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/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.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/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/cleancache.h>
42 #include <asm/uaccess.h>
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
48 #include "ext4_extents.h" /* Needed for trace points definition */
49 #include "ext4_jbd2.h"
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
57 static struct proc_dir_entry
*ext4_proc_root
;
58 static struct kset
*ext4_kset
;
59 static struct ext4_lazy_init
*ext4_li_info
;
60 static struct mutex ext4_li_mtx
;
61 static struct ext4_features
*ext4_feat
;
62 static int ext4_mballoc_ready
;
64 static int ext4_load_journal(struct super_block
*, struct ext4_super_block
*,
65 unsigned long journal_devnum
);
66 static int ext4_show_options(struct seq_file
*seq
, struct dentry
*root
);
67 static int ext4_commit_super(struct super_block
*sb
, int sync
);
68 static void ext4_mark_recovery_complete(struct super_block
*sb
,
69 struct ext4_super_block
*es
);
70 static void ext4_clear_journal_err(struct super_block
*sb
,
71 struct ext4_super_block
*es
);
72 static int ext4_sync_fs(struct super_block
*sb
, int wait
);
73 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
);
74 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
75 static int ext4_unfreeze(struct super_block
*sb
);
76 static int ext4_freeze(struct super_block
*sb
);
77 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
78 const char *dev_name
, void *data
);
79 static inline int ext2_feature_set_ok(struct super_block
*sb
);
80 static inline int ext3_feature_set_ok(struct super_block
*sb
);
81 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
);
82 static void ext4_destroy_lazyinit_thread(void);
83 static void ext4_unregister_li_request(struct super_block
*sb
);
84 static void ext4_clear_request_list(void);
85 static int ext4_reserve_clusters(struct ext4_sb_info
*, ext4_fsblk_t
);
87 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88 static struct file_system_type ext2_fs_type
= {
92 .kill_sb
= kill_block_super
,
93 .fs_flags
= FS_REQUIRES_DEV
,
95 MODULE_ALIAS_FS("ext2");
97 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
99 #define IS_EXT2_SB(sb) (0)
103 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
104 static struct file_system_type ext3_fs_type
= {
105 .owner
= THIS_MODULE
,
108 .kill_sb
= kill_block_super
,
109 .fs_flags
= FS_REQUIRES_DEV
,
111 MODULE_ALIAS_FS("ext3");
112 MODULE_ALIAS("ext3");
113 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
115 #define IS_EXT3_SB(sb) (0)
118 static int ext4_verify_csum_type(struct super_block
*sb
,
119 struct ext4_super_block
*es
)
121 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
,
122 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
))
125 return es
->s_checksum_type
== EXT4_CRC32C_CHKSUM
;
128 static __le32
ext4_superblock_csum(struct super_block
*sb
,
129 struct ext4_super_block
*es
)
131 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
132 int offset
= offsetof(struct ext4_super_block
, s_checksum
);
135 csum
= ext4_chksum(sbi
, ~0, (char *)es
, offset
);
137 return cpu_to_le32(csum
);
140 static int ext4_superblock_csum_verify(struct super_block
*sb
,
141 struct ext4_super_block
*es
)
143 if (!ext4_has_metadata_csum(sb
))
146 return es
->s_checksum
== ext4_superblock_csum(sb
, es
);
149 void ext4_superblock_csum_set(struct super_block
*sb
)
151 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
153 if (!ext4_has_metadata_csum(sb
))
156 es
->s_checksum
= ext4_superblock_csum(sb
, es
);
159 void *ext4_kvmalloc(size_t size
, gfp_t flags
)
163 ret
= kmalloc(size
, flags
| __GFP_NOWARN
);
165 ret
= __vmalloc(size
, flags
, PAGE_KERNEL
);
169 void *ext4_kvzalloc(size_t size
, gfp_t flags
)
173 ret
= kzalloc(size
, flags
| __GFP_NOWARN
);
175 ret
= __vmalloc(size
, flags
| __GFP_ZERO
, PAGE_KERNEL
);
179 void ext4_kvfree(void *ptr
)
181 if (is_vmalloc_addr(ptr
))
188 ext4_fsblk_t
ext4_block_bitmap(struct super_block
*sb
,
189 struct ext4_group_desc
*bg
)
191 return le32_to_cpu(bg
->bg_block_bitmap_lo
) |
192 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
193 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_block_bitmap_hi
) << 32 : 0);
196 ext4_fsblk_t
ext4_inode_bitmap(struct super_block
*sb
,
197 struct ext4_group_desc
*bg
)
199 return le32_to_cpu(bg
->bg_inode_bitmap_lo
) |
200 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
201 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_bitmap_hi
) << 32 : 0);
204 ext4_fsblk_t
ext4_inode_table(struct super_block
*sb
,
205 struct ext4_group_desc
*bg
)
207 return le32_to_cpu(bg
->bg_inode_table_lo
) |
208 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
209 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_table_hi
) << 32 : 0);
212 __u32
ext4_free_group_clusters(struct super_block
*sb
,
213 struct ext4_group_desc
*bg
)
215 return le16_to_cpu(bg
->bg_free_blocks_count_lo
) |
216 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
217 (__u32
)le16_to_cpu(bg
->bg_free_blocks_count_hi
) << 16 : 0);
220 __u32
ext4_free_inodes_count(struct super_block
*sb
,
221 struct ext4_group_desc
*bg
)
223 return le16_to_cpu(bg
->bg_free_inodes_count_lo
) |
224 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
225 (__u32
)le16_to_cpu(bg
->bg_free_inodes_count_hi
) << 16 : 0);
228 __u32
ext4_used_dirs_count(struct super_block
*sb
,
229 struct ext4_group_desc
*bg
)
231 return le16_to_cpu(bg
->bg_used_dirs_count_lo
) |
232 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
233 (__u32
)le16_to_cpu(bg
->bg_used_dirs_count_hi
) << 16 : 0);
236 __u32
ext4_itable_unused_count(struct super_block
*sb
,
237 struct ext4_group_desc
*bg
)
239 return le16_to_cpu(bg
->bg_itable_unused_lo
) |
240 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
241 (__u32
)le16_to_cpu(bg
->bg_itable_unused_hi
) << 16 : 0);
244 void ext4_block_bitmap_set(struct super_block
*sb
,
245 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
247 bg
->bg_block_bitmap_lo
= cpu_to_le32((u32
)blk
);
248 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
249 bg
->bg_block_bitmap_hi
= cpu_to_le32(blk
>> 32);
252 void ext4_inode_bitmap_set(struct super_block
*sb
,
253 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
255 bg
->bg_inode_bitmap_lo
= cpu_to_le32((u32
)blk
);
256 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
257 bg
->bg_inode_bitmap_hi
= cpu_to_le32(blk
>> 32);
260 void ext4_inode_table_set(struct super_block
*sb
,
261 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
263 bg
->bg_inode_table_lo
= cpu_to_le32((u32
)blk
);
264 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
265 bg
->bg_inode_table_hi
= cpu_to_le32(blk
>> 32);
268 void ext4_free_group_clusters_set(struct super_block
*sb
,
269 struct ext4_group_desc
*bg
, __u32 count
)
271 bg
->bg_free_blocks_count_lo
= cpu_to_le16((__u16
)count
);
272 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
273 bg
->bg_free_blocks_count_hi
= cpu_to_le16(count
>> 16);
276 void ext4_free_inodes_set(struct super_block
*sb
,
277 struct ext4_group_desc
*bg
, __u32 count
)
279 bg
->bg_free_inodes_count_lo
= cpu_to_le16((__u16
)count
);
280 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
281 bg
->bg_free_inodes_count_hi
= cpu_to_le16(count
>> 16);
284 void ext4_used_dirs_set(struct super_block
*sb
,
285 struct ext4_group_desc
*bg
, __u32 count
)
287 bg
->bg_used_dirs_count_lo
= cpu_to_le16((__u16
)count
);
288 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
289 bg
->bg_used_dirs_count_hi
= cpu_to_le16(count
>> 16);
292 void ext4_itable_unused_set(struct super_block
*sb
,
293 struct ext4_group_desc
*bg
, __u32 count
)
295 bg
->bg_itable_unused_lo
= cpu_to_le16((__u16
)count
);
296 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
297 bg
->bg_itable_unused_hi
= cpu_to_le16(count
>> 16);
301 static void __save_error_info(struct super_block
*sb
, const char *func
,
304 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
306 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
307 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
308 es
->s_last_error_time
= cpu_to_le32(get_seconds());
309 strncpy(es
->s_last_error_func
, func
, sizeof(es
->s_last_error_func
));
310 es
->s_last_error_line
= cpu_to_le32(line
);
311 if (!es
->s_first_error_time
) {
312 es
->s_first_error_time
= es
->s_last_error_time
;
313 strncpy(es
->s_first_error_func
, func
,
314 sizeof(es
->s_first_error_func
));
315 es
->s_first_error_line
= cpu_to_le32(line
);
316 es
->s_first_error_ino
= es
->s_last_error_ino
;
317 es
->s_first_error_block
= es
->s_last_error_block
;
320 * Start the daily error reporting function if it hasn't been
323 if (!es
->s_error_count
)
324 mod_timer(&EXT4_SB(sb
)->s_err_report
, jiffies
+ 24*60*60*HZ
);
325 le32_add_cpu(&es
->s_error_count
, 1);
328 static void save_error_info(struct super_block
*sb
, const char *func
,
331 __save_error_info(sb
, func
, line
);
332 ext4_commit_super(sb
, 1);
336 * The del_gendisk() function uninitializes the disk-specific data
337 * structures, including the bdi structure, without telling anyone
338 * else. Once this happens, any attempt to call mark_buffer_dirty()
339 * (for example, by ext4_commit_super), will cause a kernel OOPS.
340 * This is a kludge to prevent these oops until we can put in a proper
341 * hook in del_gendisk() to inform the VFS and file system layers.
343 static int block_device_ejected(struct super_block
*sb
)
345 struct inode
*bd_inode
= sb
->s_bdev
->bd_inode
;
346 struct backing_dev_info
*bdi
= bd_inode
->i_mapping
->backing_dev_info
;
348 return bdi
->dev
== NULL
;
351 static void ext4_journal_commit_callback(journal_t
*journal
, transaction_t
*txn
)
353 struct super_block
*sb
= journal
->j_private
;
354 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
355 int error
= is_journal_aborted(journal
);
356 struct ext4_journal_cb_entry
*jce
;
358 BUG_ON(txn
->t_state
== T_FINISHED
);
359 spin_lock(&sbi
->s_md_lock
);
360 while (!list_empty(&txn
->t_private_list
)) {
361 jce
= list_entry(txn
->t_private_list
.next
,
362 struct ext4_journal_cb_entry
, jce_list
);
363 list_del_init(&jce
->jce_list
);
364 spin_unlock(&sbi
->s_md_lock
);
365 jce
->jce_func(sb
, jce
, error
);
366 spin_lock(&sbi
->s_md_lock
);
368 spin_unlock(&sbi
->s_md_lock
);
371 /* Deal with the reporting of failure conditions on a filesystem such as
372 * inconsistencies detected or read IO failures.
374 * On ext2, we can store the error state of the filesystem in the
375 * superblock. That is not possible on ext4, because we may have other
376 * write ordering constraints on the superblock which prevent us from
377 * writing it out straight away; and given that the journal is about to
378 * be aborted, we can't rely on the current, or future, transactions to
379 * write out the superblock safely.
381 * We'll just use the jbd2_journal_abort() error code to record an error in
382 * the journal instead. On recovery, the journal will complain about
383 * that error until we've noted it down and cleared it.
386 static void ext4_handle_error(struct super_block
*sb
)
388 if (sb
->s_flags
& MS_RDONLY
)
391 if (!test_opt(sb
, ERRORS_CONT
)) {
392 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
394 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
396 jbd2_journal_abort(journal
, -EIO
);
398 if (test_opt(sb
, ERRORS_RO
)) {
399 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
401 * Make sure updated value of ->s_mount_flags will be visible
402 * before ->s_flags update
405 sb
->s_flags
|= MS_RDONLY
;
407 if (test_opt(sb
, ERRORS_PANIC
))
408 panic("EXT4-fs (device %s): panic forced after error\n",
412 #define ext4_error_ratelimit(sb) \
413 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
416 void __ext4_error(struct super_block
*sb
, const char *function
,
417 unsigned int line
, const char *fmt
, ...)
419 struct va_format vaf
;
422 if (ext4_error_ratelimit(sb
)) {
427 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
428 sb
->s_id
, function
, line
, current
->comm
, &vaf
);
431 save_error_info(sb
, function
, line
);
432 ext4_handle_error(sb
);
435 void __ext4_error_inode(struct inode
*inode
, const char *function
,
436 unsigned int line
, ext4_fsblk_t block
,
437 const char *fmt
, ...)
440 struct va_format vaf
;
441 struct ext4_super_block
*es
= EXT4_SB(inode
->i_sb
)->s_es
;
443 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
444 es
->s_last_error_block
= cpu_to_le64(block
);
445 if (ext4_error_ratelimit(inode
->i_sb
)) {
450 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
451 "inode #%lu: block %llu: comm %s: %pV\n",
452 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
453 block
, current
->comm
, &vaf
);
455 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
456 "inode #%lu: comm %s: %pV\n",
457 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
458 current
->comm
, &vaf
);
461 save_error_info(inode
->i_sb
, function
, line
);
462 ext4_handle_error(inode
->i_sb
);
465 void __ext4_error_file(struct file
*file
, const char *function
,
466 unsigned int line
, ext4_fsblk_t block
,
467 const char *fmt
, ...)
470 struct va_format vaf
;
471 struct ext4_super_block
*es
;
472 struct inode
*inode
= file_inode(file
);
473 char pathname
[80], *path
;
475 es
= EXT4_SB(inode
->i_sb
)->s_es
;
476 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
477 if (ext4_error_ratelimit(inode
->i_sb
)) {
478 path
= d_path(&(file
->f_path
), pathname
, sizeof(pathname
));
486 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
487 "block %llu: comm %s: path %s: %pV\n",
488 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
489 block
, current
->comm
, path
, &vaf
);
492 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
493 "comm %s: path %s: %pV\n",
494 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
495 current
->comm
, path
, &vaf
);
498 save_error_info(inode
->i_sb
, function
, line
);
499 ext4_handle_error(inode
->i_sb
);
502 const char *ext4_decode_error(struct super_block
*sb
, int errno
,
509 errstr
= "IO failure";
512 errstr
= "Out of memory";
515 if (!sb
|| (EXT4_SB(sb
)->s_journal
&&
516 EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
))
517 errstr
= "Journal has aborted";
519 errstr
= "Readonly filesystem";
522 /* If the caller passed in an extra buffer for unknown
523 * errors, textualise them now. Else we just return
526 /* Check for truncated error codes... */
527 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
536 /* __ext4_std_error decodes expected errors from journaling functions
537 * automatically and invokes the appropriate error response. */
539 void __ext4_std_error(struct super_block
*sb
, const char *function
,
540 unsigned int line
, int errno
)
545 /* Special case: if the error is EROFS, and we're not already
546 * inside a transaction, then there's really no point in logging
548 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
549 (sb
->s_flags
& MS_RDONLY
))
552 if (ext4_error_ratelimit(sb
)) {
553 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
554 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s:%d: %s\n",
555 sb
->s_id
, function
, line
, errstr
);
558 save_error_info(sb
, function
, line
);
559 ext4_handle_error(sb
);
563 * ext4_abort is a much stronger failure handler than ext4_error. The
564 * abort function may be used to deal with unrecoverable failures such
565 * as journal IO errors or ENOMEM at a critical moment in log management.
567 * We unconditionally force the filesystem into an ABORT|READONLY state,
568 * unless the error response on the fs has been set to panic in which
569 * case we take the easy way out and panic immediately.
572 void __ext4_abort(struct super_block
*sb
, const char *function
,
573 unsigned int line
, const char *fmt
, ...)
577 save_error_info(sb
, function
, line
);
579 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: ", sb
->s_id
,
585 if ((sb
->s_flags
& MS_RDONLY
) == 0) {
586 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
587 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
589 * Make sure updated value of ->s_mount_flags will be visible
590 * before ->s_flags update
593 sb
->s_flags
|= MS_RDONLY
;
594 if (EXT4_SB(sb
)->s_journal
)
595 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
596 save_error_info(sb
, function
, line
);
598 if (test_opt(sb
, ERRORS_PANIC
))
599 panic("EXT4-fs panic from previous error\n");
602 void __ext4_msg(struct super_block
*sb
,
603 const char *prefix
, const char *fmt
, ...)
605 struct va_format vaf
;
608 if (!___ratelimit(&(EXT4_SB(sb
)->s_msg_ratelimit_state
), "EXT4-fs"))
614 printk("%sEXT4-fs (%s): %pV\n", prefix
, sb
->s_id
, &vaf
);
618 void __ext4_warning(struct super_block
*sb
, const char *function
,
619 unsigned int line
, const char *fmt
, ...)
621 struct va_format vaf
;
624 if (!___ratelimit(&(EXT4_SB(sb
)->s_warning_ratelimit_state
),
631 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: %pV\n",
632 sb
->s_id
, function
, line
, &vaf
);
636 void __ext4_grp_locked_error(const char *function
, unsigned int line
,
637 struct super_block
*sb
, ext4_group_t grp
,
638 unsigned long ino
, ext4_fsblk_t block
,
639 const char *fmt
, ...)
643 struct va_format vaf
;
645 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
647 es
->s_last_error_ino
= cpu_to_le32(ino
);
648 es
->s_last_error_block
= cpu_to_le64(block
);
649 __save_error_info(sb
, function
, line
);
651 if (ext4_error_ratelimit(sb
)) {
655 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: group %u, ",
656 sb
->s_id
, function
, line
, grp
);
658 printk(KERN_CONT
"inode %lu: ", ino
);
660 printk(KERN_CONT
"block %llu:",
661 (unsigned long long) block
);
662 printk(KERN_CONT
"%pV\n", &vaf
);
666 if (test_opt(sb
, ERRORS_CONT
)) {
667 ext4_commit_super(sb
, 0);
671 ext4_unlock_group(sb
, grp
);
672 ext4_handle_error(sb
);
674 * We only get here in the ERRORS_RO case; relocking the group
675 * may be dangerous, but nothing bad will happen since the
676 * filesystem will have already been marked read/only and the
677 * journal has been aborted. We return 1 as a hint to callers
678 * who might what to use the return value from
679 * ext4_grp_locked_error() to distinguish between the
680 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
681 * aggressively from the ext4 function in question, with a
682 * more appropriate error code.
684 ext4_lock_group(sb
, grp
);
688 void ext4_update_dynamic_rev(struct super_block
*sb
)
690 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
692 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
696 "updating to rev %d because of new feature flag, "
697 "running e2fsck is recommended",
700 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
701 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
702 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
703 /* leave es->s_feature_*compat flags alone */
704 /* es->s_uuid will be set by e2fsck if empty */
707 * The rest of the superblock fields should be zero, and if not it
708 * means they are likely already in use, so leave them alone. We
709 * can leave it up to e2fsck to clean up any inconsistencies there.
714 * Open the external journal device
716 static struct block_device
*ext4_blkdev_get(dev_t dev
, struct super_block
*sb
)
718 struct block_device
*bdev
;
719 char b
[BDEVNAME_SIZE
];
721 bdev
= blkdev_get_by_dev(dev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
, sb
);
727 ext4_msg(sb
, KERN_ERR
, "failed to open journal device %s: %ld",
728 __bdevname(dev
, b
), PTR_ERR(bdev
));
733 * Release the journal device
735 static void ext4_blkdev_put(struct block_device
*bdev
)
737 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
);
740 static void ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
742 struct block_device
*bdev
;
743 bdev
= sbi
->journal_bdev
;
745 ext4_blkdev_put(bdev
);
746 sbi
->journal_bdev
= NULL
;
750 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
752 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
755 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
759 ext4_msg(sb
, KERN_ERR
, "sb orphan head is %d",
760 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
762 printk(KERN_ERR
"sb_info orphan list:\n");
763 list_for_each(l
, &sbi
->s_orphan
) {
764 struct inode
*inode
= orphan_list_entry(l
);
766 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
767 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
768 inode
->i_mode
, inode
->i_nlink
,
773 static void ext4_put_super(struct super_block
*sb
)
775 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
776 struct ext4_super_block
*es
= sbi
->s_es
;
779 ext4_unregister_li_request(sb
);
780 dquot_disable(sb
, -1, DQUOT_USAGE_ENABLED
| DQUOT_LIMITS_ENABLED
);
782 flush_workqueue(sbi
->rsv_conversion_wq
);
783 destroy_workqueue(sbi
->rsv_conversion_wq
);
785 if (sbi
->s_journal
) {
786 err
= jbd2_journal_destroy(sbi
->s_journal
);
787 sbi
->s_journal
= NULL
;
789 ext4_abort(sb
, "Couldn't clean up the journal");
792 ext4_es_unregister_shrinker(sbi
);
793 del_timer_sync(&sbi
->s_err_report
);
794 ext4_release_system_zone(sb
);
796 ext4_ext_release(sb
);
797 ext4_xattr_put_super(sb
);
799 if (!(sb
->s_flags
& MS_RDONLY
)) {
800 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
801 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
803 if (!(sb
->s_flags
& MS_RDONLY
))
804 ext4_commit_super(sb
, 1);
807 remove_proc_entry("options", sbi
->s_proc
);
808 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
810 kobject_del(&sbi
->s_kobj
);
812 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
813 brelse(sbi
->s_group_desc
[i
]);
814 ext4_kvfree(sbi
->s_group_desc
);
815 ext4_kvfree(sbi
->s_flex_groups
);
816 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
817 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
818 percpu_counter_destroy(&sbi
->s_dirs_counter
);
819 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
822 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
823 kfree(sbi
->s_qf_names
[i
]);
826 /* Debugging code just in case the in-memory inode orphan list
827 * isn't empty. The on-disk one can be non-empty if we've
828 * detected an error and taken the fs readonly, but the
829 * in-memory list had better be clean by this point. */
830 if (!list_empty(&sbi
->s_orphan
))
831 dump_orphan_list(sb
, sbi
);
832 J_ASSERT(list_empty(&sbi
->s_orphan
));
834 invalidate_bdev(sb
->s_bdev
);
835 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
837 * Invalidate the journal device's buffers. We don't want them
838 * floating about in memory - the physical journal device may
839 * hotswapped, and it breaks the `ro-after' testing code.
841 sync_blockdev(sbi
->journal_bdev
);
842 invalidate_bdev(sbi
->journal_bdev
);
843 ext4_blkdev_remove(sbi
);
845 if (sbi
->s_mb_cache
) {
846 ext4_xattr_destroy_cache(sbi
->s_mb_cache
);
847 sbi
->s_mb_cache
= NULL
;
850 kthread_stop(sbi
->s_mmp_tsk
);
851 sb
->s_fs_info
= NULL
;
853 * Now that we are completely done shutting down the
854 * superblock, we need to actually destroy the kobject.
856 kobject_put(&sbi
->s_kobj
);
857 wait_for_completion(&sbi
->s_kobj_unregister
);
858 if (sbi
->s_chksum_driver
)
859 crypto_free_shash(sbi
->s_chksum_driver
);
860 kfree(sbi
->s_blockgroup_lock
);
864 static struct kmem_cache
*ext4_inode_cachep
;
867 * Called inside transaction, so use GFP_NOFS
869 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
871 struct ext4_inode_info
*ei
;
873 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
877 ei
->vfs_inode
.i_version
= 1;
878 spin_lock_init(&ei
->i_raw_lock
);
879 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
880 spin_lock_init(&ei
->i_prealloc_lock
);
881 ext4_es_init_tree(&ei
->i_es_tree
);
882 rwlock_init(&ei
->i_es_lock
);
883 INIT_LIST_HEAD(&ei
->i_es_lru
);
886 ei
->i_touch_when
= 0;
887 ei
->i_reserved_data_blocks
= 0;
888 ei
->i_reserved_meta_blocks
= 0;
889 ei
->i_allocated_meta_blocks
= 0;
890 ei
->i_da_metadata_calc_len
= 0;
891 ei
->i_da_metadata_calc_last_lblock
= 0;
892 spin_lock_init(&(ei
->i_block_reservation_lock
));
894 ei
->i_reserved_quota
= 0;
895 memset(&ei
->i_dquot
, 0, sizeof(ei
->i_dquot
));
898 INIT_LIST_HEAD(&ei
->i_rsv_conversion_list
);
899 spin_lock_init(&ei
->i_completed_io_lock
);
901 ei
->i_datasync_tid
= 0;
902 atomic_set(&ei
->i_ioend_count
, 0);
903 atomic_set(&ei
->i_unwritten
, 0);
904 INIT_WORK(&ei
->i_rsv_conversion_work
, ext4_end_io_rsv_work
);
906 return &ei
->vfs_inode
;
909 static int ext4_drop_inode(struct inode
*inode
)
911 int drop
= generic_drop_inode(inode
);
913 trace_ext4_drop_inode(inode
, drop
);
917 static void ext4_i_callback(struct rcu_head
*head
)
919 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
920 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
923 static void ext4_destroy_inode(struct inode
*inode
)
925 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
926 ext4_msg(inode
->i_sb
, KERN_ERR
,
927 "Inode %lu (%p): orphan list check failed!",
928 inode
->i_ino
, EXT4_I(inode
));
929 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
930 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
934 call_rcu(&inode
->i_rcu
, ext4_i_callback
);
937 static void init_once(void *foo
)
939 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
941 INIT_LIST_HEAD(&ei
->i_orphan
);
942 init_rwsem(&ei
->xattr_sem
);
943 init_rwsem(&ei
->i_data_sem
);
944 inode_init_once(&ei
->vfs_inode
);
947 static int __init
init_inodecache(void)
949 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
950 sizeof(struct ext4_inode_info
),
951 0, (SLAB_RECLAIM_ACCOUNT
|
954 if (ext4_inode_cachep
== NULL
)
959 static void destroy_inodecache(void)
962 * Make sure all delayed rcu free inodes are flushed before we
966 kmem_cache_destroy(ext4_inode_cachep
);
969 void ext4_clear_inode(struct inode
*inode
)
971 invalidate_inode_buffers(inode
);
974 ext4_discard_preallocations(inode
);
975 ext4_es_remove_extent(inode
, 0, EXT_MAX_BLOCKS
);
976 ext4_es_lru_del(inode
);
977 if (EXT4_I(inode
)->jinode
) {
978 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode
),
979 EXT4_I(inode
)->jinode
);
980 jbd2_free_inode(EXT4_I(inode
)->jinode
);
981 EXT4_I(inode
)->jinode
= NULL
;
985 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
986 u64 ino
, u32 generation
)
990 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
991 return ERR_PTR(-ESTALE
);
992 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
993 return ERR_PTR(-ESTALE
);
995 /* iget isn't really right if the inode is currently unallocated!!
997 * ext4_read_inode will return a bad_inode if the inode had been
998 * deleted, so we should be safe.
1000 * Currently we don't know the generation for parent directory, so
1001 * a generation of 0 means "accept any"
1003 inode
= ext4_iget_normal(sb
, ino
);
1005 return ERR_CAST(inode
);
1006 if (generation
&& inode
->i_generation
!= generation
) {
1008 return ERR_PTR(-ESTALE
);
1014 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1015 int fh_len
, int fh_type
)
1017 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
1018 ext4_nfs_get_inode
);
1021 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1022 int fh_len
, int fh_type
)
1024 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1025 ext4_nfs_get_inode
);
1029 * Try to release metadata pages (indirect blocks, directories) which are
1030 * mapped via the block device. Since these pages could have journal heads
1031 * which would prevent try_to_free_buffers() from freeing them, we must use
1032 * jbd2 layer's try_to_free_buffers() function to release them.
1034 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
1037 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
1039 WARN_ON(PageChecked(page
));
1040 if (!page_has_buffers(page
))
1043 return jbd2_journal_try_to_free_buffers(journal
, page
,
1044 wait
& ~__GFP_WAIT
);
1045 return try_to_free_buffers(page
);
1049 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1050 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1052 static int ext4_write_dquot(struct dquot
*dquot
);
1053 static int ext4_acquire_dquot(struct dquot
*dquot
);
1054 static int ext4_release_dquot(struct dquot
*dquot
);
1055 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
1056 static int ext4_write_info(struct super_block
*sb
, int type
);
1057 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
1059 static int ext4_quota_on_sysfile(struct super_block
*sb
, int type
,
1061 static int ext4_quota_off(struct super_block
*sb
, int type
);
1062 static int ext4_quota_off_sysfile(struct super_block
*sb
, int type
);
1063 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
1064 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
1065 size_t len
, loff_t off
);
1066 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
1067 const char *data
, size_t len
, loff_t off
);
1068 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
1069 unsigned int flags
);
1070 static int ext4_enable_quotas(struct super_block
*sb
);
1072 static struct dquot
**ext4_get_dquots(struct inode
*inode
)
1074 return EXT4_I(inode
)->i_dquot
;
1077 static const struct dquot_operations ext4_quota_operations
= {
1078 .get_reserved_space
= ext4_get_reserved_space
,
1079 .write_dquot
= ext4_write_dquot
,
1080 .acquire_dquot
= ext4_acquire_dquot
,
1081 .release_dquot
= ext4_release_dquot
,
1082 .mark_dirty
= ext4_mark_dquot_dirty
,
1083 .write_info
= ext4_write_info
,
1084 .alloc_dquot
= dquot_alloc
,
1085 .destroy_dquot
= dquot_destroy
,
1088 static const struct quotactl_ops ext4_qctl_operations
= {
1089 .quota_on
= ext4_quota_on
,
1090 .quota_off
= ext4_quota_off
,
1091 .quota_sync
= dquot_quota_sync
,
1092 .get_info
= dquot_get_dqinfo
,
1093 .set_info
= dquot_set_dqinfo
,
1094 .get_dqblk
= dquot_get_dqblk
,
1095 .set_dqblk
= dquot_set_dqblk
1098 static const struct quotactl_ops ext4_qctl_sysfile_operations
= {
1099 .quota_on_meta
= ext4_quota_on_sysfile
,
1100 .quota_off
= ext4_quota_off_sysfile
,
1101 .quota_sync
= dquot_quota_sync
,
1102 .get_info
= dquot_get_dqinfo
,
1103 .set_info
= dquot_set_dqinfo
,
1104 .get_dqblk
= dquot_get_dqblk
,
1105 .set_dqblk
= dquot_set_dqblk
1109 static const struct super_operations ext4_sops
= {
1110 .alloc_inode
= ext4_alloc_inode
,
1111 .destroy_inode
= ext4_destroy_inode
,
1112 .write_inode
= ext4_write_inode
,
1113 .dirty_inode
= ext4_dirty_inode
,
1114 .drop_inode
= ext4_drop_inode
,
1115 .evict_inode
= ext4_evict_inode
,
1116 .put_super
= ext4_put_super
,
1117 .sync_fs
= ext4_sync_fs
,
1118 .freeze_fs
= ext4_freeze
,
1119 .unfreeze_fs
= ext4_unfreeze
,
1120 .statfs
= ext4_statfs
,
1121 .remount_fs
= ext4_remount
,
1122 .show_options
= ext4_show_options
,
1124 .quota_read
= ext4_quota_read
,
1125 .quota_write
= ext4_quota_write
,
1126 .get_dquots
= ext4_get_dquots
,
1128 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1131 static const struct export_operations ext4_export_ops
= {
1132 .fh_to_dentry
= ext4_fh_to_dentry
,
1133 .fh_to_parent
= ext4_fh_to_parent
,
1134 .get_parent
= ext4_get_parent
,
1138 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1139 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1140 Opt_nouid32
, Opt_debug
, Opt_removed
,
1141 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1142 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
,
1143 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
, Opt_journal_dev
,
1144 Opt_journal_path
, Opt_journal_checksum
, Opt_journal_async_commit
,
1145 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1146 Opt_data_err_abort
, Opt_data_err_ignore
,
1147 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1148 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
1149 Opt_noquota
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
1150 Opt_usrquota
, Opt_grpquota
, Opt_i_version
,
1151 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
, Opt_mblk_io_submit
,
1152 Opt_nomblk_io_submit
, Opt_block_validity
, Opt_noblock_validity
,
1153 Opt_inode_readahead_blks
, Opt_journal_ioprio
,
1154 Opt_dioread_nolock
, Opt_dioread_lock
,
1155 Opt_discard
, Opt_nodiscard
, Opt_init_itable
, Opt_noinit_itable
,
1156 Opt_max_dir_size_kb
,
1159 static const match_table_t tokens
= {
1160 {Opt_bsd_df
, "bsddf"},
1161 {Opt_minix_df
, "minixdf"},
1162 {Opt_grpid
, "grpid"},
1163 {Opt_grpid
, "bsdgroups"},
1164 {Opt_nogrpid
, "nogrpid"},
1165 {Opt_nogrpid
, "sysvgroups"},
1166 {Opt_resgid
, "resgid=%u"},
1167 {Opt_resuid
, "resuid=%u"},
1169 {Opt_err_cont
, "errors=continue"},
1170 {Opt_err_panic
, "errors=panic"},
1171 {Opt_err_ro
, "errors=remount-ro"},
1172 {Opt_nouid32
, "nouid32"},
1173 {Opt_debug
, "debug"},
1174 {Opt_removed
, "oldalloc"},
1175 {Opt_removed
, "orlov"},
1176 {Opt_user_xattr
, "user_xattr"},
1177 {Opt_nouser_xattr
, "nouser_xattr"},
1179 {Opt_noacl
, "noacl"},
1180 {Opt_noload
, "norecovery"},
1181 {Opt_noload
, "noload"},
1182 {Opt_removed
, "nobh"},
1183 {Opt_removed
, "bh"},
1184 {Opt_commit
, "commit=%u"},
1185 {Opt_min_batch_time
, "min_batch_time=%u"},
1186 {Opt_max_batch_time
, "max_batch_time=%u"},
1187 {Opt_journal_dev
, "journal_dev=%u"},
1188 {Opt_journal_path
, "journal_path=%s"},
1189 {Opt_journal_checksum
, "journal_checksum"},
1190 {Opt_journal_async_commit
, "journal_async_commit"},
1191 {Opt_abort
, "abort"},
1192 {Opt_data_journal
, "data=journal"},
1193 {Opt_data_ordered
, "data=ordered"},
1194 {Opt_data_writeback
, "data=writeback"},
1195 {Opt_data_err_abort
, "data_err=abort"},
1196 {Opt_data_err_ignore
, "data_err=ignore"},
1197 {Opt_offusrjquota
, "usrjquota="},
1198 {Opt_usrjquota
, "usrjquota=%s"},
1199 {Opt_offgrpjquota
, "grpjquota="},
1200 {Opt_grpjquota
, "grpjquota=%s"},
1201 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1202 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1203 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
1204 {Opt_grpquota
, "grpquota"},
1205 {Opt_noquota
, "noquota"},
1206 {Opt_quota
, "quota"},
1207 {Opt_usrquota
, "usrquota"},
1208 {Opt_barrier
, "barrier=%u"},
1209 {Opt_barrier
, "barrier"},
1210 {Opt_nobarrier
, "nobarrier"},
1211 {Opt_i_version
, "i_version"},
1212 {Opt_stripe
, "stripe=%u"},
1213 {Opt_delalloc
, "delalloc"},
1214 {Opt_nodelalloc
, "nodelalloc"},
1215 {Opt_removed
, "mblk_io_submit"},
1216 {Opt_removed
, "nomblk_io_submit"},
1217 {Opt_block_validity
, "block_validity"},
1218 {Opt_noblock_validity
, "noblock_validity"},
1219 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1220 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1221 {Opt_auto_da_alloc
, "auto_da_alloc=%u"},
1222 {Opt_auto_da_alloc
, "auto_da_alloc"},
1223 {Opt_noauto_da_alloc
, "noauto_da_alloc"},
1224 {Opt_dioread_nolock
, "dioread_nolock"},
1225 {Opt_dioread_lock
, "dioread_lock"},
1226 {Opt_discard
, "discard"},
1227 {Opt_nodiscard
, "nodiscard"},
1228 {Opt_init_itable
, "init_itable=%u"},
1229 {Opt_init_itable
, "init_itable"},
1230 {Opt_noinit_itable
, "noinit_itable"},
1231 {Opt_max_dir_size_kb
, "max_dir_size_kb=%u"},
1232 {Opt_removed
, "check=none"}, /* mount option from ext2/3 */
1233 {Opt_removed
, "nocheck"}, /* mount option from ext2/3 */
1234 {Opt_removed
, "reservation"}, /* mount option from ext2/3 */
1235 {Opt_removed
, "noreservation"}, /* mount option from ext2/3 */
1236 {Opt_removed
, "journal=%u"}, /* mount option from ext2/3 */
1240 static ext4_fsblk_t
get_sb_block(void **data
)
1242 ext4_fsblk_t sb_block
;
1243 char *options
= (char *) *data
;
1245 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1246 return 1; /* Default location */
1249 /* TODO: use simple_strtoll with >32bit ext4 */
1250 sb_block
= simple_strtoul(options
, &options
, 0);
1251 if (*options
&& *options
!= ',') {
1252 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1256 if (*options
== ',')
1258 *data
= (void *) options
;
1263 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1264 static char deprecated_msg
[] = "Mount option \"%s\" will be removed by %s\n"
1265 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1268 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
1270 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1274 if (sb_any_quota_loaded(sb
) &&
1275 !sbi
->s_qf_names
[qtype
]) {
1276 ext4_msg(sb
, KERN_ERR
,
1277 "Cannot change journaled "
1278 "quota options when quota turned on");
1281 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
)) {
1282 ext4_msg(sb
, KERN_ERR
, "Cannot set journaled quota options "
1283 "when QUOTA feature is enabled");
1286 qname
= match_strdup(args
);
1288 ext4_msg(sb
, KERN_ERR
,
1289 "Not enough memory for storing quotafile name");
1292 if (sbi
->s_qf_names
[qtype
]) {
1293 if (strcmp(sbi
->s_qf_names
[qtype
], qname
) == 0)
1296 ext4_msg(sb
, KERN_ERR
,
1297 "%s quota file already specified",
1301 if (strchr(qname
, '/')) {
1302 ext4_msg(sb
, KERN_ERR
,
1303 "quotafile must be on filesystem root");
1306 sbi
->s_qf_names
[qtype
] = qname
;
1314 static int clear_qf_name(struct super_block
*sb
, int qtype
)
1317 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1319 if (sb_any_quota_loaded(sb
) &&
1320 sbi
->s_qf_names
[qtype
]) {
1321 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
1322 " when quota turned on");
1325 kfree(sbi
->s_qf_names
[qtype
]);
1326 sbi
->s_qf_names
[qtype
] = NULL
;
1331 #define MOPT_SET 0x0001
1332 #define MOPT_CLEAR 0x0002
1333 #define MOPT_NOSUPPORT 0x0004
1334 #define MOPT_EXPLICIT 0x0008
1335 #define MOPT_CLEAR_ERR 0x0010
1336 #define MOPT_GTE0 0x0020
1339 #define MOPT_QFMT 0x0040
1341 #define MOPT_Q MOPT_NOSUPPORT
1342 #define MOPT_QFMT MOPT_NOSUPPORT
1344 #define MOPT_DATAJ 0x0080
1345 #define MOPT_NO_EXT2 0x0100
1346 #define MOPT_NO_EXT3 0x0200
1347 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1348 #define MOPT_STRING 0x0400
1350 static const struct mount_opts
{
1354 } ext4_mount_opts
[] = {
1355 {Opt_minix_df
, EXT4_MOUNT_MINIX_DF
, MOPT_SET
},
1356 {Opt_bsd_df
, EXT4_MOUNT_MINIX_DF
, MOPT_CLEAR
},
1357 {Opt_grpid
, EXT4_MOUNT_GRPID
, MOPT_SET
},
1358 {Opt_nogrpid
, EXT4_MOUNT_GRPID
, MOPT_CLEAR
},
1359 {Opt_block_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_SET
},
1360 {Opt_noblock_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_CLEAR
},
1361 {Opt_dioread_nolock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1362 MOPT_EXT4_ONLY
| MOPT_SET
},
1363 {Opt_dioread_lock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1364 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1365 {Opt_discard
, EXT4_MOUNT_DISCARD
, MOPT_SET
},
1366 {Opt_nodiscard
, EXT4_MOUNT_DISCARD
, MOPT_CLEAR
},
1367 {Opt_delalloc
, EXT4_MOUNT_DELALLOC
,
1368 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1369 {Opt_nodelalloc
, EXT4_MOUNT_DELALLOC
,
1370 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1371 {Opt_journal_checksum
, EXT4_MOUNT_JOURNAL_CHECKSUM
,
1372 MOPT_EXT4_ONLY
| MOPT_SET
},
1373 {Opt_journal_async_commit
, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT
|
1374 EXT4_MOUNT_JOURNAL_CHECKSUM
),
1375 MOPT_EXT4_ONLY
| MOPT_SET
},
1376 {Opt_noload
, EXT4_MOUNT_NOLOAD
, MOPT_NO_EXT2
| MOPT_SET
},
1377 {Opt_err_panic
, EXT4_MOUNT_ERRORS_PANIC
, MOPT_SET
| MOPT_CLEAR_ERR
},
1378 {Opt_err_ro
, EXT4_MOUNT_ERRORS_RO
, MOPT_SET
| MOPT_CLEAR_ERR
},
1379 {Opt_err_cont
, EXT4_MOUNT_ERRORS_CONT
, MOPT_SET
| MOPT_CLEAR_ERR
},
1380 {Opt_data_err_abort
, EXT4_MOUNT_DATA_ERR_ABORT
,
1381 MOPT_NO_EXT2
| MOPT_SET
},
1382 {Opt_data_err_ignore
, EXT4_MOUNT_DATA_ERR_ABORT
,
1383 MOPT_NO_EXT2
| MOPT_CLEAR
},
1384 {Opt_barrier
, EXT4_MOUNT_BARRIER
, MOPT_SET
},
1385 {Opt_nobarrier
, EXT4_MOUNT_BARRIER
, MOPT_CLEAR
},
1386 {Opt_noauto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_SET
},
1387 {Opt_auto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_CLEAR
},
1388 {Opt_noinit_itable
, EXT4_MOUNT_INIT_INODE_TABLE
, MOPT_CLEAR
},
1389 {Opt_commit
, 0, MOPT_GTE0
},
1390 {Opt_max_batch_time
, 0, MOPT_GTE0
},
1391 {Opt_min_batch_time
, 0, MOPT_GTE0
},
1392 {Opt_inode_readahead_blks
, 0, MOPT_GTE0
},
1393 {Opt_init_itable
, 0, MOPT_GTE0
},
1394 {Opt_stripe
, 0, MOPT_GTE0
},
1395 {Opt_resuid
, 0, MOPT_GTE0
},
1396 {Opt_resgid
, 0, MOPT_GTE0
},
1397 {Opt_journal_dev
, 0, MOPT_GTE0
},
1398 {Opt_journal_path
, 0, MOPT_STRING
},
1399 {Opt_journal_ioprio
, 0, MOPT_GTE0
},
1400 {Opt_data_journal
, EXT4_MOUNT_JOURNAL_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1401 {Opt_data_ordered
, EXT4_MOUNT_ORDERED_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1402 {Opt_data_writeback
, EXT4_MOUNT_WRITEBACK_DATA
,
1403 MOPT_NO_EXT2
| MOPT_DATAJ
},
1404 {Opt_user_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_SET
},
1405 {Opt_nouser_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_CLEAR
},
1406 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1407 {Opt_acl
, EXT4_MOUNT_POSIX_ACL
, MOPT_SET
},
1408 {Opt_noacl
, EXT4_MOUNT_POSIX_ACL
, MOPT_CLEAR
},
1410 {Opt_acl
, 0, MOPT_NOSUPPORT
},
1411 {Opt_noacl
, 0, MOPT_NOSUPPORT
},
1413 {Opt_nouid32
, EXT4_MOUNT_NO_UID32
, MOPT_SET
},
1414 {Opt_debug
, EXT4_MOUNT_DEBUG
, MOPT_SET
},
1415 {Opt_quota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
, MOPT_SET
| MOPT_Q
},
1416 {Opt_usrquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
,
1418 {Opt_grpquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_GRPQUOTA
,
1420 {Opt_noquota
, (EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
|
1421 EXT4_MOUNT_GRPQUOTA
), MOPT_CLEAR
| MOPT_Q
},
1422 {Opt_usrjquota
, 0, MOPT_Q
},
1423 {Opt_grpjquota
, 0, MOPT_Q
},
1424 {Opt_offusrjquota
, 0, MOPT_Q
},
1425 {Opt_offgrpjquota
, 0, MOPT_Q
},
1426 {Opt_jqfmt_vfsold
, QFMT_VFS_OLD
, MOPT_QFMT
},
1427 {Opt_jqfmt_vfsv0
, QFMT_VFS_V0
, MOPT_QFMT
},
1428 {Opt_jqfmt_vfsv1
, QFMT_VFS_V1
, MOPT_QFMT
},
1429 {Opt_max_dir_size_kb
, 0, MOPT_GTE0
},
1433 static int handle_mount_opt(struct super_block
*sb
, char *opt
, int token
,
1434 substring_t
*args
, unsigned long *journal_devnum
,
1435 unsigned int *journal_ioprio
, int is_remount
)
1437 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1438 const struct mount_opts
*m
;
1444 if (token
== Opt_usrjquota
)
1445 return set_qf_name(sb
, USRQUOTA
, &args
[0]);
1446 else if (token
== Opt_grpjquota
)
1447 return set_qf_name(sb
, GRPQUOTA
, &args
[0]);
1448 else if (token
== Opt_offusrjquota
)
1449 return clear_qf_name(sb
, USRQUOTA
);
1450 else if (token
== Opt_offgrpjquota
)
1451 return clear_qf_name(sb
, GRPQUOTA
);
1455 case Opt_nouser_xattr
:
1456 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, opt
, "3.5");
1459 return 1; /* handled by get_sb_block() */
1461 ext4_msg(sb
, KERN_WARNING
, "Ignoring removed %s option", opt
);
1464 sbi
->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
1467 sb
->s_flags
|= MS_I_VERSION
;
1471 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++)
1472 if (token
== m
->token
)
1475 if (m
->token
== Opt_err
) {
1476 ext4_msg(sb
, KERN_ERR
, "Unrecognized mount option \"%s\" "
1477 "or missing value", opt
);
1481 if ((m
->flags
& MOPT_NO_EXT2
) && IS_EXT2_SB(sb
)) {
1482 ext4_msg(sb
, KERN_ERR
,
1483 "Mount option \"%s\" incompatible with ext2", opt
);
1486 if ((m
->flags
& MOPT_NO_EXT3
) && IS_EXT3_SB(sb
)) {
1487 ext4_msg(sb
, KERN_ERR
,
1488 "Mount option \"%s\" incompatible with ext3", opt
);
1492 if (args
->from
&& !(m
->flags
& MOPT_STRING
) && match_int(args
, &arg
))
1494 if (args
->from
&& (m
->flags
& MOPT_GTE0
) && (arg
< 0))
1496 if (m
->flags
& MOPT_EXPLICIT
)
1497 set_opt2(sb
, EXPLICIT_DELALLOC
);
1498 if (m
->flags
& MOPT_CLEAR_ERR
)
1499 clear_opt(sb
, ERRORS_MASK
);
1500 if (token
== Opt_noquota
&& sb_any_quota_loaded(sb
)) {
1501 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
1502 "options when quota turned on");
1506 if (m
->flags
& MOPT_NOSUPPORT
) {
1507 ext4_msg(sb
, KERN_ERR
, "%s option not supported", opt
);
1508 } else if (token
== Opt_commit
) {
1510 arg
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1511 sbi
->s_commit_interval
= HZ
* arg
;
1512 } else if (token
== Opt_max_batch_time
) {
1513 sbi
->s_max_batch_time
= arg
;
1514 } else if (token
== Opt_min_batch_time
) {
1515 sbi
->s_min_batch_time
= arg
;
1516 } else if (token
== Opt_inode_readahead_blks
) {
1517 if (arg
&& (arg
> (1 << 30) || !is_power_of_2(arg
))) {
1518 ext4_msg(sb
, KERN_ERR
,
1519 "EXT4-fs: inode_readahead_blks must be "
1520 "0 or a power of 2 smaller than 2^31");
1523 sbi
->s_inode_readahead_blks
= arg
;
1524 } else if (token
== Opt_init_itable
) {
1525 set_opt(sb
, INIT_INODE_TABLE
);
1527 arg
= EXT4_DEF_LI_WAIT_MULT
;
1528 sbi
->s_li_wait_mult
= arg
;
1529 } else if (token
== Opt_max_dir_size_kb
) {
1530 sbi
->s_max_dir_size_kb
= arg
;
1531 } else if (token
== Opt_stripe
) {
1532 sbi
->s_stripe
= arg
;
1533 } else if (token
== Opt_resuid
) {
1534 uid
= make_kuid(current_user_ns(), arg
);
1535 if (!uid_valid(uid
)) {
1536 ext4_msg(sb
, KERN_ERR
, "Invalid uid value %d", arg
);
1539 sbi
->s_resuid
= uid
;
1540 } else if (token
== Opt_resgid
) {
1541 gid
= make_kgid(current_user_ns(), arg
);
1542 if (!gid_valid(gid
)) {
1543 ext4_msg(sb
, KERN_ERR
, "Invalid gid value %d", arg
);
1546 sbi
->s_resgid
= gid
;
1547 } else if (token
== Opt_journal_dev
) {
1549 ext4_msg(sb
, KERN_ERR
,
1550 "Cannot specify journal on remount");
1553 *journal_devnum
= arg
;
1554 } else if (token
== Opt_journal_path
) {
1556 struct inode
*journal_inode
;
1561 ext4_msg(sb
, KERN_ERR
,
1562 "Cannot specify journal on remount");
1565 journal_path
= match_strdup(&args
[0]);
1566 if (!journal_path
) {
1567 ext4_msg(sb
, KERN_ERR
, "error: could not dup "
1568 "journal device string");
1572 error
= kern_path(journal_path
, LOOKUP_FOLLOW
, &path
);
1574 ext4_msg(sb
, KERN_ERR
, "error: could not find "
1575 "journal device path: error %d", error
);
1576 kfree(journal_path
);
1580 journal_inode
= path
.dentry
->d_inode
;
1581 if (!S_ISBLK(journal_inode
->i_mode
)) {
1582 ext4_msg(sb
, KERN_ERR
, "error: journal path %s "
1583 "is not a block device", journal_path
);
1585 kfree(journal_path
);
1589 *journal_devnum
= new_encode_dev(journal_inode
->i_rdev
);
1591 kfree(journal_path
);
1592 } else if (token
== Opt_journal_ioprio
) {
1594 ext4_msg(sb
, KERN_ERR
, "Invalid journal IO priority"
1599 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
, arg
);
1600 } else if (m
->flags
& MOPT_DATAJ
) {
1602 if (!sbi
->s_journal
)
1603 ext4_msg(sb
, KERN_WARNING
, "Remounting file system with no journal so ignoring journalled data option");
1604 else if (test_opt(sb
, DATA_FLAGS
) != m
->mount_opt
) {
1605 ext4_msg(sb
, KERN_ERR
,
1606 "Cannot change data mode on remount");
1610 clear_opt(sb
, DATA_FLAGS
);
1611 sbi
->s_mount_opt
|= m
->mount_opt
;
1614 } else if (m
->flags
& MOPT_QFMT
) {
1615 if (sb_any_quota_loaded(sb
) &&
1616 sbi
->s_jquota_fmt
!= m
->mount_opt
) {
1617 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled "
1618 "quota options when quota turned on");
1621 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
1622 EXT4_FEATURE_RO_COMPAT_QUOTA
)) {
1623 ext4_msg(sb
, KERN_ERR
,
1624 "Cannot set journaled quota options "
1625 "when QUOTA feature is enabled");
1628 sbi
->s_jquota_fmt
= m
->mount_opt
;
1633 if (m
->flags
& MOPT_CLEAR
)
1635 else if (unlikely(!(m
->flags
& MOPT_SET
))) {
1636 ext4_msg(sb
, KERN_WARNING
,
1637 "buggy handling of option %s", opt
);
1642 sbi
->s_mount_opt
|= m
->mount_opt
;
1644 sbi
->s_mount_opt
&= ~m
->mount_opt
;
1649 static int parse_options(char *options
, struct super_block
*sb
,
1650 unsigned long *journal_devnum
,
1651 unsigned int *journal_ioprio
,
1654 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1656 substring_t args
[MAX_OPT_ARGS
];
1662 while ((p
= strsep(&options
, ",")) != NULL
) {
1666 * Initialize args struct so we know whether arg was
1667 * found; some options take optional arguments.
1669 args
[0].to
= args
[0].from
= NULL
;
1670 token
= match_token(p
, tokens
, args
);
1671 if (handle_mount_opt(sb
, p
, token
, args
, journal_devnum
,
1672 journal_ioprio
, is_remount
) < 0)
1676 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
) &&
1677 (test_opt(sb
, USRQUOTA
) || test_opt(sb
, GRPQUOTA
))) {
1678 ext4_msg(sb
, KERN_ERR
, "Cannot set quota options when QUOTA "
1679 "feature is enabled");
1682 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1683 if (test_opt(sb
, USRQUOTA
) && sbi
->s_qf_names
[USRQUOTA
])
1684 clear_opt(sb
, USRQUOTA
);
1686 if (test_opt(sb
, GRPQUOTA
) && sbi
->s_qf_names
[GRPQUOTA
])
1687 clear_opt(sb
, GRPQUOTA
);
1689 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
1690 ext4_msg(sb
, KERN_ERR
, "old and new quota "
1695 if (!sbi
->s_jquota_fmt
) {
1696 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1702 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
1704 BLOCK_SIZE
<< le32_to_cpu(sbi
->s_es
->s_log_block_size
);
1706 if (blocksize
< PAGE_CACHE_SIZE
) {
1707 ext4_msg(sb
, KERN_ERR
, "can't mount with "
1708 "dioread_nolock if block size != PAGE_SIZE");
1715 static inline void ext4_show_quota_options(struct seq_file
*seq
,
1716 struct super_block
*sb
)
1718 #if defined(CONFIG_QUOTA)
1719 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1721 if (sbi
->s_jquota_fmt
) {
1724 switch (sbi
->s_jquota_fmt
) {
1735 seq_printf(seq
, ",jqfmt=%s", fmtname
);
1738 if (sbi
->s_qf_names
[USRQUOTA
])
1739 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
1741 if (sbi
->s_qf_names
[GRPQUOTA
])
1742 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
1746 static const char *token2str(int token
)
1748 const struct match_token
*t
;
1750 for (t
= tokens
; t
->token
!= Opt_err
; t
++)
1751 if (t
->token
== token
&& !strchr(t
->pattern
, '='))
1758 * - it's set to a non-default value OR
1759 * - if the per-sb default is different from the global default
1761 static int _ext4_show_options(struct seq_file
*seq
, struct super_block
*sb
,
1764 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1765 struct ext4_super_block
*es
= sbi
->s_es
;
1766 int def_errors
, def_mount_opt
= nodefs
? 0 : sbi
->s_def_mount_opt
;
1767 const struct mount_opts
*m
;
1768 char sep
= nodefs
? '\n' : ',';
1770 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1771 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1773 if (sbi
->s_sb_block
!= 1)
1774 SEQ_OPTS_PRINT("sb=%llu", sbi
->s_sb_block
);
1776 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++) {
1777 int want_set
= m
->flags
& MOPT_SET
;
1778 if (((m
->flags
& (MOPT_SET
|MOPT_CLEAR
)) == 0) ||
1779 (m
->flags
& MOPT_CLEAR_ERR
))
1781 if (!(m
->mount_opt
& (sbi
->s_mount_opt
^ def_mount_opt
)))
1782 continue; /* skip if same as the default */
1784 (sbi
->s_mount_opt
& m
->mount_opt
) != m
->mount_opt
) ||
1785 (!want_set
&& (sbi
->s_mount_opt
& m
->mount_opt
)))
1786 continue; /* select Opt_noFoo vs Opt_Foo */
1787 SEQ_OPTS_PRINT("%s", token2str(m
->token
));
1790 if (nodefs
|| !uid_eq(sbi
->s_resuid
, make_kuid(&init_user_ns
, EXT4_DEF_RESUID
)) ||
1791 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
)
1792 SEQ_OPTS_PRINT("resuid=%u",
1793 from_kuid_munged(&init_user_ns
, sbi
->s_resuid
));
1794 if (nodefs
|| !gid_eq(sbi
->s_resgid
, make_kgid(&init_user_ns
, EXT4_DEF_RESGID
)) ||
1795 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
)
1796 SEQ_OPTS_PRINT("resgid=%u",
1797 from_kgid_munged(&init_user_ns
, sbi
->s_resgid
));
1798 def_errors
= nodefs
? -1 : le16_to_cpu(es
->s_errors
);
1799 if (test_opt(sb
, ERRORS_RO
) && def_errors
!= EXT4_ERRORS_RO
)
1800 SEQ_OPTS_PUTS("errors=remount-ro");
1801 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
1802 SEQ_OPTS_PUTS("errors=continue");
1803 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
1804 SEQ_OPTS_PUTS("errors=panic");
1805 if (nodefs
|| sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
)
1806 SEQ_OPTS_PRINT("commit=%lu", sbi
->s_commit_interval
/ HZ
);
1807 if (nodefs
|| sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
)
1808 SEQ_OPTS_PRINT("min_batch_time=%u", sbi
->s_min_batch_time
);
1809 if (nodefs
|| sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
)
1810 SEQ_OPTS_PRINT("max_batch_time=%u", sbi
->s_max_batch_time
);
1811 if (sb
->s_flags
& MS_I_VERSION
)
1812 SEQ_OPTS_PUTS("i_version");
1813 if (nodefs
|| sbi
->s_stripe
)
1814 SEQ_OPTS_PRINT("stripe=%lu", sbi
->s_stripe
);
1815 if (EXT4_MOUNT_DATA_FLAGS
& (sbi
->s_mount_opt
^ def_mount_opt
)) {
1816 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
1817 SEQ_OPTS_PUTS("data=journal");
1818 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
1819 SEQ_OPTS_PUTS("data=ordered");
1820 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
1821 SEQ_OPTS_PUTS("data=writeback");
1824 sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
1825 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1826 sbi
->s_inode_readahead_blks
);
1828 if (nodefs
|| (test_opt(sb
, INIT_INODE_TABLE
) &&
1829 (sbi
->s_li_wait_mult
!= EXT4_DEF_LI_WAIT_MULT
)))
1830 SEQ_OPTS_PRINT("init_itable=%u", sbi
->s_li_wait_mult
);
1831 if (nodefs
|| sbi
->s_max_dir_size_kb
)
1832 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi
->s_max_dir_size_kb
);
1834 ext4_show_quota_options(seq
, sb
);
1838 static int ext4_show_options(struct seq_file
*seq
, struct dentry
*root
)
1840 return _ext4_show_options(seq
, root
->d_sb
, 0);
1843 static int options_seq_show(struct seq_file
*seq
, void *offset
)
1845 struct super_block
*sb
= seq
->private;
1848 seq_puts(seq
, (sb
->s_flags
& MS_RDONLY
) ? "ro" : "rw");
1849 rc
= _ext4_show_options(seq
, sb
, 1);
1850 seq_puts(seq
, "\n");
1854 static int options_open_fs(struct inode
*inode
, struct file
*file
)
1856 return single_open(file
, options_seq_show
, PDE_DATA(inode
));
1859 static const struct file_operations ext4_seq_options_fops
= {
1860 .owner
= THIS_MODULE
,
1861 .open
= options_open_fs
,
1863 .llseek
= seq_lseek
,
1864 .release
= single_release
,
1867 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
1870 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1873 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
1874 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
1875 "forcing read-only mode");
1880 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
1881 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
1882 "running e2fsck is recommended");
1883 else if (sbi
->s_mount_state
& EXT4_ERROR_FS
)
1884 ext4_msg(sb
, KERN_WARNING
,
1885 "warning: mounting fs with errors, "
1886 "running e2fsck is recommended");
1887 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) > 0 &&
1888 le16_to_cpu(es
->s_mnt_count
) >=
1889 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1890 ext4_msg(sb
, KERN_WARNING
,
1891 "warning: maximal mount count reached, "
1892 "running e2fsck is recommended");
1893 else if (le32_to_cpu(es
->s_checkinterval
) &&
1894 (le32_to_cpu(es
->s_lastcheck
) +
1895 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1896 ext4_msg(sb
, KERN_WARNING
,
1897 "warning: checktime reached, "
1898 "running e2fsck is recommended");
1899 if (!sbi
->s_journal
)
1900 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
1901 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1902 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
1903 le16_add_cpu(&es
->s_mnt_count
, 1);
1904 es
->s_mtime
= cpu_to_le32(get_seconds());
1905 ext4_update_dynamic_rev(sb
);
1907 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
1909 ext4_commit_super(sb
, 1);
1911 if (test_opt(sb
, DEBUG
))
1912 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
1913 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1915 sbi
->s_groups_count
,
1916 EXT4_BLOCKS_PER_GROUP(sb
),
1917 EXT4_INODES_PER_GROUP(sb
),
1918 sbi
->s_mount_opt
, sbi
->s_mount_opt2
);
1920 cleancache_init_fs(sb
);
1924 int ext4_alloc_flex_bg_array(struct super_block
*sb
, ext4_group_t ngroup
)
1926 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1927 struct flex_groups
*new_groups
;
1930 if (!sbi
->s_log_groups_per_flex
)
1933 size
= ext4_flex_group(sbi
, ngroup
- 1) + 1;
1934 if (size
<= sbi
->s_flex_groups_allocated
)
1937 size
= roundup_pow_of_two(size
* sizeof(struct flex_groups
));
1938 new_groups
= ext4_kvzalloc(size
, GFP_KERNEL
);
1940 ext4_msg(sb
, KERN_ERR
, "not enough memory for %d flex groups",
1941 size
/ (int) sizeof(struct flex_groups
));
1945 if (sbi
->s_flex_groups
) {
1946 memcpy(new_groups
, sbi
->s_flex_groups
,
1947 (sbi
->s_flex_groups_allocated
*
1948 sizeof(struct flex_groups
)));
1949 ext4_kvfree(sbi
->s_flex_groups
);
1951 sbi
->s_flex_groups
= new_groups
;
1952 sbi
->s_flex_groups_allocated
= size
/ sizeof(struct flex_groups
);
1956 static int ext4_fill_flex_info(struct super_block
*sb
)
1958 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1959 struct ext4_group_desc
*gdp
= NULL
;
1960 ext4_group_t flex_group
;
1963 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
1964 if (sbi
->s_log_groups_per_flex
< 1 || sbi
->s_log_groups_per_flex
> 31) {
1965 sbi
->s_log_groups_per_flex
= 0;
1969 err
= ext4_alloc_flex_bg_array(sb
, sbi
->s_groups_count
);
1973 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1974 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1976 flex_group
= ext4_flex_group(sbi
, i
);
1977 atomic_add(ext4_free_inodes_count(sb
, gdp
),
1978 &sbi
->s_flex_groups
[flex_group
].free_inodes
);
1979 atomic64_add(ext4_free_group_clusters(sb
, gdp
),
1980 &sbi
->s_flex_groups
[flex_group
].free_clusters
);
1981 atomic_add(ext4_used_dirs_count(sb
, gdp
),
1982 &sbi
->s_flex_groups
[flex_group
].used_dirs
);
1990 static __le16
ext4_group_desc_csum(struct ext4_sb_info
*sbi
, __u32 block_group
,
1991 struct ext4_group_desc
*gdp
)
1995 __le32 le_group
= cpu_to_le32(block_group
);
1997 if (ext4_has_metadata_csum(sbi
->s_sb
)) {
1998 /* Use new metadata_csum algorithm */
2002 save_csum
= gdp
->bg_checksum
;
2003 gdp
->bg_checksum
= 0;
2004 csum32
= ext4_chksum(sbi
, sbi
->s_csum_seed
, (__u8
*)&le_group
,
2006 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)gdp
,
2008 gdp
->bg_checksum
= save_csum
;
2010 crc
= csum32
& 0xFFFF;
2014 /* old crc16 code */
2015 if (!(sbi
->s_es
->s_feature_ro_compat
&
2016 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)))
2019 offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
2021 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
2022 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
2023 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
2024 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
2025 /* for checksum of struct ext4_group_desc do the rest...*/
2026 if ((sbi
->s_es
->s_feature_incompat
&
2027 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT
)) &&
2028 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
2029 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
2030 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
2034 return cpu_to_le16(crc
);
2037 int ext4_group_desc_csum_verify(struct super_block
*sb
, __u32 block_group
,
2038 struct ext4_group_desc
*gdp
)
2040 if (ext4_has_group_desc_csum(sb
) &&
2041 (gdp
->bg_checksum
!= ext4_group_desc_csum(EXT4_SB(sb
),
2048 void ext4_group_desc_csum_set(struct super_block
*sb
, __u32 block_group
,
2049 struct ext4_group_desc
*gdp
)
2051 if (!ext4_has_group_desc_csum(sb
))
2053 gdp
->bg_checksum
= ext4_group_desc_csum(EXT4_SB(sb
), block_group
, gdp
);
2056 /* Called at mount-time, super-block is locked */
2057 static int ext4_check_descriptors(struct super_block
*sb
,
2058 ext4_group_t
*first_not_zeroed
)
2060 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2061 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
2062 ext4_fsblk_t last_block
;
2063 ext4_fsblk_t block_bitmap
;
2064 ext4_fsblk_t inode_bitmap
;
2065 ext4_fsblk_t inode_table
;
2066 int flexbg_flag
= 0;
2067 ext4_group_t i
, grp
= sbi
->s_groups_count
;
2069 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
2072 ext4_debug("Checking group descriptors");
2074 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2075 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2077 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
2078 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
2080 last_block
= first_block
+
2081 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2083 if ((grp
== sbi
->s_groups_count
) &&
2084 !(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2087 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
2088 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
2089 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2090 "Block bitmap for group %u not in group "
2091 "(block %llu)!", i
, block_bitmap
);
2094 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
2095 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
2096 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2097 "Inode bitmap for group %u not in group "
2098 "(block %llu)!", i
, inode_bitmap
);
2101 inode_table
= ext4_inode_table(sb
, gdp
);
2102 if (inode_table
< first_block
||
2103 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
2104 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2105 "Inode table for group %u not in group "
2106 "(block %llu)!", i
, inode_table
);
2109 ext4_lock_group(sb
, i
);
2110 if (!ext4_group_desc_csum_verify(sb
, i
, gdp
)) {
2111 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2112 "Checksum for group %u failed (%u!=%u)",
2113 i
, le16_to_cpu(ext4_group_desc_csum(sbi
, i
,
2114 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
2115 if (!(sb
->s_flags
& MS_RDONLY
)) {
2116 ext4_unlock_group(sb
, i
);
2120 ext4_unlock_group(sb
, i
);
2122 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
2124 if (NULL
!= first_not_zeroed
)
2125 *first_not_zeroed
= grp
;
2129 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2130 * the superblock) which were deleted from all directories, but held open by
2131 * a process at the time of a crash. We walk the list and try to delete these
2132 * inodes at recovery time (only with a read-write filesystem).
2134 * In order to keep the orphan inode chain consistent during traversal (in
2135 * case of crash during recovery), we link each inode into the superblock
2136 * orphan list_head and handle it the same way as an inode deletion during
2137 * normal operation (which journals the operations for us).
2139 * We only do an iget() and an iput() on each inode, which is very safe if we
2140 * accidentally point at an in-use or already deleted inode. The worst that
2141 * can happen in this case is that we get a "bit already cleared" message from
2142 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2143 * e2fsck was run on this filesystem, and it must have already done the orphan
2144 * inode cleanup for us, so we can safely abort without any further action.
2146 static void ext4_orphan_cleanup(struct super_block
*sb
,
2147 struct ext4_super_block
*es
)
2149 unsigned int s_flags
= sb
->s_flags
;
2150 int nr_orphans
= 0, nr_truncates
= 0;
2154 if (!es
->s_last_orphan
) {
2155 jbd_debug(4, "no orphan inodes to clean up\n");
2159 if (bdev_read_only(sb
->s_bdev
)) {
2160 ext4_msg(sb
, KERN_ERR
, "write access "
2161 "unavailable, skipping orphan cleanup");
2165 /* Check if feature set would not allow a r/w mount */
2166 if (!ext4_feature_set_ok(sb
, 0)) {
2167 ext4_msg(sb
, KERN_INFO
, "Skipping orphan cleanup due to "
2168 "unknown ROCOMPAT features");
2172 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2173 /* don't clear list on RO mount w/ errors */
2174 if (es
->s_last_orphan
&& !(s_flags
& MS_RDONLY
)) {
2175 ext4_msg(sb
, KERN_INFO
, "Errors on filesystem, "
2176 "clearing orphan list.\n");
2177 es
->s_last_orphan
= 0;
2179 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2183 if (s_flags
& MS_RDONLY
) {
2184 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
2185 sb
->s_flags
&= ~MS_RDONLY
;
2188 /* Needed for iput() to work correctly and not trash data */
2189 sb
->s_flags
|= MS_ACTIVE
;
2190 /* Turn on quotas so that they are updated correctly */
2191 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
2192 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
2193 int ret
= ext4_quota_on_mount(sb
, i
);
2195 ext4_msg(sb
, KERN_ERR
,
2196 "Cannot turn on journaled "
2197 "quota: error %d", ret
);
2202 while (es
->s_last_orphan
) {
2203 struct inode
*inode
;
2205 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
2206 if (IS_ERR(inode
)) {
2207 es
->s_last_orphan
= 0;
2211 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
2212 dquot_initialize(inode
);
2213 if (inode
->i_nlink
) {
2214 if (test_opt(sb
, DEBUG
))
2215 ext4_msg(sb
, KERN_DEBUG
,
2216 "%s: truncating inode %lu to %lld bytes",
2217 __func__
, inode
->i_ino
, inode
->i_size
);
2218 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2219 inode
->i_ino
, inode
->i_size
);
2220 mutex_lock(&inode
->i_mutex
);
2221 truncate_inode_pages(inode
->i_mapping
, inode
->i_size
);
2222 ext4_truncate(inode
);
2223 mutex_unlock(&inode
->i_mutex
);
2226 if (test_opt(sb
, DEBUG
))
2227 ext4_msg(sb
, KERN_DEBUG
,
2228 "%s: deleting unreferenced inode %lu",
2229 __func__
, inode
->i_ino
);
2230 jbd_debug(2, "deleting unreferenced inode %lu\n",
2234 iput(inode
); /* The delete magic happens here! */
2237 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2240 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
2241 PLURAL(nr_orphans
));
2243 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
2244 PLURAL(nr_truncates
));
2246 /* Turn quotas off */
2247 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
2248 if (sb_dqopt(sb
)->files
[i
])
2249 dquot_quota_off(sb
, i
);
2252 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
2256 * Maximal extent format file size.
2257 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2258 * extent format containers, within a sector_t, and within i_blocks
2259 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2260 * so that won't be a limiting factor.
2262 * However there is other limiting factor. We do store extents in the form
2263 * of starting block and length, hence the resulting length of the extent
2264 * covering maximum file size must fit into on-disk format containers as
2265 * well. Given that length is always by 1 unit bigger than max unit (because
2266 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2268 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2270 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
2273 loff_t upper_limit
= MAX_LFS_FILESIZE
;
2275 /* small i_blocks in vfs inode? */
2276 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2278 * CONFIG_LBDAF is not enabled implies the inode
2279 * i_block represent total blocks in 512 bytes
2280 * 32 == size of vfs inode i_blocks * 8
2282 upper_limit
= (1LL << 32) - 1;
2284 /* total blocks in file system block size */
2285 upper_limit
>>= (blkbits
- 9);
2286 upper_limit
<<= blkbits
;
2290 * 32-bit extent-start container, ee_block. We lower the maxbytes
2291 * by one fs block, so ee_len can cover the extent of maximum file
2294 res
= (1LL << 32) - 1;
2297 /* Sanity check against vm- & vfs- imposed limits */
2298 if (res
> upper_limit
)
2305 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2306 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2307 * We need to be 1 filesystem block less than the 2^48 sector limit.
2309 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
2311 loff_t res
= EXT4_NDIR_BLOCKS
;
2314 /* This is calculated to be the largest file size for a dense, block
2315 * mapped file such that the file's total number of 512-byte sectors,
2316 * including data and all indirect blocks, does not exceed (2^48 - 1).
2318 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2319 * number of 512-byte sectors of the file.
2322 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2324 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2325 * the inode i_block field represents total file blocks in
2326 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2328 upper_limit
= (1LL << 32) - 1;
2330 /* total blocks in file system block size */
2331 upper_limit
>>= (bits
- 9);
2335 * We use 48 bit ext4_inode i_blocks
2336 * With EXT4_HUGE_FILE_FL set the i_blocks
2337 * represent total number of blocks in
2338 * file system block size
2340 upper_limit
= (1LL << 48) - 1;
2344 /* indirect blocks */
2346 /* double indirect blocks */
2347 meta_blocks
+= 1 + (1LL << (bits
-2));
2348 /* tripple indirect blocks */
2349 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
2351 upper_limit
-= meta_blocks
;
2352 upper_limit
<<= bits
;
2354 res
+= 1LL << (bits
-2);
2355 res
+= 1LL << (2*(bits
-2));
2356 res
+= 1LL << (3*(bits
-2));
2358 if (res
> upper_limit
)
2361 if (res
> MAX_LFS_FILESIZE
)
2362 res
= MAX_LFS_FILESIZE
;
2367 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
2368 ext4_fsblk_t logical_sb_block
, int nr
)
2370 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2371 ext4_group_t bg
, first_meta_bg
;
2374 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
2376 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_META_BG
) ||
2378 return logical_sb_block
+ nr
+ 1;
2379 bg
= sbi
->s_desc_per_block
* nr
;
2380 if (ext4_bg_has_super(sb
, bg
))
2384 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2385 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2386 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2389 if (sb
->s_blocksize
== 1024 && nr
== 0 &&
2390 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_first_data_block
) == 0)
2393 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
2397 * ext4_get_stripe_size: Get the stripe size.
2398 * @sbi: In memory super block info
2400 * If we have specified it via mount option, then
2401 * use the mount option value. If the value specified at mount time is
2402 * greater than the blocks per group use the super block value.
2403 * If the super block value is greater than blocks per group return 0.
2404 * Allocator needs it be less than blocks per group.
2407 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
2409 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
2410 unsigned long stripe_width
=
2411 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
2414 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
2415 ret
= sbi
->s_stripe
;
2416 else if (stripe_width
<= sbi
->s_blocks_per_group
)
2418 else if (stride
<= sbi
->s_blocks_per_group
)
2424 * If the stripe width is 1, this makes no sense and
2425 * we set it to 0 to turn off stripe handling code.
2436 struct attribute attr
;
2437 ssize_t (*show
)(struct ext4_attr
*, struct ext4_sb_info
*, char *);
2438 ssize_t (*store
)(struct ext4_attr
*, struct ext4_sb_info
*,
2439 const char *, size_t);
2446 static int parse_strtoull(const char *buf
,
2447 unsigned long long max
, unsigned long long *value
)
2451 ret
= kstrtoull(skip_spaces(buf
), 0, value
);
2452 if (!ret
&& *value
> max
)
2457 static ssize_t
delayed_allocation_blocks_show(struct ext4_attr
*a
,
2458 struct ext4_sb_info
*sbi
,
2461 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2463 percpu_counter_sum(&sbi
->s_dirtyclusters_counter
)));
2466 static ssize_t
session_write_kbytes_show(struct ext4_attr
*a
,
2467 struct ext4_sb_info
*sbi
, char *buf
)
2469 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2471 if (!sb
->s_bdev
->bd_part
)
2472 return snprintf(buf
, PAGE_SIZE
, "0\n");
2473 return snprintf(buf
, PAGE_SIZE
, "%lu\n",
2474 (part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2475 sbi
->s_sectors_written_start
) >> 1);
2478 static ssize_t
lifetime_write_kbytes_show(struct ext4_attr
*a
,
2479 struct ext4_sb_info
*sbi
, char *buf
)
2481 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2483 if (!sb
->s_bdev
->bd_part
)
2484 return snprintf(buf
, PAGE_SIZE
, "0\n");
2485 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2486 (unsigned long long)(sbi
->s_kbytes_written
+
2487 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2488 EXT4_SB(sb
)->s_sectors_written_start
) >> 1)));
2491 static ssize_t
inode_readahead_blks_store(struct ext4_attr
*a
,
2492 struct ext4_sb_info
*sbi
,
2493 const char *buf
, size_t count
)
2498 ret
= kstrtoul(skip_spaces(buf
), 0, &t
);
2502 if (t
&& (!is_power_of_2(t
) || t
> 0x40000000))
2505 sbi
->s_inode_readahead_blks
= t
;
2509 static ssize_t
sbi_ui_show(struct ext4_attr
*a
,
2510 struct ext4_sb_info
*sbi
, char *buf
)
2512 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->u
.offset
);
2514 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
2517 static ssize_t
sbi_ui_store(struct ext4_attr
*a
,
2518 struct ext4_sb_info
*sbi
,
2519 const char *buf
, size_t count
)
2521 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->u
.offset
);
2525 ret
= kstrtoul(skip_spaces(buf
), 0, &t
);
2532 static ssize_t
es_ui_show(struct ext4_attr
*a
,
2533 struct ext4_sb_info
*sbi
, char *buf
)
2536 unsigned int *ui
= (unsigned int *) (((char *) sbi
->s_es
) +
2539 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
2542 static ssize_t
reserved_clusters_show(struct ext4_attr
*a
,
2543 struct ext4_sb_info
*sbi
, char *buf
)
2545 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2546 (unsigned long long) atomic64_read(&sbi
->s_resv_clusters
));
2549 static ssize_t
reserved_clusters_store(struct ext4_attr
*a
,
2550 struct ext4_sb_info
*sbi
,
2551 const char *buf
, size_t count
)
2553 unsigned long long val
;
2556 if (parse_strtoull(buf
, -1ULL, &val
))
2558 ret
= ext4_reserve_clusters(sbi
, val
);
2560 return ret
? ret
: count
;
2563 static ssize_t
trigger_test_error(struct ext4_attr
*a
,
2564 struct ext4_sb_info
*sbi
,
2565 const char *buf
, size_t count
)
2569 if (!capable(CAP_SYS_ADMIN
))
2572 if (len
&& buf
[len
-1] == '\n')
2576 ext4_error(sbi
->s_sb
, "%.*s", len
, buf
);
2580 static ssize_t
sbi_deprecated_show(struct ext4_attr
*a
,
2581 struct ext4_sb_info
*sbi
, char *buf
)
2583 return snprintf(buf
, PAGE_SIZE
, "%d\n", a
->u
.deprecated_val
);
2586 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2587 static struct ext4_attr ext4_attr_##_name = { \
2588 .attr = {.name = __stringify(_name), .mode = _mode }, \
2592 .offset = offsetof(struct ext4_sb_info, _elname),\
2596 #define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname) \
2597 static struct ext4_attr ext4_attr_##_name = { \
2598 .attr = {.name = __stringify(_name), .mode = _mode }, \
2602 .offset = offsetof(struct ext4_super_block, _elname), \
2606 #define EXT4_ATTR(name, mode, show, store) \
2607 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2609 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2610 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2611 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2613 #define EXT4_RO_ATTR_ES_UI(name, elname) \
2614 EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
2615 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2616 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2618 #define ATTR_LIST(name) &ext4_attr_##name.attr
2619 #define EXT4_DEPRECATED_ATTR(_name, _val) \
2620 static struct ext4_attr ext4_attr_##_name = { \
2621 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2622 .show = sbi_deprecated_show, \
2624 .deprecated_val = _val, \
2628 EXT4_RO_ATTR(delayed_allocation_blocks
);
2629 EXT4_RO_ATTR(session_write_kbytes
);
2630 EXT4_RO_ATTR(lifetime_write_kbytes
);
2631 EXT4_RW_ATTR(reserved_clusters
);
2632 EXT4_ATTR_OFFSET(inode_readahead_blks
, 0644, sbi_ui_show
,
2633 inode_readahead_blks_store
, s_inode_readahead_blks
);
2634 EXT4_RW_ATTR_SBI_UI(inode_goal
, s_inode_goal
);
2635 EXT4_RW_ATTR_SBI_UI(mb_stats
, s_mb_stats
);
2636 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan
, s_mb_max_to_scan
);
2637 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan
, s_mb_min_to_scan
);
2638 EXT4_RW_ATTR_SBI_UI(mb_order2_req
, s_mb_order2_reqs
);
2639 EXT4_RW_ATTR_SBI_UI(mb_stream_req
, s_mb_stream_request
);
2640 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc
, s_mb_group_prealloc
);
2641 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump
, 128);
2642 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb
, s_extent_max_zeroout_kb
);
2643 EXT4_ATTR(trigger_fs_error
, 0200, NULL
, trigger_test_error
);
2644 EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms
, s_err_ratelimit_state
.interval
);
2645 EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst
, s_err_ratelimit_state
.burst
);
2646 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms
, s_warning_ratelimit_state
.interval
);
2647 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst
, s_warning_ratelimit_state
.burst
);
2648 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms
, s_msg_ratelimit_state
.interval
);
2649 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst
, s_msg_ratelimit_state
.burst
);
2650 EXT4_RO_ATTR_ES_UI(errors_count
, s_error_count
);
2651 EXT4_RO_ATTR_ES_UI(first_error_time
, s_first_error_time
);
2652 EXT4_RO_ATTR_ES_UI(last_error_time
, s_last_error_time
);
2654 static struct attribute
*ext4_attrs
[] = {
2655 ATTR_LIST(delayed_allocation_blocks
),
2656 ATTR_LIST(session_write_kbytes
),
2657 ATTR_LIST(lifetime_write_kbytes
),
2658 ATTR_LIST(reserved_clusters
),
2659 ATTR_LIST(inode_readahead_blks
),
2660 ATTR_LIST(inode_goal
),
2661 ATTR_LIST(mb_stats
),
2662 ATTR_LIST(mb_max_to_scan
),
2663 ATTR_LIST(mb_min_to_scan
),
2664 ATTR_LIST(mb_order2_req
),
2665 ATTR_LIST(mb_stream_req
),
2666 ATTR_LIST(mb_group_prealloc
),
2667 ATTR_LIST(max_writeback_mb_bump
),
2668 ATTR_LIST(extent_max_zeroout_kb
),
2669 ATTR_LIST(trigger_fs_error
),
2670 ATTR_LIST(err_ratelimit_interval_ms
),
2671 ATTR_LIST(err_ratelimit_burst
),
2672 ATTR_LIST(warning_ratelimit_interval_ms
),
2673 ATTR_LIST(warning_ratelimit_burst
),
2674 ATTR_LIST(msg_ratelimit_interval_ms
),
2675 ATTR_LIST(msg_ratelimit_burst
),
2676 ATTR_LIST(errors_count
),
2677 ATTR_LIST(first_error_time
),
2678 ATTR_LIST(last_error_time
),
2682 /* Features this copy of ext4 supports */
2683 EXT4_INFO_ATTR(lazy_itable_init
);
2684 EXT4_INFO_ATTR(batched_discard
);
2685 EXT4_INFO_ATTR(meta_bg_resize
);
2687 static struct attribute
*ext4_feat_attrs
[] = {
2688 ATTR_LIST(lazy_itable_init
),
2689 ATTR_LIST(batched_discard
),
2690 ATTR_LIST(meta_bg_resize
),
2694 static ssize_t
ext4_attr_show(struct kobject
*kobj
,
2695 struct attribute
*attr
, char *buf
)
2697 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2699 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2701 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
2704 static ssize_t
ext4_attr_store(struct kobject
*kobj
,
2705 struct attribute
*attr
,
2706 const char *buf
, size_t len
)
2708 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2710 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2712 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
2715 static void ext4_sb_release(struct kobject
*kobj
)
2717 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2719 complete(&sbi
->s_kobj_unregister
);
2722 static const struct sysfs_ops ext4_attr_ops
= {
2723 .show
= ext4_attr_show
,
2724 .store
= ext4_attr_store
,
2727 static struct kobj_type ext4_ktype
= {
2728 .default_attrs
= ext4_attrs
,
2729 .sysfs_ops
= &ext4_attr_ops
,
2730 .release
= ext4_sb_release
,
2733 static void ext4_feat_release(struct kobject
*kobj
)
2735 complete(&ext4_feat
->f_kobj_unregister
);
2738 static ssize_t
ext4_feat_show(struct kobject
*kobj
,
2739 struct attribute
*attr
, char *buf
)
2741 return snprintf(buf
, PAGE_SIZE
, "supported\n");
2745 * We can not use ext4_attr_show/store because it relies on the kobject
2746 * being embedded in the ext4_sb_info structure which is definitely not
2747 * true in this case.
2749 static const struct sysfs_ops ext4_feat_ops
= {
2750 .show
= ext4_feat_show
,
2754 static struct kobj_type ext4_feat_ktype
= {
2755 .default_attrs
= ext4_feat_attrs
,
2756 .sysfs_ops
= &ext4_feat_ops
,
2757 .release
= ext4_feat_release
,
2761 * Check whether this filesystem can be mounted based on
2762 * the features present and the RDONLY/RDWR mount requested.
2763 * Returns 1 if this filesystem can be mounted as requested,
2764 * 0 if it cannot be.
2766 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
2768 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT4_FEATURE_INCOMPAT_SUPP
)) {
2769 ext4_msg(sb
, KERN_ERR
,
2770 "Couldn't mount because of "
2771 "unsupported optional features (%x)",
2772 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2773 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2780 /* Check that feature set is OK for a read-write mount */
2781 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT4_FEATURE_RO_COMPAT_SUPP
)) {
2782 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
2783 "unsupported optional features (%x)",
2784 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2785 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2789 * Large file size enabled file system can only be mounted
2790 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2792 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_HUGE_FILE
)) {
2793 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
2794 ext4_msg(sb
, KERN_ERR
, "Filesystem with huge files "
2795 "cannot be mounted RDWR without "
2800 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_BIGALLOC
) &&
2801 !EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_EXTENTS
)) {
2802 ext4_msg(sb
, KERN_ERR
,
2803 "Can't support bigalloc feature without "
2804 "extents feature\n");
2808 #ifndef CONFIG_QUOTA
2809 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
) &&
2811 ext4_msg(sb
, KERN_ERR
,
2812 "Filesystem with quota feature cannot be mounted RDWR "
2813 "without CONFIG_QUOTA");
2816 #endif /* CONFIG_QUOTA */
2821 * This function is called once a day if we have errors logged
2822 * on the file system
2824 static void print_daily_error_info(unsigned long arg
)
2826 struct super_block
*sb
= (struct super_block
*) arg
;
2827 struct ext4_sb_info
*sbi
;
2828 struct ext4_super_block
*es
;
2833 if (es
->s_error_count
)
2834 /* fsck newer than v1.41.13 is needed to clean this condition. */
2835 ext4_msg(sb
, KERN_NOTICE
, "error count since last fsck: %u",
2836 le32_to_cpu(es
->s_error_count
));
2837 if (es
->s_first_error_time
) {
2838 printk(KERN_NOTICE
"EXT4-fs (%s): initial error at time %u: %.*s:%d",
2839 sb
->s_id
, le32_to_cpu(es
->s_first_error_time
),
2840 (int) sizeof(es
->s_first_error_func
),
2841 es
->s_first_error_func
,
2842 le32_to_cpu(es
->s_first_error_line
));
2843 if (es
->s_first_error_ino
)
2844 printk(": inode %u",
2845 le32_to_cpu(es
->s_first_error_ino
));
2846 if (es
->s_first_error_block
)
2847 printk(": block %llu", (unsigned long long)
2848 le64_to_cpu(es
->s_first_error_block
));
2851 if (es
->s_last_error_time
) {
2852 printk(KERN_NOTICE
"EXT4-fs (%s): last error at time %u: %.*s:%d",
2853 sb
->s_id
, le32_to_cpu(es
->s_last_error_time
),
2854 (int) sizeof(es
->s_last_error_func
),
2855 es
->s_last_error_func
,
2856 le32_to_cpu(es
->s_last_error_line
));
2857 if (es
->s_last_error_ino
)
2858 printk(": inode %u",
2859 le32_to_cpu(es
->s_last_error_ino
));
2860 if (es
->s_last_error_block
)
2861 printk(": block %llu", (unsigned long long)
2862 le64_to_cpu(es
->s_last_error_block
));
2865 mod_timer(&sbi
->s_err_report
, jiffies
+ 24*60*60*HZ
); /* Once a day */
2868 /* Find next suitable group and run ext4_init_inode_table */
2869 static int ext4_run_li_request(struct ext4_li_request
*elr
)
2871 struct ext4_group_desc
*gdp
= NULL
;
2872 ext4_group_t group
, ngroups
;
2873 struct super_block
*sb
;
2874 unsigned long timeout
= 0;
2878 ngroups
= EXT4_SB(sb
)->s_groups_count
;
2881 for (group
= elr
->lr_next_group
; group
< ngroups
; group
++) {
2882 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2888 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2892 if (group
>= ngroups
)
2897 ret
= ext4_init_inode_table(sb
, group
,
2898 elr
->lr_timeout
? 0 : 1);
2899 if (elr
->lr_timeout
== 0) {
2900 timeout
= (jiffies
- timeout
) *
2901 elr
->lr_sbi
->s_li_wait_mult
;
2902 elr
->lr_timeout
= timeout
;
2904 elr
->lr_next_sched
= jiffies
+ elr
->lr_timeout
;
2905 elr
->lr_next_group
= group
+ 1;
2913 * Remove lr_request from the list_request and free the
2914 * request structure. Should be called with li_list_mtx held
2916 static void ext4_remove_li_request(struct ext4_li_request
*elr
)
2918 struct ext4_sb_info
*sbi
;
2925 list_del(&elr
->lr_request
);
2926 sbi
->s_li_request
= NULL
;
2930 static void ext4_unregister_li_request(struct super_block
*sb
)
2932 mutex_lock(&ext4_li_mtx
);
2933 if (!ext4_li_info
) {
2934 mutex_unlock(&ext4_li_mtx
);
2938 mutex_lock(&ext4_li_info
->li_list_mtx
);
2939 ext4_remove_li_request(EXT4_SB(sb
)->s_li_request
);
2940 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2941 mutex_unlock(&ext4_li_mtx
);
2944 static struct task_struct
*ext4_lazyinit_task
;
2947 * This is the function where ext4lazyinit thread lives. It walks
2948 * through the request list searching for next scheduled filesystem.
2949 * When such a fs is found, run the lazy initialization request
2950 * (ext4_rn_li_request) and keep track of the time spend in this
2951 * function. Based on that time we compute next schedule time of
2952 * the request. When walking through the list is complete, compute
2953 * next waking time and put itself into sleep.
2955 static int ext4_lazyinit_thread(void *arg
)
2957 struct ext4_lazy_init
*eli
= (struct ext4_lazy_init
*)arg
;
2958 struct list_head
*pos
, *n
;
2959 struct ext4_li_request
*elr
;
2960 unsigned long next_wakeup
, cur
;
2962 BUG_ON(NULL
== eli
);
2966 next_wakeup
= MAX_JIFFY_OFFSET
;
2968 mutex_lock(&eli
->li_list_mtx
);
2969 if (list_empty(&eli
->li_request_list
)) {
2970 mutex_unlock(&eli
->li_list_mtx
);
2974 list_for_each_safe(pos
, n
, &eli
->li_request_list
) {
2975 elr
= list_entry(pos
, struct ext4_li_request
,
2978 if (time_after_eq(jiffies
, elr
->lr_next_sched
)) {
2979 if (ext4_run_li_request(elr
) != 0) {
2980 /* error, remove the lazy_init job */
2981 ext4_remove_li_request(elr
);
2986 if (time_before(elr
->lr_next_sched
, next_wakeup
))
2987 next_wakeup
= elr
->lr_next_sched
;
2989 mutex_unlock(&eli
->li_list_mtx
);
2994 if ((time_after_eq(cur
, next_wakeup
)) ||
2995 (MAX_JIFFY_OFFSET
== next_wakeup
)) {
3000 schedule_timeout_interruptible(next_wakeup
- cur
);
3002 if (kthread_should_stop()) {
3003 ext4_clear_request_list();
3010 * It looks like the request list is empty, but we need
3011 * to check it under the li_list_mtx lock, to prevent any
3012 * additions into it, and of course we should lock ext4_li_mtx
3013 * to atomically free the list and ext4_li_info, because at
3014 * this point another ext4 filesystem could be registering
3017 mutex_lock(&ext4_li_mtx
);
3018 mutex_lock(&eli
->li_list_mtx
);
3019 if (!list_empty(&eli
->li_request_list
)) {
3020 mutex_unlock(&eli
->li_list_mtx
);
3021 mutex_unlock(&ext4_li_mtx
);
3024 mutex_unlock(&eli
->li_list_mtx
);
3025 kfree(ext4_li_info
);
3026 ext4_li_info
= NULL
;
3027 mutex_unlock(&ext4_li_mtx
);
3032 static void ext4_clear_request_list(void)
3034 struct list_head
*pos
, *n
;
3035 struct ext4_li_request
*elr
;
3037 mutex_lock(&ext4_li_info
->li_list_mtx
);
3038 list_for_each_safe(pos
, n
, &ext4_li_info
->li_request_list
) {
3039 elr
= list_entry(pos
, struct ext4_li_request
,
3041 ext4_remove_li_request(elr
);
3043 mutex_unlock(&ext4_li_info
->li_list_mtx
);
3046 static int ext4_run_lazyinit_thread(void)
3048 ext4_lazyinit_task
= kthread_run(ext4_lazyinit_thread
,
3049 ext4_li_info
, "ext4lazyinit");
3050 if (IS_ERR(ext4_lazyinit_task
)) {
3051 int err
= PTR_ERR(ext4_lazyinit_task
);
3052 ext4_clear_request_list();
3053 kfree(ext4_li_info
);
3054 ext4_li_info
= NULL
;
3055 printk(KERN_CRIT
"EXT4-fs: error %d creating inode table "
3056 "initialization thread\n",
3060 ext4_li_info
->li_state
|= EXT4_LAZYINIT_RUNNING
;
3065 * Check whether it make sense to run itable init. thread or not.
3066 * If there is at least one uninitialized inode table, return
3067 * corresponding group number, else the loop goes through all
3068 * groups and return total number of groups.
3070 static ext4_group_t
ext4_has_uninit_itable(struct super_block
*sb
)
3072 ext4_group_t group
, ngroups
= EXT4_SB(sb
)->s_groups_count
;
3073 struct ext4_group_desc
*gdp
= NULL
;
3075 for (group
= 0; group
< ngroups
; group
++) {
3076 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
3080 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
3087 static int ext4_li_info_new(void)
3089 struct ext4_lazy_init
*eli
= NULL
;
3091 eli
= kzalloc(sizeof(*eli
), GFP_KERNEL
);
3095 INIT_LIST_HEAD(&eli
->li_request_list
);
3096 mutex_init(&eli
->li_list_mtx
);
3098 eli
->li_state
|= EXT4_LAZYINIT_QUIT
;
3105 static struct ext4_li_request
*ext4_li_request_new(struct super_block
*sb
,
3108 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3109 struct ext4_li_request
*elr
;
3111 elr
= kzalloc(sizeof(*elr
), GFP_KERNEL
);
3117 elr
->lr_next_group
= start
;
3120 * Randomize first schedule time of the request to
3121 * spread the inode table initialization requests
3124 elr
->lr_next_sched
= jiffies
+ (prandom_u32() %
3125 (EXT4_DEF_LI_MAX_START_DELAY
* HZ
));
3129 int ext4_register_li_request(struct super_block
*sb
,
3130 ext4_group_t first_not_zeroed
)
3132 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3133 struct ext4_li_request
*elr
= NULL
;
3134 ext4_group_t ngroups
= EXT4_SB(sb
)->s_groups_count
;
3137 mutex_lock(&ext4_li_mtx
);
3138 if (sbi
->s_li_request
!= NULL
) {
3140 * Reset timeout so it can be computed again, because
3141 * s_li_wait_mult might have changed.
3143 sbi
->s_li_request
->lr_timeout
= 0;
3147 if (first_not_zeroed
== ngroups
||
3148 (sb
->s_flags
& MS_RDONLY
) ||
3149 !test_opt(sb
, INIT_INODE_TABLE
))
3152 elr
= ext4_li_request_new(sb
, first_not_zeroed
);
3158 if (NULL
== ext4_li_info
) {
3159 ret
= ext4_li_info_new();
3164 mutex_lock(&ext4_li_info
->li_list_mtx
);
3165 list_add(&elr
->lr_request
, &ext4_li_info
->li_request_list
);
3166 mutex_unlock(&ext4_li_info
->li_list_mtx
);
3168 sbi
->s_li_request
= elr
;
3170 * set elr to NULL here since it has been inserted to
3171 * the request_list and the removal and free of it is
3172 * handled by ext4_clear_request_list from now on.
3176 if (!(ext4_li_info
->li_state
& EXT4_LAZYINIT_RUNNING
)) {
3177 ret
= ext4_run_lazyinit_thread();
3182 mutex_unlock(&ext4_li_mtx
);
3189 * We do not need to lock anything since this is called on
3192 static void ext4_destroy_lazyinit_thread(void)
3195 * If thread exited earlier
3196 * there's nothing to be done.
3198 if (!ext4_li_info
|| !ext4_lazyinit_task
)
3201 kthread_stop(ext4_lazyinit_task
);
3204 static int set_journal_csum_feature_set(struct super_block
*sb
)
3207 int compat
, incompat
;
3208 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3210 if (ext4_has_metadata_csum(sb
)) {
3211 /* journal checksum v3 */
3213 incompat
= JBD2_FEATURE_INCOMPAT_CSUM_V3
;
3215 /* journal checksum v1 */
3216 compat
= JBD2_FEATURE_COMPAT_CHECKSUM
;
3220 jbd2_journal_clear_features(sbi
->s_journal
,
3221 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3222 JBD2_FEATURE_INCOMPAT_CSUM_V3
|
3223 JBD2_FEATURE_INCOMPAT_CSUM_V2
);
3224 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
3225 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3227 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
|
3229 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
3230 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3233 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3234 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3236 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3237 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3244 * Note: calculating the overhead so we can be compatible with
3245 * historical BSD practice is quite difficult in the face of
3246 * clusters/bigalloc. This is because multiple metadata blocks from
3247 * different block group can end up in the same allocation cluster.
3248 * Calculating the exact overhead in the face of clustered allocation
3249 * requires either O(all block bitmaps) in memory or O(number of block
3250 * groups**2) in time. We will still calculate the superblock for
3251 * older file systems --- and if we come across with a bigalloc file
3252 * system with zero in s_overhead_clusters the estimate will be close to
3253 * correct especially for very large cluster sizes --- but for newer
3254 * file systems, it's better to calculate this figure once at mkfs
3255 * time, and store it in the superblock. If the superblock value is
3256 * present (even for non-bigalloc file systems), we will use it.
3258 static int count_overhead(struct super_block
*sb
, ext4_group_t grp
,
3261 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3262 struct ext4_group_desc
*gdp
;
3263 ext4_fsblk_t first_block
, last_block
, b
;
3264 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3265 int s
, j
, count
= 0;
3267 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_BIGALLOC
))
3268 return (ext4_bg_has_super(sb
, grp
) + ext4_bg_num_gdb(sb
, grp
) +
3269 sbi
->s_itb_per_group
+ 2);
3271 first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
) +
3272 (grp
* EXT4_BLOCKS_PER_GROUP(sb
));
3273 last_block
= first_block
+ EXT4_BLOCKS_PER_GROUP(sb
) - 1;
3274 for (i
= 0; i
< ngroups
; i
++) {
3275 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
3276 b
= ext4_block_bitmap(sb
, gdp
);
3277 if (b
>= first_block
&& b
<= last_block
) {
3278 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3281 b
= ext4_inode_bitmap(sb
, gdp
);
3282 if (b
>= first_block
&& b
<= last_block
) {
3283 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3286 b
= ext4_inode_table(sb
, gdp
);
3287 if (b
>= first_block
&& b
+ sbi
->s_itb_per_group
<= last_block
)
3288 for (j
= 0; j
< sbi
->s_itb_per_group
; j
++, b
++) {
3289 int c
= EXT4_B2C(sbi
, b
- first_block
);
3290 ext4_set_bit(c
, buf
);
3296 if (ext4_bg_has_super(sb
, grp
)) {
3297 ext4_set_bit(s
++, buf
);
3300 for (j
= ext4_bg_num_gdb(sb
, grp
); j
> 0; j
--) {
3301 ext4_set_bit(EXT4_B2C(sbi
, s
++), buf
);
3307 return EXT4_CLUSTERS_PER_GROUP(sb
) -
3308 ext4_count_free(buf
, EXT4_CLUSTERS_PER_GROUP(sb
) / 8);
3312 * Compute the overhead and stash it in sbi->s_overhead
3314 int ext4_calculate_overhead(struct super_block
*sb
)
3316 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3317 struct ext4_super_block
*es
= sbi
->s_es
;
3318 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3319 ext4_fsblk_t overhead
= 0;
3320 char *buf
= (char *) get_zeroed_page(GFP_KERNEL
);
3326 * Compute the overhead (FS structures). This is constant
3327 * for a given filesystem unless the number of block groups
3328 * changes so we cache the previous value until it does.
3332 * All of the blocks before first_data_block are overhead
3334 overhead
= EXT4_B2C(sbi
, le32_to_cpu(es
->s_first_data_block
));
3337 * Add the overhead found in each block group
3339 for (i
= 0; i
< ngroups
; i
++) {
3342 blks
= count_overhead(sb
, i
, buf
);
3345 memset(buf
, 0, PAGE_SIZE
);
3348 /* Add the journal blocks as well */
3350 overhead
+= EXT4_NUM_B2C(sbi
, sbi
->s_journal
->j_maxlen
);
3352 sbi
->s_overhead
= overhead
;
3354 free_page((unsigned long) buf
);
3359 static ext4_fsblk_t
ext4_calculate_resv_clusters(struct super_block
*sb
)
3361 ext4_fsblk_t resv_clusters
;
3364 * There's no need to reserve anything when we aren't using extents.
3365 * The space estimates are exact, there are no unwritten extents,
3366 * hole punching doesn't need new metadata... This is needed especially
3367 * to keep ext2/3 backward compatibility.
3369 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_EXTENTS
))
3372 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3373 * This should cover the situations where we can not afford to run
3374 * out of space like for example punch hole, or converting
3375 * unwritten extents in delalloc path. In most cases such
3376 * allocation would require 1, or 2 blocks, higher numbers are
3379 resv_clusters
= ext4_blocks_count(EXT4_SB(sb
)->s_es
) >>
3380 EXT4_SB(sb
)->s_cluster_bits
;
3382 do_div(resv_clusters
, 50);
3383 resv_clusters
= min_t(ext4_fsblk_t
, resv_clusters
, 4096);
3385 return resv_clusters
;
3389 static int ext4_reserve_clusters(struct ext4_sb_info
*sbi
, ext4_fsblk_t count
)
3391 ext4_fsblk_t clusters
= ext4_blocks_count(sbi
->s_es
) >>
3392 sbi
->s_cluster_bits
;
3394 if (count
>= clusters
)
3397 atomic64_set(&sbi
->s_resv_clusters
, count
);
3401 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
3403 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
3404 struct buffer_head
*bh
;
3405 struct ext4_super_block
*es
= NULL
;
3406 struct ext4_sb_info
*sbi
;
3408 ext4_fsblk_t sb_block
= get_sb_block(&data
);
3409 ext4_fsblk_t logical_sb_block
;
3410 unsigned long offset
= 0;
3411 unsigned long journal_devnum
= 0;
3412 unsigned long def_mount_opts
;
3417 int blocksize
, clustersize
;
3418 unsigned int db_count
;
3420 int needs_recovery
, has_huge_files
, has_bigalloc
;
3423 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
3424 ext4_group_t first_not_zeroed
;
3426 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
3430 sbi
->s_blockgroup_lock
=
3431 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
3432 if (!sbi
->s_blockgroup_lock
) {
3436 sb
->s_fs_info
= sbi
;
3438 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
3439 sbi
->s_sb_block
= sb_block
;
3440 if (sb
->s_bdev
->bd_part
)
3441 sbi
->s_sectors_written_start
=
3442 part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]);
3444 /* Cleanup superblock name */
3445 for (cp
= sb
->s_id
; (cp
= strchr(cp
, '/'));)
3448 /* -EINVAL is default */
3450 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
3452 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
3457 * The ext4 superblock will not be buffer aligned for other than 1kB
3458 * block sizes. We need to calculate the offset from buffer start.
3460 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
3461 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3462 offset
= do_div(logical_sb_block
, blocksize
);
3464 logical_sb_block
= sb_block
;
3467 if (!(bh
= sb_bread_unmovable(sb
, logical_sb_block
))) {
3468 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
3472 * Note: s_es must be initialized as soon as possible because
3473 * some ext4 macro-instructions depend on its value
3475 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
3477 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
3478 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
3480 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
3482 /* Warn if metadata_csum and gdt_csum are both set. */
3483 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3484 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
) &&
3485 EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_GDT_CSUM
))
3486 ext4_warning(sb
, KERN_INFO
"metadata_csum and uninit_bg are "
3487 "redundant flags; please run fsck.");
3489 /* Check for a known checksum algorithm */
3490 if (!ext4_verify_csum_type(sb
, es
)) {
3491 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3492 "unknown checksum algorithm.");
3497 /* Load the checksum driver */
3498 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3499 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
)) {
3500 sbi
->s_chksum_driver
= crypto_alloc_shash("crc32c", 0, 0);
3501 if (IS_ERR(sbi
->s_chksum_driver
)) {
3502 ext4_msg(sb
, KERN_ERR
, "Cannot load crc32c driver.");
3503 ret
= PTR_ERR(sbi
->s_chksum_driver
);
3504 sbi
->s_chksum_driver
= NULL
;
3509 /* Check superblock checksum */
3510 if (!ext4_superblock_csum_verify(sb
, es
)) {
3511 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3512 "invalid superblock checksum. Run e2fsck?");
3517 /* Precompute checksum seed for all metadata */
3518 if (ext4_has_metadata_csum(sb
))
3519 sbi
->s_csum_seed
= ext4_chksum(sbi
, ~0, es
->s_uuid
,
3520 sizeof(es
->s_uuid
));
3522 /* Set defaults before we parse the mount options */
3523 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
3524 set_opt(sb
, INIT_INODE_TABLE
);
3525 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
3527 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
)
3529 if (def_mount_opts
& EXT4_DEFM_UID16
)
3530 set_opt(sb
, NO_UID32
);
3531 /* xattr user namespace & acls are now defaulted on */
3532 set_opt(sb
, XATTR_USER
);
3533 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3534 set_opt(sb
, POSIX_ACL
);
3536 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3537 if (ext4_has_metadata_csum(sb
))
3538 set_opt(sb
, JOURNAL_CHECKSUM
);
3540 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
3541 set_opt(sb
, JOURNAL_DATA
);
3542 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
3543 set_opt(sb
, ORDERED_DATA
);
3544 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
3545 set_opt(sb
, WRITEBACK_DATA
);
3547 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
3548 set_opt(sb
, ERRORS_PANIC
);
3549 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
3550 set_opt(sb
, ERRORS_CONT
);
3552 set_opt(sb
, ERRORS_RO
);
3553 /* block_validity enabled by default; disable with noblock_validity */
3554 set_opt(sb
, BLOCK_VALIDITY
);
3555 if (def_mount_opts
& EXT4_DEFM_DISCARD
)
3556 set_opt(sb
, DISCARD
);
3558 sbi
->s_resuid
= make_kuid(&init_user_ns
, le16_to_cpu(es
->s_def_resuid
));
3559 sbi
->s_resgid
= make_kgid(&init_user_ns
, le16_to_cpu(es
->s_def_resgid
));
3560 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
3561 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
3562 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
3564 if ((def_mount_opts
& EXT4_DEFM_NOBARRIER
) == 0)
3565 set_opt(sb
, BARRIER
);
3568 * enable delayed allocation by default
3569 * Use -o nodelalloc to turn it off
3571 if (!IS_EXT3_SB(sb
) && !IS_EXT2_SB(sb
) &&
3572 ((def_mount_opts
& EXT4_DEFM_NODELALLOC
) == 0))
3573 set_opt(sb
, DELALLOC
);
3576 * set default s_li_wait_mult for lazyinit, for the case there is
3577 * no mount option specified.
3579 sbi
->s_li_wait_mult
= EXT4_DEF_LI_WAIT_MULT
;
3581 if (!parse_options((char *) sbi
->s_es
->s_mount_opts
, sb
,
3582 &journal_devnum
, &journal_ioprio
, 0)) {
3583 ext4_msg(sb
, KERN_WARNING
,
3584 "failed to parse options in superblock: %s",
3585 sbi
->s_es
->s_mount_opts
);
3587 sbi
->s_def_mount_opt
= sbi
->s_mount_opt
;
3588 if (!parse_options((char *) data
, sb
, &journal_devnum
,
3589 &journal_ioprio
, 0))
3592 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
3593 printk_once(KERN_WARNING
"EXT4-fs: Warning: mounting "
3594 "with data=journal disables delayed "
3595 "allocation and O_DIRECT support!\n");
3596 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
3597 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3598 "both data=journal and delalloc");
3601 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
3602 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3603 "both data=journal and dioread_nolock");
3606 if (test_opt(sb
, DELALLOC
))
3607 clear_opt(sb
, DELALLOC
);
3610 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3611 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
3613 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
3614 (EXT4_HAS_COMPAT_FEATURE(sb
, ~0U) ||
3615 EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
3616 EXT4_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
3617 ext4_msg(sb
, KERN_WARNING
,
3618 "feature flags set on rev 0 fs, "
3619 "running e2fsck is recommended");
3621 if (es
->s_creator_os
== cpu_to_le32(EXT4_OS_HURD
)) {
3622 set_opt2(sb
, HURD_COMPAT
);
3623 if (EXT4_HAS_INCOMPAT_FEATURE(sb
,
3624 EXT4_FEATURE_INCOMPAT_64BIT
)) {
3625 ext4_msg(sb
, KERN_ERR
,
3626 "The Hurd can't support 64-bit file systems");
3631 if (IS_EXT2_SB(sb
)) {
3632 if (ext2_feature_set_ok(sb
))
3633 ext4_msg(sb
, KERN_INFO
, "mounting ext2 file system "
3634 "using the ext4 subsystem");
3636 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext2 due "
3637 "to feature incompatibilities");
3642 if (IS_EXT3_SB(sb
)) {
3643 if (ext3_feature_set_ok(sb
))
3644 ext4_msg(sb
, KERN_INFO
, "mounting ext3 file system "
3645 "using the ext4 subsystem");
3647 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext3 due "
3648 "to feature incompatibilities");
3654 * Check feature flags regardless of the revision level, since we
3655 * previously didn't change the revision level when setting the flags,
3656 * so there is a chance incompat flags are set on a rev 0 filesystem.
3658 if (!ext4_feature_set_ok(sb
, (sb
->s_flags
& MS_RDONLY
)))
3661 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
3662 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
3663 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
3664 ext4_msg(sb
, KERN_ERR
,
3665 "Unsupported filesystem blocksize %d", blocksize
);
3669 if (sb
->s_blocksize
!= blocksize
) {
3670 /* Validate the filesystem blocksize */
3671 if (!sb_set_blocksize(sb
, blocksize
)) {
3672 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
3678 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3679 offset
= do_div(logical_sb_block
, blocksize
);
3680 bh
= sb_bread_unmovable(sb
, logical_sb_block
);
3682 ext4_msg(sb
, KERN_ERR
,
3683 "Can't read superblock on 2nd try");
3686 es
= (struct ext4_super_block
*)(bh
->b_data
+ offset
);
3688 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
3689 ext4_msg(sb
, KERN_ERR
,
3690 "Magic mismatch, very weird!");
3695 has_huge_files
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3696 EXT4_FEATURE_RO_COMPAT_HUGE_FILE
);
3697 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
3699 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
3701 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
3702 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
3703 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
3705 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
3706 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
3707 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
3708 (!is_power_of_2(sbi
->s_inode_size
)) ||
3709 (sbi
->s_inode_size
> blocksize
)) {
3710 ext4_msg(sb
, KERN_ERR
,
3711 "unsupported inode size: %d",
3715 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
3716 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
3719 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
3720 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
)) {
3721 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
3722 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
3723 !is_power_of_2(sbi
->s_desc_size
)) {
3724 ext4_msg(sb
, KERN_ERR
,
3725 "unsupported descriptor size %lu",
3730 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
3732 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
3733 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
3734 if (EXT4_INODE_SIZE(sb
) == 0 || EXT4_INODES_PER_GROUP(sb
) == 0)
3737 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
3738 if (sbi
->s_inodes_per_block
== 0)
3740 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
3741 sbi
->s_inodes_per_block
;
3742 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
3744 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3745 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
3746 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
3748 for (i
= 0; i
< 4; i
++)
3749 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
3750 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
3751 if (EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_DIR_INDEX
)) {
3752 i
= le32_to_cpu(es
->s_flags
);
3753 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
3754 sbi
->s_hash_unsigned
= 3;
3755 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
3756 #ifdef __CHAR_UNSIGNED__
3757 if (!(sb
->s_flags
& MS_RDONLY
))
3759 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
3760 sbi
->s_hash_unsigned
= 3;
3762 if (!(sb
->s_flags
& MS_RDONLY
))
3764 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
3769 /* Handle clustersize */
3770 clustersize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_cluster_size
);
3771 has_bigalloc
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3772 EXT4_FEATURE_RO_COMPAT_BIGALLOC
);
3774 if (clustersize
< blocksize
) {
3775 ext4_msg(sb
, KERN_ERR
,
3776 "cluster size (%d) smaller than "
3777 "block size (%d)", clustersize
, blocksize
);
3780 sbi
->s_cluster_bits
= le32_to_cpu(es
->s_log_cluster_size
) -
3781 le32_to_cpu(es
->s_log_block_size
);
3782 sbi
->s_clusters_per_group
=
3783 le32_to_cpu(es
->s_clusters_per_group
);
3784 if (sbi
->s_clusters_per_group
> blocksize
* 8) {
3785 ext4_msg(sb
, KERN_ERR
,
3786 "#clusters per group too big: %lu",
3787 sbi
->s_clusters_per_group
);
3790 if (sbi
->s_blocks_per_group
!=
3791 (sbi
->s_clusters_per_group
* (clustersize
/ blocksize
))) {
3792 ext4_msg(sb
, KERN_ERR
, "blocks per group (%lu) and "
3793 "clusters per group (%lu) inconsistent",
3794 sbi
->s_blocks_per_group
,
3795 sbi
->s_clusters_per_group
);
3799 if (clustersize
!= blocksize
) {
3800 ext4_warning(sb
, "fragment/cluster size (%d) != "
3801 "block size (%d)", clustersize
,
3803 clustersize
= blocksize
;
3805 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
3806 ext4_msg(sb
, KERN_ERR
,
3807 "#blocks per group too big: %lu",
3808 sbi
->s_blocks_per_group
);
3811 sbi
->s_clusters_per_group
= sbi
->s_blocks_per_group
;
3812 sbi
->s_cluster_bits
= 0;
3814 sbi
->s_cluster_ratio
= clustersize
/ blocksize
;
3816 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
3817 ext4_msg(sb
, KERN_ERR
,
3818 "#inodes per group too big: %lu",
3819 sbi
->s_inodes_per_group
);
3823 /* Do we have standard group size of clustersize * 8 blocks ? */
3824 if (sbi
->s_blocks_per_group
== clustersize
<< 3)
3825 set_opt2(sb
, STD_GROUP_SIZE
);
3828 * Test whether we have more sectors than will fit in sector_t,
3829 * and whether the max offset is addressable by the page cache.
3831 err
= generic_check_addressable(sb
->s_blocksize_bits
,
3832 ext4_blocks_count(es
));
3834 ext4_msg(sb
, KERN_ERR
, "filesystem"
3835 " too large to mount safely on this system");
3836 if (sizeof(sector_t
) < 8)
3837 ext4_msg(sb
, KERN_WARNING
, "CONFIG_LBDAF not enabled");
3841 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
3844 /* check blocks count against device size */
3845 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
3846 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
3847 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
3848 "exceeds size of device (%llu blocks)",
3849 ext4_blocks_count(es
), blocks_count
);
3854 * It makes no sense for the first data block to be beyond the end
3855 * of the filesystem.
3857 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
3858 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data "
3859 "block %u is beyond end of filesystem (%llu)",
3860 le32_to_cpu(es
->s_first_data_block
),
3861 ext4_blocks_count(es
));
3864 blocks_count
= (ext4_blocks_count(es
) -
3865 le32_to_cpu(es
->s_first_data_block
) +
3866 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
3867 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
3868 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
3869 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %u "
3870 "(block count %llu, first data block %u, "
3871 "blocks per group %lu)", sbi
->s_groups_count
,
3872 ext4_blocks_count(es
),
3873 le32_to_cpu(es
->s_first_data_block
),
3874 EXT4_BLOCKS_PER_GROUP(sb
));
3877 sbi
->s_groups_count
= blocks_count
;
3878 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
3879 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
3880 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
3881 EXT4_DESC_PER_BLOCK(sb
);
3882 sbi
->s_group_desc
= ext4_kvmalloc(db_count
*
3883 sizeof(struct buffer_head
*),
3885 if (sbi
->s_group_desc
== NULL
) {
3886 ext4_msg(sb
, KERN_ERR
, "not enough memory");
3892 sbi
->s_proc
= proc_mkdir(sb
->s_id
, ext4_proc_root
);
3895 proc_create_data("options", S_IRUGO
, sbi
->s_proc
,
3896 &ext4_seq_options_fops
, sb
);
3898 bgl_lock_init(sbi
->s_blockgroup_lock
);
3900 for (i
= 0; i
< db_count
; i
++) {
3901 block
= descriptor_loc(sb
, logical_sb_block
, i
);
3902 sbi
->s_group_desc
[i
] = sb_bread_unmovable(sb
, block
);
3903 if (!sbi
->s_group_desc
[i
]) {
3904 ext4_msg(sb
, KERN_ERR
,
3905 "can't read group descriptor %d", i
);
3910 if (!ext4_check_descriptors(sb
, &first_not_zeroed
)) {
3911 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
3915 sbi
->s_gdb_count
= db_count
;
3916 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
3917 spin_lock_init(&sbi
->s_next_gen_lock
);
3919 init_timer(&sbi
->s_err_report
);
3920 sbi
->s_err_report
.function
= print_daily_error_info
;
3921 sbi
->s_err_report
.data
= (unsigned long) sb
;
3923 /* Register extent status tree shrinker */
3924 if (ext4_es_register_shrinker(sbi
))
3927 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
3928 sbi
->s_extent_max_zeroout_kb
= 32;
3931 * set up enough so that it can read an inode
3933 sb
->s_op
= &ext4_sops
;
3934 sb
->s_export_op
= &ext4_export_ops
;
3935 sb
->s_xattr
= ext4_xattr_handlers
;
3937 sb
->dq_op
= &ext4_quota_operations
;
3938 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
))
3939 sb
->s_qcop
= &ext4_qctl_sysfile_operations
;
3941 sb
->s_qcop
= &ext4_qctl_operations
;
3942 sb
->s_quota_types
= QTYPE_MASK_USR
| QTYPE_MASK_GRP
;
3944 memcpy(sb
->s_uuid
, es
->s_uuid
, sizeof(es
->s_uuid
));
3946 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
3947 mutex_init(&sbi
->s_orphan_lock
);
3951 needs_recovery
= (es
->s_last_orphan
!= 0 ||
3952 EXT4_HAS_INCOMPAT_FEATURE(sb
,
3953 EXT4_FEATURE_INCOMPAT_RECOVER
));
3955 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_MMP
) &&
3956 !(sb
->s_flags
& MS_RDONLY
))
3957 if (ext4_multi_mount_protect(sb
, le64_to_cpu(es
->s_mmp_block
)))
3958 goto failed_mount3a
;
3961 * The first inode we look at is the journal inode. Don't try
3962 * root first: it may be modified in the journal!
3964 if (!test_opt(sb
, NOLOAD
) &&
3965 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
3966 if (ext4_load_journal(sb
, es
, journal_devnum
))
3967 goto failed_mount3a
;
3968 } else if (test_opt(sb
, NOLOAD
) && !(sb
->s_flags
& MS_RDONLY
) &&
3969 EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
3970 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
3971 "suppressed and not mounted read-only");
3972 goto failed_mount_wq
;
3974 clear_opt(sb
, DATA_FLAGS
);
3975 sbi
->s_journal
= NULL
;
3980 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
) &&
3981 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
3982 JBD2_FEATURE_INCOMPAT_64BIT
)) {
3983 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
3984 goto failed_mount_wq
;
3987 if (!set_journal_csum_feature_set(sb
)) {
3988 ext4_msg(sb
, KERN_ERR
, "Failed to set journal checksum "
3990 goto failed_mount_wq
;
3993 /* We have now updated the journal if required, so we can
3994 * validate the data journaling mode. */
3995 switch (test_opt(sb
, DATA_FLAGS
)) {
3997 /* No mode set, assume a default based on the journal
3998 * capabilities: ORDERED_DATA if the journal can
3999 * cope, else JOURNAL_DATA
4001 if (jbd2_journal_check_available_features
4002 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
4003 set_opt(sb
, ORDERED_DATA
);
4005 set_opt(sb
, JOURNAL_DATA
);
4008 case EXT4_MOUNT_ORDERED_DATA
:
4009 case EXT4_MOUNT_WRITEBACK_DATA
:
4010 if (!jbd2_journal_check_available_features
4011 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
4012 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
4013 "requested data journaling mode");
4014 goto failed_mount_wq
;
4019 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
4021 sbi
->s_journal
->j_commit_callback
= ext4_journal_commit_callback
;
4024 if (ext4_mballoc_ready
) {
4025 sbi
->s_mb_cache
= ext4_xattr_create_cache(sb
->s_id
);
4026 if (!sbi
->s_mb_cache
) {
4027 ext4_msg(sb
, KERN_ERR
, "Failed to create an mb_cache");
4028 goto failed_mount_wq
;
4033 * Get the # of file system overhead blocks from the
4034 * superblock if present.
4036 if (es
->s_overhead_clusters
)
4037 sbi
->s_overhead
= le32_to_cpu(es
->s_overhead_clusters
);
4039 err
= ext4_calculate_overhead(sb
);
4041 goto failed_mount_wq
;
4045 * The maximum number of concurrent works can be high and
4046 * concurrency isn't really necessary. Limit it to 1.
4048 EXT4_SB(sb
)->rsv_conversion_wq
=
4049 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM
| WQ_UNBOUND
, 1);
4050 if (!EXT4_SB(sb
)->rsv_conversion_wq
) {
4051 printk(KERN_ERR
"EXT4-fs: failed to create workqueue\n");
4057 * The jbd2_journal_load will have done any necessary log recovery,
4058 * so we can safely mount the rest of the filesystem now.
4061 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
4063 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
4064 ret
= PTR_ERR(root
);
4068 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
4069 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
4073 sb
->s_root
= d_make_root(root
);
4075 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
4080 if (ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
))
4081 sb
->s_flags
|= MS_RDONLY
;
4083 /* determine the minimum size of new large inodes, if present */
4084 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
4085 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
4086 EXT4_GOOD_OLD_INODE_SIZE
;
4087 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
4088 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE
)) {
4089 if (sbi
->s_want_extra_isize
<
4090 le16_to_cpu(es
->s_want_extra_isize
))
4091 sbi
->s_want_extra_isize
=
4092 le16_to_cpu(es
->s_want_extra_isize
);
4093 if (sbi
->s_want_extra_isize
<
4094 le16_to_cpu(es
->s_min_extra_isize
))
4095 sbi
->s_want_extra_isize
=
4096 le16_to_cpu(es
->s_min_extra_isize
);
4099 /* Check if enough inode space is available */
4100 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
4101 sbi
->s_inode_size
) {
4102 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
4103 EXT4_GOOD_OLD_INODE_SIZE
;
4104 ext4_msg(sb
, KERN_INFO
, "required extra inode space not"
4108 err
= ext4_reserve_clusters(sbi
, ext4_calculate_resv_clusters(sb
));
4110 ext4_msg(sb
, KERN_ERR
, "failed to reserve %llu clusters for "
4111 "reserved pool", ext4_calculate_resv_clusters(sb
));
4112 goto failed_mount4a
;
4115 err
= ext4_setup_system_zone(sb
);
4117 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
4119 goto failed_mount4a
;
4123 err
= ext4_mb_init(sb
);
4125 ext4_msg(sb
, KERN_ERR
, "failed to initialize mballoc (%d)",
4130 block
= ext4_count_free_clusters(sb
);
4131 ext4_free_blocks_count_set(sbi
->s_es
,
4132 EXT4_C2B(sbi
, block
));
4133 err
= percpu_counter_init(&sbi
->s_freeclusters_counter
, block
,
4136 unsigned long freei
= ext4_count_free_inodes(sb
);
4137 sbi
->s_es
->s_free_inodes_count
= cpu_to_le32(freei
);
4138 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
, freei
,
4142 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
4143 ext4_count_dirs(sb
), GFP_KERNEL
);
4145 err
= percpu_counter_init(&sbi
->s_dirtyclusters_counter
, 0,
4148 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
4152 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
4153 if (!ext4_fill_flex_info(sb
)) {
4154 ext4_msg(sb
, KERN_ERR
,
4155 "unable to initialize "
4156 "flex_bg meta info!");
4160 err
= ext4_register_li_request(sb
, first_not_zeroed
);
4164 sbi
->s_kobj
.kset
= ext4_kset
;
4165 init_completion(&sbi
->s_kobj_unregister
);
4166 err
= kobject_init_and_add(&sbi
->s_kobj
, &ext4_ktype
, NULL
,
4172 /* Enable quota usage during mount. */
4173 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
) &&
4174 !(sb
->s_flags
& MS_RDONLY
)) {
4175 err
= ext4_enable_quotas(sb
);
4179 #endif /* CONFIG_QUOTA */
4181 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
4182 ext4_orphan_cleanup(sb
, es
);
4183 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
4184 if (needs_recovery
) {
4185 ext4_msg(sb
, KERN_INFO
, "recovery complete");
4186 ext4_mark_recovery_complete(sb
, es
);
4188 if (EXT4_SB(sb
)->s_journal
) {
4189 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
4190 descr
= " journalled data mode";
4191 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
4192 descr
= " ordered data mode";
4194 descr
= " writeback data mode";
4196 descr
= "out journal";
4198 if (test_opt(sb
, DISCARD
)) {
4199 struct request_queue
*q
= bdev_get_queue(sb
->s_bdev
);
4200 if (!blk_queue_discard(q
))
4201 ext4_msg(sb
, KERN_WARNING
,
4202 "mounting with \"discard\" option, but "
4203 "the device does not support discard");
4206 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s. "
4207 "Opts: %s%s%s", descr
, sbi
->s_es
->s_mount_opts
,
4208 *sbi
->s_es
->s_mount_opts
? "; " : "", orig_data
);
4210 if (es
->s_error_count
)
4211 mod_timer(&sbi
->s_err_report
, jiffies
+ 300*HZ
); /* 5 minutes */
4213 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4214 ratelimit_state_init(&sbi
->s_err_ratelimit_state
, 5 * HZ
, 10);
4215 ratelimit_state_init(&sbi
->s_warning_ratelimit_state
, 5 * HZ
, 10);
4216 ratelimit_state_init(&sbi
->s_msg_ratelimit_state
, 5 * HZ
, 10);
4223 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
4228 kobject_del(&sbi
->s_kobj
);
4231 ext4_unregister_li_request(sb
);
4233 ext4_mb_release(sb
);
4234 if (sbi
->s_flex_groups
)
4235 ext4_kvfree(sbi
->s_flex_groups
);
4236 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
4237 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
4238 percpu_counter_destroy(&sbi
->s_dirs_counter
);
4239 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
4241 ext4_ext_release(sb
);
4242 ext4_release_system_zone(sb
);
4247 ext4_msg(sb
, KERN_ERR
, "mount failed");
4248 if (EXT4_SB(sb
)->rsv_conversion_wq
)
4249 destroy_workqueue(EXT4_SB(sb
)->rsv_conversion_wq
);
4251 if (sbi
->s_journal
) {
4252 jbd2_journal_destroy(sbi
->s_journal
);
4253 sbi
->s_journal
= NULL
;
4256 ext4_es_unregister_shrinker(sbi
);
4258 del_timer_sync(&sbi
->s_err_report
);
4260 kthread_stop(sbi
->s_mmp_tsk
);
4262 for (i
= 0; i
< db_count
; i
++)
4263 brelse(sbi
->s_group_desc
[i
]);
4264 ext4_kvfree(sbi
->s_group_desc
);
4266 if (sbi
->s_chksum_driver
)
4267 crypto_free_shash(sbi
->s_chksum_driver
);
4269 remove_proc_entry("options", sbi
->s_proc
);
4270 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
4273 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
4274 kfree(sbi
->s_qf_names
[i
]);
4276 ext4_blkdev_remove(sbi
);
4279 sb
->s_fs_info
= NULL
;
4280 kfree(sbi
->s_blockgroup_lock
);
4284 return err
? err
: ret
;
4288 * Setup any per-fs journal parameters now. We'll do this both on
4289 * initial mount, once the journal has been initialised but before we've
4290 * done any recovery; and again on any subsequent remount.
4292 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
4294 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4296 journal
->j_commit_interval
= sbi
->s_commit_interval
;
4297 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
4298 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
4300 write_lock(&journal
->j_state_lock
);
4301 if (test_opt(sb
, BARRIER
))
4302 journal
->j_flags
|= JBD2_BARRIER
;
4304 journal
->j_flags
&= ~JBD2_BARRIER
;
4305 if (test_opt(sb
, DATA_ERR_ABORT
))
4306 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
4308 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
4309 write_unlock(&journal
->j_state_lock
);
4312 static journal_t
*ext4_get_journal(struct super_block
*sb
,
4313 unsigned int journal_inum
)
4315 struct inode
*journal_inode
;
4318 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4320 /* First, test for the existence of a valid inode on disk. Bad
4321 * things happen if we iget() an unused inode, as the subsequent
4322 * iput() will try to delete it. */
4324 journal_inode
= ext4_iget(sb
, journal_inum
);
4325 if (IS_ERR(journal_inode
)) {
4326 ext4_msg(sb
, KERN_ERR
, "no journal found");
4329 if (!journal_inode
->i_nlink
) {
4330 make_bad_inode(journal_inode
);
4331 iput(journal_inode
);
4332 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
4336 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4337 journal_inode
, journal_inode
->i_size
);
4338 if (!S_ISREG(journal_inode
->i_mode
)) {
4339 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
4340 iput(journal_inode
);
4344 journal
= jbd2_journal_init_inode(journal_inode
);
4346 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
4347 iput(journal_inode
);
4350 journal
->j_private
= sb
;
4351 ext4_init_journal_params(sb
, journal
);
4355 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
4358 struct buffer_head
*bh
;
4362 int hblock
, blocksize
;
4363 ext4_fsblk_t sb_block
;
4364 unsigned long offset
;
4365 struct ext4_super_block
*es
;
4366 struct block_device
*bdev
;
4368 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4370 bdev
= ext4_blkdev_get(j_dev
, sb
);
4374 blocksize
= sb
->s_blocksize
;
4375 hblock
= bdev_logical_block_size(bdev
);
4376 if (blocksize
< hblock
) {
4377 ext4_msg(sb
, KERN_ERR
,
4378 "blocksize too small for journal device");
4382 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
4383 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
4384 set_blocksize(bdev
, blocksize
);
4385 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
4386 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
4387 "external journal");
4391 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
4392 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
4393 !(le32_to_cpu(es
->s_feature_incompat
) &
4394 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
4395 ext4_msg(sb
, KERN_ERR
, "external journal has "
4401 if ((le32_to_cpu(es
->s_feature_ro_compat
) &
4402 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
) &&
4403 es
->s_checksum
!= ext4_superblock_csum(sb
, es
)) {
4404 ext4_msg(sb
, KERN_ERR
, "external journal has "
4405 "corrupt superblock");
4410 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
4411 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
4416 len
= ext4_blocks_count(es
);
4417 start
= sb_block
+ 1;
4418 brelse(bh
); /* we're done with the superblock */
4420 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
4421 start
, len
, blocksize
);
4423 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
4426 journal
->j_private
= sb
;
4427 ll_rw_block(READ
| REQ_META
| REQ_PRIO
, 1, &journal
->j_sb_buffer
);
4428 wait_on_buffer(journal
->j_sb_buffer
);
4429 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
4430 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
4433 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
4434 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
4435 "user (unsupported) - %d",
4436 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
4439 EXT4_SB(sb
)->journal_bdev
= bdev
;
4440 ext4_init_journal_params(sb
, journal
);
4444 jbd2_journal_destroy(journal
);
4446 ext4_blkdev_put(bdev
);
4450 static int ext4_load_journal(struct super_block
*sb
,
4451 struct ext4_super_block
*es
,
4452 unsigned long journal_devnum
)
4455 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
4458 int really_read_only
;
4460 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4462 if (journal_devnum
&&
4463 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4464 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
4465 "numbers have changed");
4466 journal_dev
= new_decode_dev(journal_devnum
);
4468 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
4470 really_read_only
= bdev_read_only(sb
->s_bdev
);
4473 * Are we loading a blank journal or performing recovery after a
4474 * crash? For recovery, we need to check in advance whether we
4475 * can get read-write access to the device.
4477 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
4478 if (sb
->s_flags
& MS_RDONLY
) {
4479 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
4480 "required on readonly filesystem");
4481 if (really_read_only
) {
4482 ext4_msg(sb
, KERN_ERR
, "write access "
4483 "unavailable, cannot proceed");
4486 ext4_msg(sb
, KERN_INFO
, "write access will "
4487 "be enabled during recovery");
4491 if (journal_inum
&& journal_dev
) {
4492 ext4_msg(sb
, KERN_ERR
, "filesystem has both journal "
4493 "and inode journals!");
4498 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
4501 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
4505 if (!(journal
->j_flags
& JBD2_BARRIER
))
4506 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
4508 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
))
4509 err
= jbd2_journal_wipe(journal
, !really_read_only
);
4511 char *save
= kmalloc(EXT4_S_ERR_LEN
, GFP_KERNEL
);
4513 memcpy(save
, ((char *) es
) +
4514 EXT4_S_ERR_START
, EXT4_S_ERR_LEN
);
4515 err
= jbd2_journal_load(journal
);
4517 memcpy(((char *) es
) + EXT4_S_ERR_START
,
4518 save
, EXT4_S_ERR_LEN
);
4523 ext4_msg(sb
, KERN_ERR
, "error loading journal");
4524 jbd2_journal_destroy(journal
);
4528 EXT4_SB(sb
)->s_journal
= journal
;
4529 ext4_clear_journal_err(sb
, es
);
4531 if (!really_read_only
&& journal_devnum
&&
4532 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4533 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
4535 /* Make sure we flush the recovery flag to disk. */
4536 ext4_commit_super(sb
, 1);
4542 static int ext4_commit_super(struct super_block
*sb
, int sync
)
4544 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
4545 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
4548 if (!sbh
|| block_device_ejected(sb
))
4550 if (buffer_write_io_error(sbh
)) {
4552 * Oh, dear. A previous attempt to write the
4553 * superblock failed. This could happen because the
4554 * USB device was yanked out. Or it could happen to
4555 * be a transient write error and maybe the block will
4556 * be remapped. Nothing we can do but to retry the
4557 * write and hope for the best.
4559 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
4560 "superblock detected");
4561 clear_buffer_write_io_error(sbh
);
4562 set_buffer_uptodate(sbh
);
4565 * If the file system is mounted read-only, don't update the
4566 * superblock write time. This avoids updating the superblock
4567 * write time when we are mounting the root file system
4568 * read/only but we need to replay the journal; at that point,
4569 * for people who are east of GMT and who make their clock
4570 * tick in localtime for Windows bug-for-bug compatibility,
4571 * the clock is set in the future, and this will cause e2fsck
4572 * to complain and force a full file system check.
4574 if (!(sb
->s_flags
& MS_RDONLY
))
4575 es
->s_wtime
= cpu_to_le32(get_seconds());
4576 if (sb
->s_bdev
->bd_part
)
4577 es
->s_kbytes_written
=
4578 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
+
4579 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
4580 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
4582 es
->s_kbytes_written
=
4583 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
);
4584 if (percpu_counter_initialized(&EXT4_SB(sb
)->s_freeclusters_counter
))
4585 ext4_free_blocks_count_set(es
,
4586 EXT4_C2B(EXT4_SB(sb
), percpu_counter_sum_positive(
4587 &EXT4_SB(sb
)->s_freeclusters_counter
)));
4588 if (percpu_counter_initialized(&EXT4_SB(sb
)->s_freeinodes_counter
))
4589 es
->s_free_inodes_count
=
4590 cpu_to_le32(percpu_counter_sum_positive(
4591 &EXT4_SB(sb
)->s_freeinodes_counter
));
4592 BUFFER_TRACE(sbh
, "marking dirty");
4593 ext4_superblock_csum_set(sb
);
4594 mark_buffer_dirty(sbh
);
4596 error
= sync_dirty_buffer(sbh
);
4600 error
= buffer_write_io_error(sbh
);
4602 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
4604 clear_buffer_write_io_error(sbh
);
4605 set_buffer_uptodate(sbh
);
4612 * Have we just finished recovery? If so, and if we are mounting (or
4613 * remounting) the filesystem readonly, then we will end up with a
4614 * consistent fs on disk. Record that fact.
4616 static void ext4_mark_recovery_complete(struct super_block
*sb
,
4617 struct ext4_super_block
*es
)
4619 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
4621 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
4622 BUG_ON(journal
!= NULL
);
4625 jbd2_journal_lock_updates(journal
);
4626 if (jbd2_journal_flush(journal
) < 0)
4629 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
) &&
4630 sb
->s_flags
& MS_RDONLY
) {
4631 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4632 ext4_commit_super(sb
, 1);
4636 jbd2_journal_unlock_updates(journal
);
4640 * If we are mounting (or read-write remounting) a filesystem whose journal
4641 * has recorded an error from a previous lifetime, move that error to the
4642 * main filesystem now.
4644 static void ext4_clear_journal_err(struct super_block
*sb
,
4645 struct ext4_super_block
*es
)
4651 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4653 journal
= EXT4_SB(sb
)->s_journal
;
4656 * Now check for any error status which may have been recorded in the
4657 * journal by a prior ext4_error() or ext4_abort()
4660 j_errno
= jbd2_journal_errno(journal
);
4664 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
4665 ext4_warning(sb
, "Filesystem error recorded "
4666 "from previous mount: %s", errstr
);
4667 ext4_warning(sb
, "Marking fs in need of filesystem check.");
4669 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
4670 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
4671 ext4_commit_super(sb
, 1);
4673 jbd2_journal_clear_err(journal
);
4674 jbd2_journal_update_sb_errno(journal
);
4679 * Force the running and committing transactions to commit,
4680 * and wait on the commit.
4682 int ext4_force_commit(struct super_block
*sb
)
4686 if (sb
->s_flags
& MS_RDONLY
)
4689 journal
= EXT4_SB(sb
)->s_journal
;
4690 return ext4_journal_force_commit(journal
);
4693 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
4697 bool needs_barrier
= false;
4698 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4700 trace_ext4_sync_fs(sb
, wait
);
4701 flush_workqueue(sbi
->rsv_conversion_wq
);
4703 * Writeback quota in non-journalled quota case - journalled quota has
4706 dquot_writeback_dquots(sb
, -1);
4708 * Data writeback is possible w/o journal transaction, so barrier must
4709 * being sent at the end of the function. But we can skip it if
4710 * transaction_commit will do it for us.
4712 if (sbi
->s_journal
) {
4713 target
= jbd2_get_latest_transaction(sbi
->s_journal
);
4714 if (wait
&& sbi
->s_journal
->j_flags
& JBD2_BARRIER
&&
4715 !jbd2_trans_will_send_data_barrier(sbi
->s_journal
, target
))
4716 needs_barrier
= true;
4718 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
4720 ret
= jbd2_log_wait_commit(sbi
->s_journal
,
4723 } else if (wait
&& test_opt(sb
, BARRIER
))
4724 needs_barrier
= true;
4725 if (needs_barrier
) {
4727 err
= blkdev_issue_flush(sb
->s_bdev
, GFP_KERNEL
, NULL
);
4736 * LVM calls this function before a (read-only) snapshot is created. This
4737 * gives us a chance to flush the journal completely and mark the fs clean.
4739 * Note that only this function cannot bring a filesystem to be in a clean
4740 * state independently. It relies on upper layer to stop all data & metadata
4743 static int ext4_freeze(struct super_block
*sb
)
4748 if (sb
->s_flags
& MS_RDONLY
)
4751 journal
= EXT4_SB(sb
)->s_journal
;
4754 /* Now we set up the journal barrier. */
4755 jbd2_journal_lock_updates(journal
);
4758 * Don't clear the needs_recovery flag if we failed to
4759 * flush the journal.
4761 error
= jbd2_journal_flush(journal
);
4766 /* Journal blocked and flushed, clear needs_recovery flag. */
4767 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4768 error
= ext4_commit_super(sb
, 1);
4771 /* we rely on upper layer to stop further updates */
4772 jbd2_journal_unlock_updates(journal
);
4777 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4778 * flag here, even though the filesystem is not technically dirty yet.
4780 static int ext4_unfreeze(struct super_block
*sb
)
4782 if (sb
->s_flags
& MS_RDONLY
)
4785 /* Reset the needs_recovery flag before the fs is unlocked. */
4786 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4787 ext4_commit_super(sb
, 1);
4792 * Structure to save mount options for ext4_remount's benefit
4794 struct ext4_mount_options
{
4795 unsigned long s_mount_opt
;
4796 unsigned long s_mount_opt2
;
4799 unsigned long s_commit_interval
;
4800 u32 s_min_batch_time
, s_max_batch_time
;
4803 char *s_qf_names
[EXT4_MAXQUOTAS
];
4807 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
4809 struct ext4_super_block
*es
;
4810 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4811 unsigned long old_sb_flags
;
4812 struct ext4_mount_options old_opts
;
4813 int enable_quota
= 0;
4815 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
4820 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
4822 /* Store the original options */
4823 old_sb_flags
= sb
->s_flags
;
4824 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
4825 old_opts
.s_mount_opt2
= sbi
->s_mount_opt2
;
4826 old_opts
.s_resuid
= sbi
->s_resuid
;
4827 old_opts
.s_resgid
= sbi
->s_resgid
;
4828 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
4829 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
4830 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
4832 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
4833 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
4834 if (sbi
->s_qf_names
[i
]) {
4835 old_opts
.s_qf_names
[i
] = kstrdup(sbi
->s_qf_names
[i
],
4837 if (!old_opts
.s_qf_names
[i
]) {
4838 for (j
= 0; j
< i
; j
++)
4839 kfree(old_opts
.s_qf_names
[j
]);
4844 old_opts
.s_qf_names
[i
] = NULL
;
4846 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
4847 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
4850 * Allow the "check" option to be passed as a remount option.
4852 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
, 1)) {
4857 if ((old_opts
.s_mount_opt
& EXT4_MOUNT_JOURNAL_CHECKSUM
) ^
4858 test_opt(sb
, JOURNAL_CHECKSUM
)) {
4859 ext4_msg(sb
, KERN_ERR
, "changing journal_checksum "
4860 "during remount not supported");
4865 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
4866 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
4867 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4868 "both data=journal and delalloc");
4872 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
4873 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4874 "both data=journal and dioread_nolock");
4880 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
)
4881 ext4_abort(sb
, "Abort forced by user");
4883 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
4884 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
4888 if (sbi
->s_journal
) {
4889 ext4_init_journal_params(sb
, sbi
->s_journal
);
4890 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
4893 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
)) {
4894 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
) {
4899 if (*flags
& MS_RDONLY
) {
4900 err
= sync_filesystem(sb
);
4903 err
= dquot_suspend(sb
, -1);
4908 * First of all, the unconditional stuff we have to do
4909 * to disable replay of the journal when we next remount
4911 sb
->s_flags
|= MS_RDONLY
;
4914 * OK, test if we are remounting a valid rw partition
4915 * readonly, and if so set the rdonly flag and then
4916 * mark the partition as valid again.
4918 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
4919 (sbi
->s_mount_state
& EXT4_VALID_FS
))
4920 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
4923 ext4_mark_recovery_complete(sb
, es
);
4925 /* Make sure we can mount this feature set readwrite */
4926 if (!ext4_feature_set_ok(sb
, 0)) {
4931 * Make sure the group descriptor checksums
4932 * are sane. If they aren't, refuse to remount r/w.
4934 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
4935 struct ext4_group_desc
*gdp
=
4936 ext4_get_group_desc(sb
, g
, NULL
);
4938 if (!ext4_group_desc_csum_verify(sb
, g
, gdp
)) {
4939 ext4_msg(sb
, KERN_ERR
,
4940 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4941 g
, le16_to_cpu(ext4_group_desc_csum(sbi
, g
, gdp
)),
4942 le16_to_cpu(gdp
->bg_checksum
));
4949 * If we have an unprocessed orphan list hanging
4950 * around from a previously readonly bdev mount,
4951 * require a full umount/remount for now.
4953 if (es
->s_last_orphan
) {
4954 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
4955 "remount RDWR because of unprocessed "
4956 "orphan inode list. Please "
4957 "umount/remount instead");
4963 * Mounting a RDONLY partition read-write, so reread
4964 * and store the current valid flag. (It may have
4965 * been changed by e2fsck since we originally mounted
4969 ext4_clear_journal_err(sb
, es
);
4970 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
4971 if (!ext4_setup_super(sb
, es
, 0))
4972 sb
->s_flags
&= ~MS_RDONLY
;
4973 if (EXT4_HAS_INCOMPAT_FEATURE(sb
,
4974 EXT4_FEATURE_INCOMPAT_MMP
))
4975 if (ext4_multi_mount_protect(sb
,
4976 le64_to_cpu(es
->s_mmp_block
))) {
4985 * Reinitialize lazy itable initialization thread based on
4988 if ((sb
->s_flags
& MS_RDONLY
) || !test_opt(sb
, INIT_INODE_TABLE
))
4989 ext4_unregister_li_request(sb
);
4991 ext4_group_t first_not_zeroed
;
4992 first_not_zeroed
= ext4_has_uninit_itable(sb
);
4993 ext4_register_li_request(sb
, first_not_zeroed
);
4996 ext4_setup_system_zone(sb
);
4997 if (sbi
->s_journal
== NULL
&& !(old_sb_flags
& MS_RDONLY
))
4998 ext4_commit_super(sb
, 1);
5001 /* Release old quota file names */
5002 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
5003 kfree(old_opts
.s_qf_names
[i
]);
5005 if (sb_any_quota_suspended(sb
))
5006 dquot_resume(sb
, -1);
5007 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
5008 EXT4_FEATURE_RO_COMPAT_QUOTA
)) {
5009 err
= ext4_enable_quotas(sb
);
5016 ext4_msg(sb
, KERN_INFO
, "re-mounted. Opts: %s", orig_data
);
5021 sb
->s_flags
= old_sb_flags
;
5022 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
5023 sbi
->s_mount_opt2
= old_opts
.s_mount_opt2
;
5024 sbi
->s_resuid
= old_opts
.s_resuid
;
5025 sbi
->s_resgid
= old_opts
.s_resgid
;
5026 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
5027 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
5028 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
5030 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
5031 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
5032 kfree(sbi
->s_qf_names
[i
]);
5033 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
5040 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
5042 struct super_block
*sb
= dentry
->d_sb
;
5043 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
5044 struct ext4_super_block
*es
= sbi
->s_es
;
5045 ext4_fsblk_t overhead
= 0, resv_blocks
;
5048 resv_blocks
= EXT4_C2B(sbi
, atomic64_read(&sbi
->s_resv_clusters
));
5050 if (!test_opt(sb
, MINIX_DF
))
5051 overhead
= sbi
->s_overhead
;
5053 buf
->f_type
= EXT4_SUPER_MAGIC
;
5054 buf
->f_bsize
= sb
->s_blocksize
;
5055 buf
->f_blocks
= ext4_blocks_count(es
) - EXT4_C2B(sbi
, overhead
);
5056 bfree
= percpu_counter_sum_positive(&sbi
->s_freeclusters_counter
) -
5057 percpu_counter_sum_positive(&sbi
->s_dirtyclusters_counter
);
5058 /* prevent underflow in case that few free space is available */
5059 buf
->f_bfree
= EXT4_C2B(sbi
, max_t(s64
, bfree
, 0));
5060 buf
->f_bavail
= buf
->f_bfree
-
5061 (ext4_r_blocks_count(es
) + resv_blocks
);
5062 if (buf
->f_bfree
< (ext4_r_blocks_count(es
) + resv_blocks
))
5064 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
5065 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
5066 buf
->f_namelen
= EXT4_NAME_LEN
;
5067 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
5068 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
5069 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
5070 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
5075 /* Helper function for writing quotas on sync - we need to start transaction
5076 * before quota file is locked for write. Otherwise the are possible deadlocks:
5077 * Process 1 Process 2
5078 * ext4_create() quota_sync()
5079 * jbd2_journal_start() write_dquot()
5080 * dquot_initialize() down(dqio_mutex)
5081 * down(dqio_mutex) jbd2_journal_start()
5087 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
5089 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_id
.type
];
5092 static int ext4_write_dquot(struct dquot
*dquot
)
5096 struct inode
*inode
;
5098 inode
= dquot_to_inode(dquot
);
5099 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
,
5100 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
5102 return PTR_ERR(handle
);
5103 ret
= dquot_commit(dquot
);
5104 err
= ext4_journal_stop(handle
);
5110 static int ext4_acquire_dquot(struct dquot
*dquot
)
5115 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
5116 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
5118 return PTR_ERR(handle
);
5119 ret
= dquot_acquire(dquot
);
5120 err
= ext4_journal_stop(handle
);
5126 static int ext4_release_dquot(struct dquot
*dquot
)
5131 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
5132 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
5133 if (IS_ERR(handle
)) {
5134 /* Release dquot anyway to avoid endless cycle in dqput() */
5135 dquot_release(dquot
);
5136 return PTR_ERR(handle
);
5138 ret
= dquot_release(dquot
);
5139 err
= ext4_journal_stop(handle
);
5145 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
5147 struct super_block
*sb
= dquot
->dq_sb
;
5148 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
5150 /* Are we journaling quotas? */
5151 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
) ||
5152 sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
5153 dquot_mark_dquot_dirty(dquot
);
5154 return ext4_write_dquot(dquot
);
5156 return dquot_mark_dquot_dirty(dquot
);
5160 static int ext4_write_info(struct super_block
*sb
, int type
)
5165 /* Data block + inode block */
5166 handle
= ext4_journal_start(sb
->s_root
->d_inode
, EXT4_HT_QUOTA
, 2);
5168 return PTR_ERR(handle
);
5169 ret
= dquot_commit_info(sb
, type
);
5170 err
= ext4_journal_stop(handle
);
5177 * Turn on quotas during mount time - we need to find
5178 * the quota file and such...
5180 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
5182 return dquot_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
5183 EXT4_SB(sb
)->s_jquota_fmt
, type
);
5187 * Standard function to be called on quota_on
5189 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
5194 if (!test_opt(sb
, QUOTA
))
5197 /* Quotafile not on the same filesystem? */
5198 if (path
->dentry
->d_sb
!= sb
)
5200 /* Journaling quota? */
5201 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
5202 /* Quotafile not in fs root? */
5203 if (path
->dentry
->d_parent
!= sb
->s_root
)
5204 ext4_msg(sb
, KERN_WARNING
,
5205 "Quota file not on filesystem root. "
5206 "Journaled quota will not work");
5210 * When we journal data on quota file, we have to flush journal to see
5211 * all updates to the file when we bypass pagecache...
5213 if (EXT4_SB(sb
)->s_journal
&&
5214 ext4_should_journal_data(path
->dentry
->d_inode
)) {
5216 * We don't need to lock updates but journal_flush() could
5217 * otherwise be livelocked...
5219 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
5220 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
5221 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
5226 return dquot_quota_on(sb
, type
, format_id
, path
);
5229 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
5233 struct inode
*qf_inode
;
5234 unsigned long qf_inums
[EXT4_MAXQUOTAS
] = {
5235 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
5236 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
)
5239 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
));
5241 if (!qf_inums
[type
])
5244 qf_inode
= ext4_iget(sb
, qf_inums
[type
]);
5245 if (IS_ERR(qf_inode
)) {
5246 ext4_error(sb
, "Bad quota inode # %lu", qf_inums
[type
]);
5247 return PTR_ERR(qf_inode
);
5250 /* Don't account quota for quota files to avoid recursion */
5251 qf_inode
->i_flags
|= S_NOQUOTA
;
5252 err
= dquot_enable(qf_inode
, type
, format_id
, flags
);
5258 /* Enable usage tracking for all quota types. */
5259 static int ext4_enable_quotas(struct super_block
*sb
)
5262 unsigned long qf_inums
[EXT4_MAXQUOTAS
] = {
5263 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
5264 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
)
5267 sb_dqopt(sb
)->flags
|= DQUOT_QUOTA_SYS_FILE
;
5268 for (type
= 0; type
< EXT4_MAXQUOTAS
; type
++) {
5269 if (qf_inums
[type
]) {
5270 err
= ext4_quota_enable(sb
, type
, QFMT_VFS_V1
,
5271 DQUOT_USAGE_ENABLED
);
5274 "Failed to enable quota tracking "
5275 "(type=%d, err=%d). Please run "
5276 "e2fsck to fix.", type
, err
);
5285 * quota_on function that is used when QUOTA feature is set.
5287 static int ext4_quota_on_sysfile(struct super_block
*sb
, int type
,
5290 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
))
5294 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5296 return ext4_quota_enable(sb
, type
, format_id
, DQUOT_LIMITS_ENABLED
);
5299 static int ext4_quota_off(struct super_block
*sb
, int type
)
5301 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5304 /* Force all delayed allocation blocks to be allocated.
5305 * Caller already holds s_umount sem */
5306 if (test_opt(sb
, DELALLOC
))
5307 sync_filesystem(sb
);
5312 /* Update modification times of quota files when userspace can
5313 * start looking at them */
5314 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
, 1);
5317 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
5318 ext4_mark_inode_dirty(handle
, inode
);
5319 ext4_journal_stop(handle
);
5322 return dquot_quota_off(sb
, type
);
5326 * quota_off function that is used when QUOTA feature is set.
5328 static int ext4_quota_off_sysfile(struct super_block
*sb
, int type
)
5330 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
))
5333 /* Disable only the limits. */
5334 return dquot_disable(sb
, type
, DQUOT_LIMITS_ENABLED
);
5337 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5338 * acquiring the locks... As quota files are never truncated and quota code
5339 * itself serializes the operations (and no one else should touch the files)
5340 * we don't have to be afraid of races */
5341 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
5342 size_t len
, loff_t off
)
5344 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5345 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5346 int offset
= off
& (sb
->s_blocksize
- 1);
5349 struct buffer_head
*bh
;
5350 loff_t i_size
= i_size_read(inode
);
5354 if (off
+len
> i_size
)
5357 while (toread
> 0) {
5358 tocopy
= sb
->s_blocksize
- offset
< toread
?
5359 sb
->s_blocksize
- offset
: toread
;
5360 bh
= ext4_bread(NULL
, inode
, blk
, 0);
5363 if (!bh
) /* A hole? */
5364 memset(data
, 0, tocopy
);
5366 memcpy(data
, bh
->b_data
+offset
, tocopy
);
5376 /* Write to quotafile (we know the transaction is already started and has
5377 * enough credits) */
5378 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
5379 const char *data
, size_t len
, loff_t off
)
5381 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5382 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5383 int err
, offset
= off
& (sb
->s_blocksize
- 1);
5384 struct buffer_head
*bh
;
5385 handle_t
*handle
= journal_current_handle();
5387 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
5388 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5389 " cancelled because transaction is not started",
5390 (unsigned long long)off
, (unsigned long long)len
);
5394 * Since we account only one data block in transaction credits,
5395 * then it is impossible to cross a block boundary.
5397 if (sb
->s_blocksize
- offset
< len
) {
5398 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5399 " cancelled because not block aligned",
5400 (unsigned long long)off
, (unsigned long long)len
);
5404 bh
= ext4_bread(handle
, inode
, blk
, 1);
5409 BUFFER_TRACE(bh
, "get write access");
5410 err
= ext4_journal_get_write_access(handle
, bh
);
5416 memcpy(bh
->b_data
+offset
, data
, len
);
5417 flush_dcache_page(bh
->b_page
);
5419 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
5422 if (inode
->i_size
< off
+ len
) {
5423 i_size_write(inode
, off
+ len
);
5424 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
5425 ext4_mark_inode_dirty(handle
, inode
);
5432 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
5433 const char *dev_name
, void *data
)
5435 return mount_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
);
5438 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5439 static inline void register_as_ext2(void)
5441 int err
= register_filesystem(&ext2_fs_type
);
5444 "EXT4-fs: Unable to register as ext2 (%d)\n", err
);
5447 static inline void unregister_as_ext2(void)
5449 unregister_filesystem(&ext2_fs_type
);
5452 static inline int ext2_feature_set_ok(struct super_block
*sb
)
5454 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT2_FEATURE_INCOMPAT_SUPP
))
5456 if (sb
->s_flags
& MS_RDONLY
)
5458 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT2_FEATURE_RO_COMPAT_SUPP
))
5463 static inline void register_as_ext2(void) { }
5464 static inline void unregister_as_ext2(void) { }
5465 static inline int ext2_feature_set_ok(struct super_block
*sb
) { return 0; }
5468 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5469 static inline void register_as_ext3(void)
5471 int err
= register_filesystem(&ext3_fs_type
);
5474 "EXT4-fs: Unable to register as ext3 (%d)\n", err
);
5477 static inline void unregister_as_ext3(void)
5479 unregister_filesystem(&ext3_fs_type
);
5482 static inline int ext3_feature_set_ok(struct super_block
*sb
)
5484 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT3_FEATURE_INCOMPAT_SUPP
))
5486 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
5488 if (sb
->s_flags
& MS_RDONLY
)
5490 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT3_FEATURE_RO_COMPAT_SUPP
))
5495 static inline void register_as_ext3(void) { }
5496 static inline void unregister_as_ext3(void) { }
5497 static inline int ext3_feature_set_ok(struct super_block
*sb
) { return 0; }
5500 static struct file_system_type ext4_fs_type
= {
5501 .owner
= THIS_MODULE
,
5503 .mount
= ext4_mount
,
5504 .kill_sb
= kill_block_super
,
5505 .fs_flags
= FS_REQUIRES_DEV
,
5507 MODULE_ALIAS_FS("ext4");
5509 static int __init
ext4_init_feat_adverts(void)
5511 struct ext4_features
*ef
;
5514 ef
= kzalloc(sizeof(struct ext4_features
), GFP_KERNEL
);
5518 ef
->f_kobj
.kset
= ext4_kset
;
5519 init_completion(&ef
->f_kobj_unregister
);
5520 ret
= kobject_init_and_add(&ef
->f_kobj
, &ext4_feat_ktype
, NULL
,
5533 static void ext4_exit_feat_adverts(void)
5535 kobject_put(&ext4_feat
->f_kobj
);
5536 wait_for_completion(&ext4_feat
->f_kobj_unregister
);
5540 /* Shared across all ext4 file systems */
5541 wait_queue_head_t ext4__ioend_wq
[EXT4_WQ_HASH_SZ
];
5542 struct mutex ext4__aio_mutex
[EXT4_WQ_HASH_SZ
];
5544 static int __init
ext4_init_fs(void)
5548 ext4_li_info
= NULL
;
5549 mutex_init(&ext4_li_mtx
);
5551 /* Build-time check for flags consistency */
5552 ext4_check_flag_values();
5554 for (i
= 0; i
< EXT4_WQ_HASH_SZ
; i
++) {
5555 mutex_init(&ext4__aio_mutex
[i
]);
5556 init_waitqueue_head(&ext4__ioend_wq
[i
]);
5559 err
= ext4_init_es();
5563 err
= ext4_init_pageio();
5567 err
= ext4_init_system_zone();
5570 ext4_kset
= kset_create_and_add("ext4", NULL
, fs_kobj
);
5575 ext4_proc_root
= proc_mkdir("fs/ext4", NULL
);
5577 err
= ext4_init_feat_adverts();
5581 err
= ext4_init_mballoc();
5585 ext4_mballoc_ready
= 1;
5586 err
= init_inodecache();
5591 err
= register_filesystem(&ext4_fs_type
);
5597 unregister_as_ext2();
5598 unregister_as_ext3();
5599 destroy_inodecache();
5601 ext4_mballoc_ready
= 0;
5602 ext4_exit_mballoc();
5604 ext4_exit_feat_adverts();
5607 remove_proc_entry("fs/ext4", NULL
);
5608 kset_unregister(ext4_kset
);
5610 ext4_exit_system_zone();
5619 static void __exit
ext4_exit_fs(void)
5621 ext4_destroy_lazyinit_thread();
5622 unregister_as_ext2();
5623 unregister_as_ext3();
5624 unregister_filesystem(&ext4_fs_type
);
5625 destroy_inodecache();
5626 ext4_exit_mballoc();
5627 ext4_exit_feat_adverts();
5628 remove_proc_entry("fs/ext4", NULL
);
5629 kset_unregister(ext4_kset
);
5630 ext4_exit_system_zone();
5635 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5636 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5637 MODULE_LICENSE("GPL");
5638 module_init(ext4_init_fs
)
5639 module_exit(ext4_exit_fs
)