6338 if (path->dentry->d_parent != sb->s_root)
6339 ext4_msg(sb, KERN_WARNING,
6340 "Quota file not on filesystem root. "
6341 "Journaled quota will not work");
6342 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6345 * Clear the flag just in case mount options changed since
6348 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
6352 * When we journal data on quota file, we have to flush journal to see
6353 * all updates to the file when we bypass pagecache...
6355 if (EXT4_SB(sb)->s_journal &&
6356 ext4_should_journal_data(d_inode(path->dentry))) {
6358 * We don't need to lock updates but journal_flush() could
6359 * otherwise be livelocked...
6361 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
6362 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
6363 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
6368 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6369 err = dquot_quota_on(sb, type, format_id, path);
6371 struct inode *inode = d_inode(path->dentry);
6375 * Set inode flags to prevent userspace from messing with quota
6376 * files. If this fails, we return success anyway since quotas
6377 * are already enabled and this is not a hard failure.
6380 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6383 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6384 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6385 S_NOATIME | S_IMMUTABLE);
6386 err = ext4_mark_inode_dirty(handle, inode);
6387 ext4_journal_stop(handle);
6389 inode_unlock(inode);
6391 dquot_quota_off(sb, type);
6394 lockdep_set_quota_inode(path->dentry->d_inode,
6399 static inline bool ext4_check_quota_inum(int type, unsigned long qf_inum)
6403 return qf_inum == EXT4_USR_QUOTA_INO;
6405 return qf_inum == EXT4_GRP_QUOTA_INO;
6407 return qf_inum >= EXT4_GOOD_OLD_FIRST_INO;
6413 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6417 struct inode *qf_inode;
6418 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6419 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6420 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6421 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6424 BUG_ON(!ext4_has_feature_quota(sb));
6426 if (!qf_inums[type])
6429 if (!ext4_check_quota_inum(type, qf_inums[type])) {
6430 ext4_error(sb, "Bad quota inum: %lu, type: %d",
6431 qf_inums[type], type);
6435 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
6436 if (IS_ERR(qf_inode)) {
6437 ext4_error(sb, "Bad quota inode: %lu, type: %d",
6438 qf_inums[type], type);
6439 return PTR_ERR(qf_inode);
6442 /* Don't account quota for quota files to avoid recursion */
6443 qf_inode->i_flags |= S_NOQUOTA;
6444 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
6445 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
6447 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
6453 /* Enable usage tracking for all quota types. */
6454 static int ext4_enable_quotas(struct super_block *sb)
6457 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6458 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6459 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6460 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6462 bool quota_mopt[EXT4_MAXQUOTAS] = {
6463 test_opt(sb, USRQUOTA),
6464 test_opt(sb, GRPQUOTA),
6465 test_opt(sb, PRJQUOTA),
6468 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
6469 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
6470 if (qf_inums[type]) {
6471 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
6472 DQUOT_USAGE_ENABLED |
6473 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
6476 "Failed to enable quota tracking "
6477 "(type=%d, err=%d, ino=%lu). "
6478 "Please run e2fsck to fix.", type,
6479 err, qf_inums[type]);
6480 for (type--; type >= 0; type--) {
6481 struct inode *inode;
6483 inode = sb_dqopt(sb)->files[type];
6485 inode = igrab(inode);
6486 dquot_quota_off(sb, type);
6488 lockdep_set_quota_inode(inode,
6501 static int ext4_quota_off(struct super_block *sb, int type)
6503 struct inode *inode = sb_dqopt(sb)->files[type];
6507 /* Force all delayed allocation blocks to be allocated.
6508 * Caller already holds s_umount sem */
6509 if (test_opt(sb, DELALLOC))
6510 sync_filesystem(sb);
6512 if (!inode || !igrab(inode))
6515 err = dquot_quota_off(sb, type);
6516 if (err || ext4_has_feature_quota(sb))
6521 * Update modification times of quota files when userspace can
6522 * start looking at them. If we fail, we return success anyway since
6523 * this is not a hard failure and quotas are already disabled.
6525 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6526 if (IS_ERR(handle)) {
6527 err = PTR_ERR(handle);
6530 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
6531 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
6532 inode->i_mtime = inode->i_ctime = current_time(inode);
6533 err = ext4_mark_inode_dirty(handle, inode);
6534 ext4_journal_stop(handle);
6536 inode_unlock(inode);
6538 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
6542 return dquot_quota_off(sb, type);
6545 /* Read data from quotafile - avoid pagecache and such because we cannot afford
6546 * acquiring the locks... As quota files are never truncated and quota code
6547 * itself serializes the operations (and no one else should touch the files)
6548 * we don't have to be afraid of races */
6549 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
6550 size_t len, loff_t off)
6552 struct inode *inode = sb_dqopt(sb)->files[type];
6553 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6554 int offset = off & (sb->s_blocksize - 1);
6557 struct buffer_head *bh;
6558 loff_t i_size = i_size_read(inode);
6562 if (off+len > i_size)
6565 while (toread > 0) {
6566 tocopy = sb->s_blocksize - offset < toread ?
6567 sb->s_blocksize - offset : toread;
6568 bh = ext4_bread(NULL, inode, blk, 0);
6571 if (!bh) /* A hole? */
6572 memset(data, 0, tocopy);
6574 memcpy(data, bh->b_data+offset, tocopy);
6584 /* Write to quotafile (we know the transaction is already started and has
6585 * enough credits) */
6586 static ssize_t ext4_quota_write(struct super_block *sb, int type,
6587 const char *data, size_t len, loff_t off)
6589 struct inode *inode = sb_dqopt(sb)->files[type];
6590 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6591 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
6593 struct buffer_head *bh;
6594 handle_t *handle = journal_current_handle();
6597 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6598 " cancelled because transaction is not started",
6599 (unsigned long long)off, (unsigned long long)len);
6603 * Since we account only one data block in transaction credits,
6604 * then it is impossible to cross a block boundary.
6606 if (sb->s_blocksize - offset < len) {
6607 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6608 " cancelled because not block aligned",
6609 (unsigned long long)off, (unsigned long long)len);
6614 bh = ext4_bread(handle, inode, blk,
6615 EXT4_GET_BLOCKS_CREATE |
6616 EXT4_GET_BLOCKS_METADATA_NOFAIL);
6617 } while (PTR_ERR(bh) == -ENOSPC &&
6618 ext4_should_retry_alloc(inode->i_sb, &retries));
6623 BUFFER_TRACE(bh, "get write access");
6624 err = ext4_journal_get_write_access(handle, bh);
6630 memcpy(bh->b_data+offset, data, len);
6631 flush_dcache_page(bh->b_page);
6633 err = ext4_handle_dirty_metadata(handle, NULL, bh);
6636 if (inode->i_size < off + len) {
6637 i_size_write(inode, off + len);
6638 EXT4_I(inode)->i_disksize = inode->i_size;
6639 err2 = ext4_mark_inode_dirty(handle, inode);
6640 if (unlikely(err2 && !err))
6643 return err ? err : len;
6647 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6648 const char *dev_name, void *data)
6650 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
6653 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
6654 static inline void register_as_ext2(void)
6656 int err = register_filesystem(&ext2_fs_type);
6659 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6662 static inline void unregister_as_ext2(void)
6664 unregister_filesystem(&ext2_fs_type);
6667 static inline int ext2_feature_set_ok(struct super_block *sb)
6669 if (ext4_has_unknown_ext2_incompat_features(sb))
6673 if (ext4_has_unknown_ext2_ro_compat_features(sb))
6678 static inline void register_as_ext2(void) { }
6679 static inline void unregister_as_ext2(void) { }
6680 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
6683 static inline void register_as_ext3(void)
6685 int err = register_filesystem(&ext3_fs_type);
6688 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6691 static inline void unregister_as_ext3(void)
6693 unregister_filesystem(&ext3_fs_type);
6696 static inline int ext3_feature_set_ok(struct super_block *sb)
6698 if (ext4_has_unknown_ext3_incompat_features(sb))
6700 if (!ext4_has_feature_journal(sb))
6704 if (ext4_has_unknown_ext3_ro_compat_features(sb))
6709 static struct file_system_type ext4_fs_type = {
6710 .owner = THIS_MODULE,
6712 .mount = ext4_mount,
6713 .kill_sb = kill_block_super,
6714 .fs_flags = FS_REQUIRES_DEV,
6716 MODULE_ALIAS_FS("ext4");
6718 /* Shared across all ext4 file systems */
6719 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
6721 static int __init ext4_init_fs(void)
6725 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
6726 ext4_li_info = NULL;
6727 mutex_init(&ext4_li_mtx);
6729 /* Build-time check for flags consistency */
6730 ext4_check_flag_values();
6732 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
6733 init_waitqueue_head(&ext4__ioend_wq[i]);
6735 err = ext4_init_es();
6739 err = ext4_init_pending();
6743 err = ext4_init_post_read_processing();
6747 err = ext4_init_pageio();
6751 err = ext4_init_system_zone();
6755 err = ext4_init_sysfs();
6759 err = ext4_init_mballoc();
6762 err = init_inodecache();
6766 err = ext4_fc_init_dentry_cache();
6772 err = register_filesystem(&ext4_fs_type);
6778 unregister_as_ext2();
6779 unregister_as_ext3();
6780 ext4_fc_destroy_dentry_cache();
6782 destroy_inodecache();
6784 ext4_exit_mballoc();
6788 ext4_exit_system_zone();
6792 ext4_exit_post_read_processing();
6794 ext4_exit_pending();
6801 static void __exit ext4_exit_fs(void)
6803 ext4_destroy_lazyinit_thread();
6804 unregister_as_ext2();
6805 unregister_as_ext3();
6806 unregister_filesystem(&ext4_fs_type);
6807 ext4_fc_destroy_dentry_cache();
6808 destroy_inodecache();
6809 ext4_exit_mballoc();
6811 ext4_exit_system_zone();
6813 ext4_exit_post_read_processing();
6815 ext4_exit_pending();
6818 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6819 MODULE_DESCRIPTION("Fourth Extended Filesystem");
6820 MODULE_LICENSE("GPL");
6821 MODULE_SOFTDEP("pre: crc32c");
6822 module_init(ext4_init_fs)
6823 module_exit(ext4_exit_fs)