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);
85 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
86 * i_mmap_rwsem (inode->i_mmap_rwsem)!
89 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
90 * page lock -> i_data_sem (rw)
92 * buffered write path:
93 * sb_start_write -> i_mutex -> mmap_sem
94 * sb_start_write -> i_mutex -> transaction start -> page lock ->
98 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
99 * i_mmap_rwsem (w) -> page lock
100 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
101 * transaction start -> i_data_sem (rw)
104 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
105 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
106 * transaction start -> i_data_sem (rw)
109 * transaction start -> page lock(s) -> i_data_sem (rw)
112 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
113 static struct file_system_type ext2_fs_type
= {
114 .owner
= THIS_MODULE
,
117 .kill_sb
= kill_block_super
,
118 .fs_flags
= FS_REQUIRES_DEV
,
120 MODULE_ALIAS_FS("ext2");
121 MODULE_ALIAS("ext2");
122 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
124 #define IS_EXT2_SB(sb) (0)
128 static struct file_system_type ext3_fs_type
= {
129 .owner
= THIS_MODULE
,
132 .kill_sb
= kill_block_super
,
133 .fs_flags
= FS_REQUIRES_DEV
,
135 MODULE_ALIAS_FS("ext3");
136 MODULE_ALIAS("ext3");
137 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
139 static int ext4_verify_csum_type(struct super_block
*sb
,
140 struct ext4_super_block
*es
)
142 if (!ext4_has_feature_metadata_csum(sb
))
145 return es
->s_checksum_type
== EXT4_CRC32C_CHKSUM
;
148 static __le32
ext4_superblock_csum(struct super_block
*sb
,
149 struct ext4_super_block
*es
)
151 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
152 int offset
= offsetof(struct ext4_super_block
, s_checksum
);
155 csum
= ext4_chksum(sbi
, ~0, (char *)es
, offset
);
157 return cpu_to_le32(csum
);
160 static int ext4_superblock_csum_verify(struct super_block
*sb
,
161 struct ext4_super_block
*es
)
163 if (!ext4_has_metadata_csum(sb
))
166 return es
->s_checksum
== ext4_superblock_csum(sb
, es
);
169 void ext4_superblock_csum_set(struct super_block
*sb
)
171 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
173 if (!ext4_has_metadata_csum(sb
))
176 es
->s_checksum
= ext4_superblock_csum(sb
, es
);
179 void *ext4_kvmalloc(size_t size
, gfp_t flags
)
183 ret
= kmalloc(size
, flags
| __GFP_NOWARN
);
185 ret
= __vmalloc(size
, flags
, PAGE_KERNEL
);
189 void *ext4_kvzalloc(size_t size
, gfp_t flags
)
193 ret
= kzalloc(size
, flags
| __GFP_NOWARN
);
195 ret
= __vmalloc(size
, flags
| __GFP_ZERO
, PAGE_KERNEL
);
199 ext4_fsblk_t
ext4_block_bitmap(struct super_block
*sb
,
200 struct ext4_group_desc
*bg
)
202 return le32_to_cpu(bg
->bg_block_bitmap_lo
) |
203 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
204 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_block_bitmap_hi
) << 32 : 0);
207 ext4_fsblk_t
ext4_inode_bitmap(struct super_block
*sb
,
208 struct ext4_group_desc
*bg
)
210 return le32_to_cpu(bg
->bg_inode_bitmap_lo
) |
211 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
212 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_bitmap_hi
) << 32 : 0);
215 ext4_fsblk_t
ext4_inode_table(struct super_block
*sb
,
216 struct ext4_group_desc
*bg
)
218 return le32_to_cpu(bg
->bg_inode_table_lo
) |
219 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
220 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_table_hi
) << 32 : 0);
223 __u32
ext4_free_group_clusters(struct super_block
*sb
,
224 struct ext4_group_desc
*bg
)
226 return le16_to_cpu(bg
->bg_free_blocks_count_lo
) |
227 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
228 (__u32
)le16_to_cpu(bg
->bg_free_blocks_count_hi
) << 16 : 0);
231 __u32
ext4_free_inodes_count(struct super_block
*sb
,
232 struct ext4_group_desc
*bg
)
234 return le16_to_cpu(bg
->bg_free_inodes_count_lo
) |
235 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
236 (__u32
)le16_to_cpu(bg
->bg_free_inodes_count_hi
) << 16 : 0);
239 __u32
ext4_used_dirs_count(struct super_block
*sb
,
240 struct ext4_group_desc
*bg
)
242 return le16_to_cpu(bg
->bg_used_dirs_count_lo
) |
243 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
244 (__u32
)le16_to_cpu(bg
->bg_used_dirs_count_hi
) << 16 : 0);
247 __u32
ext4_itable_unused_count(struct super_block
*sb
,
248 struct ext4_group_desc
*bg
)
250 return le16_to_cpu(bg
->bg_itable_unused_lo
) |
251 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
252 (__u32
)le16_to_cpu(bg
->bg_itable_unused_hi
) << 16 : 0);
255 void ext4_block_bitmap_set(struct super_block
*sb
,
256 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
258 bg
->bg_block_bitmap_lo
= cpu_to_le32((u32
)blk
);
259 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
260 bg
->bg_block_bitmap_hi
= cpu_to_le32(blk
>> 32);
263 void ext4_inode_bitmap_set(struct super_block
*sb
,
264 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
266 bg
->bg_inode_bitmap_lo
= cpu_to_le32((u32
)blk
);
267 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
268 bg
->bg_inode_bitmap_hi
= cpu_to_le32(blk
>> 32);
271 void ext4_inode_table_set(struct super_block
*sb
,
272 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
274 bg
->bg_inode_table_lo
= cpu_to_le32((u32
)blk
);
275 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
276 bg
->bg_inode_table_hi
= cpu_to_le32(blk
>> 32);
279 void ext4_free_group_clusters_set(struct super_block
*sb
,
280 struct ext4_group_desc
*bg
, __u32 count
)
282 bg
->bg_free_blocks_count_lo
= cpu_to_le16((__u16
)count
);
283 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
284 bg
->bg_free_blocks_count_hi
= cpu_to_le16(count
>> 16);
287 void ext4_free_inodes_set(struct super_block
*sb
,
288 struct ext4_group_desc
*bg
, __u32 count
)
290 bg
->bg_free_inodes_count_lo
= cpu_to_le16((__u16
)count
);
291 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
292 bg
->bg_free_inodes_count_hi
= cpu_to_le16(count
>> 16);
295 void ext4_used_dirs_set(struct super_block
*sb
,
296 struct ext4_group_desc
*bg
, __u32 count
)
298 bg
->bg_used_dirs_count_lo
= cpu_to_le16((__u16
)count
);
299 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
300 bg
->bg_used_dirs_count_hi
= cpu_to_le16(count
>> 16);
303 void ext4_itable_unused_set(struct super_block
*sb
,
304 struct ext4_group_desc
*bg
, __u32 count
)
306 bg
->bg_itable_unused_lo
= cpu_to_le16((__u16
)count
);
307 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
308 bg
->bg_itable_unused_hi
= cpu_to_le16(count
>> 16);
312 static void __save_error_info(struct super_block
*sb
, const char *func
,
315 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
317 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
318 if (bdev_read_only(sb
->s_bdev
))
320 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
321 es
->s_last_error_time
= cpu_to_le32(get_seconds());
322 strncpy(es
->s_last_error_func
, func
, sizeof(es
->s_last_error_func
));
323 es
->s_last_error_line
= cpu_to_le32(line
);
324 if (!es
->s_first_error_time
) {
325 es
->s_first_error_time
= es
->s_last_error_time
;
326 strncpy(es
->s_first_error_func
, func
,
327 sizeof(es
->s_first_error_func
));
328 es
->s_first_error_line
= cpu_to_le32(line
);
329 es
->s_first_error_ino
= es
->s_last_error_ino
;
330 es
->s_first_error_block
= es
->s_last_error_block
;
333 * Start the daily error reporting function if it hasn't been
336 if (!es
->s_error_count
)
337 mod_timer(&EXT4_SB(sb
)->s_err_report
, jiffies
+ 24*60*60*HZ
);
338 le32_add_cpu(&es
->s_error_count
, 1);
341 static void save_error_info(struct super_block
*sb
, const char *func
,
344 __save_error_info(sb
, func
, line
);
345 ext4_commit_super(sb
, 1);
349 * The del_gendisk() function uninitializes the disk-specific data
350 * structures, including the bdi structure, without telling anyone
351 * else. Once this happens, any attempt to call mark_buffer_dirty()
352 * (for example, by ext4_commit_super), will cause a kernel OOPS.
353 * This is a kludge to prevent these oops until we can put in a proper
354 * hook in del_gendisk() to inform the VFS and file system layers.
356 static int block_device_ejected(struct super_block
*sb
)
358 struct inode
*bd_inode
= sb
->s_bdev
->bd_inode
;
359 struct backing_dev_info
*bdi
= inode_to_bdi(bd_inode
);
361 return bdi
->dev
== NULL
;
364 static void ext4_journal_commit_callback(journal_t
*journal
, transaction_t
*txn
)
366 struct super_block
*sb
= journal
->j_private
;
367 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
368 int error
= is_journal_aborted(journal
);
369 struct ext4_journal_cb_entry
*jce
;
371 BUG_ON(txn
->t_state
== T_FINISHED
);
372 spin_lock(&sbi
->s_md_lock
);
373 while (!list_empty(&txn
->t_private_list
)) {
374 jce
= list_entry(txn
->t_private_list
.next
,
375 struct ext4_journal_cb_entry
, jce_list
);
376 list_del_init(&jce
->jce_list
);
377 spin_unlock(&sbi
->s_md_lock
);
378 jce
->jce_func(sb
, jce
, error
);
379 spin_lock(&sbi
->s_md_lock
);
381 spin_unlock(&sbi
->s_md_lock
);
384 /* Deal with the reporting of failure conditions on a filesystem such as
385 * inconsistencies detected or read IO failures.
387 * On ext2, we can store the error state of the filesystem in the
388 * superblock. That is not possible on ext4, because we may have other
389 * write ordering constraints on the superblock which prevent us from
390 * writing it out straight away; and given that the journal is about to
391 * be aborted, we can't rely on the current, or future, transactions to
392 * write out the superblock safely.
394 * We'll just use the jbd2_journal_abort() error code to record an error in
395 * the journal instead. On recovery, the journal will complain about
396 * that error until we've noted it down and cleared it.
399 static void ext4_handle_error(struct super_block
*sb
)
401 if (sb
->s_flags
& MS_RDONLY
)
404 if (!test_opt(sb
, ERRORS_CONT
)) {
405 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
407 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
409 jbd2_journal_abort(journal
, -EIO
);
411 if (test_opt(sb
, ERRORS_RO
)) {
412 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
414 * Make sure updated value of ->s_mount_flags will be visible
415 * before ->s_flags update
418 sb
->s_flags
|= MS_RDONLY
;
420 if (test_opt(sb
, ERRORS_PANIC
)) {
421 if (EXT4_SB(sb
)->s_journal
&&
422 !(EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_REC_ERR
))
424 panic("EXT4-fs (device %s): panic forced after error\n",
429 #define ext4_error_ratelimit(sb) \
430 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
433 void __ext4_error(struct super_block
*sb
, const char *function
,
434 unsigned int line
, const char *fmt
, ...)
436 struct va_format vaf
;
439 if (ext4_error_ratelimit(sb
)) {
444 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
445 sb
->s_id
, function
, line
, current
->comm
, &vaf
);
448 save_error_info(sb
, function
, line
);
449 ext4_handle_error(sb
);
452 void __ext4_error_inode(struct inode
*inode
, const char *function
,
453 unsigned int line
, ext4_fsblk_t block
,
454 const char *fmt
, ...)
457 struct va_format vaf
;
458 struct ext4_super_block
*es
= EXT4_SB(inode
->i_sb
)->s_es
;
460 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
461 es
->s_last_error_block
= cpu_to_le64(block
);
462 if (ext4_error_ratelimit(inode
->i_sb
)) {
467 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
468 "inode #%lu: block %llu: comm %s: %pV\n",
469 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
470 block
, current
->comm
, &vaf
);
472 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
473 "inode #%lu: comm %s: %pV\n",
474 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
475 current
->comm
, &vaf
);
478 save_error_info(inode
->i_sb
, function
, line
);
479 ext4_handle_error(inode
->i_sb
);
482 void __ext4_error_file(struct file
*file
, const char *function
,
483 unsigned int line
, ext4_fsblk_t block
,
484 const char *fmt
, ...)
487 struct va_format vaf
;
488 struct ext4_super_block
*es
;
489 struct inode
*inode
= file_inode(file
);
490 char pathname
[80], *path
;
492 es
= EXT4_SB(inode
->i_sb
)->s_es
;
493 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
494 if (ext4_error_ratelimit(inode
->i_sb
)) {
495 path
= file_path(file
, pathname
, sizeof(pathname
));
503 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
504 "block %llu: comm %s: path %s: %pV\n",
505 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
506 block
, current
->comm
, path
, &vaf
);
509 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
510 "comm %s: path %s: %pV\n",
511 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
512 current
->comm
, path
, &vaf
);
515 save_error_info(inode
->i_sb
, function
, line
);
516 ext4_handle_error(inode
->i_sb
);
519 const char *ext4_decode_error(struct super_block
*sb
, int errno
,
526 errstr
= "Corrupt filesystem";
529 errstr
= "Filesystem failed CRC";
532 errstr
= "IO failure";
535 errstr
= "Out of memory";
538 if (!sb
|| (EXT4_SB(sb
)->s_journal
&&
539 EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
))
540 errstr
= "Journal has aborted";
542 errstr
= "Readonly filesystem";
545 /* If the caller passed in an extra buffer for unknown
546 * errors, textualise them now. Else we just return
549 /* Check for truncated error codes... */
550 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
559 /* __ext4_std_error decodes expected errors from journaling functions
560 * automatically and invokes the appropriate error response. */
562 void __ext4_std_error(struct super_block
*sb
, const char *function
,
563 unsigned int line
, int errno
)
568 /* Special case: if the error is EROFS, and we're not already
569 * inside a transaction, then there's really no point in logging
571 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
572 (sb
->s_flags
& MS_RDONLY
))
575 if (ext4_error_ratelimit(sb
)) {
576 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
577 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s:%d: %s\n",
578 sb
->s_id
, function
, line
, errstr
);
581 save_error_info(sb
, function
, line
);
582 ext4_handle_error(sb
);
586 * ext4_abort is a much stronger failure handler than ext4_error. The
587 * abort function may be used to deal with unrecoverable failures such
588 * as journal IO errors or ENOMEM at a critical moment in log management.
590 * We unconditionally force the filesystem into an ABORT|READONLY state,
591 * unless the error response on the fs has been set to panic in which
592 * case we take the easy way out and panic immediately.
595 void __ext4_abort(struct super_block
*sb
, const char *function
,
596 unsigned int line
, const char *fmt
, ...)
600 save_error_info(sb
, function
, line
);
602 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: ", sb
->s_id
,
608 if ((sb
->s_flags
& MS_RDONLY
) == 0) {
609 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
610 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
612 * Make sure updated value of ->s_mount_flags will be visible
613 * before ->s_flags update
616 sb
->s_flags
|= MS_RDONLY
;
617 if (EXT4_SB(sb
)->s_journal
)
618 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
619 save_error_info(sb
, function
, line
);
621 if (test_opt(sb
, ERRORS_PANIC
)) {
622 if (EXT4_SB(sb
)->s_journal
&&
623 !(EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_REC_ERR
))
625 panic("EXT4-fs panic from previous error\n");
629 void __ext4_msg(struct super_block
*sb
,
630 const char *prefix
, const char *fmt
, ...)
632 struct va_format vaf
;
635 if (!___ratelimit(&(EXT4_SB(sb
)->s_msg_ratelimit_state
), "EXT4-fs"))
641 printk("%sEXT4-fs (%s): %pV\n", prefix
, sb
->s_id
, &vaf
);
645 #define ext4_warning_ratelimit(sb) \
646 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
649 void __ext4_warning(struct super_block
*sb
, const char *function
,
650 unsigned int line
, const char *fmt
, ...)
652 struct va_format vaf
;
655 if (!ext4_warning_ratelimit(sb
))
661 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: %pV\n",
662 sb
->s_id
, function
, line
, &vaf
);
666 void __ext4_warning_inode(const struct inode
*inode
, const char *function
,
667 unsigned int line
, const char *fmt
, ...)
669 struct va_format vaf
;
672 if (!ext4_warning_ratelimit(inode
->i_sb
))
678 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: "
679 "inode #%lu: comm %s: %pV\n", inode
->i_sb
->s_id
,
680 function
, line
, inode
->i_ino
, current
->comm
, &vaf
);
684 void __ext4_grp_locked_error(const char *function
, unsigned int line
,
685 struct super_block
*sb
, ext4_group_t grp
,
686 unsigned long ino
, ext4_fsblk_t block
,
687 const char *fmt
, ...)
691 struct va_format vaf
;
693 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
695 es
->s_last_error_ino
= cpu_to_le32(ino
);
696 es
->s_last_error_block
= cpu_to_le64(block
);
697 __save_error_info(sb
, function
, line
);
699 if (ext4_error_ratelimit(sb
)) {
703 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: group %u, ",
704 sb
->s_id
, function
, line
, grp
);
706 printk(KERN_CONT
"inode %lu: ", ino
);
708 printk(KERN_CONT
"block %llu:",
709 (unsigned long long) block
);
710 printk(KERN_CONT
"%pV\n", &vaf
);
714 if (test_opt(sb
, ERRORS_CONT
)) {
715 ext4_commit_super(sb
, 0);
719 ext4_unlock_group(sb
, grp
);
720 ext4_handle_error(sb
);
722 * We only get here in the ERRORS_RO case; relocking the group
723 * may be dangerous, but nothing bad will happen since the
724 * filesystem will have already been marked read/only and the
725 * journal has been aborted. We return 1 as a hint to callers
726 * who might what to use the return value from
727 * ext4_grp_locked_error() to distinguish between the
728 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
729 * aggressively from the ext4 function in question, with a
730 * more appropriate error code.
732 ext4_lock_group(sb
, grp
);
736 void ext4_update_dynamic_rev(struct super_block
*sb
)
738 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
740 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
744 "updating to rev %d because of new feature flag, "
745 "running e2fsck is recommended",
748 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
749 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
750 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
751 /* leave es->s_feature_*compat flags alone */
752 /* es->s_uuid will be set by e2fsck if empty */
755 * The rest of the superblock fields should be zero, and if not it
756 * means they are likely already in use, so leave them alone. We
757 * can leave it up to e2fsck to clean up any inconsistencies there.
762 * Open the external journal device
764 static struct block_device
*ext4_blkdev_get(dev_t dev
, struct super_block
*sb
)
766 struct block_device
*bdev
;
767 char b
[BDEVNAME_SIZE
];
769 bdev
= blkdev_get_by_dev(dev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
, sb
);
775 ext4_msg(sb
, KERN_ERR
, "failed to open journal device %s: %ld",
776 __bdevname(dev
, b
), PTR_ERR(bdev
));
781 * Release the journal device
783 static void ext4_blkdev_put(struct block_device
*bdev
)
785 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
);
788 static void ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
790 struct block_device
*bdev
;
791 bdev
= sbi
->journal_bdev
;
793 ext4_blkdev_put(bdev
);
794 sbi
->journal_bdev
= NULL
;
798 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
800 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
803 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
807 ext4_msg(sb
, KERN_ERR
, "sb orphan head is %d",
808 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
810 printk(KERN_ERR
"sb_info orphan list:\n");
811 list_for_each(l
, &sbi
->s_orphan
) {
812 struct inode
*inode
= orphan_list_entry(l
);
814 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
815 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
816 inode
->i_mode
, inode
->i_nlink
,
821 static void ext4_put_super(struct super_block
*sb
)
823 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
824 struct ext4_super_block
*es
= sbi
->s_es
;
827 ext4_unregister_li_request(sb
);
828 dquot_disable(sb
, -1, DQUOT_USAGE_ENABLED
| DQUOT_LIMITS_ENABLED
);
830 flush_workqueue(sbi
->rsv_conversion_wq
);
831 destroy_workqueue(sbi
->rsv_conversion_wq
);
833 if (sbi
->s_journal
) {
834 err
= jbd2_journal_destroy(sbi
->s_journal
);
835 sbi
->s_journal
= NULL
;
837 ext4_abort(sb
, "Couldn't clean up the journal");
840 ext4_unregister_sysfs(sb
);
841 ext4_es_unregister_shrinker(sbi
);
842 del_timer_sync(&sbi
->s_err_report
);
843 ext4_release_system_zone(sb
);
845 ext4_ext_release(sb
);
847 if (!(sb
->s_flags
& MS_RDONLY
)) {
848 ext4_clear_feature_journal_needs_recovery(sb
);
849 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
851 if (!(sb
->s_flags
& MS_RDONLY
))
852 ext4_commit_super(sb
, 1);
854 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
855 brelse(sbi
->s_group_desc
[i
]);
856 kvfree(sbi
->s_group_desc
);
857 kvfree(sbi
->s_flex_groups
);
858 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
859 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
860 percpu_counter_destroy(&sbi
->s_dirs_counter
);
861 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
864 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
865 kfree(sbi
->s_qf_names
[i
]);
868 /* Debugging code just in case the in-memory inode orphan list
869 * isn't empty. The on-disk one can be non-empty if we've
870 * detected an error and taken the fs readonly, but the
871 * in-memory list had better be clean by this point. */
872 if (!list_empty(&sbi
->s_orphan
))
873 dump_orphan_list(sb
, sbi
);
874 J_ASSERT(list_empty(&sbi
->s_orphan
));
876 sync_blockdev(sb
->s_bdev
);
877 invalidate_bdev(sb
->s_bdev
);
878 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
880 * Invalidate the journal device's buffers. We don't want them
881 * floating about in memory - the physical journal device may
882 * hotswapped, and it breaks the `ro-after' testing code.
884 sync_blockdev(sbi
->journal_bdev
);
885 invalidate_bdev(sbi
->journal_bdev
);
886 ext4_blkdev_remove(sbi
);
888 if (sbi
->s_mb_cache
) {
889 ext4_xattr_destroy_cache(sbi
->s_mb_cache
);
890 sbi
->s_mb_cache
= NULL
;
893 kthread_stop(sbi
->s_mmp_tsk
);
894 sb
->s_fs_info
= NULL
;
896 * Now that we are completely done shutting down the
897 * superblock, we need to actually destroy the kobject.
899 kobject_put(&sbi
->s_kobj
);
900 wait_for_completion(&sbi
->s_kobj_unregister
);
901 if (sbi
->s_chksum_driver
)
902 crypto_free_shash(sbi
->s_chksum_driver
);
903 kfree(sbi
->s_blockgroup_lock
);
907 static struct kmem_cache
*ext4_inode_cachep
;
910 * Called inside transaction, so use GFP_NOFS
912 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
914 struct ext4_inode_info
*ei
;
916 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
920 ei
->vfs_inode
.i_version
= 1;
921 spin_lock_init(&ei
->i_raw_lock
);
922 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
923 spin_lock_init(&ei
->i_prealloc_lock
);
924 ext4_es_init_tree(&ei
->i_es_tree
);
925 rwlock_init(&ei
->i_es_lock
);
926 INIT_LIST_HEAD(&ei
->i_es_list
);
929 ei
->i_es_shrink_lblk
= 0;
930 ei
->i_reserved_data_blocks
= 0;
931 ei
->i_reserved_meta_blocks
= 0;
932 ei
->i_allocated_meta_blocks
= 0;
933 ei
->i_da_metadata_calc_len
= 0;
934 ei
->i_da_metadata_calc_last_lblock
= 0;
935 spin_lock_init(&(ei
->i_block_reservation_lock
));
937 ei
->i_reserved_quota
= 0;
938 memset(&ei
->i_dquot
, 0, sizeof(ei
->i_dquot
));
941 INIT_LIST_HEAD(&ei
->i_rsv_conversion_list
);
942 spin_lock_init(&ei
->i_completed_io_lock
);
944 ei
->i_datasync_tid
= 0;
945 atomic_set(&ei
->i_unwritten
, 0);
946 INIT_WORK(&ei
->i_rsv_conversion_work
, ext4_end_io_rsv_work
);
947 #ifdef CONFIG_EXT4_FS_ENCRYPTION
948 ei
->i_crypt_info
= NULL
;
950 return &ei
->vfs_inode
;
953 static int ext4_drop_inode(struct inode
*inode
)
955 int drop
= generic_drop_inode(inode
);
957 trace_ext4_drop_inode(inode
, drop
);
961 static void ext4_i_callback(struct rcu_head
*head
)
963 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
964 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
967 static void ext4_destroy_inode(struct inode
*inode
)
969 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
970 ext4_msg(inode
->i_sb
, KERN_ERR
,
971 "Inode %lu (%p): orphan list check failed!",
972 inode
->i_ino
, EXT4_I(inode
));
973 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
974 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
978 call_rcu(&inode
->i_rcu
, ext4_i_callback
);
981 static void init_once(void *foo
)
983 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
985 INIT_LIST_HEAD(&ei
->i_orphan
);
986 init_rwsem(&ei
->xattr_sem
);
987 init_rwsem(&ei
->i_data_sem
);
988 init_rwsem(&ei
->i_mmap_sem
);
989 inode_init_once(&ei
->vfs_inode
);
992 static int __init
init_inodecache(void)
994 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
995 sizeof(struct ext4_inode_info
),
996 0, (SLAB_RECLAIM_ACCOUNT
|
997 SLAB_MEM_SPREAD
|SLAB_ACCOUNT
),
999 if (ext4_inode_cachep
== NULL
)
1004 static void destroy_inodecache(void)
1007 * Make sure all delayed rcu free inodes are flushed before we
1011 kmem_cache_destroy(ext4_inode_cachep
);
1014 void ext4_clear_inode(struct inode
*inode
)
1016 invalidate_inode_buffers(inode
);
1019 ext4_discard_preallocations(inode
);
1020 ext4_es_remove_extent(inode
, 0, EXT_MAX_BLOCKS
);
1021 if (EXT4_I(inode
)->jinode
) {
1022 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode
),
1023 EXT4_I(inode
)->jinode
);
1024 jbd2_free_inode(EXT4_I(inode
)->jinode
);
1025 EXT4_I(inode
)->jinode
= NULL
;
1027 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1028 if (EXT4_I(inode
)->i_crypt_info
)
1029 ext4_free_encryption_info(inode
, EXT4_I(inode
)->i_crypt_info
);
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
);
1097 static char *quotatypes
[] = INITQFNAMES
;
1098 #define QTYPE2NAME(t) (quotatypes[t])
1100 static int ext4_write_dquot(struct dquot
*dquot
);
1101 static int ext4_acquire_dquot(struct dquot
*dquot
);
1102 static int ext4_release_dquot(struct dquot
*dquot
);
1103 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
1104 static int ext4_write_info(struct super_block
*sb
, int type
);
1105 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
1107 static int ext4_quota_off(struct super_block
*sb
, int type
);
1108 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
1109 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
1110 size_t len
, loff_t off
);
1111 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
1112 const char *data
, size_t len
, loff_t off
);
1113 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
1114 unsigned int flags
);
1115 static int ext4_enable_quotas(struct super_block
*sb
);
1117 static struct dquot
**ext4_get_dquots(struct inode
*inode
)
1119 return EXT4_I(inode
)->i_dquot
;
1122 static const struct dquot_operations ext4_quota_operations
= {
1123 .get_reserved_space
= ext4_get_reserved_space
,
1124 .write_dquot
= ext4_write_dquot
,
1125 .acquire_dquot
= ext4_acquire_dquot
,
1126 .release_dquot
= ext4_release_dquot
,
1127 .mark_dirty
= ext4_mark_dquot_dirty
,
1128 .write_info
= ext4_write_info
,
1129 .alloc_dquot
= dquot_alloc
,
1130 .destroy_dquot
= dquot_destroy
,
1131 .get_projid
= ext4_get_projid
,
1132 .get_next_id
= dquot_get_next_id
,
1135 static const struct quotactl_ops ext4_qctl_operations
= {
1136 .quota_on
= ext4_quota_on
,
1137 .quota_off
= ext4_quota_off
,
1138 .quota_sync
= dquot_quota_sync
,
1139 .get_state
= dquot_get_state
,
1140 .set_info
= dquot_set_dqinfo
,
1141 .get_dqblk
= dquot_get_dqblk
,
1142 .set_dqblk
= dquot_set_dqblk
,
1143 .get_nextdqblk
= dquot_get_next_dqblk
,
1147 static const struct super_operations ext4_sops
= {
1148 .alloc_inode
= ext4_alloc_inode
,
1149 .destroy_inode
= ext4_destroy_inode
,
1150 .write_inode
= ext4_write_inode
,
1151 .dirty_inode
= ext4_dirty_inode
,
1152 .drop_inode
= ext4_drop_inode
,
1153 .evict_inode
= ext4_evict_inode
,
1154 .put_super
= ext4_put_super
,
1155 .sync_fs
= ext4_sync_fs
,
1156 .freeze_fs
= ext4_freeze
,
1157 .unfreeze_fs
= ext4_unfreeze
,
1158 .statfs
= ext4_statfs
,
1159 .remount_fs
= ext4_remount
,
1160 .show_options
= ext4_show_options
,
1162 .quota_read
= ext4_quota_read
,
1163 .quota_write
= ext4_quota_write
,
1164 .get_dquots
= ext4_get_dquots
,
1166 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1169 static const struct export_operations ext4_export_ops
= {
1170 .fh_to_dentry
= ext4_fh_to_dentry
,
1171 .fh_to_parent
= ext4_fh_to_parent
,
1172 .get_parent
= ext4_get_parent
,
1176 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1177 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1178 Opt_nouid32
, Opt_debug
, Opt_removed
,
1179 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1180 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
,
1181 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
, Opt_journal_dev
,
1182 Opt_journal_path
, Opt_journal_checksum
, Opt_journal_async_commit
,
1183 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1184 Opt_data_err_abort
, Opt_data_err_ignore
, Opt_test_dummy_encryption
,
1185 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1186 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
1187 Opt_noquota
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
1188 Opt_usrquota
, Opt_grpquota
, Opt_i_version
, Opt_dax
,
1189 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
, Opt_mblk_io_submit
,
1190 Opt_lazytime
, Opt_nolazytime
,
1191 Opt_nomblk_io_submit
, Opt_block_validity
, Opt_noblock_validity
,
1192 Opt_inode_readahead_blks
, Opt_journal_ioprio
,
1193 Opt_dioread_nolock
, Opt_dioread_lock
,
1194 Opt_discard
, Opt_nodiscard
, Opt_init_itable
, Opt_noinit_itable
,
1195 Opt_max_dir_size_kb
, Opt_nojournal_checksum
,
1198 static const match_table_t tokens
= {
1199 {Opt_bsd_df
, "bsddf"},
1200 {Opt_minix_df
, "minixdf"},
1201 {Opt_grpid
, "grpid"},
1202 {Opt_grpid
, "bsdgroups"},
1203 {Opt_nogrpid
, "nogrpid"},
1204 {Opt_nogrpid
, "sysvgroups"},
1205 {Opt_resgid
, "resgid=%u"},
1206 {Opt_resuid
, "resuid=%u"},
1208 {Opt_err_cont
, "errors=continue"},
1209 {Opt_err_panic
, "errors=panic"},
1210 {Opt_err_ro
, "errors=remount-ro"},
1211 {Opt_nouid32
, "nouid32"},
1212 {Opt_debug
, "debug"},
1213 {Opt_removed
, "oldalloc"},
1214 {Opt_removed
, "orlov"},
1215 {Opt_user_xattr
, "user_xattr"},
1216 {Opt_nouser_xattr
, "nouser_xattr"},
1218 {Opt_noacl
, "noacl"},
1219 {Opt_noload
, "norecovery"},
1220 {Opt_noload
, "noload"},
1221 {Opt_removed
, "nobh"},
1222 {Opt_removed
, "bh"},
1223 {Opt_commit
, "commit=%u"},
1224 {Opt_min_batch_time
, "min_batch_time=%u"},
1225 {Opt_max_batch_time
, "max_batch_time=%u"},
1226 {Opt_journal_dev
, "journal_dev=%u"},
1227 {Opt_journal_path
, "journal_path=%s"},
1228 {Opt_journal_checksum
, "journal_checksum"},
1229 {Opt_nojournal_checksum
, "nojournal_checksum"},
1230 {Opt_journal_async_commit
, "journal_async_commit"},
1231 {Opt_abort
, "abort"},
1232 {Opt_data_journal
, "data=journal"},
1233 {Opt_data_ordered
, "data=ordered"},
1234 {Opt_data_writeback
, "data=writeback"},
1235 {Opt_data_err_abort
, "data_err=abort"},
1236 {Opt_data_err_ignore
, "data_err=ignore"},
1237 {Opt_offusrjquota
, "usrjquota="},
1238 {Opt_usrjquota
, "usrjquota=%s"},
1239 {Opt_offgrpjquota
, "grpjquota="},
1240 {Opt_grpjquota
, "grpjquota=%s"},
1241 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1242 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1243 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
1244 {Opt_grpquota
, "grpquota"},
1245 {Opt_noquota
, "noquota"},
1246 {Opt_quota
, "quota"},
1247 {Opt_usrquota
, "usrquota"},
1248 {Opt_barrier
, "barrier=%u"},
1249 {Opt_barrier
, "barrier"},
1250 {Opt_nobarrier
, "nobarrier"},
1251 {Opt_i_version
, "i_version"},
1253 {Opt_stripe
, "stripe=%u"},
1254 {Opt_delalloc
, "delalloc"},
1255 {Opt_lazytime
, "lazytime"},
1256 {Opt_nolazytime
, "nolazytime"},
1257 {Opt_nodelalloc
, "nodelalloc"},
1258 {Opt_removed
, "mblk_io_submit"},
1259 {Opt_removed
, "nomblk_io_submit"},
1260 {Opt_block_validity
, "block_validity"},
1261 {Opt_noblock_validity
, "noblock_validity"},
1262 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1263 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1264 {Opt_auto_da_alloc
, "auto_da_alloc=%u"},
1265 {Opt_auto_da_alloc
, "auto_da_alloc"},
1266 {Opt_noauto_da_alloc
, "noauto_da_alloc"},
1267 {Opt_dioread_nolock
, "dioread_nolock"},
1268 {Opt_dioread_lock
, "dioread_lock"},
1269 {Opt_discard
, "discard"},
1270 {Opt_nodiscard
, "nodiscard"},
1271 {Opt_init_itable
, "init_itable=%u"},
1272 {Opt_init_itable
, "init_itable"},
1273 {Opt_noinit_itable
, "noinit_itable"},
1274 {Opt_max_dir_size_kb
, "max_dir_size_kb=%u"},
1275 {Opt_test_dummy_encryption
, "test_dummy_encryption"},
1276 {Opt_removed
, "check=none"}, /* mount option from ext2/3 */
1277 {Opt_removed
, "nocheck"}, /* mount option from ext2/3 */
1278 {Opt_removed
, "reservation"}, /* mount option from ext2/3 */
1279 {Opt_removed
, "noreservation"}, /* mount option from ext2/3 */
1280 {Opt_removed
, "journal=%u"}, /* mount option from ext2/3 */
1284 static ext4_fsblk_t
get_sb_block(void **data
)
1286 ext4_fsblk_t sb_block
;
1287 char *options
= (char *) *data
;
1289 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1290 return 1; /* Default location */
1293 /* TODO: use simple_strtoll with >32bit ext4 */
1294 sb_block
= simple_strtoul(options
, &options
, 0);
1295 if (*options
&& *options
!= ',') {
1296 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1300 if (*options
== ',')
1302 *data
= (void *) options
;
1307 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1308 static char deprecated_msg
[] = "Mount option \"%s\" will be removed by %s\n"
1309 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1312 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
1314 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1318 if (sb_any_quota_loaded(sb
) &&
1319 !sbi
->s_qf_names
[qtype
]) {
1320 ext4_msg(sb
, KERN_ERR
,
1321 "Cannot change journaled "
1322 "quota options when quota turned on");
1325 if (ext4_has_feature_quota(sb
)) {
1326 ext4_msg(sb
, KERN_ERR
, "Cannot set journaled quota options "
1327 "when QUOTA feature is enabled");
1330 qname
= match_strdup(args
);
1332 ext4_msg(sb
, KERN_ERR
,
1333 "Not enough memory for storing quotafile name");
1336 if (sbi
->s_qf_names
[qtype
]) {
1337 if (strcmp(sbi
->s_qf_names
[qtype
], qname
) == 0)
1340 ext4_msg(sb
, KERN_ERR
,
1341 "%s quota file already specified",
1345 if (strchr(qname
, '/')) {
1346 ext4_msg(sb
, KERN_ERR
,
1347 "quotafile must be on filesystem root");
1350 sbi
->s_qf_names
[qtype
] = qname
;
1358 static int clear_qf_name(struct super_block
*sb
, int qtype
)
1361 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1363 if (sb_any_quota_loaded(sb
) &&
1364 sbi
->s_qf_names
[qtype
]) {
1365 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
1366 " when quota turned on");
1369 kfree(sbi
->s_qf_names
[qtype
]);
1370 sbi
->s_qf_names
[qtype
] = NULL
;
1375 #define MOPT_SET 0x0001
1376 #define MOPT_CLEAR 0x0002
1377 #define MOPT_NOSUPPORT 0x0004
1378 #define MOPT_EXPLICIT 0x0008
1379 #define MOPT_CLEAR_ERR 0x0010
1380 #define MOPT_GTE0 0x0020
1383 #define MOPT_QFMT 0x0040
1385 #define MOPT_Q MOPT_NOSUPPORT
1386 #define MOPT_QFMT MOPT_NOSUPPORT
1388 #define MOPT_DATAJ 0x0080
1389 #define MOPT_NO_EXT2 0x0100
1390 #define MOPT_NO_EXT3 0x0200
1391 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1392 #define MOPT_STRING 0x0400
1394 static const struct mount_opts
{
1398 } ext4_mount_opts
[] = {
1399 {Opt_minix_df
, EXT4_MOUNT_MINIX_DF
, MOPT_SET
},
1400 {Opt_bsd_df
, EXT4_MOUNT_MINIX_DF
, MOPT_CLEAR
},
1401 {Opt_grpid
, EXT4_MOUNT_GRPID
, MOPT_SET
},
1402 {Opt_nogrpid
, EXT4_MOUNT_GRPID
, MOPT_CLEAR
},
1403 {Opt_block_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_SET
},
1404 {Opt_noblock_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_CLEAR
},
1405 {Opt_dioread_nolock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1406 MOPT_EXT4_ONLY
| MOPT_SET
},
1407 {Opt_dioread_lock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1408 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1409 {Opt_discard
, EXT4_MOUNT_DISCARD
, MOPT_SET
},
1410 {Opt_nodiscard
, EXT4_MOUNT_DISCARD
, MOPT_CLEAR
},
1411 {Opt_delalloc
, EXT4_MOUNT_DELALLOC
,
1412 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1413 {Opt_nodelalloc
, EXT4_MOUNT_DELALLOC
,
1414 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1415 {Opt_nojournal_checksum
, EXT4_MOUNT_JOURNAL_CHECKSUM
,
1416 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1417 {Opt_journal_checksum
, EXT4_MOUNT_JOURNAL_CHECKSUM
,
1418 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1419 {Opt_journal_async_commit
, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT
|
1420 EXT4_MOUNT_JOURNAL_CHECKSUM
),
1421 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1422 {Opt_noload
, EXT4_MOUNT_NOLOAD
, MOPT_NO_EXT2
| MOPT_SET
},
1423 {Opt_err_panic
, EXT4_MOUNT_ERRORS_PANIC
, MOPT_SET
| MOPT_CLEAR_ERR
},
1424 {Opt_err_ro
, EXT4_MOUNT_ERRORS_RO
, MOPT_SET
| MOPT_CLEAR_ERR
},
1425 {Opt_err_cont
, EXT4_MOUNT_ERRORS_CONT
, MOPT_SET
| MOPT_CLEAR_ERR
},
1426 {Opt_data_err_abort
, EXT4_MOUNT_DATA_ERR_ABORT
,
1428 {Opt_data_err_ignore
, EXT4_MOUNT_DATA_ERR_ABORT
,
1430 {Opt_barrier
, EXT4_MOUNT_BARRIER
, MOPT_SET
},
1431 {Opt_nobarrier
, EXT4_MOUNT_BARRIER
, MOPT_CLEAR
},
1432 {Opt_noauto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_SET
},
1433 {Opt_auto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_CLEAR
},
1434 {Opt_noinit_itable
, EXT4_MOUNT_INIT_INODE_TABLE
, MOPT_CLEAR
},
1435 {Opt_commit
, 0, MOPT_GTE0
},
1436 {Opt_max_batch_time
, 0, MOPT_GTE0
},
1437 {Opt_min_batch_time
, 0, MOPT_GTE0
},
1438 {Opt_inode_readahead_blks
, 0, MOPT_GTE0
},
1439 {Opt_init_itable
, 0, MOPT_GTE0
},
1440 {Opt_dax
, EXT4_MOUNT_DAX
, MOPT_SET
},
1441 {Opt_stripe
, 0, MOPT_GTE0
},
1442 {Opt_resuid
, 0, MOPT_GTE0
},
1443 {Opt_resgid
, 0, MOPT_GTE0
},
1444 {Opt_journal_dev
, 0, MOPT_NO_EXT2
| MOPT_GTE0
},
1445 {Opt_journal_path
, 0, MOPT_NO_EXT2
| MOPT_STRING
},
1446 {Opt_journal_ioprio
, 0, MOPT_NO_EXT2
| MOPT_GTE0
},
1447 {Opt_data_journal
, EXT4_MOUNT_JOURNAL_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1448 {Opt_data_ordered
, EXT4_MOUNT_ORDERED_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1449 {Opt_data_writeback
, EXT4_MOUNT_WRITEBACK_DATA
,
1450 MOPT_NO_EXT2
| MOPT_DATAJ
},
1451 {Opt_user_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_SET
},
1452 {Opt_nouser_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_CLEAR
},
1453 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1454 {Opt_acl
, EXT4_MOUNT_POSIX_ACL
, MOPT_SET
},
1455 {Opt_noacl
, EXT4_MOUNT_POSIX_ACL
, MOPT_CLEAR
},
1457 {Opt_acl
, 0, MOPT_NOSUPPORT
},
1458 {Opt_noacl
, 0, MOPT_NOSUPPORT
},
1460 {Opt_nouid32
, EXT4_MOUNT_NO_UID32
, MOPT_SET
},
1461 {Opt_debug
, EXT4_MOUNT_DEBUG
, MOPT_SET
},
1462 {Opt_quota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
, MOPT_SET
| MOPT_Q
},
1463 {Opt_usrquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
,
1465 {Opt_grpquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_GRPQUOTA
,
1467 {Opt_noquota
, (EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
|
1468 EXT4_MOUNT_GRPQUOTA
), MOPT_CLEAR
| MOPT_Q
},
1469 {Opt_usrjquota
, 0, MOPT_Q
},
1470 {Opt_grpjquota
, 0, MOPT_Q
},
1471 {Opt_offusrjquota
, 0, MOPT_Q
},
1472 {Opt_offgrpjquota
, 0, MOPT_Q
},
1473 {Opt_jqfmt_vfsold
, QFMT_VFS_OLD
, MOPT_QFMT
},
1474 {Opt_jqfmt_vfsv0
, QFMT_VFS_V0
, MOPT_QFMT
},
1475 {Opt_jqfmt_vfsv1
, QFMT_VFS_V1
, MOPT_QFMT
},
1476 {Opt_max_dir_size_kb
, 0, MOPT_GTE0
},
1477 {Opt_test_dummy_encryption
, 0, MOPT_GTE0
},
1481 static int handle_mount_opt(struct super_block
*sb
, char *opt
, int token
,
1482 substring_t
*args
, unsigned long *journal_devnum
,
1483 unsigned int *journal_ioprio
, int is_remount
)
1485 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1486 const struct mount_opts
*m
;
1492 if (token
== Opt_usrjquota
)
1493 return set_qf_name(sb
, USRQUOTA
, &args
[0]);
1494 else if (token
== Opt_grpjquota
)
1495 return set_qf_name(sb
, GRPQUOTA
, &args
[0]);
1496 else if (token
== Opt_offusrjquota
)
1497 return clear_qf_name(sb
, USRQUOTA
);
1498 else if (token
== Opt_offgrpjquota
)
1499 return clear_qf_name(sb
, GRPQUOTA
);
1503 case Opt_nouser_xattr
:
1504 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, opt
, "3.5");
1507 return 1; /* handled by get_sb_block() */
1509 ext4_msg(sb
, KERN_WARNING
, "Ignoring removed %s option", opt
);
1512 sbi
->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
1515 sb
->s_flags
|= MS_I_VERSION
;
1518 sb
->s_flags
|= MS_LAZYTIME
;
1520 case Opt_nolazytime
:
1521 sb
->s_flags
&= ~MS_LAZYTIME
;
1525 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++)
1526 if (token
== m
->token
)
1529 if (m
->token
== Opt_err
) {
1530 ext4_msg(sb
, KERN_ERR
, "Unrecognized mount option \"%s\" "
1531 "or missing value", opt
);
1535 if ((m
->flags
& MOPT_NO_EXT2
) && IS_EXT2_SB(sb
)) {
1536 ext4_msg(sb
, KERN_ERR
,
1537 "Mount option \"%s\" incompatible with ext2", opt
);
1540 if ((m
->flags
& MOPT_NO_EXT3
) && IS_EXT3_SB(sb
)) {
1541 ext4_msg(sb
, KERN_ERR
,
1542 "Mount option \"%s\" incompatible with ext3", opt
);
1546 if (args
->from
&& !(m
->flags
& MOPT_STRING
) && match_int(args
, &arg
))
1548 if (args
->from
&& (m
->flags
& MOPT_GTE0
) && (arg
< 0))
1550 if (m
->flags
& MOPT_EXPLICIT
) {
1551 if (m
->mount_opt
& EXT4_MOUNT_DELALLOC
) {
1552 set_opt2(sb
, EXPLICIT_DELALLOC
);
1553 } else if (m
->mount_opt
& EXT4_MOUNT_JOURNAL_CHECKSUM
) {
1554 set_opt2(sb
, EXPLICIT_JOURNAL_CHECKSUM
);
1558 if (m
->flags
& MOPT_CLEAR_ERR
)
1559 clear_opt(sb
, ERRORS_MASK
);
1560 if (token
== Opt_noquota
&& sb_any_quota_loaded(sb
)) {
1561 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
1562 "options when quota turned on");
1566 if (m
->flags
& MOPT_NOSUPPORT
) {
1567 ext4_msg(sb
, KERN_ERR
, "%s option not supported", opt
);
1568 } else if (token
== Opt_commit
) {
1570 arg
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1571 sbi
->s_commit_interval
= HZ
* arg
;
1572 } else if (token
== Opt_max_batch_time
) {
1573 sbi
->s_max_batch_time
= arg
;
1574 } else if (token
== Opt_min_batch_time
) {
1575 sbi
->s_min_batch_time
= arg
;
1576 } else if (token
== Opt_inode_readahead_blks
) {
1577 if (arg
&& (arg
> (1 << 30) || !is_power_of_2(arg
))) {
1578 ext4_msg(sb
, KERN_ERR
,
1579 "EXT4-fs: inode_readahead_blks must be "
1580 "0 or a power of 2 smaller than 2^31");
1583 sbi
->s_inode_readahead_blks
= arg
;
1584 } else if (token
== Opt_init_itable
) {
1585 set_opt(sb
, INIT_INODE_TABLE
);
1587 arg
= EXT4_DEF_LI_WAIT_MULT
;
1588 sbi
->s_li_wait_mult
= arg
;
1589 } else if (token
== Opt_max_dir_size_kb
) {
1590 sbi
->s_max_dir_size_kb
= arg
;
1591 } else if (token
== Opt_stripe
) {
1592 sbi
->s_stripe
= arg
;
1593 } else if (token
== Opt_resuid
) {
1594 uid
= make_kuid(current_user_ns(), arg
);
1595 if (!uid_valid(uid
)) {
1596 ext4_msg(sb
, KERN_ERR
, "Invalid uid value %d", arg
);
1599 sbi
->s_resuid
= uid
;
1600 } else if (token
== Opt_resgid
) {
1601 gid
= make_kgid(current_user_ns(), arg
);
1602 if (!gid_valid(gid
)) {
1603 ext4_msg(sb
, KERN_ERR
, "Invalid gid value %d", arg
);
1606 sbi
->s_resgid
= gid
;
1607 } else if (token
== Opt_journal_dev
) {
1609 ext4_msg(sb
, KERN_ERR
,
1610 "Cannot specify journal on remount");
1613 *journal_devnum
= arg
;
1614 } else if (token
== Opt_journal_path
) {
1616 struct inode
*journal_inode
;
1621 ext4_msg(sb
, KERN_ERR
,
1622 "Cannot specify journal on remount");
1625 journal_path
= match_strdup(&args
[0]);
1626 if (!journal_path
) {
1627 ext4_msg(sb
, KERN_ERR
, "error: could not dup "
1628 "journal device string");
1632 error
= kern_path(journal_path
, LOOKUP_FOLLOW
, &path
);
1634 ext4_msg(sb
, KERN_ERR
, "error: could not find "
1635 "journal device path: error %d", error
);
1636 kfree(journal_path
);
1640 journal_inode
= d_inode(path
.dentry
);
1641 if (!S_ISBLK(journal_inode
->i_mode
)) {
1642 ext4_msg(sb
, KERN_ERR
, "error: journal path %s "
1643 "is not a block device", journal_path
);
1645 kfree(journal_path
);
1649 *journal_devnum
= new_encode_dev(journal_inode
->i_rdev
);
1651 kfree(journal_path
);
1652 } else if (token
== Opt_journal_ioprio
) {
1654 ext4_msg(sb
, KERN_ERR
, "Invalid journal IO priority"
1659 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
, arg
);
1660 } else if (token
== Opt_test_dummy_encryption
) {
1661 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1662 sbi
->s_mount_flags
|= EXT4_MF_TEST_DUMMY_ENCRYPTION
;
1663 ext4_msg(sb
, KERN_WARNING
,
1664 "Test dummy encryption mode enabled");
1666 ext4_msg(sb
, KERN_WARNING
,
1667 "Test dummy encryption mount option ignored");
1669 } else if (m
->flags
& MOPT_DATAJ
) {
1671 if (!sbi
->s_journal
)
1672 ext4_msg(sb
, KERN_WARNING
, "Remounting file system with no journal so ignoring journalled data option");
1673 else if (test_opt(sb
, DATA_FLAGS
) != m
->mount_opt
) {
1674 ext4_msg(sb
, KERN_ERR
,
1675 "Cannot change data mode on remount");
1679 clear_opt(sb
, DATA_FLAGS
);
1680 sbi
->s_mount_opt
|= m
->mount_opt
;
1683 } else if (m
->flags
& MOPT_QFMT
) {
1684 if (sb_any_quota_loaded(sb
) &&
1685 sbi
->s_jquota_fmt
!= m
->mount_opt
) {
1686 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled "
1687 "quota options when quota turned on");
1690 if (ext4_has_feature_quota(sb
)) {
1691 ext4_msg(sb
, KERN_ERR
,
1692 "Cannot set journaled quota options "
1693 "when QUOTA feature is enabled");
1696 sbi
->s_jquota_fmt
= m
->mount_opt
;
1698 } else if (token
== Opt_dax
) {
1699 #ifdef CONFIG_FS_DAX
1700 ext4_msg(sb
, KERN_WARNING
,
1701 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1702 sbi
->s_mount_opt
|= m
->mount_opt
;
1704 ext4_msg(sb
, KERN_INFO
, "dax option not supported");
1707 } else if (token
== Opt_data_err_abort
) {
1708 sbi
->s_mount_opt
|= m
->mount_opt
;
1709 } else if (token
== Opt_data_err_ignore
) {
1710 sbi
->s_mount_opt
&= ~m
->mount_opt
;
1714 if (m
->flags
& MOPT_CLEAR
)
1716 else if (unlikely(!(m
->flags
& MOPT_SET
))) {
1717 ext4_msg(sb
, KERN_WARNING
,
1718 "buggy handling of option %s", opt
);
1723 sbi
->s_mount_opt
|= m
->mount_opt
;
1725 sbi
->s_mount_opt
&= ~m
->mount_opt
;
1730 static int parse_options(char *options
, struct super_block
*sb
,
1731 unsigned long *journal_devnum
,
1732 unsigned int *journal_ioprio
,
1735 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1737 substring_t args
[MAX_OPT_ARGS
];
1743 while ((p
= strsep(&options
, ",")) != NULL
) {
1747 * Initialize args struct so we know whether arg was
1748 * found; some options take optional arguments.
1750 args
[0].to
= args
[0].from
= NULL
;
1751 token
= match_token(p
, tokens
, args
);
1752 if (handle_mount_opt(sb
, p
, token
, args
, journal_devnum
,
1753 journal_ioprio
, is_remount
) < 0)
1757 if (ext4_has_feature_quota(sb
) &&
1758 (test_opt(sb
, USRQUOTA
) || test_opt(sb
, GRPQUOTA
))) {
1759 ext4_msg(sb
, KERN_ERR
, "Cannot set quota options when QUOTA "
1760 "feature is enabled");
1763 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1764 if (test_opt(sb
, USRQUOTA
) && sbi
->s_qf_names
[USRQUOTA
])
1765 clear_opt(sb
, USRQUOTA
);
1767 if (test_opt(sb
, GRPQUOTA
) && sbi
->s_qf_names
[GRPQUOTA
])
1768 clear_opt(sb
, GRPQUOTA
);
1770 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
1771 ext4_msg(sb
, KERN_ERR
, "old and new quota "
1776 if (!sbi
->s_jquota_fmt
) {
1777 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1783 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
1785 BLOCK_SIZE
<< le32_to_cpu(sbi
->s_es
->s_log_block_size
);
1787 if (blocksize
< PAGE_SIZE
) {
1788 ext4_msg(sb
, KERN_ERR
, "can't mount with "
1789 "dioread_nolock if block size != PAGE_SIZE");
1793 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
&&
1794 test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
1795 ext4_msg(sb
, KERN_ERR
, "can't mount with journal_async_commit "
1796 "in data=ordered mode");
1802 static inline void ext4_show_quota_options(struct seq_file
*seq
,
1803 struct super_block
*sb
)
1805 #if defined(CONFIG_QUOTA)
1806 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1808 if (sbi
->s_jquota_fmt
) {
1811 switch (sbi
->s_jquota_fmt
) {
1822 seq_printf(seq
, ",jqfmt=%s", fmtname
);
1825 if (sbi
->s_qf_names
[USRQUOTA
])
1826 seq_show_option(seq
, "usrjquota", sbi
->s_qf_names
[USRQUOTA
]);
1828 if (sbi
->s_qf_names
[GRPQUOTA
])
1829 seq_show_option(seq
, "grpjquota", sbi
->s_qf_names
[GRPQUOTA
]);
1833 static const char *token2str(int token
)
1835 const struct match_token
*t
;
1837 for (t
= tokens
; t
->token
!= Opt_err
; t
++)
1838 if (t
->token
== token
&& !strchr(t
->pattern
, '='))
1845 * - it's set to a non-default value OR
1846 * - if the per-sb default is different from the global default
1848 static int _ext4_show_options(struct seq_file
*seq
, struct super_block
*sb
,
1851 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1852 struct ext4_super_block
*es
= sbi
->s_es
;
1853 int def_errors
, def_mount_opt
= nodefs
? 0 : sbi
->s_def_mount_opt
;
1854 const struct mount_opts
*m
;
1855 char sep
= nodefs
? '\n' : ',';
1857 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1858 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1860 if (sbi
->s_sb_block
!= 1)
1861 SEQ_OPTS_PRINT("sb=%llu", sbi
->s_sb_block
);
1863 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++) {
1864 int want_set
= m
->flags
& MOPT_SET
;
1865 if (((m
->flags
& (MOPT_SET
|MOPT_CLEAR
)) == 0) ||
1866 (m
->flags
& MOPT_CLEAR_ERR
))
1868 if (!(m
->mount_opt
& (sbi
->s_mount_opt
^ def_mount_opt
)))
1869 continue; /* skip if same as the default */
1871 (sbi
->s_mount_opt
& m
->mount_opt
) != m
->mount_opt
) ||
1872 (!want_set
&& (sbi
->s_mount_opt
& m
->mount_opt
)))
1873 continue; /* select Opt_noFoo vs Opt_Foo */
1874 SEQ_OPTS_PRINT("%s", token2str(m
->token
));
1877 if (nodefs
|| !uid_eq(sbi
->s_resuid
, make_kuid(&init_user_ns
, EXT4_DEF_RESUID
)) ||
1878 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
)
1879 SEQ_OPTS_PRINT("resuid=%u",
1880 from_kuid_munged(&init_user_ns
, sbi
->s_resuid
));
1881 if (nodefs
|| !gid_eq(sbi
->s_resgid
, make_kgid(&init_user_ns
, EXT4_DEF_RESGID
)) ||
1882 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
)
1883 SEQ_OPTS_PRINT("resgid=%u",
1884 from_kgid_munged(&init_user_ns
, sbi
->s_resgid
));
1885 def_errors
= nodefs
? -1 : le16_to_cpu(es
->s_errors
);
1886 if (test_opt(sb
, ERRORS_RO
) && def_errors
!= EXT4_ERRORS_RO
)
1887 SEQ_OPTS_PUTS("errors=remount-ro");
1888 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
1889 SEQ_OPTS_PUTS("errors=continue");
1890 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
1891 SEQ_OPTS_PUTS("errors=panic");
1892 if (nodefs
|| sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
)
1893 SEQ_OPTS_PRINT("commit=%lu", sbi
->s_commit_interval
/ HZ
);
1894 if (nodefs
|| sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
)
1895 SEQ_OPTS_PRINT("min_batch_time=%u", sbi
->s_min_batch_time
);
1896 if (nodefs
|| sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
)
1897 SEQ_OPTS_PRINT("max_batch_time=%u", sbi
->s_max_batch_time
);
1898 if (sb
->s_flags
& MS_I_VERSION
)
1899 SEQ_OPTS_PUTS("i_version");
1900 if (nodefs
|| sbi
->s_stripe
)
1901 SEQ_OPTS_PRINT("stripe=%lu", sbi
->s_stripe
);
1902 if (EXT4_MOUNT_DATA_FLAGS
& (sbi
->s_mount_opt
^ def_mount_opt
)) {
1903 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
1904 SEQ_OPTS_PUTS("data=journal");
1905 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
1906 SEQ_OPTS_PUTS("data=ordered");
1907 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
1908 SEQ_OPTS_PUTS("data=writeback");
1911 sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
1912 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1913 sbi
->s_inode_readahead_blks
);
1915 if (nodefs
|| (test_opt(sb
, INIT_INODE_TABLE
) &&
1916 (sbi
->s_li_wait_mult
!= EXT4_DEF_LI_WAIT_MULT
)))
1917 SEQ_OPTS_PRINT("init_itable=%u", sbi
->s_li_wait_mult
);
1918 if (nodefs
|| sbi
->s_max_dir_size_kb
)
1919 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi
->s_max_dir_size_kb
);
1920 if (test_opt(sb
, DATA_ERR_ABORT
))
1921 SEQ_OPTS_PUTS("data_err=abort");
1923 ext4_show_quota_options(seq
, sb
);
1927 static int ext4_show_options(struct seq_file
*seq
, struct dentry
*root
)
1929 return _ext4_show_options(seq
, root
->d_sb
, 0);
1932 int ext4_seq_options_show(struct seq_file
*seq
, void *offset
)
1934 struct super_block
*sb
= seq
->private;
1937 seq_puts(seq
, (sb
->s_flags
& MS_RDONLY
) ? "ro" : "rw");
1938 rc
= _ext4_show_options(seq
, sb
, 1);
1939 seq_puts(seq
, "\n");
1943 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
1946 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1949 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
1950 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
1951 "forcing read-only mode");
1956 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
1957 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
1958 "running e2fsck is recommended");
1959 else if (sbi
->s_mount_state
& EXT4_ERROR_FS
)
1960 ext4_msg(sb
, KERN_WARNING
,
1961 "warning: mounting fs with errors, "
1962 "running e2fsck is recommended");
1963 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) > 0 &&
1964 le16_to_cpu(es
->s_mnt_count
) >=
1965 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1966 ext4_msg(sb
, KERN_WARNING
,
1967 "warning: maximal mount count reached, "
1968 "running e2fsck is recommended");
1969 else if (le32_to_cpu(es
->s_checkinterval
) &&
1970 (le32_to_cpu(es
->s_lastcheck
) +
1971 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1972 ext4_msg(sb
, KERN_WARNING
,
1973 "warning: checktime reached, "
1974 "running e2fsck is recommended");
1975 if (!sbi
->s_journal
)
1976 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
1977 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1978 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
1979 le16_add_cpu(&es
->s_mnt_count
, 1);
1980 es
->s_mtime
= cpu_to_le32(get_seconds());
1981 ext4_update_dynamic_rev(sb
);
1983 ext4_set_feature_journal_needs_recovery(sb
);
1985 ext4_commit_super(sb
, 1);
1987 if (test_opt(sb
, DEBUG
))
1988 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
1989 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1991 sbi
->s_groups_count
,
1992 EXT4_BLOCKS_PER_GROUP(sb
),
1993 EXT4_INODES_PER_GROUP(sb
),
1994 sbi
->s_mount_opt
, sbi
->s_mount_opt2
);
1996 cleancache_init_fs(sb
);
2000 int ext4_alloc_flex_bg_array(struct super_block
*sb
, ext4_group_t ngroup
)
2002 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2003 struct flex_groups
*new_groups
;
2006 if (!sbi
->s_log_groups_per_flex
)
2009 size
= ext4_flex_group(sbi
, ngroup
- 1) + 1;
2010 if (size
<= sbi
->s_flex_groups_allocated
)
2013 size
= roundup_pow_of_two(size
* sizeof(struct flex_groups
));
2014 new_groups
= ext4_kvzalloc(size
, GFP_KERNEL
);
2016 ext4_msg(sb
, KERN_ERR
, "not enough memory for %d flex groups",
2017 size
/ (int) sizeof(struct flex_groups
));
2021 if (sbi
->s_flex_groups
) {
2022 memcpy(new_groups
, sbi
->s_flex_groups
,
2023 (sbi
->s_flex_groups_allocated
*
2024 sizeof(struct flex_groups
)));
2025 kvfree(sbi
->s_flex_groups
);
2027 sbi
->s_flex_groups
= new_groups
;
2028 sbi
->s_flex_groups_allocated
= size
/ sizeof(struct flex_groups
);
2032 static int ext4_fill_flex_info(struct super_block
*sb
)
2034 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2035 struct ext4_group_desc
*gdp
= NULL
;
2036 ext4_group_t flex_group
;
2039 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
2040 if (sbi
->s_log_groups_per_flex
< 1 || sbi
->s_log_groups_per_flex
> 31) {
2041 sbi
->s_log_groups_per_flex
= 0;
2045 err
= ext4_alloc_flex_bg_array(sb
, sbi
->s_groups_count
);
2049 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2050 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2052 flex_group
= ext4_flex_group(sbi
, i
);
2053 atomic_add(ext4_free_inodes_count(sb
, gdp
),
2054 &sbi
->s_flex_groups
[flex_group
].free_inodes
);
2055 atomic64_add(ext4_free_group_clusters(sb
, gdp
),
2056 &sbi
->s_flex_groups
[flex_group
].free_clusters
);
2057 atomic_add(ext4_used_dirs_count(sb
, gdp
),
2058 &sbi
->s_flex_groups
[flex_group
].used_dirs
);
2066 static __le16
ext4_group_desc_csum(struct super_block
*sb
, __u32 block_group
,
2067 struct ext4_group_desc
*gdp
)
2071 __le32 le_group
= cpu_to_le32(block_group
);
2072 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2074 if (ext4_has_metadata_csum(sbi
->s_sb
)) {
2075 /* Use new metadata_csum algorithm */
2079 save_csum
= gdp
->bg_checksum
;
2080 gdp
->bg_checksum
= 0;
2081 csum32
= ext4_chksum(sbi
, sbi
->s_csum_seed
, (__u8
*)&le_group
,
2083 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)gdp
,
2085 gdp
->bg_checksum
= save_csum
;
2087 crc
= csum32
& 0xFFFF;
2091 /* old crc16 code */
2092 if (!ext4_has_feature_gdt_csum(sb
))
2095 offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
2097 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
2098 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
2099 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
2100 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
2101 /* for checksum of struct ext4_group_desc do the rest...*/
2102 if (ext4_has_feature_64bit(sb
) &&
2103 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
2104 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
2105 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
2109 return cpu_to_le16(crc
);
2112 int ext4_group_desc_csum_verify(struct super_block
*sb
, __u32 block_group
,
2113 struct ext4_group_desc
*gdp
)
2115 if (ext4_has_group_desc_csum(sb
) &&
2116 (gdp
->bg_checksum
!= ext4_group_desc_csum(sb
, block_group
, gdp
)))
2122 void ext4_group_desc_csum_set(struct super_block
*sb
, __u32 block_group
,
2123 struct ext4_group_desc
*gdp
)
2125 if (!ext4_has_group_desc_csum(sb
))
2127 gdp
->bg_checksum
= ext4_group_desc_csum(sb
, block_group
, gdp
);
2130 /* Called at mount-time, super-block is locked */
2131 static int ext4_check_descriptors(struct super_block
*sb
,
2132 ext4_group_t
*first_not_zeroed
)
2134 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2135 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
2136 ext4_fsblk_t last_block
;
2137 ext4_fsblk_t block_bitmap
;
2138 ext4_fsblk_t inode_bitmap
;
2139 ext4_fsblk_t inode_table
;
2140 int flexbg_flag
= 0;
2141 ext4_group_t i
, grp
= sbi
->s_groups_count
;
2143 if (ext4_has_feature_flex_bg(sb
))
2146 ext4_debug("Checking group descriptors");
2148 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2149 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2151 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
2152 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
2154 last_block
= first_block
+
2155 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2157 if ((grp
== sbi
->s_groups_count
) &&
2158 !(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2161 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
2162 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
2163 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2164 "Block bitmap for group %u not in group "
2165 "(block %llu)!", i
, block_bitmap
);
2168 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
2169 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
2170 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2171 "Inode bitmap for group %u not in group "
2172 "(block %llu)!", i
, inode_bitmap
);
2175 inode_table
= ext4_inode_table(sb
, gdp
);
2176 if (inode_table
< first_block
||
2177 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
2178 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2179 "Inode table for group %u not in group "
2180 "(block %llu)!", i
, inode_table
);
2183 ext4_lock_group(sb
, i
);
2184 if (!ext4_group_desc_csum_verify(sb
, i
, gdp
)) {
2185 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2186 "Checksum for group %u failed (%u!=%u)",
2187 i
, le16_to_cpu(ext4_group_desc_csum(sb
, i
,
2188 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
2189 if (!(sb
->s_flags
& MS_RDONLY
)) {
2190 ext4_unlock_group(sb
, i
);
2194 ext4_unlock_group(sb
, i
);
2196 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
2198 if (NULL
!= first_not_zeroed
)
2199 *first_not_zeroed
= grp
;
2203 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2204 * the superblock) which were deleted from all directories, but held open by
2205 * a process at the time of a crash. We walk the list and try to delete these
2206 * inodes at recovery time (only with a read-write filesystem).
2208 * In order to keep the orphan inode chain consistent during traversal (in
2209 * case of crash during recovery), we link each inode into the superblock
2210 * orphan list_head and handle it the same way as an inode deletion during
2211 * normal operation (which journals the operations for us).
2213 * We only do an iget() and an iput() on each inode, which is very safe if we
2214 * accidentally point at an in-use or already deleted inode. The worst that
2215 * can happen in this case is that we get a "bit already cleared" message from
2216 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2217 * e2fsck was run on this filesystem, and it must have already done the orphan
2218 * inode cleanup for us, so we can safely abort without any further action.
2220 static void ext4_orphan_cleanup(struct super_block
*sb
,
2221 struct ext4_super_block
*es
)
2223 unsigned int s_flags
= sb
->s_flags
;
2224 int nr_orphans
= 0, nr_truncates
= 0;
2228 if (!es
->s_last_orphan
) {
2229 jbd_debug(4, "no orphan inodes to clean up\n");
2233 if (bdev_read_only(sb
->s_bdev
)) {
2234 ext4_msg(sb
, KERN_ERR
, "write access "
2235 "unavailable, skipping orphan cleanup");
2239 /* Check if feature set would not allow a r/w mount */
2240 if (!ext4_feature_set_ok(sb
, 0)) {
2241 ext4_msg(sb
, KERN_INFO
, "Skipping orphan cleanup due to "
2242 "unknown ROCOMPAT features");
2246 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2247 /* don't clear list on RO mount w/ errors */
2248 if (es
->s_last_orphan
&& !(s_flags
& MS_RDONLY
)) {
2249 ext4_msg(sb
, KERN_INFO
, "Errors on filesystem, "
2250 "clearing orphan list.\n");
2251 es
->s_last_orphan
= 0;
2253 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2257 if (s_flags
& MS_RDONLY
) {
2258 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
2259 sb
->s_flags
&= ~MS_RDONLY
;
2262 /* Needed for iput() to work correctly and not trash data */
2263 sb
->s_flags
|= MS_ACTIVE
;
2264 /* Turn on quotas so that they are updated correctly */
2265 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
2266 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
2267 int ret
= ext4_quota_on_mount(sb
, i
);
2269 ext4_msg(sb
, KERN_ERR
,
2270 "Cannot turn on journaled "
2271 "quota: error %d", ret
);
2276 while (es
->s_last_orphan
) {
2277 struct inode
*inode
;
2279 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
2280 if (IS_ERR(inode
)) {
2281 es
->s_last_orphan
= 0;
2285 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
2286 dquot_initialize(inode
);
2287 if (inode
->i_nlink
) {
2288 if (test_opt(sb
, DEBUG
))
2289 ext4_msg(sb
, KERN_DEBUG
,
2290 "%s: truncating inode %lu to %lld bytes",
2291 __func__
, inode
->i_ino
, inode
->i_size
);
2292 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2293 inode
->i_ino
, inode
->i_size
);
2295 truncate_inode_pages(inode
->i_mapping
, inode
->i_size
);
2296 ext4_truncate(inode
);
2297 inode_unlock(inode
);
2300 if (test_opt(sb
, DEBUG
))
2301 ext4_msg(sb
, KERN_DEBUG
,
2302 "%s: deleting unreferenced inode %lu",
2303 __func__
, inode
->i_ino
);
2304 jbd_debug(2, "deleting unreferenced inode %lu\n",
2308 iput(inode
); /* The delete magic happens here! */
2311 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2314 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
2315 PLURAL(nr_orphans
));
2317 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
2318 PLURAL(nr_truncates
));
2320 /* Turn quotas off */
2321 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
2322 if (sb_dqopt(sb
)->files
[i
])
2323 dquot_quota_off(sb
, i
);
2326 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
2330 * Maximal extent format file size.
2331 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2332 * extent format containers, within a sector_t, and within i_blocks
2333 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2334 * so that won't be a limiting factor.
2336 * However there is other limiting factor. We do store extents in the form
2337 * of starting block and length, hence the resulting length of the extent
2338 * covering maximum file size must fit into on-disk format containers as
2339 * well. Given that length is always by 1 unit bigger than max unit (because
2340 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2342 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2344 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
2347 loff_t upper_limit
= MAX_LFS_FILESIZE
;
2349 /* small i_blocks in vfs inode? */
2350 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2352 * CONFIG_LBDAF is not enabled implies the inode
2353 * i_block represent total blocks in 512 bytes
2354 * 32 == size of vfs inode i_blocks * 8
2356 upper_limit
= (1LL << 32) - 1;
2358 /* total blocks in file system block size */
2359 upper_limit
>>= (blkbits
- 9);
2360 upper_limit
<<= blkbits
;
2364 * 32-bit extent-start container, ee_block. We lower the maxbytes
2365 * by one fs block, so ee_len can cover the extent of maximum file
2368 res
= (1LL << 32) - 1;
2371 /* Sanity check against vm- & vfs- imposed limits */
2372 if (res
> upper_limit
)
2379 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2380 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2381 * We need to be 1 filesystem block less than the 2^48 sector limit.
2383 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
2385 loff_t res
= EXT4_NDIR_BLOCKS
;
2388 /* This is calculated to be the largest file size for a dense, block
2389 * mapped file such that the file's total number of 512-byte sectors,
2390 * including data and all indirect blocks, does not exceed (2^48 - 1).
2392 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2393 * number of 512-byte sectors of the file.
2396 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2398 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2399 * the inode i_block field represents total file blocks in
2400 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2402 upper_limit
= (1LL << 32) - 1;
2404 /* total blocks in file system block size */
2405 upper_limit
>>= (bits
- 9);
2409 * We use 48 bit ext4_inode i_blocks
2410 * With EXT4_HUGE_FILE_FL set the i_blocks
2411 * represent total number of blocks in
2412 * file system block size
2414 upper_limit
= (1LL << 48) - 1;
2418 /* indirect blocks */
2420 /* double indirect blocks */
2421 meta_blocks
+= 1 + (1LL << (bits
-2));
2422 /* tripple indirect blocks */
2423 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
2425 upper_limit
-= meta_blocks
;
2426 upper_limit
<<= bits
;
2428 res
+= 1LL << (bits
-2);
2429 res
+= 1LL << (2*(bits
-2));
2430 res
+= 1LL << (3*(bits
-2));
2432 if (res
> upper_limit
)
2435 if (res
> MAX_LFS_FILESIZE
)
2436 res
= MAX_LFS_FILESIZE
;
2441 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
2442 ext4_fsblk_t logical_sb_block
, int nr
)
2444 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2445 ext4_group_t bg
, first_meta_bg
;
2448 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
2450 if (!ext4_has_feature_meta_bg(sb
) || nr
< first_meta_bg
)
2451 return logical_sb_block
+ nr
+ 1;
2452 bg
= sbi
->s_desc_per_block
* nr
;
2453 if (ext4_bg_has_super(sb
, bg
))
2457 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2458 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2459 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2462 if (sb
->s_blocksize
== 1024 && nr
== 0 &&
2463 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_first_data_block
) == 0)
2466 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
2470 * ext4_get_stripe_size: Get the stripe size.
2471 * @sbi: In memory super block info
2473 * If we have specified it via mount option, then
2474 * use the mount option value. If the value specified at mount time is
2475 * greater than the blocks per group use the super block value.
2476 * If the super block value is greater than blocks per group return 0.
2477 * Allocator needs it be less than blocks per group.
2480 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
2482 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
2483 unsigned long stripe_width
=
2484 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
2487 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
2488 ret
= sbi
->s_stripe
;
2489 else if (stripe_width
<= sbi
->s_blocks_per_group
)
2491 else if (stride
<= sbi
->s_blocks_per_group
)
2497 * If the stripe width is 1, this makes no sense and
2498 * we set it to 0 to turn off stripe handling code.
2507 * Check whether this filesystem can be mounted based on
2508 * the features present and the RDONLY/RDWR mount requested.
2509 * Returns 1 if this filesystem can be mounted as requested,
2510 * 0 if it cannot be.
2512 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
2514 if (ext4_has_unknown_ext4_incompat_features(sb
)) {
2515 ext4_msg(sb
, KERN_ERR
,
2516 "Couldn't mount because of "
2517 "unsupported optional features (%x)",
2518 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2519 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2526 if (ext4_has_feature_readonly(sb
)) {
2527 ext4_msg(sb
, KERN_INFO
, "filesystem is read-only");
2528 sb
->s_flags
|= MS_RDONLY
;
2532 /* Check that feature set is OK for a read-write mount */
2533 if (ext4_has_unknown_ext4_ro_compat_features(sb
)) {
2534 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
2535 "unsupported optional features (%x)",
2536 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2537 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2541 * Large file size enabled file system can only be mounted
2542 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2544 if (ext4_has_feature_huge_file(sb
)) {
2545 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
2546 ext4_msg(sb
, KERN_ERR
, "Filesystem with huge files "
2547 "cannot be mounted RDWR without "
2552 if (ext4_has_feature_bigalloc(sb
) && !ext4_has_feature_extents(sb
)) {
2553 ext4_msg(sb
, KERN_ERR
,
2554 "Can't support bigalloc feature without "
2555 "extents feature\n");
2559 #ifndef CONFIG_QUOTA
2560 if (ext4_has_feature_quota(sb
) && !readonly
) {
2561 ext4_msg(sb
, KERN_ERR
,
2562 "Filesystem with quota feature cannot be mounted RDWR "
2563 "without CONFIG_QUOTA");
2566 if (ext4_has_feature_project(sb
) && !readonly
) {
2567 ext4_msg(sb
, KERN_ERR
,
2568 "Filesystem with project quota feature cannot be mounted RDWR "
2569 "without CONFIG_QUOTA");
2572 #endif /* CONFIG_QUOTA */
2577 * This function is called once a day if we have errors logged
2578 * on the file system
2580 static void print_daily_error_info(unsigned long arg
)
2582 struct super_block
*sb
= (struct super_block
*) arg
;
2583 struct ext4_sb_info
*sbi
;
2584 struct ext4_super_block
*es
;
2589 if (es
->s_error_count
)
2590 /* fsck newer than v1.41.13 is needed to clean this condition. */
2591 ext4_msg(sb
, KERN_NOTICE
, "error count since last fsck: %u",
2592 le32_to_cpu(es
->s_error_count
));
2593 if (es
->s_first_error_time
) {
2594 printk(KERN_NOTICE
"EXT4-fs (%s): initial error at time %u: %.*s:%d",
2595 sb
->s_id
, le32_to_cpu(es
->s_first_error_time
),
2596 (int) sizeof(es
->s_first_error_func
),
2597 es
->s_first_error_func
,
2598 le32_to_cpu(es
->s_first_error_line
));
2599 if (es
->s_first_error_ino
)
2600 printk(": inode %u",
2601 le32_to_cpu(es
->s_first_error_ino
));
2602 if (es
->s_first_error_block
)
2603 printk(": block %llu", (unsigned long long)
2604 le64_to_cpu(es
->s_first_error_block
));
2607 if (es
->s_last_error_time
) {
2608 printk(KERN_NOTICE
"EXT4-fs (%s): last error at time %u: %.*s:%d",
2609 sb
->s_id
, le32_to_cpu(es
->s_last_error_time
),
2610 (int) sizeof(es
->s_last_error_func
),
2611 es
->s_last_error_func
,
2612 le32_to_cpu(es
->s_last_error_line
));
2613 if (es
->s_last_error_ino
)
2614 printk(": inode %u",
2615 le32_to_cpu(es
->s_last_error_ino
));
2616 if (es
->s_last_error_block
)
2617 printk(": block %llu", (unsigned long long)
2618 le64_to_cpu(es
->s_last_error_block
));
2621 mod_timer(&sbi
->s_err_report
, jiffies
+ 24*60*60*HZ
); /* Once a day */
2624 /* Find next suitable group and run ext4_init_inode_table */
2625 static int ext4_run_li_request(struct ext4_li_request
*elr
)
2627 struct ext4_group_desc
*gdp
= NULL
;
2628 ext4_group_t group
, ngroups
;
2629 struct super_block
*sb
;
2630 unsigned long timeout
= 0;
2634 ngroups
= EXT4_SB(sb
)->s_groups_count
;
2637 for (group
= elr
->lr_next_group
; group
< ngroups
; group
++) {
2638 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2644 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2648 if (group
>= ngroups
)
2653 ret
= ext4_init_inode_table(sb
, group
,
2654 elr
->lr_timeout
? 0 : 1);
2655 if (elr
->lr_timeout
== 0) {
2656 timeout
= (jiffies
- timeout
) *
2657 elr
->lr_sbi
->s_li_wait_mult
;
2658 elr
->lr_timeout
= timeout
;
2660 elr
->lr_next_sched
= jiffies
+ elr
->lr_timeout
;
2661 elr
->lr_next_group
= group
+ 1;
2669 * Remove lr_request from the list_request and free the
2670 * request structure. Should be called with li_list_mtx held
2672 static void ext4_remove_li_request(struct ext4_li_request
*elr
)
2674 struct ext4_sb_info
*sbi
;
2681 list_del(&elr
->lr_request
);
2682 sbi
->s_li_request
= NULL
;
2686 static void ext4_unregister_li_request(struct super_block
*sb
)
2688 mutex_lock(&ext4_li_mtx
);
2689 if (!ext4_li_info
) {
2690 mutex_unlock(&ext4_li_mtx
);
2694 mutex_lock(&ext4_li_info
->li_list_mtx
);
2695 ext4_remove_li_request(EXT4_SB(sb
)->s_li_request
);
2696 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2697 mutex_unlock(&ext4_li_mtx
);
2700 static struct task_struct
*ext4_lazyinit_task
;
2703 * This is the function where ext4lazyinit thread lives. It walks
2704 * through the request list searching for next scheduled filesystem.
2705 * When such a fs is found, run the lazy initialization request
2706 * (ext4_rn_li_request) and keep track of the time spend in this
2707 * function. Based on that time we compute next schedule time of
2708 * the request. When walking through the list is complete, compute
2709 * next waking time and put itself into sleep.
2711 static int ext4_lazyinit_thread(void *arg
)
2713 struct ext4_lazy_init
*eli
= (struct ext4_lazy_init
*)arg
;
2714 struct list_head
*pos
, *n
;
2715 struct ext4_li_request
*elr
;
2716 unsigned long next_wakeup
, cur
;
2718 BUG_ON(NULL
== eli
);
2722 next_wakeup
= MAX_JIFFY_OFFSET
;
2724 mutex_lock(&eli
->li_list_mtx
);
2725 if (list_empty(&eli
->li_request_list
)) {
2726 mutex_unlock(&eli
->li_list_mtx
);
2730 list_for_each_safe(pos
, n
, &eli
->li_request_list
) {
2731 elr
= list_entry(pos
, struct ext4_li_request
,
2734 if (time_after_eq(jiffies
, elr
->lr_next_sched
)) {
2735 if (ext4_run_li_request(elr
) != 0) {
2736 /* error, remove the lazy_init job */
2737 ext4_remove_li_request(elr
);
2742 if (time_before(elr
->lr_next_sched
, next_wakeup
))
2743 next_wakeup
= elr
->lr_next_sched
;
2745 mutex_unlock(&eli
->li_list_mtx
);
2750 if ((time_after_eq(cur
, next_wakeup
)) ||
2751 (MAX_JIFFY_OFFSET
== next_wakeup
)) {
2756 schedule_timeout_interruptible(next_wakeup
- cur
);
2758 if (kthread_should_stop()) {
2759 ext4_clear_request_list();
2766 * It looks like the request list is empty, but we need
2767 * to check it under the li_list_mtx lock, to prevent any
2768 * additions into it, and of course we should lock ext4_li_mtx
2769 * to atomically free the list and ext4_li_info, because at
2770 * this point another ext4 filesystem could be registering
2773 mutex_lock(&ext4_li_mtx
);
2774 mutex_lock(&eli
->li_list_mtx
);
2775 if (!list_empty(&eli
->li_request_list
)) {
2776 mutex_unlock(&eli
->li_list_mtx
);
2777 mutex_unlock(&ext4_li_mtx
);
2780 mutex_unlock(&eli
->li_list_mtx
);
2781 kfree(ext4_li_info
);
2782 ext4_li_info
= NULL
;
2783 mutex_unlock(&ext4_li_mtx
);
2788 static void ext4_clear_request_list(void)
2790 struct list_head
*pos
, *n
;
2791 struct ext4_li_request
*elr
;
2793 mutex_lock(&ext4_li_info
->li_list_mtx
);
2794 list_for_each_safe(pos
, n
, &ext4_li_info
->li_request_list
) {
2795 elr
= list_entry(pos
, struct ext4_li_request
,
2797 ext4_remove_li_request(elr
);
2799 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2802 static int ext4_run_lazyinit_thread(void)
2804 ext4_lazyinit_task
= kthread_run(ext4_lazyinit_thread
,
2805 ext4_li_info
, "ext4lazyinit");
2806 if (IS_ERR(ext4_lazyinit_task
)) {
2807 int err
= PTR_ERR(ext4_lazyinit_task
);
2808 ext4_clear_request_list();
2809 kfree(ext4_li_info
);
2810 ext4_li_info
= NULL
;
2811 printk(KERN_CRIT
"EXT4-fs: error %d creating inode table "
2812 "initialization thread\n",
2816 ext4_li_info
->li_state
|= EXT4_LAZYINIT_RUNNING
;
2821 * Check whether it make sense to run itable init. thread or not.
2822 * If there is at least one uninitialized inode table, return
2823 * corresponding group number, else the loop goes through all
2824 * groups and return total number of groups.
2826 static ext4_group_t
ext4_has_uninit_itable(struct super_block
*sb
)
2828 ext4_group_t group
, ngroups
= EXT4_SB(sb
)->s_groups_count
;
2829 struct ext4_group_desc
*gdp
= NULL
;
2831 for (group
= 0; group
< ngroups
; group
++) {
2832 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2836 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2843 static int ext4_li_info_new(void)
2845 struct ext4_lazy_init
*eli
= NULL
;
2847 eli
= kzalloc(sizeof(*eli
), GFP_KERNEL
);
2851 INIT_LIST_HEAD(&eli
->li_request_list
);
2852 mutex_init(&eli
->li_list_mtx
);
2854 eli
->li_state
|= EXT4_LAZYINIT_QUIT
;
2861 static struct ext4_li_request
*ext4_li_request_new(struct super_block
*sb
,
2864 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2865 struct ext4_li_request
*elr
;
2867 elr
= kzalloc(sizeof(*elr
), GFP_KERNEL
);
2873 elr
->lr_next_group
= start
;
2876 * Randomize first schedule time of the request to
2877 * spread the inode table initialization requests
2880 elr
->lr_next_sched
= jiffies
+ (prandom_u32() %
2881 (EXT4_DEF_LI_MAX_START_DELAY
* HZ
));
2885 int ext4_register_li_request(struct super_block
*sb
,
2886 ext4_group_t first_not_zeroed
)
2888 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2889 struct ext4_li_request
*elr
= NULL
;
2890 ext4_group_t ngroups
= EXT4_SB(sb
)->s_groups_count
;
2893 mutex_lock(&ext4_li_mtx
);
2894 if (sbi
->s_li_request
!= NULL
) {
2896 * Reset timeout so it can be computed again, because
2897 * s_li_wait_mult might have changed.
2899 sbi
->s_li_request
->lr_timeout
= 0;
2903 if (first_not_zeroed
== ngroups
||
2904 (sb
->s_flags
& MS_RDONLY
) ||
2905 !test_opt(sb
, INIT_INODE_TABLE
))
2908 elr
= ext4_li_request_new(sb
, first_not_zeroed
);
2914 if (NULL
== ext4_li_info
) {
2915 ret
= ext4_li_info_new();
2920 mutex_lock(&ext4_li_info
->li_list_mtx
);
2921 list_add(&elr
->lr_request
, &ext4_li_info
->li_request_list
);
2922 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2924 sbi
->s_li_request
= elr
;
2926 * set elr to NULL here since it has been inserted to
2927 * the request_list and the removal and free of it is
2928 * handled by ext4_clear_request_list from now on.
2932 if (!(ext4_li_info
->li_state
& EXT4_LAZYINIT_RUNNING
)) {
2933 ret
= ext4_run_lazyinit_thread();
2938 mutex_unlock(&ext4_li_mtx
);
2945 * We do not need to lock anything since this is called on
2948 static void ext4_destroy_lazyinit_thread(void)
2951 * If thread exited earlier
2952 * there's nothing to be done.
2954 if (!ext4_li_info
|| !ext4_lazyinit_task
)
2957 kthread_stop(ext4_lazyinit_task
);
2960 static int set_journal_csum_feature_set(struct super_block
*sb
)
2963 int compat
, incompat
;
2964 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2966 if (ext4_has_metadata_csum(sb
)) {
2967 /* journal checksum v3 */
2969 incompat
= JBD2_FEATURE_INCOMPAT_CSUM_V3
;
2971 /* journal checksum v1 */
2972 compat
= JBD2_FEATURE_COMPAT_CHECKSUM
;
2976 jbd2_journal_clear_features(sbi
->s_journal
,
2977 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2978 JBD2_FEATURE_INCOMPAT_CSUM_V3
|
2979 JBD2_FEATURE_INCOMPAT_CSUM_V2
);
2980 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
2981 ret
= jbd2_journal_set_features(sbi
->s_journal
,
2983 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
|
2985 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
2986 ret
= jbd2_journal_set_features(sbi
->s_journal
,
2989 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
2990 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2992 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
2993 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3000 * Note: calculating the overhead so we can be compatible with
3001 * historical BSD practice is quite difficult in the face of
3002 * clusters/bigalloc. This is because multiple metadata blocks from
3003 * different block group can end up in the same allocation cluster.
3004 * Calculating the exact overhead in the face of clustered allocation
3005 * requires either O(all block bitmaps) in memory or O(number of block
3006 * groups**2) in time. We will still calculate the superblock for
3007 * older file systems --- and if we come across with a bigalloc file
3008 * system with zero in s_overhead_clusters the estimate will be close to
3009 * correct especially for very large cluster sizes --- but for newer
3010 * file systems, it's better to calculate this figure once at mkfs
3011 * time, and store it in the superblock. If the superblock value is
3012 * present (even for non-bigalloc file systems), we will use it.
3014 static int count_overhead(struct super_block
*sb
, ext4_group_t grp
,
3017 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3018 struct ext4_group_desc
*gdp
;
3019 ext4_fsblk_t first_block
, last_block
, b
;
3020 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3021 int s
, j
, count
= 0;
3023 if (!ext4_has_feature_bigalloc(sb
))
3024 return (ext4_bg_has_super(sb
, grp
) + ext4_bg_num_gdb(sb
, grp
) +
3025 sbi
->s_itb_per_group
+ 2);
3027 first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
) +
3028 (grp
* EXT4_BLOCKS_PER_GROUP(sb
));
3029 last_block
= first_block
+ EXT4_BLOCKS_PER_GROUP(sb
) - 1;
3030 for (i
= 0; i
< ngroups
; i
++) {
3031 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
3032 b
= ext4_block_bitmap(sb
, gdp
);
3033 if (b
>= first_block
&& b
<= last_block
) {
3034 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3037 b
= ext4_inode_bitmap(sb
, gdp
);
3038 if (b
>= first_block
&& b
<= last_block
) {
3039 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3042 b
= ext4_inode_table(sb
, gdp
);
3043 if (b
>= first_block
&& b
+ sbi
->s_itb_per_group
<= last_block
)
3044 for (j
= 0; j
< sbi
->s_itb_per_group
; j
++, b
++) {
3045 int c
= EXT4_B2C(sbi
, b
- first_block
);
3046 ext4_set_bit(c
, buf
);
3052 if (ext4_bg_has_super(sb
, grp
)) {
3053 ext4_set_bit(s
++, buf
);
3056 for (j
= ext4_bg_num_gdb(sb
, grp
); j
> 0; j
--) {
3057 ext4_set_bit(EXT4_B2C(sbi
, s
++), buf
);
3063 return EXT4_CLUSTERS_PER_GROUP(sb
) -
3064 ext4_count_free(buf
, EXT4_CLUSTERS_PER_GROUP(sb
) / 8);
3068 * Compute the overhead and stash it in sbi->s_overhead
3070 int ext4_calculate_overhead(struct super_block
*sb
)
3072 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3073 struct ext4_super_block
*es
= sbi
->s_es
;
3074 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3075 ext4_fsblk_t overhead
= 0;
3076 char *buf
= (char *) get_zeroed_page(GFP_NOFS
);
3082 * Compute the overhead (FS structures). This is constant
3083 * for a given filesystem unless the number of block groups
3084 * changes so we cache the previous value until it does.
3088 * All of the blocks before first_data_block are overhead
3090 overhead
= EXT4_B2C(sbi
, le32_to_cpu(es
->s_first_data_block
));
3093 * Add the overhead found in each block group
3095 for (i
= 0; i
< ngroups
; i
++) {
3098 blks
= count_overhead(sb
, i
, buf
);
3101 memset(buf
, 0, PAGE_SIZE
);
3104 /* Add the internal journal blocks as well */
3105 if (sbi
->s_journal
&& !sbi
->journal_bdev
)
3106 overhead
+= EXT4_NUM_B2C(sbi
, sbi
->s_journal
->j_maxlen
);
3108 sbi
->s_overhead
= overhead
;
3110 free_page((unsigned long) buf
);
3114 static void ext4_set_resv_clusters(struct super_block
*sb
)
3116 ext4_fsblk_t resv_clusters
;
3117 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3120 * There's no need to reserve anything when we aren't using extents.
3121 * The space estimates are exact, there are no unwritten extents,
3122 * hole punching doesn't need new metadata... This is needed especially
3123 * to keep ext2/3 backward compatibility.
3125 if (!ext4_has_feature_extents(sb
))
3128 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3129 * This should cover the situations where we can not afford to run
3130 * out of space like for example punch hole, or converting
3131 * unwritten extents in delalloc path. In most cases such
3132 * allocation would require 1, or 2 blocks, higher numbers are
3135 resv_clusters
= (ext4_blocks_count(sbi
->s_es
) >>
3136 sbi
->s_cluster_bits
);
3138 do_div(resv_clusters
, 50);
3139 resv_clusters
= min_t(ext4_fsblk_t
, resv_clusters
, 4096);
3141 atomic64_set(&sbi
->s_resv_clusters
, resv_clusters
);
3144 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
3146 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
3147 struct buffer_head
*bh
;
3148 struct ext4_super_block
*es
= NULL
;
3149 struct ext4_sb_info
*sbi
;
3151 ext4_fsblk_t sb_block
= get_sb_block(&data
);
3152 ext4_fsblk_t logical_sb_block
;
3153 unsigned long offset
= 0;
3154 unsigned long journal_devnum
= 0;
3155 unsigned long def_mount_opts
;
3159 int blocksize
, clustersize
;
3160 unsigned int db_count
;
3162 int needs_recovery
, has_huge_files
, has_bigalloc
;
3165 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
3166 ext4_group_t first_not_zeroed
;
3168 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
3172 sbi
->s_blockgroup_lock
=
3173 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
3174 if (!sbi
->s_blockgroup_lock
) {
3178 sb
->s_fs_info
= sbi
;
3180 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
3181 sbi
->s_sb_block
= sb_block
;
3182 if (sb
->s_bdev
->bd_part
)
3183 sbi
->s_sectors_written_start
=
3184 part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]);
3186 /* Cleanup superblock name */
3187 strreplace(sb
->s_id
, '/', '!');
3189 /* -EINVAL is default */
3191 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
3193 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
3198 * The ext4 superblock will not be buffer aligned for other than 1kB
3199 * block sizes. We need to calculate the offset from buffer start.
3201 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
3202 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3203 offset
= do_div(logical_sb_block
, blocksize
);
3205 logical_sb_block
= sb_block
;
3208 if (!(bh
= sb_bread_unmovable(sb
, logical_sb_block
))) {
3209 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
3213 * Note: s_es must be initialized as soon as possible because
3214 * some ext4 macro-instructions depend on its value
3216 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
3218 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
3219 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
3221 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
3223 /* Warn if metadata_csum and gdt_csum are both set. */
3224 if (ext4_has_feature_metadata_csum(sb
) &&
3225 ext4_has_feature_gdt_csum(sb
))
3226 ext4_warning(sb
, "metadata_csum and uninit_bg are "
3227 "redundant flags; please run fsck.");
3229 /* Check for a known checksum algorithm */
3230 if (!ext4_verify_csum_type(sb
, es
)) {
3231 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3232 "unknown checksum algorithm.");
3237 /* Load the checksum driver */
3238 if (ext4_has_feature_metadata_csum(sb
)) {
3239 sbi
->s_chksum_driver
= crypto_alloc_shash("crc32c", 0, 0);
3240 if (IS_ERR(sbi
->s_chksum_driver
)) {
3241 ext4_msg(sb
, KERN_ERR
, "Cannot load crc32c driver.");
3242 ret
= PTR_ERR(sbi
->s_chksum_driver
);
3243 sbi
->s_chksum_driver
= NULL
;
3248 /* Check superblock checksum */
3249 if (!ext4_superblock_csum_verify(sb
, es
)) {
3250 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3251 "invalid superblock checksum. Run e2fsck?");
3257 /* Precompute checksum seed for all metadata */
3258 if (ext4_has_feature_csum_seed(sb
))
3259 sbi
->s_csum_seed
= le32_to_cpu(es
->s_checksum_seed
);
3260 else if (ext4_has_metadata_csum(sb
))
3261 sbi
->s_csum_seed
= ext4_chksum(sbi
, ~0, es
->s_uuid
,
3262 sizeof(es
->s_uuid
));
3264 /* Set defaults before we parse the mount options */
3265 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
3266 set_opt(sb
, INIT_INODE_TABLE
);
3267 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
3269 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
)
3271 if (def_mount_opts
& EXT4_DEFM_UID16
)
3272 set_opt(sb
, NO_UID32
);
3273 /* xattr user namespace & acls are now defaulted on */
3274 set_opt(sb
, XATTR_USER
);
3275 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3276 set_opt(sb
, POSIX_ACL
);
3278 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3279 if (ext4_has_metadata_csum(sb
))
3280 set_opt(sb
, JOURNAL_CHECKSUM
);
3282 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
3283 set_opt(sb
, JOURNAL_DATA
);
3284 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
3285 set_opt(sb
, ORDERED_DATA
);
3286 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
3287 set_opt(sb
, WRITEBACK_DATA
);
3289 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
3290 set_opt(sb
, ERRORS_PANIC
);
3291 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
3292 set_opt(sb
, ERRORS_CONT
);
3294 set_opt(sb
, ERRORS_RO
);
3295 /* block_validity enabled by default; disable with noblock_validity */
3296 set_opt(sb
, BLOCK_VALIDITY
);
3297 if (def_mount_opts
& EXT4_DEFM_DISCARD
)
3298 set_opt(sb
, DISCARD
);
3300 sbi
->s_resuid
= make_kuid(&init_user_ns
, le16_to_cpu(es
->s_def_resuid
));
3301 sbi
->s_resgid
= make_kgid(&init_user_ns
, le16_to_cpu(es
->s_def_resgid
));
3302 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
3303 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
3304 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
3306 if ((def_mount_opts
& EXT4_DEFM_NOBARRIER
) == 0)
3307 set_opt(sb
, BARRIER
);
3310 * enable delayed allocation by default
3311 * Use -o nodelalloc to turn it off
3313 if (!IS_EXT3_SB(sb
) && !IS_EXT2_SB(sb
) &&
3314 ((def_mount_opts
& EXT4_DEFM_NODELALLOC
) == 0))
3315 set_opt(sb
, DELALLOC
);
3318 * set default s_li_wait_mult for lazyinit, for the case there is
3319 * no mount option specified.
3321 sbi
->s_li_wait_mult
= EXT4_DEF_LI_WAIT_MULT
;
3323 if (!parse_options((char *) sbi
->s_es
->s_mount_opts
, sb
,
3324 &journal_devnum
, &journal_ioprio
, 0)) {
3325 ext4_msg(sb
, KERN_WARNING
,
3326 "failed to parse options in superblock: %s",
3327 sbi
->s_es
->s_mount_opts
);
3329 sbi
->s_def_mount_opt
= sbi
->s_mount_opt
;
3330 if (!parse_options((char *) data
, sb
, &journal_devnum
,
3331 &journal_ioprio
, 0))
3334 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
3335 printk_once(KERN_WARNING
"EXT4-fs: Warning: mounting "
3336 "with data=journal disables delayed "
3337 "allocation and O_DIRECT support!\n");
3338 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
3339 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3340 "both data=journal and delalloc");
3343 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
3344 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3345 "both data=journal and dioread_nolock");
3348 if (test_opt(sb
, DAX
)) {
3349 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3350 "both data=journal and dax");
3353 if (test_opt(sb
, DELALLOC
))
3354 clear_opt(sb
, DELALLOC
);
3356 sb
->s_iflags
|= SB_I_CGROUPWB
;
3359 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3360 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
3362 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
3363 (ext4_has_compat_features(sb
) ||
3364 ext4_has_ro_compat_features(sb
) ||
3365 ext4_has_incompat_features(sb
)))
3366 ext4_msg(sb
, KERN_WARNING
,
3367 "feature flags set on rev 0 fs, "
3368 "running e2fsck is recommended");
3370 if (es
->s_creator_os
== cpu_to_le32(EXT4_OS_HURD
)) {
3371 set_opt2(sb
, HURD_COMPAT
);
3372 if (ext4_has_feature_64bit(sb
)) {
3373 ext4_msg(sb
, KERN_ERR
,
3374 "The Hurd can't support 64-bit file systems");
3379 if (IS_EXT2_SB(sb
)) {
3380 if (ext2_feature_set_ok(sb
))
3381 ext4_msg(sb
, KERN_INFO
, "mounting ext2 file system "
3382 "using the ext4 subsystem");
3384 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext2 due "
3385 "to feature incompatibilities");
3390 if (IS_EXT3_SB(sb
)) {
3391 if (ext3_feature_set_ok(sb
))
3392 ext4_msg(sb
, KERN_INFO
, "mounting ext3 file system "
3393 "using the ext4 subsystem");
3395 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext3 due "
3396 "to feature incompatibilities");
3402 * Check feature flags regardless of the revision level, since we
3403 * previously didn't change the revision level when setting the flags,
3404 * so there is a chance incompat flags are set on a rev 0 filesystem.
3406 if (!ext4_feature_set_ok(sb
, (sb
->s_flags
& MS_RDONLY
)))
3409 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
3410 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
3411 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
3412 ext4_msg(sb
, KERN_ERR
,
3413 "Unsupported filesystem blocksize %d", blocksize
);
3417 if (sbi
->s_mount_opt
& EXT4_MOUNT_DAX
) {
3418 if (blocksize
!= PAGE_SIZE
) {
3419 ext4_msg(sb
, KERN_ERR
,
3420 "error: unsupported blocksize for dax");
3423 if (!sb
->s_bdev
->bd_disk
->fops
->direct_access
) {
3424 ext4_msg(sb
, KERN_ERR
,
3425 "error: device does not support dax");
3430 if (ext4_has_feature_encrypt(sb
) && es
->s_encryption_level
) {
3431 ext4_msg(sb
, KERN_ERR
, "Unsupported encryption level %d",
3432 es
->s_encryption_level
);
3436 if (sb
->s_blocksize
!= blocksize
) {
3437 /* Validate the filesystem blocksize */
3438 if (!sb_set_blocksize(sb
, blocksize
)) {
3439 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
3445 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3446 offset
= do_div(logical_sb_block
, blocksize
);
3447 bh
= sb_bread_unmovable(sb
, logical_sb_block
);
3449 ext4_msg(sb
, KERN_ERR
,
3450 "Can't read superblock on 2nd try");
3453 es
= (struct ext4_super_block
*)(bh
->b_data
+ offset
);
3455 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
3456 ext4_msg(sb
, KERN_ERR
,
3457 "Magic mismatch, very weird!");
3462 has_huge_files
= ext4_has_feature_huge_file(sb
);
3463 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
3465 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
3467 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
3468 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
3469 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
3471 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
3472 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
3473 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
3474 (!is_power_of_2(sbi
->s_inode_size
)) ||
3475 (sbi
->s_inode_size
> blocksize
)) {
3476 ext4_msg(sb
, KERN_ERR
,
3477 "unsupported inode size: %d",
3481 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
3482 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
3485 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
3486 if (ext4_has_feature_64bit(sb
)) {
3487 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
3488 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
3489 !is_power_of_2(sbi
->s_desc_size
)) {
3490 ext4_msg(sb
, KERN_ERR
,
3491 "unsupported descriptor size %lu",
3496 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
3498 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
3499 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
3500 if (EXT4_INODE_SIZE(sb
) == 0 || EXT4_INODES_PER_GROUP(sb
) == 0)
3503 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
3504 if (sbi
->s_inodes_per_block
== 0)
3506 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
3507 sbi
->s_inodes_per_block
;
3508 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
3510 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3511 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
3512 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
3514 for (i
= 0; i
< 4; i
++)
3515 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
3516 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
3517 if (ext4_has_feature_dir_index(sb
)) {
3518 i
= le32_to_cpu(es
->s_flags
);
3519 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
3520 sbi
->s_hash_unsigned
= 3;
3521 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
3522 #ifdef __CHAR_UNSIGNED__
3523 if (!(sb
->s_flags
& MS_RDONLY
))
3525 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
3526 sbi
->s_hash_unsigned
= 3;
3528 if (!(sb
->s_flags
& MS_RDONLY
))
3530 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
3535 /* Handle clustersize */
3536 clustersize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_cluster_size
);
3537 has_bigalloc
= ext4_has_feature_bigalloc(sb
);
3539 if (clustersize
< blocksize
) {
3540 ext4_msg(sb
, KERN_ERR
,
3541 "cluster size (%d) smaller than "
3542 "block size (%d)", clustersize
, blocksize
);
3545 sbi
->s_cluster_bits
= le32_to_cpu(es
->s_log_cluster_size
) -
3546 le32_to_cpu(es
->s_log_block_size
);
3547 sbi
->s_clusters_per_group
=
3548 le32_to_cpu(es
->s_clusters_per_group
);
3549 if (sbi
->s_clusters_per_group
> blocksize
* 8) {
3550 ext4_msg(sb
, KERN_ERR
,
3551 "#clusters per group too big: %lu",
3552 sbi
->s_clusters_per_group
);
3555 if (sbi
->s_blocks_per_group
!=
3556 (sbi
->s_clusters_per_group
* (clustersize
/ blocksize
))) {
3557 ext4_msg(sb
, KERN_ERR
, "blocks per group (%lu) and "
3558 "clusters per group (%lu) inconsistent",
3559 sbi
->s_blocks_per_group
,
3560 sbi
->s_clusters_per_group
);
3564 if (clustersize
!= blocksize
) {
3565 ext4_warning(sb
, "fragment/cluster size (%d) != "
3566 "block size (%d)", clustersize
,
3568 clustersize
= blocksize
;
3570 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
3571 ext4_msg(sb
, KERN_ERR
,
3572 "#blocks per group too big: %lu",
3573 sbi
->s_blocks_per_group
);
3576 sbi
->s_clusters_per_group
= sbi
->s_blocks_per_group
;
3577 sbi
->s_cluster_bits
= 0;
3579 sbi
->s_cluster_ratio
= clustersize
/ blocksize
;
3581 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
3582 ext4_msg(sb
, KERN_ERR
,
3583 "#inodes per group too big: %lu",
3584 sbi
->s_inodes_per_group
);
3588 /* Do we have standard group size of clustersize * 8 blocks ? */
3589 if (sbi
->s_blocks_per_group
== clustersize
<< 3)
3590 set_opt2(sb
, STD_GROUP_SIZE
);
3593 * Test whether we have more sectors than will fit in sector_t,
3594 * and whether the max offset is addressable by the page cache.
3596 err
= generic_check_addressable(sb
->s_blocksize_bits
,
3597 ext4_blocks_count(es
));
3599 ext4_msg(sb
, KERN_ERR
, "filesystem"
3600 " too large to mount safely on this system");
3601 if (sizeof(sector_t
) < 8)
3602 ext4_msg(sb
, KERN_WARNING
, "CONFIG_LBDAF not enabled");
3606 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
3609 /* check blocks count against device size */
3610 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
3611 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
3612 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
3613 "exceeds size of device (%llu blocks)",
3614 ext4_blocks_count(es
), blocks_count
);
3619 * It makes no sense for the first data block to be beyond the end
3620 * of the filesystem.
3622 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
3623 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data "
3624 "block %u is beyond end of filesystem (%llu)",
3625 le32_to_cpu(es
->s_first_data_block
),
3626 ext4_blocks_count(es
));
3629 blocks_count
= (ext4_blocks_count(es
) -
3630 le32_to_cpu(es
->s_first_data_block
) +
3631 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
3632 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
3633 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
3634 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %u "
3635 "(block count %llu, first data block %u, "
3636 "blocks per group %lu)", sbi
->s_groups_count
,
3637 ext4_blocks_count(es
),
3638 le32_to_cpu(es
->s_first_data_block
),
3639 EXT4_BLOCKS_PER_GROUP(sb
));
3642 sbi
->s_groups_count
= blocks_count
;
3643 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
3644 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
3645 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
3646 EXT4_DESC_PER_BLOCK(sb
);
3647 sbi
->s_group_desc
= ext4_kvmalloc(db_count
*
3648 sizeof(struct buffer_head
*),
3650 if (sbi
->s_group_desc
== NULL
) {
3651 ext4_msg(sb
, KERN_ERR
, "not enough memory");
3656 bgl_lock_init(sbi
->s_blockgroup_lock
);
3658 for (i
= 0; i
< db_count
; i
++) {
3659 block
= descriptor_loc(sb
, logical_sb_block
, i
);
3660 sbi
->s_group_desc
[i
] = sb_bread_unmovable(sb
, block
);
3661 if (!sbi
->s_group_desc
[i
]) {
3662 ext4_msg(sb
, KERN_ERR
,
3663 "can't read group descriptor %d", i
);
3668 if (!ext4_check_descriptors(sb
, &first_not_zeroed
)) {
3669 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
3670 ret
= -EFSCORRUPTED
;
3674 sbi
->s_gdb_count
= db_count
;
3675 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
3676 spin_lock_init(&sbi
->s_next_gen_lock
);
3678 setup_timer(&sbi
->s_err_report
, print_daily_error_info
,
3679 (unsigned long) sb
);
3681 /* Register extent status tree shrinker */
3682 if (ext4_es_register_shrinker(sbi
))
3685 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
3686 sbi
->s_extent_max_zeroout_kb
= 32;
3689 * set up enough so that it can read an inode
3691 sb
->s_op
= &ext4_sops
;
3692 sb
->s_export_op
= &ext4_export_ops
;
3693 sb
->s_xattr
= ext4_xattr_handlers
;
3695 sb
->dq_op
= &ext4_quota_operations
;
3696 if (ext4_has_feature_quota(sb
))
3697 sb
->s_qcop
= &dquot_quotactl_sysfile_ops
;
3699 sb
->s_qcop
= &ext4_qctl_operations
;
3700 sb
->s_quota_types
= QTYPE_MASK_USR
| QTYPE_MASK_GRP
| QTYPE_MASK_PRJ
;
3702 memcpy(sb
->s_uuid
, es
->s_uuid
, sizeof(es
->s_uuid
));
3704 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
3705 mutex_init(&sbi
->s_orphan_lock
);
3709 needs_recovery
= (es
->s_last_orphan
!= 0 ||
3710 ext4_has_feature_journal_needs_recovery(sb
));
3712 if (ext4_has_feature_mmp(sb
) && !(sb
->s_flags
& MS_RDONLY
))
3713 if (ext4_multi_mount_protect(sb
, le64_to_cpu(es
->s_mmp_block
)))
3714 goto failed_mount3a
;
3717 * The first inode we look at is the journal inode. Don't try
3718 * root first: it may be modified in the journal!
3720 if (!test_opt(sb
, NOLOAD
) && ext4_has_feature_journal(sb
)) {
3721 if (ext4_load_journal(sb
, es
, journal_devnum
))
3722 goto failed_mount3a
;
3723 } else if (test_opt(sb
, NOLOAD
) && !(sb
->s_flags
& MS_RDONLY
) &&
3724 ext4_has_feature_journal_needs_recovery(sb
)) {
3725 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
3726 "suppressed and not mounted read-only");
3727 goto failed_mount_wq
;
3729 /* Nojournal mode, all journal mount options are illegal */
3730 if (test_opt2(sb
, EXPLICIT_JOURNAL_CHECKSUM
)) {
3731 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3732 "journal_checksum, fs mounted w/o journal");
3733 goto failed_mount_wq
;
3735 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
3736 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3737 "journal_async_commit, fs mounted w/o journal");
3738 goto failed_mount_wq
;
3740 if (sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
) {
3741 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3742 "commit=%lu, fs mounted w/o journal",
3743 sbi
->s_commit_interval
/ HZ
);
3744 goto failed_mount_wq
;
3746 if (EXT4_MOUNT_DATA_FLAGS
&
3747 (sbi
->s_mount_opt
^ sbi
->s_def_mount_opt
)) {
3748 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3749 "data=, fs mounted w/o journal");
3750 goto failed_mount_wq
;
3752 sbi
->s_def_mount_opt
&= EXT4_MOUNT_JOURNAL_CHECKSUM
;
3753 clear_opt(sb
, JOURNAL_CHECKSUM
);
3754 clear_opt(sb
, DATA_FLAGS
);
3755 sbi
->s_journal
= NULL
;
3760 if (ext4_has_feature_64bit(sb
) &&
3761 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
3762 JBD2_FEATURE_INCOMPAT_64BIT
)) {
3763 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
3764 goto failed_mount_wq
;
3767 if (!set_journal_csum_feature_set(sb
)) {
3768 ext4_msg(sb
, KERN_ERR
, "Failed to set journal checksum "
3770 goto failed_mount_wq
;
3773 /* We have now updated the journal if required, so we can
3774 * validate the data journaling mode. */
3775 switch (test_opt(sb
, DATA_FLAGS
)) {
3777 /* No mode set, assume a default based on the journal
3778 * capabilities: ORDERED_DATA if the journal can
3779 * cope, else JOURNAL_DATA
3781 if (jbd2_journal_check_available_features
3782 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
3783 set_opt(sb
, ORDERED_DATA
);
3785 set_opt(sb
, JOURNAL_DATA
);
3788 case EXT4_MOUNT_ORDERED_DATA
:
3789 case EXT4_MOUNT_WRITEBACK_DATA
:
3790 if (!jbd2_journal_check_available_features
3791 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
3792 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
3793 "requested data journaling mode");
3794 goto failed_mount_wq
;
3799 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
3801 sbi
->s_journal
->j_commit_callback
= ext4_journal_commit_callback
;
3804 sbi
->s_mb_cache
= ext4_xattr_create_cache();
3805 if (!sbi
->s_mb_cache
) {
3806 ext4_msg(sb
, KERN_ERR
, "Failed to create an mb_cache");
3807 goto failed_mount_wq
;
3810 if ((DUMMY_ENCRYPTION_ENABLED(sbi
) || ext4_has_feature_encrypt(sb
)) &&
3811 (blocksize
!= PAGE_SIZE
)) {
3812 ext4_msg(sb
, KERN_ERR
,
3813 "Unsupported blocksize for fs encryption");
3814 goto failed_mount_wq
;
3817 if (DUMMY_ENCRYPTION_ENABLED(sbi
) && !(sb
->s_flags
& MS_RDONLY
) &&
3818 !ext4_has_feature_encrypt(sb
)) {
3819 ext4_set_feature_encrypt(sb
);
3820 ext4_commit_super(sb
, 1);
3824 * Get the # of file system overhead blocks from the
3825 * superblock if present.
3827 if (es
->s_overhead_clusters
)
3828 sbi
->s_overhead
= le32_to_cpu(es
->s_overhead_clusters
);
3830 err
= ext4_calculate_overhead(sb
);
3832 goto failed_mount_wq
;
3836 * The maximum number of concurrent works can be high and
3837 * concurrency isn't really necessary. Limit it to 1.
3839 EXT4_SB(sb
)->rsv_conversion_wq
=
3840 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM
| WQ_UNBOUND
, 1);
3841 if (!EXT4_SB(sb
)->rsv_conversion_wq
) {
3842 printk(KERN_ERR
"EXT4-fs: failed to create workqueue\n");
3848 * The jbd2_journal_load will have done any necessary log recovery,
3849 * so we can safely mount the rest of the filesystem now.
3852 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
3854 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
3855 ret
= PTR_ERR(root
);
3859 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
3860 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
3864 sb
->s_root
= d_make_root(root
);
3866 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
3871 if (ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
))
3872 sb
->s_flags
|= MS_RDONLY
;
3874 /* determine the minimum size of new large inodes, if present */
3875 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
3876 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
3877 EXT4_GOOD_OLD_INODE_SIZE
;
3878 if (ext4_has_feature_extra_isize(sb
)) {
3879 if (sbi
->s_want_extra_isize
<
3880 le16_to_cpu(es
->s_want_extra_isize
))
3881 sbi
->s_want_extra_isize
=
3882 le16_to_cpu(es
->s_want_extra_isize
);
3883 if (sbi
->s_want_extra_isize
<
3884 le16_to_cpu(es
->s_min_extra_isize
))
3885 sbi
->s_want_extra_isize
=
3886 le16_to_cpu(es
->s_min_extra_isize
);
3889 /* Check if enough inode space is available */
3890 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
3891 sbi
->s_inode_size
) {
3892 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
3893 EXT4_GOOD_OLD_INODE_SIZE
;
3894 ext4_msg(sb
, KERN_INFO
, "required extra inode space not"
3898 ext4_set_resv_clusters(sb
);
3900 err
= ext4_setup_system_zone(sb
);
3902 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
3904 goto failed_mount4a
;
3908 err
= ext4_mb_init(sb
);
3910 ext4_msg(sb
, KERN_ERR
, "failed to initialize mballoc (%d)",
3915 block
= ext4_count_free_clusters(sb
);
3916 ext4_free_blocks_count_set(sbi
->s_es
,
3917 EXT4_C2B(sbi
, block
));
3918 err
= percpu_counter_init(&sbi
->s_freeclusters_counter
, block
,
3921 unsigned long freei
= ext4_count_free_inodes(sb
);
3922 sbi
->s_es
->s_free_inodes_count
= cpu_to_le32(freei
);
3923 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
, freei
,
3927 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
3928 ext4_count_dirs(sb
), GFP_KERNEL
);
3930 err
= percpu_counter_init(&sbi
->s_dirtyclusters_counter
, 0,
3933 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
3937 if (ext4_has_feature_flex_bg(sb
))
3938 if (!ext4_fill_flex_info(sb
)) {
3939 ext4_msg(sb
, KERN_ERR
,
3940 "unable to initialize "
3941 "flex_bg meta info!");
3945 err
= ext4_register_li_request(sb
, first_not_zeroed
);
3949 err
= ext4_register_sysfs(sb
);
3954 /* Enable quota usage during mount. */
3955 if (ext4_has_feature_quota(sb
) && !(sb
->s_flags
& MS_RDONLY
)) {
3956 err
= ext4_enable_quotas(sb
);
3960 #endif /* CONFIG_QUOTA */
3962 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
3963 ext4_orphan_cleanup(sb
, es
);
3964 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
3965 if (needs_recovery
) {
3966 ext4_msg(sb
, KERN_INFO
, "recovery complete");
3967 ext4_mark_recovery_complete(sb
, es
);
3969 if (EXT4_SB(sb
)->s_journal
) {
3970 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
3971 descr
= " journalled data mode";
3972 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
3973 descr
= " ordered data mode";
3975 descr
= " writeback data mode";
3977 descr
= "out journal";
3979 if (test_opt(sb
, DISCARD
)) {
3980 struct request_queue
*q
= bdev_get_queue(sb
->s_bdev
);
3981 if (!blk_queue_discard(q
))
3982 ext4_msg(sb
, KERN_WARNING
,
3983 "mounting with \"discard\" option, but "
3984 "the device does not support discard");
3987 if (___ratelimit(&ext4_mount_msg_ratelimit
, "EXT4-fs mount"))
3988 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s. "
3989 "Opts: %s%s%s", descr
, sbi
->s_es
->s_mount_opts
,
3990 *sbi
->s_es
->s_mount_opts
? "; " : "", orig_data
);
3992 if (es
->s_error_count
)
3993 mod_timer(&sbi
->s_err_report
, jiffies
+ 300*HZ
); /* 5 minutes */
3995 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
3996 ratelimit_state_init(&sbi
->s_err_ratelimit_state
, 5 * HZ
, 10);
3997 ratelimit_state_init(&sbi
->s_warning_ratelimit_state
, 5 * HZ
, 10);
3998 ratelimit_state_init(&sbi
->s_msg_ratelimit_state
, 5 * HZ
, 10);
4005 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
4010 ext4_unregister_sysfs(sb
);
4013 ext4_unregister_li_request(sb
);
4015 ext4_mb_release(sb
);
4016 if (sbi
->s_flex_groups
)
4017 kvfree(sbi
->s_flex_groups
);
4018 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
4019 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
4020 percpu_counter_destroy(&sbi
->s_dirs_counter
);
4021 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
4023 ext4_ext_release(sb
);
4024 ext4_release_system_zone(sb
);
4029 ext4_msg(sb
, KERN_ERR
, "mount failed");
4030 if (EXT4_SB(sb
)->rsv_conversion_wq
)
4031 destroy_workqueue(EXT4_SB(sb
)->rsv_conversion_wq
);
4033 if (sbi
->s_mb_cache
) {
4034 ext4_xattr_destroy_cache(sbi
->s_mb_cache
);
4035 sbi
->s_mb_cache
= NULL
;
4037 if (sbi
->s_journal
) {
4038 jbd2_journal_destroy(sbi
->s_journal
);
4039 sbi
->s_journal
= NULL
;
4042 ext4_es_unregister_shrinker(sbi
);
4044 del_timer_sync(&sbi
->s_err_report
);
4046 kthread_stop(sbi
->s_mmp_tsk
);
4048 for (i
= 0; i
< db_count
; i
++)
4049 brelse(sbi
->s_group_desc
[i
]);
4050 kvfree(sbi
->s_group_desc
);
4052 if (sbi
->s_chksum_driver
)
4053 crypto_free_shash(sbi
->s_chksum_driver
);
4055 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
4056 kfree(sbi
->s_qf_names
[i
]);
4058 ext4_blkdev_remove(sbi
);
4061 sb
->s_fs_info
= NULL
;
4062 kfree(sbi
->s_blockgroup_lock
);
4066 return err
? err
: ret
;
4070 * Setup any per-fs journal parameters now. We'll do this both on
4071 * initial mount, once the journal has been initialised but before we've
4072 * done any recovery; and again on any subsequent remount.
4074 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
4076 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4078 journal
->j_commit_interval
= sbi
->s_commit_interval
;
4079 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
4080 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
4082 write_lock(&journal
->j_state_lock
);
4083 if (test_opt(sb
, BARRIER
))
4084 journal
->j_flags
|= JBD2_BARRIER
;
4086 journal
->j_flags
&= ~JBD2_BARRIER
;
4087 if (test_opt(sb
, DATA_ERR_ABORT
))
4088 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
4090 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
4091 write_unlock(&journal
->j_state_lock
);
4094 static journal_t
*ext4_get_journal(struct super_block
*sb
,
4095 unsigned int journal_inum
)
4097 struct inode
*journal_inode
;
4100 BUG_ON(!ext4_has_feature_journal(sb
));
4102 /* First, test for the existence of a valid inode on disk. Bad
4103 * things happen if we iget() an unused inode, as the subsequent
4104 * iput() will try to delete it. */
4106 journal_inode
= ext4_iget(sb
, journal_inum
);
4107 if (IS_ERR(journal_inode
)) {
4108 ext4_msg(sb
, KERN_ERR
, "no journal found");
4111 if (!journal_inode
->i_nlink
) {
4112 make_bad_inode(journal_inode
);
4113 iput(journal_inode
);
4114 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
4118 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4119 journal_inode
, journal_inode
->i_size
);
4120 if (!S_ISREG(journal_inode
->i_mode
)) {
4121 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
4122 iput(journal_inode
);
4126 journal
= jbd2_journal_init_inode(journal_inode
);
4128 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
4129 iput(journal_inode
);
4132 journal
->j_private
= sb
;
4133 ext4_init_journal_params(sb
, journal
);
4137 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
4140 struct buffer_head
*bh
;
4144 int hblock
, blocksize
;
4145 ext4_fsblk_t sb_block
;
4146 unsigned long offset
;
4147 struct ext4_super_block
*es
;
4148 struct block_device
*bdev
;
4150 BUG_ON(!ext4_has_feature_journal(sb
));
4152 bdev
= ext4_blkdev_get(j_dev
, sb
);
4156 blocksize
= sb
->s_blocksize
;
4157 hblock
= bdev_logical_block_size(bdev
);
4158 if (blocksize
< hblock
) {
4159 ext4_msg(sb
, KERN_ERR
,
4160 "blocksize too small for journal device");
4164 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
4165 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
4166 set_blocksize(bdev
, blocksize
);
4167 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
4168 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
4169 "external journal");
4173 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
4174 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
4175 !(le32_to_cpu(es
->s_feature_incompat
) &
4176 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
4177 ext4_msg(sb
, KERN_ERR
, "external journal has "
4183 if ((le32_to_cpu(es
->s_feature_ro_compat
) &
4184 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
) &&
4185 es
->s_checksum
!= ext4_superblock_csum(sb
, es
)) {
4186 ext4_msg(sb
, KERN_ERR
, "external journal has "
4187 "corrupt superblock");
4192 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
4193 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
4198 len
= ext4_blocks_count(es
);
4199 start
= sb_block
+ 1;
4200 brelse(bh
); /* we're done with the superblock */
4202 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
4203 start
, len
, blocksize
);
4205 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
4208 journal
->j_private
= sb
;
4209 ll_rw_block(READ
| REQ_META
| REQ_PRIO
, 1, &journal
->j_sb_buffer
);
4210 wait_on_buffer(journal
->j_sb_buffer
);
4211 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
4212 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
4215 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
4216 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
4217 "user (unsupported) - %d",
4218 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
4221 EXT4_SB(sb
)->journal_bdev
= bdev
;
4222 ext4_init_journal_params(sb
, journal
);
4226 jbd2_journal_destroy(journal
);
4228 ext4_blkdev_put(bdev
);
4232 static int ext4_load_journal(struct super_block
*sb
,
4233 struct ext4_super_block
*es
,
4234 unsigned long journal_devnum
)
4237 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
4240 int really_read_only
;
4242 BUG_ON(!ext4_has_feature_journal(sb
));
4244 if (journal_devnum
&&
4245 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4246 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
4247 "numbers have changed");
4248 journal_dev
= new_decode_dev(journal_devnum
);
4250 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
4252 really_read_only
= bdev_read_only(sb
->s_bdev
);
4255 * Are we loading a blank journal or performing recovery after a
4256 * crash? For recovery, we need to check in advance whether we
4257 * can get read-write access to the device.
4259 if (ext4_has_feature_journal_needs_recovery(sb
)) {
4260 if (sb
->s_flags
& MS_RDONLY
) {
4261 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
4262 "required on readonly filesystem");
4263 if (really_read_only
) {
4264 ext4_msg(sb
, KERN_ERR
, "write access "
4265 "unavailable, cannot proceed");
4268 ext4_msg(sb
, KERN_INFO
, "write access will "
4269 "be enabled during recovery");
4273 if (journal_inum
&& journal_dev
) {
4274 ext4_msg(sb
, KERN_ERR
, "filesystem has both journal "
4275 "and inode journals!");
4280 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
4283 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
4287 if (!(journal
->j_flags
& JBD2_BARRIER
))
4288 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
4290 if (!ext4_has_feature_journal_needs_recovery(sb
))
4291 err
= jbd2_journal_wipe(journal
, !really_read_only
);
4293 char *save
= kmalloc(EXT4_S_ERR_LEN
, GFP_KERNEL
);
4295 memcpy(save
, ((char *) es
) +
4296 EXT4_S_ERR_START
, EXT4_S_ERR_LEN
);
4297 err
= jbd2_journal_load(journal
);
4299 memcpy(((char *) es
) + EXT4_S_ERR_START
,
4300 save
, EXT4_S_ERR_LEN
);
4305 ext4_msg(sb
, KERN_ERR
, "error loading journal");
4306 jbd2_journal_destroy(journal
);
4310 EXT4_SB(sb
)->s_journal
= journal
;
4311 ext4_clear_journal_err(sb
, es
);
4313 if (!really_read_only
&& journal_devnum
&&
4314 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4315 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
4317 /* Make sure we flush the recovery flag to disk. */
4318 ext4_commit_super(sb
, 1);
4324 static int ext4_commit_super(struct super_block
*sb
, int sync
)
4326 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
4327 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
4330 if (!sbh
|| block_device_ejected(sb
))
4332 if (buffer_write_io_error(sbh
)) {
4334 * Oh, dear. A previous attempt to write the
4335 * superblock failed. This could happen because the
4336 * USB device was yanked out. Or it could happen to
4337 * be a transient write error and maybe the block will
4338 * be remapped. Nothing we can do but to retry the
4339 * write and hope for the best.
4341 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
4342 "superblock detected");
4343 clear_buffer_write_io_error(sbh
);
4344 set_buffer_uptodate(sbh
);
4347 * If the file system is mounted read-only, don't update the
4348 * superblock write time. This avoids updating the superblock
4349 * write time when we are mounting the root file system
4350 * read/only but we need to replay the journal; at that point,
4351 * for people who are east of GMT and who make their clock
4352 * tick in localtime for Windows bug-for-bug compatibility,
4353 * the clock is set in the future, and this will cause e2fsck
4354 * to complain and force a full file system check.
4356 if (!(sb
->s_flags
& MS_RDONLY
))
4357 es
->s_wtime
= cpu_to_le32(get_seconds());
4358 if (sb
->s_bdev
->bd_part
)
4359 es
->s_kbytes_written
=
4360 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
+
4361 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
4362 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
4364 es
->s_kbytes_written
=
4365 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
);
4366 if (percpu_counter_initialized(&EXT4_SB(sb
)->s_freeclusters_counter
))
4367 ext4_free_blocks_count_set(es
,
4368 EXT4_C2B(EXT4_SB(sb
), percpu_counter_sum_positive(
4369 &EXT4_SB(sb
)->s_freeclusters_counter
)));
4370 if (percpu_counter_initialized(&EXT4_SB(sb
)->s_freeinodes_counter
))
4371 es
->s_free_inodes_count
=
4372 cpu_to_le32(percpu_counter_sum_positive(
4373 &EXT4_SB(sb
)->s_freeinodes_counter
));
4374 BUFFER_TRACE(sbh
, "marking dirty");
4375 ext4_superblock_csum_set(sb
);
4376 mark_buffer_dirty(sbh
);
4378 error
= __sync_dirty_buffer(sbh
,
4379 test_opt(sb
, BARRIER
) ? WRITE_FUA
: WRITE_SYNC
);
4383 error
= buffer_write_io_error(sbh
);
4385 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
4387 clear_buffer_write_io_error(sbh
);
4388 set_buffer_uptodate(sbh
);
4395 * Have we just finished recovery? If so, and if we are mounting (or
4396 * remounting) the filesystem readonly, then we will end up with a
4397 * consistent fs on disk. Record that fact.
4399 static void ext4_mark_recovery_complete(struct super_block
*sb
,
4400 struct ext4_super_block
*es
)
4402 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
4404 if (!ext4_has_feature_journal(sb
)) {
4405 BUG_ON(journal
!= NULL
);
4408 jbd2_journal_lock_updates(journal
);
4409 if (jbd2_journal_flush(journal
) < 0)
4412 if (ext4_has_feature_journal_needs_recovery(sb
) &&
4413 sb
->s_flags
& MS_RDONLY
) {
4414 ext4_clear_feature_journal_needs_recovery(sb
);
4415 ext4_commit_super(sb
, 1);
4419 jbd2_journal_unlock_updates(journal
);
4423 * If we are mounting (or read-write remounting) a filesystem whose journal
4424 * has recorded an error from a previous lifetime, move that error to the
4425 * main filesystem now.
4427 static void ext4_clear_journal_err(struct super_block
*sb
,
4428 struct ext4_super_block
*es
)
4434 BUG_ON(!ext4_has_feature_journal(sb
));
4436 journal
= EXT4_SB(sb
)->s_journal
;
4439 * Now check for any error status which may have been recorded in the
4440 * journal by a prior ext4_error() or ext4_abort()
4443 j_errno
= jbd2_journal_errno(journal
);
4447 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
4448 ext4_warning(sb
, "Filesystem error recorded "
4449 "from previous mount: %s", errstr
);
4450 ext4_warning(sb
, "Marking fs in need of filesystem check.");
4452 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
4453 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
4454 ext4_commit_super(sb
, 1);
4456 jbd2_journal_clear_err(journal
);
4457 jbd2_journal_update_sb_errno(journal
);
4462 * Force the running and committing transactions to commit,
4463 * and wait on the commit.
4465 int ext4_force_commit(struct super_block
*sb
)
4469 if (sb
->s_flags
& MS_RDONLY
)
4472 journal
= EXT4_SB(sb
)->s_journal
;
4473 return ext4_journal_force_commit(journal
);
4476 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
4480 bool needs_barrier
= false;
4481 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4483 trace_ext4_sync_fs(sb
, wait
);
4484 flush_workqueue(sbi
->rsv_conversion_wq
);
4486 * Writeback quota in non-journalled quota case - journalled quota has
4489 dquot_writeback_dquots(sb
, -1);
4491 * Data writeback is possible w/o journal transaction, so barrier must
4492 * being sent at the end of the function. But we can skip it if
4493 * transaction_commit will do it for us.
4495 if (sbi
->s_journal
) {
4496 target
= jbd2_get_latest_transaction(sbi
->s_journal
);
4497 if (wait
&& sbi
->s_journal
->j_flags
& JBD2_BARRIER
&&
4498 !jbd2_trans_will_send_data_barrier(sbi
->s_journal
, target
))
4499 needs_barrier
= true;
4501 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
4503 ret
= jbd2_log_wait_commit(sbi
->s_journal
,
4506 } else if (wait
&& test_opt(sb
, BARRIER
))
4507 needs_barrier
= true;
4508 if (needs_barrier
) {
4510 err
= blkdev_issue_flush(sb
->s_bdev
, GFP_KERNEL
, NULL
);
4519 * LVM calls this function before a (read-only) snapshot is created. This
4520 * gives us a chance to flush the journal completely and mark the fs clean.
4522 * Note that only this function cannot bring a filesystem to be in a clean
4523 * state independently. It relies on upper layer to stop all data & metadata
4526 static int ext4_freeze(struct super_block
*sb
)
4531 if (sb
->s_flags
& MS_RDONLY
)
4534 journal
= EXT4_SB(sb
)->s_journal
;
4537 /* Now we set up the journal barrier. */
4538 jbd2_journal_lock_updates(journal
);
4541 * Don't clear the needs_recovery flag if we failed to
4542 * flush the journal.
4544 error
= jbd2_journal_flush(journal
);
4548 /* Journal blocked and flushed, clear needs_recovery flag. */
4549 ext4_clear_feature_journal_needs_recovery(sb
);
4552 error
= ext4_commit_super(sb
, 1);
4555 /* we rely on upper layer to stop further updates */
4556 jbd2_journal_unlock_updates(journal
);
4561 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4562 * flag here, even though the filesystem is not technically dirty yet.
4564 static int ext4_unfreeze(struct super_block
*sb
)
4566 if (sb
->s_flags
& MS_RDONLY
)
4569 if (EXT4_SB(sb
)->s_journal
) {
4570 /* Reset the needs_recovery flag before the fs is unlocked. */
4571 ext4_set_feature_journal_needs_recovery(sb
);
4574 ext4_commit_super(sb
, 1);
4579 * Structure to save mount options for ext4_remount's benefit
4581 struct ext4_mount_options
{
4582 unsigned long s_mount_opt
;
4583 unsigned long s_mount_opt2
;
4586 unsigned long s_commit_interval
;
4587 u32 s_min_batch_time
, s_max_batch_time
;
4590 char *s_qf_names
[EXT4_MAXQUOTAS
];
4594 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
4596 struct ext4_super_block
*es
;
4597 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4598 unsigned long old_sb_flags
;
4599 struct ext4_mount_options old_opts
;
4600 int enable_quota
= 0;
4602 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
4607 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
4609 /* Store the original options */
4610 old_sb_flags
= sb
->s_flags
;
4611 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
4612 old_opts
.s_mount_opt2
= sbi
->s_mount_opt2
;
4613 old_opts
.s_resuid
= sbi
->s_resuid
;
4614 old_opts
.s_resgid
= sbi
->s_resgid
;
4615 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
4616 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
4617 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
4619 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
4620 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
4621 if (sbi
->s_qf_names
[i
]) {
4622 old_opts
.s_qf_names
[i
] = kstrdup(sbi
->s_qf_names
[i
],
4624 if (!old_opts
.s_qf_names
[i
]) {
4625 for (j
= 0; j
< i
; j
++)
4626 kfree(old_opts
.s_qf_names
[j
]);
4631 old_opts
.s_qf_names
[i
] = NULL
;
4633 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
4634 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
4636 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
, 1)) {
4641 if ((old_opts
.s_mount_opt
& EXT4_MOUNT_JOURNAL_CHECKSUM
) ^
4642 test_opt(sb
, JOURNAL_CHECKSUM
)) {
4643 ext4_msg(sb
, KERN_ERR
, "changing journal_checksum "
4644 "during remount not supported; ignoring");
4645 sbi
->s_mount_opt
^= EXT4_MOUNT_JOURNAL_CHECKSUM
;
4648 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
4649 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
4650 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4651 "both data=journal and delalloc");
4655 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
4656 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4657 "both data=journal and dioread_nolock");
4661 if (test_opt(sb
, DAX
)) {
4662 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4663 "both data=journal and dax");
4669 if ((sbi
->s_mount_opt
^ old_opts
.s_mount_opt
) & EXT4_MOUNT_DAX
) {
4670 ext4_msg(sb
, KERN_WARNING
, "warning: refusing change of "
4671 "dax flag with busy inodes while remounting");
4672 sbi
->s_mount_opt
^= EXT4_MOUNT_DAX
;
4675 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
)
4676 ext4_abort(sb
, "Abort forced by user");
4678 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
4679 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
4683 if (sbi
->s_journal
) {
4684 ext4_init_journal_params(sb
, sbi
->s_journal
);
4685 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
4688 if (*flags
& MS_LAZYTIME
)
4689 sb
->s_flags
|= MS_LAZYTIME
;
4691 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
)) {
4692 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
) {
4697 if (*flags
& MS_RDONLY
) {
4698 err
= sync_filesystem(sb
);
4701 err
= dquot_suspend(sb
, -1);
4706 * First of all, the unconditional stuff we have to do
4707 * to disable replay of the journal when we next remount
4709 sb
->s_flags
|= MS_RDONLY
;
4712 * OK, test if we are remounting a valid rw partition
4713 * readonly, and if so set the rdonly flag and then
4714 * mark the partition as valid again.
4716 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
4717 (sbi
->s_mount_state
& EXT4_VALID_FS
))
4718 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
4721 ext4_mark_recovery_complete(sb
, es
);
4723 /* Make sure we can mount this feature set readwrite */
4724 if (ext4_has_feature_readonly(sb
) ||
4725 !ext4_feature_set_ok(sb
, 0)) {
4730 * Make sure the group descriptor checksums
4731 * are sane. If they aren't, refuse to remount r/w.
4733 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
4734 struct ext4_group_desc
*gdp
=
4735 ext4_get_group_desc(sb
, g
, NULL
);
4737 if (!ext4_group_desc_csum_verify(sb
, g
, gdp
)) {
4738 ext4_msg(sb
, KERN_ERR
,
4739 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4740 g
, le16_to_cpu(ext4_group_desc_csum(sb
, g
, gdp
)),
4741 le16_to_cpu(gdp
->bg_checksum
));
4748 * If we have an unprocessed orphan list hanging
4749 * around from a previously readonly bdev mount,
4750 * require a full umount/remount for now.
4752 if (es
->s_last_orphan
) {
4753 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
4754 "remount RDWR because of unprocessed "
4755 "orphan inode list. Please "
4756 "umount/remount instead");
4762 * Mounting a RDONLY partition read-write, so reread
4763 * and store the current valid flag. (It may have
4764 * been changed by e2fsck since we originally mounted
4768 ext4_clear_journal_err(sb
, es
);
4769 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
4770 if (!ext4_setup_super(sb
, es
, 0))
4771 sb
->s_flags
&= ~MS_RDONLY
;
4772 if (ext4_has_feature_mmp(sb
))
4773 if (ext4_multi_mount_protect(sb
,
4774 le64_to_cpu(es
->s_mmp_block
))) {
4783 * Reinitialize lazy itable initialization thread based on
4786 if ((sb
->s_flags
& MS_RDONLY
) || !test_opt(sb
, INIT_INODE_TABLE
))
4787 ext4_unregister_li_request(sb
);
4789 ext4_group_t first_not_zeroed
;
4790 first_not_zeroed
= ext4_has_uninit_itable(sb
);
4791 ext4_register_li_request(sb
, first_not_zeroed
);
4794 ext4_setup_system_zone(sb
);
4795 if (sbi
->s_journal
== NULL
&& !(old_sb_flags
& MS_RDONLY
))
4796 ext4_commit_super(sb
, 1);
4799 /* Release old quota file names */
4800 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
4801 kfree(old_opts
.s_qf_names
[i
]);
4803 if (sb_any_quota_suspended(sb
))
4804 dquot_resume(sb
, -1);
4805 else if (ext4_has_feature_quota(sb
)) {
4806 err
= ext4_enable_quotas(sb
);
4813 *flags
= (*flags
& ~MS_LAZYTIME
) | (sb
->s_flags
& MS_LAZYTIME
);
4814 ext4_msg(sb
, KERN_INFO
, "re-mounted. Opts: %s", orig_data
);
4819 sb
->s_flags
= old_sb_flags
;
4820 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
4821 sbi
->s_mount_opt2
= old_opts
.s_mount_opt2
;
4822 sbi
->s_resuid
= old_opts
.s_resuid
;
4823 sbi
->s_resgid
= old_opts
.s_resgid
;
4824 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
4825 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
4826 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
4828 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
4829 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
4830 kfree(sbi
->s_qf_names
[i
]);
4831 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
4839 static int ext4_statfs_project(struct super_block
*sb
,
4840 kprojid_t projid
, struct kstatfs
*buf
)
4843 struct dquot
*dquot
;
4847 qid
= make_kqid_projid(projid
);
4848 dquot
= dqget(sb
, qid
);
4850 return PTR_ERR(dquot
);
4851 spin_lock(&dq_data_lock
);
4853 limit
= (dquot
->dq_dqb
.dqb_bsoftlimit
?
4854 dquot
->dq_dqb
.dqb_bsoftlimit
:
4855 dquot
->dq_dqb
.dqb_bhardlimit
) >> sb
->s_blocksize_bits
;
4856 if (limit
&& buf
->f_blocks
> limit
) {
4857 curblock
= dquot
->dq_dqb
.dqb_curspace
>> sb
->s_blocksize_bits
;
4858 buf
->f_blocks
= limit
;
4859 buf
->f_bfree
= buf
->f_bavail
=
4860 (buf
->f_blocks
> curblock
) ?
4861 (buf
->f_blocks
- curblock
) : 0;
4864 limit
= dquot
->dq_dqb
.dqb_isoftlimit
?
4865 dquot
->dq_dqb
.dqb_isoftlimit
:
4866 dquot
->dq_dqb
.dqb_ihardlimit
;
4867 if (limit
&& buf
->f_files
> limit
) {
4868 buf
->f_files
= limit
;
4870 (buf
->f_files
> dquot
->dq_dqb
.dqb_curinodes
) ?
4871 (buf
->f_files
- dquot
->dq_dqb
.dqb_curinodes
) : 0;
4874 spin_unlock(&dq_data_lock
);
4880 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
4882 struct super_block
*sb
= dentry
->d_sb
;
4883 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4884 struct ext4_super_block
*es
= sbi
->s_es
;
4885 ext4_fsblk_t overhead
= 0, resv_blocks
;
4888 resv_blocks
= EXT4_C2B(sbi
, atomic64_read(&sbi
->s_resv_clusters
));
4890 if (!test_opt(sb
, MINIX_DF
))
4891 overhead
= sbi
->s_overhead
;
4893 buf
->f_type
= EXT4_SUPER_MAGIC
;
4894 buf
->f_bsize
= sb
->s_blocksize
;
4895 buf
->f_blocks
= ext4_blocks_count(es
) - EXT4_C2B(sbi
, overhead
);
4896 bfree
= percpu_counter_sum_positive(&sbi
->s_freeclusters_counter
) -
4897 percpu_counter_sum_positive(&sbi
->s_dirtyclusters_counter
);
4898 /* prevent underflow in case that few free space is available */
4899 buf
->f_bfree
= EXT4_C2B(sbi
, max_t(s64
, bfree
, 0));
4900 buf
->f_bavail
= buf
->f_bfree
-
4901 (ext4_r_blocks_count(es
) + resv_blocks
);
4902 if (buf
->f_bfree
< (ext4_r_blocks_count(es
) + resv_blocks
))
4904 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
4905 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
4906 buf
->f_namelen
= EXT4_NAME_LEN
;
4907 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
4908 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
4909 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
4910 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
4913 if (ext4_test_inode_flag(dentry
->d_inode
, EXT4_INODE_PROJINHERIT
) &&
4914 sb_has_quota_limits_enabled(sb
, PRJQUOTA
))
4915 ext4_statfs_project(sb
, EXT4_I(dentry
->d_inode
)->i_projid
, buf
);
4920 /* Helper function for writing quotas on sync - we need to start transaction
4921 * before quota file is locked for write. Otherwise the are possible deadlocks:
4922 * Process 1 Process 2
4923 * ext4_create() quota_sync()
4924 * jbd2_journal_start() write_dquot()
4925 * dquot_initialize() down(dqio_mutex)
4926 * down(dqio_mutex) jbd2_journal_start()
4932 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
4934 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_id
.type
];
4937 static int ext4_write_dquot(struct dquot
*dquot
)
4941 struct inode
*inode
;
4943 inode
= dquot_to_inode(dquot
);
4944 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
,
4945 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
4947 return PTR_ERR(handle
);
4948 ret
= dquot_commit(dquot
);
4949 err
= ext4_journal_stop(handle
);
4955 static int ext4_acquire_dquot(struct dquot
*dquot
)
4960 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
4961 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
4963 return PTR_ERR(handle
);
4964 ret
= dquot_acquire(dquot
);
4965 err
= ext4_journal_stop(handle
);
4971 static int ext4_release_dquot(struct dquot
*dquot
)
4976 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
4977 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
4978 if (IS_ERR(handle
)) {
4979 /* Release dquot anyway to avoid endless cycle in dqput() */
4980 dquot_release(dquot
);
4981 return PTR_ERR(handle
);
4983 ret
= dquot_release(dquot
);
4984 err
= ext4_journal_stop(handle
);
4990 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
4992 struct super_block
*sb
= dquot
->dq_sb
;
4993 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4995 /* Are we journaling quotas? */
4996 if (ext4_has_feature_quota(sb
) ||
4997 sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
4998 dquot_mark_dquot_dirty(dquot
);
4999 return ext4_write_dquot(dquot
);
5001 return dquot_mark_dquot_dirty(dquot
);
5005 static int ext4_write_info(struct super_block
*sb
, int type
)
5010 /* Data block + inode block */
5011 handle
= ext4_journal_start(d_inode(sb
->s_root
), EXT4_HT_QUOTA
, 2);
5013 return PTR_ERR(handle
);
5014 ret
= dquot_commit_info(sb
, type
);
5015 err
= ext4_journal_stop(handle
);
5022 * Turn on quotas during mount time - we need to find
5023 * the quota file and such...
5025 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
5027 return dquot_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
5028 EXT4_SB(sb
)->s_jquota_fmt
, type
);
5032 * Standard function to be called on quota_on
5034 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
5039 if (!test_opt(sb
, QUOTA
))
5042 /* Quotafile not on the same filesystem? */
5043 if (path
->dentry
->d_sb
!= sb
)
5045 /* Journaling quota? */
5046 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
5047 /* Quotafile not in fs root? */
5048 if (path
->dentry
->d_parent
!= sb
->s_root
)
5049 ext4_msg(sb
, KERN_WARNING
,
5050 "Quota file not on filesystem root. "
5051 "Journaled quota will not work");
5055 * When we journal data on quota file, we have to flush journal to see
5056 * all updates to the file when we bypass pagecache...
5058 if (EXT4_SB(sb
)->s_journal
&&
5059 ext4_should_journal_data(d_inode(path
->dentry
))) {
5061 * We don't need to lock updates but journal_flush() could
5062 * otherwise be livelocked...
5064 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
5065 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
5066 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
5071 return dquot_quota_on(sb
, type
, format_id
, path
);
5074 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
5078 struct inode
*qf_inode
;
5079 unsigned long qf_inums
[EXT4_MAXQUOTAS
] = {
5080 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
5081 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
),
5082 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_prj_quota_inum
)
5085 BUG_ON(!ext4_has_feature_quota(sb
));
5087 if (!qf_inums
[type
])
5090 qf_inode
= ext4_iget(sb
, qf_inums
[type
]);
5091 if (IS_ERR(qf_inode
)) {
5092 ext4_error(sb
, "Bad quota inode # %lu", qf_inums
[type
]);
5093 return PTR_ERR(qf_inode
);
5096 /* Don't account quota for quota files to avoid recursion */
5097 qf_inode
->i_flags
|= S_NOQUOTA
;
5098 err
= dquot_enable(qf_inode
, type
, format_id
, flags
);
5104 /* Enable usage tracking for all quota types. */
5105 static int ext4_enable_quotas(struct super_block
*sb
)
5108 unsigned long qf_inums
[EXT4_MAXQUOTAS
] = {
5109 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
5110 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
),
5111 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_prj_quota_inum
)
5114 sb_dqopt(sb
)->flags
|= DQUOT_QUOTA_SYS_FILE
;
5115 for (type
= 0; type
< EXT4_MAXQUOTAS
; type
++) {
5116 if (qf_inums
[type
]) {
5117 err
= ext4_quota_enable(sb
, type
, QFMT_VFS_V1
,
5118 DQUOT_USAGE_ENABLED
);
5121 "Failed to enable quota tracking "
5122 "(type=%d, err=%d). Please run "
5123 "e2fsck to fix.", type
, err
);
5131 static int ext4_quota_off(struct super_block
*sb
, int type
)
5133 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5136 /* Force all delayed allocation blocks to be allocated.
5137 * Caller already holds s_umount sem */
5138 if (test_opt(sb
, DELALLOC
))
5139 sync_filesystem(sb
);
5144 /* Update modification times of quota files when userspace can
5145 * start looking at them */
5146 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
, 1);
5149 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
5150 ext4_mark_inode_dirty(handle
, inode
);
5151 ext4_journal_stop(handle
);
5154 return dquot_quota_off(sb
, type
);
5157 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5158 * acquiring the locks... As quota files are never truncated and quota code
5159 * itself serializes the operations (and no one else should touch the files)
5160 * we don't have to be afraid of races */
5161 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
5162 size_t len
, loff_t off
)
5164 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5165 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5166 int offset
= off
& (sb
->s_blocksize
- 1);
5169 struct buffer_head
*bh
;
5170 loff_t i_size
= i_size_read(inode
);
5174 if (off
+len
> i_size
)
5177 while (toread
> 0) {
5178 tocopy
= sb
->s_blocksize
- offset
< toread
?
5179 sb
->s_blocksize
- offset
: toread
;
5180 bh
= ext4_bread(NULL
, inode
, blk
, 0);
5183 if (!bh
) /* A hole? */
5184 memset(data
, 0, tocopy
);
5186 memcpy(data
, bh
->b_data
+offset
, tocopy
);
5196 /* Write to quotafile (we know the transaction is already started and has
5197 * enough credits) */
5198 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
5199 const char *data
, size_t len
, loff_t off
)
5201 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5202 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5203 int err
, offset
= off
& (sb
->s_blocksize
- 1);
5205 struct buffer_head
*bh
;
5206 handle_t
*handle
= journal_current_handle();
5208 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
5209 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5210 " cancelled because transaction is not started",
5211 (unsigned long long)off
, (unsigned long long)len
);
5215 * Since we account only one data block in transaction credits,
5216 * then it is impossible to cross a block boundary.
5218 if (sb
->s_blocksize
- offset
< len
) {
5219 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5220 " cancelled because not block aligned",
5221 (unsigned long long)off
, (unsigned long long)len
);
5226 bh
= ext4_bread(handle
, inode
, blk
,
5227 EXT4_GET_BLOCKS_CREATE
|
5228 EXT4_GET_BLOCKS_METADATA_NOFAIL
);
5229 } while (IS_ERR(bh
) && (PTR_ERR(bh
) == -ENOSPC
) &&
5230 ext4_should_retry_alloc(inode
->i_sb
, &retries
));
5235 BUFFER_TRACE(bh
, "get write access");
5236 err
= ext4_journal_get_write_access(handle
, bh
);
5242 memcpy(bh
->b_data
+offset
, data
, len
);
5243 flush_dcache_page(bh
->b_page
);
5245 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
5248 if (inode
->i_size
< off
+ len
) {
5249 i_size_write(inode
, off
+ len
);
5250 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
5251 ext4_mark_inode_dirty(handle
, inode
);
5258 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
5259 const char *dev_name
, void *data
)
5261 return mount_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
);
5264 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5265 static inline void register_as_ext2(void)
5267 int err
= register_filesystem(&ext2_fs_type
);
5270 "EXT4-fs: Unable to register as ext2 (%d)\n", err
);
5273 static inline void unregister_as_ext2(void)
5275 unregister_filesystem(&ext2_fs_type
);
5278 static inline int ext2_feature_set_ok(struct super_block
*sb
)
5280 if (ext4_has_unknown_ext2_incompat_features(sb
))
5282 if (sb
->s_flags
& MS_RDONLY
)
5284 if (ext4_has_unknown_ext2_ro_compat_features(sb
))
5289 static inline void register_as_ext2(void) { }
5290 static inline void unregister_as_ext2(void) { }
5291 static inline int ext2_feature_set_ok(struct super_block
*sb
) { return 0; }
5294 static inline void register_as_ext3(void)
5296 int err
= register_filesystem(&ext3_fs_type
);
5299 "EXT4-fs: Unable to register as ext3 (%d)\n", err
);
5302 static inline void unregister_as_ext3(void)
5304 unregister_filesystem(&ext3_fs_type
);
5307 static inline int ext3_feature_set_ok(struct super_block
*sb
)
5309 if (ext4_has_unknown_ext3_incompat_features(sb
))
5311 if (!ext4_has_feature_journal(sb
))
5313 if (sb
->s_flags
& MS_RDONLY
)
5315 if (ext4_has_unknown_ext3_ro_compat_features(sb
))
5320 static struct file_system_type ext4_fs_type
= {
5321 .owner
= THIS_MODULE
,
5323 .mount
= ext4_mount
,
5324 .kill_sb
= kill_block_super
,
5325 .fs_flags
= FS_REQUIRES_DEV
,
5327 MODULE_ALIAS_FS("ext4");
5329 /* Shared across all ext4 file systems */
5330 wait_queue_head_t ext4__ioend_wq
[EXT4_WQ_HASH_SZ
];
5332 static int __init
ext4_init_fs(void)
5336 ratelimit_state_init(&ext4_mount_msg_ratelimit
, 30 * HZ
, 64);
5337 ext4_li_info
= NULL
;
5338 mutex_init(&ext4_li_mtx
);
5340 /* Build-time check for flags consistency */
5341 ext4_check_flag_values();
5343 for (i
= 0; i
< EXT4_WQ_HASH_SZ
; i
++)
5344 init_waitqueue_head(&ext4__ioend_wq
[i
]);
5346 err
= ext4_init_es();
5350 err
= ext4_init_pageio();
5354 err
= ext4_init_system_zone();
5358 err
= ext4_init_sysfs();
5362 err
= ext4_init_mballoc();
5365 err
= init_inodecache();
5370 err
= register_filesystem(&ext4_fs_type
);
5376 unregister_as_ext2();
5377 unregister_as_ext3();
5378 destroy_inodecache();
5380 ext4_exit_mballoc();
5384 ext4_exit_system_zone();
5393 static void __exit
ext4_exit_fs(void)
5396 ext4_destroy_lazyinit_thread();
5397 unregister_as_ext2();
5398 unregister_as_ext3();
5399 unregister_filesystem(&ext4_fs_type
);
5400 destroy_inodecache();
5401 ext4_exit_mballoc();
5403 ext4_exit_system_zone();
5408 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5409 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5410 MODULE_LICENSE("GPL");
5411 module_init(ext4_init_fs
)
5412 module_exit(ext4_exit_fs
)