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/cleancache.h>
41 #include <asm/uaccess.h>
43 #include <linux/kthread.h>
44 #include <linux/freezer.h>
47 #include "ext4_extents.h" /* Needed for trace points definition */
48 #include "ext4_jbd2.h"
53 #define CREATE_TRACE_POINTS
54 #include <trace/events/ext4.h>
56 static struct ext4_lazy_init
*ext4_li_info
;
57 static struct mutex ext4_li_mtx
;
58 static struct ratelimit_state ext4_mount_msg_ratelimit
;
60 static int ext4_load_journal(struct super_block
*, struct ext4_super_block
*,
61 unsigned long journal_devnum
);
62 static int ext4_show_options(struct seq_file
*seq
, struct dentry
*root
);
63 static int ext4_commit_super(struct super_block
*sb
, int sync
);
64 static void ext4_mark_recovery_complete(struct super_block
*sb
,
65 struct ext4_super_block
*es
);
66 static void ext4_clear_journal_err(struct super_block
*sb
,
67 struct ext4_super_block
*es
);
68 static int ext4_sync_fs(struct super_block
*sb
, int wait
);
69 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
);
70 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
71 static int ext4_unfreeze(struct super_block
*sb
);
72 static int ext4_freeze(struct super_block
*sb
);
73 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
74 const char *dev_name
, void *data
);
75 static inline int ext2_feature_set_ok(struct super_block
*sb
);
76 static inline int ext3_feature_set_ok(struct super_block
*sb
);
77 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
);
78 static void ext4_destroy_lazyinit_thread(void);
79 static void ext4_unregister_li_request(struct super_block
*sb
);
80 static void ext4_clear_request_list(void);
81 static struct inode
*ext4_get_journal_inode(struct super_block
*sb
,
82 unsigned int journal_inum
);
87 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
88 * i_mmap_rwsem (inode->i_mmap_rwsem)!
91 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
92 * page lock -> i_data_sem (rw)
94 * buffered write path:
95 * sb_start_write -> i_mutex -> mmap_sem
96 * sb_start_write -> i_mutex -> transaction start -> page lock ->
100 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
101 * i_mmap_rwsem (w) -> page lock
102 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
103 * transaction start -> i_data_sem (rw)
106 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
107 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
108 * transaction start -> i_data_sem (rw)
111 * transaction start -> page lock(s) -> i_data_sem (rw)
114 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
115 static struct file_system_type ext2_fs_type
= {
116 .owner
= THIS_MODULE
,
119 .kill_sb
= kill_block_super
,
120 .fs_flags
= FS_REQUIRES_DEV
,
122 MODULE_ALIAS_FS("ext2");
123 MODULE_ALIAS("ext2");
124 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
126 #define IS_EXT2_SB(sb) (0)
130 static struct file_system_type ext3_fs_type
= {
131 .owner
= THIS_MODULE
,
134 .kill_sb
= kill_block_super
,
135 .fs_flags
= FS_REQUIRES_DEV
,
137 MODULE_ALIAS_FS("ext3");
138 MODULE_ALIAS("ext3");
139 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
141 static int ext4_verify_csum_type(struct super_block
*sb
,
142 struct ext4_super_block
*es
)
144 if (!ext4_has_feature_metadata_csum(sb
))
147 return es
->s_checksum_type
== EXT4_CRC32C_CHKSUM
;
150 static __le32
ext4_superblock_csum(struct super_block
*sb
,
151 struct ext4_super_block
*es
)
153 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
154 int offset
= offsetof(struct ext4_super_block
, s_checksum
);
157 csum
= ext4_chksum(sbi
, ~0, (char *)es
, offset
);
159 return cpu_to_le32(csum
);
162 static int ext4_superblock_csum_verify(struct super_block
*sb
,
163 struct ext4_super_block
*es
)
165 if (!ext4_has_metadata_csum(sb
))
168 return es
->s_checksum
== ext4_superblock_csum(sb
, es
);
171 void ext4_superblock_csum_set(struct super_block
*sb
)
173 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
175 if (!ext4_has_metadata_csum(sb
))
178 es
->s_checksum
= ext4_superblock_csum(sb
, es
);
181 void *ext4_kvmalloc(size_t size
, gfp_t flags
)
185 ret
= kmalloc(size
, flags
| __GFP_NOWARN
);
187 ret
= __vmalloc(size
, flags
, PAGE_KERNEL
);
191 void *ext4_kvzalloc(size_t size
, gfp_t flags
)
195 ret
= kzalloc(size
, flags
| __GFP_NOWARN
);
197 ret
= __vmalloc(size
, flags
| __GFP_ZERO
, PAGE_KERNEL
);
201 ext4_fsblk_t
ext4_block_bitmap(struct super_block
*sb
,
202 struct ext4_group_desc
*bg
)
204 return le32_to_cpu(bg
->bg_block_bitmap_lo
) |
205 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
206 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_block_bitmap_hi
) << 32 : 0);
209 ext4_fsblk_t
ext4_inode_bitmap(struct super_block
*sb
,
210 struct ext4_group_desc
*bg
)
212 return le32_to_cpu(bg
->bg_inode_bitmap_lo
) |
213 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
214 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_bitmap_hi
) << 32 : 0);
217 ext4_fsblk_t
ext4_inode_table(struct super_block
*sb
,
218 struct ext4_group_desc
*bg
)
220 return le32_to_cpu(bg
->bg_inode_table_lo
) |
221 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
222 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_table_hi
) << 32 : 0);
225 __u32
ext4_free_group_clusters(struct super_block
*sb
,
226 struct ext4_group_desc
*bg
)
228 return le16_to_cpu(bg
->bg_free_blocks_count_lo
) |
229 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
230 (__u32
)le16_to_cpu(bg
->bg_free_blocks_count_hi
) << 16 : 0);
233 __u32
ext4_free_inodes_count(struct super_block
*sb
,
234 struct ext4_group_desc
*bg
)
236 return le16_to_cpu(bg
->bg_free_inodes_count_lo
) |
237 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
238 (__u32
)le16_to_cpu(bg
->bg_free_inodes_count_hi
) << 16 : 0);
241 __u32
ext4_used_dirs_count(struct super_block
*sb
,
242 struct ext4_group_desc
*bg
)
244 return le16_to_cpu(bg
->bg_used_dirs_count_lo
) |
245 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
246 (__u32
)le16_to_cpu(bg
->bg_used_dirs_count_hi
) << 16 : 0);
249 __u32
ext4_itable_unused_count(struct super_block
*sb
,
250 struct ext4_group_desc
*bg
)
252 return le16_to_cpu(bg
->bg_itable_unused_lo
) |
253 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
254 (__u32
)le16_to_cpu(bg
->bg_itable_unused_hi
) << 16 : 0);
257 void ext4_block_bitmap_set(struct super_block
*sb
,
258 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
260 bg
->bg_block_bitmap_lo
= cpu_to_le32((u32
)blk
);
261 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
262 bg
->bg_block_bitmap_hi
= cpu_to_le32(blk
>> 32);
265 void ext4_inode_bitmap_set(struct super_block
*sb
,
266 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
268 bg
->bg_inode_bitmap_lo
= cpu_to_le32((u32
)blk
);
269 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
270 bg
->bg_inode_bitmap_hi
= cpu_to_le32(blk
>> 32);
273 void ext4_inode_table_set(struct super_block
*sb
,
274 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
276 bg
->bg_inode_table_lo
= cpu_to_le32((u32
)blk
);
277 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
278 bg
->bg_inode_table_hi
= cpu_to_le32(blk
>> 32);
281 void ext4_free_group_clusters_set(struct super_block
*sb
,
282 struct ext4_group_desc
*bg
, __u32 count
)
284 bg
->bg_free_blocks_count_lo
= cpu_to_le16((__u16
)count
);
285 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
286 bg
->bg_free_blocks_count_hi
= cpu_to_le16(count
>> 16);
289 void ext4_free_inodes_set(struct super_block
*sb
,
290 struct ext4_group_desc
*bg
, __u32 count
)
292 bg
->bg_free_inodes_count_lo
= cpu_to_le16((__u16
)count
);
293 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
294 bg
->bg_free_inodes_count_hi
= cpu_to_le16(count
>> 16);
297 void ext4_used_dirs_set(struct super_block
*sb
,
298 struct ext4_group_desc
*bg
, __u32 count
)
300 bg
->bg_used_dirs_count_lo
= cpu_to_le16((__u16
)count
);
301 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
302 bg
->bg_used_dirs_count_hi
= cpu_to_le16(count
>> 16);
305 void ext4_itable_unused_set(struct super_block
*sb
,
306 struct ext4_group_desc
*bg
, __u32 count
)
308 bg
->bg_itable_unused_lo
= cpu_to_le16((__u16
)count
);
309 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
310 bg
->bg_itable_unused_hi
= cpu_to_le16(count
>> 16);
314 static void __save_error_info(struct super_block
*sb
, const char *func
,
317 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
319 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
320 if (bdev_read_only(sb
->s_bdev
))
322 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
323 es
->s_last_error_time
= cpu_to_le32(get_seconds());
324 strncpy(es
->s_last_error_func
, func
, sizeof(es
->s_last_error_func
));
325 es
->s_last_error_line
= cpu_to_le32(line
);
326 if (!es
->s_first_error_time
) {
327 es
->s_first_error_time
= es
->s_last_error_time
;
328 strncpy(es
->s_first_error_func
, func
,
329 sizeof(es
->s_first_error_func
));
330 es
->s_first_error_line
= cpu_to_le32(line
);
331 es
->s_first_error_ino
= es
->s_last_error_ino
;
332 es
->s_first_error_block
= es
->s_last_error_block
;
335 * Start the daily error reporting function if it hasn't been
338 if (!es
->s_error_count
)
339 mod_timer(&EXT4_SB(sb
)->s_err_report
, jiffies
+ 24*60*60*HZ
);
340 le32_add_cpu(&es
->s_error_count
, 1);
343 static void save_error_info(struct super_block
*sb
, const char *func
,
346 __save_error_info(sb
, func
, line
);
347 ext4_commit_super(sb
, 1);
351 * The del_gendisk() function uninitializes the disk-specific data
352 * structures, including the bdi structure, without telling anyone
353 * else. Once this happens, any attempt to call mark_buffer_dirty()
354 * (for example, by ext4_commit_super), will cause a kernel OOPS.
355 * This is a kludge to prevent these oops until we can put in a proper
356 * hook in del_gendisk() to inform the VFS and file system layers.
358 static int block_device_ejected(struct super_block
*sb
)
360 struct inode
*bd_inode
= sb
->s_bdev
->bd_inode
;
361 struct backing_dev_info
*bdi
= inode_to_bdi(bd_inode
);
363 return bdi
->dev
== NULL
;
366 static void ext4_journal_commit_callback(journal_t
*journal
, transaction_t
*txn
)
368 struct super_block
*sb
= journal
->j_private
;
369 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
370 int error
= is_journal_aborted(journal
);
371 struct ext4_journal_cb_entry
*jce
;
373 BUG_ON(txn
->t_state
== T_FINISHED
);
374 spin_lock(&sbi
->s_md_lock
);
375 while (!list_empty(&txn
->t_private_list
)) {
376 jce
= list_entry(txn
->t_private_list
.next
,
377 struct ext4_journal_cb_entry
, jce_list
);
378 list_del_init(&jce
->jce_list
);
379 spin_unlock(&sbi
->s_md_lock
);
380 jce
->jce_func(sb
, jce
, error
);
381 spin_lock(&sbi
->s_md_lock
);
383 spin_unlock(&sbi
->s_md_lock
);
386 /* Deal with the reporting of failure conditions on a filesystem such as
387 * inconsistencies detected or read IO failures.
389 * On ext2, we can store the error state of the filesystem in the
390 * superblock. That is not possible on ext4, because we may have other
391 * write ordering constraints on the superblock which prevent us from
392 * writing it out straight away; and given that the journal is about to
393 * be aborted, we can't rely on the current, or future, transactions to
394 * write out the superblock safely.
396 * We'll just use the jbd2_journal_abort() error code to record an error in
397 * the journal instead. On recovery, the journal will complain about
398 * that error until we've noted it down and cleared it.
401 static void ext4_handle_error(struct super_block
*sb
)
403 if (sb
->s_flags
& MS_RDONLY
)
406 if (!test_opt(sb
, ERRORS_CONT
)) {
407 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
409 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
411 jbd2_journal_abort(journal
, -EIO
);
413 if (test_opt(sb
, ERRORS_RO
)) {
414 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
416 * Make sure updated value of ->s_mount_flags will be visible
417 * before ->s_flags update
420 sb
->s_flags
|= MS_RDONLY
;
422 if (test_opt(sb
, ERRORS_PANIC
)) {
423 if (EXT4_SB(sb
)->s_journal
&&
424 !(EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_REC_ERR
))
426 panic("EXT4-fs (device %s): panic forced after error\n",
431 #define ext4_error_ratelimit(sb) \
432 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
435 void __ext4_error(struct super_block
*sb
, const char *function
,
436 unsigned int line
, const char *fmt
, ...)
438 struct va_format vaf
;
441 if (ext4_error_ratelimit(sb
)) {
446 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
447 sb
->s_id
, function
, line
, current
->comm
, &vaf
);
450 save_error_info(sb
, function
, line
);
451 ext4_handle_error(sb
);
454 void __ext4_error_inode(struct inode
*inode
, const char *function
,
455 unsigned int line
, ext4_fsblk_t block
,
456 const char *fmt
, ...)
459 struct va_format vaf
;
460 struct ext4_super_block
*es
= EXT4_SB(inode
->i_sb
)->s_es
;
462 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
463 es
->s_last_error_block
= cpu_to_le64(block
);
464 if (ext4_error_ratelimit(inode
->i_sb
)) {
469 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
470 "inode #%lu: block %llu: comm %s: %pV\n",
471 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
472 block
, current
->comm
, &vaf
);
474 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
475 "inode #%lu: comm %s: %pV\n",
476 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
477 current
->comm
, &vaf
);
480 save_error_info(inode
->i_sb
, function
, line
);
481 ext4_handle_error(inode
->i_sb
);
484 void __ext4_error_file(struct file
*file
, const char *function
,
485 unsigned int line
, ext4_fsblk_t block
,
486 const char *fmt
, ...)
489 struct va_format vaf
;
490 struct ext4_super_block
*es
;
491 struct inode
*inode
= file_inode(file
);
492 char pathname
[80], *path
;
494 es
= EXT4_SB(inode
->i_sb
)->s_es
;
495 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
496 if (ext4_error_ratelimit(inode
->i_sb
)) {
497 path
= file_path(file
, pathname
, sizeof(pathname
));
505 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
506 "block %llu: comm %s: path %s: %pV\n",
507 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
508 block
, current
->comm
, path
, &vaf
);
511 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
512 "comm %s: path %s: %pV\n",
513 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
514 current
->comm
, path
, &vaf
);
517 save_error_info(inode
->i_sb
, function
, line
);
518 ext4_handle_error(inode
->i_sb
);
521 const char *ext4_decode_error(struct super_block
*sb
, int errno
,
528 errstr
= "Corrupt filesystem";
531 errstr
= "Filesystem failed CRC";
534 errstr
= "IO failure";
537 errstr
= "Out of memory";
540 if (!sb
|| (EXT4_SB(sb
)->s_journal
&&
541 EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
))
542 errstr
= "Journal has aborted";
544 errstr
= "Readonly filesystem";
547 /* If the caller passed in an extra buffer for unknown
548 * errors, textualise them now. Else we just return
551 /* Check for truncated error codes... */
552 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
561 /* __ext4_std_error decodes expected errors from journaling functions
562 * automatically and invokes the appropriate error response. */
564 void __ext4_std_error(struct super_block
*sb
, const char *function
,
565 unsigned int line
, int errno
)
570 /* Special case: if the error is EROFS, and we're not already
571 * inside a transaction, then there's really no point in logging
573 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
574 (sb
->s_flags
& MS_RDONLY
))
577 if (ext4_error_ratelimit(sb
)) {
578 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
579 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s:%d: %s\n",
580 sb
->s_id
, function
, line
, errstr
);
583 save_error_info(sb
, function
, line
);
584 ext4_handle_error(sb
);
588 * ext4_abort is a much stronger failure handler than ext4_error. The
589 * abort function may be used to deal with unrecoverable failures such
590 * as journal IO errors or ENOMEM at a critical moment in log management.
592 * We unconditionally force the filesystem into an ABORT|READONLY state,
593 * unless the error response on the fs has been set to panic in which
594 * case we take the easy way out and panic immediately.
597 void __ext4_abort(struct super_block
*sb
, const char *function
,
598 unsigned int line
, const char *fmt
, ...)
600 struct va_format vaf
;
603 save_error_info(sb
, function
, line
);
607 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: %pV\n",
608 sb
->s_id
, function
, line
, &vaf
);
611 if ((sb
->s_flags
& MS_RDONLY
) == 0) {
612 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
613 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
615 * Make sure updated value of ->s_mount_flags will be visible
616 * before ->s_flags update
619 sb
->s_flags
|= MS_RDONLY
;
620 if (EXT4_SB(sb
)->s_journal
)
621 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
622 save_error_info(sb
, function
, line
);
624 if (test_opt(sb
, ERRORS_PANIC
)) {
625 if (EXT4_SB(sb
)->s_journal
&&
626 !(EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_REC_ERR
))
628 panic("EXT4-fs panic from previous error\n");
632 void __ext4_msg(struct super_block
*sb
,
633 const char *prefix
, const char *fmt
, ...)
635 struct va_format vaf
;
638 if (!___ratelimit(&(EXT4_SB(sb
)->s_msg_ratelimit_state
), "EXT4-fs"))
644 printk("%sEXT4-fs (%s): %pV\n", prefix
, sb
->s_id
, &vaf
);
648 #define ext4_warning_ratelimit(sb) \
649 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
652 void __ext4_warning(struct super_block
*sb
, const char *function
,
653 unsigned int line
, const char *fmt
, ...)
655 struct va_format vaf
;
658 if (!ext4_warning_ratelimit(sb
))
664 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: %pV\n",
665 sb
->s_id
, function
, line
, &vaf
);
669 void __ext4_warning_inode(const struct inode
*inode
, const char *function
,
670 unsigned int line
, const char *fmt
, ...)
672 struct va_format vaf
;
675 if (!ext4_warning_ratelimit(inode
->i_sb
))
681 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: "
682 "inode #%lu: comm %s: %pV\n", inode
->i_sb
->s_id
,
683 function
, line
, inode
->i_ino
, current
->comm
, &vaf
);
687 void __ext4_grp_locked_error(const char *function
, unsigned int line
,
688 struct super_block
*sb
, ext4_group_t grp
,
689 unsigned long ino
, ext4_fsblk_t block
,
690 const char *fmt
, ...)
694 struct va_format vaf
;
696 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
698 es
->s_last_error_ino
= cpu_to_le32(ino
);
699 es
->s_last_error_block
= cpu_to_le64(block
);
700 __save_error_info(sb
, function
, line
);
702 if (ext4_error_ratelimit(sb
)) {
706 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: group %u, ",
707 sb
->s_id
, function
, line
, grp
);
709 printk(KERN_CONT
"inode %lu: ", ino
);
711 printk(KERN_CONT
"block %llu:",
712 (unsigned long long) block
);
713 printk(KERN_CONT
"%pV\n", &vaf
);
717 if (test_opt(sb
, ERRORS_CONT
)) {
718 ext4_commit_super(sb
, 0);
722 ext4_unlock_group(sb
, grp
);
723 ext4_handle_error(sb
);
725 * We only get here in the ERRORS_RO case; relocking the group
726 * may be dangerous, but nothing bad will happen since the
727 * filesystem will have already been marked read/only and the
728 * journal has been aborted. We return 1 as a hint to callers
729 * who might what to use the return value from
730 * ext4_grp_locked_error() to distinguish between the
731 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
732 * aggressively from the ext4 function in question, with a
733 * more appropriate error code.
735 ext4_lock_group(sb
, grp
);
739 void ext4_update_dynamic_rev(struct super_block
*sb
)
741 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
743 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
747 "updating to rev %d because of new feature flag, "
748 "running e2fsck is recommended",
751 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
752 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
753 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
754 /* leave es->s_feature_*compat flags alone */
755 /* es->s_uuid will be set by e2fsck if empty */
758 * The rest of the superblock fields should be zero, and if not it
759 * means they are likely already in use, so leave them alone. We
760 * can leave it up to e2fsck to clean up any inconsistencies there.
765 * Open the external journal device
767 static struct block_device
*ext4_blkdev_get(dev_t dev
, struct super_block
*sb
)
769 struct block_device
*bdev
;
770 char b
[BDEVNAME_SIZE
];
772 bdev
= blkdev_get_by_dev(dev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
, sb
);
778 ext4_msg(sb
, KERN_ERR
, "failed to open journal device %s: %ld",
779 __bdevname(dev
, b
), PTR_ERR(bdev
));
784 * Release the journal device
786 static void ext4_blkdev_put(struct block_device
*bdev
)
788 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
);
791 static void ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
793 struct block_device
*bdev
;
794 bdev
= sbi
->journal_bdev
;
796 ext4_blkdev_put(bdev
);
797 sbi
->journal_bdev
= NULL
;
801 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
803 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
806 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
810 ext4_msg(sb
, KERN_ERR
, "sb orphan head is %d",
811 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
813 printk(KERN_ERR
"sb_info orphan list:\n");
814 list_for_each(l
, &sbi
->s_orphan
) {
815 struct inode
*inode
= orphan_list_entry(l
);
817 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
818 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
819 inode
->i_mode
, inode
->i_nlink
,
824 static void ext4_put_super(struct super_block
*sb
)
826 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
827 struct ext4_super_block
*es
= sbi
->s_es
;
830 ext4_unregister_li_request(sb
);
831 dquot_disable(sb
, -1, DQUOT_USAGE_ENABLED
| DQUOT_LIMITS_ENABLED
);
833 flush_workqueue(sbi
->rsv_conversion_wq
);
834 destroy_workqueue(sbi
->rsv_conversion_wq
);
836 if (sbi
->s_journal
) {
837 err
= jbd2_journal_destroy(sbi
->s_journal
);
838 sbi
->s_journal
= NULL
;
840 ext4_abort(sb
, "Couldn't clean up the journal");
843 ext4_unregister_sysfs(sb
);
844 ext4_es_unregister_shrinker(sbi
);
845 del_timer_sync(&sbi
->s_err_report
);
846 ext4_release_system_zone(sb
);
848 ext4_ext_release(sb
);
850 if (!(sb
->s_flags
& MS_RDONLY
)) {
851 ext4_clear_feature_journal_needs_recovery(sb
);
852 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
854 if (!(sb
->s_flags
& MS_RDONLY
))
855 ext4_commit_super(sb
, 1);
857 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
858 brelse(sbi
->s_group_desc
[i
]);
859 kvfree(sbi
->s_group_desc
);
860 kvfree(sbi
->s_flex_groups
);
861 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
862 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
863 percpu_counter_destroy(&sbi
->s_dirs_counter
);
864 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
865 percpu_free_rwsem(&sbi
->s_journal_flag_rwsem
);
867 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
868 kfree(sbi
->s_qf_names
[i
]);
871 /* Debugging code just in case the in-memory inode orphan list
872 * isn't empty. The on-disk one can be non-empty if we've
873 * detected an error and taken the fs readonly, but the
874 * in-memory list had better be clean by this point. */
875 if (!list_empty(&sbi
->s_orphan
))
876 dump_orphan_list(sb
, sbi
);
877 J_ASSERT(list_empty(&sbi
->s_orphan
));
879 sync_blockdev(sb
->s_bdev
);
880 invalidate_bdev(sb
->s_bdev
);
881 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
883 * Invalidate the journal device's buffers. We don't want them
884 * floating about in memory - the physical journal device may
885 * hotswapped, and it breaks the `ro-after' testing code.
887 sync_blockdev(sbi
->journal_bdev
);
888 invalidate_bdev(sbi
->journal_bdev
);
889 ext4_blkdev_remove(sbi
);
891 if (sbi
->s_mb_cache
) {
892 ext4_xattr_destroy_cache(sbi
->s_mb_cache
);
893 sbi
->s_mb_cache
= NULL
;
896 kthread_stop(sbi
->s_mmp_tsk
);
898 sb
->s_fs_info
= NULL
;
900 * Now that we are completely done shutting down the
901 * superblock, we need to actually destroy the kobject.
903 kobject_put(&sbi
->s_kobj
);
904 wait_for_completion(&sbi
->s_kobj_unregister
);
905 if (sbi
->s_chksum_driver
)
906 crypto_free_shash(sbi
->s_chksum_driver
);
907 kfree(sbi
->s_blockgroup_lock
);
911 static struct kmem_cache
*ext4_inode_cachep
;
914 * Called inside transaction, so use GFP_NOFS
916 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
918 struct ext4_inode_info
*ei
;
920 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
924 ei
->vfs_inode
.i_version
= 1;
925 spin_lock_init(&ei
->i_raw_lock
);
926 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
927 spin_lock_init(&ei
->i_prealloc_lock
);
928 ext4_es_init_tree(&ei
->i_es_tree
);
929 rwlock_init(&ei
->i_es_lock
);
930 INIT_LIST_HEAD(&ei
->i_es_list
);
933 ei
->i_es_shrink_lblk
= 0;
934 ei
->i_reserved_data_blocks
= 0;
935 ei
->i_reserved_meta_blocks
= 0;
936 ei
->i_allocated_meta_blocks
= 0;
937 ei
->i_da_metadata_calc_len
= 0;
938 ei
->i_da_metadata_calc_last_lblock
= 0;
939 spin_lock_init(&(ei
->i_block_reservation_lock
));
941 ei
->i_reserved_quota
= 0;
942 memset(&ei
->i_dquot
, 0, sizeof(ei
->i_dquot
));
945 INIT_LIST_HEAD(&ei
->i_rsv_conversion_list
);
946 spin_lock_init(&ei
->i_completed_io_lock
);
948 ei
->i_datasync_tid
= 0;
949 atomic_set(&ei
->i_unwritten
, 0);
950 INIT_WORK(&ei
->i_rsv_conversion_work
, ext4_end_io_rsv_work
);
951 return &ei
->vfs_inode
;
954 static int ext4_drop_inode(struct inode
*inode
)
956 int drop
= generic_drop_inode(inode
);
958 trace_ext4_drop_inode(inode
, drop
);
962 static void ext4_i_callback(struct rcu_head
*head
)
964 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
965 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
968 static void ext4_destroy_inode(struct inode
*inode
)
970 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
971 ext4_msg(inode
->i_sb
, KERN_ERR
,
972 "Inode %lu (%p): orphan list check failed!",
973 inode
->i_ino
, EXT4_I(inode
));
974 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
975 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
979 call_rcu(&inode
->i_rcu
, ext4_i_callback
);
982 static void init_once(void *foo
)
984 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
986 INIT_LIST_HEAD(&ei
->i_orphan
);
987 init_rwsem(&ei
->xattr_sem
);
988 init_rwsem(&ei
->i_data_sem
);
989 init_rwsem(&ei
->i_mmap_sem
);
990 inode_init_once(&ei
->vfs_inode
);
993 static int __init
init_inodecache(void)
995 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
996 sizeof(struct ext4_inode_info
),
997 0, (SLAB_RECLAIM_ACCOUNT
|
998 SLAB_MEM_SPREAD
|SLAB_ACCOUNT
),
1000 if (ext4_inode_cachep
== NULL
)
1005 static void destroy_inodecache(void)
1008 * Make sure all delayed rcu free inodes are flushed before we
1012 kmem_cache_destroy(ext4_inode_cachep
);
1015 void ext4_clear_inode(struct inode
*inode
)
1017 invalidate_inode_buffers(inode
);
1020 ext4_discard_preallocations(inode
);
1021 ext4_es_remove_extent(inode
, 0, EXT_MAX_BLOCKS
);
1022 if (EXT4_I(inode
)->jinode
) {
1023 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode
),
1024 EXT4_I(inode
)->jinode
);
1025 jbd2_free_inode(EXT4_I(inode
)->jinode
);
1026 EXT4_I(inode
)->jinode
= NULL
;
1028 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1029 fscrypt_put_encryption_info(inode
, NULL
);
1033 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
1034 u64 ino
, u32 generation
)
1036 struct inode
*inode
;
1038 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
1039 return ERR_PTR(-ESTALE
);
1040 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
1041 return ERR_PTR(-ESTALE
);
1043 /* iget isn't really right if the inode is currently unallocated!!
1045 * ext4_read_inode will return a bad_inode if the inode had been
1046 * deleted, so we should be safe.
1048 * Currently we don't know the generation for parent directory, so
1049 * a generation of 0 means "accept any"
1051 inode
= ext4_iget_normal(sb
, ino
);
1053 return ERR_CAST(inode
);
1054 if (generation
&& inode
->i_generation
!= generation
) {
1056 return ERR_PTR(-ESTALE
);
1062 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1063 int fh_len
, int fh_type
)
1065 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
1066 ext4_nfs_get_inode
);
1069 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1070 int fh_len
, int fh_type
)
1072 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1073 ext4_nfs_get_inode
);
1077 * Try to release metadata pages (indirect blocks, directories) which are
1078 * mapped via the block device. Since these pages could have journal heads
1079 * which would prevent try_to_free_buffers() from freeing them, we must use
1080 * jbd2 layer's try_to_free_buffers() function to release them.
1082 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
1085 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
1087 WARN_ON(PageChecked(page
));
1088 if (!page_has_buffers(page
))
1091 return jbd2_journal_try_to_free_buffers(journal
, page
,
1092 wait
& ~__GFP_DIRECT_RECLAIM
);
1093 return try_to_free_buffers(page
);
1096 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1097 static int ext4_get_context(struct inode
*inode
, void *ctx
, size_t len
)
1099 return ext4_xattr_get(inode
, EXT4_XATTR_INDEX_ENCRYPTION
,
1100 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT
, ctx
, len
);
1103 static int ext4_key_prefix(struct inode
*inode
, u8
**key
)
1105 *key
= EXT4_SB(inode
->i_sb
)->key_prefix
;
1106 return EXT4_SB(inode
->i_sb
)->key_prefix_size
;
1109 static int ext4_prepare_context(struct inode
*inode
)
1111 return ext4_convert_inline_data(inode
);
1114 static int ext4_set_context(struct inode
*inode
, const void *ctx
, size_t len
,
1117 handle_t
*handle
= fs_data
;
1118 int res
, res2
, retries
= 0;
1121 * If a journal handle was specified, then the encryption context is
1122 * being set on a new inode via inheritance and is part of a larger
1123 * transaction to create the inode. Otherwise the encryption context is
1124 * being set on an existing inode in its own transaction. Only in the
1125 * latter case should the "retry on ENOSPC" logic be used.
1129 res
= ext4_xattr_set_handle(handle
, inode
,
1130 EXT4_XATTR_INDEX_ENCRYPTION
,
1131 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT
,
1134 ext4_set_inode_flag(inode
, EXT4_INODE_ENCRYPT
);
1135 ext4_clear_inode_state(inode
,
1136 EXT4_STATE_MAY_INLINE_DATA
);
1138 * Update inode->i_flags - e.g. S_DAX may get disabled
1140 ext4_set_inode_flags(inode
);
1146 handle
= ext4_journal_start(inode
, EXT4_HT_MISC
,
1147 ext4_jbd2_credits_xattr(inode
));
1149 return PTR_ERR(handle
);
1151 res
= ext4_xattr_set_handle(handle
, inode
, EXT4_XATTR_INDEX_ENCRYPTION
,
1152 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT
,
1155 ext4_set_inode_flag(inode
, EXT4_INODE_ENCRYPT
);
1156 /* Update inode->i_flags - e.g. S_DAX may get disabled */
1157 ext4_set_inode_flags(inode
);
1158 res
= ext4_mark_inode_dirty(handle
, inode
);
1160 EXT4_ERROR_INODE(inode
, "Failed to mark inode dirty");
1162 res2
= ext4_journal_stop(handle
);
1164 if (res
== -ENOSPC
&& ext4_should_retry_alloc(inode
->i_sb
, &retries
))
1171 static int ext4_dummy_context(struct inode
*inode
)
1173 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode
->i_sb
));
1176 static unsigned ext4_max_namelen(struct inode
*inode
)
1178 return S_ISLNK(inode
->i_mode
) ? inode
->i_sb
->s_blocksize
:
1182 static struct fscrypt_operations ext4_cryptops
= {
1183 .get_context
= ext4_get_context
,
1184 .key_prefix
= ext4_key_prefix
,
1185 .prepare_context
= ext4_prepare_context
,
1186 .set_context
= ext4_set_context
,
1187 .dummy_context
= ext4_dummy_context
,
1188 .is_encrypted
= ext4_encrypted_inode
,
1189 .empty_dir
= ext4_empty_dir
,
1190 .max_namelen
= ext4_max_namelen
,
1193 static struct fscrypt_operations ext4_cryptops
= {
1194 .is_encrypted
= ext4_encrypted_inode
,
1199 static char *quotatypes
[] = INITQFNAMES
;
1200 #define QTYPE2NAME(t) (quotatypes[t])
1202 static int ext4_write_dquot(struct dquot
*dquot
);
1203 static int ext4_acquire_dquot(struct dquot
*dquot
);
1204 static int ext4_release_dquot(struct dquot
*dquot
);
1205 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
1206 static int ext4_write_info(struct super_block
*sb
, int type
);
1207 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
1208 const struct path
*path
);
1209 static int ext4_quota_off(struct super_block
*sb
, int type
);
1210 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
1211 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
1212 size_t len
, loff_t off
);
1213 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
1214 const char *data
, size_t len
, loff_t off
);
1215 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
1216 unsigned int flags
);
1217 static int ext4_enable_quotas(struct super_block
*sb
);
1218 static int ext4_get_next_id(struct super_block
*sb
, struct kqid
*qid
);
1220 static struct dquot
**ext4_get_dquots(struct inode
*inode
)
1222 return EXT4_I(inode
)->i_dquot
;
1225 static const struct dquot_operations ext4_quota_operations
= {
1226 .get_reserved_space
= ext4_get_reserved_space
,
1227 .write_dquot
= ext4_write_dquot
,
1228 .acquire_dquot
= ext4_acquire_dquot
,
1229 .release_dquot
= ext4_release_dquot
,
1230 .mark_dirty
= ext4_mark_dquot_dirty
,
1231 .write_info
= ext4_write_info
,
1232 .alloc_dquot
= dquot_alloc
,
1233 .destroy_dquot
= dquot_destroy
,
1234 .get_projid
= ext4_get_projid
,
1235 .get_next_id
= ext4_get_next_id
,
1238 static const struct quotactl_ops ext4_qctl_operations
= {
1239 .quota_on
= ext4_quota_on
,
1240 .quota_off
= ext4_quota_off
,
1241 .quota_sync
= dquot_quota_sync
,
1242 .get_state
= dquot_get_state
,
1243 .set_info
= dquot_set_dqinfo
,
1244 .get_dqblk
= dquot_get_dqblk
,
1245 .set_dqblk
= dquot_set_dqblk
,
1246 .get_nextdqblk
= dquot_get_next_dqblk
,
1250 static const struct super_operations ext4_sops
= {
1251 .alloc_inode
= ext4_alloc_inode
,
1252 .destroy_inode
= ext4_destroy_inode
,
1253 .write_inode
= ext4_write_inode
,
1254 .dirty_inode
= ext4_dirty_inode
,
1255 .drop_inode
= ext4_drop_inode
,
1256 .evict_inode
= ext4_evict_inode
,
1257 .put_super
= ext4_put_super
,
1258 .sync_fs
= ext4_sync_fs
,
1259 .freeze_fs
= ext4_freeze
,
1260 .unfreeze_fs
= ext4_unfreeze
,
1261 .statfs
= ext4_statfs
,
1262 .remount_fs
= ext4_remount
,
1263 .show_options
= ext4_show_options
,
1265 .quota_read
= ext4_quota_read
,
1266 .quota_write
= ext4_quota_write
,
1267 .get_dquots
= ext4_get_dquots
,
1269 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1272 static const struct export_operations ext4_export_ops
= {
1273 .fh_to_dentry
= ext4_fh_to_dentry
,
1274 .fh_to_parent
= ext4_fh_to_parent
,
1275 .get_parent
= ext4_get_parent
,
1279 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1280 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1281 Opt_nouid32
, Opt_debug
, Opt_removed
,
1282 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1283 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
,
1284 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
, Opt_journal_dev
,
1285 Opt_journal_path
, Opt_journal_checksum
, Opt_journal_async_commit
,
1286 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1287 Opt_data_err_abort
, Opt_data_err_ignore
, Opt_test_dummy_encryption
,
1288 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1289 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
1290 Opt_noquota
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
1291 Opt_usrquota
, Opt_grpquota
, Opt_prjquota
, Opt_i_version
, Opt_dax
,
1292 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
, Opt_mblk_io_submit
,
1293 Opt_lazytime
, Opt_nolazytime
,
1294 Opt_nomblk_io_submit
, Opt_block_validity
, Opt_noblock_validity
,
1295 Opt_inode_readahead_blks
, Opt_journal_ioprio
,
1296 Opt_dioread_nolock
, Opt_dioread_lock
,
1297 Opt_discard
, Opt_nodiscard
, Opt_init_itable
, Opt_noinit_itable
,
1298 Opt_max_dir_size_kb
, Opt_nojournal_checksum
,
1301 static const match_table_t tokens
= {
1302 {Opt_bsd_df
, "bsddf"},
1303 {Opt_minix_df
, "minixdf"},
1304 {Opt_grpid
, "grpid"},
1305 {Opt_grpid
, "bsdgroups"},
1306 {Opt_nogrpid
, "nogrpid"},
1307 {Opt_nogrpid
, "sysvgroups"},
1308 {Opt_resgid
, "resgid=%u"},
1309 {Opt_resuid
, "resuid=%u"},
1311 {Opt_err_cont
, "errors=continue"},
1312 {Opt_err_panic
, "errors=panic"},
1313 {Opt_err_ro
, "errors=remount-ro"},
1314 {Opt_nouid32
, "nouid32"},
1315 {Opt_debug
, "debug"},
1316 {Opt_removed
, "oldalloc"},
1317 {Opt_removed
, "orlov"},
1318 {Opt_user_xattr
, "user_xattr"},
1319 {Opt_nouser_xattr
, "nouser_xattr"},
1321 {Opt_noacl
, "noacl"},
1322 {Opt_noload
, "norecovery"},
1323 {Opt_noload
, "noload"},
1324 {Opt_removed
, "nobh"},
1325 {Opt_removed
, "bh"},
1326 {Opt_commit
, "commit=%u"},
1327 {Opt_min_batch_time
, "min_batch_time=%u"},
1328 {Opt_max_batch_time
, "max_batch_time=%u"},
1329 {Opt_journal_dev
, "journal_dev=%u"},
1330 {Opt_journal_path
, "journal_path=%s"},
1331 {Opt_journal_checksum
, "journal_checksum"},
1332 {Opt_nojournal_checksum
, "nojournal_checksum"},
1333 {Opt_journal_async_commit
, "journal_async_commit"},
1334 {Opt_abort
, "abort"},
1335 {Opt_data_journal
, "data=journal"},
1336 {Opt_data_ordered
, "data=ordered"},
1337 {Opt_data_writeback
, "data=writeback"},
1338 {Opt_data_err_abort
, "data_err=abort"},
1339 {Opt_data_err_ignore
, "data_err=ignore"},
1340 {Opt_offusrjquota
, "usrjquota="},
1341 {Opt_usrjquota
, "usrjquota=%s"},
1342 {Opt_offgrpjquota
, "grpjquota="},
1343 {Opt_grpjquota
, "grpjquota=%s"},
1344 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1345 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1346 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
1347 {Opt_grpquota
, "grpquota"},
1348 {Opt_noquota
, "noquota"},
1349 {Opt_quota
, "quota"},
1350 {Opt_usrquota
, "usrquota"},
1351 {Opt_prjquota
, "prjquota"},
1352 {Opt_barrier
, "barrier=%u"},
1353 {Opt_barrier
, "barrier"},
1354 {Opt_nobarrier
, "nobarrier"},
1355 {Opt_i_version
, "i_version"},
1357 {Opt_stripe
, "stripe=%u"},
1358 {Opt_delalloc
, "delalloc"},
1359 {Opt_lazytime
, "lazytime"},
1360 {Opt_nolazytime
, "nolazytime"},
1361 {Opt_nodelalloc
, "nodelalloc"},
1362 {Opt_removed
, "mblk_io_submit"},
1363 {Opt_removed
, "nomblk_io_submit"},
1364 {Opt_block_validity
, "block_validity"},
1365 {Opt_noblock_validity
, "noblock_validity"},
1366 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1367 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1368 {Opt_auto_da_alloc
, "auto_da_alloc=%u"},
1369 {Opt_auto_da_alloc
, "auto_da_alloc"},
1370 {Opt_noauto_da_alloc
, "noauto_da_alloc"},
1371 {Opt_dioread_nolock
, "dioread_nolock"},
1372 {Opt_dioread_lock
, "dioread_lock"},
1373 {Opt_discard
, "discard"},
1374 {Opt_nodiscard
, "nodiscard"},
1375 {Opt_init_itable
, "init_itable=%u"},
1376 {Opt_init_itable
, "init_itable"},
1377 {Opt_noinit_itable
, "noinit_itable"},
1378 {Opt_max_dir_size_kb
, "max_dir_size_kb=%u"},
1379 {Opt_test_dummy_encryption
, "test_dummy_encryption"},
1380 {Opt_removed
, "check=none"}, /* mount option from ext2/3 */
1381 {Opt_removed
, "nocheck"}, /* mount option from ext2/3 */
1382 {Opt_removed
, "reservation"}, /* mount option from ext2/3 */
1383 {Opt_removed
, "noreservation"}, /* mount option from ext2/3 */
1384 {Opt_removed
, "journal=%u"}, /* mount option from ext2/3 */
1388 static ext4_fsblk_t
get_sb_block(void **data
)
1390 ext4_fsblk_t sb_block
;
1391 char *options
= (char *) *data
;
1393 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1394 return 1; /* Default location */
1397 /* TODO: use simple_strtoll with >32bit ext4 */
1398 sb_block
= simple_strtoul(options
, &options
, 0);
1399 if (*options
&& *options
!= ',') {
1400 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1404 if (*options
== ',')
1406 *data
= (void *) options
;
1411 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1412 static char deprecated_msg
[] = "Mount option \"%s\" will be removed by %s\n"
1413 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1416 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
1418 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1422 if (sb_any_quota_loaded(sb
) &&
1423 !sbi
->s_qf_names
[qtype
]) {
1424 ext4_msg(sb
, KERN_ERR
,
1425 "Cannot change journaled "
1426 "quota options when quota turned on");
1429 if (ext4_has_feature_quota(sb
)) {
1430 ext4_msg(sb
, KERN_INFO
, "Journaled quota options "
1431 "ignored when QUOTA feature is enabled");
1434 qname
= match_strdup(args
);
1436 ext4_msg(sb
, KERN_ERR
,
1437 "Not enough memory for storing quotafile name");
1440 if (sbi
->s_qf_names
[qtype
]) {
1441 if (strcmp(sbi
->s_qf_names
[qtype
], qname
) == 0)
1444 ext4_msg(sb
, KERN_ERR
,
1445 "%s quota file already specified",
1449 if (strchr(qname
, '/')) {
1450 ext4_msg(sb
, KERN_ERR
,
1451 "quotafile must be on filesystem root");
1454 sbi
->s_qf_names
[qtype
] = qname
;
1462 static int clear_qf_name(struct super_block
*sb
, int qtype
)
1465 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1467 if (sb_any_quota_loaded(sb
) &&
1468 sbi
->s_qf_names
[qtype
]) {
1469 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
1470 " when quota turned on");
1473 kfree(sbi
->s_qf_names
[qtype
]);
1474 sbi
->s_qf_names
[qtype
] = NULL
;
1479 #define MOPT_SET 0x0001
1480 #define MOPT_CLEAR 0x0002
1481 #define MOPT_NOSUPPORT 0x0004
1482 #define MOPT_EXPLICIT 0x0008
1483 #define MOPT_CLEAR_ERR 0x0010
1484 #define MOPT_GTE0 0x0020
1487 #define MOPT_QFMT 0x0040
1489 #define MOPT_Q MOPT_NOSUPPORT
1490 #define MOPT_QFMT MOPT_NOSUPPORT
1492 #define MOPT_DATAJ 0x0080
1493 #define MOPT_NO_EXT2 0x0100
1494 #define MOPT_NO_EXT3 0x0200
1495 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1496 #define MOPT_STRING 0x0400
1498 static const struct mount_opts
{
1502 } ext4_mount_opts
[] = {
1503 {Opt_minix_df
, EXT4_MOUNT_MINIX_DF
, MOPT_SET
},
1504 {Opt_bsd_df
, EXT4_MOUNT_MINIX_DF
, MOPT_CLEAR
},
1505 {Opt_grpid
, EXT4_MOUNT_GRPID
, MOPT_SET
},
1506 {Opt_nogrpid
, EXT4_MOUNT_GRPID
, MOPT_CLEAR
},
1507 {Opt_block_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_SET
},
1508 {Opt_noblock_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_CLEAR
},
1509 {Opt_dioread_nolock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1510 MOPT_EXT4_ONLY
| MOPT_SET
},
1511 {Opt_dioread_lock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1512 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1513 {Opt_discard
, EXT4_MOUNT_DISCARD
, MOPT_SET
},
1514 {Opt_nodiscard
, EXT4_MOUNT_DISCARD
, MOPT_CLEAR
},
1515 {Opt_delalloc
, EXT4_MOUNT_DELALLOC
,
1516 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1517 {Opt_nodelalloc
, EXT4_MOUNT_DELALLOC
,
1518 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1519 {Opt_nojournal_checksum
, EXT4_MOUNT_JOURNAL_CHECKSUM
,
1520 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1521 {Opt_journal_checksum
, EXT4_MOUNT_JOURNAL_CHECKSUM
,
1522 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1523 {Opt_journal_async_commit
, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT
|
1524 EXT4_MOUNT_JOURNAL_CHECKSUM
),
1525 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1526 {Opt_noload
, EXT4_MOUNT_NOLOAD
, MOPT_NO_EXT2
| MOPT_SET
},
1527 {Opt_err_panic
, EXT4_MOUNT_ERRORS_PANIC
, MOPT_SET
| MOPT_CLEAR_ERR
},
1528 {Opt_err_ro
, EXT4_MOUNT_ERRORS_RO
, MOPT_SET
| MOPT_CLEAR_ERR
},
1529 {Opt_err_cont
, EXT4_MOUNT_ERRORS_CONT
, MOPT_SET
| MOPT_CLEAR_ERR
},
1530 {Opt_data_err_abort
, EXT4_MOUNT_DATA_ERR_ABORT
,
1532 {Opt_data_err_ignore
, EXT4_MOUNT_DATA_ERR_ABORT
,
1534 {Opt_barrier
, EXT4_MOUNT_BARRIER
, MOPT_SET
},
1535 {Opt_nobarrier
, EXT4_MOUNT_BARRIER
, MOPT_CLEAR
},
1536 {Opt_noauto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_SET
},
1537 {Opt_auto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_CLEAR
},
1538 {Opt_noinit_itable
, EXT4_MOUNT_INIT_INODE_TABLE
, MOPT_CLEAR
},
1539 {Opt_commit
, 0, MOPT_GTE0
},
1540 {Opt_max_batch_time
, 0, MOPT_GTE0
},
1541 {Opt_min_batch_time
, 0, MOPT_GTE0
},
1542 {Opt_inode_readahead_blks
, 0, MOPT_GTE0
},
1543 {Opt_init_itable
, 0, MOPT_GTE0
},
1544 {Opt_dax
, EXT4_MOUNT_DAX
, MOPT_SET
},
1545 {Opt_stripe
, 0, MOPT_GTE0
},
1546 {Opt_resuid
, 0, MOPT_GTE0
},
1547 {Opt_resgid
, 0, MOPT_GTE0
},
1548 {Opt_journal_dev
, 0, MOPT_NO_EXT2
| MOPT_GTE0
},
1549 {Opt_journal_path
, 0, MOPT_NO_EXT2
| MOPT_STRING
},
1550 {Opt_journal_ioprio
, 0, MOPT_NO_EXT2
| MOPT_GTE0
},
1551 {Opt_data_journal
, EXT4_MOUNT_JOURNAL_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1552 {Opt_data_ordered
, EXT4_MOUNT_ORDERED_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1553 {Opt_data_writeback
, EXT4_MOUNT_WRITEBACK_DATA
,
1554 MOPT_NO_EXT2
| MOPT_DATAJ
},
1555 {Opt_user_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_SET
},
1556 {Opt_nouser_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_CLEAR
},
1557 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1558 {Opt_acl
, EXT4_MOUNT_POSIX_ACL
, MOPT_SET
},
1559 {Opt_noacl
, EXT4_MOUNT_POSIX_ACL
, MOPT_CLEAR
},
1561 {Opt_acl
, 0, MOPT_NOSUPPORT
},
1562 {Opt_noacl
, 0, MOPT_NOSUPPORT
},
1564 {Opt_nouid32
, EXT4_MOUNT_NO_UID32
, MOPT_SET
},
1565 {Opt_debug
, EXT4_MOUNT_DEBUG
, MOPT_SET
},
1566 {Opt_quota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
, MOPT_SET
| MOPT_Q
},
1567 {Opt_usrquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
,
1569 {Opt_grpquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_GRPQUOTA
,
1571 {Opt_prjquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_PRJQUOTA
,
1573 {Opt_noquota
, (EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
|
1574 EXT4_MOUNT_GRPQUOTA
| EXT4_MOUNT_PRJQUOTA
),
1575 MOPT_CLEAR
| MOPT_Q
},
1576 {Opt_usrjquota
, 0, MOPT_Q
},
1577 {Opt_grpjquota
, 0, MOPT_Q
},
1578 {Opt_offusrjquota
, 0, MOPT_Q
},
1579 {Opt_offgrpjquota
, 0, MOPT_Q
},
1580 {Opt_jqfmt_vfsold
, QFMT_VFS_OLD
, MOPT_QFMT
},
1581 {Opt_jqfmt_vfsv0
, QFMT_VFS_V0
, MOPT_QFMT
},
1582 {Opt_jqfmt_vfsv1
, QFMT_VFS_V1
, MOPT_QFMT
},
1583 {Opt_max_dir_size_kb
, 0, MOPT_GTE0
},
1584 {Opt_test_dummy_encryption
, 0, MOPT_GTE0
},
1588 static int handle_mount_opt(struct super_block
*sb
, char *opt
, int token
,
1589 substring_t
*args
, unsigned long *journal_devnum
,
1590 unsigned int *journal_ioprio
, int is_remount
)
1592 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1593 const struct mount_opts
*m
;
1599 if (token
== Opt_usrjquota
)
1600 return set_qf_name(sb
, USRQUOTA
, &args
[0]);
1601 else if (token
== Opt_grpjquota
)
1602 return set_qf_name(sb
, GRPQUOTA
, &args
[0]);
1603 else if (token
== Opt_offusrjquota
)
1604 return clear_qf_name(sb
, USRQUOTA
);
1605 else if (token
== Opt_offgrpjquota
)
1606 return clear_qf_name(sb
, GRPQUOTA
);
1610 case Opt_nouser_xattr
:
1611 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, opt
, "3.5");
1614 return 1; /* handled by get_sb_block() */
1616 ext4_msg(sb
, KERN_WARNING
, "Ignoring removed %s option", opt
);
1619 sbi
->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
1622 sb
->s_flags
|= MS_I_VERSION
;
1625 sb
->s_flags
|= MS_LAZYTIME
;
1627 case Opt_nolazytime
:
1628 sb
->s_flags
&= ~MS_LAZYTIME
;
1632 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++)
1633 if (token
== m
->token
)
1636 if (m
->token
== Opt_err
) {
1637 ext4_msg(sb
, KERN_ERR
, "Unrecognized mount option \"%s\" "
1638 "or missing value", opt
);
1642 if ((m
->flags
& MOPT_NO_EXT2
) && IS_EXT2_SB(sb
)) {
1643 ext4_msg(sb
, KERN_ERR
,
1644 "Mount option \"%s\" incompatible with ext2", opt
);
1647 if ((m
->flags
& MOPT_NO_EXT3
) && IS_EXT3_SB(sb
)) {
1648 ext4_msg(sb
, KERN_ERR
,
1649 "Mount option \"%s\" incompatible with ext3", opt
);
1653 if (args
->from
&& !(m
->flags
& MOPT_STRING
) && match_int(args
, &arg
))
1655 if (args
->from
&& (m
->flags
& MOPT_GTE0
) && (arg
< 0))
1657 if (m
->flags
& MOPT_EXPLICIT
) {
1658 if (m
->mount_opt
& EXT4_MOUNT_DELALLOC
) {
1659 set_opt2(sb
, EXPLICIT_DELALLOC
);
1660 } else if (m
->mount_opt
& EXT4_MOUNT_JOURNAL_CHECKSUM
) {
1661 set_opt2(sb
, EXPLICIT_JOURNAL_CHECKSUM
);
1665 if (m
->flags
& MOPT_CLEAR_ERR
)
1666 clear_opt(sb
, ERRORS_MASK
);
1667 if (token
== Opt_noquota
&& sb_any_quota_loaded(sb
)) {
1668 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
1669 "options when quota turned on");
1673 if (m
->flags
& MOPT_NOSUPPORT
) {
1674 ext4_msg(sb
, KERN_ERR
, "%s option not supported", opt
);
1675 } else if (token
== Opt_commit
) {
1677 arg
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1678 sbi
->s_commit_interval
= HZ
* arg
;
1679 } else if (token
== Opt_max_batch_time
) {
1680 sbi
->s_max_batch_time
= arg
;
1681 } else if (token
== Opt_min_batch_time
) {
1682 sbi
->s_min_batch_time
= arg
;
1683 } else if (token
== Opt_inode_readahead_blks
) {
1684 if (arg
&& (arg
> (1 << 30) || !is_power_of_2(arg
))) {
1685 ext4_msg(sb
, KERN_ERR
,
1686 "EXT4-fs: inode_readahead_blks must be "
1687 "0 or a power of 2 smaller than 2^31");
1690 sbi
->s_inode_readahead_blks
= arg
;
1691 } else if (token
== Opt_init_itable
) {
1692 set_opt(sb
, INIT_INODE_TABLE
);
1694 arg
= EXT4_DEF_LI_WAIT_MULT
;
1695 sbi
->s_li_wait_mult
= arg
;
1696 } else if (token
== Opt_max_dir_size_kb
) {
1697 sbi
->s_max_dir_size_kb
= arg
;
1698 } else if (token
== Opt_stripe
) {
1699 sbi
->s_stripe
= arg
;
1700 } else if (token
== Opt_resuid
) {
1701 uid
= make_kuid(current_user_ns(), arg
);
1702 if (!uid_valid(uid
)) {
1703 ext4_msg(sb
, KERN_ERR
, "Invalid uid value %d", arg
);
1706 sbi
->s_resuid
= uid
;
1707 } else if (token
== Opt_resgid
) {
1708 gid
= make_kgid(current_user_ns(), arg
);
1709 if (!gid_valid(gid
)) {
1710 ext4_msg(sb
, KERN_ERR
, "Invalid gid value %d", arg
);
1713 sbi
->s_resgid
= gid
;
1714 } else if (token
== Opt_journal_dev
) {
1716 ext4_msg(sb
, KERN_ERR
,
1717 "Cannot specify journal on remount");
1720 *journal_devnum
= arg
;
1721 } else if (token
== Opt_journal_path
) {
1723 struct inode
*journal_inode
;
1728 ext4_msg(sb
, KERN_ERR
,
1729 "Cannot specify journal on remount");
1732 journal_path
= match_strdup(&args
[0]);
1733 if (!journal_path
) {
1734 ext4_msg(sb
, KERN_ERR
, "error: could not dup "
1735 "journal device string");
1739 error
= kern_path(journal_path
, LOOKUP_FOLLOW
, &path
);
1741 ext4_msg(sb
, KERN_ERR
, "error: could not find "
1742 "journal device path: error %d", error
);
1743 kfree(journal_path
);
1747 journal_inode
= d_inode(path
.dentry
);
1748 if (!S_ISBLK(journal_inode
->i_mode
)) {
1749 ext4_msg(sb
, KERN_ERR
, "error: journal path %s "
1750 "is not a block device", journal_path
);
1752 kfree(journal_path
);
1756 *journal_devnum
= new_encode_dev(journal_inode
->i_rdev
);
1758 kfree(journal_path
);
1759 } else if (token
== Opt_journal_ioprio
) {
1761 ext4_msg(sb
, KERN_ERR
, "Invalid journal IO priority"
1766 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
, arg
);
1767 } else if (token
== Opt_test_dummy_encryption
) {
1768 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1769 sbi
->s_mount_flags
|= EXT4_MF_TEST_DUMMY_ENCRYPTION
;
1770 ext4_msg(sb
, KERN_WARNING
,
1771 "Test dummy encryption mode enabled");
1773 ext4_msg(sb
, KERN_WARNING
,
1774 "Test dummy encryption mount option ignored");
1776 } else if (m
->flags
& MOPT_DATAJ
) {
1778 if (!sbi
->s_journal
)
1779 ext4_msg(sb
, KERN_WARNING
, "Remounting file system with no journal so ignoring journalled data option");
1780 else if (test_opt(sb
, DATA_FLAGS
) != m
->mount_opt
) {
1781 ext4_msg(sb
, KERN_ERR
,
1782 "Cannot change data mode on remount");
1786 clear_opt(sb
, DATA_FLAGS
);
1787 sbi
->s_mount_opt
|= m
->mount_opt
;
1790 } else if (m
->flags
& MOPT_QFMT
) {
1791 if (sb_any_quota_loaded(sb
) &&
1792 sbi
->s_jquota_fmt
!= m
->mount_opt
) {
1793 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled "
1794 "quota options when quota turned on");
1797 if (ext4_has_feature_quota(sb
)) {
1798 ext4_msg(sb
, KERN_INFO
,
1799 "Quota format mount options ignored "
1800 "when QUOTA feature is enabled");
1803 sbi
->s_jquota_fmt
= m
->mount_opt
;
1805 } else if (token
== Opt_dax
) {
1806 #ifdef CONFIG_FS_DAX
1807 ext4_msg(sb
, KERN_WARNING
,
1808 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1809 sbi
->s_mount_opt
|= m
->mount_opt
;
1811 ext4_msg(sb
, KERN_INFO
, "dax option not supported");
1814 } else if (token
== Opt_data_err_abort
) {
1815 sbi
->s_mount_opt
|= m
->mount_opt
;
1816 } else if (token
== Opt_data_err_ignore
) {
1817 sbi
->s_mount_opt
&= ~m
->mount_opt
;
1821 if (m
->flags
& MOPT_CLEAR
)
1823 else if (unlikely(!(m
->flags
& MOPT_SET
))) {
1824 ext4_msg(sb
, KERN_WARNING
,
1825 "buggy handling of option %s", opt
);
1830 sbi
->s_mount_opt
|= m
->mount_opt
;
1832 sbi
->s_mount_opt
&= ~m
->mount_opt
;
1837 static int parse_options(char *options
, struct super_block
*sb
,
1838 unsigned long *journal_devnum
,
1839 unsigned int *journal_ioprio
,
1842 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1844 substring_t args
[MAX_OPT_ARGS
];
1850 while ((p
= strsep(&options
, ",")) != NULL
) {
1854 * Initialize args struct so we know whether arg was
1855 * found; some options take optional arguments.
1857 args
[0].to
= args
[0].from
= NULL
;
1858 token
= match_token(p
, tokens
, args
);
1859 if (handle_mount_opt(sb
, p
, token
, args
, journal_devnum
,
1860 journal_ioprio
, is_remount
) < 0)
1865 * We do the test below only for project quotas. 'usrquota' and
1866 * 'grpquota' mount options are allowed even without quota feature
1867 * to support legacy quotas in quota files.
1869 if (test_opt(sb
, PRJQUOTA
) && !ext4_has_feature_project(sb
)) {
1870 ext4_msg(sb
, KERN_ERR
, "Project quota feature not enabled. "
1871 "Cannot enable project quota enforcement.");
1874 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1875 if (test_opt(sb
, USRQUOTA
) && sbi
->s_qf_names
[USRQUOTA
])
1876 clear_opt(sb
, USRQUOTA
);
1878 if (test_opt(sb
, GRPQUOTA
) && sbi
->s_qf_names
[GRPQUOTA
])
1879 clear_opt(sb
, GRPQUOTA
);
1881 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
1882 ext4_msg(sb
, KERN_ERR
, "old and new quota "
1887 if (!sbi
->s_jquota_fmt
) {
1888 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1894 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
1896 BLOCK_SIZE
<< le32_to_cpu(sbi
->s_es
->s_log_block_size
);
1898 if (blocksize
< PAGE_SIZE
) {
1899 ext4_msg(sb
, KERN_ERR
, "can't mount with "
1900 "dioread_nolock if block size != PAGE_SIZE");
1907 static inline void ext4_show_quota_options(struct seq_file
*seq
,
1908 struct super_block
*sb
)
1910 #if defined(CONFIG_QUOTA)
1911 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1913 if (sbi
->s_jquota_fmt
) {
1916 switch (sbi
->s_jquota_fmt
) {
1927 seq_printf(seq
, ",jqfmt=%s", fmtname
);
1930 if (sbi
->s_qf_names
[USRQUOTA
])
1931 seq_show_option(seq
, "usrjquota", sbi
->s_qf_names
[USRQUOTA
]);
1933 if (sbi
->s_qf_names
[GRPQUOTA
])
1934 seq_show_option(seq
, "grpjquota", sbi
->s_qf_names
[GRPQUOTA
]);
1938 static const char *token2str(int token
)
1940 const struct match_token
*t
;
1942 for (t
= tokens
; t
->token
!= Opt_err
; t
++)
1943 if (t
->token
== token
&& !strchr(t
->pattern
, '='))
1950 * - it's set to a non-default value OR
1951 * - if the per-sb default is different from the global default
1953 static int _ext4_show_options(struct seq_file
*seq
, struct super_block
*sb
,
1956 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1957 struct ext4_super_block
*es
= sbi
->s_es
;
1958 int def_errors
, def_mount_opt
= nodefs
? 0 : sbi
->s_def_mount_opt
;
1959 const struct mount_opts
*m
;
1960 char sep
= nodefs
? '\n' : ',';
1962 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1963 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1965 if (sbi
->s_sb_block
!= 1)
1966 SEQ_OPTS_PRINT("sb=%llu", sbi
->s_sb_block
);
1968 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++) {
1969 int want_set
= m
->flags
& MOPT_SET
;
1970 if (((m
->flags
& (MOPT_SET
|MOPT_CLEAR
)) == 0) ||
1971 (m
->flags
& MOPT_CLEAR_ERR
))
1973 if (!(m
->mount_opt
& (sbi
->s_mount_opt
^ def_mount_opt
)))
1974 continue; /* skip if same as the default */
1976 (sbi
->s_mount_opt
& m
->mount_opt
) != m
->mount_opt
) ||
1977 (!want_set
&& (sbi
->s_mount_opt
& m
->mount_opt
)))
1978 continue; /* select Opt_noFoo vs Opt_Foo */
1979 SEQ_OPTS_PRINT("%s", token2str(m
->token
));
1982 if (nodefs
|| !uid_eq(sbi
->s_resuid
, make_kuid(&init_user_ns
, EXT4_DEF_RESUID
)) ||
1983 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
)
1984 SEQ_OPTS_PRINT("resuid=%u",
1985 from_kuid_munged(&init_user_ns
, sbi
->s_resuid
));
1986 if (nodefs
|| !gid_eq(sbi
->s_resgid
, make_kgid(&init_user_ns
, EXT4_DEF_RESGID
)) ||
1987 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
)
1988 SEQ_OPTS_PRINT("resgid=%u",
1989 from_kgid_munged(&init_user_ns
, sbi
->s_resgid
));
1990 def_errors
= nodefs
? -1 : le16_to_cpu(es
->s_errors
);
1991 if (test_opt(sb
, ERRORS_RO
) && def_errors
!= EXT4_ERRORS_RO
)
1992 SEQ_OPTS_PUTS("errors=remount-ro");
1993 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
1994 SEQ_OPTS_PUTS("errors=continue");
1995 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
1996 SEQ_OPTS_PUTS("errors=panic");
1997 if (nodefs
|| sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
)
1998 SEQ_OPTS_PRINT("commit=%lu", sbi
->s_commit_interval
/ HZ
);
1999 if (nodefs
|| sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
)
2000 SEQ_OPTS_PRINT("min_batch_time=%u", sbi
->s_min_batch_time
);
2001 if (nodefs
|| sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
)
2002 SEQ_OPTS_PRINT("max_batch_time=%u", sbi
->s_max_batch_time
);
2003 if (sb
->s_flags
& MS_I_VERSION
)
2004 SEQ_OPTS_PUTS("i_version");
2005 if (nodefs
|| sbi
->s_stripe
)
2006 SEQ_OPTS_PRINT("stripe=%lu", sbi
->s_stripe
);
2007 if (EXT4_MOUNT_DATA_FLAGS
& (sbi
->s_mount_opt
^ def_mount_opt
)) {
2008 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
2009 SEQ_OPTS_PUTS("data=journal");
2010 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
2011 SEQ_OPTS_PUTS("data=ordered");
2012 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
2013 SEQ_OPTS_PUTS("data=writeback");
2016 sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
2017 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2018 sbi
->s_inode_readahead_blks
);
2020 if (nodefs
|| (test_opt(sb
, INIT_INODE_TABLE
) &&
2021 (sbi
->s_li_wait_mult
!= EXT4_DEF_LI_WAIT_MULT
)))
2022 SEQ_OPTS_PRINT("init_itable=%u", sbi
->s_li_wait_mult
);
2023 if (nodefs
|| sbi
->s_max_dir_size_kb
)
2024 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi
->s_max_dir_size_kb
);
2025 if (test_opt(sb
, DATA_ERR_ABORT
))
2026 SEQ_OPTS_PUTS("data_err=abort");
2028 ext4_show_quota_options(seq
, sb
);
2032 static int ext4_show_options(struct seq_file
*seq
, struct dentry
*root
)
2034 return _ext4_show_options(seq
, root
->d_sb
, 0);
2037 int ext4_seq_options_show(struct seq_file
*seq
, void *offset
)
2039 struct super_block
*sb
= seq
->private;
2042 seq_puts(seq
, (sb
->s_flags
& MS_RDONLY
) ? "ro" : "rw");
2043 rc
= _ext4_show_options(seq
, sb
, 1);
2044 seq_puts(seq
, "\n");
2048 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
2051 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2054 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
2055 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
2056 "forcing read-only mode");
2061 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
2062 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
2063 "running e2fsck is recommended");
2064 else if (sbi
->s_mount_state
& EXT4_ERROR_FS
)
2065 ext4_msg(sb
, KERN_WARNING
,
2066 "warning: mounting fs with errors, "
2067 "running e2fsck is recommended");
2068 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) > 0 &&
2069 le16_to_cpu(es
->s_mnt_count
) >=
2070 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
2071 ext4_msg(sb
, KERN_WARNING
,
2072 "warning: maximal mount count reached, "
2073 "running e2fsck is recommended");
2074 else if (le32_to_cpu(es
->s_checkinterval
) &&
2075 (le32_to_cpu(es
->s_lastcheck
) +
2076 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
2077 ext4_msg(sb
, KERN_WARNING
,
2078 "warning: checktime reached, "
2079 "running e2fsck is recommended");
2080 if (!sbi
->s_journal
)
2081 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
2082 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
2083 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
2084 le16_add_cpu(&es
->s_mnt_count
, 1);
2085 es
->s_mtime
= cpu_to_le32(get_seconds());
2086 ext4_update_dynamic_rev(sb
);
2088 ext4_set_feature_journal_needs_recovery(sb
);
2090 ext4_commit_super(sb
, 1);
2092 if (test_opt(sb
, DEBUG
))
2093 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
2094 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2096 sbi
->s_groups_count
,
2097 EXT4_BLOCKS_PER_GROUP(sb
),
2098 EXT4_INODES_PER_GROUP(sb
),
2099 sbi
->s_mount_opt
, sbi
->s_mount_opt2
);
2101 cleancache_init_fs(sb
);
2105 int ext4_alloc_flex_bg_array(struct super_block
*sb
, ext4_group_t ngroup
)
2107 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2108 struct flex_groups
*new_groups
;
2111 if (!sbi
->s_log_groups_per_flex
)
2114 size
= ext4_flex_group(sbi
, ngroup
- 1) + 1;
2115 if (size
<= sbi
->s_flex_groups_allocated
)
2118 size
= roundup_pow_of_two(size
* sizeof(struct flex_groups
));
2119 new_groups
= ext4_kvzalloc(size
, GFP_KERNEL
);
2121 ext4_msg(sb
, KERN_ERR
, "not enough memory for %d flex groups",
2122 size
/ (int) sizeof(struct flex_groups
));
2126 if (sbi
->s_flex_groups
) {
2127 memcpy(new_groups
, sbi
->s_flex_groups
,
2128 (sbi
->s_flex_groups_allocated
*
2129 sizeof(struct flex_groups
)));
2130 kvfree(sbi
->s_flex_groups
);
2132 sbi
->s_flex_groups
= new_groups
;
2133 sbi
->s_flex_groups_allocated
= size
/ sizeof(struct flex_groups
);
2137 static int ext4_fill_flex_info(struct super_block
*sb
)
2139 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2140 struct ext4_group_desc
*gdp
= NULL
;
2141 ext4_group_t flex_group
;
2144 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
2145 if (sbi
->s_log_groups_per_flex
< 1 || sbi
->s_log_groups_per_flex
> 31) {
2146 sbi
->s_log_groups_per_flex
= 0;
2150 err
= ext4_alloc_flex_bg_array(sb
, sbi
->s_groups_count
);
2154 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2155 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2157 flex_group
= ext4_flex_group(sbi
, i
);
2158 atomic_add(ext4_free_inodes_count(sb
, gdp
),
2159 &sbi
->s_flex_groups
[flex_group
].free_inodes
);
2160 atomic64_add(ext4_free_group_clusters(sb
, gdp
),
2161 &sbi
->s_flex_groups
[flex_group
].free_clusters
);
2162 atomic_add(ext4_used_dirs_count(sb
, gdp
),
2163 &sbi
->s_flex_groups
[flex_group
].used_dirs
);
2171 static __le16
ext4_group_desc_csum(struct super_block
*sb
, __u32 block_group
,
2172 struct ext4_group_desc
*gdp
)
2174 int offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
2176 __le32 le_group
= cpu_to_le32(block_group
);
2177 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2179 if (ext4_has_metadata_csum(sbi
->s_sb
)) {
2180 /* Use new metadata_csum algorithm */
2182 __u16 dummy_csum
= 0;
2184 csum32
= ext4_chksum(sbi
, sbi
->s_csum_seed
, (__u8
*)&le_group
,
2186 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)gdp
, offset
);
2187 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)&dummy_csum
,
2188 sizeof(dummy_csum
));
2189 offset
+= sizeof(dummy_csum
);
2190 if (offset
< sbi
->s_desc_size
)
2191 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)gdp
+ offset
,
2192 sbi
->s_desc_size
- offset
);
2194 crc
= csum32
& 0xFFFF;
2198 /* old crc16 code */
2199 if (!ext4_has_feature_gdt_csum(sb
))
2202 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
2203 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
2204 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
2205 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
2206 /* for checksum of struct ext4_group_desc do the rest...*/
2207 if (ext4_has_feature_64bit(sb
) &&
2208 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
2209 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
2210 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
2214 return cpu_to_le16(crc
);
2217 int ext4_group_desc_csum_verify(struct super_block
*sb
, __u32 block_group
,
2218 struct ext4_group_desc
*gdp
)
2220 if (ext4_has_group_desc_csum(sb
) &&
2221 (gdp
->bg_checksum
!= ext4_group_desc_csum(sb
, block_group
, gdp
)))
2227 void ext4_group_desc_csum_set(struct super_block
*sb
, __u32 block_group
,
2228 struct ext4_group_desc
*gdp
)
2230 if (!ext4_has_group_desc_csum(sb
))
2232 gdp
->bg_checksum
= ext4_group_desc_csum(sb
, block_group
, gdp
);
2235 /* Called at mount-time, super-block is locked */
2236 static int ext4_check_descriptors(struct super_block
*sb
,
2237 ext4_fsblk_t sb_block
,
2238 ext4_group_t
*first_not_zeroed
)
2240 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2241 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
2242 ext4_fsblk_t last_block
;
2243 ext4_fsblk_t block_bitmap
;
2244 ext4_fsblk_t inode_bitmap
;
2245 ext4_fsblk_t inode_table
;
2246 int flexbg_flag
= 0;
2247 ext4_group_t i
, grp
= sbi
->s_groups_count
;
2249 if (ext4_has_feature_flex_bg(sb
))
2252 ext4_debug("Checking group descriptors");
2254 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2255 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2257 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
2258 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
2260 last_block
= first_block
+
2261 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2263 if ((grp
== sbi
->s_groups_count
) &&
2264 !(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2267 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
2268 if (block_bitmap
== sb_block
) {
2269 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2270 "Block bitmap for group %u overlaps "
2273 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
2274 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2275 "Block bitmap for group %u not in group "
2276 "(block %llu)!", i
, block_bitmap
);
2279 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
2280 if (inode_bitmap
== sb_block
) {
2281 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2282 "Inode bitmap for group %u overlaps "
2285 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
2286 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2287 "Inode bitmap for group %u not in group "
2288 "(block %llu)!", i
, inode_bitmap
);
2291 inode_table
= ext4_inode_table(sb
, gdp
);
2292 if (inode_table
== sb_block
) {
2293 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2294 "Inode table for group %u overlaps "
2297 if (inode_table
< first_block
||
2298 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
2299 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2300 "Inode table for group %u not in group "
2301 "(block %llu)!", i
, inode_table
);
2304 ext4_lock_group(sb
, i
);
2305 if (!ext4_group_desc_csum_verify(sb
, i
, gdp
)) {
2306 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2307 "Checksum for group %u failed (%u!=%u)",
2308 i
, le16_to_cpu(ext4_group_desc_csum(sb
, i
,
2309 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
2310 if (!(sb
->s_flags
& MS_RDONLY
)) {
2311 ext4_unlock_group(sb
, i
);
2315 ext4_unlock_group(sb
, i
);
2317 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
2319 if (NULL
!= first_not_zeroed
)
2320 *first_not_zeroed
= grp
;
2324 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2325 * the superblock) which were deleted from all directories, but held open by
2326 * a process at the time of a crash. We walk the list and try to delete these
2327 * inodes at recovery time (only with a read-write filesystem).
2329 * In order to keep the orphan inode chain consistent during traversal (in
2330 * case of crash during recovery), we link each inode into the superblock
2331 * orphan list_head and handle it the same way as an inode deletion during
2332 * normal operation (which journals the operations for us).
2334 * We only do an iget() and an iput() on each inode, which is very safe if we
2335 * accidentally point at an in-use or already deleted inode. The worst that
2336 * can happen in this case is that we get a "bit already cleared" message from
2337 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2338 * e2fsck was run on this filesystem, and it must have already done the orphan
2339 * inode cleanup for us, so we can safely abort without any further action.
2341 static void ext4_orphan_cleanup(struct super_block
*sb
,
2342 struct ext4_super_block
*es
)
2344 unsigned int s_flags
= sb
->s_flags
;
2345 int ret
, nr_orphans
= 0, nr_truncates
= 0;
2349 if (!es
->s_last_orphan
) {
2350 jbd_debug(4, "no orphan inodes to clean up\n");
2354 if (bdev_read_only(sb
->s_bdev
)) {
2355 ext4_msg(sb
, KERN_ERR
, "write access "
2356 "unavailable, skipping orphan cleanup");
2360 /* Check if feature set would not allow a r/w mount */
2361 if (!ext4_feature_set_ok(sb
, 0)) {
2362 ext4_msg(sb
, KERN_INFO
, "Skipping orphan cleanup due to "
2363 "unknown ROCOMPAT features");
2367 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2368 /* don't clear list on RO mount w/ errors */
2369 if (es
->s_last_orphan
&& !(s_flags
& MS_RDONLY
)) {
2370 ext4_msg(sb
, KERN_INFO
, "Errors on filesystem, "
2371 "clearing orphan list.\n");
2372 es
->s_last_orphan
= 0;
2374 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2378 if (s_flags
& MS_RDONLY
) {
2379 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
2380 sb
->s_flags
&= ~MS_RDONLY
;
2383 /* Needed for iput() to work correctly and not trash data */
2384 sb
->s_flags
|= MS_ACTIVE
;
2385 /* Turn on quotas so that they are updated correctly */
2386 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
2387 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
2388 int ret
= ext4_quota_on_mount(sb
, i
);
2390 ext4_msg(sb
, KERN_ERR
,
2391 "Cannot turn on journaled "
2392 "quota: error %d", ret
);
2397 while (es
->s_last_orphan
) {
2398 struct inode
*inode
;
2401 * We may have encountered an error during cleanup; if
2402 * so, skip the rest.
2404 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2405 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2406 es
->s_last_orphan
= 0;
2410 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
2411 if (IS_ERR(inode
)) {
2412 es
->s_last_orphan
= 0;
2416 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
2417 dquot_initialize(inode
);
2418 if (inode
->i_nlink
) {
2419 if (test_opt(sb
, DEBUG
))
2420 ext4_msg(sb
, KERN_DEBUG
,
2421 "%s: truncating inode %lu to %lld bytes",
2422 __func__
, inode
->i_ino
, inode
->i_size
);
2423 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2424 inode
->i_ino
, inode
->i_size
);
2426 truncate_inode_pages(inode
->i_mapping
, inode
->i_size
);
2427 ret
= ext4_truncate(inode
);
2429 ext4_std_error(inode
->i_sb
, ret
);
2430 inode_unlock(inode
);
2433 if (test_opt(sb
, DEBUG
))
2434 ext4_msg(sb
, KERN_DEBUG
,
2435 "%s: deleting unreferenced inode %lu",
2436 __func__
, inode
->i_ino
);
2437 jbd_debug(2, "deleting unreferenced inode %lu\n",
2441 iput(inode
); /* The delete magic happens here! */
2444 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2447 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
2448 PLURAL(nr_orphans
));
2450 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
2451 PLURAL(nr_truncates
));
2453 /* Turn quotas off */
2454 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
2455 if (sb_dqopt(sb
)->files
[i
])
2456 dquot_quota_off(sb
, i
);
2459 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
2463 * Maximal extent format file size.
2464 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2465 * extent format containers, within a sector_t, and within i_blocks
2466 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2467 * so that won't be a limiting factor.
2469 * However there is other limiting factor. We do store extents in the form
2470 * of starting block and length, hence the resulting length of the extent
2471 * covering maximum file size must fit into on-disk format containers as
2472 * well. Given that length is always by 1 unit bigger than max unit (because
2473 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2475 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2477 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
2480 loff_t upper_limit
= MAX_LFS_FILESIZE
;
2482 /* small i_blocks in vfs inode? */
2483 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2485 * CONFIG_LBDAF is not enabled implies the inode
2486 * i_block represent total blocks in 512 bytes
2487 * 32 == size of vfs inode i_blocks * 8
2489 upper_limit
= (1LL << 32) - 1;
2491 /* total blocks in file system block size */
2492 upper_limit
>>= (blkbits
- 9);
2493 upper_limit
<<= blkbits
;
2497 * 32-bit extent-start container, ee_block. We lower the maxbytes
2498 * by one fs block, so ee_len can cover the extent of maximum file
2501 res
= (1LL << 32) - 1;
2504 /* Sanity check against vm- & vfs- imposed limits */
2505 if (res
> upper_limit
)
2512 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2513 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2514 * We need to be 1 filesystem block less than the 2^48 sector limit.
2516 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
2518 loff_t res
= EXT4_NDIR_BLOCKS
;
2521 /* This is calculated to be the largest file size for a dense, block
2522 * mapped file such that the file's total number of 512-byte sectors,
2523 * including data and all indirect blocks, does not exceed (2^48 - 1).
2525 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2526 * number of 512-byte sectors of the file.
2529 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2531 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2532 * the inode i_block field represents total file blocks in
2533 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2535 upper_limit
= (1LL << 32) - 1;
2537 /* total blocks in file system block size */
2538 upper_limit
>>= (bits
- 9);
2542 * We use 48 bit ext4_inode i_blocks
2543 * With EXT4_HUGE_FILE_FL set the i_blocks
2544 * represent total number of blocks in
2545 * file system block size
2547 upper_limit
= (1LL << 48) - 1;
2551 /* indirect blocks */
2553 /* double indirect blocks */
2554 meta_blocks
+= 1 + (1LL << (bits
-2));
2555 /* tripple indirect blocks */
2556 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
2558 upper_limit
-= meta_blocks
;
2559 upper_limit
<<= bits
;
2561 res
+= 1LL << (bits
-2);
2562 res
+= 1LL << (2*(bits
-2));
2563 res
+= 1LL << (3*(bits
-2));
2565 if (res
> upper_limit
)
2568 if (res
> MAX_LFS_FILESIZE
)
2569 res
= MAX_LFS_FILESIZE
;
2574 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
2575 ext4_fsblk_t logical_sb_block
, int nr
)
2577 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2578 ext4_group_t bg
, first_meta_bg
;
2581 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
2583 if (!ext4_has_feature_meta_bg(sb
) || nr
< first_meta_bg
)
2584 return logical_sb_block
+ nr
+ 1;
2585 bg
= sbi
->s_desc_per_block
* nr
;
2586 if (ext4_bg_has_super(sb
, bg
))
2590 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2591 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2592 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2595 if (sb
->s_blocksize
== 1024 && nr
== 0 &&
2596 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_first_data_block
) == 0)
2599 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
2603 * ext4_get_stripe_size: Get the stripe size.
2604 * @sbi: In memory super block info
2606 * If we have specified it via mount option, then
2607 * use the mount option value. If the value specified at mount time is
2608 * greater than the blocks per group use the super block value.
2609 * If the super block value is greater than blocks per group return 0.
2610 * Allocator needs it be less than blocks per group.
2613 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
2615 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
2616 unsigned long stripe_width
=
2617 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
2620 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
2621 ret
= sbi
->s_stripe
;
2622 else if (stripe_width
<= sbi
->s_blocks_per_group
)
2624 else if (stride
<= sbi
->s_blocks_per_group
)
2630 * If the stripe width is 1, this makes no sense and
2631 * we set it to 0 to turn off stripe handling code.
2640 * Check whether this filesystem can be mounted based on
2641 * the features present and the RDONLY/RDWR mount requested.
2642 * Returns 1 if this filesystem can be mounted as requested,
2643 * 0 if it cannot be.
2645 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
2647 if (ext4_has_unknown_ext4_incompat_features(sb
)) {
2648 ext4_msg(sb
, KERN_ERR
,
2649 "Couldn't mount because of "
2650 "unsupported optional features (%x)",
2651 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2652 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2659 if (ext4_has_feature_readonly(sb
)) {
2660 ext4_msg(sb
, KERN_INFO
, "filesystem is read-only");
2661 sb
->s_flags
|= MS_RDONLY
;
2665 /* Check that feature set is OK for a read-write mount */
2666 if (ext4_has_unknown_ext4_ro_compat_features(sb
)) {
2667 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
2668 "unsupported optional features (%x)",
2669 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2670 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2674 * Large file size enabled file system can only be mounted
2675 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2677 if (ext4_has_feature_huge_file(sb
)) {
2678 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
2679 ext4_msg(sb
, KERN_ERR
, "Filesystem with huge files "
2680 "cannot be mounted RDWR without "
2685 if (ext4_has_feature_bigalloc(sb
) && !ext4_has_feature_extents(sb
)) {
2686 ext4_msg(sb
, KERN_ERR
,
2687 "Can't support bigalloc feature without "
2688 "extents feature\n");
2692 #ifndef CONFIG_QUOTA
2693 if (ext4_has_feature_quota(sb
) && !readonly
) {
2694 ext4_msg(sb
, KERN_ERR
,
2695 "Filesystem with quota feature cannot be mounted RDWR "
2696 "without CONFIG_QUOTA");
2699 if (ext4_has_feature_project(sb
) && !readonly
) {
2700 ext4_msg(sb
, KERN_ERR
,
2701 "Filesystem with project quota feature cannot be mounted RDWR "
2702 "without CONFIG_QUOTA");
2705 #endif /* CONFIG_QUOTA */
2710 * This function is called once a day if we have errors logged
2711 * on the file system
2713 static void print_daily_error_info(unsigned long arg
)
2715 struct super_block
*sb
= (struct super_block
*) arg
;
2716 struct ext4_sb_info
*sbi
;
2717 struct ext4_super_block
*es
;
2722 if (es
->s_error_count
)
2723 /* fsck newer than v1.41.13 is needed to clean this condition. */
2724 ext4_msg(sb
, KERN_NOTICE
, "error count since last fsck: %u",
2725 le32_to_cpu(es
->s_error_count
));
2726 if (es
->s_first_error_time
) {
2727 printk(KERN_NOTICE
"EXT4-fs (%s): initial error at time %u: %.*s:%d",
2728 sb
->s_id
, le32_to_cpu(es
->s_first_error_time
),
2729 (int) sizeof(es
->s_first_error_func
),
2730 es
->s_first_error_func
,
2731 le32_to_cpu(es
->s_first_error_line
));
2732 if (es
->s_first_error_ino
)
2733 printk(KERN_CONT
": inode %u",
2734 le32_to_cpu(es
->s_first_error_ino
));
2735 if (es
->s_first_error_block
)
2736 printk(KERN_CONT
": block %llu", (unsigned long long)
2737 le64_to_cpu(es
->s_first_error_block
));
2738 printk(KERN_CONT
"\n");
2740 if (es
->s_last_error_time
) {
2741 printk(KERN_NOTICE
"EXT4-fs (%s): last error at time %u: %.*s:%d",
2742 sb
->s_id
, le32_to_cpu(es
->s_last_error_time
),
2743 (int) sizeof(es
->s_last_error_func
),
2744 es
->s_last_error_func
,
2745 le32_to_cpu(es
->s_last_error_line
));
2746 if (es
->s_last_error_ino
)
2747 printk(KERN_CONT
": inode %u",
2748 le32_to_cpu(es
->s_last_error_ino
));
2749 if (es
->s_last_error_block
)
2750 printk(KERN_CONT
": block %llu", (unsigned long long)
2751 le64_to_cpu(es
->s_last_error_block
));
2752 printk(KERN_CONT
"\n");
2754 mod_timer(&sbi
->s_err_report
, jiffies
+ 24*60*60*HZ
); /* Once a day */
2757 /* Find next suitable group and run ext4_init_inode_table */
2758 static int ext4_run_li_request(struct ext4_li_request
*elr
)
2760 struct ext4_group_desc
*gdp
= NULL
;
2761 ext4_group_t group
, ngroups
;
2762 struct super_block
*sb
;
2763 unsigned long timeout
= 0;
2767 ngroups
= EXT4_SB(sb
)->s_groups_count
;
2769 for (group
= elr
->lr_next_group
; group
< ngroups
; group
++) {
2770 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2776 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2780 if (group
>= ngroups
)
2785 ret
= ext4_init_inode_table(sb
, group
,
2786 elr
->lr_timeout
? 0 : 1);
2787 if (elr
->lr_timeout
== 0) {
2788 timeout
= (jiffies
- timeout
) *
2789 elr
->lr_sbi
->s_li_wait_mult
;
2790 elr
->lr_timeout
= timeout
;
2792 elr
->lr_next_sched
= jiffies
+ elr
->lr_timeout
;
2793 elr
->lr_next_group
= group
+ 1;
2799 * Remove lr_request from the list_request and free the
2800 * request structure. Should be called with li_list_mtx held
2802 static void ext4_remove_li_request(struct ext4_li_request
*elr
)
2804 struct ext4_sb_info
*sbi
;
2811 list_del(&elr
->lr_request
);
2812 sbi
->s_li_request
= NULL
;
2816 static void ext4_unregister_li_request(struct super_block
*sb
)
2818 mutex_lock(&ext4_li_mtx
);
2819 if (!ext4_li_info
) {
2820 mutex_unlock(&ext4_li_mtx
);
2824 mutex_lock(&ext4_li_info
->li_list_mtx
);
2825 ext4_remove_li_request(EXT4_SB(sb
)->s_li_request
);
2826 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2827 mutex_unlock(&ext4_li_mtx
);
2830 static struct task_struct
*ext4_lazyinit_task
;
2833 * This is the function where ext4lazyinit thread lives. It walks
2834 * through the request list searching for next scheduled filesystem.
2835 * When such a fs is found, run the lazy initialization request
2836 * (ext4_rn_li_request) and keep track of the time spend in this
2837 * function. Based on that time we compute next schedule time of
2838 * the request. When walking through the list is complete, compute
2839 * next waking time and put itself into sleep.
2841 static int ext4_lazyinit_thread(void *arg
)
2843 struct ext4_lazy_init
*eli
= (struct ext4_lazy_init
*)arg
;
2844 struct list_head
*pos
, *n
;
2845 struct ext4_li_request
*elr
;
2846 unsigned long next_wakeup
, cur
;
2848 BUG_ON(NULL
== eli
);
2852 next_wakeup
= MAX_JIFFY_OFFSET
;
2854 mutex_lock(&eli
->li_list_mtx
);
2855 if (list_empty(&eli
->li_request_list
)) {
2856 mutex_unlock(&eli
->li_list_mtx
);
2859 list_for_each_safe(pos
, n
, &eli
->li_request_list
) {
2862 elr
= list_entry(pos
, struct ext4_li_request
,
2865 if (time_before(jiffies
, elr
->lr_next_sched
)) {
2866 if (time_before(elr
->lr_next_sched
, next_wakeup
))
2867 next_wakeup
= elr
->lr_next_sched
;
2870 if (down_read_trylock(&elr
->lr_super
->s_umount
)) {
2871 if (sb_start_write_trylock(elr
->lr_super
)) {
2874 * We hold sb->s_umount, sb can not
2875 * be removed from the list, it is
2876 * now safe to drop li_list_mtx
2878 mutex_unlock(&eli
->li_list_mtx
);
2879 err
= ext4_run_li_request(elr
);
2880 sb_end_write(elr
->lr_super
);
2881 mutex_lock(&eli
->li_list_mtx
);
2884 up_read((&elr
->lr_super
->s_umount
));
2886 /* error, remove the lazy_init job */
2888 ext4_remove_li_request(elr
);
2892 elr
->lr_next_sched
= jiffies
+
2894 % (EXT4_DEF_LI_MAX_START_DELAY
* HZ
));
2896 if (time_before(elr
->lr_next_sched
, next_wakeup
))
2897 next_wakeup
= elr
->lr_next_sched
;
2899 mutex_unlock(&eli
->li_list_mtx
);
2904 if ((time_after_eq(cur
, next_wakeup
)) ||
2905 (MAX_JIFFY_OFFSET
== next_wakeup
)) {
2910 schedule_timeout_interruptible(next_wakeup
- cur
);
2912 if (kthread_should_stop()) {
2913 ext4_clear_request_list();
2920 * It looks like the request list is empty, but we need
2921 * to check it under the li_list_mtx lock, to prevent any
2922 * additions into it, and of course we should lock ext4_li_mtx
2923 * to atomically free the list and ext4_li_info, because at
2924 * this point another ext4 filesystem could be registering
2927 mutex_lock(&ext4_li_mtx
);
2928 mutex_lock(&eli
->li_list_mtx
);
2929 if (!list_empty(&eli
->li_request_list
)) {
2930 mutex_unlock(&eli
->li_list_mtx
);
2931 mutex_unlock(&ext4_li_mtx
);
2934 mutex_unlock(&eli
->li_list_mtx
);
2935 kfree(ext4_li_info
);
2936 ext4_li_info
= NULL
;
2937 mutex_unlock(&ext4_li_mtx
);
2942 static void ext4_clear_request_list(void)
2944 struct list_head
*pos
, *n
;
2945 struct ext4_li_request
*elr
;
2947 mutex_lock(&ext4_li_info
->li_list_mtx
);
2948 list_for_each_safe(pos
, n
, &ext4_li_info
->li_request_list
) {
2949 elr
= list_entry(pos
, struct ext4_li_request
,
2951 ext4_remove_li_request(elr
);
2953 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2956 static int ext4_run_lazyinit_thread(void)
2958 ext4_lazyinit_task
= kthread_run(ext4_lazyinit_thread
,
2959 ext4_li_info
, "ext4lazyinit");
2960 if (IS_ERR(ext4_lazyinit_task
)) {
2961 int err
= PTR_ERR(ext4_lazyinit_task
);
2962 ext4_clear_request_list();
2963 kfree(ext4_li_info
);
2964 ext4_li_info
= NULL
;
2965 printk(KERN_CRIT
"EXT4-fs: error %d creating inode table "
2966 "initialization thread\n",
2970 ext4_li_info
->li_state
|= EXT4_LAZYINIT_RUNNING
;
2975 * Check whether it make sense to run itable init. thread or not.
2976 * If there is at least one uninitialized inode table, return
2977 * corresponding group number, else the loop goes through all
2978 * groups and return total number of groups.
2980 static ext4_group_t
ext4_has_uninit_itable(struct super_block
*sb
)
2982 ext4_group_t group
, ngroups
= EXT4_SB(sb
)->s_groups_count
;
2983 struct ext4_group_desc
*gdp
= NULL
;
2985 for (group
= 0; group
< ngroups
; group
++) {
2986 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2990 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2997 static int ext4_li_info_new(void)
2999 struct ext4_lazy_init
*eli
= NULL
;
3001 eli
= kzalloc(sizeof(*eli
), GFP_KERNEL
);
3005 INIT_LIST_HEAD(&eli
->li_request_list
);
3006 mutex_init(&eli
->li_list_mtx
);
3008 eli
->li_state
|= EXT4_LAZYINIT_QUIT
;
3015 static struct ext4_li_request
*ext4_li_request_new(struct super_block
*sb
,
3018 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3019 struct ext4_li_request
*elr
;
3021 elr
= kzalloc(sizeof(*elr
), GFP_KERNEL
);
3027 elr
->lr_next_group
= start
;
3030 * Randomize first schedule time of the request to
3031 * spread the inode table initialization requests
3034 elr
->lr_next_sched
= jiffies
+ (prandom_u32() %
3035 (EXT4_DEF_LI_MAX_START_DELAY
* HZ
));
3039 int ext4_register_li_request(struct super_block
*sb
,
3040 ext4_group_t first_not_zeroed
)
3042 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3043 struct ext4_li_request
*elr
= NULL
;
3044 ext4_group_t ngroups
= EXT4_SB(sb
)->s_groups_count
;
3047 mutex_lock(&ext4_li_mtx
);
3048 if (sbi
->s_li_request
!= NULL
) {
3050 * Reset timeout so it can be computed again, because
3051 * s_li_wait_mult might have changed.
3053 sbi
->s_li_request
->lr_timeout
= 0;
3057 if (first_not_zeroed
== ngroups
||
3058 (sb
->s_flags
& MS_RDONLY
) ||
3059 !test_opt(sb
, INIT_INODE_TABLE
))
3062 elr
= ext4_li_request_new(sb
, first_not_zeroed
);
3068 if (NULL
== ext4_li_info
) {
3069 ret
= ext4_li_info_new();
3074 mutex_lock(&ext4_li_info
->li_list_mtx
);
3075 list_add(&elr
->lr_request
, &ext4_li_info
->li_request_list
);
3076 mutex_unlock(&ext4_li_info
->li_list_mtx
);
3078 sbi
->s_li_request
= elr
;
3080 * set elr to NULL here since it has been inserted to
3081 * the request_list and the removal and free of it is
3082 * handled by ext4_clear_request_list from now on.
3086 if (!(ext4_li_info
->li_state
& EXT4_LAZYINIT_RUNNING
)) {
3087 ret
= ext4_run_lazyinit_thread();
3092 mutex_unlock(&ext4_li_mtx
);
3099 * We do not need to lock anything since this is called on
3102 static void ext4_destroy_lazyinit_thread(void)
3105 * If thread exited earlier
3106 * there's nothing to be done.
3108 if (!ext4_li_info
|| !ext4_lazyinit_task
)
3111 kthread_stop(ext4_lazyinit_task
);
3114 static int set_journal_csum_feature_set(struct super_block
*sb
)
3117 int compat
, incompat
;
3118 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3120 if (ext4_has_metadata_csum(sb
)) {
3121 /* journal checksum v3 */
3123 incompat
= JBD2_FEATURE_INCOMPAT_CSUM_V3
;
3125 /* journal checksum v1 */
3126 compat
= JBD2_FEATURE_COMPAT_CHECKSUM
;
3130 jbd2_journal_clear_features(sbi
->s_journal
,
3131 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3132 JBD2_FEATURE_INCOMPAT_CSUM_V3
|
3133 JBD2_FEATURE_INCOMPAT_CSUM_V2
);
3134 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
3135 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3137 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
|
3139 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
3140 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3143 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3144 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3146 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3147 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3154 * Note: calculating the overhead so we can be compatible with
3155 * historical BSD practice is quite difficult in the face of
3156 * clusters/bigalloc. This is because multiple metadata blocks from
3157 * different block group can end up in the same allocation cluster.
3158 * Calculating the exact overhead in the face of clustered allocation
3159 * requires either O(all block bitmaps) in memory or O(number of block
3160 * groups**2) in time. We will still calculate the superblock for
3161 * older file systems --- and if we come across with a bigalloc file
3162 * system with zero in s_overhead_clusters the estimate will be close to
3163 * correct especially for very large cluster sizes --- but for newer
3164 * file systems, it's better to calculate this figure once at mkfs
3165 * time, and store it in the superblock. If the superblock value is
3166 * present (even for non-bigalloc file systems), we will use it.
3168 static int count_overhead(struct super_block
*sb
, ext4_group_t grp
,
3171 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3172 struct ext4_group_desc
*gdp
;
3173 ext4_fsblk_t first_block
, last_block
, b
;
3174 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3175 int s
, j
, count
= 0;
3177 if (!ext4_has_feature_bigalloc(sb
))
3178 return (ext4_bg_has_super(sb
, grp
) + ext4_bg_num_gdb(sb
, grp
) +
3179 sbi
->s_itb_per_group
+ 2);
3181 first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
) +
3182 (grp
* EXT4_BLOCKS_PER_GROUP(sb
));
3183 last_block
= first_block
+ EXT4_BLOCKS_PER_GROUP(sb
) - 1;
3184 for (i
= 0; i
< ngroups
; i
++) {
3185 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
3186 b
= ext4_block_bitmap(sb
, gdp
);
3187 if (b
>= first_block
&& b
<= last_block
) {
3188 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3191 b
= ext4_inode_bitmap(sb
, gdp
);
3192 if (b
>= first_block
&& b
<= last_block
) {
3193 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3196 b
= ext4_inode_table(sb
, gdp
);
3197 if (b
>= first_block
&& b
+ sbi
->s_itb_per_group
<= last_block
)
3198 for (j
= 0; j
< sbi
->s_itb_per_group
; j
++, b
++) {
3199 int c
= EXT4_B2C(sbi
, b
- first_block
);
3200 ext4_set_bit(c
, buf
);
3206 if (ext4_bg_has_super(sb
, grp
)) {
3207 ext4_set_bit(s
++, buf
);
3210 j
= ext4_bg_num_gdb(sb
, grp
);
3211 if (s
+ j
> EXT4_BLOCKS_PER_GROUP(sb
)) {
3212 ext4_error(sb
, "Invalid number of block group "
3213 "descriptor blocks: %d", j
);
3214 j
= EXT4_BLOCKS_PER_GROUP(sb
) - s
;
3218 ext4_set_bit(EXT4_B2C(sbi
, s
++), buf
);
3222 return EXT4_CLUSTERS_PER_GROUP(sb
) -
3223 ext4_count_free(buf
, EXT4_CLUSTERS_PER_GROUP(sb
) / 8);
3227 * Compute the overhead and stash it in sbi->s_overhead
3229 int ext4_calculate_overhead(struct super_block
*sb
)
3231 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3232 struct ext4_super_block
*es
= sbi
->s_es
;
3233 struct inode
*j_inode
;
3234 unsigned int j_blocks
, j_inum
= le32_to_cpu(es
->s_journal_inum
);
3235 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3236 ext4_fsblk_t overhead
= 0;
3237 char *buf
= (char *) get_zeroed_page(GFP_NOFS
);
3243 * Compute the overhead (FS structures). This is constant
3244 * for a given filesystem unless the number of block groups
3245 * changes so we cache the previous value until it does.
3249 * All of the blocks before first_data_block are overhead
3251 overhead
= EXT4_B2C(sbi
, le32_to_cpu(es
->s_first_data_block
));
3254 * Add the overhead found in each block group
3256 for (i
= 0; i
< ngroups
; i
++) {
3259 blks
= count_overhead(sb
, i
, buf
);
3262 memset(buf
, 0, PAGE_SIZE
);
3267 * Add the internal journal blocks whether the journal has been
3270 if (sbi
->s_journal
&& !sbi
->journal_bdev
)
3271 overhead
+= EXT4_NUM_B2C(sbi
, sbi
->s_journal
->j_maxlen
);
3272 else if (ext4_has_feature_journal(sb
) && !sbi
->s_journal
) {
3273 j_inode
= ext4_get_journal_inode(sb
, j_inum
);
3275 j_blocks
= j_inode
->i_size
>> sb
->s_blocksize_bits
;
3276 overhead
+= EXT4_NUM_B2C(sbi
, j_blocks
);
3279 ext4_msg(sb
, KERN_ERR
, "can't get journal size");
3282 sbi
->s_overhead
= overhead
;
3284 free_page((unsigned long) buf
);
3288 static void ext4_set_resv_clusters(struct super_block
*sb
)
3290 ext4_fsblk_t resv_clusters
;
3291 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3294 * There's no need to reserve anything when we aren't using extents.
3295 * The space estimates are exact, there are no unwritten extents,
3296 * hole punching doesn't need new metadata... This is needed especially
3297 * to keep ext2/3 backward compatibility.
3299 if (!ext4_has_feature_extents(sb
))
3302 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3303 * This should cover the situations where we can not afford to run
3304 * out of space like for example punch hole, or converting
3305 * unwritten extents in delalloc path. In most cases such
3306 * allocation would require 1, or 2 blocks, higher numbers are
3309 resv_clusters
= (ext4_blocks_count(sbi
->s_es
) >>
3310 sbi
->s_cluster_bits
);
3312 do_div(resv_clusters
, 50);
3313 resv_clusters
= min_t(ext4_fsblk_t
, resv_clusters
, 4096);
3315 atomic64_set(&sbi
->s_resv_clusters
, resv_clusters
);
3318 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
3320 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
3321 struct buffer_head
*bh
;
3322 struct ext4_super_block
*es
= NULL
;
3323 struct ext4_sb_info
*sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
3325 ext4_fsblk_t sb_block
= get_sb_block(&data
);
3326 ext4_fsblk_t logical_sb_block
;
3327 unsigned long offset
= 0;
3328 unsigned long journal_devnum
= 0;
3329 unsigned long def_mount_opts
;
3333 int blocksize
, clustersize
;
3334 unsigned int db_count
;
3336 int needs_recovery
, has_huge_files
, has_bigalloc
;
3339 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
3340 ext4_group_t first_not_zeroed
;
3342 if ((data
&& !orig_data
) || !sbi
)
3345 sbi
->s_blockgroup_lock
=
3346 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
3347 if (!sbi
->s_blockgroup_lock
)
3350 sb
->s_fs_info
= sbi
;
3352 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
3353 sbi
->s_sb_block
= sb_block
;
3354 if (sb
->s_bdev
->bd_part
)
3355 sbi
->s_sectors_written_start
=
3356 part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]);
3358 /* Cleanup superblock name */
3359 strreplace(sb
->s_id
, '/', '!');
3361 /* -EINVAL is default */
3363 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
3365 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
3370 * The ext4 superblock will not be buffer aligned for other than 1kB
3371 * block sizes. We need to calculate the offset from buffer start.
3373 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
3374 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3375 offset
= do_div(logical_sb_block
, blocksize
);
3377 logical_sb_block
= sb_block
;
3380 if (!(bh
= sb_bread_unmovable(sb
, logical_sb_block
))) {
3381 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
3385 * Note: s_es must be initialized as soon as possible because
3386 * some ext4 macro-instructions depend on its value
3388 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
3390 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
3391 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
3393 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
3395 /* Warn if metadata_csum and gdt_csum are both set. */
3396 if (ext4_has_feature_metadata_csum(sb
) &&
3397 ext4_has_feature_gdt_csum(sb
))
3398 ext4_warning(sb
, "metadata_csum and uninit_bg are "
3399 "redundant flags; please run fsck.");
3401 /* Check for a known checksum algorithm */
3402 if (!ext4_verify_csum_type(sb
, es
)) {
3403 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3404 "unknown checksum algorithm.");
3409 /* Load the checksum driver */
3410 if (ext4_has_feature_metadata_csum(sb
)) {
3411 sbi
->s_chksum_driver
= crypto_alloc_shash("crc32c", 0, 0);
3412 if (IS_ERR(sbi
->s_chksum_driver
)) {
3413 ext4_msg(sb
, KERN_ERR
, "Cannot load crc32c driver.");
3414 ret
= PTR_ERR(sbi
->s_chksum_driver
);
3415 sbi
->s_chksum_driver
= NULL
;
3420 /* Check superblock checksum */
3421 if (!ext4_superblock_csum_verify(sb
, es
)) {
3422 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3423 "invalid superblock checksum. Run e2fsck?");
3429 /* Precompute checksum seed for all metadata */
3430 if (ext4_has_feature_csum_seed(sb
))
3431 sbi
->s_csum_seed
= le32_to_cpu(es
->s_checksum_seed
);
3432 else if (ext4_has_metadata_csum(sb
))
3433 sbi
->s_csum_seed
= ext4_chksum(sbi
, ~0, es
->s_uuid
,
3434 sizeof(es
->s_uuid
));
3436 /* Set defaults before we parse the mount options */
3437 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
3438 set_opt(sb
, INIT_INODE_TABLE
);
3439 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
3441 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
)
3443 if (def_mount_opts
& EXT4_DEFM_UID16
)
3444 set_opt(sb
, NO_UID32
);
3445 /* xattr user namespace & acls are now defaulted on */
3446 set_opt(sb
, XATTR_USER
);
3447 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3448 set_opt(sb
, POSIX_ACL
);
3450 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3451 if (ext4_has_metadata_csum(sb
))
3452 set_opt(sb
, JOURNAL_CHECKSUM
);
3454 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
3455 set_opt(sb
, JOURNAL_DATA
);
3456 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
3457 set_opt(sb
, ORDERED_DATA
);
3458 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
3459 set_opt(sb
, WRITEBACK_DATA
);
3461 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
3462 set_opt(sb
, ERRORS_PANIC
);
3463 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
3464 set_opt(sb
, ERRORS_CONT
);
3466 set_opt(sb
, ERRORS_RO
);
3467 /* block_validity enabled by default; disable with noblock_validity */
3468 set_opt(sb
, BLOCK_VALIDITY
);
3469 if (def_mount_opts
& EXT4_DEFM_DISCARD
)
3470 set_opt(sb
, DISCARD
);
3472 sbi
->s_resuid
= make_kuid(&init_user_ns
, le16_to_cpu(es
->s_def_resuid
));
3473 sbi
->s_resgid
= make_kgid(&init_user_ns
, le16_to_cpu(es
->s_def_resgid
));
3474 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
3475 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
3476 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
3478 if ((def_mount_opts
& EXT4_DEFM_NOBARRIER
) == 0)
3479 set_opt(sb
, BARRIER
);
3482 * enable delayed allocation by default
3483 * Use -o nodelalloc to turn it off
3485 if (!IS_EXT3_SB(sb
) && !IS_EXT2_SB(sb
) &&
3486 ((def_mount_opts
& EXT4_DEFM_NODELALLOC
) == 0))
3487 set_opt(sb
, DELALLOC
);
3490 * set default s_li_wait_mult for lazyinit, for the case there is
3491 * no mount option specified.
3493 sbi
->s_li_wait_mult
= EXT4_DEF_LI_WAIT_MULT
;
3495 if (sbi
->s_es
->s_mount_opts
[0]) {
3496 char *s_mount_opts
= kstrndup(sbi
->s_es
->s_mount_opts
,
3497 sizeof(sbi
->s_es
->s_mount_opts
),
3501 if (!parse_options(s_mount_opts
, sb
, &journal_devnum
,
3502 &journal_ioprio
, 0)) {
3503 ext4_msg(sb
, KERN_WARNING
,
3504 "failed to parse options in superblock: %s",
3507 kfree(s_mount_opts
);
3509 sbi
->s_def_mount_opt
= sbi
->s_mount_opt
;
3510 if (!parse_options((char *) data
, sb
, &journal_devnum
,
3511 &journal_ioprio
, 0))
3514 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
3515 printk_once(KERN_WARNING
"EXT4-fs: Warning: mounting "
3516 "with data=journal disables delayed "
3517 "allocation and O_DIRECT support!\n");
3518 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
3519 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3520 "both data=journal and delalloc");
3523 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
3524 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3525 "both data=journal and dioread_nolock");
3528 if (test_opt(sb
, DAX
)) {
3529 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3530 "both data=journal and dax");
3533 if (ext4_has_feature_encrypt(sb
)) {
3534 ext4_msg(sb
, KERN_WARNING
,
3535 "encrypted files will use data=ordered "
3536 "instead of data journaling mode");
3538 if (test_opt(sb
, DELALLOC
))
3539 clear_opt(sb
, DELALLOC
);
3541 sb
->s_iflags
|= SB_I_CGROUPWB
;
3544 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3545 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
3547 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
3548 (ext4_has_compat_features(sb
) ||
3549 ext4_has_ro_compat_features(sb
) ||
3550 ext4_has_incompat_features(sb
)))
3551 ext4_msg(sb
, KERN_WARNING
,
3552 "feature flags set on rev 0 fs, "
3553 "running e2fsck is recommended");
3555 if (es
->s_creator_os
== cpu_to_le32(EXT4_OS_HURD
)) {
3556 set_opt2(sb
, HURD_COMPAT
);
3557 if (ext4_has_feature_64bit(sb
)) {
3558 ext4_msg(sb
, KERN_ERR
,
3559 "The Hurd can't support 64-bit file systems");
3564 if (IS_EXT2_SB(sb
)) {
3565 if (ext2_feature_set_ok(sb
))
3566 ext4_msg(sb
, KERN_INFO
, "mounting ext2 file system "
3567 "using the ext4 subsystem");
3569 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext2 due "
3570 "to feature incompatibilities");
3575 if (IS_EXT3_SB(sb
)) {
3576 if (ext3_feature_set_ok(sb
))
3577 ext4_msg(sb
, KERN_INFO
, "mounting ext3 file system "
3578 "using the ext4 subsystem");
3580 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext3 due "
3581 "to feature incompatibilities");
3587 * Check feature flags regardless of the revision level, since we
3588 * previously didn't change the revision level when setting the flags,
3589 * so there is a chance incompat flags are set on a rev 0 filesystem.
3591 if (!ext4_feature_set_ok(sb
, (sb
->s_flags
& MS_RDONLY
)))
3594 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
3595 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
3596 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
3597 ext4_msg(sb
, KERN_ERR
,
3598 "Unsupported filesystem blocksize %d (%d log_block_size)",
3599 blocksize
, le32_to_cpu(es
->s_log_block_size
));
3602 if (le32_to_cpu(es
->s_log_block_size
) >
3603 (EXT4_MAX_BLOCK_LOG_SIZE
- EXT4_MIN_BLOCK_LOG_SIZE
)) {
3604 ext4_msg(sb
, KERN_ERR
,
3605 "Invalid log block size: %u",
3606 le32_to_cpu(es
->s_log_block_size
));
3610 if (le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
) > (blocksize
/ 4)) {
3611 ext4_msg(sb
, KERN_ERR
,
3612 "Number of reserved GDT blocks insanely large: %d",
3613 le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
));
3617 if (sbi
->s_mount_opt
& EXT4_MOUNT_DAX
) {
3618 err
= bdev_dax_supported(sb
, blocksize
);
3623 if (ext4_has_feature_encrypt(sb
) && es
->s_encryption_level
) {
3624 ext4_msg(sb
, KERN_ERR
, "Unsupported encryption level %d",
3625 es
->s_encryption_level
);
3629 if (sb
->s_blocksize
!= blocksize
) {
3630 /* Validate the filesystem blocksize */
3631 if (!sb_set_blocksize(sb
, blocksize
)) {
3632 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
3638 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3639 offset
= do_div(logical_sb_block
, blocksize
);
3640 bh
= sb_bread_unmovable(sb
, logical_sb_block
);
3642 ext4_msg(sb
, KERN_ERR
,
3643 "Can't read superblock on 2nd try");
3646 es
= (struct ext4_super_block
*)(bh
->b_data
+ offset
);
3648 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
3649 ext4_msg(sb
, KERN_ERR
,
3650 "Magic mismatch, very weird!");
3655 has_huge_files
= ext4_has_feature_huge_file(sb
);
3656 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
3658 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
3660 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
3661 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
3662 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
3664 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
3665 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
3666 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
3667 (!is_power_of_2(sbi
->s_inode_size
)) ||
3668 (sbi
->s_inode_size
> blocksize
)) {
3669 ext4_msg(sb
, KERN_ERR
,
3670 "unsupported inode size: %d",
3674 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
3675 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
3678 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
3679 if (ext4_has_feature_64bit(sb
)) {
3680 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
3681 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
3682 !is_power_of_2(sbi
->s_desc_size
)) {
3683 ext4_msg(sb
, KERN_ERR
,
3684 "unsupported descriptor size %lu",
3689 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
3691 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
3692 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
3694 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
3695 if (sbi
->s_inodes_per_block
== 0)
3697 if (sbi
->s_inodes_per_group
< sbi
->s_inodes_per_block
||
3698 sbi
->s_inodes_per_group
> blocksize
* 8) {
3699 ext4_msg(sb
, KERN_ERR
, "invalid inodes per group: %lu\n",
3700 sbi
->s_blocks_per_group
);
3703 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
3704 sbi
->s_inodes_per_block
;
3705 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
3707 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3708 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
3709 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
3711 for (i
= 0; i
< 4; i
++)
3712 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
3713 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
3714 if (ext4_has_feature_dir_index(sb
)) {
3715 i
= le32_to_cpu(es
->s_flags
);
3716 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
3717 sbi
->s_hash_unsigned
= 3;
3718 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
3719 #ifdef __CHAR_UNSIGNED__
3720 if (!(sb
->s_flags
& MS_RDONLY
))
3722 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
3723 sbi
->s_hash_unsigned
= 3;
3725 if (!(sb
->s_flags
& MS_RDONLY
))
3727 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
3732 /* Handle clustersize */
3733 clustersize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_cluster_size
);
3734 has_bigalloc
= ext4_has_feature_bigalloc(sb
);
3736 if (clustersize
< blocksize
) {
3737 ext4_msg(sb
, KERN_ERR
,
3738 "cluster size (%d) smaller than "
3739 "block size (%d)", clustersize
, blocksize
);
3742 if (le32_to_cpu(es
->s_log_cluster_size
) >
3743 (EXT4_MAX_CLUSTER_LOG_SIZE
- EXT4_MIN_BLOCK_LOG_SIZE
)) {
3744 ext4_msg(sb
, KERN_ERR
,
3745 "Invalid log cluster size: %u",
3746 le32_to_cpu(es
->s_log_cluster_size
));
3749 sbi
->s_cluster_bits
= le32_to_cpu(es
->s_log_cluster_size
) -
3750 le32_to_cpu(es
->s_log_block_size
);
3751 sbi
->s_clusters_per_group
=
3752 le32_to_cpu(es
->s_clusters_per_group
);
3753 if (sbi
->s_clusters_per_group
> blocksize
* 8) {
3754 ext4_msg(sb
, KERN_ERR
,
3755 "#clusters per group too big: %lu",
3756 sbi
->s_clusters_per_group
);
3759 if (sbi
->s_blocks_per_group
!=
3760 (sbi
->s_clusters_per_group
* (clustersize
/ blocksize
))) {
3761 ext4_msg(sb
, KERN_ERR
, "blocks per group (%lu) and "
3762 "clusters per group (%lu) inconsistent",
3763 sbi
->s_blocks_per_group
,
3764 sbi
->s_clusters_per_group
);
3768 if (clustersize
!= blocksize
) {
3769 ext4_warning(sb
, "fragment/cluster size (%d) != "
3770 "block size (%d)", clustersize
,
3772 clustersize
= blocksize
;
3774 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
3775 ext4_msg(sb
, KERN_ERR
,
3776 "#blocks per group too big: %lu",
3777 sbi
->s_blocks_per_group
);
3780 sbi
->s_clusters_per_group
= sbi
->s_blocks_per_group
;
3781 sbi
->s_cluster_bits
= 0;
3783 sbi
->s_cluster_ratio
= clustersize
/ blocksize
;
3785 /* Do we have standard group size of clustersize * 8 blocks ? */
3786 if (sbi
->s_blocks_per_group
== clustersize
<< 3)
3787 set_opt2(sb
, STD_GROUP_SIZE
);
3790 * Test whether we have more sectors than will fit in sector_t,
3791 * and whether the max offset is addressable by the page cache.
3793 err
= generic_check_addressable(sb
->s_blocksize_bits
,
3794 ext4_blocks_count(es
));
3796 ext4_msg(sb
, KERN_ERR
, "filesystem"
3797 " too large to mount safely on this system");
3798 if (sizeof(sector_t
) < 8)
3799 ext4_msg(sb
, KERN_WARNING
, "CONFIG_LBDAF not enabled");
3803 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
3806 /* check blocks count against device size */
3807 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
3808 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
3809 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
3810 "exceeds size of device (%llu blocks)",
3811 ext4_blocks_count(es
), blocks_count
);
3816 * It makes no sense for the first data block to be beyond the end
3817 * of the filesystem.
3819 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
3820 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data "
3821 "block %u is beyond end of filesystem (%llu)",
3822 le32_to_cpu(es
->s_first_data_block
),
3823 ext4_blocks_count(es
));
3826 blocks_count
= (ext4_blocks_count(es
) -
3827 le32_to_cpu(es
->s_first_data_block
) +
3828 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
3829 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
3830 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
3831 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %u "
3832 "(block count %llu, first data block %u, "
3833 "blocks per group %lu)", sbi
->s_groups_count
,
3834 ext4_blocks_count(es
),
3835 le32_to_cpu(es
->s_first_data_block
),
3836 EXT4_BLOCKS_PER_GROUP(sb
));
3839 sbi
->s_groups_count
= blocks_count
;
3840 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
3841 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
3842 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
3843 EXT4_DESC_PER_BLOCK(sb
);
3844 if (ext4_has_feature_meta_bg(sb
)) {
3845 if (le32_to_cpu(es
->s_first_meta_bg
) >= db_count
) {
3846 ext4_msg(sb
, KERN_WARNING
,
3847 "first meta block group too large: %u "
3848 "(group descriptor block count %u)",
3849 le32_to_cpu(es
->s_first_meta_bg
), db_count
);
3853 sbi
->s_group_desc
= ext4_kvmalloc(db_count
*
3854 sizeof(struct buffer_head
*),
3856 if (sbi
->s_group_desc
== NULL
) {
3857 ext4_msg(sb
, KERN_ERR
, "not enough memory");
3862 bgl_lock_init(sbi
->s_blockgroup_lock
);
3864 for (i
= 0; i
< db_count
; i
++) {
3865 block
= descriptor_loc(sb
, logical_sb_block
, i
);
3866 sbi
->s_group_desc
[i
] = sb_bread_unmovable(sb
, block
);
3867 if (!sbi
->s_group_desc
[i
]) {
3868 ext4_msg(sb
, KERN_ERR
,
3869 "can't read group descriptor %d", i
);
3874 if (!ext4_check_descriptors(sb
, logical_sb_block
, &first_not_zeroed
)) {
3875 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
3876 ret
= -EFSCORRUPTED
;
3880 sbi
->s_gdb_count
= db_count
;
3881 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
3882 spin_lock_init(&sbi
->s_next_gen_lock
);
3884 setup_timer(&sbi
->s_err_report
, print_daily_error_info
,
3885 (unsigned long) sb
);
3887 /* Register extent status tree shrinker */
3888 if (ext4_es_register_shrinker(sbi
))
3891 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
3892 sbi
->s_extent_max_zeroout_kb
= 32;
3895 * set up enough so that it can read an inode
3897 sb
->s_op
= &ext4_sops
;
3898 sb
->s_export_op
= &ext4_export_ops
;
3899 sb
->s_xattr
= ext4_xattr_handlers
;
3900 sb
->s_cop
= &ext4_cryptops
;
3902 sb
->dq_op
= &ext4_quota_operations
;
3903 if (ext4_has_feature_quota(sb
))
3904 sb
->s_qcop
= &dquot_quotactl_sysfile_ops
;
3906 sb
->s_qcop
= &ext4_qctl_operations
;
3907 sb
->s_quota_types
= QTYPE_MASK_USR
| QTYPE_MASK_GRP
| QTYPE_MASK_PRJ
;
3909 memcpy(sb
->s_uuid
, es
->s_uuid
, sizeof(es
->s_uuid
));
3911 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
3912 mutex_init(&sbi
->s_orphan_lock
);
3916 needs_recovery
= (es
->s_last_orphan
!= 0 ||
3917 ext4_has_feature_journal_needs_recovery(sb
));
3919 if (ext4_has_feature_mmp(sb
) && !(sb
->s_flags
& MS_RDONLY
))
3920 if (ext4_multi_mount_protect(sb
, le64_to_cpu(es
->s_mmp_block
)))
3921 goto failed_mount3a
;
3924 * The first inode we look at is the journal inode. Don't try
3925 * root first: it may be modified in the journal!
3927 if (!test_opt(sb
, NOLOAD
) && ext4_has_feature_journal(sb
)) {
3928 if (ext4_load_journal(sb
, es
, journal_devnum
))
3929 goto failed_mount3a
;
3930 } else if (test_opt(sb
, NOLOAD
) && !(sb
->s_flags
& MS_RDONLY
) &&
3931 ext4_has_feature_journal_needs_recovery(sb
)) {
3932 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
3933 "suppressed and not mounted read-only");
3934 goto failed_mount_wq
;
3936 /* Nojournal mode, all journal mount options are illegal */
3937 if (test_opt2(sb
, EXPLICIT_JOURNAL_CHECKSUM
)) {
3938 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3939 "journal_checksum, fs mounted w/o journal");
3940 goto failed_mount_wq
;
3942 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
3943 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3944 "journal_async_commit, fs mounted w/o journal");
3945 goto failed_mount_wq
;
3947 if (sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
) {
3948 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3949 "commit=%lu, fs mounted w/o journal",
3950 sbi
->s_commit_interval
/ HZ
);
3951 goto failed_mount_wq
;
3953 if (EXT4_MOUNT_DATA_FLAGS
&
3954 (sbi
->s_mount_opt
^ sbi
->s_def_mount_opt
)) {
3955 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3956 "data=, fs mounted w/o journal");
3957 goto failed_mount_wq
;
3959 sbi
->s_def_mount_opt
&= EXT4_MOUNT_JOURNAL_CHECKSUM
;
3960 clear_opt(sb
, JOURNAL_CHECKSUM
);
3961 clear_opt(sb
, DATA_FLAGS
);
3962 sbi
->s_journal
= NULL
;
3967 if (ext4_has_feature_64bit(sb
) &&
3968 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
3969 JBD2_FEATURE_INCOMPAT_64BIT
)) {
3970 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
3971 goto failed_mount_wq
;
3974 if (!set_journal_csum_feature_set(sb
)) {
3975 ext4_msg(sb
, KERN_ERR
, "Failed to set journal checksum "
3977 goto failed_mount_wq
;
3980 /* We have now updated the journal if required, so we can
3981 * validate the data journaling mode. */
3982 switch (test_opt(sb
, DATA_FLAGS
)) {
3984 /* No mode set, assume a default based on the journal
3985 * capabilities: ORDERED_DATA if the journal can
3986 * cope, else JOURNAL_DATA
3988 if (jbd2_journal_check_available_features
3989 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
3990 set_opt(sb
, ORDERED_DATA
);
3992 set_opt(sb
, JOURNAL_DATA
);
3995 case EXT4_MOUNT_ORDERED_DATA
:
3996 case EXT4_MOUNT_WRITEBACK_DATA
:
3997 if (!jbd2_journal_check_available_features
3998 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
3999 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
4000 "requested data journaling mode");
4001 goto failed_mount_wq
;
4007 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
&&
4008 test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
4009 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4010 "journal_async_commit in data=ordered mode");
4011 goto failed_mount_wq
;
4014 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
4016 sbi
->s_journal
->j_commit_callback
= ext4_journal_commit_callback
;
4019 sbi
->s_mb_cache
= ext4_xattr_create_cache();
4020 if (!sbi
->s_mb_cache
) {
4021 ext4_msg(sb
, KERN_ERR
, "Failed to create an mb_cache");
4022 goto failed_mount_wq
;
4025 if ((DUMMY_ENCRYPTION_ENABLED(sbi
) || ext4_has_feature_encrypt(sb
)) &&
4026 (blocksize
!= PAGE_SIZE
)) {
4027 ext4_msg(sb
, KERN_ERR
,
4028 "Unsupported blocksize for fs encryption");
4029 goto failed_mount_wq
;
4032 if (DUMMY_ENCRYPTION_ENABLED(sbi
) && !(sb
->s_flags
& MS_RDONLY
) &&
4033 !ext4_has_feature_encrypt(sb
)) {
4034 ext4_set_feature_encrypt(sb
);
4035 ext4_commit_super(sb
, 1);
4039 * Get the # of file system overhead blocks from the
4040 * superblock if present.
4042 if (es
->s_overhead_clusters
)
4043 sbi
->s_overhead
= le32_to_cpu(es
->s_overhead_clusters
);
4045 err
= ext4_calculate_overhead(sb
);
4047 goto failed_mount_wq
;
4051 * The maximum number of concurrent works can be high and
4052 * concurrency isn't really necessary. Limit it to 1.
4054 EXT4_SB(sb
)->rsv_conversion_wq
=
4055 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM
| WQ_UNBOUND
, 1);
4056 if (!EXT4_SB(sb
)->rsv_conversion_wq
) {
4057 printk(KERN_ERR
"EXT4-fs: failed to create workqueue\n");
4063 * The jbd2_journal_load will have done any necessary log recovery,
4064 * so we can safely mount the rest of the filesystem now.
4067 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
4069 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
4070 ret
= PTR_ERR(root
);
4074 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
4075 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
4079 sb
->s_root
= d_make_root(root
);
4081 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
4086 if (ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
))
4087 sb
->s_flags
|= MS_RDONLY
;
4089 /* determine the minimum size of new large inodes, if present */
4090 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
4091 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
4092 EXT4_GOOD_OLD_INODE_SIZE
;
4093 if (ext4_has_feature_extra_isize(sb
)) {
4094 if (sbi
->s_want_extra_isize
<
4095 le16_to_cpu(es
->s_want_extra_isize
))
4096 sbi
->s_want_extra_isize
=
4097 le16_to_cpu(es
->s_want_extra_isize
);
4098 if (sbi
->s_want_extra_isize
<
4099 le16_to_cpu(es
->s_min_extra_isize
))
4100 sbi
->s_want_extra_isize
=
4101 le16_to_cpu(es
->s_min_extra_isize
);
4104 /* Check if enough inode space is available */
4105 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
4106 sbi
->s_inode_size
) {
4107 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
4108 EXT4_GOOD_OLD_INODE_SIZE
;
4109 ext4_msg(sb
, KERN_INFO
, "required extra inode space not"
4113 ext4_set_resv_clusters(sb
);
4115 err
= ext4_setup_system_zone(sb
);
4117 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
4119 goto failed_mount4a
;
4123 err
= ext4_mb_init(sb
);
4125 ext4_msg(sb
, KERN_ERR
, "failed to initialize mballoc (%d)",
4130 block
= ext4_count_free_clusters(sb
);
4131 ext4_free_blocks_count_set(sbi
->s_es
,
4132 EXT4_C2B(sbi
, block
));
4133 err
= percpu_counter_init(&sbi
->s_freeclusters_counter
, block
,
4136 unsigned long freei
= ext4_count_free_inodes(sb
);
4137 sbi
->s_es
->s_free_inodes_count
= cpu_to_le32(freei
);
4138 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
, freei
,
4142 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
4143 ext4_count_dirs(sb
), GFP_KERNEL
);
4145 err
= percpu_counter_init(&sbi
->s_dirtyclusters_counter
, 0,
4148 err
= percpu_init_rwsem(&sbi
->s_journal_flag_rwsem
);
4151 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
4155 if (ext4_has_feature_flex_bg(sb
))
4156 if (!ext4_fill_flex_info(sb
)) {
4157 ext4_msg(sb
, KERN_ERR
,
4158 "unable to initialize "
4159 "flex_bg meta info!");
4163 err
= ext4_register_li_request(sb
, first_not_zeroed
);
4167 err
= ext4_register_sysfs(sb
);
4172 /* Enable quota usage during mount. */
4173 if (ext4_has_feature_quota(sb
) && !(sb
->s_flags
& MS_RDONLY
)) {
4174 err
= ext4_enable_quotas(sb
);
4178 #endif /* CONFIG_QUOTA */
4180 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
4181 ext4_orphan_cleanup(sb
, es
);
4182 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
4183 if (needs_recovery
) {
4184 ext4_msg(sb
, KERN_INFO
, "recovery complete");
4185 ext4_mark_recovery_complete(sb
, es
);
4187 if (EXT4_SB(sb
)->s_journal
) {
4188 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
4189 descr
= " journalled data mode";
4190 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
4191 descr
= " ordered data mode";
4193 descr
= " writeback data mode";
4195 descr
= "out journal";
4197 if (test_opt(sb
, DISCARD
)) {
4198 struct request_queue
*q
= bdev_get_queue(sb
->s_bdev
);
4199 if (!blk_queue_discard(q
))
4200 ext4_msg(sb
, KERN_WARNING
,
4201 "mounting with \"discard\" option, but "
4202 "the device does not support discard");
4205 if (___ratelimit(&ext4_mount_msg_ratelimit
, "EXT4-fs mount"))
4206 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s. "
4207 "Opts: %.*s%s%s", descr
,
4208 (int) sizeof(sbi
->s_es
->s_mount_opts
),
4209 sbi
->s_es
->s_mount_opts
,
4210 *sbi
->s_es
->s_mount_opts
? "; " : "", orig_data
);
4212 if (es
->s_error_count
)
4213 mod_timer(&sbi
->s_err_report
, jiffies
+ 300*HZ
); /* 5 minutes */
4215 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4216 ratelimit_state_init(&sbi
->s_err_ratelimit_state
, 5 * HZ
, 10);
4217 ratelimit_state_init(&sbi
->s_warning_ratelimit_state
, 5 * HZ
, 10);
4218 ratelimit_state_init(&sbi
->s_msg_ratelimit_state
, 5 * HZ
, 10);
4221 #ifdef CONFIG_EXT4_FS_ENCRYPTION
4222 memcpy(sbi
->key_prefix
, EXT4_KEY_DESC_PREFIX
,
4223 EXT4_KEY_DESC_PREFIX_SIZE
);
4224 sbi
->key_prefix_size
= EXT4_KEY_DESC_PREFIX_SIZE
;
4230 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
4235 ext4_unregister_sysfs(sb
);
4238 ext4_unregister_li_request(sb
);
4240 ext4_mb_release(sb
);
4241 if (sbi
->s_flex_groups
)
4242 kvfree(sbi
->s_flex_groups
);
4243 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
4244 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
4245 percpu_counter_destroy(&sbi
->s_dirs_counter
);
4246 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
4248 ext4_ext_release(sb
);
4249 ext4_release_system_zone(sb
);
4254 ext4_msg(sb
, KERN_ERR
, "mount failed");
4255 if (EXT4_SB(sb
)->rsv_conversion_wq
)
4256 destroy_workqueue(EXT4_SB(sb
)->rsv_conversion_wq
);
4258 if (sbi
->s_mb_cache
) {
4259 ext4_xattr_destroy_cache(sbi
->s_mb_cache
);
4260 sbi
->s_mb_cache
= NULL
;
4262 if (sbi
->s_journal
) {
4263 jbd2_journal_destroy(sbi
->s_journal
);
4264 sbi
->s_journal
= NULL
;
4267 ext4_es_unregister_shrinker(sbi
);
4269 del_timer_sync(&sbi
->s_err_report
);
4271 kthread_stop(sbi
->s_mmp_tsk
);
4273 for (i
= 0; i
< db_count
; i
++)
4274 brelse(sbi
->s_group_desc
[i
]);
4275 kvfree(sbi
->s_group_desc
);
4277 if (sbi
->s_chksum_driver
)
4278 crypto_free_shash(sbi
->s_chksum_driver
);
4280 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
4281 kfree(sbi
->s_qf_names
[i
]);
4283 ext4_blkdev_remove(sbi
);
4286 sb
->s_fs_info
= NULL
;
4287 kfree(sbi
->s_blockgroup_lock
);
4291 return err
? err
: ret
;
4295 * Setup any per-fs journal parameters now. We'll do this both on
4296 * initial mount, once the journal has been initialised but before we've
4297 * done any recovery; and again on any subsequent remount.
4299 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
4301 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4303 journal
->j_commit_interval
= sbi
->s_commit_interval
;
4304 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
4305 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
4307 write_lock(&journal
->j_state_lock
);
4308 if (test_opt(sb
, BARRIER
))
4309 journal
->j_flags
|= JBD2_BARRIER
;
4311 journal
->j_flags
&= ~JBD2_BARRIER
;
4312 if (test_opt(sb
, DATA_ERR_ABORT
))
4313 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
4315 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
4316 write_unlock(&journal
->j_state_lock
);
4319 static struct inode
*ext4_get_journal_inode(struct super_block
*sb
,
4320 unsigned int journal_inum
)
4322 struct inode
*journal_inode
;
4325 * Test for the existence of a valid inode on disk. Bad things
4326 * happen if we iget() an unused inode, as the subsequent iput()
4327 * will try to delete it.
4329 journal_inode
= ext4_iget(sb
, journal_inum
);
4330 if (IS_ERR(journal_inode
)) {
4331 ext4_msg(sb
, KERN_ERR
, "no journal found");
4334 if (!journal_inode
->i_nlink
) {
4335 make_bad_inode(journal_inode
);
4336 iput(journal_inode
);
4337 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
4341 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4342 journal_inode
, journal_inode
->i_size
);
4343 if (!S_ISREG(journal_inode
->i_mode
)) {
4344 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
4345 iput(journal_inode
);
4348 return journal_inode
;
4351 static journal_t
*ext4_get_journal(struct super_block
*sb
,
4352 unsigned int journal_inum
)
4354 struct inode
*journal_inode
;
4357 BUG_ON(!ext4_has_feature_journal(sb
));
4359 journal_inode
= ext4_get_journal_inode(sb
, journal_inum
);
4363 journal
= jbd2_journal_init_inode(journal_inode
);
4365 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
4366 iput(journal_inode
);
4369 journal
->j_private
= sb
;
4370 ext4_init_journal_params(sb
, journal
);
4374 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
4377 struct buffer_head
*bh
;
4381 int hblock
, blocksize
;
4382 ext4_fsblk_t sb_block
;
4383 unsigned long offset
;
4384 struct ext4_super_block
*es
;
4385 struct block_device
*bdev
;
4387 BUG_ON(!ext4_has_feature_journal(sb
));
4389 bdev
= ext4_blkdev_get(j_dev
, sb
);
4393 blocksize
= sb
->s_blocksize
;
4394 hblock
= bdev_logical_block_size(bdev
);
4395 if (blocksize
< hblock
) {
4396 ext4_msg(sb
, KERN_ERR
,
4397 "blocksize too small for journal device");
4401 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
4402 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
4403 set_blocksize(bdev
, blocksize
);
4404 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
4405 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
4406 "external journal");
4410 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
4411 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
4412 !(le32_to_cpu(es
->s_feature_incompat
) &
4413 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
4414 ext4_msg(sb
, KERN_ERR
, "external journal has "
4420 if ((le32_to_cpu(es
->s_feature_ro_compat
) &
4421 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
) &&
4422 es
->s_checksum
!= ext4_superblock_csum(sb
, es
)) {
4423 ext4_msg(sb
, KERN_ERR
, "external journal has "
4424 "corrupt superblock");
4429 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
4430 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
4435 len
= ext4_blocks_count(es
);
4436 start
= sb_block
+ 1;
4437 brelse(bh
); /* we're done with the superblock */
4439 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
4440 start
, len
, blocksize
);
4442 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
4445 journal
->j_private
= sb
;
4446 ll_rw_block(REQ_OP_READ
, REQ_META
| REQ_PRIO
, 1, &journal
->j_sb_buffer
);
4447 wait_on_buffer(journal
->j_sb_buffer
);
4448 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
4449 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
4452 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
4453 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
4454 "user (unsupported) - %d",
4455 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
4458 EXT4_SB(sb
)->journal_bdev
= bdev
;
4459 ext4_init_journal_params(sb
, journal
);
4463 jbd2_journal_destroy(journal
);
4465 ext4_blkdev_put(bdev
);
4469 static int ext4_load_journal(struct super_block
*sb
,
4470 struct ext4_super_block
*es
,
4471 unsigned long journal_devnum
)
4474 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
4477 int really_read_only
;
4479 BUG_ON(!ext4_has_feature_journal(sb
));
4481 if (journal_devnum
&&
4482 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4483 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
4484 "numbers have changed");
4485 journal_dev
= new_decode_dev(journal_devnum
);
4487 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
4489 really_read_only
= bdev_read_only(sb
->s_bdev
);
4492 * Are we loading a blank journal or performing recovery after a
4493 * crash? For recovery, we need to check in advance whether we
4494 * can get read-write access to the device.
4496 if (ext4_has_feature_journal_needs_recovery(sb
)) {
4497 if (sb
->s_flags
& MS_RDONLY
) {
4498 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
4499 "required on readonly filesystem");
4500 if (really_read_only
) {
4501 ext4_msg(sb
, KERN_ERR
, "write access "
4502 "unavailable, cannot proceed");
4505 ext4_msg(sb
, KERN_INFO
, "write access will "
4506 "be enabled during recovery");
4510 if (journal_inum
&& journal_dev
) {
4511 ext4_msg(sb
, KERN_ERR
, "filesystem has both journal "
4512 "and inode journals!");
4517 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
4520 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
4524 if (!(journal
->j_flags
& JBD2_BARRIER
))
4525 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
4527 if (!ext4_has_feature_journal_needs_recovery(sb
))
4528 err
= jbd2_journal_wipe(journal
, !really_read_only
);
4530 char *save
= kmalloc(EXT4_S_ERR_LEN
, GFP_KERNEL
);
4532 memcpy(save
, ((char *) es
) +
4533 EXT4_S_ERR_START
, EXT4_S_ERR_LEN
);
4534 err
= jbd2_journal_load(journal
);
4536 memcpy(((char *) es
) + EXT4_S_ERR_START
,
4537 save
, EXT4_S_ERR_LEN
);
4542 ext4_msg(sb
, KERN_ERR
, "error loading journal");
4543 jbd2_journal_destroy(journal
);
4547 EXT4_SB(sb
)->s_journal
= journal
;
4548 ext4_clear_journal_err(sb
, es
);
4550 if (!really_read_only
&& journal_devnum
&&
4551 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4552 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
4554 /* Make sure we flush the recovery flag to disk. */
4555 ext4_commit_super(sb
, 1);
4561 static int ext4_commit_super(struct super_block
*sb
, int sync
)
4563 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
4564 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
4567 if (!sbh
|| block_device_ejected(sb
))
4570 * If the file system is mounted read-only, don't update the
4571 * superblock write time. This avoids updating the superblock
4572 * write time when we are mounting the root file system
4573 * read/only but we need to replay the journal; at that point,
4574 * for people who are east of GMT and who make their clock
4575 * tick in localtime for Windows bug-for-bug compatibility,
4576 * the clock is set in the future, and this will cause e2fsck
4577 * to complain and force a full file system check.
4579 if (!(sb
->s_flags
& MS_RDONLY
))
4580 es
->s_wtime
= cpu_to_le32(get_seconds());
4581 if (sb
->s_bdev
->bd_part
)
4582 es
->s_kbytes_written
=
4583 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
+
4584 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
4585 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
4587 es
->s_kbytes_written
=
4588 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
);
4589 if (percpu_counter_initialized(&EXT4_SB(sb
)->s_freeclusters_counter
))
4590 ext4_free_blocks_count_set(es
,
4591 EXT4_C2B(EXT4_SB(sb
), percpu_counter_sum_positive(
4592 &EXT4_SB(sb
)->s_freeclusters_counter
)));
4593 if (percpu_counter_initialized(&EXT4_SB(sb
)->s_freeinodes_counter
))
4594 es
->s_free_inodes_count
=
4595 cpu_to_le32(percpu_counter_sum_positive(
4596 &EXT4_SB(sb
)->s_freeinodes_counter
));
4597 BUFFER_TRACE(sbh
, "marking dirty");
4598 ext4_superblock_csum_set(sb
);
4601 if (buffer_write_io_error(sbh
)) {
4603 * Oh, dear. A previous attempt to write the
4604 * superblock failed. This could happen because the
4605 * USB device was yanked out. Or it could happen to
4606 * be a transient write error and maybe the block will
4607 * be remapped. Nothing we can do but to retry the
4608 * write and hope for the best.
4610 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
4611 "superblock detected");
4612 clear_buffer_write_io_error(sbh
);
4613 set_buffer_uptodate(sbh
);
4615 mark_buffer_dirty(sbh
);
4618 error
= __sync_dirty_buffer(sbh
,
4619 test_opt(sb
, BARRIER
) ? REQ_FUA
: REQ_SYNC
);
4623 error
= buffer_write_io_error(sbh
);
4625 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
4627 clear_buffer_write_io_error(sbh
);
4628 set_buffer_uptodate(sbh
);
4635 * Have we just finished recovery? If so, and if we are mounting (or
4636 * remounting) the filesystem readonly, then we will end up with a
4637 * consistent fs on disk. Record that fact.
4639 static void ext4_mark_recovery_complete(struct super_block
*sb
,
4640 struct ext4_super_block
*es
)
4642 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
4644 if (!ext4_has_feature_journal(sb
)) {
4645 BUG_ON(journal
!= NULL
);
4648 jbd2_journal_lock_updates(journal
);
4649 if (jbd2_journal_flush(journal
) < 0)
4652 if (ext4_has_feature_journal_needs_recovery(sb
) &&
4653 sb
->s_flags
& MS_RDONLY
) {
4654 ext4_clear_feature_journal_needs_recovery(sb
);
4655 ext4_commit_super(sb
, 1);
4659 jbd2_journal_unlock_updates(journal
);
4663 * If we are mounting (or read-write remounting) a filesystem whose journal
4664 * has recorded an error from a previous lifetime, move that error to the
4665 * main filesystem now.
4667 static void ext4_clear_journal_err(struct super_block
*sb
,
4668 struct ext4_super_block
*es
)
4674 BUG_ON(!ext4_has_feature_journal(sb
));
4676 journal
= EXT4_SB(sb
)->s_journal
;
4679 * Now check for any error status which may have been recorded in the
4680 * journal by a prior ext4_error() or ext4_abort()
4683 j_errno
= jbd2_journal_errno(journal
);
4687 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
4688 ext4_warning(sb
, "Filesystem error recorded "
4689 "from previous mount: %s", errstr
);
4690 ext4_warning(sb
, "Marking fs in need of filesystem check.");
4692 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
4693 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
4694 ext4_commit_super(sb
, 1);
4696 jbd2_journal_clear_err(journal
);
4697 jbd2_journal_update_sb_errno(journal
);
4702 * Force the running and committing transactions to commit,
4703 * and wait on the commit.
4705 int ext4_force_commit(struct super_block
*sb
)
4709 if (sb
->s_flags
& MS_RDONLY
)
4712 journal
= EXT4_SB(sb
)->s_journal
;
4713 return ext4_journal_force_commit(journal
);
4716 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
4720 bool needs_barrier
= false;
4721 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4723 trace_ext4_sync_fs(sb
, wait
);
4724 flush_workqueue(sbi
->rsv_conversion_wq
);
4726 * Writeback quota in non-journalled quota case - journalled quota has
4729 dquot_writeback_dquots(sb
, -1);
4731 * Data writeback is possible w/o journal transaction, so barrier must
4732 * being sent at the end of the function. But we can skip it if
4733 * transaction_commit will do it for us.
4735 if (sbi
->s_journal
) {
4736 target
= jbd2_get_latest_transaction(sbi
->s_journal
);
4737 if (wait
&& sbi
->s_journal
->j_flags
& JBD2_BARRIER
&&
4738 !jbd2_trans_will_send_data_barrier(sbi
->s_journal
, target
))
4739 needs_barrier
= true;
4741 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
4743 ret
= jbd2_log_wait_commit(sbi
->s_journal
,
4746 } else if (wait
&& test_opt(sb
, BARRIER
))
4747 needs_barrier
= true;
4748 if (needs_barrier
) {
4750 err
= blkdev_issue_flush(sb
->s_bdev
, GFP_KERNEL
, NULL
);
4759 * LVM calls this function before a (read-only) snapshot is created. This
4760 * gives us a chance to flush the journal completely and mark the fs clean.
4762 * Note that only this function cannot bring a filesystem to be in a clean
4763 * state independently. It relies on upper layer to stop all data & metadata
4766 static int ext4_freeze(struct super_block
*sb
)
4771 if (sb
->s_flags
& MS_RDONLY
)
4774 journal
= EXT4_SB(sb
)->s_journal
;
4777 /* Now we set up the journal barrier. */
4778 jbd2_journal_lock_updates(journal
);
4781 * Don't clear the needs_recovery flag if we failed to
4782 * flush the journal.
4784 error
= jbd2_journal_flush(journal
);
4788 /* Journal blocked and flushed, clear needs_recovery flag. */
4789 ext4_clear_feature_journal_needs_recovery(sb
);
4792 error
= ext4_commit_super(sb
, 1);
4795 /* we rely on upper layer to stop further updates */
4796 jbd2_journal_unlock_updates(journal
);
4801 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4802 * flag here, even though the filesystem is not technically dirty yet.
4804 static int ext4_unfreeze(struct super_block
*sb
)
4806 if (sb
->s_flags
& MS_RDONLY
)
4809 if (EXT4_SB(sb
)->s_journal
) {
4810 /* Reset the needs_recovery flag before the fs is unlocked. */
4811 ext4_set_feature_journal_needs_recovery(sb
);
4814 ext4_commit_super(sb
, 1);
4819 * Structure to save mount options for ext4_remount's benefit
4821 struct ext4_mount_options
{
4822 unsigned long s_mount_opt
;
4823 unsigned long s_mount_opt2
;
4826 unsigned long s_commit_interval
;
4827 u32 s_min_batch_time
, s_max_batch_time
;
4830 char *s_qf_names
[EXT4_MAXQUOTAS
];
4834 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
4836 struct ext4_super_block
*es
;
4837 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4838 unsigned long old_sb_flags
;
4839 struct ext4_mount_options old_opts
;
4840 int enable_quota
= 0;
4842 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
4847 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
4849 /* Store the original options */
4850 old_sb_flags
= sb
->s_flags
;
4851 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
4852 old_opts
.s_mount_opt2
= sbi
->s_mount_opt2
;
4853 old_opts
.s_resuid
= sbi
->s_resuid
;
4854 old_opts
.s_resgid
= sbi
->s_resgid
;
4855 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
4856 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
4857 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
4859 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
4860 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
4861 if (sbi
->s_qf_names
[i
]) {
4862 old_opts
.s_qf_names
[i
] = kstrdup(sbi
->s_qf_names
[i
],
4864 if (!old_opts
.s_qf_names
[i
]) {
4865 for (j
= 0; j
< i
; j
++)
4866 kfree(old_opts
.s_qf_names
[j
]);
4871 old_opts
.s_qf_names
[i
] = NULL
;
4873 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
4874 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
4876 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
, 1)) {
4881 if ((old_opts
.s_mount_opt
& EXT4_MOUNT_JOURNAL_CHECKSUM
) ^
4882 test_opt(sb
, JOURNAL_CHECKSUM
)) {
4883 ext4_msg(sb
, KERN_ERR
, "changing journal_checksum "
4884 "during remount not supported; ignoring");
4885 sbi
->s_mount_opt
^= EXT4_MOUNT_JOURNAL_CHECKSUM
;
4888 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
4889 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
4890 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4891 "both data=journal and delalloc");
4895 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
4896 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4897 "both data=journal and dioread_nolock");
4901 if (test_opt(sb
, DAX
)) {
4902 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4903 "both data=journal and dax");
4907 } else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
) {
4908 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
4909 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4910 "journal_async_commit in data=ordered mode");
4916 if ((sbi
->s_mount_opt
^ old_opts
.s_mount_opt
) & EXT4_MOUNT_DAX
) {
4917 ext4_msg(sb
, KERN_WARNING
, "warning: refusing change of "
4918 "dax flag with busy inodes while remounting");
4919 sbi
->s_mount_opt
^= EXT4_MOUNT_DAX
;
4922 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
)
4923 ext4_abort(sb
, "Abort forced by user");
4925 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
4926 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
4930 if (sbi
->s_journal
) {
4931 ext4_init_journal_params(sb
, sbi
->s_journal
);
4932 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
4935 if (*flags
& MS_LAZYTIME
)
4936 sb
->s_flags
|= MS_LAZYTIME
;
4938 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
)) {
4939 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
) {
4944 if (*flags
& MS_RDONLY
) {
4945 err
= sync_filesystem(sb
);
4948 err
= dquot_suspend(sb
, -1);
4953 * First of all, the unconditional stuff we have to do
4954 * to disable replay of the journal when we next remount
4956 sb
->s_flags
|= MS_RDONLY
;
4959 * OK, test if we are remounting a valid rw partition
4960 * readonly, and if so set the rdonly flag and then
4961 * mark the partition as valid again.
4963 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
4964 (sbi
->s_mount_state
& EXT4_VALID_FS
))
4965 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
4968 ext4_mark_recovery_complete(sb
, es
);
4970 /* Make sure we can mount this feature set readwrite */
4971 if (ext4_has_feature_readonly(sb
) ||
4972 !ext4_feature_set_ok(sb
, 0)) {
4977 * Make sure the group descriptor checksums
4978 * are sane. If they aren't, refuse to remount r/w.
4980 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
4981 struct ext4_group_desc
*gdp
=
4982 ext4_get_group_desc(sb
, g
, NULL
);
4984 if (!ext4_group_desc_csum_verify(sb
, g
, gdp
)) {
4985 ext4_msg(sb
, KERN_ERR
,
4986 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4987 g
, le16_to_cpu(ext4_group_desc_csum(sb
, g
, gdp
)),
4988 le16_to_cpu(gdp
->bg_checksum
));
4995 * If we have an unprocessed orphan list hanging
4996 * around from a previously readonly bdev mount,
4997 * require a full umount/remount for now.
4999 if (es
->s_last_orphan
) {
5000 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
5001 "remount RDWR because of unprocessed "
5002 "orphan inode list. Please "
5003 "umount/remount instead");
5009 * Mounting a RDONLY partition read-write, so reread
5010 * and store the current valid flag. (It may have
5011 * been changed by e2fsck since we originally mounted
5015 ext4_clear_journal_err(sb
, es
);
5016 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
5017 if (!ext4_setup_super(sb
, es
, 0))
5018 sb
->s_flags
&= ~MS_RDONLY
;
5019 if (ext4_has_feature_mmp(sb
))
5020 if (ext4_multi_mount_protect(sb
,
5021 le64_to_cpu(es
->s_mmp_block
))) {
5030 * Reinitialize lazy itable initialization thread based on
5033 if ((sb
->s_flags
& MS_RDONLY
) || !test_opt(sb
, INIT_INODE_TABLE
))
5034 ext4_unregister_li_request(sb
);
5036 ext4_group_t first_not_zeroed
;
5037 first_not_zeroed
= ext4_has_uninit_itable(sb
);
5038 ext4_register_li_request(sb
, first_not_zeroed
);
5041 ext4_setup_system_zone(sb
);
5042 if (sbi
->s_journal
== NULL
&& !(old_sb_flags
& MS_RDONLY
))
5043 ext4_commit_super(sb
, 1);
5046 /* Release old quota file names */
5047 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
5048 kfree(old_opts
.s_qf_names
[i
]);
5050 if (sb_any_quota_suspended(sb
))
5051 dquot_resume(sb
, -1);
5052 else if (ext4_has_feature_quota(sb
)) {
5053 err
= ext4_enable_quotas(sb
);
5060 *flags
= (*flags
& ~MS_LAZYTIME
) | (sb
->s_flags
& MS_LAZYTIME
);
5061 ext4_msg(sb
, KERN_INFO
, "re-mounted. Opts: %s", orig_data
);
5066 sb
->s_flags
= old_sb_flags
;
5067 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
5068 sbi
->s_mount_opt2
= old_opts
.s_mount_opt2
;
5069 sbi
->s_resuid
= old_opts
.s_resuid
;
5070 sbi
->s_resgid
= old_opts
.s_resgid
;
5071 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
5072 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
5073 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
5075 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
5076 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
5077 kfree(sbi
->s_qf_names
[i
]);
5078 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
5086 static int ext4_statfs_project(struct super_block
*sb
,
5087 kprojid_t projid
, struct kstatfs
*buf
)
5090 struct dquot
*dquot
;
5094 qid
= make_kqid_projid(projid
);
5095 dquot
= dqget(sb
, qid
);
5097 return PTR_ERR(dquot
);
5098 spin_lock(&dq_data_lock
);
5100 limit
= (dquot
->dq_dqb
.dqb_bsoftlimit
?
5101 dquot
->dq_dqb
.dqb_bsoftlimit
:
5102 dquot
->dq_dqb
.dqb_bhardlimit
) >> sb
->s_blocksize_bits
;
5103 if (limit
&& buf
->f_blocks
> limit
) {
5104 curblock
= dquot
->dq_dqb
.dqb_curspace
>> sb
->s_blocksize_bits
;
5105 buf
->f_blocks
= limit
;
5106 buf
->f_bfree
= buf
->f_bavail
=
5107 (buf
->f_blocks
> curblock
) ?
5108 (buf
->f_blocks
- curblock
) : 0;
5111 limit
= dquot
->dq_dqb
.dqb_isoftlimit
?
5112 dquot
->dq_dqb
.dqb_isoftlimit
:
5113 dquot
->dq_dqb
.dqb_ihardlimit
;
5114 if (limit
&& buf
->f_files
> limit
) {
5115 buf
->f_files
= limit
;
5117 (buf
->f_files
> dquot
->dq_dqb
.dqb_curinodes
) ?
5118 (buf
->f_files
- dquot
->dq_dqb
.dqb_curinodes
) : 0;
5121 spin_unlock(&dq_data_lock
);
5127 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
5129 struct super_block
*sb
= dentry
->d_sb
;
5130 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
5131 struct ext4_super_block
*es
= sbi
->s_es
;
5132 ext4_fsblk_t overhead
= 0, resv_blocks
;
5135 resv_blocks
= EXT4_C2B(sbi
, atomic64_read(&sbi
->s_resv_clusters
));
5137 if (!test_opt(sb
, MINIX_DF
))
5138 overhead
= sbi
->s_overhead
;
5140 buf
->f_type
= EXT4_SUPER_MAGIC
;
5141 buf
->f_bsize
= sb
->s_blocksize
;
5142 buf
->f_blocks
= ext4_blocks_count(es
) - EXT4_C2B(sbi
, overhead
);
5143 bfree
= percpu_counter_sum_positive(&sbi
->s_freeclusters_counter
) -
5144 percpu_counter_sum_positive(&sbi
->s_dirtyclusters_counter
);
5145 /* prevent underflow in case that few free space is available */
5146 buf
->f_bfree
= EXT4_C2B(sbi
, max_t(s64
, bfree
, 0));
5147 buf
->f_bavail
= buf
->f_bfree
-
5148 (ext4_r_blocks_count(es
) + resv_blocks
);
5149 if (buf
->f_bfree
< (ext4_r_blocks_count(es
) + resv_blocks
))
5151 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
5152 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
5153 buf
->f_namelen
= EXT4_NAME_LEN
;
5154 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
5155 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
5156 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
5157 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
5160 if (ext4_test_inode_flag(dentry
->d_inode
, EXT4_INODE_PROJINHERIT
) &&
5161 sb_has_quota_limits_enabled(sb
, PRJQUOTA
))
5162 ext4_statfs_project(sb
, EXT4_I(dentry
->d_inode
)->i_projid
, buf
);
5167 /* Helper function for writing quotas on sync - we need to start transaction
5168 * before quota file is locked for write. Otherwise the are possible deadlocks:
5169 * Process 1 Process 2
5170 * ext4_create() quota_sync()
5171 * jbd2_journal_start() write_dquot()
5172 * dquot_initialize() down(dqio_mutex)
5173 * down(dqio_mutex) jbd2_journal_start()
5179 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
5181 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_id
.type
];
5184 static int ext4_write_dquot(struct dquot
*dquot
)
5188 struct inode
*inode
;
5190 inode
= dquot_to_inode(dquot
);
5191 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
,
5192 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
5194 return PTR_ERR(handle
);
5195 ret
= dquot_commit(dquot
);
5196 err
= ext4_journal_stop(handle
);
5202 static int ext4_acquire_dquot(struct dquot
*dquot
)
5207 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
5208 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
5210 return PTR_ERR(handle
);
5211 ret
= dquot_acquire(dquot
);
5212 err
= ext4_journal_stop(handle
);
5218 static int ext4_release_dquot(struct dquot
*dquot
)
5223 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
5224 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
5225 if (IS_ERR(handle
)) {
5226 /* Release dquot anyway to avoid endless cycle in dqput() */
5227 dquot_release(dquot
);
5228 return PTR_ERR(handle
);
5230 ret
= dquot_release(dquot
);
5231 err
= ext4_journal_stop(handle
);
5237 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
5239 struct super_block
*sb
= dquot
->dq_sb
;
5240 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
5242 /* Are we journaling quotas? */
5243 if (ext4_has_feature_quota(sb
) ||
5244 sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
5245 dquot_mark_dquot_dirty(dquot
);
5246 return ext4_write_dquot(dquot
);
5248 return dquot_mark_dquot_dirty(dquot
);
5252 static int ext4_write_info(struct super_block
*sb
, int type
)
5257 /* Data block + inode block */
5258 handle
= ext4_journal_start(d_inode(sb
->s_root
), EXT4_HT_QUOTA
, 2);
5260 return PTR_ERR(handle
);
5261 ret
= dquot_commit_info(sb
, type
);
5262 err
= ext4_journal_stop(handle
);
5269 * Turn on quotas during mount time - we need to find
5270 * the quota file and such...
5272 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
5274 return dquot_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
5275 EXT4_SB(sb
)->s_jquota_fmt
, type
);
5278 static void lockdep_set_quota_inode(struct inode
*inode
, int subclass
)
5280 struct ext4_inode_info
*ei
= EXT4_I(inode
);
5282 /* The first argument of lockdep_set_subclass has to be
5283 * *exactly* the same as the argument to init_rwsem() --- in
5284 * this case, in init_once() --- or lockdep gets unhappy
5285 * because the name of the lock is set using the
5286 * stringification of the argument to init_rwsem().
5288 (void) ei
; /* shut up clang warning if !CONFIG_LOCKDEP */
5289 lockdep_set_subclass(&ei
->i_data_sem
, subclass
);
5293 * Standard function to be called on quota_on
5295 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
5296 const struct path
*path
)
5300 if (!test_opt(sb
, QUOTA
))
5303 /* Quotafile not on the same filesystem? */
5304 if (path
->dentry
->d_sb
!= sb
)
5306 /* Journaling quota? */
5307 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
5308 /* Quotafile not in fs root? */
5309 if (path
->dentry
->d_parent
!= sb
->s_root
)
5310 ext4_msg(sb
, KERN_WARNING
,
5311 "Quota file not on filesystem root. "
5312 "Journaled quota will not work");
5316 * When we journal data on quota file, we have to flush journal to see
5317 * all updates to the file when we bypass pagecache...
5319 if (EXT4_SB(sb
)->s_journal
&&
5320 ext4_should_journal_data(d_inode(path
->dentry
))) {
5322 * We don't need to lock updates but journal_flush() could
5323 * otherwise be livelocked...
5325 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
5326 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
5327 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
5331 lockdep_set_quota_inode(path
->dentry
->d_inode
, I_DATA_SEM_QUOTA
);
5332 err
= dquot_quota_on(sb
, type
, format_id
, path
);
5334 lockdep_set_quota_inode(path
->dentry
->d_inode
,
5339 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
5343 struct inode
*qf_inode
;
5344 unsigned long qf_inums
[EXT4_MAXQUOTAS
] = {
5345 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
5346 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
),
5347 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_prj_quota_inum
)
5350 BUG_ON(!ext4_has_feature_quota(sb
));
5352 if (!qf_inums
[type
])
5355 qf_inode
= ext4_iget(sb
, qf_inums
[type
]);
5356 if (IS_ERR(qf_inode
)) {
5357 ext4_error(sb
, "Bad quota inode # %lu", qf_inums
[type
]);
5358 return PTR_ERR(qf_inode
);
5361 /* Don't account quota for quota files to avoid recursion */
5362 qf_inode
->i_flags
|= S_NOQUOTA
;
5363 lockdep_set_quota_inode(qf_inode
, I_DATA_SEM_QUOTA
);
5364 err
= dquot_enable(qf_inode
, type
, format_id
, flags
);
5367 lockdep_set_quota_inode(qf_inode
, I_DATA_SEM_NORMAL
);
5372 /* Enable usage tracking for all quota types. */
5373 static int ext4_enable_quotas(struct super_block
*sb
)
5376 unsigned long qf_inums
[EXT4_MAXQUOTAS
] = {
5377 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
5378 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
),
5379 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_prj_quota_inum
)
5381 bool quota_mopt
[EXT4_MAXQUOTAS
] = {
5382 test_opt(sb
, USRQUOTA
),
5383 test_opt(sb
, GRPQUOTA
),
5384 test_opt(sb
, PRJQUOTA
),
5387 sb_dqopt(sb
)->flags
|= DQUOT_QUOTA_SYS_FILE
;
5388 for (type
= 0; type
< EXT4_MAXQUOTAS
; type
++) {
5389 if (qf_inums
[type
]) {
5390 err
= ext4_quota_enable(sb
, type
, QFMT_VFS_V1
,
5391 DQUOT_USAGE_ENABLED
|
5392 (quota_mopt
[type
] ? DQUOT_LIMITS_ENABLED
: 0));
5395 "Failed to enable quota tracking "
5396 "(type=%d, err=%d). Please run "
5397 "e2fsck to fix.", type
, err
);
5405 static int ext4_quota_off(struct super_block
*sb
, int type
)
5407 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5410 /* Force all delayed allocation blocks to be allocated.
5411 * Caller already holds s_umount sem */
5412 if (test_opt(sb
, DELALLOC
))
5413 sync_filesystem(sb
);
5418 /* Update modification times of quota files when userspace can
5419 * start looking at them */
5420 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
, 1);
5423 inode
->i_mtime
= inode
->i_ctime
= current_time(inode
);
5424 ext4_mark_inode_dirty(handle
, inode
);
5425 ext4_journal_stop(handle
);
5428 return dquot_quota_off(sb
, type
);
5431 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5432 * acquiring the locks... As quota files are never truncated and quota code
5433 * itself serializes the operations (and no one else should touch the files)
5434 * we don't have to be afraid of races */
5435 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
5436 size_t len
, loff_t off
)
5438 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5439 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5440 int offset
= off
& (sb
->s_blocksize
- 1);
5443 struct buffer_head
*bh
;
5444 loff_t i_size
= i_size_read(inode
);
5448 if (off
+len
> i_size
)
5451 while (toread
> 0) {
5452 tocopy
= sb
->s_blocksize
- offset
< toread
?
5453 sb
->s_blocksize
- offset
: toread
;
5454 bh
= ext4_bread(NULL
, inode
, blk
, 0);
5457 if (!bh
) /* A hole? */
5458 memset(data
, 0, tocopy
);
5460 memcpy(data
, bh
->b_data
+offset
, tocopy
);
5470 /* Write to quotafile (we know the transaction is already started and has
5471 * enough credits) */
5472 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
5473 const char *data
, size_t len
, loff_t off
)
5475 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5476 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5477 int err
, offset
= off
& (sb
->s_blocksize
- 1);
5479 struct buffer_head
*bh
;
5480 handle_t
*handle
= journal_current_handle();
5482 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
5483 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5484 " cancelled because transaction is not started",
5485 (unsigned long long)off
, (unsigned long long)len
);
5489 * Since we account only one data block in transaction credits,
5490 * then it is impossible to cross a block boundary.
5492 if (sb
->s_blocksize
- offset
< len
) {
5493 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5494 " cancelled because not block aligned",
5495 (unsigned long long)off
, (unsigned long long)len
);
5500 bh
= ext4_bread(handle
, inode
, blk
,
5501 EXT4_GET_BLOCKS_CREATE
|
5502 EXT4_GET_BLOCKS_METADATA_NOFAIL
);
5503 } while (IS_ERR(bh
) && (PTR_ERR(bh
) == -ENOSPC
) &&
5504 ext4_should_retry_alloc(inode
->i_sb
, &retries
));
5509 BUFFER_TRACE(bh
, "get write access");
5510 err
= ext4_journal_get_write_access(handle
, bh
);
5516 memcpy(bh
->b_data
+offset
, data
, len
);
5517 flush_dcache_page(bh
->b_page
);
5519 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
5522 if (inode
->i_size
< off
+ len
) {
5523 i_size_write(inode
, off
+ len
);
5524 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
5525 ext4_mark_inode_dirty(handle
, inode
);
5530 static int ext4_get_next_id(struct super_block
*sb
, struct kqid
*qid
)
5532 const struct quota_format_ops
*ops
;
5534 if (!sb_has_quota_loaded(sb
, qid
->type
))
5536 ops
= sb_dqopt(sb
)->ops
[qid
->type
];
5537 if (!ops
|| !ops
->get_next_id
)
5539 return dquot_get_next_id(sb
, qid
);
5543 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
5544 const char *dev_name
, void *data
)
5546 return mount_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
);
5549 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5550 static inline void register_as_ext2(void)
5552 int err
= register_filesystem(&ext2_fs_type
);
5555 "EXT4-fs: Unable to register as ext2 (%d)\n", err
);
5558 static inline void unregister_as_ext2(void)
5560 unregister_filesystem(&ext2_fs_type
);
5563 static inline int ext2_feature_set_ok(struct super_block
*sb
)
5565 if (ext4_has_unknown_ext2_incompat_features(sb
))
5567 if (sb
->s_flags
& MS_RDONLY
)
5569 if (ext4_has_unknown_ext2_ro_compat_features(sb
))
5574 static inline void register_as_ext2(void) { }
5575 static inline void unregister_as_ext2(void) { }
5576 static inline int ext2_feature_set_ok(struct super_block
*sb
) { return 0; }
5579 static inline void register_as_ext3(void)
5581 int err
= register_filesystem(&ext3_fs_type
);
5584 "EXT4-fs: Unable to register as ext3 (%d)\n", err
);
5587 static inline void unregister_as_ext3(void)
5589 unregister_filesystem(&ext3_fs_type
);
5592 static inline int ext3_feature_set_ok(struct super_block
*sb
)
5594 if (ext4_has_unknown_ext3_incompat_features(sb
))
5596 if (!ext4_has_feature_journal(sb
))
5598 if (sb
->s_flags
& MS_RDONLY
)
5600 if (ext4_has_unknown_ext3_ro_compat_features(sb
))
5605 static struct file_system_type ext4_fs_type
= {
5606 .owner
= THIS_MODULE
,
5608 .mount
= ext4_mount
,
5609 .kill_sb
= kill_block_super
,
5610 .fs_flags
= FS_REQUIRES_DEV
,
5612 MODULE_ALIAS_FS("ext4");
5614 /* Shared across all ext4 file systems */
5615 wait_queue_head_t ext4__ioend_wq
[EXT4_WQ_HASH_SZ
];
5617 static int __init
ext4_init_fs(void)
5621 ratelimit_state_init(&ext4_mount_msg_ratelimit
, 30 * HZ
, 64);
5622 ext4_li_info
= NULL
;
5623 mutex_init(&ext4_li_mtx
);
5625 /* Build-time check for flags consistency */
5626 ext4_check_flag_values();
5628 for (i
= 0; i
< EXT4_WQ_HASH_SZ
; i
++)
5629 init_waitqueue_head(&ext4__ioend_wq
[i
]);
5631 err
= ext4_init_es();
5635 err
= ext4_init_pageio();
5639 err
= ext4_init_system_zone();
5643 err
= ext4_init_sysfs();
5647 err
= ext4_init_mballoc();
5650 err
= init_inodecache();
5655 err
= register_filesystem(&ext4_fs_type
);
5661 unregister_as_ext2();
5662 unregister_as_ext3();
5663 destroy_inodecache();
5665 ext4_exit_mballoc();
5669 ext4_exit_system_zone();
5678 static void __exit
ext4_exit_fs(void)
5680 ext4_destroy_lazyinit_thread();
5681 unregister_as_ext2();
5682 unregister_as_ext3();
5683 unregister_filesystem(&ext4_fs_type
);
5684 destroy_inodecache();
5685 ext4_exit_mballoc();
5687 ext4_exit_system_zone();
5692 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5693 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5694 MODULE_LICENSE("GPL");
5695 module_init(ext4_init_fs
)
5696 module_exit(ext4_exit_fs
)