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>
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 void ext4_mark_recovery_complete(struct super_block *sb,
70 struct ext4_super_block *es);
71 static void 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_sem -> 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_sem
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_sem
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 * This works like sb_bread() except it uses ERR_PTR for error
146 * returns. Currently with sb_bread it's impossible to distinguish
147 * between ENOMEM and EIO situations (since both result in a NULL
151 ext4_sb_bread(struct super_block *sb, sector_t block, int op_flags)
153 struct buffer_head *bh = sb_getblk(sb, block);
156 return ERR_PTR(-ENOMEM);
157 if (buffer_uptodate(bh))
159 ll_rw_block(REQ_OP_READ, REQ_META | op_flags, 1, &bh);
161 if (buffer_uptodate(bh))
164 return ERR_PTR(-EIO);
167 static int ext4_verify_csum_type(struct super_block *sb,
168 struct ext4_super_block *es)
170 if (!ext4_has_feature_metadata_csum(sb))
173 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
176 static __le32 ext4_superblock_csum(struct super_block *sb,
177 struct ext4_super_block *es)
179 struct ext4_sb_info *sbi = EXT4_SB(sb);
180 int offset = offsetof(struct ext4_super_block, s_checksum);
183 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
185 return cpu_to_le32(csum);
188 static int ext4_superblock_csum_verify(struct super_block *sb,
189 struct ext4_super_block *es)
191 if (!ext4_has_metadata_csum(sb))
194 return es->s_checksum == ext4_superblock_csum(sb, es);
197 void ext4_superblock_csum_set(struct super_block *sb)
199 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
201 if (!ext4_has_metadata_csum(sb))
204 es->s_checksum = ext4_superblock_csum(sb, es);
207 void *ext4_kvmalloc(size_t size, gfp_t flags)
211 ret = kmalloc(size, flags | __GFP_NOWARN);
213 ret = __vmalloc(size, flags, PAGE_KERNEL);
217 void *ext4_kvzalloc(size_t size, gfp_t flags)
221 ret = kzalloc(size, flags | __GFP_NOWARN);
223 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
227 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
228 struct ext4_group_desc *bg)
230 return le32_to_cpu(bg->bg_block_bitmap_lo) |
231 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
232 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
235 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
236 struct ext4_group_desc *bg)
238 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
239 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
240 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
243 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
244 struct ext4_group_desc *bg)
246 return le32_to_cpu(bg->bg_inode_table_lo) |
247 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
248 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
251 __u32 ext4_free_group_clusters(struct super_block *sb,
252 struct ext4_group_desc *bg)
254 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
255 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
256 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
259 __u32 ext4_free_inodes_count(struct super_block *sb,
260 struct ext4_group_desc *bg)
262 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
263 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
264 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
267 __u32 ext4_used_dirs_count(struct super_block *sb,
268 struct ext4_group_desc *bg)
270 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
271 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
272 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
275 __u32 ext4_itable_unused_count(struct super_block *sb,
276 struct ext4_group_desc *bg)
278 return le16_to_cpu(bg->bg_itable_unused_lo) |
279 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
280 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
283 void ext4_block_bitmap_set(struct super_block *sb,
284 struct ext4_group_desc *bg, ext4_fsblk_t blk)
286 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
287 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
288 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
291 void ext4_inode_bitmap_set(struct super_block *sb,
292 struct ext4_group_desc *bg, ext4_fsblk_t blk)
294 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
295 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
296 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
299 void ext4_inode_table_set(struct super_block *sb,
300 struct ext4_group_desc *bg, ext4_fsblk_t blk)
302 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
303 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
304 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
307 void ext4_free_group_clusters_set(struct super_block *sb,
308 struct ext4_group_desc *bg, __u32 count)
310 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
311 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
312 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
315 void ext4_free_inodes_set(struct super_block *sb,
316 struct ext4_group_desc *bg, __u32 count)
318 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
319 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
320 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
323 void ext4_used_dirs_set(struct super_block *sb,
324 struct ext4_group_desc *bg, __u32 count)
326 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
327 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
328 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
331 void ext4_itable_unused_set(struct super_block *sb,
332 struct ext4_group_desc *bg, __u32 count)
334 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
335 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
336 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
339 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi)
341 time64_t now = ktime_get_real_seconds();
343 now = clamp_val(now, 0, (1ull << 40) - 1);
345 *lo = cpu_to_le32(lower_32_bits(now));
346 *hi = upper_32_bits(now);
349 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
351 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
353 #define ext4_update_tstamp(es, tstamp) \
354 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
355 #define ext4_get_tstamp(es, tstamp) \
356 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
358 static void __save_error_info(struct super_block *sb, const char *func,
361 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
363 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
364 if (bdev_read_only(sb->s_bdev))
366 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
367 ext4_update_tstamp(es, s_last_error_time);
368 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
369 es->s_last_error_line = cpu_to_le32(line);
370 if (!es->s_first_error_time) {
371 es->s_first_error_time = es->s_last_error_time;
372 es->s_first_error_time_hi = es->s_last_error_time_hi;
373 strncpy(es->s_first_error_func, func,
374 sizeof(es->s_first_error_func));
375 es->s_first_error_line = cpu_to_le32(line);
376 es->s_first_error_ino = es->s_last_error_ino;
377 es->s_first_error_block = es->s_last_error_block;
380 * Start the daily error reporting function if it hasn't been
383 if (!es->s_error_count)
384 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
385 le32_add_cpu(&es->s_error_count, 1);
388 static void save_error_info(struct super_block *sb, const char *func,
391 __save_error_info(sb, func, line);
392 ext4_commit_super(sb, 1);
396 * The del_gendisk() function uninitializes the disk-specific data
397 * structures, including the bdi structure, without telling anyone
398 * else. Once this happens, any attempt to call mark_buffer_dirty()
399 * (for example, by ext4_commit_super), will cause a kernel OOPS.
400 * This is a kludge to prevent these oops until we can put in a proper
401 * hook in del_gendisk() to inform the VFS and file system layers.
403 static int block_device_ejected(struct super_block *sb)
405 struct inode *bd_inode = sb->s_bdev->bd_inode;
406 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
408 return bdi->dev == NULL;
411 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
413 struct super_block *sb = journal->j_private;
414 struct ext4_sb_info *sbi = EXT4_SB(sb);
415 int error = is_journal_aborted(journal);
416 struct ext4_journal_cb_entry *jce;
418 BUG_ON(txn->t_state == T_FINISHED);
420 ext4_process_freed_data(sb, txn->t_tid);
422 spin_lock(&sbi->s_md_lock);
423 while (!list_empty(&txn->t_private_list)) {
424 jce = list_entry(txn->t_private_list.next,
425 struct ext4_journal_cb_entry, jce_list);
426 list_del_init(&jce->jce_list);
427 spin_unlock(&sbi->s_md_lock);
428 jce->jce_func(sb, jce, error);
429 spin_lock(&sbi->s_md_lock);
431 spin_unlock(&sbi->s_md_lock);
434 static bool system_going_down(void)
436 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
437 || system_state == SYSTEM_RESTART;
440 /* Deal with the reporting of failure conditions on a filesystem such as
441 * inconsistencies detected or read IO failures.
443 * On ext2, we can store the error state of the filesystem in the
444 * superblock. That is not possible on ext4, because we may have other
445 * write ordering constraints on the superblock which prevent us from
446 * writing it out straight away; and given that the journal is about to
447 * be aborted, we can't rely on the current, or future, transactions to
448 * write out the superblock safely.
450 * We'll just use the jbd2_journal_abort() error code to record an error in
451 * the journal instead. On recovery, the journal will complain about
452 * that error until we've noted it down and cleared it.
455 static void ext4_handle_error(struct super_block *sb)
457 if (test_opt(sb, WARN_ON_ERROR))
463 if (!test_opt(sb, ERRORS_CONT)) {
464 journal_t *journal = EXT4_SB(sb)->s_journal;
466 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
468 jbd2_journal_abort(journal, -EIO);
471 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
472 * could panic during 'reboot -f' as the underlying device got already
475 if (test_opt(sb, ERRORS_RO) || system_going_down()) {
476 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
478 * Make sure updated value of ->s_mount_flags will be visible
479 * before ->s_flags update
482 sb->s_flags |= SB_RDONLY;
483 } else if (test_opt(sb, ERRORS_PANIC)) {
484 if (EXT4_SB(sb)->s_journal &&
485 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
487 panic("EXT4-fs (device %s): panic forced after error\n",
492 #define ext4_error_ratelimit(sb) \
493 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
496 void __ext4_error(struct super_block *sb, const char *function,
497 unsigned int line, const char *fmt, ...)
499 struct va_format vaf;
502 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
505 trace_ext4_error(sb, function, line);
506 if (ext4_error_ratelimit(sb)) {
511 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
512 sb->s_id, function, line, current->comm, &vaf);
515 save_error_info(sb, function, line);
516 ext4_handle_error(sb);
519 void __ext4_error_inode(struct inode *inode, const char *function,
520 unsigned int line, ext4_fsblk_t block,
521 const char *fmt, ...)
524 struct va_format vaf;
525 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
527 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
530 trace_ext4_error(inode->i_sb, function, line);
531 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
532 es->s_last_error_block = cpu_to_le64(block);
533 if (ext4_error_ratelimit(inode->i_sb)) {
538 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
539 "inode #%lu: block %llu: comm %s: %pV\n",
540 inode->i_sb->s_id, function, line, inode->i_ino,
541 block, current->comm, &vaf);
543 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
544 "inode #%lu: comm %s: %pV\n",
545 inode->i_sb->s_id, function, line, inode->i_ino,
546 current->comm, &vaf);
549 save_error_info(inode->i_sb, function, line);
550 ext4_handle_error(inode->i_sb);
553 void __ext4_error_file(struct file *file, const char *function,
554 unsigned int line, ext4_fsblk_t block,
555 const char *fmt, ...)
558 struct va_format vaf;
559 struct ext4_super_block *es;
560 struct inode *inode = file_inode(file);
561 char pathname[80], *path;
563 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
566 trace_ext4_error(inode->i_sb, function, line);
567 es = EXT4_SB(inode->i_sb)->s_es;
568 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
569 if (ext4_error_ratelimit(inode->i_sb)) {
570 path = file_path(file, pathname, sizeof(pathname));
578 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
579 "block %llu: comm %s: path %s: %pV\n",
580 inode->i_sb->s_id, function, line, inode->i_ino,
581 block, current->comm, path, &vaf);
584 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
585 "comm %s: path %s: %pV\n",
586 inode->i_sb->s_id, function, line, inode->i_ino,
587 current->comm, path, &vaf);
590 save_error_info(inode->i_sb, function, line);
591 ext4_handle_error(inode->i_sb);
594 const char *ext4_decode_error(struct super_block *sb, int errno,
601 errstr = "Corrupt filesystem";
604 errstr = "Filesystem failed CRC";
607 errstr = "IO failure";
610 errstr = "Out of memory";
613 if (!sb || (EXT4_SB(sb)->s_journal &&
614 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
615 errstr = "Journal has aborted";
617 errstr = "Readonly filesystem";
620 /* If the caller passed in an extra buffer for unknown
621 * errors, textualise them now. Else we just return
624 /* Check for truncated error codes... */
625 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
634 /* __ext4_std_error decodes expected errors from journaling functions
635 * automatically and invokes the appropriate error response. */
637 void __ext4_std_error(struct super_block *sb, const char *function,
638 unsigned int line, int errno)
643 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
646 /* Special case: if the error is EROFS, and we're not already
647 * inside a transaction, then there's really no point in logging
649 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
652 if (ext4_error_ratelimit(sb)) {
653 errstr = ext4_decode_error(sb, errno, nbuf);
654 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
655 sb->s_id, function, line, errstr);
658 save_error_info(sb, function, line);
659 ext4_handle_error(sb);
663 * ext4_abort is a much stronger failure handler than ext4_error. The
664 * abort function may be used to deal with unrecoverable failures such
665 * as journal IO errors or ENOMEM at a critical moment in log management.
667 * We unconditionally force the filesystem into an ABORT|READONLY state,
668 * unless the error response on the fs has been set to panic in which
669 * case we take the easy way out and panic immediately.
672 void __ext4_abort(struct super_block *sb, const char *function,
673 unsigned int line, const char *fmt, ...)
675 struct va_format vaf;
678 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
681 save_error_info(sb, function, line);
685 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
686 sb->s_id, function, line, &vaf);
689 if (sb_rdonly(sb) == 0) {
690 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
691 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
693 * Make sure updated value of ->s_mount_flags will be visible
694 * before ->s_flags update
697 sb->s_flags |= SB_RDONLY;
698 if (EXT4_SB(sb)->s_journal)
699 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
700 save_error_info(sb, function, line);
702 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
703 if (EXT4_SB(sb)->s_journal &&
704 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
706 panic("EXT4-fs panic from previous error\n");
710 void __ext4_msg(struct super_block *sb,
711 const char *prefix, const char *fmt, ...)
713 struct va_format vaf;
716 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
722 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
726 #define ext4_warning_ratelimit(sb) \
727 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
730 void __ext4_warning(struct super_block *sb, const char *function,
731 unsigned int line, const char *fmt, ...)
733 struct va_format vaf;
736 if (!ext4_warning_ratelimit(sb))
742 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
743 sb->s_id, function, line, &vaf);
747 void __ext4_warning_inode(const struct inode *inode, const char *function,
748 unsigned int line, const char *fmt, ...)
750 struct va_format vaf;
753 if (!ext4_warning_ratelimit(inode->i_sb))
759 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
760 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
761 function, line, inode->i_ino, current->comm, &vaf);
765 void __ext4_grp_locked_error(const char *function, unsigned int line,
766 struct super_block *sb, ext4_group_t grp,
767 unsigned long ino, ext4_fsblk_t block,
768 const char *fmt, ...)
772 struct va_format vaf;
774 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
776 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
779 trace_ext4_error(sb, function, line);
780 es->s_last_error_ino = cpu_to_le32(ino);
781 es->s_last_error_block = cpu_to_le64(block);
782 __save_error_info(sb, function, line);
784 if (ext4_error_ratelimit(sb)) {
788 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
789 sb->s_id, function, line, grp);
791 printk(KERN_CONT "inode %lu: ", ino);
793 printk(KERN_CONT "block %llu:",
794 (unsigned long long) block);
795 printk(KERN_CONT "%pV\n", &vaf);
799 if (test_opt(sb, WARN_ON_ERROR))
802 if (test_opt(sb, ERRORS_CONT)) {
803 ext4_commit_super(sb, 0);
807 ext4_unlock_group(sb, grp);
808 ext4_commit_super(sb, 1);
809 ext4_handle_error(sb);
811 * We only get here in the ERRORS_RO case; relocking the group
812 * may be dangerous, but nothing bad will happen since the
813 * filesystem will have already been marked read/only and the
814 * journal has been aborted. We return 1 as a hint to callers
815 * who might what to use the return value from
816 * ext4_grp_locked_error() to distinguish between the
817 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
818 * aggressively from the ext4 function in question, with a
819 * more appropriate error code.
821 ext4_lock_group(sb, grp);
825 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
829 struct ext4_sb_info *sbi = EXT4_SB(sb);
830 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
831 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
834 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
835 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
838 percpu_counter_sub(&sbi->s_freeclusters_counter,
842 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
843 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
848 count = ext4_free_inodes_count(sb, gdp);
849 percpu_counter_sub(&sbi->s_freeinodes_counter,
855 void ext4_update_dynamic_rev(struct super_block *sb)
857 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
859 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
863 "updating to rev %d because of new feature flag, "
864 "running e2fsck is recommended",
867 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
868 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
869 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
870 /* leave es->s_feature_*compat flags alone */
871 /* es->s_uuid will be set by e2fsck if empty */
874 * The rest of the superblock fields should be zero, and if not it
875 * means they are likely already in use, so leave them alone. We
876 * can leave it up to e2fsck to clean up any inconsistencies there.
881 * Open the external journal device
883 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
885 struct block_device *bdev;
886 char b[BDEVNAME_SIZE];
888 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
894 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
895 __bdevname(dev, b), PTR_ERR(bdev));
900 * Release the journal device
902 static void ext4_blkdev_put(struct block_device *bdev)
904 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
907 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
909 struct block_device *bdev;
910 bdev = sbi->journal_bdev;
912 ext4_blkdev_put(bdev);
913 sbi->journal_bdev = NULL;
917 static inline struct inode *orphan_list_entry(struct list_head *l)
919 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
922 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
926 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
927 le32_to_cpu(sbi->s_es->s_last_orphan));
929 printk(KERN_ERR "sb_info orphan list:\n");
930 list_for_each(l, &sbi->s_orphan) {
931 struct inode *inode = orphan_list_entry(l);
933 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
934 inode->i_sb->s_id, inode->i_ino, inode,
935 inode->i_mode, inode->i_nlink,
941 static int ext4_quota_off(struct super_block *sb, int type);
943 static inline void ext4_quota_off_umount(struct super_block *sb)
947 /* Use our quota_off function to clear inode flags etc. */
948 for (type = 0; type < EXT4_MAXQUOTAS; type++)
949 ext4_quota_off(sb, type);
953 * This is a helper function which is used in the mount/remount
954 * codepaths (which holds s_umount) to fetch the quota file name.
956 static inline char *get_qf_name(struct super_block *sb,
957 struct ext4_sb_info *sbi,
960 return rcu_dereference_protected(sbi->s_qf_names[type],
961 lockdep_is_held(&sb->s_umount));
964 static inline void ext4_quota_off_umount(struct super_block *sb)
969 static void ext4_put_super(struct super_block *sb)
971 struct ext4_sb_info *sbi = EXT4_SB(sb);
972 struct ext4_super_block *es = sbi->s_es;
976 ext4_unregister_li_request(sb);
977 ext4_quota_off_umount(sb);
979 destroy_workqueue(sbi->rsv_conversion_wq);
981 if (sbi->s_journal) {
982 aborted = is_journal_aborted(sbi->s_journal);
983 err = jbd2_journal_destroy(sbi->s_journal);
984 sbi->s_journal = NULL;
985 if ((err < 0) && !aborted)
986 ext4_abort(sb, "Couldn't clean up the journal");
989 ext4_unregister_sysfs(sb);
990 ext4_es_unregister_shrinker(sbi);
991 del_timer_sync(&sbi->s_err_report);
992 ext4_release_system_zone(sb);
994 ext4_ext_release(sb);
996 if (!sb_rdonly(sb) && !aborted) {
997 ext4_clear_feature_journal_needs_recovery(sb);
998 es->s_state = cpu_to_le16(sbi->s_mount_state);
1001 ext4_commit_super(sb, 1);
1003 for (i = 0; i < sbi->s_gdb_count; i++)
1004 brelse(sbi->s_group_desc[i]);
1005 kvfree(sbi->s_group_desc);
1006 kvfree(sbi->s_flex_groups);
1007 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1008 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1009 percpu_counter_destroy(&sbi->s_dirs_counter);
1010 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1011 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
1013 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1014 kfree(get_qf_name(sb, sbi, i));
1017 /* Debugging code just in case the in-memory inode orphan list
1018 * isn't empty. The on-disk one can be non-empty if we've
1019 * detected an error and taken the fs readonly, but the
1020 * in-memory list had better be clean by this point. */
1021 if (!list_empty(&sbi->s_orphan))
1022 dump_orphan_list(sb, sbi);
1023 J_ASSERT(list_empty(&sbi->s_orphan));
1025 sync_blockdev(sb->s_bdev);
1026 invalidate_bdev(sb->s_bdev);
1027 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
1029 * Invalidate the journal device's buffers. We don't want them
1030 * floating about in memory - the physical journal device may
1031 * hotswapped, and it breaks the `ro-after' testing code.
1033 sync_blockdev(sbi->journal_bdev);
1034 invalidate_bdev(sbi->journal_bdev);
1035 ext4_blkdev_remove(sbi);
1038 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1039 sbi->s_ea_inode_cache = NULL;
1041 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1042 sbi->s_ea_block_cache = NULL;
1045 kthread_stop(sbi->s_mmp_tsk);
1047 sb->s_fs_info = NULL;
1049 * Now that we are completely done shutting down the
1050 * superblock, we need to actually destroy the kobject.
1052 kobject_put(&sbi->s_kobj);
1053 wait_for_completion(&sbi->s_kobj_unregister);
1054 if (sbi->s_chksum_driver)
1055 crypto_free_shash(sbi->s_chksum_driver);
1056 kfree(sbi->s_blockgroup_lock);
1057 fs_put_dax(sbi->s_daxdev);
1058 #ifdef CONFIG_UNICODE
1059 utf8_unload(sbi->s_encoding);
1064 static struct kmem_cache *ext4_inode_cachep;
1067 * Called inside transaction, so use GFP_NOFS
1069 static struct inode *ext4_alloc_inode(struct super_block *sb)
1071 struct ext4_inode_info *ei;
1073 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1077 inode_set_iversion(&ei->vfs_inode, 1);
1078 spin_lock_init(&ei->i_raw_lock);
1079 INIT_LIST_HEAD(&ei->i_prealloc_list);
1080 spin_lock_init(&ei->i_prealloc_lock);
1081 ext4_es_init_tree(&ei->i_es_tree);
1082 rwlock_init(&ei->i_es_lock);
1083 INIT_LIST_HEAD(&ei->i_es_list);
1084 ei->i_es_all_nr = 0;
1085 ei->i_es_shk_nr = 0;
1086 ei->i_es_shrink_lblk = 0;
1087 ei->i_reserved_data_blocks = 0;
1088 ei->i_da_metadata_calc_len = 0;
1089 ei->i_da_metadata_calc_last_lblock = 0;
1090 spin_lock_init(&(ei->i_block_reservation_lock));
1091 ext4_init_pending_tree(&ei->i_pending_tree);
1093 ei->i_reserved_quota = 0;
1094 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1097 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1098 spin_lock_init(&ei->i_completed_io_lock);
1100 ei->i_datasync_tid = 0;
1101 atomic_set(&ei->i_unwritten, 0);
1102 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1103 return &ei->vfs_inode;
1106 static int ext4_drop_inode(struct inode *inode)
1108 int drop = generic_drop_inode(inode);
1111 drop = fscrypt_drop_inode(inode);
1113 trace_ext4_drop_inode(inode, drop);
1117 static void ext4_free_in_core_inode(struct inode *inode)
1119 fscrypt_free_inode(inode);
1120 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1123 static void ext4_destroy_inode(struct inode *inode)
1125 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1126 ext4_msg(inode->i_sb, KERN_ERR,
1127 "Inode %lu (%p): orphan list check failed!",
1128 inode->i_ino, EXT4_I(inode));
1129 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1130 EXT4_I(inode), sizeof(struct ext4_inode_info),
1136 static void init_once(void *foo)
1138 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1140 INIT_LIST_HEAD(&ei->i_orphan);
1141 init_rwsem(&ei->xattr_sem);
1142 init_rwsem(&ei->i_data_sem);
1143 init_rwsem(&ei->i_mmap_sem);
1144 inode_init_once(&ei->vfs_inode);
1147 static int __init init_inodecache(void)
1149 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1150 sizeof(struct ext4_inode_info), 0,
1151 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1153 offsetof(struct ext4_inode_info, i_data),
1154 sizeof_field(struct ext4_inode_info, i_data),
1156 if (ext4_inode_cachep == NULL)
1161 static void destroy_inodecache(void)
1164 * Make sure all delayed rcu free inodes are flushed before we
1168 kmem_cache_destroy(ext4_inode_cachep);
1171 void ext4_clear_inode(struct inode *inode)
1173 invalidate_inode_buffers(inode);
1176 ext4_discard_preallocations(inode);
1177 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1178 if (EXT4_I(inode)->jinode) {
1179 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1180 EXT4_I(inode)->jinode);
1181 jbd2_free_inode(EXT4_I(inode)->jinode);
1182 EXT4_I(inode)->jinode = NULL;
1184 fscrypt_put_encryption_info(inode);
1187 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1188 u64 ino, u32 generation)
1190 struct inode *inode;
1193 * Currently we don't know the generation for parent directory, so
1194 * a generation of 0 means "accept any"
1196 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1198 return ERR_CAST(inode);
1199 if (generation && inode->i_generation != generation) {
1201 return ERR_PTR(-ESTALE);
1207 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1208 int fh_len, int fh_type)
1210 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1211 ext4_nfs_get_inode);
1214 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1215 int fh_len, int fh_type)
1217 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1218 ext4_nfs_get_inode);
1221 static int ext4_nfs_commit_metadata(struct inode *inode)
1223 struct writeback_control wbc = {
1224 .sync_mode = WB_SYNC_ALL
1227 trace_ext4_nfs_commit_metadata(inode);
1228 return ext4_write_inode(inode, &wbc);
1232 * Try to release metadata pages (indirect blocks, directories) which are
1233 * mapped via the block device. Since these pages could have journal heads
1234 * which would prevent try_to_free_buffers() from freeing them, we must use
1235 * jbd2 layer's try_to_free_buffers() function to release them.
1237 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1240 journal_t *journal = EXT4_SB(sb)->s_journal;
1242 WARN_ON(PageChecked(page));
1243 if (!page_has_buffers(page))
1246 return jbd2_journal_try_to_free_buffers(journal, page,
1247 wait & ~__GFP_DIRECT_RECLAIM);
1248 return try_to_free_buffers(page);
1251 #ifdef CONFIG_FS_ENCRYPTION
1252 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1254 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1255 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1258 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1261 handle_t *handle = fs_data;
1262 int res, res2, credits, retries = 0;
1265 * Encrypting the root directory is not allowed because e2fsck expects
1266 * lost+found to exist and be unencrypted, and encrypting the root
1267 * directory would imply encrypting the lost+found directory as well as
1268 * the filename "lost+found" itself.
1270 if (inode->i_ino == EXT4_ROOT_INO)
1273 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1276 res = ext4_convert_inline_data(inode);
1281 * If a journal handle was specified, then the encryption context is
1282 * being set on a new inode via inheritance and is part of a larger
1283 * transaction to create the inode. Otherwise the encryption context is
1284 * being set on an existing inode in its own transaction. Only in the
1285 * latter case should the "retry on ENOSPC" logic be used.
1289 res = ext4_xattr_set_handle(handle, inode,
1290 EXT4_XATTR_INDEX_ENCRYPTION,
1291 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1294 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1295 ext4_clear_inode_state(inode,
1296 EXT4_STATE_MAY_INLINE_DATA);
1298 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1299 * S_DAX may be disabled
1301 ext4_set_inode_flags(inode);
1306 res = dquot_initialize(inode);
1310 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1315 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1317 return PTR_ERR(handle);
1319 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1320 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1323 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1325 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1326 * S_DAX may be disabled
1328 ext4_set_inode_flags(inode);
1329 res = ext4_mark_inode_dirty(handle, inode);
1331 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1333 res2 = ext4_journal_stop(handle);
1335 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1342 static bool ext4_dummy_context(struct inode *inode)
1344 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1347 static const struct fscrypt_operations ext4_cryptops = {
1348 .key_prefix = "ext4:",
1349 .get_context = ext4_get_context,
1350 .set_context = ext4_set_context,
1351 .dummy_context = ext4_dummy_context,
1352 .empty_dir = ext4_empty_dir,
1353 .max_namelen = EXT4_NAME_LEN,
1358 static const char * const quotatypes[] = INITQFNAMES;
1359 #define QTYPE2NAME(t) (quotatypes[t])
1361 static int ext4_write_dquot(struct dquot *dquot);
1362 static int ext4_acquire_dquot(struct dquot *dquot);
1363 static int ext4_release_dquot(struct dquot *dquot);
1364 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1365 static int ext4_write_info(struct super_block *sb, int type);
1366 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1367 const struct path *path);
1368 static int ext4_quota_on_mount(struct super_block *sb, int type);
1369 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1370 size_t len, loff_t off);
1371 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1372 const char *data, size_t len, loff_t off);
1373 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1374 unsigned int flags);
1375 static int ext4_enable_quotas(struct super_block *sb);
1376 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1378 static struct dquot **ext4_get_dquots(struct inode *inode)
1380 return EXT4_I(inode)->i_dquot;
1383 static const struct dquot_operations ext4_quota_operations = {
1384 .get_reserved_space = ext4_get_reserved_space,
1385 .write_dquot = ext4_write_dquot,
1386 .acquire_dquot = ext4_acquire_dquot,
1387 .release_dquot = ext4_release_dquot,
1388 .mark_dirty = ext4_mark_dquot_dirty,
1389 .write_info = ext4_write_info,
1390 .alloc_dquot = dquot_alloc,
1391 .destroy_dquot = dquot_destroy,
1392 .get_projid = ext4_get_projid,
1393 .get_inode_usage = ext4_get_inode_usage,
1394 .get_next_id = ext4_get_next_id,
1397 static const struct quotactl_ops ext4_qctl_operations = {
1398 .quota_on = ext4_quota_on,
1399 .quota_off = ext4_quota_off,
1400 .quota_sync = dquot_quota_sync,
1401 .get_state = dquot_get_state,
1402 .set_info = dquot_set_dqinfo,
1403 .get_dqblk = dquot_get_dqblk,
1404 .set_dqblk = dquot_set_dqblk,
1405 .get_nextdqblk = dquot_get_next_dqblk,
1409 static const struct super_operations ext4_sops = {
1410 .alloc_inode = ext4_alloc_inode,
1411 .free_inode = ext4_free_in_core_inode,
1412 .destroy_inode = ext4_destroy_inode,
1413 .write_inode = ext4_write_inode,
1414 .dirty_inode = ext4_dirty_inode,
1415 .drop_inode = ext4_drop_inode,
1416 .evict_inode = ext4_evict_inode,
1417 .put_super = ext4_put_super,
1418 .sync_fs = ext4_sync_fs,
1419 .freeze_fs = ext4_freeze,
1420 .unfreeze_fs = ext4_unfreeze,
1421 .statfs = ext4_statfs,
1422 .remount_fs = ext4_remount,
1423 .show_options = ext4_show_options,
1425 .quota_read = ext4_quota_read,
1426 .quota_write = ext4_quota_write,
1427 .get_dquots = ext4_get_dquots,
1429 .bdev_try_to_free_page = bdev_try_to_free_page,
1432 static const struct export_operations ext4_export_ops = {
1433 .fh_to_dentry = ext4_fh_to_dentry,
1434 .fh_to_parent = ext4_fh_to_parent,
1435 .get_parent = ext4_get_parent,
1436 .commit_metadata = ext4_nfs_commit_metadata,
1440 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1441 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1442 Opt_nouid32, Opt_debug, Opt_removed,
1443 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1444 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1445 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1446 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1447 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1448 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1449 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1450 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1451 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1452 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1453 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1454 Opt_nowarn_on_error, Opt_mblk_io_submit,
1455 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1456 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1457 Opt_inode_readahead_blks, Opt_journal_ioprio,
1458 Opt_dioread_nolock, Opt_dioread_lock,
1459 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1460 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1463 static const match_table_t tokens = {
1464 {Opt_bsd_df, "bsddf"},
1465 {Opt_minix_df, "minixdf"},
1466 {Opt_grpid, "grpid"},
1467 {Opt_grpid, "bsdgroups"},
1468 {Opt_nogrpid, "nogrpid"},
1469 {Opt_nogrpid, "sysvgroups"},
1470 {Opt_resgid, "resgid=%u"},
1471 {Opt_resuid, "resuid=%u"},
1473 {Opt_err_cont, "errors=continue"},
1474 {Opt_err_panic, "errors=panic"},
1475 {Opt_err_ro, "errors=remount-ro"},
1476 {Opt_nouid32, "nouid32"},
1477 {Opt_debug, "debug"},
1478 {Opt_removed, "oldalloc"},
1479 {Opt_removed, "orlov"},
1480 {Opt_user_xattr, "user_xattr"},
1481 {Opt_nouser_xattr, "nouser_xattr"},
1483 {Opt_noacl, "noacl"},
1484 {Opt_noload, "norecovery"},
1485 {Opt_noload, "noload"},
1486 {Opt_removed, "nobh"},
1487 {Opt_removed, "bh"},
1488 {Opt_commit, "commit=%u"},
1489 {Opt_min_batch_time, "min_batch_time=%u"},
1490 {Opt_max_batch_time, "max_batch_time=%u"},
1491 {Opt_journal_dev, "journal_dev=%u"},
1492 {Opt_journal_path, "journal_path=%s"},
1493 {Opt_journal_checksum, "journal_checksum"},
1494 {Opt_nojournal_checksum, "nojournal_checksum"},
1495 {Opt_journal_async_commit, "journal_async_commit"},
1496 {Opt_abort, "abort"},
1497 {Opt_data_journal, "data=journal"},
1498 {Opt_data_ordered, "data=ordered"},
1499 {Opt_data_writeback, "data=writeback"},
1500 {Opt_data_err_abort, "data_err=abort"},
1501 {Opt_data_err_ignore, "data_err=ignore"},
1502 {Opt_offusrjquota, "usrjquota="},
1503 {Opt_usrjquota, "usrjquota=%s"},
1504 {Opt_offgrpjquota, "grpjquota="},
1505 {Opt_grpjquota, "grpjquota=%s"},
1506 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1507 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1508 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1509 {Opt_grpquota, "grpquota"},
1510 {Opt_noquota, "noquota"},
1511 {Opt_quota, "quota"},
1512 {Opt_usrquota, "usrquota"},
1513 {Opt_prjquota, "prjquota"},
1514 {Opt_barrier, "barrier=%u"},
1515 {Opt_barrier, "barrier"},
1516 {Opt_nobarrier, "nobarrier"},
1517 {Opt_i_version, "i_version"},
1519 {Opt_stripe, "stripe=%u"},
1520 {Opt_delalloc, "delalloc"},
1521 {Opt_warn_on_error, "warn_on_error"},
1522 {Opt_nowarn_on_error, "nowarn_on_error"},
1523 {Opt_lazytime, "lazytime"},
1524 {Opt_nolazytime, "nolazytime"},
1525 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1526 {Opt_nodelalloc, "nodelalloc"},
1527 {Opt_removed, "mblk_io_submit"},
1528 {Opt_removed, "nomblk_io_submit"},
1529 {Opt_block_validity, "block_validity"},
1530 {Opt_noblock_validity, "noblock_validity"},
1531 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1532 {Opt_journal_ioprio, "journal_ioprio=%u"},
1533 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1534 {Opt_auto_da_alloc, "auto_da_alloc"},
1535 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1536 {Opt_dioread_nolock, "dioread_nolock"},
1537 {Opt_dioread_lock, "dioread_lock"},
1538 {Opt_discard, "discard"},
1539 {Opt_nodiscard, "nodiscard"},
1540 {Opt_init_itable, "init_itable=%u"},
1541 {Opt_init_itable, "init_itable"},
1542 {Opt_noinit_itable, "noinit_itable"},
1543 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1544 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1545 {Opt_nombcache, "nombcache"},
1546 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1547 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1548 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1549 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1550 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1551 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1555 static ext4_fsblk_t get_sb_block(void **data)
1557 ext4_fsblk_t sb_block;
1558 char *options = (char *) *data;
1560 if (!options || strncmp(options, "sb=", 3) != 0)
1561 return 1; /* Default location */
1564 /* TODO: use simple_strtoll with >32bit ext4 */
1565 sb_block = simple_strtoul(options, &options, 0);
1566 if (*options && *options != ',') {
1567 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1571 if (*options == ',')
1573 *data = (void *) options;
1578 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1579 static const char deprecated_msg[] =
1580 "Mount option \"%s\" will be removed by %s\n"
1581 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1584 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1586 struct ext4_sb_info *sbi = EXT4_SB(sb);
1587 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
1590 if (sb_any_quota_loaded(sb) && !old_qname) {
1591 ext4_msg(sb, KERN_ERR,
1592 "Cannot change journaled "
1593 "quota options when quota turned on");
1596 if (ext4_has_feature_quota(sb)) {
1597 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1598 "ignored when QUOTA feature is enabled");
1601 qname = match_strdup(args);
1603 ext4_msg(sb, KERN_ERR,
1604 "Not enough memory for storing quotafile name");
1608 if (strcmp(old_qname, qname) == 0)
1611 ext4_msg(sb, KERN_ERR,
1612 "%s quota file already specified",
1616 if (strchr(qname, '/')) {
1617 ext4_msg(sb, KERN_ERR,
1618 "quotafile must be on filesystem root");
1621 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
1629 static int clear_qf_name(struct super_block *sb, int qtype)
1632 struct ext4_sb_info *sbi = EXT4_SB(sb);
1633 char *old_qname = get_qf_name(sb, sbi, qtype);
1635 if (sb_any_quota_loaded(sb) && old_qname) {
1636 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1637 " when quota turned on");
1640 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1647 #define MOPT_SET 0x0001
1648 #define MOPT_CLEAR 0x0002
1649 #define MOPT_NOSUPPORT 0x0004
1650 #define MOPT_EXPLICIT 0x0008
1651 #define MOPT_CLEAR_ERR 0x0010
1652 #define MOPT_GTE0 0x0020
1655 #define MOPT_QFMT 0x0040
1657 #define MOPT_Q MOPT_NOSUPPORT
1658 #define MOPT_QFMT MOPT_NOSUPPORT
1660 #define MOPT_DATAJ 0x0080
1661 #define MOPT_NO_EXT2 0x0100
1662 #define MOPT_NO_EXT3 0x0200
1663 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1664 #define MOPT_STRING 0x0400
1666 static const struct mount_opts {
1670 } ext4_mount_opts[] = {
1671 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1672 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1673 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1674 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1675 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1676 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1677 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1678 MOPT_EXT4_ONLY | MOPT_SET},
1679 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1680 MOPT_EXT4_ONLY | MOPT_CLEAR},
1681 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1682 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1683 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1684 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1685 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1686 MOPT_EXT4_ONLY | MOPT_CLEAR},
1687 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1688 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1689 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1690 MOPT_EXT4_ONLY | MOPT_CLEAR},
1691 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1692 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1693 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1694 EXT4_MOUNT_JOURNAL_CHECKSUM),
1695 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1696 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1697 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1698 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1699 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1700 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1702 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1704 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1705 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1706 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1707 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1708 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1709 {Opt_commit, 0, MOPT_GTE0},
1710 {Opt_max_batch_time, 0, MOPT_GTE0},
1711 {Opt_min_batch_time, 0, MOPT_GTE0},
1712 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1713 {Opt_init_itable, 0, MOPT_GTE0},
1714 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1715 {Opt_stripe, 0, MOPT_GTE0},
1716 {Opt_resuid, 0, MOPT_GTE0},
1717 {Opt_resgid, 0, MOPT_GTE0},
1718 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1719 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1720 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1721 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1722 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1723 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1724 MOPT_NO_EXT2 | MOPT_DATAJ},
1725 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1726 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1727 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1728 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1729 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1731 {Opt_acl, 0, MOPT_NOSUPPORT},
1732 {Opt_noacl, 0, MOPT_NOSUPPORT},
1734 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1735 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1736 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1737 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1738 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1740 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1742 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1744 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1745 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1746 MOPT_CLEAR | MOPT_Q},
1747 {Opt_usrjquota, 0, MOPT_Q},
1748 {Opt_grpjquota, 0, MOPT_Q},
1749 {Opt_offusrjquota, 0, MOPT_Q},
1750 {Opt_offgrpjquota, 0, MOPT_Q},
1751 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1752 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1753 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1754 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1755 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1756 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1760 #ifdef CONFIG_UNICODE
1761 static const struct ext4_sb_encodings {
1765 } ext4_sb_encoding_map[] = {
1766 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
1769 static int ext4_sb_read_encoding(const struct ext4_super_block *es,
1770 const struct ext4_sb_encodings **encoding,
1773 __u16 magic = le16_to_cpu(es->s_encoding);
1776 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1777 if (magic == ext4_sb_encoding_map[i].magic)
1780 if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
1783 *encoding = &ext4_sb_encoding_map[i];
1784 *flags = le16_to_cpu(es->s_encoding_flags);
1790 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1791 substring_t *args, unsigned long *journal_devnum,
1792 unsigned int *journal_ioprio, int is_remount)
1794 struct ext4_sb_info *sbi = EXT4_SB(sb);
1795 const struct mount_opts *m;
1801 if (token == Opt_usrjquota)
1802 return set_qf_name(sb, USRQUOTA, &args[0]);
1803 else if (token == Opt_grpjquota)
1804 return set_qf_name(sb, GRPQUOTA, &args[0]);
1805 else if (token == Opt_offusrjquota)
1806 return clear_qf_name(sb, USRQUOTA);
1807 else if (token == Opt_offgrpjquota)
1808 return clear_qf_name(sb, GRPQUOTA);
1812 case Opt_nouser_xattr:
1813 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1816 return 1; /* handled by get_sb_block() */
1818 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1821 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1824 sb->s_flags |= SB_I_VERSION;
1827 sb->s_flags |= SB_LAZYTIME;
1829 case Opt_nolazytime:
1830 sb->s_flags &= ~SB_LAZYTIME;
1834 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1835 if (token == m->token)
1838 if (m->token == Opt_err) {
1839 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1840 "or missing value", opt);
1844 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1845 ext4_msg(sb, KERN_ERR,
1846 "Mount option \"%s\" incompatible with ext2", opt);
1849 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1850 ext4_msg(sb, KERN_ERR,
1851 "Mount option \"%s\" incompatible with ext3", opt);
1855 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1857 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1859 if (m->flags & MOPT_EXPLICIT) {
1860 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1861 set_opt2(sb, EXPLICIT_DELALLOC);
1862 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1863 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1867 if (m->flags & MOPT_CLEAR_ERR)
1868 clear_opt(sb, ERRORS_MASK);
1869 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1870 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1871 "options when quota turned on");
1875 if (m->flags & MOPT_NOSUPPORT) {
1876 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1877 } else if (token == Opt_commit) {
1879 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1880 sbi->s_commit_interval = HZ * arg;
1881 } else if (token == Opt_debug_want_extra_isize) {
1882 sbi->s_want_extra_isize = arg;
1883 } else if (token == Opt_max_batch_time) {
1884 sbi->s_max_batch_time = arg;
1885 } else if (token == Opt_min_batch_time) {
1886 sbi->s_min_batch_time = arg;
1887 } else if (token == Opt_inode_readahead_blks) {
1888 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1889 ext4_msg(sb, KERN_ERR,
1890 "EXT4-fs: inode_readahead_blks must be "
1891 "0 or a power of 2 smaller than 2^31");
1894 sbi->s_inode_readahead_blks = arg;
1895 } else if (token == Opt_init_itable) {
1896 set_opt(sb, INIT_INODE_TABLE);
1898 arg = EXT4_DEF_LI_WAIT_MULT;
1899 sbi->s_li_wait_mult = arg;
1900 } else if (token == Opt_max_dir_size_kb) {
1901 sbi->s_max_dir_size_kb = arg;
1902 } else if (token == Opt_stripe) {
1903 sbi->s_stripe = arg;
1904 } else if (token == Opt_resuid) {
1905 uid = make_kuid(current_user_ns(), arg);
1906 if (!uid_valid(uid)) {
1907 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1910 sbi->s_resuid = uid;
1911 } else if (token == Opt_resgid) {
1912 gid = make_kgid(current_user_ns(), arg);
1913 if (!gid_valid(gid)) {
1914 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1917 sbi->s_resgid = gid;
1918 } else if (token == Opt_journal_dev) {
1920 ext4_msg(sb, KERN_ERR,
1921 "Cannot specify journal on remount");
1924 *journal_devnum = arg;
1925 } else if (token == Opt_journal_path) {
1927 struct inode *journal_inode;
1932 ext4_msg(sb, KERN_ERR,
1933 "Cannot specify journal on remount");
1936 journal_path = match_strdup(&args[0]);
1937 if (!journal_path) {
1938 ext4_msg(sb, KERN_ERR, "error: could not dup "
1939 "journal device string");
1943 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1945 ext4_msg(sb, KERN_ERR, "error: could not find "
1946 "journal device path: error %d", error);
1947 kfree(journal_path);
1951 journal_inode = d_inode(path.dentry);
1952 if (!S_ISBLK(journal_inode->i_mode)) {
1953 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1954 "is not a block device", journal_path);
1956 kfree(journal_path);
1960 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1962 kfree(journal_path);
1963 } else if (token == Opt_journal_ioprio) {
1965 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1970 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1971 } else if (token == Opt_test_dummy_encryption) {
1972 #ifdef CONFIG_FS_ENCRYPTION
1973 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1974 ext4_msg(sb, KERN_WARNING,
1975 "Test dummy encryption mode enabled");
1977 ext4_msg(sb, KERN_WARNING,
1978 "Test dummy encryption mount option ignored");
1980 } else if (m->flags & MOPT_DATAJ) {
1982 if (!sbi->s_journal)
1983 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1984 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1985 ext4_msg(sb, KERN_ERR,
1986 "Cannot change data mode on remount");
1990 clear_opt(sb, DATA_FLAGS);
1991 sbi->s_mount_opt |= m->mount_opt;
1994 } else if (m->flags & MOPT_QFMT) {
1995 if (sb_any_quota_loaded(sb) &&
1996 sbi->s_jquota_fmt != m->mount_opt) {
1997 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1998 "quota options when quota turned on");
2001 if (ext4_has_feature_quota(sb)) {
2002 ext4_msg(sb, KERN_INFO,
2003 "Quota format mount options ignored "
2004 "when QUOTA feature is enabled");
2007 sbi->s_jquota_fmt = m->mount_opt;
2009 } else if (token == Opt_dax) {
2010 #ifdef CONFIG_FS_DAX
2011 ext4_msg(sb, KERN_WARNING,
2012 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2013 sbi->s_mount_opt |= m->mount_opt;
2015 ext4_msg(sb, KERN_INFO, "dax option not supported");
2018 } else if (token == Opt_data_err_abort) {
2019 sbi->s_mount_opt |= m->mount_opt;
2020 } else if (token == Opt_data_err_ignore) {
2021 sbi->s_mount_opt &= ~m->mount_opt;
2025 if (m->flags & MOPT_CLEAR)
2027 else if (unlikely(!(m->flags & MOPT_SET))) {
2028 ext4_msg(sb, KERN_WARNING,
2029 "buggy handling of option %s", opt);
2034 sbi->s_mount_opt |= m->mount_opt;
2036 sbi->s_mount_opt &= ~m->mount_opt;
2041 static int parse_options(char *options, struct super_block *sb,
2042 unsigned long *journal_devnum,
2043 unsigned int *journal_ioprio,
2046 struct ext4_sb_info *sbi = EXT4_SB(sb);
2047 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
2048 substring_t args[MAX_OPT_ARGS];
2054 while ((p = strsep(&options, ",")) != NULL) {
2058 * Initialize args struct so we know whether arg was
2059 * found; some options take optional arguments.
2061 args[0].to = args[0].from = NULL;
2062 token = match_token(p, tokens, args);
2063 if (handle_mount_opt(sb, p, token, args, journal_devnum,
2064 journal_ioprio, is_remount) < 0)
2069 * We do the test below only for project quotas. 'usrquota' and
2070 * 'grpquota' mount options are allowed even without quota feature
2071 * to support legacy quotas in quota files.
2073 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2074 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2075 "Cannot enable project quota enforcement.");
2078 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2079 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2080 if (usr_qf_name || grp_qf_name) {
2081 if (test_opt(sb, USRQUOTA) && usr_qf_name)
2082 clear_opt(sb, USRQUOTA);
2084 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
2085 clear_opt(sb, GRPQUOTA);
2087 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
2088 ext4_msg(sb, KERN_ERR, "old and new quota "
2093 if (!sbi->s_jquota_fmt) {
2094 ext4_msg(sb, KERN_ERR, "journaled quota format "
2100 if (test_opt(sb, DIOREAD_NOLOCK)) {
2102 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2104 if (blocksize < PAGE_SIZE) {
2105 ext4_msg(sb, KERN_ERR, "can't mount with "
2106 "dioread_nolock if block size != PAGE_SIZE");
2113 static inline void ext4_show_quota_options(struct seq_file *seq,
2114 struct super_block *sb)
2116 #if defined(CONFIG_QUOTA)
2117 struct ext4_sb_info *sbi = EXT4_SB(sb);
2118 char *usr_qf_name, *grp_qf_name;
2120 if (sbi->s_jquota_fmt) {
2123 switch (sbi->s_jquota_fmt) {
2134 seq_printf(seq, ",jqfmt=%s", fmtname);
2138 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2139 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2141 seq_show_option(seq, "usrjquota", usr_qf_name);
2143 seq_show_option(seq, "grpjquota", grp_qf_name);
2148 static const char *token2str(int token)
2150 const struct match_token *t;
2152 for (t = tokens; t->token != Opt_err; t++)
2153 if (t->token == token && !strchr(t->pattern, '='))
2160 * - it's set to a non-default value OR
2161 * - if the per-sb default is different from the global default
2163 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2166 struct ext4_sb_info *sbi = EXT4_SB(sb);
2167 struct ext4_super_block *es = sbi->s_es;
2168 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2169 const struct mount_opts *m;
2170 char sep = nodefs ? '\n' : ',';
2172 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2173 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2175 if (sbi->s_sb_block != 1)
2176 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2178 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2179 int want_set = m->flags & MOPT_SET;
2180 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2181 (m->flags & MOPT_CLEAR_ERR))
2183 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2184 continue; /* skip if same as the default */
2186 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2187 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2188 continue; /* select Opt_noFoo vs Opt_Foo */
2189 SEQ_OPTS_PRINT("%s", token2str(m->token));
2192 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2193 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2194 SEQ_OPTS_PRINT("resuid=%u",
2195 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2196 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2197 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2198 SEQ_OPTS_PRINT("resgid=%u",
2199 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2200 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2201 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2202 SEQ_OPTS_PUTS("errors=remount-ro");
2203 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2204 SEQ_OPTS_PUTS("errors=continue");
2205 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2206 SEQ_OPTS_PUTS("errors=panic");
2207 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2208 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2209 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2210 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2211 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2212 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2213 if (sb->s_flags & SB_I_VERSION)
2214 SEQ_OPTS_PUTS("i_version");
2215 if (nodefs || sbi->s_stripe)
2216 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2217 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2218 (sbi->s_mount_opt ^ def_mount_opt)) {
2219 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2220 SEQ_OPTS_PUTS("data=journal");
2221 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2222 SEQ_OPTS_PUTS("data=ordered");
2223 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2224 SEQ_OPTS_PUTS("data=writeback");
2227 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2228 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2229 sbi->s_inode_readahead_blks);
2231 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2232 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2233 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2234 if (nodefs || sbi->s_max_dir_size_kb)
2235 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2236 if (test_opt(sb, DATA_ERR_ABORT))
2237 SEQ_OPTS_PUTS("data_err=abort");
2238 if (DUMMY_ENCRYPTION_ENABLED(sbi))
2239 SEQ_OPTS_PUTS("test_dummy_encryption");
2241 ext4_show_quota_options(seq, sb);
2245 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2247 return _ext4_show_options(seq, root->d_sb, 0);
2250 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2252 struct super_block *sb = seq->private;
2255 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2256 rc = _ext4_show_options(seq, sb, 1);
2257 seq_puts(seq, "\n");
2261 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2264 struct ext4_sb_info *sbi = EXT4_SB(sb);
2267 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2268 ext4_msg(sb, KERN_ERR, "revision level too high, "
2269 "forcing read-only mode");
2274 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2275 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2276 "running e2fsck is recommended");
2277 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2278 ext4_msg(sb, KERN_WARNING,
2279 "warning: mounting fs with errors, "
2280 "running e2fsck is recommended");
2281 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2282 le16_to_cpu(es->s_mnt_count) >=
2283 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2284 ext4_msg(sb, KERN_WARNING,
2285 "warning: maximal mount count reached, "
2286 "running e2fsck is recommended");
2287 else if (le32_to_cpu(es->s_checkinterval) &&
2288 (ext4_get_tstamp(es, s_lastcheck) +
2289 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
2290 ext4_msg(sb, KERN_WARNING,
2291 "warning: checktime reached, "
2292 "running e2fsck is recommended");
2293 if (!sbi->s_journal)
2294 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2295 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2296 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2297 le16_add_cpu(&es->s_mnt_count, 1);
2298 ext4_update_tstamp(es, s_mtime);
2300 ext4_set_feature_journal_needs_recovery(sb);
2302 err = ext4_commit_super(sb, 1);
2304 if (test_opt(sb, DEBUG))
2305 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2306 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2308 sbi->s_groups_count,
2309 EXT4_BLOCKS_PER_GROUP(sb),
2310 EXT4_INODES_PER_GROUP(sb),
2311 sbi->s_mount_opt, sbi->s_mount_opt2);
2313 cleancache_init_fs(sb);
2317 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2319 struct ext4_sb_info *sbi = EXT4_SB(sb);
2320 struct flex_groups *new_groups;
2323 if (!sbi->s_log_groups_per_flex)
2326 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2327 if (size <= sbi->s_flex_groups_allocated)
2330 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2331 new_groups = kvzalloc(size, GFP_KERNEL);
2333 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2334 size / (int) sizeof(struct flex_groups));
2338 if (sbi->s_flex_groups) {
2339 memcpy(new_groups, sbi->s_flex_groups,
2340 (sbi->s_flex_groups_allocated *
2341 sizeof(struct flex_groups)));
2342 kvfree(sbi->s_flex_groups);
2344 sbi->s_flex_groups = new_groups;
2345 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2349 static int ext4_fill_flex_info(struct super_block *sb)
2351 struct ext4_sb_info *sbi = EXT4_SB(sb);
2352 struct ext4_group_desc *gdp = NULL;
2353 ext4_group_t flex_group;
2356 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2357 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2358 sbi->s_log_groups_per_flex = 0;
2362 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2366 for (i = 0; i < sbi->s_groups_count; i++) {
2367 gdp = ext4_get_group_desc(sb, i, NULL);
2369 flex_group = ext4_flex_group(sbi, i);
2370 atomic_add(ext4_free_inodes_count(sb, gdp),
2371 &sbi->s_flex_groups[flex_group].free_inodes);
2372 atomic64_add(ext4_free_group_clusters(sb, gdp),
2373 &sbi->s_flex_groups[flex_group].free_clusters);
2374 atomic_add(ext4_used_dirs_count(sb, gdp),
2375 &sbi->s_flex_groups[flex_group].used_dirs);
2383 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2384 struct ext4_group_desc *gdp)
2386 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2388 __le32 le_group = cpu_to_le32(block_group);
2389 struct ext4_sb_info *sbi = EXT4_SB(sb);
2391 if (ext4_has_metadata_csum(sbi->s_sb)) {
2392 /* Use new metadata_csum algorithm */
2394 __u16 dummy_csum = 0;
2396 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2398 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2399 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2400 sizeof(dummy_csum));
2401 offset += sizeof(dummy_csum);
2402 if (offset < sbi->s_desc_size)
2403 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2404 sbi->s_desc_size - offset);
2406 crc = csum32 & 0xFFFF;
2410 /* old crc16 code */
2411 if (!ext4_has_feature_gdt_csum(sb))
2414 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2415 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2416 crc = crc16(crc, (__u8 *)gdp, offset);
2417 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2418 /* for checksum of struct ext4_group_desc do the rest...*/
2419 if (ext4_has_feature_64bit(sb) &&
2420 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2421 crc = crc16(crc, (__u8 *)gdp + offset,
2422 le16_to_cpu(sbi->s_es->s_desc_size) -
2426 return cpu_to_le16(crc);
2429 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2430 struct ext4_group_desc *gdp)
2432 if (ext4_has_group_desc_csum(sb) &&
2433 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2439 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2440 struct ext4_group_desc *gdp)
2442 if (!ext4_has_group_desc_csum(sb))
2444 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2447 /* Called at mount-time, super-block is locked */
2448 static int ext4_check_descriptors(struct super_block *sb,
2449 ext4_fsblk_t sb_block,
2450 ext4_group_t *first_not_zeroed)
2452 struct ext4_sb_info *sbi = EXT4_SB(sb);
2453 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2454 ext4_fsblk_t last_block;
2455 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2456 ext4_fsblk_t block_bitmap;
2457 ext4_fsblk_t inode_bitmap;
2458 ext4_fsblk_t inode_table;
2459 int flexbg_flag = 0;
2460 ext4_group_t i, grp = sbi->s_groups_count;
2462 if (ext4_has_feature_flex_bg(sb))
2465 ext4_debug("Checking group descriptors");
2467 for (i = 0; i < sbi->s_groups_count; i++) {
2468 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2470 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2471 last_block = ext4_blocks_count(sbi->s_es) - 1;
2473 last_block = first_block +
2474 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2476 if ((grp == sbi->s_groups_count) &&
2477 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2480 block_bitmap = ext4_block_bitmap(sb, gdp);
2481 if (block_bitmap == sb_block) {
2482 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2483 "Block bitmap for group %u overlaps "
2488 if (block_bitmap >= sb_block + 1 &&
2489 block_bitmap <= last_bg_block) {
2490 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2491 "Block bitmap for group %u overlaps "
2492 "block group descriptors", i);
2496 if (block_bitmap < first_block || block_bitmap > last_block) {
2497 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2498 "Block bitmap for group %u not in group "
2499 "(block %llu)!", i, block_bitmap);
2502 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2503 if (inode_bitmap == sb_block) {
2504 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2505 "Inode bitmap for group %u overlaps "
2510 if (inode_bitmap >= sb_block + 1 &&
2511 inode_bitmap <= last_bg_block) {
2512 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2513 "Inode bitmap for group %u overlaps "
2514 "block group descriptors", i);
2518 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2519 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2520 "Inode bitmap for group %u not in group "
2521 "(block %llu)!", i, inode_bitmap);
2524 inode_table = ext4_inode_table(sb, gdp);
2525 if (inode_table == sb_block) {
2526 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2527 "Inode table for group %u overlaps "
2532 if (inode_table >= sb_block + 1 &&
2533 inode_table <= last_bg_block) {
2534 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2535 "Inode table for group %u overlaps "
2536 "block group descriptors", i);
2540 if (inode_table < first_block ||
2541 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2542 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2543 "Inode table for group %u not in group "
2544 "(block %llu)!", i, inode_table);
2547 ext4_lock_group(sb, i);
2548 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2549 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2550 "Checksum for group %u failed (%u!=%u)",
2551 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2552 gdp)), le16_to_cpu(gdp->bg_checksum));
2553 if (!sb_rdonly(sb)) {
2554 ext4_unlock_group(sb, i);
2558 ext4_unlock_group(sb, i);
2560 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2562 if (NULL != first_not_zeroed)
2563 *first_not_zeroed = grp;
2567 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2568 * the superblock) which were deleted from all directories, but held open by
2569 * a process at the time of a crash. We walk the list and try to delete these
2570 * inodes at recovery time (only with a read-write filesystem).
2572 * In order to keep the orphan inode chain consistent during traversal (in
2573 * case of crash during recovery), we link each inode into the superblock
2574 * orphan list_head and handle it the same way as an inode deletion during
2575 * normal operation (which journals the operations for us).
2577 * We only do an iget() and an iput() on each inode, which is very safe if we
2578 * accidentally point at an in-use or already deleted inode. The worst that
2579 * can happen in this case is that we get a "bit already cleared" message from
2580 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2581 * e2fsck was run on this filesystem, and it must have already done the orphan
2582 * inode cleanup for us, so we can safely abort without any further action.
2584 static void ext4_orphan_cleanup(struct super_block *sb,
2585 struct ext4_super_block *es)
2587 unsigned int s_flags = sb->s_flags;
2588 int ret, nr_orphans = 0, nr_truncates = 0;
2590 int quota_update = 0;
2593 if (!es->s_last_orphan) {
2594 jbd_debug(4, "no orphan inodes to clean up\n");
2598 if (bdev_read_only(sb->s_bdev)) {
2599 ext4_msg(sb, KERN_ERR, "write access "
2600 "unavailable, skipping orphan cleanup");
2604 /* Check if feature set would not allow a r/w mount */
2605 if (!ext4_feature_set_ok(sb, 0)) {
2606 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2607 "unknown ROCOMPAT features");
2611 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2612 /* don't clear list on RO mount w/ errors */
2613 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
2614 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2615 "clearing orphan list.\n");
2616 es->s_last_orphan = 0;
2618 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2622 if (s_flags & SB_RDONLY) {
2623 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2624 sb->s_flags &= ~SB_RDONLY;
2627 /* Needed for iput() to work correctly and not trash data */
2628 sb->s_flags |= SB_ACTIVE;
2631 * Turn on quotas which were not enabled for read-only mounts if
2632 * filesystem has quota feature, so that they are updated correctly.
2634 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
2635 int ret = ext4_enable_quotas(sb);
2640 ext4_msg(sb, KERN_ERR,
2641 "Cannot turn on quotas: error %d", ret);
2644 /* Turn on journaled quotas used for old sytle */
2645 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2646 if (EXT4_SB(sb)->s_qf_names[i]) {
2647 int ret = ext4_quota_on_mount(sb, i);
2652 ext4_msg(sb, KERN_ERR,
2653 "Cannot turn on journaled "
2654 "quota: type %d: error %d", i, ret);
2659 while (es->s_last_orphan) {
2660 struct inode *inode;
2663 * We may have encountered an error during cleanup; if
2664 * so, skip the rest.
2666 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2667 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2668 es->s_last_orphan = 0;
2672 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2673 if (IS_ERR(inode)) {
2674 es->s_last_orphan = 0;
2678 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2679 dquot_initialize(inode);
2680 if (inode->i_nlink) {
2681 if (test_opt(sb, DEBUG))
2682 ext4_msg(sb, KERN_DEBUG,
2683 "%s: truncating inode %lu to %lld bytes",
2684 __func__, inode->i_ino, inode->i_size);
2685 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2686 inode->i_ino, inode->i_size);
2688 truncate_inode_pages(inode->i_mapping, inode->i_size);
2689 ret = ext4_truncate(inode);
2691 ext4_std_error(inode->i_sb, ret);
2692 inode_unlock(inode);
2695 if (test_opt(sb, DEBUG))
2696 ext4_msg(sb, KERN_DEBUG,
2697 "%s: deleting unreferenced inode %lu",
2698 __func__, inode->i_ino);
2699 jbd_debug(2, "deleting unreferenced inode %lu\n",
2703 iput(inode); /* The delete magic happens here! */
2706 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2709 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2710 PLURAL(nr_orphans));
2712 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2713 PLURAL(nr_truncates));
2715 /* Turn off quotas if they were enabled for orphan cleanup */
2717 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2718 if (sb_dqopt(sb)->files[i])
2719 dquot_quota_off(sb, i);
2723 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2727 * Maximal extent format file size.
2728 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2729 * extent format containers, within a sector_t, and within i_blocks
2730 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2731 * so that won't be a limiting factor.
2733 * However there is other limiting factor. We do store extents in the form
2734 * of starting block and length, hence the resulting length of the extent
2735 * covering maximum file size must fit into on-disk format containers as
2736 * well. Given that length is always by 1 unit bigger than max unit (because
2737 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2739 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2741 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2744 loff_t upper_limit = MAX_LFS_FILESIZE;
2746 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
2748 if (!has_huge_files) {
2749 upper_limit = (1LL << 32) - 1;
2751 /* total blocks in file system block size */
2752 upper_limit >>= (blkbits - 9);
2753 upper_limit <<= blkbits;
2757 * 32-bit extent-start container, ee_block. We lower the maxbytes
2758 * by one fs block, so ee_len can cover the extent of maximum file
2761 res = (1LL << 32) - 1;
2764 /* Sanity check against vm- & vfs- imposed limits */
2765 if (res > upper_limit)
2772 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2773 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2774 * We need to be 1 filesystem block less than the 2^48 sector limit.
2776 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2778 loff_t res = EXT4_NDIR_BLOCKS;
2781 /* This is calculated to be the largest file size for a dense, block
2782 * mapped file such that the file's total number of 512-byte sectors,
2783 * including data and all indirect blocks, does not exceed (2^48 - 1).
2785 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2786 * number of 512-byte sectors of the file.
2789 if (!has_huge_files) {
2791 * !has_huge_files or implies that the inode i_block field
2792 * represents total file blocks in 2^32 512-byte sectors ==
2793 * size of vfs inode i_blocks * 8
2795 upper_limit = (1LL << 32) - 1;
2797 /* total blocks in file system block size */
2798 upper_limit >>= (bits - 9);
2802 * We use 48 bit ext4_inode i_blocks
2803 * With EXT4_HUGE_FILE_FL set the i_blocks
2804 * represent total number of blocks in
2805 * file system block size
2807 upper_limit = (1LL << 48) - 1;
2811 /* indirect blocks */
2813 /* double indirect blocks */
2814 meta_blocks += 1 + (1LL << (bits-2));
2815 /* tripple indirect blocks */
2816 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2818 upper_limit -= meta_blocks;
2819 upper_limit <<= bits;
2821 res += 1LL << (bits-2);
2822 res += 1LL << (2*(bits-2));
2823 res += 1LL << (3*(bits-2));
2825 if (res > upper_limit)
2828 if (res > MAX_LFS_FILESIZE)
2829 res = MAX_LFS_FILESIZE;
2834 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2835 ext4_fsblk_t logical_sb_block, int nr)
2837 struct ext4_sb_info *sbi = EXT4_SB(sb);
2838 ext4_group_t bg, first_meta_bg;
2841 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2843 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2844 return logical_sb_block + nr + 1;
2845 bg = sbi->s_desc_per_block * nr;
2846 if (ext4_bg_has_super(sb, bg))
2850 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2851 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2852 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2855 if (sb->s_blocksize == 1024 && nr == 0 &&
2856 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
2859 return (has_super + ext4_group_first_block_no(sb, bg));
2863 * ext4_get_stripe_size: Get the stripe size.
2864 * @sbi: In memory super block info
2866 * If we have specified it via mount option, then
2867 * use the mount option value. If the value specified at mount time is
2868 * greater than the blocks per group use the super block value.
2869 * If the super block value is greater than blocks per group return 0.
2870 * Allocator needs it be less than blocks per group.
2873 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2875 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2876 unsigned long stripe_width =
2877 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2880 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2881 ret = sbi->s_stripe;
2882 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
2884 else if (stride && stride <= sbi->s_blocks_per_group)
2890 * If the stripe width is 1, this makes no sense and
2891 * we set it to 0 to turn off stripe handling code.
2900 * Check whether this filesystem can be mounted based on
2901 * the features present and the RDONLY/RDWR mount requested.
2902 * Returns 1 if this filesystem can be mounted as requested,
2903 * 0 if it cannot be.
2905 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2907 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2908 ext4_msg(sb, KERN_ERR,
2909 "Couldn't mount because of "
2910 "unsupported optional features (%x)",
2911 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2912 ~EXT4_FEATURE_INCOMPAT_SUPP));
2916 #ifndef CONFIG_UNICODE
2917 if (ext4_has_feature_casefold(sb)) {
2918 ext4_msg(sb, KERN_ERR,
2919 "Filesystem with casefold feature cannot be "
2920 "mounted without CONFIG_UNICODE");
2928 if (ext4_has_feature_readonly(sb)) {
2929 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2930 sb->s_flags |= SB_RDONLY;
2934 /* Check that feature set is OK for a read-write mount */
2935 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2936 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2937 "unsupported optional features (%x)",
2938 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2939 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2942 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2943 ext4_msg(sb, KERN_ERR,
2944 "Can't support bigalloc feature without "
2945 "extents feature\n");
2949 #ifndef CONFIG_QUOTA
2950 if (ext4_has_feature_quota(sb) && !readonly) {
2951 ext4_msg(sb, KERN_ERR,
2952 "Filesystem with quota feature cannot be mounted RDWR "
2953 "without CONFIG_QUOTA");
2956 if (ext4_has_feature_project(sb) && !readonly) {
2957 ext4_msg(sb, KERN_ERR,
2958 "Filesystem with project quota feature cannot be mounted RDWR "
2959 "without CONFIG_QUOTA");
2962 #endif /* CONFIG_QUOTA */
2967 * This function is called once a day if we have errors logged
2968 * on the file system
2970 static void print_daily_error_info(struct timer_list *t)
2972 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
2973 struct super_block *sb = sbi->s_sb;
2974 struct ext4_super_block *es = sbi->s_es;
2976 if (es->s_error_count)
2977 /* fsck newer than v1.41.13 is needed to clean this condition. */
2978 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2979 le32_to_cpu(es->s_error_count));
2980 if (es->s_first_error_time) {
2981 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
2983 ext4_get_tstamp(es, s_first_error_time),
2984 (int) sizeof(es->s_first_error_func),
2985 es->s_first_error_func,
2986 le32_to_cpu(es->s_first_error_line));
2987 if (es->s_first_error_ino)
2988 printk(KERN_CONT ": inode %u",
2989 le32_to_cpu(es->s_first_error_ino));
2990 if (es->s_first_error_block)
2991 printk(KERN_CONT ": block %llu", (unsigned long long)
2992 le64_to_cpu(es->s_first_error_block));
2993 printk(KERN_CONT "\n");
2995 if (es->s_last_error_time) {
2996 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
2998 ext4_get_tstamp(es, s_last_error_time),
2999 (int) sizeof(es->s_last_error_func),
3000 es->s_last_error_func,
3001 le32_to_cpu(es->s_last_error_line));
3002 if (es->s_last_error_ino)
3003 printk(KERN_CONT ": inode %u",
3004 le32_to_cpu(es->s_last_error_ino));
3005 if (es->s_last_error_block)
3006 printk(KERN_CONT ": block %llu", (unsigned long long)
3007 le64_to_cpu(es->s_last_error_block));
3008 printk(KERN_CONT "\n");
3010 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3013 /* Find next suitable group and run ext4_init_inode_table */
3014 static int ext4_run_li_request(struct ext4_li_request *elr)
3016 struct ext4_group_desc *gdp = NULL;
3017 ext4_group_t group, ngroups;
3018 struct super_block *sb;
3019 unsigned long timeout = 0;
3023 ngroups = EXT4_SB(sb)->s_groups_count;
3025 for (group = elr->lr_next_group; group < ngroups; group++) {
3026 gdp = ext4_get_group_desc(sb, group, NULL);
3032 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3036 if (group >= ngroups)
3041 ret = ext4_init_inode_table(sb, group,
3042 elr->lr_timeout ? 0 : 1);
3043 if (elr->lr_timeout == 0) {
3044 timeout = (jiffies - timeout) *
3045 elr->lr_sbi->s_li_wait_mult;
3046 elr->lr_timeout = timeout;
3048 elr->lr_next_sched = jiffies + elr->lr_timeout;
3049 elr->lr_next_group = group + 1;
3055 * Remove lr_request from the list_request and free the
3056 * request structure. Should be called with li_list_mtx held
3058 static void ext4_remove_li_request(struct ext4_li_request *elr)
3060 struct ext4_sb_info *sbi;
3067 list_del(&elr->lr_request);
3068 sbi->s_li_request = NULL;
3072 static void ext4_unregister_li_request(struct super_block *sb)
3074 mutex_lock(&ext4_li_mtx);
3075 if (!ext4_li_info) {
3076 mutex_unlock(&ext4_li_mtx);
3080 mutex_lock(&ext4_li_info->li_list_mtx);
3081 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3082 mutex_unlock(&ext4_li_info->li_list_mtx);
3083 mutex_unlock(&ext4_li_mtx);
3086 static struct task_struct *ext4_lazyinit_task;
3089 * This is the function where ext4lazyinit thread lives. It walks
3090 * through the request list searching for next scheduled filesystem.
3091 * When such a fs is found, run the lazy initialization request
3092 * (ext4_rn_li_request) and keep track of the time spend in this
3093 * function. Based on that time we compute next schedule time of
3094 * the request. When walking through the list is complete, compute
3095 * next waking time and put itself into sleep.
3097 static int ext4_lazyinit_thread(void *arg)
3099 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3100 struct list_head *pos, *n;
3101 struct ext4_li_request *elr;
3102 unsigned long next_wakeup, cur;
3104 BUG_ON(NULL == eli);
3108 next_wakeup = MAX_JIFFY_OFFSET;
3110 mutex_lock(&eli->li_list_mtx);
3111 if (list_empty(&eli->li_request_list)) {
3112 mutex_unlock(&eli->li_list_mtx);
3115 list_for_each_safe(pos, n, &eli->li_request_list) {
3118 elr = list_entry(pos, struct ext4_li_request,
3121 if (time_before(jiffies, elr->lr_next_sched)) {
3122 if (time_before(elr->lr_next_sched, next_wakeup))
3123 next_wakeup = elr->lr_next_sched;
3126 if (down_read_trylock(&elr->lr_super->s_umount)) {
3127 if (sb_start_write_trylock(elr->lr_super)) {
3130 * We hold sb->s_umount, sb can not
3131 * be removed from the list, it is
3132 * now safe to drop li_list_mtx
3134 mutex_unlock(&eli->li_list_mtx);
3135 err = ext4_run_li_request(elr);
3136 sb_end_write(elr->lr_super);
3137 mutex_lock(&eli->li_list_mtx);
3140 up_read((&elr->lr_super->s_umount));
3142 /* error, remove the lazy_init job */
3144 ext4_remove_li_request(elr);
3148 elr->lr_next_sched = jiffies +
3150 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3152 if (time_before(elr->lr_next_sched, next_wakeup))
3153 next_wakeup = elr->lr_next_sched;
3155 mutex_unlock(&eli->li_list_mtx);
3160 if ((time_after_eq(cur, next_wakeup)) ||
3161 (MAX_JIFFY_OFFSET == next_wakeup)) {
3166 schedule_timeout_interruptible(next_wakeup - cur);
3168 if (kthread_should_stop()) {
3169 ext4_clear_request_list();
3176 * It looks like the request list is empty, but we need
3177 * to check it under the li_list_mtx lock, to prevent any
3178 * additions into it, and of course we should lock ext4_li_mtx
3179 * to atomically free the list and ext4_li_info, because at
3180 * this point another ext4 filesystem could be registering
3183 mutex_lock(&ext4_li_mtx);
3184 mutex_lock(&eli->li_list_mtx);
3185 if (!list_empty(&eli->li_request_list)) {
3186 mutex_unlock(&eli->li_list_mtx);
3187 mutex_unlock(&ext4_li_mtx);
3190 mutex_unlock(&eli->li_list_mtx);
3191 kfree(ext4_li_info);
3192 ext4_li_info = NULL;
3193 mutex_unlock(&ext4_li_mtx);
3198 static void ext4_clear_request_list(void)
3200 struct list_head *pos, *n;
3201 struct ext4_li_request *elr;
3203 mutex_lock(&ext4_li_info->li_list_mtx);
3204 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3205 elr = list_entry(pos, struct ext4_li_request,
3207 ext4_remove_li_request(elr);
3209 mutex_unlock(&ext4_li_info->li_list_mtx);
3212 static int ext4_run_lazyinit_thread(void)
3214 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3215 ext4_li_info, "ext4lazyinit");
3216 if (IS_ERR(ext4_lazyinit_task)) {
3217 int err = PTR_ERR(ext4_lazyinit_task);
3218 ext4_clear_request_list();
3219 kfree(ext4_li_info);
3220 ext4_li_info = NULL;
3221 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3222 "initialization thread\n",
3226 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3231 * Check whether it make sense to run itable init. thread or not.
3232 * If there is at least one uninitialized inode table, return
3233 * corresponding group number, else the loop goes through all
3234 * groups and return total number of groups.
3236 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3238 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3239 struct ext4_group_desc *gdp = NULL;
3241 if (!ext4_has_group_desc_csum(sb))
3244 for (group = 0; group < ngroups; group++) {
3245 gdp = ext4_get_group_desc(sb, group, NULL);
3249 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3256 static int ext4_li_info_new(void)
3258 struct ext4_lazy_init *eli = NULL;
3260 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3264 INIT_LIST_HEAD(&eli->li_request_list);
3265 mutex_init(&eli->li_list_mtx);
3267 eli->li_state |= EXT4_LAZYINIT_QUIT;
3274 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3277 struct ext4_sb_info *sbi = EXT4_SB(sb);
3278 struct ext4_li_request *elr;
3280 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3286 elr->lr_next_group = start;
3289 * Randomize first schedule time of the request to
3290 * spread the inode table initialization requests
3293 elr->lr_next_sched = jiffies + (prandom_u32() %
3294 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3298 int ext4_register_li_request(struct super_block *sb,
3299 ext4_group_t first_not_zeroed)
3301 struct ext4_sb_info *sbi = EXT4_SB(sb);
3302 struct ext4_li_request *elr = NULL;
3303 ext4_group_t ngroups = sbi->s_groups_count;
3306 mutex_lock(&ext4_li_mtx);
3307 if (sbi->s_li_request != NULL) {
3309 * Reset timeout so it can be computed again, because
3310 * s_li_wait_mult might have changed.
3312 sbi->s_li_request->lr_timeout = 0;
3316 if (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3317 !test_opt(sb, INIT_INODE_TABLE))
3320 elr = ext4_li_request_new(sb, first_not_zeroed);
3326 if (NULL == ext4_li_info) {
3327 ret = ext4_li_info_new();
3332 mutex_lock(&ext4_li_info->li_list_mtx);
3333 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3334 mutex_unlock(&ext4_li_info->li_list_mtx);
3336 sbi->s_li_request = elr;
3338 * set elr to NULL here since it has been inserted to
3339 * the request_list and the removal and free of it is
3340 * handled by ext4_clear_request_list from now on.
3344 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3345 ret = ext4_run_lazyinit_thread();
3350 mutex_unlock(&ext4_li_mtx);
3357 * We do not need to lock anything since this is called on
3360 static void ext4_destroy_lazyinit_thread(void)
3363 * If thread exited earlier
3364 * there's nothing to be done.
3366 if (!ext4_li_info || !ext4_lazyinit_task)
3369 kthread_stop(ext4_lazyinit_task);
3372 static int set_journal_csum_feature_set(struct super_block *sb)
3375 int compat, incompat;
3376 struct ext4_sb_info *sbi = EXT4_SB(sb);
3378 if (ext4_has_metadata_csum(sb)) {
3379 /* journal checksum v3 */
3381 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3383 /* journal checksum v1 */
3384 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3388 jbd2_journal_clear_features(sbi->s_journal,
3389 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3390 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3391 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3392 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3393 ret = jbd2_journal_set_features(sbi->s_journal,
3395 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3397 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3398 ret = jbd2_journal_set_features(sbi->s_journal,
3401 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3402 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3404 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3405 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3412 * Note: calculating the overhead so we can be compatible with
3413 * historical BSD practice is quite difficult in the face of
3414 * clusters/bigalloc. This is because multiple metadata blocks from
3415 * different block group can end up in the same allocation cluster.
3416 * Calculating the exact overhead in the face of clustered allocation
3417 * requires either O(all block bitmaps) in memory or O(number of block
3418 * groups**2) in time. We will still calculate the superblock for
3419 * older file systems --- and if we come across with a bigalloc file
3420 * system with zero in s_overhead_clusters the estimate will be close to
3421 * correct especially for very large cluster sizes --- but for newer
3422 * file systems, it's better to calculate this figure once at mkfs
3423 * time, and store it in the superblock. If the superblock value is
3424 * present (even for non-bigalloc file systems), we will use it.
3426 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3429 struct ext4_sb_info *sbi = EXT4_SB(sb);
3430 struct ext4_group_desc *gdp;
3431 ext4_fsblk_t first_block, last_block, b;
3432 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3433 int s, j, count = 0;
3435 if (!ext4_has_feature_bigalloc(sb))
3436 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3437 sbi->s_itb_per_group + 2);
3439 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3440 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3441 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3442 for (i = 0; i < ngroups; i++) {
3443 gdp = ext4_get_group_desc(sb, i, NULL);
3444 b = ext4_block_bitmap(sb, gdp);
3445 if (b >= first_block && b <= last_block) {
3446 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3449 b = ext4_inode_bitmap(sb, gdp);
3450 if (b >= first_block && b <= last_block) {
3451 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3454 b = ext4_inode_table(sb, gdp);
3455 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3456 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3457 int c = EXT4_B2C(sbi, b - first_block);
3458 ext4_set_bit(c, buf);
3464 if (ext4_bg_has_super(sb, grp)) {
3465 ext4_set_bit(s++, buf);
3468 j = ext4_bg_num_gdb(sb, grp);
3469 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3470 ext4_error(sb, "Invalid number of block group "
3471 "descriptor blocks: %d", j);
3472 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3476 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3480 return EXT4_CLUSTERS_PER_GROUP(sb) -
3481 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3485 * Compute the overhead and stash it in sbi->s_overhead
3487 int ext4_calculate_overhead(struct super_block *sb)
3489 struct ext4_sb_info *sbi = EXT4_SB(sb);
3490 struct ext4_super_block *es = sbi->s_es;
3491 struct inode *j_inode;
3492 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3493 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3494 ext4_fsblk_t overhead = 0;
3495 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3501 * Compute the overhead (FS structures). This is constant
3502 * for a given filesystem unless the number of block groups
3503 * changes so we cache the previous value until it does.
3507 * All of the blocks before first_data_block are overhead
3509 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3512 * Add the overhead found in each block group
3514 for (i = 0; i < ngroups; i++) {
3517 blks = count_overhead(sb, i, buf);
3520 memset(buf, 0, PAGE_SIZE);
3525 * Add the internal journal blocks whether the journal has been
3528 if (sbi->s_journal && !sbi->journal_bdev)
3529 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3530 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3531 j_inode = ext4_get_journal_inode(sb, j_inum);
3533 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3534 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3537 ext4_msg(sb, KERN_ERR, "can't get journal size");
3540 sbi->s_overhead = overhead;
3542 free_page((unsigned long) buf);
3546 static void ext4_clamp_want_extra_isize(struct super_block *sb)
3548 struct ext4_sb_info *sbi = EXT4_SB(sb);
3549 struct ext4_super_block *es = sbi->s_es;
3551 /* determine the minimum size of new large inodes, if present */
3552 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE &&
3553 sbi->s_want_extra_isize == 0) {
3554 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3555 EXT4_GOOD_OLD_INODE_SIZE;
3556 if (ext4_has_feature_extra_isize(sb)) {
3557 if (sbi->s_want_extra_isize <
3558 le16_to_cpu(es->s_want_extra_isize))
3559 sbi->s_want_extra_isize =
3560 le16_to_cpu(es->s_want_extra_isize);
3561 if (sbi->s_want_extra_isize <
3562 le16_to_cpu(es->s_min_extra_isize))
3563 sbi->s_want_extra_isize =
3564 le16_to_cpu(es->s_min_extra_isize);
3567 /* Check if enough inode space is available */
3568 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3569 sbi->s_inode_size) {
3570 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3571 EXT4_GOOD_OLD_INODE_SIZE;
3572 ext4_msg(sb, KERN_INFO,
3573 "required extra inode space not available");
3577 static void ext4_set_resv_clusters(struct super_block *sb)
3579 ext4_fsblk_t resv_clusters;
3580 struct ext4_sb_info *sbi = EXT4_SB(sb);
3583 * There's no need to reserve anything when we aren't using extents.
3584 * The space estimates are exact, there are no unwritten extents,
3585 * hole punching doesn't need new metadata... This is needed especially
3586 * to keep ext2/3 backward compatibility.
3588 if (!ext4_has_feature_extents(sb))
3591 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3592 * This should cover the situations where we can not afford to run
3593 * out of space like for example punch hole, or converting
3594 * unwritten extents in delalloc path. In most cases such
3595 * allocation would require 1, or 2 blocks, higher numbers are
3598 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3599 sbi->s_cluster_bits);
3601 do_div(resv_clusters, 50);
3602 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3604 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3607 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3609 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
3610 char *orig_data = kstrdup(data, GFP_KERNEL);
3611 struct buffer_head *bh;
3612 struct ext4_super_block *es = NULL;
3613 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3615 ext4_fsblk_t sb_block = get_sb_block(&data);
3616 ext4_fsblk_t logical_sb_block;
3617 unsigned long offset = 0;
3618 unsigned long journal_devnum = 0;
3619 unsigned long def_mount_opts;
3623 int blocksize, clustersize;
3624 unsigned int db_count;
3626 int needs_recovery, has_huge_files, has_bigalloc;
3629 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3630 ext4_group_t first_not_zeroed;
3632 if ((data && !orig_data) || !sbi)
3635 sbi->s_daxdev = dax_dev;
3636 sbi->s_blockgroup_lock =
3637 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3638 if (!sbi->s_blockgroup_lock)
3641 sb->s_fs_info = sbi;
3643 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3644 sbi->s_sb_block = sb_block;
3645 if (sb->s_bdev->bd_part)
3646 sbi->s_sectors_written_start =
3647 part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]);
3649 /* Cleanup superblock name */
3650 strreplace(sb->s_id, '/', '!');
3652 /* -EINVAL is default */
3654 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3656 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3661 * The ext4 superblock will not be buffer aligned for other than 1kB
3662 * block sizes. We need to calculate the offset from buffer start.
3664 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3665 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3666 offset = do_div(logical_sb_block, blocksize);
3668 logical_sb_block = sb_block;
3671 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3672 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3676 * Note: s_es must be initialized as soon as possible because
3677 * some ext4 macro-instructions depend on its value
3679 es = (struct ext4_super_block *) (bh->b_data + offset);
3681 sb->s_magic = le16_to_cpu(es->s_magic);
3682 if (sb->s_magic != EXT4_SUPER_MAGIC)
3684 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3686 /* Warn if metadata_csum and gdt_csum are both set. */
3687 if (ext4_has_feature_metadata_csum(sb) &&
3688 ext4_has_feature_gdt_csum(sb))
3689 ext4_warning(sb, "metadata_csum and uninit_bg are "
3690 "redundant flags; please run fsck.");
3692 /* Check for a known checksum algorithm */
3693 if (!ext4_verify_csum_type(sb, es)) {
3694 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3695 "unknown checksum algorithm.");
3700 /* Load the checksum driver */
3701 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3702 if (IS_ERR(sbi->s_chksum_driver)) {
3703 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3704 ret = PTR_ERR(sbi->s_chksum_driver);
3705 sbi->s_chksum_driver = NULL;
3709 /* Check superblock checksum */
3710 if (!ext4_superblock_csum_verify(sb, es)) {
3711 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3712 "invalid superblock checksum. Run e2fsck?");
3718 /* Precompute checksum seed for all metadata */
3719 if (ext4_has_feature_csum_seed(sb))
3720 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3721 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
3722 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3723 sizeof(es->s_uuid));
3725 /* Set defaults before we parse the mount options */
3726 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3727 set_opt(sb, INIT_INODE_TABLE);
3728 if (def_mount_opts & EXT4_DEFM_DEBUG)
3730 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3732 if (def_mount_opts & EXT4_DEFM_UID16)
3733 set_opt(sb, NO_UID32);
3734 /* xattr user namespace & acls are now defaulted on */
3735 set_opt(sb, XATTR_USER);
3736 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3737 set_opt(sb, POSIX_ACL);
3739 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3740 if (ext4_has_metadata_csum(sb))
3741 set_opt(sb, JOURNAL_CHECKSUM);
3743 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3744 set_opt(sb, JOURNAL_DATA);
3745 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3746 set_opt(sb, ORDERED_DATA);
3747 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3748 set_opt(sb, WRITEBACK_DATA);
3750 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3751 set_opt(sb, ERRORS_PANIC);
3752 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3753 set_opt(sb, ERRORS_CONT);
3755 set_opt(sb, ERRORS_RO);
3756 /* block_validity enabled by default; disable with noblock_validity */
3757 set_opt(sb, BLOCK_VALIDITY);
3758 if (def_mount_opts & EXT4_DEFM_DISCARD)
3759 set_opt(sb, DISCARD);
3761 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3762 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3763 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3764 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3765 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3767 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3768 set_opt(sb, BARRIER);
3771 * enable delayed allocation by default
3772 * Use -o nodelalloc to turn it off
3774 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3775 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3776 set_opt(sb, DELALLOC);
3779 * set default s_li_wait_mult for lazyinit, for the case there is
3780 * no mount option specified.
3782 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3784 if (sbi->s_es->s_mount_opts[0]) {
3785 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3786 sizeof(sbi->s_es->s_mount_opts),
3790 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3791 &journal_ioprio, 0)) {
3792 ext4_msg(sb, KERN_WARNING,
3793 "failed to parse options in superblock: %s",
3796 kfree(s_mount_opts);
3798 sbi->s_def_mount_opt = sbi->s_mount_opt;
3799 if (!parse_options((char *) data, sb, &journal_devnum,
3800 &journal_ioprio, 0))
3803 #ifdef CONFIG_UNICODE
3804 if (ext4_has_feature_casefold(sb) && !sbi->s_encoding) {
3805 const struct ext4_sb_encodings *encoding_info;
3806 struct unicode_map *encoding;
3807 __u16 encoding_flags;
3809 if (ext4_has_feature_encrypt(sb)) {
3810 ext4_msg(sb, KERN_ERR,
3811 "Can't mount with encoding and encryption");
3815 if (ext4_sb_read_encoding(es, &encoding_info,
3817 ext4_msg(sb, KERN_ERR,
3818 "Encoding requested by superblock is unknown");
3822 encoding = utf8_load(encoding_info->version);
3823 if (IS_ERR(encoding)) {
3824 ext4_msg(sb, KERN_ERR,
3825 "can't mount with superblock charset: %s-%s "
3826 "not supported by the kernel. flags: 0x%x.",
3827 encoding_info->name, encoding_info->version,
3831 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
3832 "%s-%s with flags 0x%hx", encoding_info->name,
3833 encoding_info->version?:"\b", encoding_flags);
3835 sbi->s_encoding = encoding;
3836 sbi->s_encoding_flags = encoding_flags;
3840 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3841 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3842 "with data=journal disables delayed "
3843 "allocation and O_DIRECT support!\n");
3844 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3845 ext4_msg(sb, KERN_ERR, "can't mount with "
3846 "both data=journal and delalloc");
3849 if (test_opt(sb, DIOREAD_NOLOCK)) {
3850 ext4_msg(sb, KERN_ERR, "can't mount with "
3851 "both data=journal and dioread_nolock");
3854 if (test_opt(sb, DAX)) {
3855 ext4_msg(sb, KERN_ERR, "can't mount with "
3856 "both data=journal and dax");
3859 if (ext4_has_feature_encrypt(sb)) {
3860 ext4_msg(sb, KERN_WARNING,
3861 "encrypted files will use data=ordered "
3862 "instead of data journaling mode");
3864 if (test_opt(sb, DELALLOC))
3865 clear_opt(sb, DELALLOC);
3867 sb->s_iflags |= SB_I_CGROUPWB;
3870 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3871 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
3873 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3874 (ext4_has_compat_features(sb) ||
3875 ext4_has_ro_compat_features(sb) ||
3876 ext4_has_incompat_features(sb)))
3877 ext4_msg(sb, KERN_WARNING,
3878 "feature flags set on rev 0 fs, "
3879 "running e2fsck is recommended");
3881 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3882 set_opt2(sb, HURD_COMPAT);
3883 if (ext4_has_feature_64bit(sb)) {
3884 ext4_msg(sb, KERN_ERR,
3885 "The Hurd can't support 64-bit file systems");
3890 * ea_inode feature uses l_i_version field which is not
3891 * available in HURD_COMPAT mode.
3893 if (ext4_has_feature_ea_inode(sb)) {
3894 ext4_msg(sb, KERN_ERR,
3895 "ea_inode feature is not supported for Hurd");
3900 if (IS_EXT2_SB(sb)) {
3901 if (ext2_feature_set_ok(sb))
3902 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3903 "using the ext4 subsystem");
3906 * If we're probing be silent, if this looks like
3907 * it's actually an ext[34] filesystem.
3909 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
3911 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3912 "to feature incompatibilities");
3917 if (IS_EXT3_SB(sb)) {
3918 if (ext3_feature_set_ok(sb))
3919 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3920 "using the ext4 subsystem");
3923 * If we're probing be silent, if this looks like
3924 * it's actually an ext4 filesystem.
3926 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
3928 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3929 "to feature incompatibilities");
3935 * Check feature flags regardless of the revision level, since we
3936 * previously didn't change the revision level when setting the flags,
3937 * so there is a chance incompat flags are set on a rev 0 filesystem.
3939 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
3942 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3943 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3944 blocksize > EXT4_MAX_BLOCK_SIZE) {
3945 ext4_msg(sb, KERN_ERR,
3946 "Unsupported filesystem blocksize %d (%d log_block_size)",
3947 blocksize, le32_to_cpu(es->s_log_block_size));
3950 if (le32_to_cpu(es->s_log_block_size) >
3951 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3952 ext4_msg(sb, KERN_ERR,
3953 "Invalid log block size: %u",
3954 le32_to_cpu(es->s_log_block_size));
3957 if (le32_to_cpu(es->s_log_cluster_size) >
3958 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3959 ext4_msg(sb, KERN_ERR,
3960 "Invalid log cluster size: %u",
3961 le32_to_cpu(es->s_log_cluster_size));
3965 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3966 ext4_msg(sb, KERN_ERR,
3967 "Number of reserved GDT blocks insanely large: %d",
3968 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3972 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3973 if (ext4_has_feature_inline_data(sb)) {
3974 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
3975 " that may contain inline data");
3978 if (!bdev_dax_supported(sb->s_bdev, blocksize)) {
3979 ext4_msg(sb, KERN_ERR,
3980 "DAX unsupported by block device.");
3985 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3986 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3987 es->s_encryption_level);
3991 if (sb->s_blocksize != blocksize) {
3992 /* Validate the filesystem blocksize */
3993 if (!sb_set_blocksize(sb, blocksize)) {
3994 ext4_msg(sb, KERN_ERR, "bad block size %d",
4000 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4001 offset = do_div(logical_sb_block, blocksize);
4002 bh = sb_bread_unmovable(sb, logical_sb_block);
4004 ext4_msg(sb, KERN_ERR,
4005 "Can't read superblock on 2nd try");
4008 es = (struct ext4_super_block *)(bh->b_data + offset);
4010 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
4011 ext4_msg(sb, KERN_ERR,
4012 "Magic mismatch, very weird!");
4017 has_huge_files = ext4_has_feature_huge_file(sb);
4018 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4020 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
4022 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4023 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4024 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4026 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4027 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4028 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4029 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4033 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4034 (!is_power_of_2(sbi->s_inode_size)) ||
4035 (sbi->s_inode_size > blocksize)) {
4036 ext4_msg(sb, KERN_ERR,
4037 "unsupported inode size: %d",
4041 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
4042 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
4045 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
4046 if (ext4_has_feature_64bit(sb)) {
4047 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
4048 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
4049 !is_power_of_2(sbi->s_desc_size)) {
4050 ext4_msg(sb, KERN_ERR,
4051 "unsupported descriptor size %lu",
4056 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
4058 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
4059 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
4061 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
4062 if (sbi->s_inodes_per_block == 0)
4064 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4065 sbi->s_inodes_per_group > blocksize * 8) {
4066 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
4067 sbi->s_blocks_per_group);
4070 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4071 sbi->s_inodes_per_block;
4072 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
4074 sbi->s_mount_state = le16_to_cpu(es->s_state);
4075 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4076 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
4078 for (i = 0; i < 4; i++)
4079 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4080 sbi->s_def_hash_version = es->s_def_hash_version;
4081 if (ext4_has_feature_dir_index(sb)) {
4082 i = le32_to_cpu(es->s_flags);
4083 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4084 sbi->s_hash_unsigned = 3;
4085 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
4086 #ifdef __CHAR_UNSIGNED__
4089 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4090 sbi->s_hash_unsigned = 3;
4094 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
4099 /* Handle clustersize */
4100 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4101 has_bigalloc = ext4_has_feature_bigalloc(sb);
4103 if (clustersize < blocksize) {
4104 ext4_msg(sb, KERN_ERR,
4105 "cluster size (%d) smaller than "
4106 "block size (%d)", clustersize, blocksize);
4109 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4110 le32_to_cpu(es->s_log_block_size);
4111 sbi->s_clusters_per_group =
4112 le32_to_cpu(es->s_clusters_per_group);
4113 if (sbi->s_clusters_per_group > blocksize * 8) {
4114 ext4_msg(sb, KERN_ERR,
4115 "#clusters per group too big: %lu",
4116 sbi->s_clusters_per_group);
4119 if (sbi->s_blocks_per_group !=
4120 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4121 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4122 "clusters per group (%lu) inconsistent",
4123 sbi->s_blocks_per_group,
4124 sbi->s_clusters_per_group);
4128 if (clustersize != blocksize) {
4129 ext4_msg(sb, KERN_ERR,
4130 "fragment/cluster size (%d) != "
4131 "block size (%d)", clustersize, blocksize);
4134 if (sbi->s_blocks_per_group > blocksize * 8) {
4135 ext4_msg(sb, KERN_ERR,
4136 "#blocks per group too big: %lu",
4137 sbi->s_blocks_per_group);
4140 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4141 sbi->s_cluster_bits = 0;
4143 sbi->s_cluster_ratio = clustersize / blocksize;
4145 /* Do we have standard group size of clustersize * 8 blocks ? */
4146 if (sbi->s_blocks_per_group == clustersize << 3)
4147 set_opt2(sb, STD_GROUP_SIZE);
4150 * Test whether we have more sectors than will fit in sector_t,
4151 * and whether the max offset is addressable by the page cache.
4153 err = generic_check_addressable(sb->s_blocksize_bits,
4154 ext4_blocks_count(es));
4156 ext4_msg(sb, KERN_ERR, "filesystem"
4157 " too large to mount safely on this system");
4161 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4164 /* check blocks count against device size */
4165 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4166 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4167 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4168 "exceeds size of device (%llu blocks)",
4169 ext4_blocks_count(es), blocks_count);
4174 * It makes no sense for the first data block to be beyond the end
4175 * of the filesystem.
4177 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4178 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4179 "block %u is beyond end of filesystem (%llu)",
4180 le32_to_cpu(es->s_first_data_block),
4181 ext4_blocks_count(es));
4184 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4185 (sbi->s_cluster_ratio == 1)) {
4186 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4187 "block is 0 with a 1k block and cluster size");
4191 blocks_count = (ext4_blocks_count(es) -
4192 le32_to_cpu(es->s_first_data_block) +
4193 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4194 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4195 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4196 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4197 "(block count %llu, first data block %u, "
4198 "blocks per group %lu)", sbi->s_groups_count,
4199 ext4_blocks_count(es),
4200 le32_to_cpu(es->s_first_data_block),
4201 EXT4_BLOCKS_PER_GROUP(sb));
4204 sbi->s_groups_count = blocks_count;
4205 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4206 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4207 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4208 le32_to_cpu(es->s_inodes_count)) {
4209 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4210 le32_to_cpu(es->s_inodes_count),
4211 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4215 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4216 EXT4_DESC_PER_BLOCK(sb);
4217 if (ext4_has_feature_meta_bg(sb)) {
4218 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4219 ext4_msg(sb, KERN_WARNING,
4220 "first meta block group too large: %u "
4221 "(group descriptor block count %u)",
4222 le32_to_cpu(es->s_first_meta_bg), db_count);
4226 sbi->s_group_desc = kvmalloc_array(db_count,
4227 sizeof(struct buffer_head *),
4229 if (sbi->s_group_desc == NULL) {
4230 ext4_msg(sb, KERN_ERR, "not enough memory");
4235 bgl_lock_init(sbi->s_blockgroup_lock);
4237 /* Pre-read the descriptors into the buffer cache */
4238 for (i = 0; i < db_count; i++) {
4239 block = descriptor_loc(sb, logical_sb_block, i);
4240 sb_breadahead(sb, block);
4243 for (i = 0; i < db_count; i++) {
4244 block = descriptor_loc(sb, logical_sb_block, i);
4245 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
4246 if (!sbi->s_group_desc[i]) {
4247 ext4_msg(sb, KERN_ERR,
4248 "can't read group descriptor %d", i);
4253 sbi->s_gdb_count = db_count;
4254 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4255 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4256 ret = -EFSCORRUPTED;
4260 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
4262 /* Register extent status tree shrinker */
4263 if (ext4_es_register_shrinker(sbi))
4266 sbi->s_stripe = ext4_get_stripe_size(sbi);
4267 sbi->s_extent_max_zeroout_kb = 32;
4270 * set up enough so that it can read an inode
4272 sb->s_op = &ext4_sops;
4273 sb->s_export_op = &ext4_export_ops;
4274 sb->s_xattr = ext4_xattr_handlers;
4275 #ifdef CONFIG_FS_ENCRYPTION
4276 sb->s_cop = &ext4_cryptops;
4279 sb->dq_op = &ext4_quota_operations;
4280 if (ext4_has_feature_quota(sb))
4281 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4283 sb->s_qcop = &ext4_qctl_operations;
4284 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4286 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4288 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4289 mutex_init(&sbi->s_orphan_lock);
4293 needs_recovery = (es->s_last_orphan != 0 ||
4294 ext4_has_feature_journal_needs_recovery(sb));
4296 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
4297 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4298 goto failed_mount3a;
4301 * The first inode we look at is the journal inode. Don't try
4302 * root first: it may be modified in the journal!
4304 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4305 err = ext4_load_journal(sb, es, journal_devnum);
4307 goto failed_mount3a;
4308 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4309 ext4_has_feature_journal_needs_recovery(sb)) {
4310 ext4_msg(sb, KERN_ERR, "required journal recovery "
4311 "suppressed and not mounted read-only");
4312 goto failed_mount_wq;
4314 /* Nojournal mode, all journal mount options are illegal */
4315 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4316 ext4_msg(sb, KERN_ERR, "can't mount with "
4317 "journal_checksum, fs mounted w/o journal");
4318 goto failed_mount_wq;
4320 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4321 ext4_msg(sb, KERN_ERR, "can't mount with "
4322 "journal_async_commit, fs mounted w/o journal");
4323 goto failed_mount_wq;
4325 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4326 ext4_msg(sb, KERN_ERR, "can't mount with "
4327 "commit=%lu, fs mounted w/o journal",
4328 sbi->s_commit_interval / HZ);
4329 goto failed_mount_wq;
4331 if (EXT4_MOUNT_DATA_FLAGS &
4332 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4333 ext4_msg(sb, KERN_ERR, "can't mount with "
4334 "data=, fs mounted w/o journal");
4335 goto failed_mount_wq;
4337 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
4338 clear_opt(sb, JOURNAL_CHECKSUM);
4339 clear_opt(sb, DATA_FLAGS);
4340 sbi->s_journal = NULL;
4345 if (ext4_has_feature_64bit(sb) &&
4346 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4347 JBD2_FEATURE_INCOMPAT_64BIT)) {
4348 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4349 goto failed_mount_wq;
4352 if (!set_journal_csum_feature_set(sb)) {
4353 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4355 goto failed_mount_wq;
4358 /* We have now updated the journal if required, so we can
4359 * validate the data journaling mode. */
4360 switch (test_opt(sb, DATA_FLAGS)) {
4362 /* No mode set, assume a default based on the journal
4363 * capabilities: ORDERED_DATA if the journal can
4364 * cope, else JOURNAL_DATA
4366 if (jbd2_journal_check_available_features
4367 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4368 set_opt(sb, ORDERED_DATA);
4369 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4371 set_opt(sb, JOURNAL_DATA);
4372 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4376 case EXT4_MOUNT_ORDERED_DATA:
4377 case EXT4_MOUNT_WRITEBACK_DATA:
4378 if (!jbd2_journal_check_available_features
4379 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4380 ext4_msg(sb, KERN_ERR, "Journal does not support "
4381 "requested data journaling mode");
4382 goto failed_mount_wq;
4388 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4389 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4390 ext4_msg(sb, KERN_ERR, "can't mount with "
4391 "journal_async_commit in data=ordered mode");
4392 goto failed_mount_wq;
4395 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4397 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4400 if (!test_opt(sb, NO_MBCACHE)) {
4401 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4402 if (!sbi->s_ea_block_cache) {
4403 ext4_msg(sb, KERN_ERR,
4404 "Failed to create ea_block_cache");
4405 goto failed_mount_wq;
4408 if (ext4_has_feature_ea_inode(sb)) {
4409 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4410 if (!sbi->s_ea_inode_cache) {
4411 ext4_msg(sb, KERN_ERR,
4412 "Failed to create ea_inode_cache");
4413 goto failed_mount_wq;
4418 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
4419 (blocksize != PAGE_SIZE)) {
4420 ext4_msg(sb, KERN_ERR,
4421 "Unsupported blocksize for fs encryption");
4422 goto failed_mount_wq;
4425 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
4426 !ext4_has_feature_encrypt(sb)) {
4427 ext4_set_feature_encrypt(sb);
4428 ext4_commit_super(sb, 1);
4432 * Get the # of file system overhead blocks from the
4433 * superblock if present.
4435 if (es->s_overhead_clusters)
4436 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4438 err = ext4_calculate_overhead(sb);
4440 goto failed_mount_wq;
4444 * The maximum number of concurrent works can be high and
4445 * concurrency isn't really necessary. Limit it to 1.
4447 EXT4_SB(sb)->rsv_conversion_wq =
4448 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4449 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4450 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4456 * The jbd2_journal_load will have done any necessary log recovery,
4457 * so we can safely mount the rest of the filesystem now.
4460 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
4462 ext4_msg(sb, KERN_ERR, "get root inode failed");
4463 ret = PTR_ERR(root);
4467 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4468 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4473 #ifdef CONFIG_UNICODE
4474 if (sbi->s_encoding)
4475 sb->s_d_op = &ext4_dentry_ops;
4478 sb->s_root = d_make_root(root);
4480 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4485 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4486 if (ret == -EROFS) {
4487 sb->s_flags |= SB_RDONLY;
4490 goto failed_mount4a;
4492 ext4_clamp_want_extra_isize(sb);
4494 ext4_set_resv_clusters(sb);
4496 err = ext4_setup_system_zone(sb);
4498 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4500 goto failed_mount4a;
4504 err = ext4_mb_init(sb);
4506 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4511 block = ext4_count_free_clusters(sb);
4512 ext4_free_blocks_count_set(sbi->s_es,
4513 EXT4_C2B(sbi, block));
4514 ext4_superblock_csum_set(sb);
4515 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4518 unsigned long freei = ext4_count_free_inodes(sb);
4519 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4520 ext4_superblock_csum_set(sb);
4521 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4525 err = percpu_counter_init(&sbi->s_dirs_counter,
4526 ext4_count_dirs(sb), GFP_KERNEL);
4528 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4531 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4534 ext4_msg(sb, KERN_ERR, "insufficient memory");
4538 if (ext4_has_feature_flex_bg(sb))
4539 if (!ext4_fill_flex_info(sb)) {
4540 ext4_msg(sb, KERN_ERR,
4541 "unable to initialize "
4542 "flex_bg meta info!");
4546 err = ext4_register_li_request(sb, first_not_zeroed);
4550 err = ext4_register_sysfs(sb);
4555 /* Enable quota usage during mount. */
4556 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
4557 err = ext4_enable_quotas(sb);
4561 #endif /* CONFIG_QUOTA */
4563 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4564 ext4_orphan_cleanup(sb, es);
4565 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4566 if (needs_recovery) {
4567 ext4_msg(sb, KERN_INFO, "recovery complete");
4568 ext4_mark_recovery_complete(sb, es);
4570 if (EXT4_SB(sb)->s_journal) {
4571 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4572 descr = " journalled data mode";
4573 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4574 descr = " ordered data mode";
4576 descr = " writeback data mode";
4578 descr = "out journal";
4580 if (test_opt(sb, DISCARD)) {
4581 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4582 if (!blk_queue_discard(q))
4583 ext4_msg(sb, KERN_WARNING,
4584 "mounting with \"discard\" option, but "
4585 "the device does not support discard");
4588 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4589 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4590 "Opts: %.*s%s%s", descr,
4591 (int) sizeof(sbi->s_es->s_mount_opts),
4592 sbi->s_es->s_mount_opts,
4593 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4595 if (es->s_error_count)
4596 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4598 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4599 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4600 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4601 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4608 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4613 ext4_unregister_sysfs(sb);
4616 ext4_unregister_li_request(sb);
4618 ext4_mb_release(sb);
4619 if (sbi->s_flex_groups)
4620 kvfree(sbi->s_flex_groups);
4621 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4622 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4623 percpu_counter_destroy(&sbi->s_dirs_counter);
4624 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4625 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
4627 ext4_ext_release(sb);
4628 ext4_release_system_zone(sb);
4633 ext4_msg(sb, KERN_ERR, "mount failed");
4634 if (EXT4_SB(sb)->rsv_conversion_wq)
4635 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4637 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4638 sbi->s_ea_inode_cache = NULL;
4640 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4641 sbi->s_ea_block_cache = NULL;
4643 if (sbi->s_journal) {
4644 jbd2_journal_destroy(sbi->s_journal);
4645 sbi->s_journal = NULL;
4648 ext4_es_unregister_shrinker(sbi);
4650 del_timer_sync(&sbi->s_err_report);
4652 kthread_stop(sbi->s_mmp_tsk);
4654 for (i = 0; i < db_count; i++)
4655 brelse(sbi->s_group_desc[i]);
4656 kvfree(sbi->s_group_desc);
4658 if (sbi->s_chksum_driver)
4659 crypto_free_shash(sbi->s_chksum_driver);
4661 #ifdef CONFIG_UNICODE
4662 utf8_unload(sbi->s_encoding);
4666 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4667 kfree(get_qf_name(sb, sbi, i));
4669 ext4_blkdev_remove(sbi);
4672 sb->s_fs_info = NULL;
4673 kfree(sbi->s_blockgroup_lock);
4677 fs_put_dax(dax_dev);
4678 return err ? err : ret;
4682 * Setup any per-fs journal parameters now. We'll do this both on
4683 * initial mount, once the journal has been initialised but before we've
4684 * done any recovery; and again on any subsequent remount.
4686 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4688 struct ext4_sb_info *sbi = EXT4_SB(sb);
4690 journal->j_commit_interval = sbi->s_commit_interval;
4691 journal->j_min_batch_time = sbi->s_min_batch_time;
4692 journal->j_max_batch_time = sbi->s_max_batch_time;
4694 write_lock(&journal->j_state_lock);
4695 if (test_opt(sb, BARRIER))
4696 journal->j_flags |= JBD2_BARRIER;
4698 journal->j_flags &= ~JBD2_BARRIER;
4699 if (test_opt(sb, DATA_ERR_ABORT))
4700 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4702 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4703 write_unlock(&journal->j_state_lock);
4706 static struct inode *ext4_get_journal_inode(struct super_block *sb,
4707 unsigned int journal_inum)
4709 struct inode *journal_inode;
4712 * Test for the existence of a valid inode on disk. Bad things
4713 * happen if we iget() an unused inode, as the subsequent iput()
4714 * will try to delete it.
4716 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
4717 if (IS_ERR(journal_inode)) {
4718 ext4_msg(sb, KERN_ERR, "no journal found");
4721 if (!journal_inode->i_nlink) {
4722 make_bad_inode(journal_inode);
4723 iput(journal_inode);
4724 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4728 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4729 journal_inode, journal_inode->i_size);
4730 if (!S_ISREG(journal_inode->i_mode)) {
4731 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4732 iput(journal_inode);
4735 return journal_inode;
4738 static journal_t *ext4_get_journal(struct super_block *sb,
4739 unsigned int journal_inum)
4741 struct inode *journal_inode;
4744 BUG_ON(!ext4_has_feature_journal(sb));
4746 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4750 journal = jbd2_journal_init_inode(journal_inode);
4752 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4753 iput(journal_inode);
4756 journal->j_private = sb;
4757 ext4_init_journal_params(sb, journal);
4761 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4764 struct buffer_head *bh;
4768 int hblock, blocksize;
4769 ext4_fsblk_t sb_block;
4770 unsigned long offset;
4771 struct ext4_super_block *es;
4772 struct block_device *bdev;
4774 BUG_ON(!ext4_has_feature_journal(sb));
4776 bdev = ext4_blkdev_get(j_dev, sb);
4780 blocksize = sb->s_blocksize;
4781 hblock = bdev_logical_block_size(bdev);
4782 if (blocksize < hblock) {
4783 ext4_msg(sb, KERN_ERR,
4784 "blocksize too small for journal device");
4788 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4789 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4790 set_blocksize(bdev, blocksize);
4791 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4792 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4793 "external journal");
4797 es = (struct ext4_super_block *) (bh->b_data + offset);
4798 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4799 !(le32_to_cpu(es->s_feature_incompat) &
4800 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4801 ext4_msg(sb, KERN_ERR, "external journal has "
4807 if ((le32_to_cpu(es->s_feature_ro_compat) &
4808 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4809 es->s_checksum != ext4_superblock_csum(sb, es)) {
4810 ext4_msg(sb, KERN_ERR, "external journal has "
4811 "corrupt superblock");
4816 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4817 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4822 len = ext4_blocks_count(es);
4823 start = sb_block + 1;
4824 brelse(bh); /* we're done with the superblock */
4826 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4827 start, len, blocksize);
4829 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4832 journal->j_private = sb;
4833 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4834 wait_on_buffer(journal->j_sb_buffer);
4835 if (!buffer_uptodate(journal->j_sb_buffer)) {
4836 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4839 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4840 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4841 "user (unsupported) - %d",
4842 be32_to_cpu(journal->j_superblock->s_nr_users));
4845 EXT4_SB(sb)->journal_bdev = bdev;
4846 ext4_init_journal_params(sb, journal);
4850 jbd2_journal_destroy(journal);
4852 ext4_blkdev_put(bdev);
4856 static int ext4_load_journal(struct super_block *sb,
4857 struct ext4_super_block *es,
4858 unsigned long journal_devnum)
4861 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4864 int really_read_only;
4866 BUG_ON(!ext4_has_feature_journal(sb));
4868 if (journal_devnum &&
4869 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4870 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4871 "numbers have changed");
4872 journal_dev = new_decode_dev(journal_devnum);
4874 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4876 really_read_only = bdev_read_only(sb->s_bdev);
4879 * Are we loading a blank journal or performing recovery after a
4880 * crash? For recovery, we need to check in advance whether we
4881 * can get read-write access to the device.
4883 if (ext4_has_feature_journal_needs_recovery(sb)) {
4884 if (sb_rdonly(sb)) {
4885 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4886 "required on readonly filesystem");
4887 if (really_read_only) {
4888 ext4_msg(sb, KERN_ERR, "write access "
4889 "unavailable, cannot proceed "
4890 "(try mounting with noload)");
4893 ext4_msg(sb, KERN_INFO, "write access will "
4894 "be enabled during recovery");
4898 if (journal_inum && journal_dev) {
4899 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4900 "and inode journals!");
4905 if (!(journal = ext4_get_journal(sb, journal_inum)))
4908 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4912 if (!(journal->j_flags & JBD2_BARRIER))
4913 ext4_msg(sb, KERN_INFO, "barriers disabled");
4915 if (!ext4_has_feature_journal_needs_recovery(sb))
4916 err = jbd2_journal_wipe(journal, !really_read_only);
4918 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4920 memcpy(save, ((char *) es) +
4921 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4922 err = jbd2_journal_load(journal);
4924 memcpy(((char *) es) + EXT4_S_ERR_START,
4925 save, EXT4_S_ERR_LEN);
4930 ext4_msg(sb, KERN_ERR, "error loading journal");
4931 jbd2_journal_destroy(journal);
4935 EXT4_SB(sb)->s_journal = journal;
4936 ext4_clear_journal_err(sb, es);
4938 if (!really_read_only && journal_devnum &&
4939 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4940 es->s_journal_dev = cpu_to_le32(journal_devnum);
4942 /* Make sure we flush the recovery flag to disk. */
4943 ext4_commit_super(sb, 1);
4949 static int ext4_commit_super(struct super_block *sb, int sync)
4951 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4952 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4955 if (!sbh || block_device_ejected(sb))
4959 * The superblock bh should be mapped, but it might not be if the
4960 * device was hot-removed. Not much we can do but fail the I/O.
4962 if (!buffer_mapped(sbh))
4966 * If the file system is mounted read-only, don't update the
4967 * superblock write time. This avoids updating the superblock
4968 * write time when we are mounting the root file system
4969 * read/only but we need to replay the journal; at that point,
4970 * for people who are east of GMT and who make their clock
4971 * tick in localtime for Windows bug-for-bug compatibility,
4972 * the clock is set in the future, and this will cause e2fsck
4973 * to complain and force a full file system check.
4975 if (!(sb->s_flags & SB_RDONLY))
4976 ext4_update_tstamp(es, s_wtime);
4977 if (sb->s_bdev->bd_part)
4978 es->s_kbytes_written =
4979 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4980 ((part_stat_read(sb->s_bdev->bd_part,
4981 sectors[STAT_WRITE]) -
4982 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4984 es->s_kbytes_written =
4985 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4986 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4987 ext4_free_blocks_count_set(es,
4988 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4989 &EXT4_SB(sb)->s_freeclusters_counter)));
4990 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4991 es->s_free_inodes_count =
4992 cpu_to_le32(percpu_counter_sum_positive(
4993 &EXT4_SB(sb)->s_freeinodes_counter));
4994 BUFFER_TRACE(sbh, "marking dirty");
4995 ext4_superblock_csum_set(sb);
4998 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
5000 * Oh, dear. A previous attempt to write the
5001 * superblock failed. This could happen because the
5002 * USB device was yanked out. Or it could happen to
5003 * be a transient write error and maybe the block will
5004 * be remapped. Nothing we can do but to retry the
5005 * write and hope for the best.
5007 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5008 "superblock detected");
5009 clear_buffer_write_io_error(sbh);
5010 set_buffer_uptodate(sbh);
5012 mark_buffer_dirty(sbh);
5015 error = __sync_dirty_buffer(sbh,
5016 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
5017 if (buffer_write_io_error(sbh)) {
5018 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5020 clear_buffer_write_io_error(sbh);
5021 set_buffer_uptodate(sbh);
5028 * Have we just finished recovery? If so, and if we are mounting (or
5029 * remounting) the filesystem readonly, then we will end up with a
5030 * consistent fs on disk. Record that fact.
5032 static void ext4_mark_recovery_complete(struct super_block *sb,
5033 struct ext4_super_block *es)
5035 journal_t *journal = EXT4_SB(sb)->s_journal;
5037 if (!ext4_has_feature_journal(sb)) {
5038 BUG_ON(journal != NULL);
5041 jbd2_journal_lock_updates(journal);
5042 if (jbd2_journal_flush(journal) < 0)
5045 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
5046 ext4_clear_feature_journal_needs_recovery(sb);
5047 ext4_commit_super(sb, 1);
5051 jbd2_journal_unlock_updates(journal);
5055 * If we are mounting (or read-write remounting) a filesystem whose journal
5056 * has recorded an error from a previous lifetime, move that error to the
5057 * main filesystem now.
5059 static void ext4_clear_journal_err(struct super_block *sb,
5060 struct ext4_super_block *es)
5066 BUG_ON(!ext4_has_feature_journal(sb));
5068 journal = EXT4_SB(sb)->s_journal;
5071 * Now check for any error status which may have been recorded in the
5072 * journal by a prior ext4_error() or ext4_abort()
5075 j_errno = jbd2_journal_errno(journal);
5079 errstr = ext4_decode_error(sb, j_errno, nbuf);
5080 ext4_warning(sb, "Filesystem error recorded "
5081 "from previous mount: %s", errstr);
5082 ext4_warning(sb, "Marking fs in need of filesystem check.");
5084 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5085 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5086 ext4_commit_super(sb, 1);
5088 jbd2_journal_clear_err(journal);
5089 jbd2_journal_update_sb_errno(journal);
5094 * Force the running and committing transactions to commit,
5095 * and wait on the commit.
5097 int ext4_force_commit(struct super_block *sb)
5104 journal = EXT4_SB(sb)->s_journal;
5105 return ext4_journal_force_commit(journal);
5108 static int ext4_sync_fs(struct super_block *sb, int wait)
5112 bool needs_barrier = false;
5113 struct ext4_sb_info *sbi = EXT4_SB(sb);
5115 if (unlikely(ext4_forced_shutdown(sbi)))
5118 trace_ext4_sync_fs(sb, wait);
5119 flush_workqueue(sbi->rsv_conversion_wq);
5121 * Writeback quota in non-journalled quota case - journalled quota has
5124 dquot_writeback_dquots(sb, -1);
5126 * Data writeback is possible w/o journal transaction, so barrier must
5127 * being sent at the end of the function. But we can skip it if
5128 * transaction_commit will do it for us.
5130 if (sbi->s_journal) {
5131 target = jbd2_get_latest_transaction(sbi->s_journal);
5132 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5133 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5134 needs_barrier = true;
5136 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5138 ret = jbd2_log_wait_commit(sbi->s_journal,
5141 } else if (wait && test_opt(sb, BARRIER))
5142 needs_barrier = true;
5143 if (needs_barrier) {
5145 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
5154 * LVM calls this function before a (read-only) snapshot is created. This
5155 * gives us a chance to flush the journal completely and mark the fs clean.
5157 * Note that only this function cannot bring a filesystem to be in a clean
5158 * state independently. It relies on upper layer to stop all data & metadata
5161 static int ext4_freeze(struct super_block *sb)
5169 journal = EXT4_SB(sb)->s_journal;
5172 /* Now we set up the journal barrier. */
5173 jbd2_journal_lock_updates(journal);
5176 * Don't clear the needs_recovery flag if we failed to
5177 * flush the journal.
5179 error = jbd2_journal_flush(journal);
5183 /* Journal blocked and flushed, clear needs_recovery flag. */
5184 ext4_clear_feature_journal_needs_recovery(sb);
5187 error = ext4_commit_super(sb, 1);
5190 /* we rely on upper layer to stop further updates */
5191 jbd2_journal_unlock_updates(journal);
5196 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5197 * flag here, even though the filesystem is not technically dirty yet.
5199 static int ext4_unfreeze(struct super_block *sb)
5201 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
5204 if (EXT4_SB(sb)->s_journal) {
5205 /* Reset the needs_recovery flag before the fs is unlocked. */
5206 ext4_set_feature_journal_needs_recovery(sb);
5209 ext4_commit_super(sb, 1);
5214 * Structure to save mount options for ext4_remount's benefit
5216 struct ext4_mount_options {
5217 unsigned long s_mount_opt;
5218 unsigned long s_mount_opt2;
5221 unsigned long s_commit_interval;
5222 u32 s_min_batch_time, s_max_batch_time;
5225 char *s_qf_names[EXT4_MAXQUOTAS];
5229 static int ext4_remount(struct super_block *sb, int *flags, char *data)
5231 struct ext4_super_block *es;
5232 struct ext4_sb_info *sbi = EXT4_SB(sb);
5233 unsigned long old_sb_flags;
5234 struct ext4_mount_options old_opts;
5235 int enable_quota = 0;
5237 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5241 char *to_free[EXT4_MAXQUOTAS];
5243 char *orig_data = kstrdup(data, GFP_KERNEL);
5245 if (data && !orig_data)
5248 /* Store the original options */
5249 old_sb_flags = sb->s_flags;
5250 old_opts.s_mount_opt = sbi->s_mount_opt;
5251 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
5252 old_opts.s_resuid = sbi->s_resuid;
5253 old_opts.s_resgid = sbi->s_resgid;
5254 old_opts.s_commit_interval = sbi->s_commit_interval;
5255 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5256 old_opts.s_max_batch_time = sbi->s_max_batch_time;
5258 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
5259 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5260 if (sbi->s_qf_names[i]) {
5261 char *qf_name = get_qf_name(sb, sbi, i);
5263 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
5264 if (!old_opts.s_qf_names[i]) {
5265 for (j = 0; j < i; j++)
5266 kfree(old_opts.s_qf_names[j]);
5271 old_opts.s_qf_names[i] = NULL;
5273 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5274 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
5276 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
5281 ext4_clamp_want_extra_isize(sb);
5283 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5284 test_opt(sb, JOURNAL_CHECKSUM)) {
5285 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5286 "during remount not supported; ignoring");
5287 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5290 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5291 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5292 ext4_msg(sb, KERN_ERR, "can't mount with "
5293 "both data=journal and delalloc");
5297 if (test_opt(sb, DIOREAD_NOLOCK)) {
5298 ext4_msg(sb, KERN_ERR, "can't mount with "
5299 "both data=journal and dioread_nolock");
5303 if (test_opt(sb, DAX)) {
5304 ext4_msg(sb, KERN_ERR, "can't mount with "
5305 "both data=journal and dax");
5309 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5310 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5311 ext4_msg(sb, KERN_ERR, "can't mount with "
5312 "journal_async_commit in data=ordered mode");
5318 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5319 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5324 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
5325 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
5326 "dax flag with busy inodes while remounting");
5327 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
5330 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
5331 ext4_abort(sb, "Abort forced by user");
5333 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5334 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5338 if (sbi->s_journal) {
5339 ext4_init_journal_params(sb, sbi->s_journal);
5340 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5343 if (*flags & SB_LAZYTIME)
5344 sb->s_flags |= SB_LAZYTIME;
5346 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
5347 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
5352 if (*flags & SB_RDONLY) {
5353 err = sync_filesystem(sb);
5356 err = dquot_suspend(sb, -1);
5361 * First of all, the unconditional stuff we have to do
5362 * to disable replay of the journal when we next remount
5364 sb->s_flags |= SB_RDONLY;
5367 * OK, test if we are remounting a valid rw partition
5368 * readonly, and if so set the rdonly flag and then
5369 * mark the partition as valid again.
5371 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5372 (sbi->s_mount_state & EXT4_VALID_FS))
5373 es->s_state = cpu_to_le16(sbi->s_mount_state);
5376 ext4_mark_recovery_complete(sb, es);
5378 kthread_stop(sbi->s_mmp_tsk);
5380 /* Make sure we can mount this feature set readwrite */
5381 if (ext4_has_feature_readonly(sb) ||
5382 !ext4_feature_set_ok(sb, 0)) {
5387 * Make sure the group descriptor checksums
5388 * are sane. If they aren't, refuse to remount r/w.
5390 for (g = 0; g < sbi->s_groups_count; g++) {
5391 struct ext4_group_desc *gdp =
5392 ext4_get_group_desc(sb, g, NULL);
5394 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5395 ext4_msg(sb, KERN_ERR,
5396 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5397 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5398 le16_to_cpu(gdp->bg_checksum));
5405 * If we have an unprocessed orphan list hanging
5406 * around from a previously readonly bdev mount,
5407 * require a full umount/remount for now.
5409 if (es->s_last_orphan) {
5410 ext4_msg(sb, KERN_WARNING, "Couldn't "
5411 "remount RDWR because of unprocessed "
5412 "orphan inode list. Please "
5413 "umount/remount instead");
5419 * Mounting a RDONLY partition read-write, so reread
5420 * and store the current valid flag. (It may have
5421 * been changed by e2fsck since we originally mounted
5425 ext4_clear_journal_err(sb, es);
5426 sbi->s_mount_state = le16_to_cpu(es->s_state);
5428 err = ext4_setup_super(sb, es, 0);
5432 sb->s_flags &= ~SB_RDONLY;
5433 if (ext4_has_feature_mmp(sb))
5434 if (ext4_multi_mount_protect(sb,
5435 le64_to_cpu(es->s_mmp_block))) {
5444 * Reinitialize lazy itable initialization thread based on
5447 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
5448 ext4_unregister_li_request(sb);
5450 ext4_group_t first_not_zeroed;
5451 first_not_zeroed = ext4_has_uninit_itable(sb);
5452 ext4_register_li_request(sb, first_not_zeroed);
5455 ext4_setup_system_zone(sb);
5456 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5457 err = ext4_commit_super(sb, 1);
5463 /* Release old quota file names */
5464 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5465 kfree(old_opts.s_qf_names[i]);
5467 if (sb_any_quota_suspended(sb))
5468 dquot_resume(sb, -1);
5469 else if (ext4_has_feature_quota(sb)) {
5470 err = ext4_enable_quotas(sb);
5477 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
5478 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5483 sb->s_flags = old_sb_flags;
5484 sbi->s_mount_opt = old_opts.s_mount_opt;
5485 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5486 sbi->s_resuid = old_opts.s_resuid;
5487 sbi->s_resgid = old_opts.s_resgid;
5488 sbi->s_commit_interval = old_opts.s_commit_interval;
5489 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5490 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5492 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5493 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5494 to_free[i] = get_qf_name(sb, sbi, i);
5495 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
5498 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5506 static int ext4_statfs_project(struct super_block *sb,
5507 kprojid_t projid, struct kstatfs *buf)
5510 struct dquot *dquot;
5514 qid = make_kqid_projid(projid);
5515 dquot = dqget(sb, qid);
5517 return PTR_ERR(dquot);
5518 spin_lock(&dquot->dq_dqb_lock);
5520 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5521 dquot->dq_dqb.dqb_bsoftlimit :
5522 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5523 if (limit && buf->f_blocks > limit) {
5524 curblock = (dquot->dq_dqb.dqb_curspace +
5525 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
5526 buf->f_blocks = limit;
5527 buf->f_bfree = buf->f_bavail =
5528 (buf->f_blocks > curblock) ?
5529 (buf->f_blocks - curblock) : 0;
5532 limit = dquot->dq_dqb.dqb_isoftlimit ?
5533 dquot->dq_dqb.dqb_isoftlimit :
5534 dquot->dq_dqb.dqb_ihardlimit;
5535 if (limit && buf->f_files > limit) {
5536 buf->f_files = limit;
5538 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5539 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5542 spin_unlock(&dquot->dq_dqb_lock);
5548 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5550 struct super_block *sb = dentry->d_sb;
5551 struct ext4_sb_info *sbi = EXT4_SB(sb);
5552 struct ext4_super_block *es = sbi->s_es;
5553 ext4_fsblk_t overhead = 0, resv_blocks;
5556 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5558 if (!test_opt(sb, MINIX_DF))
5559 overhead = sbi->s_overhead;
5561 buf->f_type = EXT4_SUPER_MAGIC;
5562 buf->f_bsize = sb->s_blocksize;
5563 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5564 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5565 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5566 /* prevent underflow in case that few free space is available */
5567 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5568 buf->f_bavail = buf->f_bfree -
5569 (ext4_r_blocks_count(es) + resv_blocks);
5570 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5572 buf->f_files = le32_to_cpu(es->s_inodes_count);
5573 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5574 buf->f_namelen = EXT4_NAME_LEN;
5575 fsid = le64_to_cpup((void *)es->s_uuid) ^
5576 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5577 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5578 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5581 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5582 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5583 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5592 * Helper functions so that transaction is started before we acquire dqio_sem
5593 * to keep correct lock ordering of transaction > dqio_sem
5595 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5597 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5600 static int ext4_write_dquot(struct dquot *dquot)
5604 struct inode *inode;
5606 inode = dquot_to_inode(dquot);
5607 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5608 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5610 return PTR_ERR(handle);
5611 ret = dquot_commit(dquot);
5612 err = ext4_journal_stop(handle);
5618 static int ext4_acquire_dquot(struct dquot *dquot)
5623 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5624 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5626 return PTR_ERR(handle);
5627 ret = dquot_acquire(dquot);
5628 err = ext4_journal_stop(handle);
5634 static int ext4_release_dquot(struct dquot *dquot)
5639 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5640 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5641 if (IS_ERR(handle)) {
5642 /* Release dquot anyway to avoid endless cycle in dqput() */
5643 dquot_release(dquot);
5644 return PTR_ERR(handle);
5646 ret = dquot_release(dquot);
5647 err = ext4_journal_stop(handle);
5653 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5655 struct super_block *sb = dquot->dq_sb;
5656 struct ext4_sb_info *sbi = EXT4_SB(sb);
5658 /* Are we journaling quotas? */
5659 if (ext4_has_feature_quota(sb) ||
5660 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5661 dquot_mark_dquot_dirty(dquot);
5662 return ext4_write_dquot(dquot);
5664 return dquot_mark_dquot_dirty(dquot);
5668 static int ext4_write_info(struct super_block *sb, int type)
5673 /* Data block + inode block */
5674 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5676 return PTR_ERR(handle);
5677 ret = dquot_commit_info(sb, type);
5678 err = ext4_journal_stop(handle);
5685 * Turn on quotas during mount time - we need to find
5686 * the quota file and such...
5688 static int ext4_quota_on_mount(struct super_block *sb, int type)
5690 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
5691 EXT4_SB(sb)->s_jquota_fmt, type);
5694 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5696 struct ext4_inode_info *ei = EXT4_I(inode);
5698 /* The first argument of lockdep_set_subclass has to be
5699 * *exactly* the same as the argument to init_rwsem() --- in
5700 * this case, in init_once() --- or lockdep gets unhappy
5701 * because the name of the lock is set using the
5702 * stringification of the argument to init_rwsem().
5704 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5705 lockdep_set_subclass(&ei->i_data_sem, subclass);
5709 * Standard function to be called on quota_on
5711 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5712 const struct path *path)
5716 if (!test_opt(sb, QUOTA))
5719 /* Quotafile not on the same filesystem? */
5720 if (path->dentry->d_sb != sb)
5722 /* Journaling quota? */
5723 if (EXT4_SB(sb)->s_qf_names[type]) {
5724 /* Quotafile not in fs root? */
5725 if (path->dentry->d_parent != sb->s_root)
5726 ext4_msg(sb, KERN_WARNING,
5727 "Quota file not on filesystem root. "
5728 "Journaled quota will not work");
5729 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
5732 * Clear the flag just in case mount options changed since
5735 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
5739 * When we journal data on quota file, we have to flush journal to see
5740 * all updates to the file when we bypass pagecache...
5742 if (EXT4_SB(sb)->s_journal &&
5743 ext4_should_journal_data(d_inode(path->dentry))) {
5745 * We don't need to lock updates but journal_flush() could
5746 * otherwise be livelocked...
5748 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5749 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5750 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5755 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5756 err = dquot_quota_on(sb, type, format_id, path);
5758 lockdep_set_quota_inode(path->dentry->d_inode,
5761 struct inode *inode = d_inode(path->dentry);
5765 * Set inode flags to prevent userspace from messing with quota
5766 * files. If this fails, we return success anyway since quotas
5767 * are already enabled and this is not a hard failure.
5770 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5773 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5774 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5775 S_NOATIME | S_IMMUTABLE);
5776 ext4_mark_inode_dirty(handle, inode);
5777 ext4_journal_stop(handle);
5779 inode_unlock(inode);
5784 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5788 struct inode *qf_inode;
5789 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5790 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5791 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5792 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5795 BUG_ON(!ext4_has_feature_quota(sb));
5797 if (!qf_inums[type])
5800 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
5801 if (IS_ERR(qf_inode)) {
5802 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5803 return PTR_ERR(qf_inode);
5806 /* Don't account quota for quota files to avoid recursion */
5807 qf_inode->i_flags |= S_NOQUOTA;
5808 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5809 err = dquot_enable(qf_inode, type, format_id, flags);
5811 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5817 /* Enable usage tracking for all quota types. */
5818 static int ext4_enable_quotas(struct super_block *sb)
5821 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5822 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5823 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5824 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5826 bool quota_mopt[EXT4_MAXQUOTAS] = {
5827 test_opt(sb, USRQUOTA),
5828 test_opt(sb, GRPQUOTA),
5829 test_opt(sb, PRJQUOTA),
5832 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
5833 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5834 if (qf_inums[type]) {
5835 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5836 DQUOT_USAGE_ENABLED |
5837 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
5840 "Failed to enable quota tracking "
5841 "(type=%d, err=%d). Please run "
5842 "e2fsck to fix.", type, err);
5843 for (type--; type >= 0; type--)
5844 dquot_quota_off(sb, type);
5853 static int ext4_quota_off(struct super_block *sb, int type)
5855 struct inode *inode = sb_dqopt(sb)->files[type];
5859 /* Force all delayed allocation blocks to be allocated.
5860 * Caller already holds s_umount sem */
5861 if (test_opt(sb, DELALLOC))
5862 sync_filesystem(sb);
5864 if (!inode || !igrab(inode))
5867 err = dquot_quota_off(sb, type);
5868 if (err || ext4_has_feature_quota(sb))
5873 * Update modification times of quota files when userspace can
5874 * start looking at them. If we fail, we return success anyway since
5875 * this is not a hard failure and quotas are already disabled.
5877 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5880 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5881 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
5882 inode->i_mtime = inode->i_ctime = current_time(inode);
5883 ext4_mark_inode_dirty(handle, inode);
5884 ext4_journal_stop(handle);
5886 inode_unlock(inode);
5888 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
5892 return dquot_quota_off(sb, type);
5895 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5896 * acquiring the locks... As quota files are never truncated and quota code
5897 * itself serializes the operations (and no one else should touch the files)
5898 * we don't have to be afraid of races */
5899 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5900 size_t len, loff_t off)
5902 struct inode *inode = sb_dqopt(sb)->files[type];
5903 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5904 int offset = off & (sb->s_blocksize - 1);
5907 struct buffer_head *bh;
5908 loff_t i_size = i_size_read(inode);
5912 if (off+len > i_size)
5915 while (toread > 0) {
5916 tocopy = sb->s_blocksize - offset < toread ?
5917 sb->s_blocksize - offset : toread;
5918 bh = ext4_bread(NULL, inode, blk, 0);
5921 if (!bh) /* A hole? */
5922 memset(data, 0, tocopy);
5924 memcpy(data, bh->b_data+offset, tocopy);
5934 /* Write to quotafile (we know the transaction is already started and has
5935 * enough credits) */
5936 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5937 const char *data, size_t len, loff_t off)
5939 struct inode *inode = sb_dqopt(sb)->files[type];
5940 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5941 int err, offset = off & (sb->s_blocksize - 1);
5943 struct buffer_head *bh;
5944 handle_t *handle = journal_current_handle();
5946 if (EXT4_SB(sb)->s_journal && !handle) {
5947 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5948 " cancelled because transaction is not started",
5949 (unsigned long long)off, (unsigned long long)len);
5953 * Since we account only one data block in transaction credits,
5954 * then it is impossible to cross a block boundary.
5956 if (sb->s_blocksize - offset < len) {
5957 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5958 " cancelled because not block aligned",
5959 (unsigned long long)off, (unsigned long long)len);
5964 bh = ext4_bread(handle, inode, blk,
5965 EXT4_GET_BLOCKS_CREATE |
5966 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5967 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5968 ext4_should_retry_alloc(inode->i_sb, &retries));
5973 BUFFER_TRACE(bh, "get write access");
5974 err = ext4_journal_get_write_access(handle, bh);
5980 memcpy(bh->b_data+offset, data, len);
5981 flush_dcache_page(bh->b_page);
5983 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5986 if (inode->i_size < off + len) {
5987 i_size_write(inode, off + len);
5988 EXT4_I(inode)->i_disksize = inode->i_size;
5989 ext4_mark_inode_dirty(handle, inode);
5994 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5996 const struct quota_format_ops *ops;
5998 if (!sb_has_quota_loaded(sb, qid->type))
6000 ops = sb_dqopt(sb)->ops[qid->type];
6001 if (!ops || !ops->get_next_id)
6003 return dquot_get_next_id(sb, qid);
6007 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6008 const char *dev_name, void *data)
6010 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
6013 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
6014 static inline void register_as_ext2(void)
6016 int err = register_filesystem(&ext2_fs_type);
6019 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6022 static inline void unregister_as_ext2(void)
6024 unregister_filesystem(&ext2_fs_type);
6027 static inline int ext2_feature_set_ok(struct super_block *sb)
6029 if (ext4_has_unknown_ext2_incompat_features(sb))
6033 if (ext4_has_unknown_ext2_ro_compat_features(sb))
6038 static inline void register_as_ext2(void) { }
6039 static inline void unregister_as_ext2(void) { }
6040 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
6043 static inline void register_as_ext3(void)
6045 int err = register_filesystem(&ext3_fs_type);
6048 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6051 static inline void unregister_as_ext3(void)
6053 unregister_filesystem(&ext3_fs_type);
6056 static inline int ext3_feature_set_ok(struct super_block *sb)
6058 if (ext4_has_unknown_ext3_incompat_features(sb))
6060 if (!ext4_has_feature_journal(sb))
6064 if (ext4_has_unknown_ext3_ro_compat_features(sb))
6069 static struct file_system_type ext4_fs_type = {
6070 .owner = THIS_MODULE,
6072 .mount = ext4_mount,
6073 .kill_sb = kill_block_super,
6074 .fs_flags = FS_REQUIRES_DEV,
6076 MODULE_ALIAS_FS("ext4");
6078 /* Shared across all ext4 file systems */
6079 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
6081 static int __init ext4_init_fs(void)
6085 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
6086 ext4_li_info = NULL;
6087 mutex_init(&ext4_li_mtx);
6089 /* Build-time check for flags consistency */
6090 ext4_check_flag_values();
6092 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
6093 init_waitqueue_head(&ext4__ioend_wq[i]);
6095 err = ext4_init_es();
6099 err = ext4_init_pending();
6103 err = ext4_init_pageio();
6107 err = ext4_init_system_zone();
6111 err = ext4_init_sysfs();
6115 err = ext4_init_mballoc();
6118 err = init_inodecache();
6123 err = register_filesystem(&ext4_fs_type);
6129 unregister_as_ext2();
6130 unregister_as_ext3();
6131 destroy_inodecache();
6133 ext4_exit_mballoc();
6137 ext4_exit_system_zone();
6141 ext4_exit_pending();
6148 static void __exit ext4_exit_fs(void)
6150 ext4_destroy_lazyinit_thread();
6151 unregister_as_ext2();
6152 unregister_as_ext3();
6153 unregister_filesystem(&ext4_fs_type);
6154 destroy_inodecache();
6155 ext4_exit_mballoc();
6157 ext4_exit_system_zone();
6160 ext4_exit_pending();
6163 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6164 MODULE_DESCRIPTION("Fourth Extended Filesystem");
6165 MODULE_LICENSE("GPL");
6166 MODULE_SOFTDEP("pre: crc32c");
6167 module_init(ext4_init_fs)
6168 module_exit(ext4_exit_fs)