2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/backing-dev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/ctype.h>
38 #include <linux/log2.h>
39 #include <linux/crc16.h>
40 #include <linux/dax.h>
41 #include <linux/cleancache.h>
42 #include <linux/uaccess.h>
43 #include <linux/user_namespace.h>
45 #include <linux/kthread.h>
46 #include <linux/freezer.h>
49 #include "ext4_extents.h" /* Needed for trace points definition */
50 #include "ext4_jbd2.h"
56 #define CREATE_TRACE_POINTS
57 #include <trace/events/ext4.h>
59 static struct ext4_lazy_init
*ext4_li_info
;
60 static struct mutex ext4_li_mtx
;
61 static struct ratelimit_state ext4_mount_msg_ratelimit
;
63 static int ext4_load_journal(struct super_block
*, struct ext4_super_block
*,
64 unsigned long journal_devnum
);
65 static int ext4_show_options(struct seq_file
*seq
, struct dentry
*root
);
66 static int ext4_commit_super(struct super_block
*sb
, int sync
);
67 static void ext4_mark_recovery_complete(struct super_block
*sb
,
68 struct ext4_super_block
*es
);
69 static void ext4_clear_journal_err(struct super_block
*sb
,
70 struct ext4_super_block
*es
);
71 static int ext4_sync_fs(struct super_block
*sb
, int wait
);
72 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
);
73 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
74 static int ext4_unfreeze(struct super_block
*sb
);
75 static int ext4_freeze(struct super_block
*sb
);
76 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
77 const char *dev_name
, void *data
);
78 static inline int ext2_feature_set_ok(struct super_block
*sb
);
79 static inline int ext3_feature_set_ok(struct super_block
*sb
);
80 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
);
81 static void ext4_destroy_lazyinit_thread(void);
82 static void ext4_unregister_li_request(struct super_block
*sb
);
83 static void ext4_clear_request_list(void);
84 static struct inode
*ext4_get_journal_inode(struct super_block
*sb
,
85 unsigned int journal_inum
);
90 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
91 * i_mmap_rwsem (inode->i_mmap_rwsem)!
94 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
95 * page lock -> i_data_sem (rw)
97 * buffered write path:
98 * sb_start_write -> i_mutex -> mmap_sem
99 * sb_start_write -> i_mutex -> transaction start -> page lock ->
103 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
104 * i_mmap_rwsem (w) -> page lock
105 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
106 * transaction start -> i_data_sem (rw)
109 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
110 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
111 * transaction start -> i_data_sem (rw)
114 * transaction start -> page lock(s) -> i_data_sem (rw)
117 static bool userns_mounts
= false;
118 module_param(userns_mounts
, bool, 0644);
119 MODULE_PARM_DESC(userns_mounts
, "Allow mounts from unprivileged user namespaces");
121 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
122 static struct file_system_type ext2_fs_type
= {
123 .owner
= THIS_MODULE
,
126 .kill_sb
= kill_block_super
,
127 .fs_flags
= FS_REQUIRES_DEV
| FS_USERNS_MOUNT
,
129 MODULE_ALIAS_FS("ext2");
130 MODULE_ALIAS("ext2");
131 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
133 #define IS_EXT2_SB(sb) (0)
137 static struct file_system_type ext3_fs_type
= {
138 .owner
= THIS_MODULE
,
141 .kill_sb
= kill_block_super
,
142 .fs_flags
= FS_REQUIRES_DEV
| FS_USERNS_MOUNT
,
144 MODULE_ALIAS_FS("ext3");
145 MODULE_ALIAS("ext3");
146 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
149 * This works like sb_bread() except it uses ERR_PTR for error
150 * returns. Currently with sb_bread it's impossible to distinguish
151 * between ENOMEM and EIO situations (since both result in a NULL
155 ext4_sb_bread(struct super_block
*sb
, sector_t block
, int op_flags
)
157 struct buffer_head
*bh
= sb_getblk(sb
, block
);
160 return ERR_PTR(-ENOMEM
);
161 if (buffer_uptodate(bh
))
163 ll_rw_block(REQ_OP_READ
, REQ_META
| op_flags
, 1, &bh
);
165 if (buffer_uptodate(bh
))
168 return ERR_PTR(-EIO
);
171 static int ext4_verify_csum_type(struct super_block
*sb
,
172 struct ext4_super_block
*es
)
174 if (!ext4_has_feature_metadata_csum(sb
))
177 return es
->s_checksum_type
== EXT4_CRC32C_CHKSUM
;
180 static __le32
ext4_superblock_csum(struct super_block
*sb
,
181 struct ext4_super_block
*es
)
183 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
184 int offset
= offsetof(struct ext4_super_block
, s_checksum
);
187 csum
= ext4_chksum(sbi
, ~0, (char *)es
, offset
);
189 return cpu_to_le32(csum
);
192 static int ext4_superblock_csum_verify(struct super_block
*sb
,
193 struct ext4_super_block
*es
)
195 if (!ext4_has_metadata_csum(sb
))
198 return es
->s_checksum
== ext4_superblock_csum(sb
, es
);
201 void ext4_superblock_csum_set(struct super_block
*sb
)
203 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
205 if (!ext4_has_metadata_csum(sb
))
208 es
->s_checksum
= ext4_superblock_csum(sb
, es
);
211 void *ext4_kvmalloc(size_t size
, gfp_t flags
)
215 ret
= kmalloc(size
, flags
| __GFP_NOWARN
);
217 ret
= __vmalloc(size
, flags
, PAGE_KERNEL
);
221 void *ext4_kvzalloc(size_t size
, gfp_t flags
)
225 ret
= kzalloc(size
, flags
| __GFP_NOWARN
);
227 ret
= __vmalloc(size
, flags
| __GFP_ZERO
, PAGE_KERNEL
);
231 ext4_fsblk_t
ext4_block_bitmap(struct super_block
*sb
,
232 struct ext4_group_desc
*bg
)
234 return le32_to_cpu(bg
->bg_block_bitmap_lo
) |
235 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
236 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_block_bitmap_hi
) << 32 : 0);
239 ext4_fsblk_t
ext4_inode_bitmap(struct super_block
*sb
,
240 struct ext4_group_desc
*bg
)
242 return le32_to_cpu(bg
->bg_inode_bitmap_lo
) |
243 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
244 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_bitmap_hi
) << 32 : 0);
247 ext4_fsblk_t
ext4_inode_table(struct super_block
*sb
,
248 struct ext4_group_desc
*bg
)
250 return le32_to_cpu(bg
->bg_inode_table_lo
) |
251 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
252 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_table_hi
) << 32 : 0);
255 __u32
ext4_free_group_clusters(struct super_block
*sb
,
256 struct ext4_group_desc
*bg
)
258 return le16_to_cpu(bg
->bg_free_blocks_count_lo
) |
259 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
260 (__u32
)le16_to_cpu(bg
->bg_free_blocks_count_hi
) << 16 : 0);
263 __u32
ext4_free_inodes_count(struct super_block
*sb
,
264 struct ext4_group_desc
*bg
)
266 return le16_to_cpu(bg
->bg_free_inodes_count_lo
) |
267 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
268 (__u32
)le16_to_cpu(bg
->bg_free_inodes_count_hi
) << 16 : 0);
271 __u32
ext4_used_dirs_count(struct super_block
*sb
,
272 struct ext4_group_desc
*bg
)
274 return le16_to_cpu(bg
->bg_used_dirs_count_lo
) |
275 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
276 (__u32
)le16_to_cpu(bg
->bg_used_dirs_count_hi
) << 16 : 0);
279 __u32
ext4_itable_unused_count(struct super_block
*sb
,
280 struct ext4_group_desc
*bg
)
282 return le16_to_cpu(bg
->bg_itable_unused_lo
) |
283 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
284 (__u32
)le16_to_cpu(bg
->bg_itable_unused_hi
) << 16 : 0);
287 void ext4_block_bitmap_set(struct super_block
*sb
,
288 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
290 bg
->bg_block_bitmap_lo
= cpu_to_le32((u32
)blk
);
291 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
292 bg
->bg_block_bitmap_hi
= cpu_to_le32(blk
>> 32);
295 void ext4_inode_bitmap_set(struct super_block
*sb
,
296 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
298 bg
->bg_inode_bitmap_lo
= cpu_to_le32((u32
)blk
);
299 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
300 bg
->bg_inode_bitmap_hi
= cpu_to_le32(blk
>> 32);
303 void ext4_inode_table_set(struct super_block
*sb
,
304 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
306 bg
->bg_inode_table_lo
= cpu_to_le32((u32
)blk
);
307 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
308 bg
->bg_inode_table_hi
= cpu_to_le32(blk
>> 32);
311 void ext4_free_group_clusters_set(struct super_block
*sb
,
312 struct ext4_group_desc
*bg
, __u32 count
)
314 bg
->bg_free_blocks_count_lo
= cpu_to_le16((__u16
)count
);
315 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
316 bg
->bg_free_blocks_count_hi
= cpu_to_le16(count
>> 16);
319 void ext4_free_inodes_set(struct super_block
*sb
,
320 struct ext4_group_desc
*bg
, __u32 count
)
322 bg
->bg_free_inodes_count_lo
= cpu_to_le16((__u16
)count
);
323 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
324 bg
->bg_free_inodes_count_hi
= cpu_to_le16(count
>> 16);
327 void ext4_used_dirs_set(struct super_block
*sb
,
328 struct ext4_group_desc
*bg
, __u32 count
)
330 bg
->bg_used_dirs_count_lo
= cpu_to_le16((__u16
)count
);
331 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
332 bg
->bg_used_dirs_count_hi
= cpu_to_le16(count
>> 16);
335 void ext4_itable_unused_set(struct super_block
*sb
,
336 struct ext4_group_desc
*bg
, __u32 count
)
338 bg
->bg_itable_unused_lo
= cpu_to_le16((__u16
)count
);
339 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
340 bg
->bg_itable_unused_hi
= cpu_to_le16(count
>> 16);
344 static void __save_error_info(struct super_block
*sb
, const char *func
,
347 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
349 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
350 if (bdev_read_only(sb
->s_bdev
))
352 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
353 es
->s_last_error_time
= cpu_to_le32(get_seconds());
354 strncpy(es
->s_last_error_func
, func
, sizeof(es
->s_last_error_func
));
355 es
->s_last_error_line
= cpu_to_le32(line
);
356 if (!es
->s_first_error_time
) {
357 es
->s_first_error_time
= es
->s_last_error_time
;
358 strncpy(es
->s_first_error_func
, func
,
359 sizeof(es
->s_first_error_func
));
360 es
->s_first_error_line
= cpu_to_le32(line
);
361 es
->s_first_error_ino
= es
->s_last_error_ino
;
362 es
->s_first_error_block
= es
->s_last_error_block
;
365 * Start the daily error reporting function if it hasn't been
368 if (!es
->s_error_count
)
369 mod_timer(&EXT4_SB(sb
)->s_err_report
, jiffies
+ 24*60*60*HZ
);
370 le32_add_cpu(&es
->s_error_count
, 1);
373 static void save_error_info(struct super_block
*sb
, const char *func
,
376 __save_error_info(sb
, func
, line
);
377 ext4_commit_super(sb
, 1);
381 * The del_gendisk() function uninitializes the disk-specific data
382 * structures, including the bdi structure, without telling anyone
383 * else. Once this happens, any attempt to call mark_buffer_dirty()
384 * (for example, by ext4_commit_super), will cause a kernel OOPS.
385 * This is a kludge to prevent these oops until we can put in a proper
386 * hook in del_gendisk() to inform the VFS and file system layers.
388 static int block_device_ejected(struct super_block
*sb
)
390 struct inode
*bd_inode
= sb
->s_bdev
->bd_inode
;
391 struct backing_dev_info
*bdi
= inode_to_bdi(bd_inode
);
393 return bdi
->dev
== NULL
;
396 static void ext4_journal_commit_callback(journal_t
*journal
, transaction_t
*txn
)
398 struct super_block
*sb
= journal
->j_private
;
399 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
400 int error
= is_journal_aborted(journal
);
401 struct ext4_journal_cb_entry
*jce
;
403 BUG_ON(txn
->t_state
== T_FINISHED
);
405 ext4_process_freed_data(sb
, txn
->t_tid
);
407 spin_lock(&sbi
->s_md_lock
);
408 while (!list_empty(&txn
->t_private_list
)) {
409 jce
= list_entry(txn
->t_private_list
.next
,
410 struct ext4_journal_cb_entry
, jce_list
);
411 list_del_init(&jce
->jce_list
);
412 spin_unlock(&sbi
->s_md_lock
);
413 jce
->jce_func(sb
, jce
, error
);
414 spin_lock(&sbi
->s_md_lock
);
416 spin_unlock(&sbi
->s_md_lock
);
419 static bool system_going_down(void)
421 return system_state
== SYSTEM_HALT
|| system_state
== SYSTEM_POWER_OFF
422 || system_state
== SYSTEM_RESTART
;
425 /* Deal with the reporting of failure conditions on a filesystem such as
426 * inconsistencies detected or read IO failures.
428 * On ext2, we can store the error state of the filesystem in the
429 * superblock. That is not possible on ext4, because we may have other
430 * write ordering constraints on the superblock which prevent us from
431 * writing it out straight away; and given that the journal is about to
432 * be aborted, we can't rely on the current, or future, transactions to
433 * write out the superblock safely.
435 * We'll just use the jbd2_journal_abort() error code to record an error in
436 * the journal instead. On recovery, the journal will complain about
437 * that error until we've noted it down and cleared it.
440 static void ext4_handle_error(struct super_block
*sb
)
445 if (!test_opt(sb
, ERRORS_CONT
)) {
446 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
448 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
450 jbd2_journal_abort(journal
, -EIO
);
453 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
454 * could panic during 'reboot -f' as the underlying device got already
457 if (test_opt(sb
, ERRORS_RO
) || system_going_down()) {
458 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
460 * Make sure updated value of ->s_mount_flags will be visible
461 * before ->s_flags update
464 sb
->s_flags
|= SB_RDONLY
;
465 } else if (test_opt(sb
, ERRORS_PANIC
)) {
466 if (EXT4_SB(sb
)->s_journal
&&
467 !(EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_REC_ERR
))
469 panic("EXT4-fs (device %s): panic forced after error\n",
474 #define ext4_error_ratelimit(sb) \
475 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
478 void __ext4_error(struct super_block
*sb
, const char *function
,
479 unsigned int line
, const char *fmt
, ...)
481 struct va_format vaf
;
484 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb
))))
487 if (ext4_error_ratelimit(sb
)) {
492 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
493 sb
->s_id
, function
, line
, current
->comm
, &vaf
);
496 save_error_info(sb
, function
, line
);
497 ext4_handle_error(sb
);
500 void __ext4_error_inode(struct inode
*inode
, const char *function
,
501 unsigned int line
, ext4_fsblk_t block
,
502 const char *fmt
, ...)
505 struct va_format vaf
;
506 struct ext4_super_block
*es
= EXT4_SB(inode
->i_sb
)->s_es
;
508 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode
->i_sb
))))
511 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
512 es
->s_last_error_block
= cpu_to_le64(block
);
513 if (ext4_error_ratelimit(inode
->i_sb
)) {
518 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
519 "inode #%lu: block %llu: comm %s: %pV\n",
520 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
521 block
, current
->comm
, &vaf
);
523 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
524 "inode #%lu: comm %s: %pV\n",
525 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
526 current
->comm
, &vaf
);
529 save_error_info(inode
->i_sb
, function
, line
);
530 ext4_handle_error(inode
->i_sb
);
533 void __ext4_error_file(struct file
*file
, const char *function
,
534 unsigned int line
, ext4_fsblk_t block
,
535 const char *fmt
, ...)
538 struct va_format vaf
;
539 struct ext4_super_block
*es
;
540 struct inode
*inode
= file_inode(file
);
541 char pathname
[80], *path
;
543 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode
->i_sb
))))
546 es
= EXT4_SB(inode
->i_sb
)->s_es
;
547 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
548 if (ext4_error_ratelimit(inode
->i_sb
)) {
549 path
= file_path(file
, pathname
, sizeof(pathname
));
557 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
558 "block %llu: comm %s: path %s: %pV\n",
559 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
560 block
, current
->comm
, path
, &vaf
);
563 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
564 "comm %s: path %s: %pV\n",
565 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
566 current
->comm
, path
, &vaf
);
569 save_error_info(inode
->i_sb
, function
, line
);
570 ext4_handle_error(inode
->i_sb
);
573 const char *ext4_decode_error(struct super_block
*sb
, int errno
,
580 errstr
= "Corrupt filesystem";
583 errstr
= "Filesystem failed CRC";
586 errstr
= "IO failure";
589 errstr
= "Out of memory";
592 if (!sb
|| (EXT4_SB(sb
)->s_journal
&&
593 EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
))
594 errstr
= "Journal has aborted";
596 errstr
= "Readonly filesystem";
599 /* If the caller passed in an extra buffer for unknown
600 * errors, textualise them now. Else we just return
603 /* Check for truncated error codes... */
604 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
613 /* __ext4_std_error decodes expected errors from journaling functions
614 * automatically and invokes the appropriate error response. */
616 void __ext4_std_error(struct super_block
*sb
, const char *function
,
617 unsigned int line
, int errno
)
622 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb
))))
625 /* Special case: if the error is EROFS, and we're not already
626 * inside a transaction, then there's really no point in logging
628 if (errno
== -EROFS
&& journal_current_handle() == NULL
&& sb_rdonly(sb
))
631 if (ext4_error_ratelimit(sb
)) {
632 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
633 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s:%d: %s\n",
634 sb
->s_id
, function
, line
, errstr
);
637 save_error_info(sb
, function
, line
);
638 ext4_handle_error(sb
);
642 * ext4_abort is a much stronger failure handler than ext4_error. The
643 * abort function may be used to deal with unrecoverable failures such
644 * as journal IO errors or ENOMEM at a critical moment in log management.
646 * We unconditionally force the filesystem into an ABORT|READONLY state,
647 * unless the error response on the fs has been set to panic in which
648 * case we take the easy way out and panic immediately.
651 void __ext4_abort(struct super_block
*sb
, const char *function
,
652 unsigned int line
, const char *fmt
, ...)
654 struct va_format vaf
;
657 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb
))))
660 save_error_info(sb
, function
, line
);
664 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: %pV\n",
665 sb
->s_id
, function
, line
, &vaf
);
668 if (sb_rdonly(sb
) == 0) {
669 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
670 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
672 * Make sure updated value of ->s_mount_flags will be visible
673 * before ->s_flags update
676 sb
->s_flags
|= SB_RDONLY
;
677 if (EXT4_SB(sb
)->s_journal
)
678 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
679 save_error_info(sb
, function
, line
);
681 if (test_opt(sb
, ERRORS_PANIC
) && !system_going_down()) {
682 if (EXT4_SB(sb
)->s_journal
&&
683 !(EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_REC_ERR
))
685 panic("EXT4-fs panic from previous error\n");
689 void __ext4_msg(struct super_block
*sb
,
690 const char *prefix
, const char *fmt
, ...)
692 struct va_format vaf
;
695 if (!___ratelimit(&(EXT4_SB(sb
)->s_msg_ratelimit_state
), "EXT4-fs"))
701 printk("%sEXT4-fs (%s): %pV\n", prefix
, sb
->s_id
, &vaf
);
705 #define ext4_warning_ratelimit(sb) \
706 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
709 void __ext4_warning(struct super_block
*sb
, const char *function
,
710 unsigned int line
, const char *fmt
, ...)
712 struct va_format vaf
;
715 if (!ext4_warning_ratelimit(sb
))
721 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: %pV\n",
722 sb
->s_id
, function
, line
, &vaf
);
726 void __ext4_warning_inode(const struct inode
*inode
, const char *function
,
727 unsigned int line
, const char *fmt
, ...)
729 struct va_format vaf
;
732 if (!ext4_warning_ratelimit(inode
->i_sb
))
738 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: "
739 "inode #%lu: comm %s: %pV\n", inode
->i_sb
->s_id
,
740 function
, line
, inode
->i_ino
, current
->comm
, &vaf
);
744 void __ext4_grp_locked_error(const char *function
, unsigned int line
,
745 struct super_block
*sb
, ext4_group_t grp
,
746 unsigned long ino
, ext4_fsblk_t block
,
747 const char *fmt
, ...)
751 struct va_format vaf
;
753 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
755 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb
))))
758 es
->s_last_error_ino
= cpu_to_le32(ino
);
759 es
->s_last_error_block
= cpu_to_le64(block
);
760 __save_error_info(sb
, function
, line
);
762 if (ext4_error_ratelimit(sb
)) {
766 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: group %u, ",
767 sb
->s_id
, function
, line
, grp
);
769 printk(KERN_CONT
"inode %lu: ", ino
);
771 printk(KERN_CONT
"block %llu:",
772 (unsigned long long) block
);
773 printk(KERN_CONT
"%pV\n", &vaf
);
777 if (test_opt(sb
, ERRORS_CONT
)) {
778 ext4_commit_super(sb
, 0);
782 ext4_unlock_group(sb
, grp
);
783 ext4_commit_super(sb
, 1);
784 ext4_handle_error(sb
);
786 * We only get here in the ERRORS_RO case; relocking the group
787 * may be dangerous, but nothing bad will happen since the
788 * filesystem will have already been marked read/only and the
789 * journal has been aborted. We return 1 as a hint to callers
790 * who might what to use the return value from
791 * ext4_grp_locked_error() to distinguish between the
792 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
793 * aggressively from the ext4 function in question, with a
794 * more appropriate error code.
796 ext4_lock_group(sb
, grp
);
800 void ext4_update_dynamic_rev(struct super_block
*sb
)
802 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
804 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
808 "updating to rev %d because of new feature flag, "
809 "running e2fsck is recommended",
812 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
813 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
814 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
815 /* leave es->s_feature_*compat flags alone */
816 /* es->s_uuid will be set by e2fsck if empty */
819 * The rest of the superblock fields should be zero, and if not it
820 * means they are likely already in use, so leave them alone. We
821 * can leave it up to e2fsck to clean up any inconsistencies there.
826 * Open the external journal device
828 static struct block_device
*ext4_blkdev_get(dev_t dev
, struct super_block
*sb
)
830 struct block_device
*bdev
;
831 char b
[BDEVNAME_SIZE
];
833 bdev
= blkdev_get_by_dev(dev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
, sb
);
839 ext4_msg(sb
, KERN_ERR
, "failed to open journal device %s: %ld",
840 __bdevname(dev
, b
), PTR_ERR(bdev
));
845 * Release the journal device
847 static void ext4_blkdev_put(struct block_device
*bdev
)
849 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
);
852 static void ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
854 struct block_device
*bdev
;
855 bdev
= sbi
->journal_bdev
;
857 ext4_blkdev_put(bdev
);
858 sbi
->journal_bdev
= NULL
;
862 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
864 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
867 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
871 ext4_msg(sb
, KERN_ERR
, "sb orphan head is %d",
872 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
874 printk(KERN_ERR
"sb_info orphan list:\n");
875 list_for_each(l
, &sbi
->s_orphan
) {
876 struct inode
*inode
= orphan_list_entry(l
);
878 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
879 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
880 inode
->i_mode
, inode
->i_nlink
,
886 static int ext4_quota_off(struct super_block
*sb
, int type
);
888 static inline void ext4_quota_off_umount(struct super_block
*sb
)
892 /* Use our quota_off function to clear inode flags etc. */
893 for (type
= 0; type
< EXT4_MAXQUOTAS
; type
++)
894 ext4_quota_off(sb
, type
);
898 * This is a helper function which is used in the mount/remount
899 * codepaths (which holds s_umount) to fetch the quota file name.
901 static inline char *get_qf_name(struct super_block
*sb
,
902 struct ext4_sb_info
*sbi
,
905 return rcu_dereference_protected(sbi
->s_qf_names
[type
],
906 lockdep_is_held(&sb
->s_umount
));
909 static inline void ext4_quota_off_umount(struct super_block
*sb
)
914 static void ext4_put_super(struct super_block
*sb
)
916 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
917 struct ext4_super_block
*es
= sbi
->s_es
;
921 ext4_unregister_li_request(sb
);
922 ext4_quota_off_umount(sb
);
924 flush_workqueue(sbi
->rsv_conversion_wq
);
925 destroy_workqueue(sbi
->rsv_conversion_wq
);
927 if (sbi
->s_journal
) {
928 aborted
= is_journal_aborted(sbi
->s_journal
);
929 err
= jbd2_journal_destroy(sbi
->s_journal
);
930 sbi
->s_journal
= NULL
;
931 if ((err
< 0) && !aborted
)
932 ext4_abort(sb
, "Couldn't clean up the journal");
935 ext4_unregister_sysfs(sb
);
936 ext4_es_unregister_shrinker(sbi
);
937 del_timer_sync(&sbi
->s_err_report
);
938 ext4_release_system_zone(sb
);
940 ext4_ext_release(sb
);
942 if (!sb_rdonly(sb
) && !aborted
) {
943 ext4_clear_feature_journal_needs_recovery(sb
);
944 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
947 ext4_commit_super(sb
, 1);
949 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
950 brelse(sbi
->s_group_desc
[i
]);
951 kvfree(sbi
->s_group_desc
);
952 kvfree(sbi
->s_flex_groups
);
953 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
954 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
955 percpu_counter_destroy(&sbi
->s_dirs_counter
);
956 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
957 percpu_free_rwsem(&sbi
->s_journal_flag_rwsem
);
959 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
960 kfree(get_qf_name(sb
, sbi
, i
));
963 /* Debugging code just in case the in-memory inode orphan list
964 * isn't empty. The on-disk one can be non-empty if we've
965 * detected an error and taken the fs readonly, but the
966 * in-memory list had better be clean by this point. */
967 if (!list_empty(&sbi
->s_orphan
))
968 dump_orphan_list(sb
, sbi
);
969 J_ASSERT(list_empty(&sbi
->s_orphan
));
971 sync_blockdev(sb
->s_bdev
);
972 invalidate_bdev(sb
->s_bdev
);
973 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
975 * Invalidate the journal device's buffers. We don't want them
976 * floating about in memory - the physical journal device may
977 * hotswapped, and it breaks the `ro-after' testing code.
979 sync_blockdev(sbi
->journal_bdev
);
980 invalidate_bdev(sbi
->journal_bdev
);
981 ext4_blkdev_remove(sbi
);
983 if (sbi
->s_ea_inode_cache
) {
984 ext4_xattr_destroy_cache(sbi
->s_ea_inode_cache
);
985 sbi
->s_ea_inode_cache
= NULL
;
987 if (sbi
->s_ea_block_cache
) {
988 ext4_xattr_destroy_cache(sbi
->s_ea_block_cache
);
989 sbi
->s_ea_block_cache
= NULL
;
992 kthread_stop(sbi
->s_mmp_tsk
);
994 sb
->s_fs_info
= NULL
;
996 * Now that we are completely done shutting down the
997 * superblock, we need to actually destroy the kobject.
999 kobject_put(&sbi
->s_kobj
);
1000 wait_for_completion(&sbi
->s_kobj_unregister
);
1001 if (sbi
->s_chksum_driver
)
1002 crypto_free_shash(sbi
->s_chksum_driver
);
1003 kfree(sbi
->s_blockgroup_lock
);
1004 fs_put_dax(sbi
->s_daxdev
);
1008 static struct kmem_cache
*ext4_inode_cachep
;
1011 * Called inside transaction, so use GFP_NOFS
1013 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
1015 struct ext4_inode_info
*ei
;
1017 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
1021 ei
->vfs_inode
.i_version
= 1;
1022 spin_lock_init(&ei
->i_raw_lock
);
1023 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
1024 spin_lock_init(&ei
->i_prealloc_lock
);
1025 ext4_es_init_tree(&ei
->i_es_tree
);
1026 rwlock_init(&ei
->i_es_lock
);
1027 INIT_LIST_HEAD(&ei
->i_es_list
);
1028 ei
->i_es_all_nr
= 0;
1029 ei
->i_es_shk_nr
= 0;
1030 ei
->i_es_shrink_lblk
= 0;
1031 ei
->i_reserved_data_blocks
= 0;
1032 ei
->i_da_metadata_calc_len
= 0;
1033 ei
->i_da_metadata_calc_last_lblock
= 0;
1034 spin_lock_init(&(ei
->i_block_reservation_lock
));
1036 ei
->i_reserved_quota
= 0;
1037 memset(&ei
->i_dquot
, 0, sizeof(ei
->i_dquot
));
1040 INIT_LIST_HEAD(&ei
->i_rsv_conversion_list
);
1041 spin_lock_init(&ei
->i_completed_io_lock
);
1043 ei
->i_datasync_tid
= 0;
1044 atomic_set(&ei
->i_unwritten
, 0);
1045 INIT_WORK(&ei
->i_rsv_conversion_work
, ext4_end_io_rsv_work
);
1046 return &ei
->vfs_inode
;
1049 static int ext4_drop_inode(struct inode
*inode
)
1051 int drop
= generic_drop_inode(inode
);
1053 trace_ext4_drop_inode(inode
, drop
);
1057 static void ext4_i_callback(struct rcu_head
*head
)
1059 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
1060 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
1063 static void ext4_destroy_inode(struct inode
*inode
)
1065 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
1066 ext4_msg(inode
->i_sb
, KERN_ERR
,
1067 "Inode %lu (%p): orphan list check failed!",
1068 inode
->i_ino
, EXT4_I(inode
));
1069 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
1070 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
1074 call_rcu(&inode
->i_rcu
, ext4_i_callback
);
1077 static void init_once(void *foo
)
1079 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
1081 INIT_LIST_HEAD(&ei
->i_orphan
);
1082 init_rwsem(&ei
->xattr_sem
);
1083 init_rwsem(&ei
->i_data_sem
);
1084 init_rwsem(&ei
->i_mmap_sem
);
1085 inode_init_once(&ei
->vfs_inode
);
1088 static int __init
init_inodecache(void)
1090 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
1091 sizeof(struct ext4_inode_info
),
1092 0, (SLAB_RECLAIM_ACCOUNT
|
1093 SLAB_MEM_SPREAD
|SLAB_ACCOUNT
),
1095 if (ext4_inode_cachep
== NULL
)
1100 static void destroy_inodecache(void)
1103 * Make sure all delayed rcu free inodes are flushed before we
1107 kmem_cache_destroy(ext4_inode_cachep
);
1110 void ext4_clear_inode(struct inode
*inode
)
1112 invalidate_inode_buffers(inode
);
1115 ext4_discard_preallocations(inode
);
1116 ext4_es_remove_extent(inode
, 0, EXT_MAX_BLOCKS
);
1117 if (EXT4_I(inode
)->jinode
) {
1118 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode
),
1119 EXT4_I(inode
)->jinode
);
1120 jbd2_free_inode(EXT4_I(inode
)->jinode
);
1121 EXT4_I(inode
)->jinode
= NULL
;
1123 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1124 fscrypt_put_encryption_info(inode
, NULL
);
1128 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
1129 u64 ino
, u32 generation
)
1131 struct inode
*inode
;
1134 * Currently we don't know the generation for parent directory, so
1135 * a generation of 0 means "accept any"
1137 inode
= ext4_iget(sb
, ino
, EXT4_IGET_HANDLE
);
1139 return ERR_CAST(inode
);
1140 if (generation
&& inode
->i_generation
!= generation
) {
1142 return ERR_PTR(-ESTALE
);
1148 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1149 int fh_len
, int fh_type
)
1151 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
1152 ext4_nfs_get_inode
);
1155 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1156 int fh_len
, int fh_type
)
1158 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1159 ext4_nfs_get_inode
);
1162 static int ext4_nfs_commit_metadata(struct inode
*inode
)
1164 struct writeback_control wbc
= {
1165 .sync_mode
= WB_SYNC_ALL
1168 trace_ext4_nfs_commit_metadata(inode
);
1169 return ext4_write_inode(inode
, &wbc
);
1173 * Try to release metadata pages (indirect blocks, directories) which are
1174 * mapped via the block device. Since these pages could have journal heads
1175 * which would prevent try_to_free_buffers() from freeing them, we must use
1176 * jbd2 layer's try_to_free_buffers() function to release them.
1178 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
1181 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
1183 WARN_ON(PageChecked(page
));
1184 if (!page_has_buffers(page
))
1187 return jbd2_journal_try_to_free_buffers(journal
, page
,
1188 wait
& ~__GFP_DIRECT_RECLAIM
);
1189 return try_to_free_buffers(page
);
1192 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1193 static int ext4_get_context(struct inode
*inode
, void *ctx
, size_t len
)
1195 return ext4_xattr_get(inode
, EXT4_XATTR_INDEX_ENCRYPTION
,
1196 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT
, ctx
, len
);
1199 static int ext4_set_context(struct inode
*inode
, const void *ctx
, size_t len
,
1202 handle_t
*handle
= fs_data
;
1203 int res
, res2
, credits
, retries
= 0;
1206 * Encrypting the root directory is not allowed because e2fsck expects
1207 * lost+found to exist and be unencrypted, and encrypting the root
1208 * directory would imply encrypting the lost+found directory as well as
1209 * the filename "lost+found" itself.
1211 if (inode
->i_ino
== EXT4_ROOT_INO
)
1214 if (WARN_ON_ONCE(IS_DAX(inode
) && i_size_read(inode
)))
1217 res
= ext4_convert_inline_data(inode
);
1222 * If a journal handle was specified, then the encryption context is
1223 * being set on a new inode via inheritance and is part of a larger
1224 * transaction to create the inode. Otherwise the encryption context is
1225 * being set on an existing inode in its own transaction. Only in the
1226 * latter case should the "retry on ENOSPC" logic be used.
1230 res
= ext4_xattr_set_handle(handle
, inode
,
1231 EXT4_XATTR_INDEX_ENCRYPTION
,
1232 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT
,
1235 ext4_set_inode_flag(inode
, EXT4_INODE_ENCRYPT
);
1236 ext4_clear_inode_state(inode
,
1237 EXT4_STATE_MAY_INLINE_DATA
);
1239 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1240 * S_DAX may be disabled
1242 ext4_set_inode_flags(inode
);
1247 res
= dquot_initialize(inode
);
1251 res
= ext4_xattr_set_credits(inode
, len
, false /* is_create */,
1256 handle
= ext4_journal_start(inode
, EXT4_HT_MISC
, credits
);
1258 return PTR_ERR(handle
);
1260 res
= ext4_xattr_set_handle(handle
, inode
, EXT4_XATTR_INDEX_ENCRYPTION
,
1261 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT
,
1264 ext4_set_inode_flag(inode
, EXT4_INODE_ENCRYPT
);
1266 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1267 * S_DAX may be disabled
1269 ext4_set_inode_flags(inode
);
1270 res
= ext4_mark_inode_dirty(handle
, inode
);
1272 EXT4_ERROR_INODE(inode
, "Failed to mark inode dirty");
1274 res2
= ext4_journal_stop(handle
);
1276 if (res
== -ENOSPC
&& ext4_should_retry_alloc(inode
->i_sb
, &retries
))
1283 static bool ext4_dummy_context(struct inode
*inode
)
1285 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode
->i_sb
));
1288 static unsigned ext4_max_namelen(struct inode
*inode
)
1290 return S_ISLNK(inode
->i_mode
) ? inode
->i_sb
->s_blocksize
:
1294 static const struct fscrypt_operations ext4_cryptops
= {
1295 .key_prefix
= "ext4:",
1296 .get_context
= ext4_get_context
,
1297 .set_context
= ext4_set_context
,
1298 .dummy_context
= ext4_dummy_context
,
1299 .empty_dir
= ext4_empty_dir
,
1300 .max_namelen
= ext4_max_namelen
,
1305 static const char * const quotatypes
[] = INITQFNAMES
;
1306 #define QTYPE2NAME(t) (quotatypes[t])
1308 static int ext4_write_dquot(struct dquot
*dquot
);
1309 static int ext4_acquire_dquot(struct dquot
*dquot
);
1310 static int ext4_release_dquot(struct dquot
*dquot
);
1311 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
1312 static int ext4_write_info(struct super_block
*sb
, int type
);
1313 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
1314 const struct path
*path
);
1315 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
1316 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
1317 size_t len
, loff_t off
);
1318 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
1319 const char *data
, size_t len
, loff_t off
);
1320 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
1321 unsigned int flags
);
1322 static int ext4_enable_quotas(struct super_block
*sb
);
1323 static int ext4_get_next_id(struct super_block
*sb
, struct kqid
*qid
);
1325 static struct dquot
**ext4_get_dquots(struct inode
*inode
)
1327 return EXT4_I(inode
)->i_dquot
;
1330 static const struct dquot_operations ext4_quota_operations
= {
1331 .get_reserved_space
= ext4_get_reserved_space
,
1332 .write_dquot
= ext4_write_dquot
,
1333 .acquire_dquot
= ext4_acquire_dquot
,
1334 .release_dquot
= ext4_release_dquot
,
1335 .mark_dirty
= ext4_mark_dquot_dirty
,
1336 .write_info
= ext4_write_info
,
1337 .alloc_dquot
= dquot_alloc
,
1338 .destroy_dquot
= dquot_destroy
,
1339 .get_projid
= ext4_get_projid
,
1340 .get_inode_usage
= ext4_get_inode_usage
,
1341 .get_next_id
= ext4_get_next_id
,
1344 static const struct quotactl_ops ext4_qctl_operations
= {
1345 .quota_on
= ext4_quota_on
,
1346 .quota_off
= ext4_quota_off
,
1347 .quota_sync
= dquot_quota_sync
,
1348 .get_state
= dquot_get_state
,
1349 .set_info
= dquot_set_dqinfo
,
1350 .get_dqblk
= dquot_get_dqblk
,
1351 .set_dqblk
= dquot_set_dqblk
,
1352 .get_nextdqblk
= dquot_get_next_dqblk
,
1356 static const struct super_operations ext4_sops
= {
1357 .alloc_inode
= ext4_alloc_inode
,
1358 .destroy_inode
= ext4_destroy_inode
,
1359 .write_inode
= ext4_write_inode
,
1360 .dirty_inode
= ext4_dirty_inode
,
1361 .drop_inode
= ext4_drop_inode
,
1362 .evict_inode
= ext4_evict_inode
,
1363 .put_super
= ext4_put_super
,
1364 .sync_fs
= ext4_sync_fs
,
1365 .freeze_fs
= ext4_freeze
,
1366 .unfreeze_fs
= ext4_unfreeze
,
1367 .statfs
= ext4_statfs
,
1368 .remount_fs
= ext4_remount
,
1369 .show_options
= ext4_show_options
,
1371 .quota_read
= ext4_quota_read
,
1372 .quota_write
= ext4_quota_write
,
1373 .get_dquots
= ext4_get_dquots
,
1375 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1378 static const struct export_operations ext4_export_ops
= {
1379 .fh_to_dentry
= ext4_fh_to_dentry
,
1380 .fh_to_parent
= ext4_fh_to_parent
,
1381 .get_parent
= ext4_get_parent
,
1382 .commit_metadata
= ext4_nfs_commit_metadata
,
1386 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1387 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1388 Opt_nouid32
, Opt_debug
, Opt_removed
,
1389 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1390 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
,
1391 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
, Opt_journal_dev
,
1392 Opt_journal_path
, Opt_journal_checksum
, Opt_journal_async_commit
,
1393 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1394 Opt_data_err_abort
, Opt_data_err_ignore
, Opt_test_dummy_encryption
,
1395 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1396 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
1397 Opt_noquota
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
1398 Opt_usrquota
, Opt_grpquota
, Opt_prjquota
, Opt_i_version
, Opt_dax
,
1399 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
, Opt_mblk_io_submit
,
1400 Opt_lazytime
, Opt_nolazytime
, Opt_debug_want_extra_isize
,
1401 Opt_nomblk_io_submit
, Opt_block_validity
, Opt_noblock_validity
,
1402 Opt_inode_readahead_blks
, Opt_journal_ioprio
,
1403 Opt_dioread_nolock
, Opt_dioread_lock
,
1404 Opt_discard
, Opt_nodiscard
, Opt_init_itable
, Opt_noinit_itable
,
1405 Opt_max_dir_size_kb
, Opt_nojournal_checksum
, Opt_nombcache
,
1408 static const match_table_t tokens
= {
1409 {Opt_bsd_df
, "bsddf"},
1410 {Opt_minix_df
, "minixdf"},
1411 {Opt_grpid
, "grpid"},
1412 {Opt_grpid
, "bsdgroups"},
1413 {Opt_nogrpid
, "nogrpid"},
1414 {Opt_nogrpid
, "sysvgroups"},
1415 {Opt_resgid
, "resgid=%u"},
1416 {Opt_resuid
, "resuid=%u"},
1418 {Opt_err_cont
, "errors=continue"},
1419 {Opt_err_panic
, "errors=panic"},
1420 {Opt_err_ro
, "errors=remount-ro"},
1421 {Opt_nouid32
, "nouid32"},
1422 {Opt_debug
, "debug"},
1423 {Opt_removed
, "oldalloc"},
1424 {Opt_removed
, "orlov"},
1425 {Opt_user_xattr
, "user_xattr"},
1426 {Opt_nouser_xattr
, "nouser_xattr"},
1428 {Opt_noacl
, "noacl"},
1429 {Opt_noload
, "norecovery"},
1430 {Opt_noload
, "noload"},
1431 {Opt_removed
, "nobh"},
1432 {Opt_removed
, "bh"},
1433 {Opt_commit
, "commit=%u"},
1434 {Opt_min_batch_time
, "min_batch_time=%u"},
1435 {Opt_max_batch_time
, "max_batch_time=%u"},
1436 {Opt_journal_dev
, "journal_dev=%u"},
1437 {Opt_journal_path
, "journal_path=%s"},
1438 {Opt_journal_checksum
, "journal_checksum"},
1439 {Opt_nojournal_checksum
, "nojournal_checksum"},
1440 {Opt_journal_async_commit
, "journal_async_commit"},
1441 {Opt_abort
, "abort"},
1442 {Opt_data_journal
, "data=journal"},
1443 {Opt_data_ordered
, "data=ordered"},
1444 {Opt_data_writeback
, "data=writeback"},
1445 {Opt_data_err_abort
, "data_err=abort"},
1446 {Opt_data_err_ignore
, "data_err=ignore"},
1447 {Opt_offusrjquota
, "usrjquota="},
1448 {Opt_usrjquota
, "usrjquota=%s"},
1449 {Opt_offgrpjquota
, "grpjquota="},
1450 {Opt_grpjquota
, "grpjquota=%s"},
1451 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1452 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1453 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
1454 {Opt_grpquota
, "grpquota"},
1455 {Opt_noquota
, "noquota"},
1456 {Opt_quota
, "quota"},
1457 {Opt_usrquota
, "usrquota"},
1458 {Opt_prjquota
, "prjquota"},
1459 {Opt_barrier
, "barrier=%u"},
1460 {Opt_barrier
, "barrier"},
1461 {Opt_nobarrier
, "nobarrier"},
1462 {Opt_i_version
, "i_version"},
1464 {Opt_stripe
, "stripe=%u"},
1465 {Opt_delalloc
, "delalloc"},
1466 {Opt_lazytime
, "lazytime"},
1467 {Opt_nolazytime
, "nolazytime"},
1468 {Opt_debug_want_extra_isize
, "debug_want_extra_isize=%u"},
1469 {Opt_nodelalloc
, "nodelalloc"},
1470 {Opt_removed
, "mblk_io_submit"},
1471 {Opt_removed
, "nomblk_io_submit"},
1472 {Opt_block_validity
, "block_validity"},
1473 {Opt_noblock_validity
, "noblock_validity"},
1474 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1475 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1476 {Opt_auto_da_alloc
, "auto_da_alloc=%u"},
1477 {Opt_auto_da_alloc
, "auto_da_alloc"},
1478 {Opt_noauto_da_alloc
, "noauto_da_alloc"},
1479 {Opt_dioread_nolock
, "dioread_nolock"},
1480 {Opt_dioread_lock
, "dioread_lock"},
1481 {Opt_discard
, "discard"},
1482 {Opt_nodiscard
, "nodiscard"},
1483 {Opt_init_itable
, "init_itable=%u"},
1484 {Opt_init_itable
, "init_itable"},
1485 {Opt_noinit_itable
, "noinit_itable"},
1486 {Opt_max_dir_size_kb
, "max_dir_size_kb=%u"},
1487 {Opt_test_dummy_encryption
, "test_dummy_encryption"},
1488 {Opt_nombcache
, "nombcache"},
1489 {Opt_nombcache
, "no_mbcache"}, /* for backward compatibility */
1490 {Opt_removed
, "check=none"}, /* mount option from ext2/3 */
1491 {Opt_removed
, "nocheck"}, /* mount option from ext2/3 */
1492 {Opt_removed
, "reservation"}, /* mount option from ext2/3 */
1493 {Opt_removed
, "noreservation"}, /* mount option from ext2/3 */
1494 {Opt_removed
, "journal=%u"}, /* mount option from ext2/3 */
1498 static ext4_fsblk_t
get_sb_block(void **data
)
1500 ext4_fsblk_t sb_block
;
1501 char *options
= (char *) *data
;
1503 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1504 return 1; /* Default location */
1507 /* TODO: use simple_strtoll with >32bit ext4 */
1508 sb_block
= simple_strtoul(options
, &options
, 0);
1509 if (*options
&& *options
!= ',') {
1510 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1514 if (*options
== ',')
1516 *data
= (void *) options
;
1521 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1522 static const char deprecated_msg
[] =
1523 "Mount option \"%s\" will be removed by %s\n"
1524 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1527 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
1529 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1530 char *qname
, *old_qname
= get_qf_name(sb
, sbi
, qtype
);
1533 if (sb_any_quota_loaded(sb
) && !old_qname
) {
1534 ext4_msg(sb
, KERN_ERR
,
1535 "Cannot change journaled "
1536 "quota options when quota turned on");
1539 if (ext4_has_feature_quota(sb
)) {
1540 ext4_msg(sb
, KERN_INFO
, "Journaled quota options "
1541 "ignored when QUOTA feature is enabled");
1544 qname
= match_strdup(args
);
1546 ext4_msg(sb
, KERN_ERR
,
1547 "Not enough memory for storing quotafile name");
1551 if (strcmp(old_qname
, qname
) == 0)
1554 ext4_msg(sb
, KERN_ERR
,
1555 "%s quota file already specified",
1559 if (strchr(qname
, '/')) {
1560 ext4_msg(sb
, KERN_ERR
,
1561 "quotafile must be on filesystem root");
1564 rcu_assign_pointer(sbi
->s_qf_names
[qtype
], qname
);
1572 static int clear_qf_name(struct super_block
*sb
, int qtype
)
1575 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1576 char *old_qname
= get_qf_name(sb
, sbi
, qtype
);
1578 if (sb_any_quota_loaded(sb
) && old_qname
) {
1579 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
1580 " when quota turned on");
1583 rcu_assign_pointer(sbi
->s_qf_names
[qtype
], NULL
);
1590 #define MOPT_SET 0x0001
1591 #define MOPT_CLEAR 0x0002
1592 #define MOPT_NOSUPPORT 0x0004
1593 #define MOPT_EXPLICIT 0x0008
1594 #define MOPT_CLEAR_ERR 0x0010
1595 #define MOPT_GTE0 0x0020
1598 #define MOPT_QFMT 0x0040
1600 #define MOPT_Q MOPT_NOSUPPORT
1601 #define MOPT_QFMT MOPT_NOSUPPORT
1603 #define MOPT_DATAJ 0x0080
1604 #define MOPT_NO_EXT2 0x0100
1605 #define MOPT_NO_EXT3 0x0200
1606 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1607 #define MOPT_STRING 0x0400
1609 static const struct mount_opts
{
1613 } ext4_mount_opts
[] = {
1614 {Opt_minix_df
, EXT4_MOUNT_MINIX_DF
, MOPT_SET
},
1615 {Opt_bsd_df
, EXT4_MOUNT_MINIX_DF
, MOPT_CLEAR
},
1616 {Opt_grpid
, EXT4_MOUNT_GRPID
, MOPT_SET
},
1617 {Opt_nogrpid
, EXT4_MOUNT_GRPID
, MOPT_CLEAR
},
1618 {Opt_block_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_SET
},
1619 {Opt_noblock_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_CLEAR
},
1620 {Opt_dioread_nolock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1621 MOPT_EXT4_ONLY
| MOPT_SET
},
1622 {Opt_dioread_lock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1623 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1624 {Opt_discard
, EXT4_MOUNT_DISCARD
, MOPT_SET
},
1625 {Opt_nodiscard
, EXT4_MOUNT_DISCARD
, MOPT_CLEAR
},
1626 {Opt_delalloc
, EXT4_MOUNT_DELALLOC
,
1627 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1628 {Opt_nodelalloc
, EXT4_MOUNT_DELALLOC
,
1629 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1630 {Opt_nojournal_checksum
, EXT4_MOUNT_JOURNAL_CHECKSUM
,
1631 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1632 {Opt_journal_checksum
, EXT4_MOUNT_JOURNAL_CHECKSUM
,
1633 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1634 {Opt_journal_async_commit
, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT
|
1635 EXT4_MOUNT_JOURNAL_CHECKSUM
),
1636 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1637 {Opt_noload
, EXT4_MOUNT_NOLOAD
, MOPT_NO_EXT2
| MOPT_SET
},
1638 {Opt_err_panic
, EXT4_MOUNT_ERRORS_PANIC
, MOPT_SET
| MOPT_CLEAR_ERR
},
1639 {Opt_err_ro
, EXT4_MOUNT_ERRORS_RO
, MOPT_SET
| MOPT_CLEAR_ERR
},
1640 {Opt_err_cont
, EXT4_MOUNT_ERRORS_CONT
, MOPT_SET
| MOPT_CLEAR_ERR
},
1641 {Opt_data_err_abort
, EXT4_MOUNT_DATA_ERR_ABORT
,
1643 {Opt_data_err_ignore
, EXT4_MOUNT_DATA_ERR_ABORT
,
1645 {Opt_barrier
, EXT4_MOUNT_BARRIER
, MOPT_SET
},
1646 {Opt_nobarrier
, EXT4_MOUNT_BARRIER
, MOPT_CLEAR
},
1647 {Opt_noauto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_SET
},
1648 {Opt_auto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_CLEAR
},
1649 {Opt_noinit_itable
, EXT4_MOUNT_INIT_INODE_TABLE
, MOPT_CLEAR
},
1650 {Opt_commit
, 0, MOPT_GTE0
},
1651 {Opt_max_batch_time
, 0, MOPT_GTE0
},
1652 {Opt_min_batch_time
, 0, MOPT_GTE0
},
1653 {Opt_inode_readahead_blks
, 0, MOPT_GTE0
},
1654 {Opt_init_itable
, 0, MOPT_GTE0
},
1655 {Opt_dax
, EXT4_MOUNT_DAX
, MOPT_SET
},
1656 {Opt_stripe
, 0, MOPT_GTE0
},
1657 {Opt_resuid
, 0, MOPT_GTE0
},
1658 {Opt_resgid
, 0, MOPT_GTE0
},
1659 {Opt_journal_dev
, 0, MOPT_NO_EXT2
| MOPT_GTE0
},
1660 {Opt_journal_path
, 0, MOPT_NO_EXT2
| MOPT_STRING
},
1661 {Opt_journal_ioprio
, 0, MOPT_NO_EXT2
| MOPT_GTE0
},
1662 {Opt_data_journal
, EXT4_MOUNT_JOURNAL_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1663 {Opt_data_ordered
, EXT4_MOUNT_ORDERED_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1664 {Opt_data_writeback
, EXT4_MOUNT_WRITEBACK_DATA
,
1665 MOPT_NO_EXT2
| MOPT_DATAJ
},
1666 {Opt_user_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_SET
},
1667 {Opt_nouser_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_CLEAR
},
1668 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1669 {Opt_acl
, EXT4_MOUNT_POSIX_ACL
, MOPT_SET
},
1670 {Opt_noacl
, EXT4_MOUNT_POSIX_ACL
, MOPT_CLEAR
},
1672 {Opt_acl
, 0, MOPT_NOSUPPORT
},
1673 {Opt_noacl
, 0, MOPT_NOSUPPORT
},
1675 {Opt_nouid32
, EXT4_MOUNT_NO_UID32
, MOPT_SET
},
1676 {Opt_debug
, EXT4_MOUNT_DEBUG
, MOPT_SET
},
1677 {Opt_debug_want_extra_isize
, 0, MOPT_GTE0
},
1678 {Opt_quota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
, MOPT_SET
| MOPT_Q
},
1679 {Opt_usrquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
,
1681 {Opt_grpquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_GRPQUOTA
,
1683 {Opt_prjquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_PRJQUOTA
,
1685 {Opt_noquota
, (EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
|
1686 EXT4_MOUNT_GRPQUOTA
| EXT4_MOUNT_PRJQUOTA
),
1687 MOPT_CLEAR
| MOPT_Q
},
1688 {Opt_usrjquota
, 0, MOPT_Q
},
1689 {Opt_grpjquota
, 0, MOPT_Q
},
1690 {Opt_offusrjquota
, 0, MOPT_Q
},
1691 {Opt_offgrpjquota
, 0, MOPT_Q
},
1692 {Opt_jqfmt_vfsold
, QFMT_VFS_OLD
, MOPT_QFMT
},
1693 {Opt_jqfmt_vfsv0
, QFMT_VFS_V0
, MOPT_QFMT
},
1694 {Opt_jqfmt_vfsv1
, QFMT_VFS_V1
, MOPT_QFMT
},
1695 {Opt_max_dir_size_kb
, 0, MOPT_GTE0
},
1696 {Opt_test_dummy_encryption
, 0, MOPT_GTE0
},
1697 {Opt_nombcache
, EXT4_MOUNT_NO_MBCACHE
, MOPT_SET
},
1701 static int handle_mount_opt(struct super_block
*sb
, char *opt
, int token
,
1702 substring_t
*args
, unsigned long *journal_devnum
,
1703 unsigned int *journal_ioprio
, int is_remount
)
1705 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1706 const struct mount_opts
*m
;
1712 if (token
== Opt_usrjquota
)
1713 return set_qf_name(sb
, USRQUOTA
, &args
[0]);
1714 else if (token
== Opt_grpjquota
)
1715 return set_qf_name(sb
, GRPQUOTA
, &args
[0]);
1716 else if (token
== Opt_offusrjquota
)
1717 return clear_qf_name(sb
, USRQUOTA
);
1718 else if (token
== Opt_offgrpjquota
)
1719 return clear_qf_name(sb
, GRPQUOTA
);
1723 case Opt_nouser_xattr
:
1724 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, opt
, "3.5");
1727 return 1; /* handled by get_sb_block() */
1729 ext4_msg(sb
, KERN_WARNING
, "Ignoring removed %s option", opt
);
1732 sbi
->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
1735 sb
->s_flags
|= SB_I_VERSION
;
1738 sb
->s_flags
|= SB_LAZYTIME
;
1740 case Opt_nolazytime
:
1741 sb
->s_flags
&= ~SB_LAZYTIME
;
1745 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++)
1746 if (token
== m
->token
)
1749 if (m
->token
== Opt_err
) {
1750 ext4_msg(sb
, KERN_ERR
, "Unrecognized mount option \"%s\" "
1751 "or missing value", opt
);
1755 if ((m
->flags
& MOPT_NO_EXT2
) && IS_EXT2_SB(sb
)) {
1756 ext4_msg(sb
, KERN_ERR
,
1757 "Mount option \"%s\" incompatible with ext2", opt
);
1760 if ((m
->flags
& MOPT_NO_EXT3
) && IS_EXT3_SB(sb
)) {
1761 ext4_msg(sb
, KERN_ERR
,
1762 "Mount option \"%s\" incompatible with ext3", opt
);
1766 if (token
== Opt_err_panic
&& !capable(CAP_SYS_ADMIN
)) {
1767 ext4_msg(sb
, KERN_ERR
,
1768 "Mount option \"%s\" not allowed for unprivileged mounts",
1773 if (args
->from
&& !(m
->flags
& MOPT_STRING
) && match_int(args
, &arg
))
1775 if (args
->from
&& (m
->flags
& MOPT_GTE0
) && (arg
< 0))
1777 if (m
->flags
& MOPT_EXPLICIT
) {
1778 if (m
->mount_opt
& EXT4_MOUNT_DELALLOC
) {
1779 set_opt2(sb
, EXPLICIT_DELALLOC
);
1780 } else if (m
->mount_opt
& EXT4_MOUNT_JOURNAL_CHECKSUM
) {
1781 set_opt2(sb
, EXPLICIT_JOURNAL_CHECKSUM
);
1785 if (m
->flags
& MOPT_CLEAR_ERR
)
1786 clear_opt(sb
, ERRORS_MASK
);
1787 if (token
== Opt_noquota
&& sb_any_quota_loaded(sb
)) {
1788 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
1789 "options when quota turned on");
1793 if (m
->flags
& MOPT_NOSUPPORT
) {
1794 ext4_msg(sb
, KERN_ERR
, "%s option not supported", opt
);
1795 } else if (token
== Opt_commit
) {
1797 arg
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1798 sbi
->s_commit_interval
= HZ
* arg
;
1799 } else if (token
== Opt_debug_want_extra_isize
) {
1802 (arg
> (sbi
->s_inode_size
- EXT4_GOOD_OLD_INODE_SIZE
))) {
1803 ext4_msg(sb
, KERN_ERR
,
1804 "Invalid want_extra_isize %d", arg
);
1807 sbi
->s_want_extra_isize
= arg
;
1808 } else if (token
== Opt_max_batch_time
) {
1809 sbi
->s_max_batch_time
= arg
;
1810 } else if (token
== Opt_min_batch_time
) {
1811 sbi
->s_min_batch_time
= arg
;
1812 } else if (token
== Opt_inode_readahead_blks
) {
1813 if (arg
&& (arg
> (1 << 30) || !is_power_of_2(arg
))) {
1814 ext4_msg(sb
, KERN_ERR
,
1815 "EXT4-fs: inode_readahead_blks must be "
1816 "0 or a power of 2 smaller than 2^31");
1819 sbi
->s_inode_readahead_blks
= arg
;
1820 } else if (token
== Opt_init_itable
) {
1821 set_opt(sb
, INIT_INODE_TABLE
);
1823 arg
= EXT4_DEF_LI_WAIT_MULT
;
1824 sbi
->s_li_wait_mult
= arg
;
1825 } else if (token
== Opt_max_dir_size_kb
) {
1826 sbi
->s_max_dir_size_kb
= arg
;
1827 } else if (token
== Opt_stripe
) {
1828 sbi
->s_stripe
= arg
;
1829 } else if (token
== Opt_resuid
) {
1830 uid
= make_kuid(sb
->s_user_ns
, arg
);
1831 if (!uid_valid(uid
)) {
1832 ext4_msg(sb
, KERN_ERR
, "Invalid uid value %d", arg
);
1835 sbi
->s_resuid
= uid
;
1836 } else if (token
== Opt_resgid
) {
1837 gid
= make_kgid(sb
->s_user_ns
, arg
);
1838 if (!gid_valid(gid
)) {
1839 ext4_msg(sb
, KERN_ERR
, "Invalid gid value %d", arg
);
1842 sbi
->s_resgid
= gid
;
1843 } else if (token
== Opt_journal_dev
) {
1845 ext4_msg(sb
, KERN_ERR
,
1846 "Cannot specify journal on remount");
1849 *journal_devnum
= arg
;
1850 } else if (token
== Opt_journal_path
) {
1852 struct inode
*journal_inode
;
1857 ext4_msg(sb
, KERN_ERR
,
1858 "Cannot specify journal on remount");
1861 journal_path
= match_strdup(&args
[0]);
1862 if (!journal_path
) {
1863 ext4_msg(sb
, KERN_ERR
, "error: could not dup "
1864 "journal device string");
1868 error
= kern_path(journal_path
, LOOKUP_FOLLOW
, &path
);
1870 ext4_msg(sb
, KERN_ERR
, "error: could not find "
1871 "journal device path: error %d", error
);
1872 kfree(journal_path
);
1877 * Refuse access for unprivileged mounts if the user does
1878 * not have rw access to the journal device via the supplied
1881 if (!capable(CAP_SYS_ADMIN
) &&
1882 inode_permission(d_inode(path
.dentry
), MAY_READ
|MAY_WRITE
)) {
1883 ext4_msg(sb
, KERN_ERR
,
1884 "error: Insufficient access to journal path %s",
1889 journal_inode
= d_inode(path
.dentry
);
1890 if (!S_ISBLK(journal_inode
->i_mode
)) {
1891 ext4_msg(sb
, KERN_ERR
, "error: journal path %s "
1892 "is not a block device", journal_path
);
1894 kfree(journal_path
);
1898 *journal_devnum
= new_encode_dev(journal_inode
->i_rdev
);
1900 kfree(journal_path
);
1901 } else if (token
== Opt_journal_ioprio
) {
1903 ext4_msg(sb
, KERN_ERR
, "Invalid journal IO priority"
1908 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
, arg
);
1909 } else if (token
== Opt_test_dummy_encryption
) {
1910 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1911 sbi
->s_mount_flags
|= EXT4_MF_TEST_DUMMY_ENCRYPTION
;
1912 ext4_msg(sb
, KERN_WARNING
,
1913 "Test dummy encryption mode enabled");
1915 ext4_msg(sb
, KERN_WARNING
,
1916 "Test dummy encryption mount option ignored");
1918 } else if (m
->flags
& MOPT_DATAJ
) {
1920 if (!sbi
->s_journal
)
1921 ext4_msg(sb
, KERN_WARNING
, "Remounting file system with no journal so ignoring journalled data option");
1922 else if (test_opt(sb
, DATA_FLAGS
) != m
->mount_opt
) {
1923 ext4_msg(sb
, KERN_ERR
,
1924 "Cannot change data mode on remount");
1928 clear_opt(sb
, DATA_FLAGS
);
1929 sbi
->s_mount_opt
|= m
->mount_opt
;
1932 } else if (m
->flags
& MOPT_QFMT
) {
1933 if (sb_any_quota_loaded(sb
) &&
1934 sbi
->s_jquota_fmt
!= m
->mount_opt
) {
1935 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled "
1936 "quota options when quota turned on");
1939 if (ext4_has_feature_quota(sb
)) {
1940 ext4_msg(sb
, KERN_INFO
,
1941 "Quota format mount options ignored "
1942 "when QUOTA feature is enabled");
1945 sbi
->s_jquota_fmt
= m
->mount_opt
;
1947 } else if (token
== Opt_dax
) {
1948 #ifdef CONFIG_FS_DAX
1949 ext4_msg(sb
, KERN_WARNING
,
1950 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1951 sbi
->s_mount_opt
|= m
->mount_opt
;
1953 ext4_msg(sb
, KERN_INFO
, "dax option not supported");
1956 } else if (token
== Opt_data_err_abort
) {
1957 sbi
->s_mount_opt
|= m
->mount_opt
;
1958 } else if (token
== Opt_data_err_ignore
) {
1959 sbi
->s_mount_opt
&= ~m
->mount_opt
;
1963 if (m
->flags
& MOPT_CLEAR
)
1965 else if (unlikely(!(m
->flags
& MOPT_SET
))) {
1966 ext4_msg(sb
, KERN_WARNING
,
1967 "buggy handling of option %s", opt
);
1972 sbi
->s_mount_opt
|= m
->mount_opt
;
1974 sbi
->s_mount_opt
&= ~m
->mount_opt
;
1979 static int parse_options(char *options
, struct super_block
*sb
,
1980 unsigned long *journal_devnum
,
1981 unsigned int *journal_ioprio
,
1984 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1985 char *p
, __maybe_unused
*usr_qf_name
, __maybe_unused
*grp_qf_name
;
1986 substring_t args
[MAX_OPT_ARGS
];
1992 while ((p
= strsep(&options
, ",")) != NULL
) {
1996 * Initialize args struct so we know whether arg was
1997 * found; some options take optional arguments.
1999 args
[0].to
= args
[0].from
= NULL
;
2000 token
= match_token(p
, tokens
, args
);
2001 if (handle_mount_opt(sb
, p
, token
, args
, journal_devnum
,
2002 journal_ioprio
, is_remount
) < 0)
2007 * We do the test below only for project quotas. 'usrquota' and
2008 * 'grpquota' mount options are allowed even without quota feature
2009 * to support legacy quotas in quota files.
2011 if (test_opt(sb
, PRJQUOTA
) && !ext4_has_feature_project(sb
)) {
2012 ext4_msg(sb
, KERN_ERR
, "Project quota feature not enabled. "
2013 "Cannot enable project quota enforcement.");
2016 usr_qf_name
= get_qf_name(sb
, sbi
, USRQUOTA
);
2017 grp_qf_name
= get_qf_name(sb
, sbi
, GRPQUOTA
);
2018 if (usr_qf_name
|| grp_qf_name
) {
2019 if (test_opt(sb
, USRQUOTA
) && usr_qf_name
)
2020 clear_opt(sb
, USRQUOTA
);
2022 if (test_opt(sb
, GRPQUOTA
) && grp_qf_name
)
2023 clear_opt(sb
, GRPQUOTA
);
2025 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
2026 ext4_msg(sb
, KERN_ERR
, "old and new quota "
2031 if (!sbi
->s_jquota_fmt
) {
2032 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
2038 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
2040 BLOCK_SIZE
<< le32_to_cpu(sbi
->s_es
->s_log_block_size
);
2042 if (blocksize
< PAGE_SIZE
) {
2043 ext4_msg(sb
, KERN_ERR
, "can't mount with "
2044 "dioread_nolock if block size != PAGE_SIZE");
2051 static inline void ext4_show_quota_options(struct seq_file
*seq
,
2052 struct super_block
*sb
)
2054 #if defined(CONFIG_QUOTA)
2055 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2056 char *usr_qf_name
, *grp_qf_name
;
2058 if (sbi
->s_jquota_fmt
) {
2061 switch (sbi
->s_jquota_fmt
) {
2072 seq_printf(seq
, ",jqfmt=%s", fmtname
);
2076 usr_qf_name
= rcu_dereference(sbi
->s_qf_names
[USRQUOTA
]);
2077 grp_qf_name
= rcu_dereference(sbi
->s_qf_names
[GRPQUOTA
]);
2079 seq_show_option(seq
, "usrjquota", usr_qf_name
);
2081 seq_show_option(seq
, "grpjquota", grp_qf_name
);
2086 static const char *token2str(int token
)
2088 const struct match_token
*t
;
2090 for (t
= tokens
; t
->token
!= Opt_err
; t
++)
2091 if (t
->token
== token
&& !strchr(t
->pattern
, '='))
2098 * - it's set to a non-default value OR
2099 * - if the per-sb default is different from the global default
2101 static int _ext4_show_options(struct seq_file
*seq
, struct super_block
*sb
,
2104 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2105 struct ext4_super_block
*es
= sbi
->s_es
;
2106 int def_errors
, def_mount_opt
= nodefs
? 0 : sbi
->s_def_mount_opt
;
2107 const struct mount_opts
*m
;
2108 char sep
= nodefs
? '\n' : ',';
2110 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2111 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2113 if (sbi
->s_sb_block
!= 1)
2114 SEQ_OPTS_PRINT("sb=%llu", sbi
->s_sb_block
);
2116 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++) {
2117 int want_set
= m
->flags
& MOPT_SET
;
2118 if (((m
->flags
& (MOPT_SET
|MOPT_CLEAR
)) == 0) ||
2119 (m
->flags
& MOPT_CLEAR_ERR
))
2121 if (!(m
->mount_opt
& (sbi
->s_mount_opt
^ def_mount_opt
)))
2122 continue; /* skip if same as the default */
2124 (sbi
->s_mount_opt
& m
->mount_opt
) != m
->mount_opt
) ||
2125 (!want_set
&& (sbi
->s_mount_opt
& m
->mount_opt
)))
2126 continue; /* select Opt_noFoo vs Opt_Foo */
2127 SEQ_OPTS_PRINT("%s", token2str(m
->token
));
2130 if (nodefs
|| !uid_eq(sbi
->s_resuid
, make_kuid(sb
->s_user_ns
, EXT4_DEF_RESUID
)) ||
2131 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
)
2132 SEQ_OPTS_PRINT("resuid=%u",
2133 from_kuid_munged(sb
->s_user_ns
, sbi
->s_resuid
));
2134 if (nodefs
|| !gid_eq(sbi
->s_resgid
, make_kgid(sb
->s_user_ns
, EXT4_DEF_RESGID
)) ||
2135 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
)
2136 SEQ_OPTS_PRINT("resgid=%u",
2137 from_kgid_munged(sb
->s_user_ns
, sbi
->s_resgid
));
2138 def_errors
= nodefs
? -1 : le16_to_cpu(es
->s_errors
);
2139 if (test_opt(sb
, ERRORS_RO
) && def_errors
!= EXT4_ERRORS_RO
)
2140 SEQ_OPTS_PUTS("errors=remount-ro");
2141 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
2142 SEQ_OPTS_PUTS("errors=continue");
2143 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
2144 SEQ_OPTS_PUTS("errors=panic");
2145 if (nodefs
|| sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
)
2146 SEQ_OPTS_PRINT("commit=%lu", sbi
->s_commit_interval
/ HZ
);
2147 if (nodefs
|| sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
)
2148 SEQ_OPTS_PRINT("min_batch_time=%u", sbi
->s_min_batch_time
);
2149 if (nodefs
|| sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
)
2150 SEQ_OPTS_PRINT("max_batch_time=%u", sbi
->s_max_batch_time
);
2151 if (sb
->s_flags
& SB_I_VERSION
)
2152 SEQ_OPTS_PUTS("i_version");
2153 if (nodefs
|| sbi
->s_stripe
)
2154 SEQ_OPTS_PRINT("stripe=%lu", sbi
->s_stripe
);
2155 if (EXT4_MOUNT_DATA_FLAGS
& (sbi
->s_mount_opt
^ def_mount_opt
)) {
2156 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
2157 SEQ_OPTS_PUTS("data=journal");
2158 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
2159 SEQ_OPTS_PUTS("data=ordered");
2160 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
2161 SEQ_OPTS_PUTS("data=writeback");
2164 sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
2165 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2166 sbi
->s_inode_readahead_blks
);
2168 if (nodefs
|| (test_opt(sb
, INIT_INODE_TABLE
) &&
2169 (sbi
->s_li_wait_mult
!= EXT4_DEF_LI_WAIT_MULT
)))
2170 SEQ_OPTS_PRINT("init_itable=%u", sbi
->s_li_wait_mult
);
2171 if (nodefs
|| sbi
->s_max_dir_size_kb
)
2172 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi
->s_max_dir_size_kb
);
2173 if (test_opt(sb
, DATA_ERR_ABORT
))
2174 SEQ_OPTS_PUTS("data_err=abort");
2175 if (DUMMY_ENCRYPTION_ENABLED(sbi
))
2176 SEQ_OPTS_PUTS("test_dummy_encryption");
2178 ext4_show_quota_options(seq
, sb
);
2182 static int ext4_show_options(struct seq_file
*seq
, struct dentry
*root
)
2184 return _ext4_show_options(seq
, root
->d_sb
, 0);
2187 int ext4_seq_options_show(struct seq_file
*seq
, void *offset
)
2189 struct super_block
*sb
= seq
->private;
2192 seq_puts(seq
, sb_rdonly(sb
) ? "ro" : "rw");
2193 rc
= _ext4_show_options(seq
, sb
, 1);
2194 seq_puts(seq
, "\n");
2198 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
2201 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2204 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
2205 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
2206 "forcing read-only mode");
2211 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
2212 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
2213 "running e2fsck is recommended");
2214 else if (sbi
->s_mount_state
& EXT4_ERROR_FS
)
2215 ext4_msg(sb
, KERN_WARNING
,
2216 "warning: mounting fs with errors, "
2217 "running e2fsck is recommended");
2218 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) > 0 &&
2219 le16_to_cpu(es
->s_mnt_count
) >=
2220 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
2221 ext4_msg(sb
, KERN_WARNING
,
2222 "warning: maximal mount count reached, "
2223 "running e2fsck is recommended");
2224 else if (le32_to_cpu(es
->s_checkinterval
) &&
2225 (le32_to_cpu(es
->s_lastcheck
) +
2226 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
2227 ext4_msg(sb
, KERN_WARNING
,
2228 "warning: checktime reached, "
2229 "running e2fsck is recommended");
2230 if (!sbi
->s_journal
)
2231 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
2232 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
2233 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
2234 le16_add_cpu(&es
->s_mnt_count
, 1);
2235 es
->s_mtime
= cpu_to_le32(get_seconds());
2237 ext4_set_feature_journal_needs_recovery(sb
);
2239 ext4_commit_super(sb
, 1);
2241 if (test_opt(sb
, DEBUG
))
2242 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
2243 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2245 sbi
->s_groups_count
,
2246 EXT4_BLOCKS_PER_GROUP(sb
),
2247 EXT4_INODES_PER_GROUP(sb
),
2248 sbi
->s_mount_opt
, sbi
->s_mount_opt2
);
2250 cleancache_init_fs(sb
);
2254 int ext4_alloc_flex_bg_array(struct super_block
*sb
, ext4_group_t ngroup
)
2256 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2257 struct flex_groups
*new_groups
;
2260 if (!sbi
->s_log_groups_per_flex
)
2263 size
= ext4_flex_group(sbi
, ngroup
- 1) + 1;
2264 if (size
<= sbi
->s_flex_groups_allocated
)
2267 size
= roundup_pow_of_two(size
* sizeof(struct flex_groups
));
2268 new_groups
= kvzalloc(size
, GFP_KERNEL
);
2270 ext4_msg(sb
, KERN_ERR
, "not enough memory for %d flex groups",
2271 size
/ (int) sizeof(struct flex_groups
));
2275 if (sbi
->s_flex_groups
) {
2276 memcpy(new_groups
, sbi
->s_flex_groups
,
2277 (sbi
->s_flex_groups_allocated
*
2278 sizeof(struct flex_groups
)));
2279 kvfree(sbi
->s_flex_groups
);
2281 sbi
->s_flex_groups
= new_groups
;
2282 sbi
->s_flex_groups_allocated
= size
/ sizeof(struct flex_groups
);
2286 static int ext4_fill_flex_info(struct super_block
*sb
)
2288 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2289 struct ext4_group_desc
*gdp
= NULL
;
2290 ext4_group_t flex_group
;
2293 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
2294 if (sbi
->s_log_groups_per_flex
< 1 || sbi
->s_log_groups_per_flex
> 31) {
2295 sbi
->s_log_groups_per_flex
= 0;
2299 err
= ext4_alloc_flex_bg_array(sb
, sbi
->s_groups_count
);
2303 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2304 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2306 flex_group
= ext4_flex_group(sbi
, i
);
2307 atomic_add(ext4_free_inodes_count(sb
, gdp
),
2308 &sbi
->s_flex_groups
[flex_group
].free_inodes
);
2309 atomic64_add(ext4_free_group_clusters(sb
, gdp
),
2310 &sbi
->s_flex_groups
[flex_group
].free_clusters
);
2311 atomic_add(ext4_used_dirs_count(sb
, gdp
),
2312 &sbi
->s_flex_groups
[flex_group
].used_dirs
);
2320 static __le16
ext4_group_desc_csum(struct super_block
*sb
, __u32 block_group
,
2321 struct ext4_group_desc
*gdp
)
2323 int offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
2325 __le32 le_group
= cpu_to_le32(block_group
);
2326 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2328 if (ext4_has_metadata_csum(sbi
->s_sb
)) {
2329 /* Use new metadata_csum algorithm */
2331 __u16 dummy_csum
= 0;
2333 csum32
= ext4_chksum(sbi
, sbi
->s_csum_seed
, (__u8
*)&le_group
,
2335 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)gdp
, offset
);
2336 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)&dummy_csum
,
2337 sizeof(dummy_csum
));
2338 offset
+= sizeof(dummy_csum
);
2339 if (offset
< sbi
->s_desc_size
)
2340 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)gdp
+ offset
,
2341 sbi
->s_desc_size
- offset
);
2343 crc
= csum32
& 0xFFFF;
2347 /* old crc16 code */
2348 if (!ext4_has_feature_gdt_csum(sb
))
2351 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
2352 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
2353 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
2354 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
2355 /* for checksum of struct ext4_group_desc do the rest...*/
2356 if (ext4_has_feature_64bit(sb
) &&
2357 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
2358 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
2359 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
2363 return cpu_to_le16(crc
);
2366 int ext4_group_desc_csum_verify(struct super_block
*sb
, __u32 block_group
,
2367 struct ext4_group_desc
*gdp
)
2369 if (ext4_has_group_desc_csum(sb
) &&
2370 (gdp
->bg_checksum
!= ext4_group_desc_csum(sb
, block_group
, gdp
)))
2376 void ext4_group_desc_csum_set(struct super_block
*sb
, __u32 block_group
,
2377 struct ext4_group_desc
*gdp
)
2379 if (!ext4_has_group_desc_csum(sb
))
2381 gdp
->bg_checksum
= ext4_group_desc_csum(sb
, block_group
, gdp
);
2384 /* Called at mount-time, super-block is locked */
2385 static int ext4_check_descriptors(struct super_block
*sb
,
2386 ext4_fsblk_t sb_block
,
2387 ext4_group_t
*first_not_zeroed
)
2389 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2390 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
2391 ext4_fsblk_t last_block
;
2392 ext4_fsblk_t last_bg_block
= sb_block
+ ext4_bg_num_gdb(sb
, 0);
2393 ext4_fsblk_t block_bitmap
;
2394 ext4_fsblk_t inode_bitmap
;
2395 ext4_fsblk_t inode_table
;
2396 int flexbg_flag
= 0;
2397 ext4_group_t i
, grp
= sbi
->s_groups_count
;
2399 if (ext4_has_feature_flex_bg(sb
))
2402 ext4_debug("Checking group descriptors");
2404 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2405 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2407 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
2408 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
2410 last_block
= first_block
+
2411 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2413 if ((grp
== sbi
->s_groups_count
) &&
2414 !(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2417 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
2418 if (block_bitmap
== sb_block
) {
2419 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2420 "Block bitmap for group %u overlaps "
2425 if (block_bitmap
>= sb_block
+ 1 &&
2426 block_bitmap
<= last_bg_block
) {
2427 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2428 "Block bitmap for group %u overlaps "
2429 "block group descriptors", i
);
2433 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
2434 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2435 "Block bitmap for group %u not in group "
2436 "(block %llu)!", i
, block_bitmap
);
2439 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
2440 if (inode_bitmap
== sb_block
) {
2441 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2442 "Inode bitmap for group %u overlaps "
2447 if (inode_bitmap
>= sb_block
+ 1 &&
2448 inode_bitmap
<= last_bg_block
) {
2449 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2450 "Inode bitmap for group %u overlaps "
2451 "block group descriptors", i
);
2455 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
2456 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2457 "Inode bitmap for group %u not in group "
2458 "(block %llu)!", i
, inode_bitmap
);
2461 inode_table
= ext4_inode_table(sb
, gdp
);
2462 if (inode_table
== sb_block
) {
2463 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2464 "Inode table for group %u overlaps "
2469 if (inode_table
>= sb_block
+ 1 &&
2470 inode_table
<= last_bg_block
) {
2471 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2472 "Inode table for group %u overlaps "
2473 "block group descriptors", i
);
2477 if (inode_table
< first_block
||
2478 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
2479 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2480 "Inode table for group %u not in group "
2481 "(block %llu)!", i
, inode_table
);
2484 ext4_lock_group(sb
, i
);
2485 if (!ext4_group_desc_csum_verify(sb
, i
, gdp
)) {
2486 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2487 "Checksum for group %u failed (%u!=%u)",
2488 i
, le16_to_cpu(ext4_group_desc_csum(sb
, i
,
2489 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
2490 if (!sb_rdonly(sb
)) {
2491 ext4_unlock_group(sb
, i
);
2495 ext4_unlock_group(sb
, i
);
2497 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
2499 if (NULL
!= first_not_zeroed
)
2500 *first_not_zeroed
= grp
;
2504 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2505 * the superblock) which were deleted from all directories, but held open by
2506 * a process at the time of a crash. We walk the list and try to delete these
2507 * inodes at recovery time (only with a read-write filesystem).
2509 * In order to keep the orphan inode chain consistent during traversal (in
2510 * case of crash during recovery), we link each inode into the superblock
2511 * orphan list_head and handle it the same way as an inode deletion during
2512 * normal operation (which journals the operations for us).
2514 * We only do an iget() and an iput() on each inode, which is very safe if we
2515 * accidentally point at an in-use or already deleted inode. The worst that
2516 * can happen in this case is that we get a "bit already cleared" message from
2517 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2518 * e2fsck was run on this filesystem, and it must have already done the orphan
2519 * inode cleanup for us, so we can safely abort without any further action.
2521 static void ext4_orphan_cleanup(struct super_block
*sb
,
2522 struct ext4_super_block
*es
)
2524 unsigned int s_flags
= sb
->s_flags
;
2525 int ret
, nr_orphans
= 0, nr_truncates
= 0;
2527 int quota_update
= 0;
2530 if (!es
->s_last_orphan
) {
2531 jbd_debug(4, "no orphan inodes to clean up\n");
2535 if (bdev_read_only(sb
->s_bdev
)) {
2536 ext4_msg(sb
, KERN_ERR
, "write access "
2537 "unavailable, skipping orphan cleanup");
2541 /* Check if feature set would not allow a r/w mount */
2542 if (!ext4_feature_set_ok(sb
, 0)) {
2543 ext4_msg(sb
, KERN_INFO
, "Skipping orphan cleanup due to "
2544 "unknown ROCOMPAT features");
2548 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2549 /* don't clear list on RO mount w/ errors */
2550 if (es
->s_last_orphan
&& !(s_flags
& SB_RDONLY
)) {
2551 ext4_msg(sb
, KERN_INFO
, "Errors on filesystem, "
2552 "clearing orphan list.\n");
2553 es
->s_last_orphan
= 0;
2555 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2559 if (s_flags
& SB_RDONLY
) {
2560 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
2561 sb
->s_flags
&= ~SB_RDONLY
;
2564 /* Needed for iput() to work correctly and not trash data */
2565 sb
->s_flags
|= SB_ACTIVE
;
2568 * Turn on quotas which were not enabled for read-only mounts if
2569 * filesystem has quota feature, so that they are updated correctly.
2571 if (ext4_has_feature_quota(sb
) && (s_flags
& SB_RDONLY
)) {
2572 int ret
= ext4_enable_quotas(sb
);
2577 ext4_msg(sb
, KERN_ERR
,
2578 "Cannot turn on quotas: error %d", ret
);
2581 /* Turn on journaled quotas used for old sytle */
2582 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
2583 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
2584 int ret
= ext4_quota_on_mount(sb
, i
);
2589 ext4_msg(sb
, KERN_ERR
,
2590 "Cannot turn on journaled "
2591 "quota: type %d: error %d", i
, ret
);
2596 while (es
->s_last_orphan
) {
2597 struct inode
*inode
;
2600 * We may have encountered an error during cleanup; if
2601 * so, skip the rest.
2603 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2604 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2605 es
->s_last_orphan
= 0;
2609 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
2610 if (IS_ERR(inode
)) {
2611 es
->s_last_orphan
= 0;
2615 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
2616 dquot_initialize(inode
);
2617 if (inode
->i_nlink
) {
2618 if (test_opt(sb
, DEBUG
))
2619 ext4_msg(sb
, KERN_DEBUG
,
2620 "%s: truncating inode %lu to %lld bytes",
2621 __func__
, inode
->i_ino
, inode
->i_size
);
2622 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2623 inode
->i_ino
, inode
->i_size
);
2625 truncate_inode_pages(inode
->i_mapping
, inode
->i_size
);
2626 ret
= ext4_truncate(inode
);
2628 ext4_std_error(inode
->i_sb
, ret
);
2629 inode_unlock(inode
);
2632 if (test_opt(sb
, DEBUG
))
2633 ext4_msg(sb
, KERN_DEBUG
,
2634 "%s: deleting unreferenced inode %lu",
2635 __func__
, inode
->i_ino
);
2636 jbd_debug(2, "deleting unreferenced inode %lu\n",
2640 iput(inode
); /* The delete magic happens here! */
2643 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2646 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
2647 PLURAL(nr_orphans
));
2649 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
2650 PLURAL(nr_truncates
));
2652 /* Turn off quotas if they were enabled for orphan cleanup */
2654 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
2655 if (sb_dqopt(sb
)->files
[i
])
2656 dquot_quota_off(sb
, i
);
2660 sb
->s_flags
= s_flags
; /* Restore SB_RDONLY status */
2664 * Maximal extent format file size.
2665 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2666 * extent format containers, within a sector_t, and within i_blocks
2667 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2668 * so that won't be a limiting factor.
2670 * However there is other limiting factor. We do store extents in the form
2671 * of starting block and length, hence the resulting length of the extent
2672 * covering maximum file size must fit into on-disk format containers as
2673 * well. Given that length is always by 1 unit bigger than max unit (because
2674 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2676 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2678 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
2681 loff_t upper_limit
= MAX_LFS_FILESIZE
;
2683 /* small i_blocks in vfs inode? */
2684 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2686 * CONFIG_LBDAF is not enabled implies the inode
2687 * i_block represent total blocks in 512 bytes
2688 * 32 == size of vfs inode i_blocks * 8
2690 upper_limit
= (1LL << 32) - 1;
2692 /* total blocks in file system block size */
2693 upper_limit
>>= (blkbits
- 9);
2694 upper_limit
<<= blkbits
;
2698 * 32-bit extent-start container, ee_block. We lower the maxbytes
2699 * by one fs block, so ee_len can cover the extent of maximum file
2702 res
= (1LL << 32) - 1;
2705 /* Sanity check against vm- & vfs- imposed limits */
2706 if (res
> upper_limit
)
2713 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2714 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2715 * We need to be 1 filesystem block less than the 2^48 sector limit.
2717 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
2719 loff_t res
= EXT4_NDIR_BLOCKS
;
2722 /* This is calculated to be the largest file size for a dense, block
2723 * mapped file such that the file's total number of 512-byte sectors,
2724 * including data and all indirect blocks, does not exceed (2^48 - 1).
2726 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2727 * number of 512-byte sectors of the file.
2730 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2732 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2733 * the inode i_block field represents total file blocks in
2734 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2736 upper_limit
= (1LL << 32) - 1;
2738 /* total blocks in file system block size */
2739 upper_limit
>>= (bits
- 9);
2743 * We use 48 bit ext4_inode i_blocks
2744 * With EXT4_HUGE_FILE_FL set the i_blocks
2745 * represent total number of blocks in
2746 * file system block size
2748 upper_limit
= (1LL << 48) - 1;
2752 /* indirect blocks */
2754 /* double indirect blocks */
2755 meta_blocks
+= 1 + (1LL << (bits
-2));
2756 /* tripple indirect blocks */
2757 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
2759 upper_limit
-= meta_blocks
;
2760 upper_limit
<<= bits
;
2762 res
+= 1LL << (bits
-2);
2763 res
+= 1LL << (2*(bits
-2));
2764 res
+= 1LL << (3*(bits
-2));
2766 if (res
> upper_limit
)
2769 if (res
> MAX_LFS_FILESIZE
)
2770 res
= MAX_LFS_FILESIZE
;
2775 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
2776 ext4_fsblk_t logical_sb_block
, int nr
)
2778 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2779 ext4_group_t bg
, first_meta_bg
;
2782 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
2784 if (!ext4_has_feature_meta_bg(sb
) || nr
< first_meta_bg
)
2785 return logical_sb_block
+ nr
+ 1;
2786 bg
= sbi
->s_desc_per_block
* nr
;
2787 if (ext4_bg_has_super(sb
, bg
))
2791 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2792 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2793 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2796 if (sb
->s_blocksize
== 1024 && nr
== 0 &&
2797 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_first_data_block
) == 0)
2800 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
2804 * ext4_get_stripe_size: Get the stripe size.
2805 * @sbi: In memory super block info
2807 * If we have specified it via mount option, then
2808 * use the mount option value. If the value specified at mount time is
2809 * greater than the blocks per group use the super block value.
2810 * If the super block value is greater than blocks per group return 0.
2811 * Allocator needs it be less than blocks per group.
2814 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
2816 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
2817 unsigned long stripe_width
=
2818 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
2821 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
2822 ret
= sbi
->s_stripe
;
2823 else if (stripe_width
&& stripe_width
<= sbi
->s_blocks_per_group
)
2825 else if (stride
&& stride
<= sbi
->s_blocks_per_group
)
2831 * If the stripe width is 1, this makes no sense and
2832 * we set it to 0 to turn off stripe handling code.
2841 * Check whether this filesystem can be mounted based on
2842 * the features present and the RDONLY/RDWR mount requested.
2843 * Returns 1 if this filesystem can be mounted as requested,
2844 * 0 if it cannot be.
2846 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
2848 if (ext4_has_unknown_ext4_incompat_features(sb
)) {
2849 ext4_msg(sb
, KERN_ERR
,
2850 "Couldn't mount because of "
2851 "unsupported optional features (%x)",
2852 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2853 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2860 if (ext4_has_feature_readonly(sb
)) {
2861 ext4_msg(sb
, KERN_INFO
, "filesystem is read-only");
2862 sb
->s_flags
|= SB_RDONLY
;
2866 /* Check that feature set is OK for a read-write mount */
2867 if (ext4_has_unknown_ext4_ro_compat_features(sb
)) {
2868 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
2869 "unsupported optional features (%x)",
2870 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2871 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2875 * Large file size enabled file system can only be mounted
2876 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2878 if (ext4_has_feature_huge_file(sb
)) {
2879 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
2880 ext4_msg(sb
, KERN_ERR
, "Filesystem with huge files "
2881 "cannot be mounted RDWR without "
2886 if (ext4_has_feature_bigalloc(sb
) && !ext4_has_feature_extents(sb
)) {
2887 ext4_msg(sb
, KERN_ERR
,
2888 "Can't support bigalloc feature without "
2889 "extents feature\n");
2893 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
2894 if (!readonly
&& (ext4_has_feature_quota(sb
) ||
2895 ext4_has_feature_project(sb
))) {
2896 ext4_msg(sb
, KERN_ERR
,
2897 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
2900 #endif /* CONFIG_QUOTA */
2905 * This function is called once a day if we have errors logged
2906 * on the file system
2908 static void print_daily_error_info(struct timer_list
*t
)
2910 struct ext4_sb_info
*sbi
= from_timer(sbi
, t
, s_err_report
);
2911 struct super_block
*sb
= sbi
->s_sb
;
2912 struct ext4_super_block
*es
= sbi
->s_es
;
2914 if (es
->s_error_count
)
2915 /* fsck newer than v1.41.13 is needed to clean this condition. */
2916 ext4_msg(sb
, KERN_NOTICE
, "error count since last fsck: %u",
2917 le32_to_cpu(es
->s_error_count
));
2918 if (es
->s_first_error_time
) {
2919 printk(KERN_NOTICE
"EXT4-fs (%s): initial error at time %u: %.*s:%d",
2920 sb
->s_id
, le32_to_cpu(es
->s_first_error_time
),
2921 (int) sizeof(es
->s_first_error_func
),
2922 es
->s_first_error_func
,
2923 le32_to_cpu(es
->s_first_error_line
));
2924 if (es
->s_first_error_ino
)
2925 printk(KERN_CONT
": inode %u",
2926 le32_to_cpu(es
->s_first_error_ino
));
2927 if (es
->s_first_error_block
)
2928 printk(KERN_CONT
": block %llu", (unsigned long long)
2929 le64_to_cpu(es
->s_first_error_block
));
2930 printk(KERN_CONT
"\n");
2932 if (es
->s_last_error_time
) {
2933 printk(KERN_NOTICE
"EXT4-fs (%s): last error at time %u: %.*s:%d",
2934 sb
->s_id
, le32_to_cpu(es
->s_last_error_time
),
2935 (int) sizeof(es
->s_last_error_func
),
2936 es
->s_last_error_func
,
2937 le32_to_cpu(es
->s_last_error_line
));
2938 if (es
->s_last_error_ino
)
2939 printk(KERN_CONT
": inode %u",
2940 le32_to_cpu(es
->s_last_error_ino
));
2941 if (es
->s_last_error_block
)
2942 printk(KERN_CONT
": block %llu", (unsigned long long)
2943 le64_to_cpu(es
->s_last_error_block
));
2944 printk(KERN_CONT
"\n");
2946 mod_timer(&sbi
->s_err_report
, jiffies
+ 24*60*60*HZ
); /* Once a day */
2949 /* Find next suitable group and run ext4_init_inode_table */
2950 static int ext4_run_li_request(struct ext4_li_request
*elr
)
2952 struct ext4_group_desc
*gdp
= NULL
;
2953 ext4_group_t group
, ngroups
;
2954 struct super_block
*sb
;
2955 unsigned long timeout
= 0;
2959 ngroups
= EXT4_SB(sb
)->s_groups_count
;
2961 for (group
= elr
->lr_next_group
; group
< ngroups
; group
++) {
2962 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2968 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2972 if (group
>= ngroups
)
2977 ret
= ext4_init_inode_table(sb
, group
,
2978 elr
->lr_timeout
? 0 : 1);
2979 if (elr
->lr_timeout
== 0) {
2980 timeout
= (jiffies
- timeout
) *
2981 elr
->lr_sbi
->s_li_wait_mult
;
2982 elr
->lr_timeout
= timeout
;
2984 elr
->lr_next_sched
= jiffies
+ elr
->lr_timeout
;
2985 elr
->lr_next_group
= group
+ 1;
2991 * Remove lr_request from the list_request and free the
2992 * request structure. Should be called with li_list_mtx held
2994 static void ext4_remove_li_request(struct ext4_li_request
*elr
)
2996 struct ext4_sb_info
*sbi
;
3003 list_del(&elr
->lr_request
);
3004 sbi
->s_li_request
= NULL
;
3008 static void ext4_unregister_li_request(struct super_block
*sb
)
3010 mutex_lock(&ext4_li_mtx
);
3011 if (!ext4_li_info
) {
3012 mutex_unlock(&ext4_li_mtx
);
3016 mutex_lock(&ext4_li_info
->li_list_mtx
);
3017 ext4_remove_li_request(EXT4_SB(sb
)->s_li_request
);
3018 mutex_unlock(&ext4_li_info
->li_list_mtx
);
3019 mutex_unlock(&ext4_li_mtx
);
3022 static struct task_struct
*ext4_lazyinit_task
;
3025 * This is the function where ext4lazyinit thread lives. It walks
3026 * through the request list searching for next scheduled filesystem.
3027 * When such a fs is found, run the lazy initialization request
3028 * (ext4_rn_li_request) and keep track of the time spend in this
3029 * function. Based on that time we compute next schedule time of
3030 * the request. When walking through the list is complete, compute
3031 * next waking time and put itself into sleep.
3033 static int ext4_lazyinit_thread(void *arg
)
3035 struct ext4_lazy_init
*eli
= (struct ext4_lazy_init
*)arg
;
3036 struct list_head
*pos
, *n
;
3037 struct ext4_li_request
*elr
;
3038 unsigned long next_wakeup
, cur
;
3040 BUG_ON(NULL
== eli
);
3044 next_wakeup
= MAX_JIFFY_OFFSET
;
3046 mutex_lock(&eli
->li_list_mtx
);
3047 if (list_empty(&eli
->li_request_list
)) {
3048 mutex_unlock(&eli
->li_list_mtx
);
3051 list_for_each_safe(pos
, n
, &eli
->li_request_list
) {
3054 elr
= list_entry(pos
, struct ext4_li_request
,
3057 if (time_before(jiffies
, elr
->lr_next_sched
)) {
3058 if (time_before(elr
->lr_next_sched
, next_wakeup
))
3059 next_wakeup
= elr
->lr_next_sched
;
3062 if (down_read_trylock(&elr
->lr_super
->s_umount
)) {
3063 if (sb_start_write_trylock(elr
->lr_super
)) {
3066 * We hold sb->s_umount, sb can not
3067 * be removed from the list, it is
3068 * now safe to drop li_list_mtx
3070 mutex_unlock(&eli
->li_list_mtx
);
3071 err
= ext4_run_li_request(elr
);
3072 sb_end_write(elr
->lr_super
);
3073 mutex_lock(&eli
->li_list_mtx
);
3076 up_read((&elr
->lr_super
->s_umount
));
3078 /* error, remove the lazy_init job */
3080 ext4_remove_li_request(elr
);
3084 elr
->lr_next_sched
= jiffies
+
3086 % (EXT4_DEF_LI_MAX_START_DELAY
* HZ
));
3088 if (time_before(elr
->lr_next_sched
, next_wakeup
))
3089 next_wakeup
= elr
->lr_next_sched
;
3091 mutex_unlock(&eli
->li_list_mtx
);
3096 if ((time_after_eq(cur
, next_wakeup
)) ||
3097 (MAX_JIFFY_OFFSET
== next_wakeup
)) {
3102 schedule_timeout_interruptible(next_wakeup
- cur
);
3104 if (kthread_should_stop()) {
3105 ext4_clear_request_list();
3112 * It looks like the request list is empty, but we need
3113 * to check it under the li_list_mtx lock, to prevent any
3114 * additions into it, and of course we should lock ext4_li_mtx
3115 * to atomically free the list and ext4_li_info, because at
3116 * this point another ext4 filesystem could be registering
3119 mutex_lock(&ext4_li_mtx
);
3120 mutex_lock(&eli
->li_list_mtx
);
3121 if (!list_empty(&eli
->li_request_list
)) {
3122 mutex_unlock(&eli
->li_list_mtx
);
3123 mutex_unlock(&ext4_li_mtx
);
3126 mutex_unlock(&eli
->li_list_mtx
);
3127 kfree(ext4_li_info
);
3128 ext4_li_info
= NULL
;
3129 mutex_unlock(&ext4_li_mtx
);
3134 static void ext4_clear_request_list(void)
3136 struct list_head
*pos
, *n
;
3137 struct ext4_li_request
*elr
;
3139 mutex_lock(&ext4_li_info
->li_list_mtx
);
3140 list_for_each_safe(pos
, n
, &ext4_li_info
->li_request_list
) {
3141 elr
= list_entry(pos
, struct ext4_li_request
,
3143 ext4_remove_li_request(elr
);
3145 mutex_unlock(&ext4_li_info
->li_list_mtx
);
3148 static int ext4_run_lazyinit_thread(void)
3150 ext4_lazyinit_task
= kthread_run(ext4_lazyinit_thread
,
3151 ext4_li_info
, "ext4lazyinit");
3152 if (IS_ERR(ext4_lazyinit_task
)) {
3153 int err
= PTR_ERR(ext4_lazyinit_task
);
3154 ext4_clear_request_list();
3155 kfree(ext4_li_info
);
3156 ext4_li_info
= NULL
;
3157 printk(KERN_CRIT
"EXT4-fs: error %d creating inode table "
3158 "initialization thread\n",
3162 ext4_li_info
->li_state
|= EXT4_LAZYINIT_RUNNING
;
3167 * Check whether it make sense to run itable init. thread or not.
3168 * If there is at least one uninitialized inode table, return
3169 * corresponding group number, else the loop goes through all
3170 * groups and return total number of groups.
3172 static ext4_group_t
ext4_has_uninit_itable(struct super_block
*sb
)
3174 ext4_group_t group
, ngroups
= EXT4_SB(sb
)->s_groups_count
;
3175 struct ext4_group_desc
*gdp
= NULL
;
3177 if (!ext4_has_group_desc_csum(sb
))
3180 for (group
= 0; group
< ngroups
; group
++) {
3181 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
3185 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
3192 static int ext4_li_info_new(void)
3194 struct ext4_lazy_init
*eli
= NULL
;
3196 eli
= kzalloc(sizeof(*eli
), GFP_KERNEL
);
3200 INIT_LIST_HEAD(&eli
->li_request_list
);
3201 mutex_init(&eli
->li_list_mtx
);
3203 eli
->li_state
|= EXT4_LAZYINIT_QUIT
;
3210 static struct ext4_li_request
*ext4_li_request_new(struct super_block
*sb
,
3213 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3214 struct ext4_li_request
*elr
;
3216 elr
= kzalloc(sizeof(*elr
), GFP_KERNEL
);
3222 elr
->lr_next_group
= start
;
3225 * Randomize first schedule time of the request to
3226 * spread the inode table initialization requests
3229 elr
->lr_next_sched
= jiffies
+ (prandom_u32() %
3230 (EXT4_DEF_LI_MAX_START_DELAY
* HZ
));
3234 int ext4_register_li_request(struct super_block
*sb
,
3235 ext4_group_t first_not_zeroed
)
3237 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3238 struct ext4_li_request
*elr
= NULL
;
3239 ext4_group_t ngroups
= EXT4_SB(sb
)->s_groups_count
;
3242 mutex_lock(&ext4_li_mtx
);
3243 if (sbi
->s_li_request
!= NULL
) {
3245 * Reset timeout so it can be computed again, because
3246 * s_li_wait_mult might have changed.
3248 sbi
->s_li_request
->lr_timeout
= 0;
3252 if (first_not_zeroed
== ngroups
|| sb_rdonly(sb
) ||
3253 !test_opt(sb
, INIT_INODE_TABLE
))
3256 elr
= ext4_li_request_new(sb
, first_not_zeroed
);
3262 if (NULL
== ext4_li_info
) {
3263 ret
= ext4_li_info_new();
3268 mutex_lock(&ext4_li_info
->li_list_mtx
);
3269 list_add(&elr
->lr_request
, &ext4_li_info
->li_request_list
);
3270 mutex_unlock(&ext4_li_info
->li_list_mtx
);
3272 sbi
->s_li_request
= elr
;
3274 * set elr to NULL here since it has been inserted to
3275 * the request_list and the removal and free of it is
3276 * handled by ext4_clear_request_list from now on.
3280 if (!(ext4_li_info
->li_state
& EXT4_LAZYINIT_RUNNING
)) {
3281 ret
= ext4_run_lazyinit_thread();
3286 mutex_unlock(&ext4_li_mtx
);
3293 * We do not need to lock anything since this is called on
3296 static void ext4_destroy_lazyinit_thread(void)
3299 * If thread exited earlier
3300 * there's nothing to be done.
3302 if (!ext4_li_info
|| !ext4_lazyinit_task
)
3305 kthread_stop(ext4_lazyinit_task
);
3308 static int set_journal_csum_feature_set(struct super_block
*sb
)
3311 int compat
, incompat
;
3312 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3314 if (ext4_has_metadata_csum(sb
)) {
3315 /* journal checksum v3 */
3317 incompat
= JBD2_FEATURE_INCOMPAT_CSUM_V3
;
3319 /* journal checksum v1 */
3320 compat
= JBD2_FEATURE_COMPAT_CHECKSUM
;
3324 jbd2_journal_clear_features(sbi
->s_journal
,
3325 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3326 JBD2_FEATURE_INCOMPAT_CSUM_V3
|
3327 JBD2_FEATURE_INCOMPAT_CSUM_V2
);
3328 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
3329 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3331 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
|
3333 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
3334 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3337 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3338 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3340 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3341 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3348 * Note: calculating the overhead so we can be compatible with
3349 * historical BSD practice is quite difficult in the face of
3350 * clusters/bigalloc. This is because multiple metadata blocks from
3351 * different block group can end up in the same allocation cluster.
3352 * Calculating the exact overhead in the face of clustered allocation
3353 * requires either O(all block bitmaps) in memory or O(number of block
3354 * groups**2) in time. We will still calculate the superblock for
3355 * older file systems --- and if we come across with a bigalloc file
3356 * system with zero in s_overhead_clusters the estimate will be close to
3357 * correct especially for very large cluster sizes --- but for newer
3358 * file systems, it's better to calculate this figure once at mkfs
3359 * time, and store it in the superblock. If the superblock value is
3360 * present (even for non-bigalloc file systems), we will use it.
3362 static int count_overhead(struct super_block
*sb
, ext4_group_t grp
,
3365 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3366 struct ext4_group_desc
*gdp
;
3367 ext4_fsblk_t first_block
, last_block
, b
;
3368 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3369 int s
, j
, count
= 0;
3371 if (!ext4_has_feature_bigalloc(sb
))
3372 return (ext4_bg_has_super(sb
, grp
) + ext4_bg_num_gdb(sb
, grp
) +
3373 sbi
->s_itb_per_group
+ 2);
3375 first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
) +
3376 (grp
* EXT4_BLOCKS_PER_GROUP(sb
));
3377 last_block
= first_block
+ EXT4_BLOCKS_PER_GROUP(sb
) - 1;
3378 for (i
= 0; i
< ngroups
; i
++) {
3379 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
3380 b
= ext4_block_bitmap(sb
, gdp
);
3381 if (b
>= first_block
&& b
<= last_block
) {
3382 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3385 b
= ext4_inode_bitmap(sb
, gdp
);
3386 if (b
>= first_block
&& b
<= last_block
) {
3387 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3390 b
= ext4_inode_table(sb
, gdp
);
3391 if (b
>= first_block
&& b
+ sbi
->s_itb_per_group
<= last_block
)
3392 for (j
= 0; j
< sbi
->s_itb_per_group
; j
++, b
++) {
3393 int c
= EXT4_B2C(sbi
, b
- first_block
);
3394 ext4_set_bit(c
, buf
);
3400 if (ext4_bg_has_super(sb
, grp
)) {
3401 ext4_set_bit(s
++, buf
);
3404 j
= ext4_bg_num_gdb(sb
, grp
);
3405 if (s
+ j
> EXT4_BLOCKS_PER_GROUP(sb
)) {
3406 ext4_error(sb
, "Invalid number of block group "
3407 "descriptor blocks: %d", j
);
3408 j
= EXT4_BLOCKS_PER_GROUP(sb
) - s
;
3412 ext4_set_bit(EXT4_B2C(sbi
, s
++), buf
);
3416 return EXT4_CLUSTERS_PER_GROUP(sb
) -
3417 ext4_count_free(buf
, EXT4_CLUSTERS_PER_GROUP(sb
) / 8);
3421 * Compute the overhead and stash it in sbi->s_overhead
3423 int ext4_calculate_overhead(struct super_block
*sb
)
3425 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3426 struct ext4_super_block
*es
= sbi
->s_es
;
3427 struct inode
*j_inode
;
3428 unsigned int j_blocks
, j_inum
= le32_to_cpu(es
->s_journal_inum
);
3429 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3430 ext4_fsblk_t overhead
= 0;
3431 char *buf
= (char *) get_zeroed_page(GFP_NOFS
);
3437 * Compute the overhead (FS structures). This is constant
3438 * for a given filesystem unless the number of block groups
3439 * changes so we cache the previous value until it does.
3443 * All of the blocks before first_data_block are overhead
3445 overhead
= EXT4_B2C(sbi
, le32_to_cpu(es
->s_first_data_block
));
3448 * Add the overhead found in each block group
3450 for (i
= 0; i
< ngroups
; i
++) {
3453 blks
= count_overhead(sb
, i
, buf
);
3456 memset(buf
, 0, PAGE_SIZE
);
3461 * Add the internal journal blocks whether the journal has been
3464 if (sbi
->s_journal
&& !sbi
->journal_bdev
)
3465 overhead
+= EXT4_NUM_B2C(sbi
, sbi
->s_journal
->j_maxlen
);
3466 else if (ext4_has_feature_journal(sb
) && !sbi
->s_journal
) {
3467 j_inode
= ext4_get_journal_inode(sb
, j_inum
);
3469 j_blocks
= j_inode
->i_size
>> sb
->s_blocksize_bits
;
3470 overhead
+= EXT4_NUM_B2C(sbi
, j_blocks
);
3473 ext4_msg(sb
, KERN_ERR
, "can't get journal size");
3476 sbi
->s_overhead
= overhead
;
3478 free_page((unsigned long) buf
);
3482 static void ext4_set_resv_clusters(struct super_block
*sb
)
3484 ext4_fsblk_t resv_clusters
;
3485 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3488 * There's no need to reserve anything when we aren't using extents.
3489 * The space estimates are exact, there are no unwritten extents,
3490 * hole punching doesn't need new metadata... This is needed especially
3491 * to keep ext2/3 backward compatibility.
3493 if (!ext4_has_feature_extents(sb
))
3496 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3497 * This should cover the situations where we can not afford to run
3498 * out of space like for example punch hole, or converting
3499 * unwritten extents in delalloc path. In most cases such
3500 * allocation would require 1, or 2 blocks, higher numbers are
3503 resv_clusters
= (ext4_blocks_count(sbi
->s_es
) >>
3504 sbi
->s_cluster_bits
);
3506 do_div(resv_clusters
, 50);
3507 resv_clusters
= min_t(ext4_fsblk_t
, resv_clusters
, 4096);
3509 atomic64_set(&sbi
->s_resv_clusters
, resv_clusters
);
3512 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
3514 struct dax_device
*dax_dev
= fs_dax_get_by_bdev(sb
->s_bdev
);
3515 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
3516 struct buffer_head
*bh
;
3517 struct ext4_super_block
*es
= NULL
;
3518 struct ext4_sb_info
*sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
3520 ext4_fsblk_t sb_block
= get_sb_block(&data
);
3521 ext4_fsblk_t logical_sb_block
;
3522 unsigned long offset
= 0;
3523 unsigned long journal_devnum
= 0;
3524 unsigned long def_mount_opts
;
3528 int blocksize
, clustersize
;
3529 unsigned int db_count
;
3531 int needs_recovery
, has_huge_files
, has_bigalloc
;
3534 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
3535 ext4_group_t first_not_zeroed
;
3537 if ((data
&& !orig_data
) || !sbi
)
3540 if (!userns_mounts
&& !capable(CAP_SYS_ADMIN
)) {
3545 sbi
->s_daxdev
= dax_dev
;
3546 sbi
->s_blockgroup_lock
=
3547 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
3548 if (!sbi
->s_blockgroup_lock
)
3551 sb
->s_fs_info
= sbi
;
3553 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
3554 sbi
->s_sb_block
= sb_block
;
3555 if (sb
->s_bdev
->bd_part
)
3556 sbi
->s_sectors_written_start
=
3557 part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]);
3559 /* Cleanup superblock name */
3560 strreplace(sb
->s_id
, '/', '!');
3562 /* -EINVAL is default */
3564 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
3566 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
3571 * The ext4 superblock will not be buffer aligned for other than 1kB
3572 * block sizes. We need to calculate the offset from buffer start.
3574 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
3575 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3576 offset
= do_div(logical_sb_block
, blocksize
);
3578 logical_sb_block
= sb_block
;
3581 if (!(bh
= sb_bread_unmovable(sb
, logical_sb_block
))) {
3582 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
3586 * Note: s_es must be initialized as soon as possible because
3587 * some ext4 macro-instructions depend on its value
3589 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
3591 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
3592 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
3594 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
3596 /* Warn if metadata_csum and gdt_csum are both set. */
3597 if (ext4_has_feature_metadata_csum(sb
) &&
3598 ext4_has_feature_gdt_csum(sb
))
3599 ext4_warning(sb
, "metadata_csum and uninit_bg are "
3600 "redundant flags; please run fsck.");
3602 /* Check for a known checksum algorithm */
3603 if (!ext4_verify_csum_type(sb
, es
)) {
3604 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3605 "unknown checksum algorithm.");
3610 /* Load the checksum driver */
3611 sbi
->s_chksum_driver
= crypto_alloc_shash("crc32c", 0, 0);
3612 if (IS_ERR(sbi
->s_chksum_driver
)) {
3613 ext4_msg(sb
, KERN_ERR
, "Cannot load crc32c driver.");
3614 ret
= PTR_ERR(sbi
->s_chksum_driver
);
3615 sbi
->s_chksum_driver
= NULL
;
3619 /* Check superblock checksum */
3620 if (!ext4_superblock_csum_verify(sb
, es
)) {
3621 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3622 "invalid superblock checksum. Run e2fsck?");
3628 /* Precompute checksum seed for all metadata */
3629 if (ext4_has_feature_csum_seed(sb
))
3630 sbi
->s_csum_seed
= le32_to_cpu(es
->s_checksum_seed
);
3631 else if (ext4_has_metadata_csum(sb
) || ext4_has_feature_ea_inode(sb
))
3632 sbi
->s_csum_seed
= ext4_chksum(sbi
, ~0, es
->s_uuid
,
3633 sizeof(es
->s_uuid
));
3635 /* Set defaults before we parse the mount options */
3636 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
3637 set_opt(sb
, INIT_INODE_TABLE
);
3638 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
3640 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
)
3642 if (def_mount_opts
& EXT4_DEFM_UID16
)
3643 set_opt(sb
, NO_UID32
);
3644 /* xattr user namespace & acls are now defaulted on */
3645 set_opt(sb
, XATTR_USER
);
3646 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3647 set_opt(sb
, POSIX_ACL
);
3649 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3650 if (ext4_has_metadata_csum(sb
))
3651 set_opt(sb
, JOURNAL_CHECKSUM
);
3653 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
3654 set_opt(sb
, JOURNAL_DATA
);
3655 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
3656 set_opt(sb
, ORDERED_DATA
);
3657 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
3658 set_opt(sb
, WRITEBACK_DATA
);
3660 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
) {
3661 if (!capable(CAP_SYS_ADMIN
))
3663 set_opt(sb
, ERRORS_PANIC
);
3664 } else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
) {
3665 set_opt(sb
, ERRORS_CONT
);
3667 set_opt(sb
, ERRORS_RO
);
3669 /* block_validity enabled by default; disable with noblock_validity */
3670 set_opt(sb
, BLOCK_VALIDITY
);
3671 if (def_mount_opts
& EXT4_DEFM_DISCARD
)
3672 set_opt(sb
, DISCARD
);
3674 sbi
->s_resuid
= make_kuid(sb
->s_user_ns
, le16_to_cpu(es
->s_def_resuid
));
3675 if (!uid_valid(sbi
->s_resuid
))
3676 sbi
->s_resuid
= make_kuid(sb
->s_user_ns
, EXT4_DEF_RESUID
);
3677 sbi
->s_resgid
= make_kgid(sb
->s_user_ns
, le16_to_cpu(es
->s_def_resgid
));
3678 if (!gid_valid(sbi
->s_resgid
))
3679 sbi
->s_resgid
= make_kgid(sb
->s_user_ns
, EXT4_DEF_RESGID
);
3680 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
3681 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
3682 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
3684 if ((def_mount_opts
& EXT4_DEFM_NOBARRIER
) == 0)
3685 set_opt(sb
, BARRIER
);
3688 * enable delayed allocation by default
3689 * Use -o nodelalloc to turn it off
3691 if (!IS_EXT3_SB(sb
) && !IS_EXT2_SB(sb
) &&
3692 ((def_mount_opts
& EXT4_DEFM_NODELALLOC
) == 0))
3693 set_opt(sb
, DELALLOC
);
3696 * set default s_li_wait_mult for lazyinit, for the case there is
3697 * no mount option specified.
3699 sbi
->s_li_wait_mult
= EXT4_DEF_LI_WAIT_MULT
;
3701 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
3702 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
3703 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
3704 ext4_msg(sb
, KERN_ERR
,
3705 "Unsupported filesystem blocksize %d (%d log_block_size)",
3706 blocksize
, le32_to_cpu(es
->s_log_block_size
));
3710 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
3711 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
3712 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
3714 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
3715 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
3716 if (sbi
->s_first_ino
< EXT4_GOOD_OLD_FIRST_INO
) {
3717 ext4_msg(sb
, KERN_ERR
, "invalid first ino: %u",
3721 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
3722 (!is_power_of_2(sbi
->s_inode_size
)) ||
3723 (sbi
->s_inode_size
> blocksize
)) {
3724 ext4_msg(sb
, KERN_ERR
,
3725 "unsupported inode size: %d",
3727 ext4_msg(sb
, KERN_ERR
, "blocksize: %d", blocksize
);
3731 * i_atime_extra is the last extra field available for
3732 * [acm]times in struct ext4_inode. Checking for that
3733 * field should suffice to ensure we have extra space
3736 if (sbi
->s_inode_size
>= offsetof(struct ext4_inode
, i_atime_extra
) +
3737 sizeof(((struct ext4_inode
*)0)->i_atime_extra
)) {
3738 sb
->s_time_gran
= 1;
3740 sb
->s_time_gran
= NSEC_PER_SEC
;
3743 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
3744 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
3745 EXT4_GOOD_OLD_INODE_SIZE
;
3746 if (ext4_has_feature_extra_isize(sb
)) {
3747 unsigned v
, max
= (sbi
->s_inode_size
-
3748 EXT4_GOOD_OLD_INODE_SIZE
);
3750 v
= le16_to_cpu(es
->s_want_extra_isize
);
3752 ext4_msg(sb
, KERN_ERR
,
3753 "bad s_want_extra_isize: %d", v
);
3756 if (sbi
->s_want_extra_isize
< v
)
3757 sbi
->s_want_extra_isize
= v
;
3759 v
= le16_to_cpu(es
->s_min_extra_isize
);
3761 ext4_msg(sb
, KERN_ERR
,
3762 "bad s_min_extra_isize: %d", v
);
3765 if (sbi
->s_want_extra_isize
< v
)
3766 sbi
->s_want_extra_isize
= v
;
3770 if (sbi
->s_es
->s_mount_opts
[0]) {
3771 char *s_mount_opts
= kstrndup(sbi
->s_es
->s_mount_opts
,
3772 sizeof(sbi
->s_es
->s_mount_opts
),
3776 if (!parse_options(s_mount_opts
, sb
, &journal_devnum
,
3777 &journal_ioprio
, 0)) {
3778 ext4_msg(sb
, KERN_WARNING
,
3779 "failed to parse options in superblock: %s",
3782 kfree(s_mount_opts
);
3784 sbi
->s_def_mount_opt
= sbi
->s_mount_opt
;
3785 if (!parse_options((char *) data
, sb
, &journal_devnum
,
3786 &journal_ioprio
, 0))
3789 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
3790 printk_once(KERN_WARNING
"EXT4-fs: Warning: mounting "
3791 "with data=journal disables delayed "
3792 "allocation and O_DIRECT support!\n");
3793 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
3794 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3795 "both data=journal and delalloc");
3798 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
3799 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3800 "both data=journal and dioread_nolock");
3803 if (test_opt(sb
, DAX
)) {
3804 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3805 "both data=journal and dax");
3808 if (ext4_has_feature_encrypt(sb
)) {
3809 ext4_msg(sb
, KERN_WARNING
,
3810 "encrypted files will use data=ordered "
3811 "instead of data journaling mode");
3813 if (test_opt(sb
, DELALLOC
))
3814 clear_opt(sb
, DELALLOC
);
3816 sb
->s_iflags
|= SB_I_CGROUPWB
;
3819 sb
->s_flags
= (sb
->s_flags
& ~SB_POSIXACL
) |
3820 (test_opt(sb
, POSIX_ACL
) ? SB_POSIXACL
: 0);
3822 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
3823 (ext4_has_compat_features(sb
) ||
3824 ext4_has_ro_compat_features(sb
) ||
3825 ext4_has_incompat_features(sb
)))
3826 ext4_msg(sb
, KERN_WARNING
,
3827 "feature flags set on rev 0 fs, "
3828 "running e2fsck is recommended");
3830 if (es
->s_creator_os
== cpu_to_le32(EXT4_OS_HURD
)) {
3831 set_opt2(sb
, HURD_COMPAT
);
3832 if (ext4_has_feature_64bit(sb
)) {
3833 ext4_msg(sb
, KERN_ERR
,
3834 "The Hurd can't support 64-bit file systems");
3839 * ea_inode feature uses l_i_version field which is not
3840 * available in HURD_COMPAT mode.
3842 if (ext4_has_feature_ea_inode(sb
)) {
3843 ext4_msg(sb
, KERN_ERR
,
3844 "ea_inode feature is not supported for Hurd");
3849 if (IS_EXT2_SB(sb
)) {
3850 if (ext2_feature_set_ok(sb
))
3851 ext4_msg(sb
, KERN_INFO
, "mounting ext2 file system "
3852 "using the ext4 subsystem");
3855 * If we're probing be silent, if this looks like
3856 * it's actually an ext[34] filesystem.
3858 if (silent
&& ext4_feature_set_ok(sb
, sb_rdonly(sb
)))
3860 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext2 due "
3861 "to feature incompatibilities");
3866 if (IS_EXT3_SB(sb
)) {
3867 if (ext3_feature_set_ok(sb
))
3868 ext4_msg(sb
, KERN_INFO
, "mounting ext3 file system "
3869 "using the ext4 subsystem");
3872 * If we're probing be silent, if this looks like
3873 * it's actually an ext4 filesystem.
3875 if (silent
&& ext4_feature_set_ok(sb
, sb_rdonly(sb
)))
3877 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext3 due "
3878 "to feature incompatibilities");
3884 * Check feature flags regardless of the revision level, since we
3885 * previously didn't change the revision level when setting the flags,
3886 * so there is a chance incompat flags are set on a rev 0 filesystem.
3888 if (!ext4_feature_set_ok(sb
, (sb_rdonly(sb
))))
3891 if (le32_to_cpu(es
->s_log_block_size
) >
3892 (EXT4_MAX_BLOCK_LOG_SIZE
- EXT4_MIN_BLOCK_LOG_SIZE
)) {
3893 ext4_msg(sb
, KERN_ERR
,
3894 "Invalid log block size: %u",
3895 le32_to_cpu(es
->s_log_block_size
));
3898 if (le32_to_cpu(es
->s_log_cluster_size
) >
3899 (EXT4_MAX_CLUSTER_LOG_SIZE
- EXT4_MIN_BLOCK_LOG_SIZE
)) {
3900 ext4_msg(sb
, KERN_ERR
,
3901 "Invalid log cluster size: %u",
3902 le32_to_cpu(es
->s_log_cluster_size
));
3906 if (le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
) > (blocksize
/ 4)) {
3907 ext4_msg(sb
, KERN_ERR
,
3908 "Number of reserved GDT blocks insanely large: %d",
3909 le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
));
3913 if (sbi
->s_mount_opt
& EXT4_MOUNT_DAX
) {
3914 if (ext4_has_feature_inline_data(sb
)) {
3915 ext4_msg(sb
, KERN_ERR
, "Cannot use DAX on a filesystem"
3916 " that may contain inline data");
3917 sbi
->s_mount_opt
&= ~EXT4_MOUNT_DAX
;
3919 if (!bdev_dax_supported(sb
->s_bdev
, blocksize
)) {
3920 ext4_msg(sb
, KERN_ERR
,
3921 "DAX unsupported by block device. Turning off DAX.");
3922 sbi
->s_mount_opt
&= ~EXT4_MOUNT_DAX
;
3926 if (ext4_has_feature_encrypt(sb
) && es
->s_encryption_level
) {
3927 ext4_msg(sb
, KERN_ERR
, "Unsupported encryption level %d",
3928 es
->s_encryption_level
);
3932 if (sb
->s_blocksize
!= blocksize
) {
3933 /* Validate the filesystem blocksize */
3934 if (!sb_set_blocksize(sb
, blocksize
)) {
3935 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
3941 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3942 offset
= do_div(logical_sb_block
, blocksize
);
3943 bh
= sb_bread_unmovable(sb
, logical_sb_block
);
3945 ext4_msg(sb
, KERN_ERR
,
3946 "Can't read superblock on 2nd try");
3949 es
= (struct ext4_super_block
*)(bh
->b_data
+ offset
);
3951 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
3952 ext4_msg(sb
, KERN_ERR
,
3953 "Magic mismatch, very weird!");
3958 has_huge_files
= ext4_has_feature_huge_file(sb
);
3959 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
3961 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
3963 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
3964 if (ext4_has_feature_64bit(sb
)) {
3965 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
3966 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
3967 !is_power_of_2(sbi
->s_desc_size
)) {
3968 ext4_msg(sb
, KERN_ERR
,
3969 "unsupported descriptor size %lu",
3974 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
3976 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
3977 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
3979 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
3980 if (sbi
->s_inodes_per_block
== 0)
3982 if (sbi
->s_inodes_per_group
< sbi
->s_inodes_per_block
||
3983 sbi
->s_inodes_per_group
> blocksize
* 8) {
3984 ext4_msg(sb
, KERN_ERR
, "invalid inodes per group: %lu\n",
3985 sbi
->s_blocks_per_group
);
3988 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
3989 sbi
->s_inodes_per_block
;
3990 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
3992 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3993 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
3994 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
3996 for (i
= 0; i
< 4; i
++)
3997 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
3998 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
3999 if (ext4_has_feature_dir_index(sb
)) {
4000 i
= le32_to_cpu(es
->s_flags
);
4001 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
4002 sbi
->s_hash_unsigned
= 3;
4003 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
4004 #ifdef __CHAR_UNSIGNED__
4007 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
4008 sbi
->s_hash_unsigned
= 3;
4012 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
4017 /* Handle clustersize */
4018 clustersize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_cluster_size
);
4019 has_bigalloc
= ext4_has_feature_bigalloc(sb
);
4021 if (clustersize
< blocksize
) {
4022 ext4_msg(sb
, KERN_ERR
,
4023 "cluster size (%d) smaller than "
4024 "block size (%d)", clustersize
, blocksize
);
4027 sbi
->s_cluster_bits
= le32_to_cpu(es
->s_log_cluster_size
) -
4028 le32_to_cpu(es
->s_log_block_size
);
4029 sbi
->s_clusters_per_group
=
4030 le32_to_cpu(es
->s_clusters_per_group
);
4031 if (sbi
->s_clusters_per_group
> blocksize
* 8) {
4032 ext4_msg(sb
, KERN_ERR
,
4033 "#clusters per group too big: %lu",
4034 sbi
->s_clusters_per_group
);
4037 if (sbi
->s_blocks_per_group
!=
4038 (sbi
->s_clusters_per_group
* (clustersize
/ blocksize
))) {
4039 ext4_msg(sb
, KERN_ERR
, "blocks per group (%lu) and "
4040 "clusters per group (%lu) inconsistent",
4041 sbi
->s_blocks_per_group
,
4042 sbi
->s_clusters_per_group
);
4046 if (clustersize
!= blocksize
) {
4047 ext4_msg(sb
, KERN_ERR
,
4048 "fragment/cluster size (%d) != "
4049 "block size (%d)", clustersize
, blocksize
);
4052 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
4053 ext4_msg(sb
, KERN_ERR
,
4054 "#blocks per group too big: %lu",
4055 sbi
->s_blocks_per_group
);
4058 sbi
->s_clusters_per_group
= sbi
->s_blocks_per_group
;
4059 sbi
->s_cluster_bits
= 0;
4061 sbi
->s_cluster_ratio
= clustersize
/ blocksize
;
4063 /* Do we have standard group size of clustersize * 8 blocks ? */
4064 if (sbi
->s_blocks_per_group
== clustersize
<< 3)
4065 set_opt2(sb
, STD_GROUP_SIZE
);
4068 * Test whether we have more sectors than will fit in sector_t,
4069 * and whether the max offset is addressable by the page cache.
4071 err
= generic_check_addressable(sb
->s_blocksize_bits
,
4072 ext4_blocks_count(es
));
4074 ext4_msg(sb
, KERN_ERR
, "filesystem"
4075 " too large to mount safely on this system");
4076 if (sizeof(sector_t
) < 8)
4077 ext4_msg(sb
, KERN_WARNING
, "CONFIG_LBDAF not enabled");
4081 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
4084 /* check blocks count against device size */
4085 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
4086 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
4087 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
4088 "exceeds size of device (%llu blocks)",
4089 ext4_blocks_count(es
), blocks_count
);
4094 * It makes no sense for the first data block to be beyond the end
4095 * of the filesystem.
4097 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
4098 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data "
4099 "block %u is beyond end of filesystem (%llu)",
4100 le32_to_cpu(es
->s_first_data_block
),
4101 ext4_blocks_count(es
));
4104 if ((es
->s_first_data_block
== 0) && (es
->s_log_block_size
== 0) &&
4105 (sbi
->s_cluster_ratio
== 1)) {
4106 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data "
4107 "block is 0 with a 1k block and cluster size");
4111 blocks_count
= (ext4_blocks_count(es
) -
4112 le32_to_cpu(es
->s_first_data_block
) +
4113 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
4114 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
4115 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
4116 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %u "
4117 "(block count %llu, first data block %u, "
4118 "blocks per group %lu)", sbi
->s_groups_count
,
4119 ext4_blocks_count(es
),
4120 le32_to_cpu(es
->s_first_data_block
),
4121 EXT4_BLOCKS_PER_GROUP(sb
));
4124 sbi
->s_groups_count
= blocks_count
;
4125 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
4126 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
4127 if (((u64
)sbi
->s_groups_count
* sbi
->s_inodes_per_group
) !=
4128 le32_to_cpu(es
->s_inodes_count
)) {
4129 ext4_msg(sb
, KERN_ERR
, "inodes count not valid: %u vs %llu",
4130 le32_to_cpu(es
->s_inodes_count
),
4131 ((u64
)sbi
->s_groups_count
* sbi
->s_inodes_per_group
));
4135 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
4136 EXT4_DESC_PER_BLOCK(sb
);
4137 if (ext4_has_feature_meta_bg(sb
)) {
4138 if (le32_to_cpu(es
->s_first_meta_bg
) > db_count
) {
4139 ext4_msg(sb
, KERN_WARNING
,
4140 "first meta block group too large: %u "
4141 "(group descriptor block count %u)",
4142 le32_to_cpu(es
->s_first_meta_bg
), db_count
);
4146 sbi
->s_group_desc
= kvmalloc(db_count
*
4147 sizeof(struct buffer_head
*),
4149 if (sbi
->s_group_desc
== NULL
) {
4150 ext4_msg(sb
, KERN_ERR
, "not enough memory");
4155 bgl_lock_init(sbi
->s_blockgroup_lock
);
4157 /* Pre-read the descriptors into the buffer cache */
4158 for (i
= 0; i
< db_count
; i
++) {
4159 block
= descriptor_loc(sb
, logical_sb_block
, i
);
4160 sb_breadahead(sb
, block
);
4163 for (i
= 0; i
< db_count
; i
++) {
4164 block
= descriptor_loc(sb
, logical_sb_block
, i
);
4165 sbi
->s_group_desc
[i
] = sb_bread_unmovable(sb
, block
);
4166 if (!sbi
->s_group_desc
[i
]) {
4167 ext4_msg(sb
, KERN_ERR
,
4168 "can't read group descriptor %d", i
);
4173 sbi
->s_gdb_count
= db_count
;
4174 if (!ext4_check_descriptors(sb
, logical_sb_block
, &first_not_zeroed
)) {
4175 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
4176 ret
= -EFSCORRUPTED
;
4180 timer_setup(&sbi
->s_err_report
, print_daily_error_info
, 0);
4182 /* Register extent status tree shrinker */
4183 if (ext4_es_register_shrinker(sbi
))
4186 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
4187 sbi
->s_extent_max_zeroout_kb
= 32;
4190 * set up enough so that it can read an inode
4192 sb
->s_op
= &ext4_sops
;
4193 sb
->s_export_op
= &ext4_export_ops
;
4194 sb
->s_xattr
= ext4_xattr_handlers
;
4195 #ifdef CONFIG_EXT4_FS_ENCRYPTION
4196 sb
->s_cop
= &ext4_cryptops
;
4199 sb
->dq_op
= &ext4_quota_operations
;
4200 if (ext4_has_feature_quota(sb
))
4201 sb
->s_qcop
= &dquot_quotactl_sysfile_ops
;
4203 sb
->s_qcop
= &ext4_qctl_operations
;
4204 sb
->s_quota_types
= QTYPE_MASK_USR
| QTYPE_MASK_GRP
| QTYPE_MASK_PRJ
;
4206 memcpy(&sb
->s_uuid
, es
->s_uuid
, sizeof(es
->s_uuid
));
4208 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
4209 mutex_init(&sbi
->s_orphan_lock
);
4213 needs_recovery
= (es
->s_last_orphan
!= 0 ||
4214 ext4_has_feature_journal_needs_recovery(sb
));
4216 if (ext4_has_feature_mmp(sb
) && !sb_rdonly(sb
))
4217 if (ext4_multi_mount_protect(sb
, le64_to_cpu(es
->s_mmp_block
)))
4218 goto failed_mount3a
;
4221 * The first inode we look at is the journal inode. Don't try
4222 * root first: it may be modified in the journal!
4224 if (!test_opt(sb
, NOLOAD
) && ext4_has_feature_journal(sb
)) {
4225 err
= ext4_load_journal(sb
, es
, journal_devnum
);
4227 goto failed_mount3a
;
4228 } else if (test_opt(sb
, NOLOAD
) && !sb_rdonly(sb
) &&
4229 ext4_has_feature_journal_needs_recovery(sb
)) {
4230 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
4231 "suppressed and not mounted read-only");
4232 goto failed_mount_wq
;
4234 /* Nojournal mode, all journal mount options are illegal */
4235 if (test_opt2(sb
, EXPLICIT_JOURNAL_CHECKSUM
)) {
4236 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4237 "journal_checksum, fs mounted w/o journal");
4238 goto failed_mount_wq
;
4240 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
4241 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4242 "journal_async_commit, fs mounted w/o journal");
4243 goto failed_mount_wq
;
4245 if (sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
) {
4246 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4247 "commit=%lu, fs mounted w/o journal",
4248 sbi
->s_commit_interval
/ HZ
);
4249 goto failed_mount_wq
;
4251 if (EXT4_MOUNT_DATA_FLAGS
&
4252 (sbi
->s_mount_opt
^ sbi
->s_def_mount_opt
)) {
4253 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4254 "data=, fs mounted w/o journal");
4255 goto failed_mount_wq
;
4257 sbi
->s_def_mount_opt
&= ~EXT4_MOUNT_JOURNAL_CHECKSUM
;
4258 clear_opt(sb
, JOURNAL_CHECKSUM
);
4259 clear_opt(sb
, DATA_FLAGS
);
4260 sbi
->s_journal
= NULL
;
4265 if (ext4_has_feature_64bit(sb
) &&
4266 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
4267 JBD2_FEATURE_INCOMPAT_64BIT
)) {
4268 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
4269 goto failed_mount_wq
;
4272 if (!set_journal_csum_feature_set(sb
)) {
4273 ext4_msg(sb
, KERN_ERR
, "Failed to set journal checksum "
4275 goto failed_mount_wq
;
4278 /* We have now updated the journal if required, so we can
4279 * validate the data journaling mode. */
4280 switch (test_opt(sb
, DATA_FLAGS
)) {
4282 /* No mode set, assume a default based on the journal
4283 * capabilities: ORDERED_DATA if the journal can
4284 * cope, else JOURNAL_DATA
4286 if (jbd2_journal_check_available_features
4287 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
4288 set_opt(sb
, ORDERED_DATA
);
4290 set_opt(sb
, JOURNAL_DATA
);
4293 case EXT4_MOUNT_ORDERED_DATA
:
4294 case EXT4_MOUNT_WRITEBACK_DATA
:
4295 if (!jbd2_journal_check_available_features
4296 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
4297 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
4298 "requested data journaling mode");
4299 goto failed_mount_wq
;
4305 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
&&
4306 test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
4307 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4308 "journal_async_commit in data=ordered mode");
4309 goto failed_mount_wq
;
4312 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
4314 sbi
->s_journal
->j_commit_callback
= ext4_journal_commit_callback
;
4317 if (!test_opt(sb
, NO_MBCACHE
)) {
4318 sbi
->s_ea_block_cache
= ext4_xattr_create_cache();
4319 if (!sbi
->s_ea_block_cache
) {
4320 ext4_msg(sb
, KERN_ERR
,
4321 "Failed to create ea_block_cache");
4322 goto failed_mount_wq
;
4325 if (ext4_has_feature_ea_inode(sb
)) {
4326 sbi
->s_ea_inode_cache
= ext4_xattr_create_cache();
4327 if (!sbi
->s_ea_inode_cache
) {
4328 ext4_msg(sb
, KERN_ERR
,
4329 "Failed to create ea_inode_cache");
4330 goto failed_mount_wq
;
4335 if ((DUMMY_ENCRYPTION_ENABLED(sbi
) || ext4_has_feature_encrypt(sb
)) &&
4336 (blocksize
!= PAGE_SIZE
)) {
4337 ext4_msg(sb
, KERN_ERR
,
4338 "Unsupported blocksize for fs encryption");
4339 goto failed_mount_wq
;
4342 if (DUMMY_ENCRYPTION_ENABLED(sbi
) && !sb_rdonly(sb
) &&
4343 !ext4_has_feature_encrypt(sb
)) {
4344 ext4_set_feature_encrypt(sb
);
4345 ext4_commit_super(sb
, 1);
4349 * Get the # of file system overhead blocks from the
4350 * superblock if present.
4352 if (es
->s_overhead_clusters
)
4353 sbi
->s_overhead
= le32_to_cpu(es
->s_overhead_clusters
);
4355 err
= ext4_calculate_overhead(sb
);
4357 goto failed_mount_wq
;
4361 * The maximum number of concurrent works can be high and
4362 * concurrency isn't really necessary. Limit it to 1.
4364 EXT4_SB(sb
)->rsv_conversion_wq
=
4365 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM
| WQ_UNBOUND
, 1);
4366 if (!EXT4_SB(sb
)->rsv_conversion_wq
) {
4367 printk(KERN_ERR
"EXT4-fs: failed to create workqueue\n");
4373 * The jbd2_journal_load will have done any necessary log recovery,
4374 * so we can safely mount the rest of the filesystem now.
4377 root
= ext4_iget(sb
, EXT4_ROOT_INO
, EXT4_IGET_SPECIAL
);
4379 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
4380 ret
= PTR_ERR(root
);
4384 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
4385 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
4389 sb
->s_root
= d_make_root(root
);
4391 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
4396 if (ext4_setup_super(sb
, es
, sb_rdonly(sb
)))
4397 sb
->s_flags
|= SB_RDONLY
;
4399 ext4_set_resv_clusters(sb
);
4401 err
= ext4_setup_system_zone(sb
);
4403 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
4405 goto failed_mount4a
;
4409 err
= ext4_mb_init(sb
);
4411 ext4_msg(sb
, KERN_ERR
, "failed to initialize mballoc (%d)",
4416 block
= ext4_count_free_clusters(sb
);
4417 ext4_free_blocks_count_set(sbi
->s_es
,
4418 EXT4_C2B(sbi
, block
));
4419 ext4_superblock_csum_set(sb
);
4420 err
= percpu_counter_init(&sbi
->s_freeclusters_counter
, block
,
4423 unsigned long freei
= ext4_count_free_inodes(sb
);
4424 sbi
->s_es
->s_free_inodes_count
= cpu_to_le32(freei
);
4425 ext4_superblock_csum_set(sb
);
4426 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
, freei
,
4430 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
4431 ext4_count_dirs(sb
), GFP_KERNEL
);
4433 err
= percpu_counter_init(&sbi
->s_dirtyclusters_counter
, 0,
4436 err
= percpu_init_rwsem(&sbi
->s_journal_flag_rwsem
);
4439 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
4443 if (ext4_has_feature_flex_bg(sb
))
4444 if (!ext4_fill_flex_info(sb
)) {
4445 ext4_msg(sb
, KERN_ERR
,
4446 "unable to initialize "
4447 "flex_bg meta info!");
4451 err
= ext4_register_li_request(sb
, first_not_zeroed
);
4455 err
= ext4_register_sysfs(sb
);
4460 /* Enable quota usage during mount. */
4461 if (ext4_has_feature_quota(sb
) && !sb_rdonly(sb
)) {
4462 err
= ext4_enable_quotas(sb
);
4466 #endif /* CONFIG_QUOTA */
4468 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
4469 ext4_orphan_cleanup(sb
, es
);
4470 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
4471 if (needs_recovery
) {
4472 ext4_msg(sb
, KERN_INFO
, "recovery complete");
4473 ext4_mark_recovery_complete(sb
, es
);
4475 if (EXT4_SB(sb
)->s_journal
) {
4476 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
4477 descr
= " journalled data mode";
4478 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
4479 descr
= " ordered data mode";
4481 descr
= " writeback data mode";
4483 descr
= "out journal";
4485 if (test_opt(sb
, DISCARD
)) {
4486 struct request_queue
*q
= bdev_get_queue(sb
->s_bdev
);
4487 if (!blk_queue_discard(q
))
4488 ext4_msg(sb
, KERN_WARNING
,
4489 "mounting with \"discard\" option, but "
4490 "the device does not support discard");
4493 if (___ratelimit(&ext4_mount_msg_ratelimit
, "EXT4-fs mount"))
4494 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s. "
4495 "Opts: %.*s%s%s", descr
,
4496 (int) sizeof(sbi
->s_es
->s_mount_opts
),
4497 sbi
->s_es
->s_mount_opts
,
4498 *sbi
->s_es
->s_mount_opts
? "; " : "", orig_data
);
4500 if (es
->s_error_count
)
4501 mod_timer(&sbi
->s_err_report
, jiffies
+ 300*HZ
); /* 5 minutes */
4503 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4504 ratelimit_state_init(&sbi
->s_err_ratelimit_state
, 5 * HZ
, 10);
4505 ratelimit_state_init(&sbi
->s_warning_ratelimit_state
, 5 * HZ
, 10);
4506 ratelimit_state_init(&sbi
->s_msg_ratelimit_state
, 5 * HZ
, 10);
4513 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
4518 ext4_unregister_sysfs(sb
);
4521 ext4_unregister_li_request(sb
);
4523 ext4_mb_release(sb
);
4524 if (sbi
->s_flex_groups
)
4525 kvfree(sbi
->s_flex_groups
);
4526 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
4527 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
4528 percpu_counter_destroy(&sbi
->s_dirs_counter
);
4529 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
4530 percpu_free_rwsem(&sbi
->s_journal_flag_rwsem
);
4532 ext4_ext_release(sb
);
4533 ext4_release_system_zone(sb
);
4538 ext4_msg(sb
, KERN_ERR
, "mount failed");
4539 if (EXT4_SB(sb
)->rsv_conversion_wq
)
4540 destroy_workqueue(EXT4_SB(sb
)->rsv_conversion_wq
);
4542 if (sbi
->s_ea_inode_cache
) {
4543 ext4_xattr_destroy_cache(sbi
->s_ea_inode_cache
);
4544 sbi
->s_ea_inode_cache
= NULL
;
4546 if (sbi
->s_ea_block_cache
) {
4547 ext4_xattr_destroy_cache(sbi
->s_ea_block_cache
);
4548 sbi
->s_ea_block_cache
= NULL
;
4550 if (sbi
->s_journal
) {
4551 jbd2_journal_destroy(sbi
->s_journal
);
4552 sbi
->s_journal
= NULL
;
4555 ext4_es_unregister_shrinker(sbi
);
4557 del_timer_sync(&sbi
->s_err_report
);
4559 kthread_stop(sbi
->s_mmp_tsk
);
4561 for (i
= 0; i
< db_count
; i
++)
4562 brelse(sbi
->s_group_desc
[i
]);
4563 kvfree(sbi
->s_group_desc
);
4565 if (sbi
->s_chksum_driver
)
4566 crypto_free_shash(sbi
->s_chksum_driver
);
4568 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
4569 kfree(sbi
->s_qf_names
[i
]);
4571 ext4_blkdev_remove(sbi
);
4574 /* sb->s_user_ns will be put when sb is destroyed */
4575 sb
->s_fs_info
= NULL
;
4576 kfree(sbi
->s_blockgroup_lock
);
4580 fs_put_dax(dax_dev
);
4581 return err
? err
: ret
;
4585 * Setup any per-fs journal parameters now. We'll do this both on
4586 * initial mount, once the journal has been initialised but before we've
4587 * done any recovery; and again on any subsequent remount.
4589 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
4591 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4593 journal
->j_commit_interval
= sbi
->s_commit_interval
;
4594 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
4595 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
4597 write_lock(&journal
->j_state_lock
);
4598 if (test_opt(sb
, BARRIER
))
4599 journal
->j_flags
|= JBD2_BARRIER
;
4601 journal
->j_flags
&= ~JBD2_BARRIER
;
4602 if (test_opt(sb
, DATA_ERR_ABORT
))
4603 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
4605 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
4606 write_unlock(&journal
->j_state_lock
);
4609 static struct inode
*ext4_get_journal_inode(struct super_block
*sb
,
4610 unsigned int journal_inum
)
4612 struct inode
*journal_inode
;
4615 * Test for the existence of a valid inode on disk. Bad things
4616 * happen if we iget() an unused inode, as the subsequent iput()
4617 * will try to delete it.
4619 journal_inode
= ext4_iget(sb
, journal_inum
, EXT4_IGET_SPECIAL
);
4620 if (IS_ERR(journal_inode
)) {
4621 ext4_msg(sb
, KERN_ERR
, "no journal found");
4624 if (!journal_inode
->i_nlink
) {
4625 make_bad_inode(journal_inode
);
4626 iput(journal_inode
);
4627 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
4631 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4632 journal_inode
, journal_inode
->i_size
);
4633 if (!S_ISREG(journal_inode
->i_mode
)) {
4634 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
4635 iput(journal_inode
);
4638 return journal_inode
;
4641 static journal_t
*ext4_get_journal(struct super_block
*sb
,
4642 unsigned int journal_inum
)
4644 struct inode
*journal_inode
;
4647 BUG_ON(!ext4_has_feature_journal(sb
));
4649 journal_inode
= ext4_get_journal_inode(sb
, journal_inum
);
4653 journal
= jbd2_journal_init_inode(journal_inode
);
4655 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
4656 iput(journal_inode
);
4659 journal
->j_private
= sb
;
4660 ext4_init_journal_params(sb
, journal
);
4664 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
4667 struct buffer_head
*bh
;
4671 int hblock
, blocksize
;
4672 ext4_fsblk_t sb_block
;
4673 unsigned long offset
;
4674 struct ext4_super_block
*es
;
4675 struct block_device
*bdev
;
4677 BUG_ON(!ext4_has_feature_journal(sb
));
4679 bdev
= ext4_blkdev_get(j_dev
, sb
);
4683 blocksize
= sb
->s_blocksize
;
4684 hblock
= bdev_logical_block_size(bdev
);
4685 if (blocksize
< hblock
) {
4686 ext4_msg(sb
, KERN_ERR
,
4687 "blocksize too small for journal device");
4691 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
4692 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
4693 set_blocksize(bdev
, blocksize
);
4694 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
4695 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
4696 "external journal");
4700 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
4701 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
4702 !(le32_to_cpu(es
->s_feature_incompat
) &
4703 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
4704 ext4_msg(sb
, KERN_ERR
, "external journal has "
4710 if ((le32_to_cpu(es
->s_feature_ro_compat
) &
4711 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
) &&
4712 es
->s_checksum
!= ext4_superblock_csum(sb
, es
)) {
4713 ext4_msg(sb
, KERN_ERR
, "external journal has "
4714 "corrupt superblock");
4719 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
4720 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
4725 len
= ext4_blocks_count(es
);
4726 start
= sb_block
+ 1;
4727 brelse(bh
); /* we're done with the superblock */
4729 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
4730 start
, len
, blocksize
);
4732 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
4735 journal
->j_private
= sb
;
4736 ll_rw_block(REQ_OP_READ
, REQ_META
| REQ_PRIO
, 1, &journal
->j_sb_buffer
);
4737 wait_on_buffer(journal
->j_sb_buffer
);
4738 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
4739 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
4742 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
4743 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
4744 "user (unsupported) - %d",
4745 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
4748 EXT4_SB(sb
)->journal_bdev
= bdev
;
4749 ext4_init_journal_params(sb
, journal
);
4753 jbd2_journal_destroy(journal
);
4755 ext4_blkdev_put(bdev
);
4759 static int ext4_load_journal(struct super_block
*sb
,
4760 struct ext4_super_block
*es
,
4761 unsigned long journal_devnum
)
4764 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
4767 int really_read_only
;
4769 BUG_ON(!ext4_has_feature_journal(sb
));
4771 if (journal_devnum
&&
4772 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4773 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
4774 "numbers have changed");
4775 journal_dev
= new_decode_dev(journal_devnum
);
4777 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
4779 really_read_only
= bdev_read_only(sb
->s_bdev
);
4782 * Are we loading a blank journal or performing recovery after a
4783 * crash? For recovery, we need to check in advance whether we
4784 * can get read-write access to the device.
4786 if (ext4_has_feature_journal_needs_recovery(sb
)) {
4787 if (sb_rdonly(sb
)) {
4788 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
4789 "required on readonly filesystem");
4790 if (really_read_only
) {
4791 ext4_msg(sb
, KERN_ERR
, "write access "
4792 "unavailable, cannot proceed "
4793 "(try mounting with noload)");
4796 ext4_msg(sb
, KERN_INFO
, "write access will "
4797 "be enabled during recovery");
4801 if (journal_inum
&& journal_dev
) {
4802 ext4_msg(sb
, KERN_ERR
, "filesystem has both journal "
4803 "and inode journals!");
4808 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
4811 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
4815 if (!(journal
->j_flags
& JBD2_BARRIER
))
4816 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
4818 if (!ext4_has_feature_journal_needs_recovery(sb
))
4819 err
= jbd2_journal_wipe(journal
, !really_read_only
);
4821 char *save
= kmalloc(EXT4_S_ERR_LEN
, GFP_KERNEL
);
4823 memcpy(save
, ((char *) es
) +
4824 EXT4_S_ERR_START
, EXT4_S_ERR_LEN
);
4825 err
= jbd2_journal_load(journal
);
4827 memcpy(((char *) es
) + EXT4_S_ERR_START
,
4828 save
, EXT4_S_ERR_LEN
);
4833 ext4_msg(sb
, KERN_ERR
, "error loading journal");
4834 jbd2_journal_destroy(journal
);
4838 EXT4_SB(sb
)->s_journal
= journal
;
4839 ext4_clear_journal_err(sb
, es
);
4841 if (!really_read_only
&& journal_devnum
&&
4842 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4843 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
4845 /* Make sure we flush the recovery flag to disk. */
4846 ext4_commit_super(sb
, 1);
4852 static int ext4_commit_super(struct super_block
*sb
, int sync
)
4854 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
4855 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
4858 if (!sbh
|| block_device_ejected(sb
))
4862 * The superblock bh should be mapped, but it might not be if the
4863 * device was hot-removed. Not much we can do but fail the I/O.
4865 if (!buffer_mapped(sbh
))
4869 * If the file system is mounted read-only, don't update the
4870 * superblock write time. This avoids updating the superblock
4871 * write time when we are mounting the root file system
4872 * read/only but we need to replay the journal; at that point,
4873 * for people who are east of GMT and who make their clock
4874 * tick in localtime for Windows bug-for-bug compatibility,
4875 * the clock is set in the future, and this will cause e2fsck
4876 * to complain and force a full file system check.
4878 if (!(sb
->s_flags
& SB_RDONLY
))
4879 es
->s_wtime
= cpu_to_le32(get_seconds());
4880 if (sb
->s_bdev
->bd_part
)
4881 es
->s_kbytes_written
=
4882 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
+
4883 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
4884 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
4886 es
->s_kbytes_written
=
4887 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
);
4888 if (percpu_counter_initialized(&EXT4_SB(sb
)->s_freeclusters_counter
))
4889 ext4_free_blocks_count_set(es
,
4890 EXT4_C2B(EXT4_SB(sb
), percpu_counter_sum_positive(
4891 &EXT4_SB(sb
)->s_freeclusters_counter
)));
4892 if (percpu_counter_initialized(&EXT4_SB(sb
)->s_freeinodes_counter
))
4893 es
->s_free_inodes_count
=
4894 cpu_to_le32(percpu_counter_sum_positive(
4895 &EXT4_SB(sb
)->s_freeinodes_counter
));
4896 BUFFER_TRACE(sbh
, "marking dirty");
4897 ext4_superblock_csum_set(sb
);
4900 if (buffer_write_io_error(sbh
) || !buffer_uptodate(sbh
)) {
4902 * Oh, dear. A previous attempt to write the
4903 * superblock failed. This could happen because the
4904 * USB device was yanked out. Or it could happen to
4905 * be a transient write error and maybe the block will
4906 * be remapped. Nothing we can do but to retry the
4907 * write and hope for the best.
4909 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
4910 "superblock detected");
4911 clear_buffer_write_io_error(sbh
);
4912 set_buffer_uptodate(sbh
);
4914 mark_buffer_dirty(sbh
);
4917 error
= __sync_dirty_buffer(sbh
,
4918 REQ_SYNC
| (test_opt(sb
, BARRIER
) ? REQ_FUA
: 0));
4922 error
= buffer_write_io_error(sbh
);
4924 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
4926 clear_buffer_write_io_error(sbh
);
4927 set_buffer_uptodate(sbh
);
4934 * Have we just finished recovery? If so, and if we are mounting (or
4935 * remounting) the filesystem readonly, then we will end up with a
4936 * consistent fs on disk. Record that fact.
4938 static void ext4_mark_recovery_complete(struct super_block
*sb
,
4939 struct ext4_super_block
*es
)
4941 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
4943 if (!ext4_has_feature_journal(sb
)) {
4944 BUG_ON(journal
!= NULL
);
4947 jbd2_journal_lock_updates(journal
);
4948 if (jbd2_journal_flush(journal
) < 0)
4951 if (ext4_has_feature_journal_needs_recovery(sb
) && sb_rdonly(sb
)) {
4952 ext4_clear_feature_journal_needs_recovery(sb
);
4953 ext4_commit_super(sb
, 1);
4957 jbd2_journal_unlock_updates(journal
);
4961 * If we are mounting (or read-write remounting) a filesystem whose journal
4962 * has recorded an error from a previous lifetime, move that error to the
4963 * main filesystem now.
4965 static void ext4_clear_journal_err(struct super_block
*sb
,
4966 struct ext4_super_block
*es
)
4972 BUG_ON(!ext4_has_feature_journal(sb
));
4974 journal
= EXT4_SB(sb
)->s_journal
;
4977 * Now check for any error status which may have been recorded in the
4978 * journal by a prior ext4_error() or ext4_abort()
4981 j_errno
= jbd2_journal_errno(journal
);
4985 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
4986 ext4_warning(sb
, "Filesystem error recorded "
4987 "from previous mount: %s", errstr
);
4988 ext4_warning(sb
, "Marking fs in need of filesystem check.");
4990 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
4991 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
4992 ext4_commit_super(sb
, 1);
4994 jbd2_journal_clear_err(journal
);
4995 jbd2_journal_update_sb_errno(journal
);
5000 * Force the running and committing transactions to commit,
5001 * and wait on the commit.
5003 int ext4_force_commit(struct super_block
*sb
)
5010 journal
= EXT4_SB(sb
)->s_journal
;
5011 return ext4_journal_force_commit(journal
);
5014 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
5018 bool needs_barrier
= false;
5019 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
5021 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb
))))
5024 trace_ext4_sync_fs(sb
, wait
);
5025 flush_workqueue(sbi
->rsv_conversion_wq
);
5027 * Writeback quota in non-journalled quota case - journalled quota has
5030 dquot_writeback_dquots(sb
, -1);
5032 * Data writeback is possible w/o journal transaction, so barrier must
5033 * being sent at the end of the function. But we can skip it if
5034 * transaction_commit will do it for us.
5036 if (sbi
->s_journal
) {
5037 target
= jbd2_get_latest_transaction(sbi
->s_journal
);
5038 if (wait
&& sbi
->s_journal
->j_flags
& JBD2_BARRIER
&&
5039 !jbd2_trans_will_send_data_barrier(sbi
->s_journal
, target
))
5040 needs_barrier
= true;
5042 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
5044 ret
= jbd2_log_wait_commit(sbi
->s_journal
,
5047 } else if (wait
&& test_opt(sb
, BARRIER
))
5048 needs_barrier
= true;
5049 if (needs_barrier
) {
5051 err
= blkdev_issue_flush(sb
->s_bdev
, GFP_KERNEL
, NULL
);
5060 * LVM calls this function before a (read-only) snapshot is created. This
5061 * gives us a chance to flush the journal completely and mark the fs clean.
5063 * Note that only this function cannot bring a filesystem to be in a clean
5064 * state independently. It relies on upper layer to stop all data & metadata
5067 static int ext4_freeze(struct super_block
*sb
)
5075 journal
= EXT4_SB(sb
)->s_journal
;
5078 /* Now we set up the journal barrier. */
5079 jbd2_journal_lock_updates(journal
);
5082 * Don't clear the needs_recovery flag if we failed to
5083 * flush the journal.
5085 error
= jbd2_journal_flush(journal
);
5089 /* Journal blocked and flushed, clear needs_recovery flag. */
5090 ext4_clear_feature_journal_needs_recovery(sb
);
5093 error
= ext4_commit_super(sb
, 1);
5096 /* we rely on upper layer to stop further updates */
5097 jbd2_journal_unlock_updates(journal
);
5102 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5103 * flag here, even though the filesystem is not technically dirty yet.
5105 static int ext4_unfreeze(struct super_block
*sb
)
5107 if (sb_rdonly(sb
) || ext4_forced_shutdown(EXT4_SB(sb
)))
5110 if (EXT4_SB(sb
)->s_journal
) {
5111 /* Reset the needs_recovery flag before the fs is unlocked. */
5112 ext4_set_feature_journal_needs_recovery(sb
);
5115 ext4_commit_super(sb
, 1);
5120 * Structure to save mount options for ext4_remount's benefit
5122 struct ext4_mount_options
{
5123 unsigned long s_mount_opt
;
5124 unsigned long s_mount_opt2
;
5127 unsigned long s_commit_interval
;
5128 u32 s_min_batch_time
, s_max_batch_time
;
5131 char *s_qf_names
[EXT4_MAXQUOTAS
];
5135 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
5137 struct ext4_super_block
*es
;
5138 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
5139 unsigned long old_sb_flags
;
5140 struct ext4_mount_options old_opts
;
5141 int enable_quota
= 0;
5143 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
5147 char *to_free
[EXT4_MAXQUOTAS
];
5149 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
5151 /* Store the original options */
5152 old_sb_flags
= sb
->s_flags
;
5153 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
5154 old_opts
.s_mount_opt2
= sbi
->s_mount_opt2
;
5155 old_opts
.s_resuid
= sbi
->s_resuid
;
5156 old_opts
.s_resgid
= sbi
->s_resgid
;
5157 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
5158 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
5159 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
5161 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
5162 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
5163 if (sbi
->s_qf_names
[i
]) {
5164 char *qf_name
= get_qf_name(sb
, sbi
, i
);
5166 old_opts
.s_qf_names
[i
] = kstrdup(qf_name
, GFP_KERNEL
);
5167 if (!old_opts
.s_qf_names
[i
]) {
5168 for (j
= 0; j
< i
; j
++)
5169 kfree(old_opts
.s_qf_names
[j
]);
5174 old_opts
.s_qf_names
[i
] = NULL
;
5176 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
5177 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
5179 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
, 1)) {
5184 if ((old_opts
.s_mount_opt
& EXT4_MOUNT_JOURNAL_CHECKSUM
) ^
5185 test_opt(sb
, JOURNAL_CHECKSUM
)) {
5186 ext4_msg(sb
, KERN_ERR
, "changing journal_checksum "
5187 "during remount not supported; ignoring");
5188 sbi
->s_mount_opt
^= EXT4_MOUNT_JOURNAL_CHECKSUM
;
5191 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
5192 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
5193 ext4_msg(sb
, KERN_ERR
, "can't mount with "
5194 "both data=journal and delalloc");
5198 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
5199 ext4_msg(sb
, KERN_ERR
, "can't mount with "
5200 "both data=journal and dioread_nolock");
5204 if (test_opt(sb
, DAX
)) {
5205 ext4_msg(sb
, KERN_ERR
, "can't mount with "
5206 "both data=journal and dax");
5210 } else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
) {
5211 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
5212 ext4_msg(sb
, KERN_ERR
, "can't mount with "
5213 "journal_async_commit in data=ordered mode");
5219 if ((sbi
->s_mount_opt
^ old_opts
.s_mount_opt
) & EXT4_MOUNT_NO_MBCACHE
) {
5220 ext4_msg(sb
, KERN_ERR
, "can't enable nombcache during remount");
5225 if ((sbi
->s_mount_opt
^ old_opts
.s_mount_opt
) & EXT4_MOUNT_DAX
) {
5226 ext4_msg(sb
, KERN_WARNING
, "warning: refusing change of "
5227 "dax flag with busy inodes while remounting");
5228 sbi
->s_mount_opt
^= EXT4_MOUNT_DAX
;
5231 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
)
5232 ext4_abort(sb
, "Abort forced by user");
5234 sb
->s_flags
= (sb
->s_flags
& ~SB_POSIXACL
) |
5235 (test_opt(sb
, POSIX_ACL
) ? SB_POSIXACL
: 0);
5239 if (sbi
->s_journal
) {
5240 ext4_init_journal_params(sb
, sbi
->s_journal
);
5241 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
5244 if (*flags
& SB_LAZYTIME
)
5245 sb
->s_flags
|= SB_LAZYTIME
;
5247 if ((bool)(*flags
& SB_RDONLY
) != sb_rdonly(sb
)) {
5248 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
) {
5253 if (*flags
& SB_RDONLY
) {
5254 err
= sync_filesystem(sb
);
5257 err
= dquot_suspend(sb
, -1);
5262 * First of all, the unconditional stuff we have to do
5263 * to disable replay of the journal when we next remount
5265 sb
->s_flags
|= SB_RDONLY
;
5268 * OK, test if we are remounting a valid rw partition
5269 * readonly, and if so set the rdonly flag and then
5270 * mark the partition as valid again.
5272 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
5273 (sbi
->s_mount_state
& EXT4_VALID_FS
))
5274 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
5277 ext4_mark_recovery_complete(sb
, es
);
5279 kthread_stop(sbi
->s_mmp_tsk
);
5281 /* Make sure we can mount this feature set readwrite */
5282 if (ext4_has_feature_readonly(sb
) ||
5283 !ext4_feature_set_ok(sb
, 0)) {
5288 * Make sure the group descriptor checksums
5289 * are sane. If they aren't, refuse to remount r/w.
5291 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
5292 struct ext4_group_desc
*gdp
=
5293 ext4_get_group_desc(sb
, g
, NULL
);
5295 if (!ext4_group_desc_csum_verify(sb
, g
, gdp
)) {
5296 ext4_msg(sb
, KERN_ERR
,
5297 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5298 g
, le16_to_cpu(ext4_group_desc_csum(sb
, g
, gdp
)),
5299 le16_to_cpu(gdp
->bg_checksum
));
5306 * If we have an unprocessed orphan list hanging
5307 * around from a previously readonly bdev mount,
5308 * require a full umount/remount for now.
5310 if (es
->s_last_orphan
) {
5311 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
5312 "remount RDWR because of unprocessed "
5313 "orphan inode list. Please "
5314 "umount/remount instead");
5320 * Mounting a RDONLY partition read-write, so reread
5321 * and store the current valid flag. (It may have
5322 * been changed by e2fsck since we originally mounted
5326 ext4_clear_journal_err(sb
, es
);
5327 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
5328 if (!ext4_setup_super(sb
, es
, 0))
5329 sb
->s_flags
&= ~SB_RDONLY
;
5330 if (ext4_has_feature_mmp(sb
))
5331 if (ext4_multi_mount_protect(sb
,
5332 le64_to_cpu(es
->s_mmp_block
))) {
5341 * Reinitialize lazy itable initialization thread based on
5344 if (sb_rdonly(sb
) || !test_opt(sb
, INIT_INODE_TABLE
))
5345 ext4_unregister_li_request(sb
);
5347 ext4_group_t first_not_zeroed
;
5348 first_not_zeroed
= ext4_has_uninit_itable(sb
);
5349 ext4_register_li_request(sb
, first_not_zeroed
);
5352 ext4_setup_system_zone(sb
);
5353 if (sbi
->s_journal
== NULL
&& !(old_sb_flags
& SB_RDONLY
))
5354 ext4_commit_super(sb
, 1);
5357 /* Release old quota file names */
5358 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
5359 kfree(old_opts
.s_qf_names
[i
]);
5361 if (sb_any_quota_suspended(sb
))
5362 dquot_resume(sb
, -1);
5363 else if (ext4_has_feature_quota(sb
)) {
5364 err
= ext4_enable_quotas(sb
);
5371 *flags
= (*flags
& ~SB_LAZYTIME
) | (sb
->s_flags
& SB_LAZYTIME
);
5372 ext4_msg(sb
, KERN_INFO
, "re-mounted. Opts: %s", orig_data
);
5377 sb
->s_flags
= old_sb_flags
;
5378 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
5379 sbi
->s_mount_opt2
= old_opts
.s_mount_opt2
;
5380 sbi
->s_resuid
= old_opts
.s_resuid
;
5381 sbi
->s_resgid
= old_opts
.s_resgid
;
5382 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
5383 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
5384 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
5386 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
5387 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
5388 to_free
[i
] = get_qf_name(sb
, sbi
, i
);
5389 rcu_assign_pointer(sbi
->s_qf_names
[i
], old_opts
.s_qf_names
[i
]);
5392 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
5400 static int ext4_statfs_project(struct super_block
*sb
,
5401 kprojid_t projid
, struct kstatfs
*buf
)
5404 struct dquot
*dquot
;
5408 qid
= make_kqid_projid(projid
);
5409 dquot
= dqget(sb
, qid
);
5411 return PTR_ERR(dquot
);
5412 spin_lock(&dquot
->dq_dqb_lock
);
5414 limit
= (dquot
->dq_dqb
.dqb_bsoftlimit
?
5415 dquot
->dq_dqb
.dqb_bsoftlimit
:
5416 dquot
->dq_dqb
.dqb_bhardlimit
) >> sb
->s_blocksize_bits
;
5417 if (limit
&& buf
->f_blocks
> limit
) {
5418 curblock
= dquot
->dq_dqb
.dqb_curspace
>> sb
->s_blocksize_bits
;
5419 buf
->f_blocks
= limit
;
5420 buf
->f_bfree
= buf
->f_bavail
=
5421 (buf
->f_blocks
> curblock
) ?
5422 (buf
->f_blocks
- curblock
) : 0;
5425 limit
= dquot
->dq_dqb
.dqb_isoftlimit
?
5426 dquot
->dq_dqb
.dqb_isoftlimit
:
5427 dquot
->dq_dqb
.dqb_ihardlimit
;
5428 if (limit
&& buf
->f_files
> limit
) {
5429 buf
->f_files
= limit
;
5431 (buf
->f_files
> dquot
->dq_dqb
.dqb_curinodes
) ?
5432 (buf
->f_files
- dquot
->dq_dqb
.dqb_curinodes
) : 0;
5435 spin_unlock(&dquot
->dq_dqb_lock
);
5441 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
5443 struct super_block
*sb
= dentry
->d_sb
;
5444 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
5445 struct ext4_super_block
*es
= sbi
->s_es
;
5446 ext4_fsblk_t overhead
= 0, resv_blocks
;
5449 resv_blocks
= EXT4_C2B(sbi
, atomic64_read(&sbi
->s_resv_clusters
));
5451 if (!test_opt(sb
, MINIX_DF
))
5452 overhead
= sbi
->s_overhead
;
5454 buf
->f_type
= EXT4_SUPER_MAGIC
;
5455 buf
->f_bsize
= sb
->s_blocksize
;
5456 buf
->f_blocks
= ext4_blocks_count(es
) - EXT4_C2B(sbi
, overhead
);
5457 bfree
= percpu_counter_sum_positive(&sbi
->s_freeclusters_counter
) -
5458 percpu_counter_sum_positive(&sbi
->s_dirtyclusters_counter
);
5459 /* prevent underflow in case that few free space is available */
5460 buf
->f_bfree
= EXT4_C2B(sbi
, max_t(s64
, bfree
, 0));
5461 buf
->f_bavail
= buf
->f_bfree
-
5462 (ext4_r_blocks_count(es
) + resv_blocks
);
5463 if (buf
->f_bfree
< (ext4_r_blocks_count(es
) + resv_blocks
))
5465 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
5466 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
5467 buf
->f_namelen
= EXT4_NAME_LEN
;
5468 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
5469 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
5470 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
5471 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
5474 if (ext4_test_inode_flag(dentry
->d_inode
, EXT4_INODE_PROJINHERIT
) &&
5475 sb_has_quota_limits_enabled(sb
, PRJQUOTA
))
5476 ext4_statfs_project(sb
, EXT4_I(dentry
->d_inode
)->i_projid
, buf
);
5485 * Helper functions so that transaction is started before we acquire dqio_sem
5486 * to keep correct lock ordering of transaction > dqio_sem
5488 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
5490 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_id
.type
];
5493 static int ext4_write_dquot(struct dquot
*dquot
)
5497 struct inode
*inode
;
5499 inode
= dquot_to_inode(dquot
);
5500 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
,
5501 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
5503 return PTR_ERR(handle
);
5504 ret
= dquot_commit(dquot
);
5505 err
= ext4_journal_stop(handle
);
5511 static int ext4_acquire_dquot(struct dquot
*dquot
)
5516 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
5517 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
5519 return PTR_ERR(handle
);
5520 ret
= dquot_acquire(dquot
);
5521 err
= ext4_journal_stop(handle
);
5527 static int ext4_release_dquot(struct dquot
*dquot
)
5532 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
5533 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
5534 if (IS_ERR(handle
)) {
5535 /* Release dquot anyway to avoid endless cycle in dqput() */
5536 dquot_release(dquot
);
5537 return PTR_ERR(handle
);
5539 ret
= dquot_release(dquot
);
5540 err
= ext4_journal_stop(handle
);
5546 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
5548 struct super_block
*sb
= dquot
->dq_sb
;
5549 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
5551 /* Are we journaling quotas? */
5552 if (ext4_has_feature_quota(sb
) ||
5553 sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
5554 dquot_mark_dquot_dirty(dquot
);
5555 return ext4_write_dquot(dquot
);
5557 return dquot_mark_dquot_dirty(dquot
);
5561 static int ext4_write_info(struct super_block
*sb
, int type
)
5566 /* Data block + inode block */
5567 handle
= ext4_journal_start(d_inode(sb
->s_root
), EXT4_HT_QUOTA
, 2);
5569 return PTR_ERR(handle
);
5570 ret
= dquot_commit_info(sb
, type
);
5571 err
= ext4_journal_stop(handle
);
5578 * Turn on quotas during mount time - we need to find
5579 * the quota file and such...
5581 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
5583 return dquot_quota_on_mount(sb
, get_qf_name(sb
, EXT4_SB(sb
), type
),
5584 EXT4_SB(sb
)->s_jquota_fmt
, type
);
5587 static void lockdep_set_quota_inode(struct inode
*inode
, int subclass
)
5589 struct ext4_inode_info
*ei
= EXT4_I(inode
);
5591 /* The first argument of lockdep_set_subclass has to be
5592 * *exactly* the same as the argument to init_rwsem() --- in
5593 * this case, in init_once() --- or lockdep gets unhappy
5594 * because the name of the lock is set using the
5595 * stringification of the argument to init_rwsem().
5597 (void) ei
; /* shut up clang warning if !CONFIG_LOCKDEP */
5598 lockdep_set_subclass(&ei
->i_data_sem
, subclass
);
5602 * Standard function to be called on quota_on
5604 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
5605 const struct path
*path
)
5609 if (!test_opt(sb
, QUOTA
))
5612 /* Quotafile not on the same filesystem? */
5613 if (path
->dentry
->d_sb
!= sb
)
5615 /* Journaling quota? */
5616 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
5617 /* Quotafile not in fs root? */
5618 if (path
->dentry
->d_parent
!= sb
->s_root
)
5619 ext4_msg(sb
, KERN_WARNING
,
5620 "Quota file not on filesystem root. "
5621 "Journaled quota will not work");
5622 sb_dqopt(sb
)->flags
|= DQUOT_NOLIST_DIRTY
;
5625 * Clear the flag just in case mount options changed since
5628 sb_dqopt(sb
)->flags
&= ~DQUOT_NOLIST_DIRTY
;
5632 * When we journal data on quota file, we have to flush journal to see
5633 * all updates to the file when we bypass pagecache...
5635 if (EXT4_SB(sb
)->s_journal
&&
5636 ext4_should_journal_data(d_inode(path
->dentry
))) {
5638 * We don't need to lock updates but journal_flush() could
5639 * otherwise be livelocked...
5641 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
5642 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
5643 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
5648 lockdep_set_quota_inode(path
->dentry
->d_inode
, I_DATA_SEM_QUOTA
);
5649 err
= dquot_quota_on(sb
, type
, format_id
, path
);
5651 lockdep_set_quota_inode(path
->dentry
->d_inode
,
5654 struct inode
*inode
= d_inode(path
->dentry
);
5658 * Set inode flags to prevent userspace from messing with quota
5659 * files. If this fails, we return success anyway since quotas
5660 * are already enabled and this is not a hard failure.
5663 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
, 1);
5666 EXT4_I(inode
)->i_flags
|= EXT4_NOATIME_FL
| EXT4_IMMUTABLE_FL
;
5667 inode_set_flags(inode
, S_NOATIME
| S_IMMUTABLE
,
5668 S_NOATIME
| S_IMMUTABLE
);
5669 ext4_mark_inode_dirty(handle
, inode
);
5670 ext4_journal_stop(handle
);
5672 inode_unlock(inode
);
5677 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
5681 struct inode
*qf_inode
;
5682 unsigned long qf_inums
[EXT4_MAXQUOTAS
] = {
5683 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
5684 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
),
5685 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_prj_quota_inum
)
5688 BUG_ON(!ext4_has_feature_quota(sb
));
5690 if (!qf_inums
[type
])
5693 qf_inode
= ext4_iget(sb
, qf_inums
[type
], EXT4_IGET_SPECIAL
);
5694 if (IS_ERR(qf_inode
)) {
5695 ext4_error(sb
, "Bad quota inode # %lu", qf_inums
[type
]);
5696 return PTR_ERR(qf_inode
);
5699 /* Don't account quota for quota files to avoid recursion */
5700 qf_inode
->i_flags
|= S_NOQUOTA
;
5701 lockdep_set_quota_inode(qf_inode
, I_DATA_SEM_QUOTA
);
5702 err
= dquot_enable(qf_inode
, type
, format_id
, flags
);
5704 lockdep_set_quota_inode(qf_inode
, I_DATA_SEM_NORMAL
);
5710 /* Enable usage tracking for all quota types. */
5711 static int ext4_enable_quotas(struct super_block
*sb
)
5714 unsigned long qf_inums
[EXT4_MAXQUOTAS
] = {
5715 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
5716 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
),
5717 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_prj_quota_inum
)
5719 bool quota_mopt
[EXT4_MAXQUOTAS
] = {
5720 test_opt(sb
, USRQUOTA
),
5721 test_opt(sb
, GRPQUOTA
),
5722 test_opt(sb
, PRJQUOTA
),
5725 sb_dqopt(sb
)->flags
|= DQUOT_QUOTA_SYS_FILE
| DQUOT_NOLIST_DIRTY
;
5726 for (type
= 0; type
< EXT4_MAXQUOTAS
; type
++) {
5727 if (qf_inums
[type
]) {
5728 err
= ext4_quota_enable(sb
, type
, QFMT_VFS_V1
,
5729 DQUOT_USAGE_ENABLED
|
5730 (quota_mopt
[type
] ? DQUOT_LIMITS_ENABLED
: 0));
5732 for (type
--; type
>= 0; type
--)
5733 dquot_quota_off(sb
, type
);
5736 "Failed to enable quota tracking "
5737 "(type=%d, err=%d). Please run "
5738 "e2fsck to fix.", type
, err
);
5746 static int ext4_quota_off(struct super_block
*sb
, int type
)
5748 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5752 /* Force all delayed allocation blocks to be allocated.
5753 * Caller already holds s_umount sem */
5754 if (test_opt(sb
, DELALLOC
))
5755 sync_filesystem(sb
);
5757 if (!inode
|| !igrab(inode
))
5760 err
= dquot_quota_off(sb
, type
);
5761 if (err
|| ext4_has_feature_quota(sb
))
5766 * Update modification times of quota files when userspace can
5767 * start looking at them. If we fail, we return success anyway since
5768 * this is not a hard failure and quotas are already disabled.
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
, 0, S_NOATIME
| S_IMMUTABLE
);
5775 inode
->i_mtime
= inode
->i_ctime
= current_time(inode
);
5776 ext4_mark_inode_dirty(handle
, inode
);
5777 ext4_journal_stop(handle
);
5779 inode_unlock(inode
);
5781 lockdep_set_quota_inode(inode
, I_DATA_SEM_NORMAL
);
5785 return dquot_quota_off(sb
, type
);
5788 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5789 * acquiring the locks... As quota files are never truncated and quota code
5790 * itself serializes the operations (and no one else should touch the files)
5791 * we don't have to be afraid of races */
5792 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
5793 size_t len
, loff_t off
)
5795 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5796 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5797 int offset
= off
& (sb
->s_blocksize
- 1);
5800 struct buffer_head
*bh
;
5801 loff_t i_size
= i_size_read(inode
);
5805 if (off
+len
> i_size
)
5808 while (toread
> 0) {
5809 tocopy
= sb
->s_blocksize
- offset
< toread
?
5810 sb
->s_blocksize
- offset
: toread
;
5811 bh
= ext4_bread(NULL
, inode
, blk
, 0);
5814 if (!bh
) /* A hole? */
5815 memset(data
, 0, tocopy
);
5817 memcpy(data
, bh
->b_data
+offset
, tocopy
);
5827 /* Write to quotafile (we know the transaction is already started and has
5828 * enough credits) */
5829 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
5830 const char *data
, size_t len
, loff_t off
)
5832 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5833 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5834 int err
, offset
= off
& (sb
->s_blocksize
- 1);
5836 struct buffer_head
*bh
;
5837 handle_t
*handle
= journal_current_handle();
5839 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
5840 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5841 " cancelled because transaction is not started",
5842 (unsigned long long)off
, (unsigned long long)len
);
5846 * Since we account only one data block in transaction credits,
5847 * then it is impossible to cross a block boundary.
5849 if (sb
->s_blocksize
- offset
< len
) {
5850 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5851 " cancelled because not block aligned",
5852 (unsigned long long)off
, (unsigned long long)len
);
5857 bh
= ext4_bread(handle
, inode
, blk
,
5858 EXT4_GET_BLOCKS_CREATE
|
5859 EXT4_GET_BLOCKS_METADATA_NOFAIL
);
5860 } while (IS_ERR(bh
) && (PTR_ERR(bh
) == -ENOSPC
) &&
5861 ext4_should_retry_alloc(inode
->i_sb
, &retries
));
5866 BUFFER_TRACE(bh
, "get write access");
5867 err
= ext4_journal_get_write_access(handle
, bh
);
5873 memcpy(bh
->b_data
+offset
, data
, len
);
5874 flush_dcache_page(bh
->b_page
);
5876 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
5879 if (inode
->i_size
< off
+ len
) {
5880 i_size_write(inode
, off
+ len
);
5881 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
5882 ext4_mark_inode_dirty(handle
, inode
);
5887 static int ext4_get_next_id(struct super_block
*sb
, struct kqid
*qid
)
5889 const struct quota_format_ops
*ops
;
5891 if (!sb_has_quota_loaded(sb
, qid
->type
))
5893 ops
= sb_dqopt(sb
)->ops
[qid
->type
];
5894 if (!ops
|| !ops
->get_next_id
)
5896 return dquot_get_next_id(sb
, qid
);
5900 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
5901 const char *dev_name
, void *data
)
5903 return mount_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
);
5906 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5907 static inline void register_as_ext2(void)
5909 int err
= register_filesystem(&ext2_fs_type
);
5912 "EXT4-fs: Unable to register as ext2 (%d)\n", err
);
5915 static inline void unregister_as_ext2(void)
5917 unregister_filesystem(&ext2_fs_type
);
5920 static inline int ext2_feature_set_ok(struct super_block
*sb
)
5922 if (ext4_has_unknown_ext2_incompat_features(sb
))
5926 if (ext4_has_unknown_ext2_ro_compat_features(sb
))
5931 static inline void register_as_ext2(void) { }
5932 static inline void unregister_as_ext2(void) { }
5933 static inline int ext2_feature_set_ok(struct super_block
*sb
) { return 0; }
5936 static inline void register_as_ext3(void)
5938 int err
= register_filesystem(&ext3_fs_type
);
5941 "EXT4-fs: Unable to register as ext3 (%d)\n", err
);
5944 static inline void unregister_as_ext3(void)
5946 unregister_filesystem(&ext3_fs_type
);
5949 static inline int ext3_feature_set_ok(struct super_block
*sb
)
5951 if (ext4_has_unknown_ext3_incompat_features(sb
))
5953 if (!ext4_has_feature_journal(sb
))
5957 if (ext4_has_unknown_ext3_ro_compat_features(sb
))
5962 static struct file_system_type ext4_fs_type
= {
5963 .owner
= THIS_MODULE
,
5965 .mount
= ext4_mount
,
5966 .kill_sb
= kill_block_super
,
5967 .fs_flags
= FS_REQUIRES_DEV
| FS_USERNS_MOUNT
,
5969 MODULE_ALIAS_FS("ext4");
5971 /* Shared across all ext4 file systems */
5972 wait_queue_head_t ext4__ioend_wq
[EXT4_WQ_HASH_SZ
];
5974 static int __init
ext4_init_fs(void)
5978 ratelimit_state_init(&ext4_mount_msg_ratelimit
, 30 * HZ
, 64);
5979 ext4_li_info
= NULL
;
5980 mutex_init(&ext4_li_mtx
);
5982 /* Build-time check for flags consistency */
5983 ext4_check_flag_values();
5985 for (i
= 0; i
< EXT4_WQ_HASH_SZ
; i
++)
5986 init_waitqueue_head(&ext4__ioend_wq
[i
]);
5988 err
= ext4_init_es();
5992 err
= ext4_init_pageio();
5996 err
= ext4_init_system_zone();
6000 err
= ext4_init_sysfs();
6004 err
= ext4_init_mballoc();
6007 err
= init_inodecache();
6012 err
= register_filesystem(&ext4_fs_type
);
6018 unregister_as_ext2();
6019 unregister_as_ext3();
6020 destroy_inodecache();
6022 ext4_exit_mballoc();
6026 ext4_exit_system_zone();
6035 static void __exit
ext4_exit_fs(void)
6037 ext4_destroy_lazyinit_thread();
6038 unregister_as_ext2();
6039 unregister_as_ext3();
6040 unregister_filesystem(&ext4_fs_type
);
6041 destroy_inodecache();
6042 ext4_exit_mballoc();
6044 ext4_exit_system_zone();
6049 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6050 MODULE_DESCRIPTION("Fourth Extended Filesystem");
6051 MODULE_LICENSE("GPL");
6052 MODULE_SOFTDEP("pre: crc32c");
6053 module_init(ext4_init_fs
)
6054 module_exit(ext4_exit_fs
)