2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/parser.h>
28 #include <linux/buffer_head.h>
29 #include <linux/exportfs.h>
30 #include <linux/vfs.h>
31 #include <linux/random.h>
32 #include <linux/mount.h>
33 #include <linux/namei.h>
34 #include <linux/quotaops.h>
35 #include <linux/seq_file.h>
36 #include <linux/proc_fs.h>
37 #include <linux/ctype.h>
38 #include <linux/log2.h>
39 #include <linux/crc16.h>
40 #include <linux/cleancache.h>
41 #include <asm/uaccess.h>
43 #include <linux/kthread.h>
44 #include <linux/freezer.h>
47 #include "ext4_extents.h" /* Needed for trace points definition */
48 #include "ext4_jbd2.h"
53 #define CREATE_TRACE_POINTS
54 #include <trace/events/ext4.h>
56 static struct proc_dir_entry
*ext4_proc_root
;
57 static struct kset
*ext4_kset
;
58 static struct ext4_lazy_init
*ext4_li_info
;
59 static struct mutex ext4_li_mtx
;
60 static struct ext4_features
*ext4_feat
;
61 static int ext4_mballoc_ready
;
63 static int ext4_load_journal(struct super_block
*, struct ext4_super_block
*,
64 unsigned long journal_devnum
);
65 static int ext4_show_options(struct seq_file
*seq
, struct dentry
*root
);
66 static int ext4_commit_super(struct super_block
*sb
, int sync
);
67 static void ext4_mark_recovery_complete(struct super_block
*sb
,
68 struct ext4_super_block
*es
);
69 static void ext4_clear_journal_err(struct super_block
*sb
,
70 struct ext4_super_block
*es
);
71 static int ext4_sync_fs(struct super_block
*sb
, int wait
);
72 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
);
73 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
74 static int ext4_unfreeze(struct super_block
*sb
);
75 static int ext4_freeze(struct super_block
*sb
);
76 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
77 const char *dev_name
, void *data
);
78 static inline int ext2_feature_set_ok(struct super_block
*sb
);
79 static inline int ext3_feature_set_ok(struct super_block
*sb
);
80 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
);
81 static void ext4_destroy_lazyinit_thread(void);
82 static void ext4_unregister_li_request(struct super_block
*sb
);
83 static void ext4_clear_request_list(void);
84 static int ext4_reserve_clusters(struct ext4_sb_info
*, ext4_fsblk_t
);
86 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
87 static struct file_system_type ext2_fs_type
= {
91 .kill_sb
= kill_block_super
,
92 .fs_flags
= FS_REQUIRES_DEV
,
94 MODULE_ALIAS_FS("ext2");
96 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
98 #define IS_EXT2_SB(sb) (0)
102 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
103 static struct file_system_type ext3_fs_type
= {
104 .owner
= THIS_MODULE
,
107 .kill_sb
= kill_block_super
,
108 .fs_flags
= FS_REQUIRES_DEV
,
110 MODULE_ALIAS_FS("ext3");
111 MODULE_ALIAS("ext3");
112 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
114 #define IS_EXT3_SB(sb) (0)
117 static int ext4_verify_csum_type(struct super_block
*sb
,
118 struct ext4_super_block
*es
)
120 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
,
121 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
))
124 return es
->s_checksum_type
== EXT4_CRC32C_CHKSUM
;
127 static __le32
ext4_superblock_csum(struct super_block
*sb
,
128 struct ext4_super_block
*es
)
130 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
131 int offset
= offsetof(struct ext4_super_block
, s_checksum
);
134 csum
= ext4_chksum(sbi
, ~0, (char *)es
, offset
);
136 return cpu_to_le32(csum
);
139 static int ext4_superblock_csum_verify(struct super_block
*sb
,
140 struct ext4_super_block
*es
)
142 if (!ext4_has_metadata_csum(sb
))
145 return es
->s_checksum
== ext4_superblock_csum(sb
, es
);
148 void ext4_superblock_csum_set(struct super_block
*sb
)
150 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
152 if (!ext4_has_metadata_csum(sb
))
155 es
->s_checksum
= ext4_superblock_csum(sb
, es
);
158 void *ext4_kvmalloc(size_t size
, gfp_t flags
)
162 ret
= kmalloc(size
, flags
| __GFP_NOWARN
);
164 ret
= __vmalloc(size
, flags
, PAGE_KERNEL
);
168 void *ext4_kvzalloc(size_t size
, gfp_t flags
)
172 ret
= kzalloc(size
, flags
| __GFP_NOWARN
);
174 ret
= __vmalloc(size
, flags
| __GFP_ZERO
, PAGE_KERNEL
);
178 ext4_fsblk_t
ext4_block_bitmap(struct super_block
*sb
,
179 struct ext4_group_desc
*bg
)
181 return le32_to_cpu(bg
->bg_block_bitmap_lo
) |
182 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
183 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_block_bitmap_hi
) << 32 : 0);
186 ext4_fsblk_t
ext4_inode_bitmap(struct super_block
*sb
,
187 struct ext4_group_desc
*bg
)
189 return le32_to_cpu(bg
->bg_inode_bitmap_lo
) |
190 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
191 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_bitmap_hi
) << 32 : 0);
194 ext4_fsblk_t
ext4_inode_table(struct super_block
*sb
,
195 struct ext4_group_desc
*bg
)
197 return le32_to_cpu(bg
->bg_inode_table_lo
) |
198 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
199 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_table_hi
) << 32 : 0);
202 __u32
ext4_free_group_clusters(struct super_block
*sb
,
203 struct ext4_group_desc
*bg
)
205 return le16_to_cpu(bg
->bg_free_blocks_count_lo
) |
206 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
207 (__u32
)le16_to_cpu(bg
->bg_free_blocks_count_hi
) << 16 : 0);
210 __u32
ext4_free_inodes_count(struct super_block
*sb
,
211 struct ext4_group_desc
*bg
)
213 return le16_to_cpu(bg
->bg_free_inodes_count_lo
) |
214 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
215 (__u32
)le16_to_cpu(bg
->bg_free_inodes_count_hi
) << 16 : 0);
218 __u32
ext4_used_dirs_count(struct super_block
*sb
,
219 struct ext4_group_desc
*bg
)
221 return le16_to_cpu(bg
->bg_used_dirs_count_lo
) |
222 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
223 (__u32
)le16_to_cpu(bg
->bg_used_dirs_count_hi
) << 16 : 0);
226 __u32
ext4_itable_unused_count(struct super_block
*sb
,
227 struct ext4_group_desc
*bg
)
229 return le16_to_cpu(bg
->bg_itable_unused_lo
) |
230 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
231 (__u32
)le16_to_cpu(bg
->bg_itable_unused_hi
) << 16 : 0);
234 void ext4_block_bitmap_set(struct super_block
*sb
,
235 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
237 bg
->bg_block_bitmap_lo
= cpu_to_le32((u32
)blk
);
238 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
239 bg
->bg_block_bitmap_hi
= cpu_to_le32(blk
>> 32);
242 void ext4_inode_bitmap_set(struct super_block
*sb
,
243 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
245 bg
->bg_inode_bitmap_lo
= cpu_to_le32((u32
)blk
);
246 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
247 bg
->bg_inode_bitmap_hi
= cpu_to_le32(blk
>> 32);
250 void ext4_inode_table_set(struct super_block
*sb
,
251 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
253 bg
->bg_inode_table_lo
= cpu_to_le32((u32
)blk
);
254 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
255 bg
->bg_inode_table_hi
= cpu_to_le32(blk
>> 32);
258 void ext4_free_group_clusters_set(struct super_block
*sb
,
259 struct ext4_group_desc
*bg
, __u32 count
)
261 bg
->bg_free_blocks_count_lo
= cpu_to_le16((__u16
)count
);
262 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
263 bg
->bg_free_blocks_count_hi
= cpu_to_le16(count
>> 16);
266 void ext4_free_inodes_set(struct super_block
*sb
,
267 struct ext4_group_desc
*bg
, __u32 count
)
269 bg
->bg_free_inodes_count_lo
= cpu_to_le16((__u16
)count
);
270 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
271 bg
->bg_free_inodes_count_hi
= cpu_to_le16(count
>> 16);
274 void ext4_used_dirs_set(struct super_block
*sb
,
275 struct ext4_group_desc
*bg
, __u32 count
)
277 bg
->bg_used_dirs_count_lo
= cpu_to_le16((__u16
)count
);
278 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
279 bg
->bg_used_dirs_count_hi
= cpu_to_le16(count
>> 16);
282 void ext4_itable_unused_set(struct super_block
*sb
,
283 struct ext4_group_desc
*bg
, __u32 count
)
285 bg
->bg_itable_unused_lo
= cpu_to_le16((__u16
)count
);
286 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
287 bg
->bg_itable_unused_hi
= cpu_to_le16(count
>> 16);
291 static void __save_error_info(struct super_block
*sb
, const char *func
,
294 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
296 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
297 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
298 es
->s_last_error_time
= cpu_to_le32(get_seconds());
299 strncpy(es
->s_last_error_func
, func
, sizeof(es
->s_last_error_func
));
300 es
->s_last_error_line
= cpu_to_le32(line
);
301 if (!es
->s_first_error_time
) {
302 es
->s_first_error_time
= es
->s_last_error_time
;
303 strncpy(es
->s_first_error_func
, func
,
304 sizeof(es
->s_first_error_func
));
305 es
->s_first_error_line
= cpu_to_le32(line
);
306 es
->s_first_error_ino
= es
->s_last_error_ino
;
307 es
->s_first_error_block
= es
->s_last_error_block
;
310 * Start the daily error reporting function if it hasn't been
313 if (!es
->s_error_count
)
314 mod_timer(&EXT4_SB(sb
)->s_err_report
, jiffies
+ 24*60*60*HZ
);
315 le32_add_cpu(&es
->s_error_count
, 1);
318 static void save_error_info(struct super_block
*sb
, const char *func
,
321 __save_error_info(sb
, func
, line
);
322 ext4_commit_super(sb
, 1);
326 * The del_gendisk() function uninitializes the disk-specific data
327 * structures, including the bdi structure, without telling anyone
328 * else. Once this happens, any attempt to call mark_buffer_dirty()
329 * (for example, by ext4_commit_super), will cause a kernel OOPS.
330 * This is a kludge to prevent these oops until we can put in a proper
331 * hook in del_gendisk() to inform the VFS and file system layers.
333 static int block_device_ejected(struct super_block
*sb
)
335 struct inode
*bd_inode
= sb
->s_bdev
->bd_inode
;
336 struct backing_dev_info
*bdi
= inode_to_bdi(bd_inode
);
338 return bdi
->dev
== NULL
;
341 static void ext4_journal_commit_callback(journal_t
*journal
, transaction_t
*txn
)
343 struct super_block
*sb
= journal
->j_private
;
344 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
345 int error
= is_journal_aborted(journal
);
346 struct ext4_journal_cb_entry
*jce
;
348 BUG_ON(txn
->t_state
== T_FINISHED
);
349 spin_lock(&sbi
->s_md_lock
);
350 while (!list_empty(&txn
->t_private_list
)) {
351 jce
= list_entry(txn
->t_private_list
.next
,
352 struct ext4_journal_cb_entry
, jce_list
);
353 list_del_init(&jce
->jce_list
);
354 spin_unlock(&sbi
->s_md_lock
);
355 jce
->jce_func(sb
, jce
, error
);
356 spin_lock(&sbi
->s_md_lock
);
358 spin_unlock(&sbi
->s_md_lock
);
361 /* Deal with the reporting of failure conditions on a filesystem such as
362 * inconsistencies detected or read IO failures.
364 * On ext2, we can store the error state of the filesystem in the
365 * superblock. That is not possible on ext4, because we may have other
366 * write ordering constraints on the superblock which prevent us from
367 * writing it out straight away; and given that the journal is about to
368 * be aborted, we can't rely on the current, or future, transactions to
369 * write out the superblock safely.
371 * We'll just use the jbd2_journal_abort() error code to record an error in
372 * the journal instead. On recovery, the journal will complain about
373 * that error until we've noted it down and cleared it.
376 static void ext4_handle_error(struct super_block
*sb
)
378 if (sb
->s_flags
& MS_RDONLY
)
381 if (!test_opt(sb
, ERRORS_CONT
)) {
382 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
384 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
386 jbd2_journal_abort(journal
, -EIO
);
388 if (test_opt(sb
, ERRORS_RO
)) {
389 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
391 * Make sure updated value of ->s_mount_flags will be visible
392 * before ->s_flags update
395 sb
->s_flags
|= MS_RDONLY
;
397 if (test_opt(sb
, ERRORS_PANIC
))
398 panic("EXT4-fs (device %s): panic forced after error\n",
402 #define ext4_error_ratelimit(sb) \
403 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
406 void __ext4_error(struct super_block
*sb
, const char *function
,
407 unsigned int line
, const char *fmt
, ...)
409 struct va_format vaf
;
412 if (ext4_error_ratelimit(sb
)) {
417 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
418 sb
->s_id
, function
, line
, current
->comm
, &vaf
);
421 save_error_info(sb
, function
, line
);
422 ext4_handle_error(sb
);
425 void __ext4_error_inode(struct inode
*inode
, const char *function
,
426 unsigned int line
, ext4_fsblk_t block
,
427 const char *fmt
, ...)
430 struct va_format vaf
;
431 struct ext4_super_block
*es
= EXT4_SB(inode
->i_sb
)->s_es
;
433 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
434 es
->s_last_error_block
= cpu_to_le64(block
);
435 if (ext4_error_ratelimit(inode
->i_sb
)) {
440 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
441 "inode #%lu: block %llu: comm %s: %pV\n",
442 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
443 block
, current
->comm
, &vaf
);
445 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
446 "inode #%lu: comm %s: %pV\n",
447 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
448 current
->comm
, &vaf
);
451 save_error_info(inode
->i_sb
, function
, line
);
452 ext4_handle_error(inode
->i_sb
);
455 void __ext4_error_file(struct file
*file
, const char *function
,
456 unsigned int line
, ext4_fsblk_t block
,
457 const char *fmt
, ...)
460 struct va_format vaf
;
461 struct ext4_super_block
*es
;
462 struct inode
*inode
= file_inode(file
);
463 char pathname
[80], *path
;
465 es
= EXT4_SB(inode
->i_sb
)->s_es
;
466 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
467 if (ext4_error_ratelimit(inode
->i_sb
)) {
468 path
= d_path(&(file
->f_path
), pathname
, sizeof(pathname
));
476 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
477 "block %llu: comm %s: path %s: %pV\n",
478 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
479 block
, current
->comm
, path
, &vaf
);
482 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
483 "comm %s: path %s: %pV\n",
484 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
485 current
->comm
, path
, &vaf
);
488 save_error_info(inode
->i_sb
, function
, line
);
489 ext4_handle_error(inode
->i_sb
);
492 const char *ext4_decode_error(struct super_block
*sb
, int errno
,
499 errstr
= "IO failure";
502 errstr
= "Out of memory";
505 if (!sb
|| (EXT4_SB(sb
)->s_journal
&&
506 EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
))
507 errstr
= "Journal has aborted";
509 errstr
= "Readonly filesystem";
512 /* If the caller passed in an extra buffer for unknown
513 * errors, textualise them now. Else we just return
516 /* Check for truncated error codes... */
517 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
526 /* __ext4_std_error decodes expected errors from journaling functions
527 * automatically and invokes the appropriate error response. */
529 void __ext4_std_error(struct super_block
*sb
, const char *function
,
530 unsigned int line
, int errno
)
535 /* Special case: if the error is EROFS, and we're not already
536 * inside a transaction, then there's really no point in logging
538 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
539 (sb
->s_flags
& MS_RDONLY
))
542 if (ext4_error_ratelimit(sb
)) {
543 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
544 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s:%d: %s\n",
545 sb
->s_id
, function
, line
, errstr
);
548 save_error_info(sb
, function
, line
);
549 ext4_handle_error(sb
);
553 * ext4_abort is a much stronger failure handler than ext4_error. The
554 * abort function may be used to deal with unrecoverable failures such
555 * as journal IO errors or ENOMEM at a critical moment in log management.
557 * We unconditionally force the filesystem into an ABORT|READONLY state,
558 * unless the error response on the fs has been set to panic in which
559 * case we take the easy way out and panic immediately.
562 void __ext4_abort(struct super_block
*sb
, const char *function
,
563 unsigned int line
, const char *fmt
, ...)
567 save_error_info(sb
, function
, line
);
569 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: ", sb
->s_id
,
575 if ((sb
->s_flags
& MS_RDONLY
) == 0) {
576 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
577 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
579 * Make sure updated value of ->s_mount_flags will be visible
580 * before ->s_flags update
583 sb
->s_flags
|= MS_RDONLY
;
584 if (EXT4_SB(sb
)->s_journal
)
585 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
586 save_error_info(sb
, function
, line
);
588 if (test_opt(sb
, ERRORS_PANIC
))
589 panic("EXT4-fs panic from previous error\n");
592 void __ext4_msg(struct super_block
*sb
,
593 const char *prefix
, const char *fmt
, ...)
595 struct va_format vaf
;
598 if (!___ratelimit(&(EXT4_SB(sb
)->s_msg_ratelimit_state
), "EXT4-fs"))
604 printk("%sEXT4-fs (%s): %pV\n", prefix
, sb
->s_id
, &vaf
);
608 void __ext4_warning(struct super_block
*sb
, const char *function
,
609 unsigned int line
, const char *fmt
, ...)
611 struct va_format vaf
;
614 if (!___ratelimit(&(EXT4_SB(sb
)->s_warning_ratelimit_state
),
621 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: %pV\n",
622 sb
->s_id
, function
, line
, &vaf
);
626 void __ext4_grp_locked_error(const char *function
, unsigned int line
,
627 struct super_block
*sb
, ext4_group_t grp
,
628 unsigned long ino
, ext4_fsblk_t block
,
629 const char *fmt
, ...)
633 struct va_format vaf
;
635 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
637 es
->s_last_error_ino
= cpu_to_le32(ino
);
638 es
->s_last_error_block
= cpu_to_le64(block
);
639 __save_error_info(sb
, function
, line
);
641 if (ext4_error_ratelimit(sb
)) {
645 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: group %u, ",
646 sb
->s_id
, function
, line
, grp
);
648 printk(KERN_CONT
"inode %lu: ", ino
);
650 printk(KERN_CONT
"block %llu:",
651 (unsigned long long) block
);
652 printk(KERN_CONT
"%pV\n", &vaf
);
656 if (test_opt(sb
, ERRORS_CONT
)) {
657 ext4_commit_super(sb
, 0);
661 ext4_unlock_group(sb
, grp
);
662 ext4_handle_error(sb
);
664 * We only get here in the ERRORS_RO case; relocking the group
665 * may be dangerous, but nothing bad will happen since the
666 * filesystem will have already been marked read/only and the
667 * journal has been aborted. We return 1 as a hint to callers
668 * who might what to use the return value from
669 * ext4_grp_locked_error() to distinguish between the
670 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
671 * aggressively from the ext4 function in question, with a
672 * more appropriate error code.
674 ext4_lock_group(sb
, grp
);
678 void ext4_update_dynamic_rev(struct super_block
*sb
)
680 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
682 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
686 "updating to rev %d because of new feature flag, "
687 "running e2fsck is recommended",
690 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
691 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
692 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
693 /* leave es->s_feature_*compat flags alone */
694 /* es->s_uuid will be set by e2fsck if empty */
697 * The rest of the superblock fields should be zero, and if not it
698 * means they are likely already in use, so leave them alone. We
699 * can leave it up to e2fsck to clean up any inconsistencies there.
704 * Open the external journal device
706 static struct block_device
*ext4_blkdev_get(dev_t dev
, struct super_block
*sb
)
708 struct block_device
*bdev
;
709 char b
[BDEVNAME_SIZE
];
711 bdev
= blkdev_get_by_dev(dev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
, sb
);
717 ext4_msg(sb
, KERN_ERR
, "failed to open journal device %s: %ld",
718 __bdevname(dev
, b
), PTR_ERR(bdev
));
723 * Release the journal device
725 static void ext4_blkdev_put(struct block_device
*bdev
)
727 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
);
730 static void ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
732 struct block_device
*bdev
;
733 bdev
= sbi
->journal_bdev
;
735 ext4_blkdev_put(bdev
);
736 sbi
->journal_bdev
= NULL
;
740 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
742 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
745 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
749 ext4_msg(sb
, KERN_ERR
, "sb orphan head is %d",
750 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
752 printk(KERN_ERR
"sb_info orphan list:\n");
753 list_for_each(l
, &sbi
->s_orphan
) {
754 struct inode
*inode
= orphan_list_entry(l
);
756 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
757 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
758 inode
->i_mode
, inode
->i_nlink
,
763 static void ext4_put_super(struct super_block
*sb
)
765 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
766 struct ext4_super_block
*es
= sbi
->s_es
;
769 ext4_unregister_li_request(sb
);
770 dquot_disable(sb
, -1, DQUOT_USAGE_ENABLED
| DQUOT_LIMITS_ENABLED
);
772 flush_workqueue(sbi
->rsv_conversion_wq
);
773 destroy_workqueue(sbi
->rsv_conversion_wq
);
775 if (sbi
->s_journal
) {
776 err
= jbd2_journal_destroy(sbi
->s_journal
);
777 sbi
->s_journal
= NULL
;
779 ext4_abort(sb
, "Couldn't clean up the journal");
782 ext4_es_unregister_shrinker(sbi
);
783 del_timer_sync(&sbi
->s_err_report
);
784 ext4_release_system_zone(sb
);
786 ext4_ext_release(sb
);
787 ext4_xattr_put_super(sb
);
789 if (!(sb
->s_flags
& MS_RDONLY
)) {
790 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
791 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
793 if (!(sb
->s_flags
& MS_RDONLY
))
794 ext4_commit_super(sb
, 1);
797 remove_proc_entry("options", sbi
->s_proc
);
798 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
800 kobject_del(&sbi
->s_kobj
);
802 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
803 brelse(sbi
->s_group_desc
[i
]);
804 kvfree(sbi
->s_group_desc
);
805 kvfree(sbi
->s_flex_groups
);
806 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
807 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
808 percpu_counter_destroy(&sbi
->s_dirs_counter
);
809 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
812 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
813 kfree(sbi
->s_qf_names
[i
]);
816 /* Debugging code just in case the in-memory inode orphan list
817 * isn't empty. The on-disk one can be non-empty if we've
818 * detected an error and taken the fs readonly, but the
819 * in-memory list had better be clean by this point. */
820 if (!list_empty(&sbi
->s_orphan
))
821 dump_orphan_list(sb
, sbi
);
822 J_ASSERT(list_empty(&sbi
->s_orphan
));
824 invalidate_bdev(sb
->s_bdev
);
825 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
827 * Invalidate the journal device's buffers. We don't want them
828 * floating about in memory - the physical journal device may
829 * hotswapped, and it breaks the `ro-after' testing code.
831 sync_blockdev(sbi
->journal_bdev
);
832 invalidate_bdev(sbi
->journal_bdev
);
833 ext4_blkdev_remove(sbi
);
835 if (sbi
->s_mb_cache
) {
836 ext4_xattr_destroy_cache(sbi
->s_mb_cache
);
837 sbi
->s_mb_cache
= NULL
;
840 kthread_stop(sbi
->s_mmp_tsk
);
841 sb
->s_fs_info
= NULL
;
843 * Now that we are completely done shutting down the
844 * superblock, we need to actually destroy the kobject.
846 kobject_put(&sbi
->s_kobj
);
847 wait_for_completion(&sbi
->s_kobj_unregister
);
848 if (sbi
->s_chksum_driver
)
849 crypto_free_shash(sbi
->s_chksum_driver
);
850 kfree(sbi
->s_blockgroup_lock
);
854 static struct kmem_cache
*ext4_inode_cachep
;
857 * Called inside transaction, so use GFP_NOFS
859 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
861 struct ext4_inode_info
*ei
;
863 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
867 ei
->vfs_inode
.i_version
= 1;
868 spin_lock_init(&ei
->i_raw_lock
);
869 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
870 spin_lock_init(&ei
->i_prealloc_lock
);
871 ext4_es_init_tree(&ei
->i_es_tree
);
872 rwlock_init(&ei
->i_es_lock
);
873 INIT_LIST_HEAD(&ei
->i_es_list
);
876 ei
->i_es_shrink_lblk
= 0;
877 ei
->i_reserved_data_blocks
= 0;
878 ei
->i_reserved_meta_blocks
= 0;
879 ei
->i_allocated_meta_blocks
= 0;
880 ei
->i_da_metadata_calc_len
= 0;
881 ei
->i_da_metadata_calc_last_lblock
= 0;
882 spin_lock_init(&(ei
->i_block_reservation_lock
));
884 ei
->i_reserved_quota
= 0;
885 memset(&ei
->i_dquot
, 0, sizeof(ei
->i_dquot
));
888 INIT_LIST_HEAD(&ei
->i_rsv_conversion_list
);
889 spin_lock_init(&ei
->i_completed_io_lock
);
891 ei
->i_datasync_tid
= 0;
892 atomic_set(&ei
->i_ioend_count
, 0);
893 atomic_set(&ei
->i_unwritten
, 0);
894 INIT_WORK(&ei
->i_rsv_conversion_work
, ext4_end_io_rsv_work
);
896 return &ei
->vfs_inode
;
899 static int ext4_drop_inode(struct inode
*inode
)
901 int drop
= generic_drop_inode(inode
);
903 trace_ext4_drop_inode(inode
, drop
);
907 static void ext4_i_callback(struct rcu_head
*head
)
909 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
910 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
913 static void ext4_destroy_inode(struct inode
*inode
)
915 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
916 ext4_msg(inode
->i_sb
, KERN_ERR
,
917 "Inode %lu (%p): orphan list check failed!",
918 inode
->i_ino
, EXT4_I(inode
));
919 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
920 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
924 call_rcu(&inode
->i_rcu
, ext4_i_callback
);
927 static void init_once(void *foo
)
929 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
931 INIT_LIST_HEAD(&ei
->i_orphan
);
932 init_rwsem(&ei
->xattr_sem
);
933 init_rwsem(&ei
->i_data_sem
);
934 inode_init_once(&ei
->vfs_inode
);
937 static int __init
init_inodecache(void)
939 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
940 sizeof(struct ext4_inode_info
),
941 0, (SLAB_RECLAIM_ACCOUNT
|
944 if (ext4_inode_cachep
== NULL
)
949 static void destroy_inodecache(void)
952 * Make sure all delayed rcu free inodes are flushed before we
956 kmem_cache_destroy(ext4_inode_cachep
);
959 void ext4_clear_inode(struct inode
*inode
)
961 invalidate_inode_buffers(inode
);
964 ext4_discard_preallocations(inode
);
965 ext4_es_remove_extent(inode
, 0, EXT_MAX_BLOCKS
);
966 if (EXT4_I(inode
)->jinode
) {
967 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode
),
968 EXT4_I(inode
)->jinode
);
969 jbd2_free_inode(EXT4_I(inode
)->jinode
);
970 EXT4_I(inode
)->jinode
= NULL
;
974 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
975 u64 ino
, u32 generation
)
979 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
980 return ERR_PTR(-ESTALE
);
981 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
982 return ERR_PTR(-ESTALE
);
984 /* iget isn't really right if the inode is currently unallocated!!
986 * ext4_read_inode will return a bad_inode if the inode had been
987 * deleted, so we should be safe.
989 * Currently we don't know the generation for parent directory, so
990 * a generation of 0 means "accept any"
992 inode
= ext4_iget_normal(sb
, ino
);
994 return ERR_CAST(inode
);
995 if (generation
&& inode
->i_generation
!= generation
) {
997 return ERR_PTR(-ESTALE
);
1003 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1004 int fh_len
, int fh_type
)
1006 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
1007 ext4_nfs_get_inode
);
1010 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1011 int fh_len
, int fh_type
)
1013 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1014 ext4_nfs_get_inode
);
1018 * Try to release metadata pages (indirect blocks, directories) which are
1019 * mapped via the block device. Since these pages could have journal heads
1020 * which would prevent try_to_free_buffers() from freeing them, we must use
1021 * jbd2 layer's try_to_free_buffers() function to release them.
1023 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
1026 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
1028 WARN_ON(PageChecked(page
));
1029 if (!page_has_buffers(page
))
1032 return jbd2_journal_try_to_free_buffers(journal
, page
,
1033 wait
& ~__GFP_WAIT
);
1034 return try_to_free_buffers(page
);
1038 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1039 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1041 static int ext4_write_dquot(struct dquot
*dquot
);
1042 static int ext4_acquire_dquot(struct dquot
*dquot
);
1043 static int ext4_release_dquot(struct dquot
*dquot
);
1044 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
1045 static int ext4_write_info(struct super_block
*sb
, int type
);
1046 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
1048 static int ext4_quota_off(struct super_block
*sb
, int type
);
1049 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
1050 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
1051 size_t len
, loff_t off
);
1052 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
1053 const char *data
, size_t len
, loff_t off
);
1054 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
1055 unsigned int flags
);
1056 static int ext4_enable_quotas(struct super_block
*sb
);
1058 static struct dquot
**ext4_get_dquots(struct inode
*inode
)
1060 return EXT4_I(inode
)->i_dquot
;
1063 static const struct dquot_operations ext4_quota_operations
= {
1064 .get_reserved_space
= ext4_get_reserved_space
,
1065 .write_dquot
= ext4_write_dquot
,
1066 .acquire_dquot
= ext4_acquire_dquot
,
1067 .release_dquot
= ext4_release_dquot
,
1068 .mark_dirty
= ext4_mark_dquot_dirty
,
1069 .write_info
= ext4_write_info
,
1070 .alloc_dquot
= dquot_alloc
,
1071 .destroy_dquot
= dquot_destroy
,
1074 static const struct quotactl_ops ext4_qctl_operations
= {
1075 .quota_on
= ext4_quota_on
,
1076 .quota_off
= ext4_quota_off
,
1077 .quota_sync
= dquot_quota_sync
,
1078 .get_info
= dquot_get_dqinfo
,
1079 .set_info
= dquot_set_dqinfo
,
1080 .get_dqblk
= dquot_get_dqblk
,
1081 .set_dqblk
= dquot_set_dqblk
1085 static const struct super_operations ext4_sops
= {
1086 .alloc_inode
= ext4_alloc_inode
,
1087 .destroy_inode
= ext4_destroy_inode
,
1088 .write_inode
= ext4_write_inode
,
1089 .dirty_inode
= ext4_dirty_inode
,
1090 .drop_inode
= ext4_drop_inode
,
1091 .evict_inode
= ext4_evict_inode
,
1092 .put_super
= ext4_put_super
,
1093 .sync_fs
= ext4_sync_fs
,
1094 .freeze_fs
= ext4_freeze
,
1095 .unfreeze_fs
= ext4_unfreeze
,
1096 .statfs
= ext4_statfs
,
1097 .remount_fs
= ext4_remount
,
1098 .show_options
= ext4_show_options
,
1100 .quota_read
= ext4_quota_read
,
1101 .quota_write
= ext4_quota_write
,
1102 .get_dquots
= ext4_get_dquots
,
1104 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1107 static const struct export_operations ext4_export_ops
= {
1108 .fh_to_dentry
= ext4_fh_to_dentry
,
1109 .fh_to_parent
= ext4_fh_to_parent
,
1110 .get_parent
= ext4_get_parent
,
1114 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1115 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1116 Opt_nouid32
, Opt_debug
, Opt_removed
,
1117 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1118 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
,
1119 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
, Opt_journal_dev
,
1120 Opt_journal_path
, Opt_journal_checksum
, Opt_journal_async_commit
,
1121 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1122 Opt_data_err_abort
, Opt_data_err_ignore
,
1123 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1124 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
1125 Opt_noquota
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
1126 Opt_usrquota
, Opt_grpquota
, Opt_i_version
, Opt_dax
,
1127 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
, Opt_mblk_io_submit
,
1128 Opt_lazytime
, Opt_nolazytime
,
1129 Opt_nomblk_io_submit
, Opt_block_validity
, Opt_noblock_validity
,
1130 Opt_inode_readahead_blks
, Opt_journal_ioprio
,
1131 Opt_dioread_nolock
, Opt_dioread_lock
,
1132 Opt_discard
, Opt_nodiscard
, Opt_init_itable
, Opt_noinit_itable
,
1133 Opt_max_dir_size_kb
, Opt_nojournal_checksum
,
1136 static const match_table_t tokens
= {
1137 {Opt_bsd_df
, "bsddf"},
1138 {Opt_minix_df
, "minixdf"},
1139 {Opt_grpid
, "grpid"},
1140 {Opt_grpid
, "bsdgroups"},
1141 {Opt_nogrpid
, "nogrpid"},
1142 {Opt_nogrpid
, "sysvgroups"},
1143 {Opt_resgid
, "resgid=%u"},
1144 {Opt_resuid
, "resuid=%u"},
1146 {Opt_err_cont
, "errors=continue"},
1147 {Opt_err_panic
, "errors=panic"},
1148 {Opt_err_ro
, "errors=remount-ro"},
1149 {Opt_nouid32
, "nouid32"},
1150 {Opt_debug
, "debug"},
1151 {Opt_removed
, "oldalloc"},
1152 {Opt_removed
, "orlov"},
1153 {Opt_user_xattr
, "user_xattr"},
1154 {Opt_nouser_xattr
, "nouser_xattr"},
1156 {Opt_noacl
, "noacl"},
1157 {Opt_noload
, "norecovery"},
1158 {Opt_noload
, "noload"},
1159 {Opt_removed
, "nobh"},
1160 {Opt_removed
, "bh"},
1161 {Opt_commit
, "commit=%u"},
1162 {Opt_min_batch_time
, "min_batch_time=%u"},
1163 {Opt_max_batch_time
, "max_batch_time=%u"},
1164 {Opt_journal_dev
, "journal_dev=%u"},
1165 {Opt_journal_path
, "journal_path=%s"},
1166 {Opt_journal_checksum
, "journal_checksum"},
1167 {Opt_nojournal_checksum
, "nojournal_checksum"},
1168 {Opt_journal_async_commit
, "journal_async_commit"},
1169 {Opt_abort
, "abort"},
1170 {Opt_data_journal
, "data=journal"},
1171 {Opt_data_ordered
, "data=ordered"},
1172 {Opt_data_writeback
, "data=writeback"},
1173 {Opt_data_err_abort
, "data_err=abort"},
1174 {Opt_data_err_ignore
, "data_err=ignore"},
1175 {Opt_offusrjquota
, "usrjquota="},
1176 {Opt_usrjquota
, "usrjquota=%s"},
1177 {Opt_offgrpjquota
, "grpjquota="},
1178 {Opt_grpjquota
, "grpjquota=%s"},
1179 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1180 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1181 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
1182 {Opt_grpquota
, "grpquota"},
1183 {Opt_noquota
, "noquota"},
1184 {Opt_quota
, "quota"},
1185 {Opt_usrquota
, "usrquota"},
1186 {Opt_barrier
, "barrier=%u"},
1187 {Opt_barrier
, "barrier"},
1188 {Opt_nobarrier
, "nobarrier"},
1189 {Opt_i_version
, "i_version"},
1191 {Opt_stripe
, "stripe=%u"},
1192 {Opt_delalloc
, "delalloc"},
1193 {Opt_lazytime
, "lazytime"},
1194 {Opt_nolazytime
, "nolazytime"},
1195 {Opt_nodelalloc
, "nodelalloc"},
1196 {Opt_removed
, "mblk_io_submit"},
1197 {Opt_removed
, "nomblk_io_submit"},
1198 {Opt_block_validity
, "block_validity"},
1199 {Opt_noblock_validity
, "noblock_validity"},
1200 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1201 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1202 {Opt_auto_da_alloc
, "auto_da_alloc=%u"},
1203 {Opt_auto_da_alloc
, "auto_da_alloc"},
1204 {Opt_noauto_da_alloc
, "noauto_da_alloc"},
1205 {Opt_dioread_nolock
, "dioread_nolock"},
1206 {Opt_dioread_lock
, "dioread_lock"},
1207 {Opt_discard
, "discard"},
1208 {Opt_nodiscard
, "nodiscard"},
1209 {Opt_init_itable
, "init_itable=%u"},
1210 {Opt_init_itable
, "init_itable"},
1211 {Opt_noinit_itable
, "noinit_itable"},
1212 {Opt_max_dir_size_kb
, "max_dir_size_kb=%u"},
1213 {Opt_removed
, "check=none"}, /* mount option from ext2/3 */
1214 {Opt_removed
, "nocheck"}, /* mount option from ext2/3 */
1215 {Opt_removed
, "reservation"}, /* mount option from ext2/3 */
1216 {Opt_removed
, "noreservation"}, /* mount option from ext2/3 */
1217 {Opt_removed
, "journal=%u"}, /* mount option from ext2/3 */
1221 static ext4_fsblk_t
get_sb_block(void **data
)
1223 ext4_fsblk_t sb_block
;
1224 char *options
= (char *) *data
;
1226 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1227 return 1; /* Default location */
1230 /* TODO: use simple_strtoll with >32bit ext4 */
1231 sb_block
= simple_strtoul(options
, &options
, 0);
1232 if (*options
&& *options
!= ',') {
1233 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1237 if (*options
== ',')
1239 *data
= (void *) options
;
1244 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1245 static char deprecated_msg
[] = "Mount option \"%s\" will be removed by %s\n"
1246 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1249 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
1251 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1255 if (sb_any_quota_loaded(sb
) &&
1256 !sbi
->s_qf_names
[qtype
]) {
1257 ext4_msg(sb
, KERN_ERR
,
1258 "Cannot change journaled "
1259 "quota options when quota turned on");
1262 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
)) {
1263 ext4_msg(sb
, KERN_ERR
, "Cannot set journaled quota options "
1264 "when QUOTA feature is enabled");
1267 qname
= match_strdup(args
);
1269 ext4_msg(sb
, KERN_ERR
,
1270 "Not enough memory for storing quotafile name");
1273 if (sbi
->s_qf_names
[qtype
]) {
1274 if (strcmp(sbi
->s_qf_names
[qtype
], qname
) == 0)
1277 ext4_msg(sb
, KERN_ERR
,
1278 "%s quota file already specified",
1282 if (strchr(qname
, '/')) {
1283 ext4_msg(sb
, KERN_ERR
,
1284 "quotafile must be on filesystem root");
1287 sbi
->s_qf_names
[qtype
] = qname
;
1295 static int clear_qf_name(struct super_block
*sb
, int qtype
)
1298 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1300 if (sb_any_quota_loaded(sb
) &&
1301 sbi
->s_qf_names
[qtype
]) {
1302 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
1303 " when quota turned on");
1306 kfree(sbi
->s_qf_names
[qtype
]);
1307 sbi
->s_qf_names
[qtype
] = NULL
;
1312 #define MOPT_SET 0x0001
1313 #define MOPT_CLEAR 0x0002
1314 #define MOPT_NOSUPPORT 0x0004
1315 #define MOPT_EXPLICIT 0x0008
1316 #define MOPT_CLEAR_ERR 0x0010
1317 #define MOPT_GTE0 0x0020
1320 #define MOPT_QFMT 0x0040
1322 #define MOPT_Q MOPT_NOSUPPORT
1323 #define MOPT_QFMT MOPT_NOSUPPORT
1325 #define MOPT_DATAJ 0x0080
1326 #define MOPT_NO_EXT2 0x0100
1327 #define MOPT_NO_EXT3 0x0200
1328 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1329 #define MOPT_STRING 0x0400
1331 static const struct mount_opts
{
1335 } ext4_mount_opts
[] = {
1336 {Opt_minix_df
, EXT4_MOUNT_MINIX_DF
, MOPT_SET
},
1337 {Opt_bsd_df
, EXT4_MOUNT_MINIX_DF
, MOPT_CLEAR
},
1338 {Opt_grpid
, EXT4_MOUNT_GRPID
, MOPT_SET
},
1339 {Opt_nogrpid
, EXT4_MOUNT_GRPID
, MOPT_CLEAR
},
1340 {Opt_block_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_SET
},
1341 {Opt_noblock_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_CLEAR
},
1342 {Opt_dioread_nolock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1343 MOPT_EXT4_ONLY
| MOPT_SET
},
1344 {Opt_dioread_lock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1345 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1346 {Opt_discard
, EXT4_MOUNT_DISCARD
, MOPT_SET
},
1347 {Opt_nodiscard
, EXT4_MOUNT_DISCARD
, MOPT_CLEAR
},
1348 {Opt_delalloc
, EXT4_MOUNT_DELALLOC
,
1349 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1350 {Opt_nodelalloc
, EXT4_MOUNT_DELALLOC
,
1351 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1352 {Opt_nojournal_checksum
, EXT4_MOUNT_JOURNAL_CHECKSUM
,
1353 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1354 {Opt_journal_checksum
, EXT4_MOUNT_JOURNAL_CHECKSUM
,
1355 MOPT_EXT4_ONLY
| MOPT_SET
},
1356 {Opt_journal_async_commit
, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT
|
1357 EXT4_MOUNT_JOURNAL_CHECKSUM
),
1358 MOPT_EXT4_ONLY
| MOPT_SET
},
1359 {Opt_noload
, EXT4_MOUNT_NOLOAD
, MOPT_NO_EXT2
| MOPT_SET
},
1360 {Opt_err_panic
, EXT4_MOUNT_ERRORS_PANIC
, MOPT_SET
| MOPT_CLEAR_ERR
},
1361 {Opt_err_ro
, EXT4_MOUNT_ERRORS_RO
, MOPT_SET
| MOPT_CLEAR_ERR
},
1362 {Opt_err_cont
, EXT4_MOUNT_ERRORS_CONT
, MOPT_SET
| MOPT_CLEAR_ERR
},
1363 {Opt_data_err_abort
, EXT4_MOUNT_DATA_ERR_ABORT
,
1364 MOPT_NO_EXT2
| MOPT_SET
},
1365 {Opt_data_err_ignore
, EXT4_MOUNT_DATA_ERR_ABORT
,
1366 MOPT_NO_EXT2
| MOPT_CLEAR
},
1367 {Opt_barrier
, EXT4_MOUNT_BARRIER
, MOPT_SET
},
1368 {Opt_nobarrier
, EXT4_MOUNT_BARRIER
, MOPT_CLEAR
},
1369 {Opt_noauto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_SET
},
1370 {Opt_auto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_CLEAR
},
1371 {Opt_noinit_itable
, EXT4_MOUNT_INIT_INODE_TABLE
, MOPT_CLEAR
},
1372 {Opt_commit
, 0, MOPT_GTE0
},
1373 {Opt_max_batch_time
, 0, MOPT_GTE0
},
1374 {Opt_min_batch_time
, 0, MOPT_GTE0
},
1375 {Opt_inode_readahead_blks
, 0, MOPT_GTE0
},
1376 {Opt_init_itable
, 0, MOPT_GTE0
},
1377 {Opt_dax
, EXT4_MOUNT_DAX
, MOPT_SET
},
1378 {Opt_stripe
, 0, MOPT_GTE0
},
1379 {Opt_resuid
, 0, MOPT_GTE0
},
1380 {Opt_resgid
, 0, MOPT_GTE0
},
1381 {Opt_journal_dev
, 0, MOPT_GTE0
},
1382 {Opt_journal_path
, 0, MOPT_STRING
},
1383 {Opt_journal_ioprio
, 0, MOPT_GTE0
},
1384 {Opt_data_journal
, EXT4_MOUNT_JOURNAL_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1385 {Opt_data_ordered
, EXT4_MOUNT_ORDERED_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1386 {Opt_data_writeback
, EXT4_MOUNT_WRITEBACK_DATA
,
1387 MOPT_NO_EXT2
| MOPT_DATAJ
},
1388 {Opt_user_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_SET
},
1389 {Opt_nouser_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_CLEAR
},
1390 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1391 {Opt_acl
, EXT4_MOUNT_POSIX_ACL
, MOPT_SET
},
1392 {Opt_noacl
, EXT4_MOUNT_POSIX_ACL
, MOPT_CLEAR
},
1394 {Opt_acl
, 0, MOPT_NOSUPPORT
},
1395 {Opt_noacl
, 0, MOPT_NOSUPPORT
},
1397 {Opt_nouid32
, EXT4_MOUNT_NO_UID32
, MOPT_SET
},
1398 {Opt_debug
, EXT4_MOUNT_DEBUG
, MOPT_SET
},
1399 {Opt_quota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
, MOPT_SET
| MOPT_Q
},
1400 {Opt_usrquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
,
1402 {Opt_grpquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_GRPQUOTA
,
1404 {Opt_noquota
, (EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
|
1405 EXT4_MOUNT_GRPQUOTA
), MOPT_CLEAR
| MOPT_Q
},
1406 {Opt_usrjquota
, 0, MOPT_Q
},
1407 {Opt_grpjquota
, 0, MOPT_Q
},
1408 {Opt_offusrjquota
, 0, MOPT_Q
},
1409 {Opt_offgrpjquota
, 0, MOPT_Q
},
1410 {Opt_jqfmt_vfsold
, QFMT_VFS_OLD
, MOPT_QFMT
},
1411 {Opt_jqfmt_vfsv0
, QFMT_VFS_V0
, MOPT_QFMT
},
1412 {Opt_jqfmt_vfsv1
, QFMT_VFS_V1
, MOPT_QFMT
},
1413 {Opt_max_dir_size_kb
, 0, MOPT_GTE0
},
1417 static int handle_mount_opt(struct super_block
*sb
, char *opt
, int token
,
1418 substring_t
*args
, unsigned long *journal_devnum
,
1419 unsigned int *journal_ioprio
, int is_remount
)
1421 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1422 const struct mount_opts
*m
;
1428 if (token
== Opt_usrjquota
)
1429 return set_qf_name(sb
, USRQUOTA
, &args
[0]);
1430 else if (token
== Opt_grpjquota
)
1431 return set_qf_name(sb
, GRPQUOTA
, &args
[0]);
1432 else if (token
== Opt_offusrjquota
)
1433 return clear_qf_name(sb
, USRQUOTA
);
1434 else if (token
== Opt_offgrpjquota
)
1435 return clear_qf_name(sb
, GRPQUOTA
);
1439 case Opt_nouser_xattr
:
1440 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, opt
, "3.5");
1443 return 1; /* handled by get_sb_block() */
1445 ext4_msg(sb
, KERN_WARNING
, "Ignoring removed %s option", opt
);
1448 sbi
->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
1451 sb
->s_flags
|= MS_I_VERSION
;
1454 sb
->s_flags
|= MS_LAZYTIME
;
1456 case Opt_nolazytime
:
1457 sb
->s_flags
&= ~MS_LAZYTIME
;
1461 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++)
1462 if (token
== m
->token
)
1465 if (m
->token
== Opt_err
) {
1466 ext4_msg(sb
, KERN_ERR
, "Unrecognized mount option \"%s\" "
1467 "or missing value", opt
);
1471 if ((m
->flags
& MOPT_NO_EXT2
) && IS_EXT2_SB(sb
)) {
1472 ext4_msg(sb
, KERN_ERR
,
1473 "Mount option \"%s\" incompatible with ext2", opt
);
1476 if ((m
->flags
& MOPT_NO_EXT3
) && IS_EXT3_SB(sb
)) {
1477 ext4_msg(sb
, KERN_ERR
,
1478 "Mount option \"%s\" incompatible with ext3", opt
);
1482 if (args
->from
&& !(m
->flags
& MOPT_STRING
) && match_int(args
, &arg
))
1484 if (args
->from
&& (m
->flags
& MOPT_GTE0
) && (arg
< 0))
1486 if (m
->flags
& MOPT_EXPLICIT
)
1487 set_opt2(sb
, EXPLICIT_DELALLOC
);
1488 if (m
->flags
& MOPT_CLEAR_ERR
)
1489 clear_opt(sb
, ERRORS_MASK
);
1490 if (token
== Opt_noquota
&& sb_any_quota_loaded(sb
)) {
1491 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
1492 "options when quota turned on");
1496 if (m
->flags
& MOPT_NOSUPPORT
) {
1497 ext4_msg(sb
, KERN_ERR
, "%s option not supported", opt
);
1498 } else if (token
== Opt_commit
) {
1500 arg
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1501 sbi
->s_commit_interval
= HZ
* arg
;
1502 } else if (token
== Opt_max_batch_time
) {
1503 sbi
->s_max_batch_time
= arg
;
1504 } else if (token
== Opt_min_batch_time
) {
1505 sbi
->s_min_batch_time
= arg
;
1506 } else if (token
== Opt_inode_readahead_blks
) {
1507 if (arg
&& (arg
> (1 << 30) || !is_power_of_2(arg
))) {
1508 ext4_msg(sb
, KERN_ERR
,
1509 "EXT4-fs: inode_readahead_blks must be "
1510 "0 or a power of 2 smaller than 2^31");
1513 sbi
->s_inode_readahead_blks
= arg
;
1514 } else if (token
== Opt_init_itable
) {
1515 set_opt(sb
, INIT_INODE_TABLE
);
1517 arg
= EXT4_DEF_LI_WAIT_MULT
;
1518 sbi
->s_li_wait_mult
= arg
;
1519 } else if (token
== Opt_max_dir_size_kb
) {
1520 sbi
->s_max_dir_size_kb
= arg
;
1521 } else if (token
== Opt_stripe
) {
1522 sbi
->s_stripe
= arg
;
1523 } else if (token
== Opt_resuid
) {
1524 uid
= make_kuid(current_user_ns(), arg
);
1525 if (!uid_valid(uid
)) {
1526 ext4_msg(sb
, KERN_ERR
, "Invalid uid value %d", arg
);
1529 sbi
->s_resuid
= uid
;
1530 } else if (token
== Opt_resgid
) {
1531 gid
= make_kgid(current_user_ns(), arg
);
1532 if (!gid_valid(gid
)) {
1533 ext4_msg(sb
, KERN_ERR
, "Invalid gid value %d", arg
);
1536 sbi
->s_resgid
= gid
;
1537 } else if (token
== Opt_journal_dev
) {
1539 ext4_msg(sb
, KERN_ERR
,
1540 "Cannot specify journal on remount");
1543 *journal_devnum
= arg
;
1544 } else if (token
== Opt_journal_path
) {
1546 struct inode
*journal_inode
;
1551 ext4_msg(sb
, KERN_ERR
,
1552 "Cannot specify journal on remount");
1555 journal_path
= match_strdup(&args
[0]);
1556 if (!journal_path
) {
1557 ext4_msg(sb
, KERN_ERR
, "error: could not dup "
1558 "journal device string");
1562 error
= kern_path(journal_path
, LOOKUP_FOLLOW
, &path
);
1564 ext4_msg(sb
, KERN_ERR
, "error: could not find "
1565 "journal device path: error %d", error
);
1566 kfree(journal_path
);
1570 journal_inode
= path
.dentry
->d_inode
;
1571 if (!S_ISBLK(journal_inode
->i_mode
)) {
1572 ext4_msg(sb
, KERN_ERR
, "error: journal path %s "
1573 "is not a block device", journal_path
);
1575 kfree(journal_path
);
1579 *journal_devnum
= new_encode_dev(journal_inode
->i_rdev
);
1581 kfree(journal_path
);
1582 } else if (token
== Opt_journal_ioprio
) {
1584 ext4_msg(sb
, KERN_ERR
, "Invalid journal IO priority"
1589 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
, arg
);
1590 } else if (m
->flags
& MOPT_DATAJ
) {
1592 if (!sbi
->s_journal
)
1593 ext4_msg(sb
, KERN_WARNING
, "Remounting file system with no journal so ignoring journalled data option");
1594 else if (test_opt(sb
, DATA_FLAGS
) != m
->mount_opt
) {
1595 ext4_msg(sb
, KERN_ERR
,
1596 "Cannot change data mode on remount");
1600 clear_opt(sb
, DATA_FLAGS
);
1601 sbi
->s_mount_opt
|= m
->mount_opt
;
1604 } else if (m
->flags
& MOPT_QFMT
) {
1605 if (sb_any_quota_loaded(sb
) &&
1606 sbi
->s_jquota_fmt
!= m
->mount_opt
) {
1607 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled "
1608 "quota options when quota turned on");
1611 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
1612 EXT4_FEATURE_RO_COMPAT_QUOTA
)) {
1613 ext4_msg(sb
, KERN_ERR
,
1614 "Cannot set journaled quota options "
1615 "when QUOTA feature is enabled");
1618 sbi
->s_jquota_fmt
= m
->mount_opt
;
1620 #ifndef CONFIG_FS_DAX
1621 } else if (token
== Opt_dax
) {
1622 ext4_msg(sb
, KERN_INFO
, "dax option not supported");
1628 if (m
->flags
& MOPT_CLEAR
)
1630 else if (unlikely(!(m
->flags
& MOPT_SET
))) {
1631 ext4_msg(sb
, KERN_WARNING
,
1632 "buggy handling of option %s", opt
);
1637 sbi
->s_mount_opt
|= m
->mount_opt
;
1639 sbi
->s_mount_opt
&= ~m
->mount_opt
;
1644 static int parse_options(char *options
, struct super_block
*sb
,
1645 unsigned long *journal_devnum
,
1646 unsigned int *journal_ioprio
,
1649 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1651 substring_t args
[MAX_OPT_ARGS
];
1657 while ((p
= strsep(&options
, ",")) != NULL
) {
1661 * Initialize args struct so we know whether arg was
1662 * found; some options take optional arguments.
1664 args
[0].to
= args
[0].from
= NULL
;
1665 token
= match_token(p
, tokens
, args
);
1666 if (handle_mount_opt(sb
, p
, token
, args
, journal_devnum
,
1667 journal_ioprio
, is_remount
) < 0)
1671 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
) &&
1672 (test_opt(sb
, USRQUOTA
) || test_opt(sb
, GRPQUOTA
))) {
1673 ext4_msg(sb
, KERN_ERR
, "Cannot set quota options when QUOTA "
1674 "feature is enabled");
1677 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1678 if (test_opt(sb
, USRQUOTA
) && sbi
->s_qf_names
[USRQUOTA
])
1679 clear_opt(sb
, USRQUOTA
);
1681 if (test_opt(sb
, GRPQUOTA
) && sbi
->s_qf_names
[GRPQUOTA
])
1682 clear_opt(sb
, GRPQUOTA
);
1684 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
1685 ext4_msg(sb
, KERN_ERR
, "old and new quota "
1690 if (!sbi
->s_jquota_fmt
) {
1691 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1697 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
1699 BLOCK_SIZE
<< le32_to_cpu(sbi
->s_es
->s_log_block_size
);
1701 if (blocksize
< PAGE_CACHE_SIZE
) {
1702 ext4_msg(sb
, KERN_ERR
, "can't mount with "
1703 "dioread_nolock if block size != PAGE_SIZE");
1707 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
&&
1708 test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
1709 ext4_msg(sb
, KERN_ERR
, "can't mount with journal_async_commit "
1710 "in data=ordered mode");
1716 static inline void ext4_show_quota_options(struct seq_file
*seq
,
1717 struct super_block
*sb
)
1719 #if defined(CONFIG_QUOTA)
1720 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1722 if (sbi
->s_jquota_fmt
) {
1725 switch (sbi
->s_jquota_fmt
) {
1736 seq_printf(seq
, ",jqfmt=%s", fmtname
);
1739 if (sbi
->s_qf_names
[USRQUOTA
])
1740 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
1742 if (sbi
->s_qf_names
[GRPQUOTA
])
1743 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
1747 static const char *token2str(int token
)
1749 const struct match_token
*t
;
1751 for (t
= tokens
; t
->token
!= Opt_err
; t
++)
1752 if (t
->token
== token
&& !strchr(t
->pattern
, '='))
1759 * - it's set to a non-default value OR
1760 * - if the per-sb default is different from the global default
1762 static int _ext4_show_options(struct seq_file
*seq
, struct super_block
*sb
,
1765 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1766 struct ext4_super_block
*es
= sbi
->s_es
;
1767 int def_errors
, def_mount_opt
= nodefs
? 0 : sbi
->s_def_mount_opt
;
1768 const struct mount_opts
*m
;
1769 char sep
= nodefs
? '\n' : ',';
1771 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1772 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1774 if (sbi
->s_sb_block
!= 1)
1775 SEQ_OPTS_PRINT("sb=%llu", sbi
->s_sb_block
);
1777 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++) {
1778 int want_set
= m
->flags
& MOPT_SET
;
1779 if (((m
->flags
& (MOPT_SET
|MOPT_CLEAR
)) == 0) ||
1780 (m
->flags
& MOPT_CLEAR_ERR
))
1782 if (!(m
->mount_opt
& (sbi
->s_mount_opt
^ def_mount_opt
)))
1783 continue; /* skip if same as the default */
1785 (sbi
->s_mount_opt
& m
->mount_opt
) != m
->mount_opt
) ||
1786 (!want_set
&& (sbi
->s_mount_opt
& m
->mount_opt
)))
1787 continue; /* select Opt_noFoo vs Opt_Foo */
1788 SEQ_OPTS_PRINT("%s", token2str(m
->token
));
1791 if (nodefs
|| !uid_eq(sbi
->s_resuid
, make_kuid(&init_user_ns
, EXT4_DEF_RESUID
)) ||
1792 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
)
1793 SEQ_OPTS_PRINT("resuid=%u",
1794 from_kuid_munged(&init_user_ns
, sbi
->s_resuid
));
1795 if (nodefs
|| !gid_eq(sbi
->s_resgid
, make_kgid(&init_user_ns
, EXT4_DEF_RESGID
)) ||
1796 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
)
1797 SEQ_OPTS_PRINT("resgid=%u",
1798 from_kgid_munged(&init_user_ns
, sbi
->s_resgid
));
1799 def_errors
= nodefs
? -1 : le16_to_cpu(es
->s_errors
);
1800 if (test_opt(sb
, ERRORS_RO
) && def_errors
!= EXT4_ERRORS_RO
)
1801 SEQ_OPTS_PUTS("errors=remount-ro");
1802 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
1803 SEQ_OPTS_PUTS("errors=continue");
1804 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
1805 SEQ_OPTS_PUTS("errors=panic");
1806 if (nodefs
|| sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
)
1807 SEQ_OPTS_PRINT("commit=%lu", sbi
->s_commit_interval
/ HZ
);
1808 if (nodefs
|| sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
)
1809 SEQ_OPTS_PRINT("min_batch_time=%u", sbi
->s_min_batch_time
);
1810 if (nodefs
|| sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
)
1811 SEQ_OPTS_PRINT("max_batch_time=%u", sbi
->s_max_batch_time
);
1812 if (sb
->s_flags
& MS_I_VERSION
)
1813 SEQ_OPTS_PUTS("i_version");
1814 if (nodefs
|| sbi
->s_stripe
)
1815 SEQ_OPTS_PRINT("stripe=%lu", sbi
->s_stripe
);
1816 if (EXT4_MOUNT_DATA_FLAGS
& (sbi
->s_mount_opt
^ def_mount_opt
)) {
1817 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
1818 SEQ_OPTS_PUTS("data=journal");
1819 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
1820 SEQ_OPTS_PUTS("data=ordered");
1821 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
1822 SEQ_OPTS_PUTS("data=writeback");
1825 sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
1826 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1827 sbi
->s_inode_readahead_blks
);
1829 if (nodefs
|| (test_opt(sb
, INIT_INODE_TABLE
) &&
1830 (sbi
->s_li_wait_mult
!= EXT4_DEF_LI_WAIT_MULT
)))
1831 SEQ_OPTS_PRINT("init_itable=%u", sbi
->s_li_wait_mult
);
1832 if (nodefs
|| sbi
->s_max_dir_size_kb
)
1833 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi
->s_max_dir_size_kb
);
1835 ext4_show_quota_options(seq
, sb
);
1839 static int ext4_show_options(struct seq_file
*seq
, struct dentry
*root
)
1841 return _ext4_show_options(seq
, root
->d_sb
, 0);
1844 static int options_seq_show(struct seq_file
*seq
, void *offset
)
1846 struct super_block
*sb
= seq
->private;
1849 seq_puts(seq
, (sb
->s_flags
& MS_RDONLY
) ? "ro" : "rw");
1850 rc
= _ext4_show_options(seq
, sb
, 1);
1851 seq_puts(seq
, "\n");
1855 static int options_open_fs(struct inode
*inode
, struct file
*file
)
1857 return single_open(file
, options_seq_show
, PDE_DATA(inode
));
1860 static const struct file_operations ext4_seq_options_fops
= {
1861 .owner
= THIS_MODULE
,
1862 .open
= options_open_fs
,
1864 .llseek
= seq_lseek
,
1865 .release
= single_release
,
1868 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
1871 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1874 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
1875 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
1876 "forcing read-only mode");
1881 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
1882 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
1883 "running e2fsck is recommended");
1884 else if (sbi
->s_mount_state
& EXT4_ERROR_FS
)
1885 ext4_msg(sb
, KERN_WARNING
,
1886 "warning: mounting fs with errors, "
1887 "running e2fsck is recommended");
1888 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) > 0 &&
1889 le16_to_cpu(es
->s_mnt_count
) >=
1890 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1891 ext4_msg(sb
, KERN_WARNING
,
1892 "warning: maximal mount count reached, "
1893 "running e2fsck is recommended");
1894 else if (le32_to_cpu(es
->s_checkinterval
) &&
1895 (le32_to_cpu(es
->s_lastcheck
) +
1896 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1897 ext4_msg(sb
, KERN_WARNING
,
1898 "warning: checktime reached, "
1899 "running e2fsck is recommended");
1900 if (!sbi
->s_journal
)
1901 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
1902 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1903 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
1904 le16_add_cpu(&es
->s_mnt_count
, 1);
1905 es
->s_mtime
= cpu_to_le32(get_seconds());
1906 ext4_update_dynamic_rev(sb
);
1908 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
1910 ext4_commit_super(sb
, 1);
1912 if (test_opt(sb
, DEBUG
))
1913 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
1914 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1916 sbi
->s_groups_count
,
1917 EXT4_BLOCKS_PER_GROUP(sb
),
1918 EXT4_INODES_PER_GROUP(sb
),
1919 sbi
->s_mount_opt
, sbi
->s_mount_opt2
);
1921 cleancache_init_fs(sb
);
1925 int ext4_alloc_flex_bg_array(struct super_block
*sb
, ext4_group_t ngroup
)
1927 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1928 struct flex_groups
*new_groups
;
1931 if (!sbi
->s_log_groups_per_flex
)
1934 size
= ext4_flex_group(sbi
, ngroup
- 1) + 1;
1935 if (size
<= sbi
->s_flex_groups_allocated
)
1938 size
= roundup_pow_of_two(size
* sizeof(struct flex_groups
));
1939 new_groups
= ext4_kvzalloc(size
, GFP_KERNEL
);
1941 ext4_msg(sb
, KERN_ERR
, "not enough memory for %d flex groups",
1942 size
/ (int) sizeof(struct flex_groups
));
1946 if (sbi
->s_flex_groups
) {
1947 memcpy(new_groups
, sbi
->s_flex_groups
,
1948 (sbi
->s_flex_groups_allocated
*
1949 sizeof(struct flex_groups
)));
1950 kvfree(sbi
->s_flex_groups
);
1952 sbi
->s_flex_groups
= new_groups
;
1953 sbi
->s_flex_groups_allocated
= size
/ sizeof(struct flex_groups
);
1957 static int ext4_fill_flex_info(struct super_block
*sb
)
1959 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1960 struct ext4_group_desc
*gdp
= NULL
;
1961 ext4_group_t flex_group
;
1964 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
1965 if (sbi
->s_log_groups_per_flex
< 1 || sbi
->s_log_groups_per_flex
> 31) {
1966 sbi
->s_log_groups_per_flex
= 0;
1970 err
= ext4_alloc_flex_bg_array(sb
, sbi
->s_groups_count
);
1974 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1975 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1977 flex_group
= ext4_flex_group(sbi
, i
);
1978 atomic_add(ext4_free_inodes_count(sb
, gdp
),
1979 &sbi
->s_flex_groups
[flex_group
].free_inodes
);
1980 atomic64_add(ext4_free_group_clusters(sb
, gdp
),
1981 &sbi
->s_flex_groups
[flex_group
].free_clusters
);
1982 atomic_add(ext4_used_dirs_count(sb
, gdp
),
1983 &sbi
->s_flex_groups
[flex_group
].used_dirs
);
1991 static __le16
ext4_group_desc_csum(struct ext4_sb_info
*sbi
, __u32 block_group
,
1992 struct ext4_group_desc
*gdp
)
1996 __le32 le_group
= cpu_to_le32(block_group
);
1998 if (ext4_has_metadata_csum(sbi
->s_sb
)) {
1999 /* Use new metadata_csum algorithm */
2003 save_csum
= gdp
->bg_checksum
;
2004 gdp
->bg_checksum
= 0;
2005 csum32
= ext4_chksum(sbi
, sbi
->s_csum_seed
, (__u8
*)&le_group
,
2007 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)gdp
,
2009 gdp
->bg_checksum
= save_csum
;
2011 crc
= csum32
& 0xFFFF;
2015 /* old crc16 code */
2016 if (!(sbi
->s_es
->s_feature_ro_compat
&
2017 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)))
2020 offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
2022 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
2023 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
2024 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
2025 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
2026 /* for checksum of struct ext4_group_desc do the rest...*/
2027 if ((sbi
->s_es
->s_feature_incompat
&
2028 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT
)) &&
2029 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
2030 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
2031 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
2035 return cpu_to_le16(crc
);
2038 int ext4_group_desc_csum_verify(struct super_block
*sb
, __u32 block_group
,
2039 struct ext4_group_desc
*gdp
)
2041 if (ext4_has_group_desc_csum(sb
) &&
2042 (gdp
->bg_checksum
!= ext4_group_desc_csum(EXT4_SB(sb
),
2049 void ext4_group_desc_csum_set(struct super_block
*sb
, __u32 block_group
,
2050 struct ext4_group_desc
*gdp
)
2052 if (!ext4_has_group_desc_csum(sb
))
2054 gdp
->bg_checksum
= ext4_group_desc_csum(EXT4_SB(sb
), block_group
, gdp
);
2057 /* Called at mount-time, super-block is locked */
2058 static int ext4_check_descriptors(struct super_block
*sb
,
2059 ext4_group_t
*first_not_zeroed
)
2061 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2062 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
2063 ext4_fsblk_t last_block
;
2064 ext4_fsblk_t block_bitmap
;
2065 ext4_fsblk_t inode_bitmap
;
2066 ext4_fsblk_t inode_table
;
2067 int flexbg_flag
= 0;
2068 ext4_group_t i
, grp
= sbi
->s_groups_count
;
2070 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
2073 ext4_debug("Checking group descriptors");
2075 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2076 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2078 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
2079 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
2081 last_block
= first_block
+
2082 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2084 if ((grp
== sbi
->s_groups_count
) &&
2085 !(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2088 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
2089 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
2090 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2091 "Block bitmap for group %u not in group "
2092 "(block %llu)!", i
, block_bitmap
);
2095 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
2096 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
2097 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2098 "Inode bitmap for group %u not in group "
2099 "(block %llu)!", i
, inode_bitmap
);
2102 inode_table
= ext4_inode_table(sb
, gdp
);
2103 if (inode_table
< first_block
||
2104 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
2105 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2106 "Inode table for group %u not in group "
2107 "(block %llu)!", i
, inode_table
);
2110 ext4_lock_group(sb
, i
);
2111 if (!ext4_group_desc_csum_verify(sb
, i
, gdp
)) {
2112 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2113 "Checksum for group %u failed (%u!=%u)",
2114 i
, le16_to_cpu(ext4_group_desc_csum(sbi
, i
,
2115 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
2116 if (!(sb
->s_flags
& MS_RDONLY
)) {
2117 ext4_unlock_group(sb
, i
);
2121 ext4_unlock_group(sb
, i
);
2123 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
2125 if (NULL
!= first_not_zeroed
)
2126 *first_not_zeroed
= grp
;
2130 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2131 * the superblock) which were deleted from all directories, but held open by
2132 * a process at the time of a crash. We walk the list and try to delete these
2133 * inodes at recovery time (only with a read-write filesystem).
2135 * In order to keep the orphan inode chain consistent during traversal (in
2136 * case of crash during recovery), we link each inode into the superblock
2137 * orphan list_head and handle it the same way as an inode deletion during
2138 * normal operation (which journals the operations for us).
2140 * We only do an iget() and an iput() on each inode, which is very safe if we
2141 * accidentally point at an in-use or already deleted inode. The worst that
2142 * can happen in this case is that we get a "bit already cleared" message from
2143 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2144 * e2fsck was run on this filesystem, and it must have already done the orphan
2145 * inode cleanup for us, so we can safely abort without any further action.
2147 static void ext4_orphan_cleanup(struct super_block
*sb
,
2148 struct ext4_super_block
*es
)
2150 unsigned int s_flags
= sb
->s_flags
;
2151 int nr_orphans
= 0, nr_truncates
= 0;
2155 if (!es
->s_last_orphan
) {
2156 jbd_debug(4, "no orphan inodes to clean up\n");
2160 if (bdev_read_only(sb
->s_bdev
)) {
2161 ext4_msg(sb
, KERN_ERR
, "write access "
2162 "unavailable, skipping orphan cleanup");
2166 /* Check if feature set would not allow a r/w mount */
2167 if (!ext4_feature_set_ok(sb
, 0)) {
2168 ext4_msg(sb
, KERN_INFO
, "Skipping orphan cleanup due to "
2169 "unknown ROCOMPAT features");
2173 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2174 /* don't clear list on RO mount w/ errors */
2175 if (es
->s_last_orphan
&& !(s_flags
& MS_RDONLY
)) {
2176 ext4_msg(sb
, KERN_INFO
, "Errors on filesystem, "
2177 "clearing orphan list.\n");
2178 es
->s_last_orphan
= 0;
2180 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2184 if (s_flags
& MS_RDONLY
) {
2185 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
2186 sb
->s_flags
&= ~MS_RDONLY
;
2189 /* Needed for iput() to work correctly and not trash data */
2190 sb
->s_flags
|= MS_ACTIVE
;
2191 /* Turn on quotas so that they are updated correctly */
2192 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
2193 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
2194 int ret
= ext4_quota_on_mount(sb
, i
);
2196 ext4_msg(sb
, KERN_ERR
,
2197 "Cannot turn on journaled "
2198 "quota: error %d", ret
);
2203 while (es
->s_last_orphan
) {
2204 struct inode
*inode
;
2206 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
2207 if (IS_ERR(inode
)) {
2208 es
->s_last_orphan
= 0;
2212 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
2213 dquot_initialize(inode
);
2214 if (inode
->i_nlink
) {
2215 if (test_opt(sb
, DEBUG
))
2216 ext4_msg(sb
, KERN_DEBUG
,
2217 "%s: truncating inode %lu to %lld bytes",
2218 __func__
, inode
->i_ino
, inode
->i_size
);
2219 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2220 inode
->i_ino
, inode
->i_size
);
2221 mutex_lock(&inode
->i_mutex
);
2222 truncate_inode_pages(inode
->i_mapping
, inode
->i_size
);
2223 ext4_truncate(inode
);
2224 mutex_unlock(&inode
->i_mutex
);
2227 if (test_opt(sb
, DEBUG
))
2228 ext4_msg(sb
, KERN_DEBUG
,
2229 "%s: deleting unreferenced inode %lu",
2230 __func__
, inode
->i_ino
);
2231 jbd_debug(2, "deleting unreferenced inode %lu\n",
2235 iput(inode
); /* The delete magic happens here! */
2238 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2241 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
2242 PLURAL(nr_orphans
));
2244 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
2245 PLURAL(nr_truncates
));
2247 /* Turn quotas off */
2248 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
2249 if (sb_dqopt(sb
)->files
[i
])
2250 dquot_quota_off(sb
, i
);
2253 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
2257 * Maximal extent format file size.
2258 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2259 * extent format containers, within a sector_t, and within i_blocks
2260 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2261 * so that won't be a limiting factor.
2263 * However there is other limiting factor. We do store extents in the form
2264 * of starting block and length, hence the resulting length of the extent
2265 * covering maximum file size must fit into on-disk format containers as
2266 * well. Given that length is always by 1 unit bigger than max unit (because
2267 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2269 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2271 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
2274 loff_t upper_limit
= MAX_LFS_FILESIZE
;
2276 /* small i_blocks in vfs inode? */
2277 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2279 * CONFIG_LBDAF is not enabled implies the inode
2280 * i_block represent total blocks in 512 bytes
2281 * 32 == size of vfs inode i_blocks * 8
2283 upper_limit
= (1LL << 32) - 1;
2285 /* total blocks in file system block size */
2286 upper_limit
>>= (blkbits
- 9);
2287 upper_limit
<<= blkbits
;
2291 * 32-bit extent-start container, ee_block. We lower the maxbytes
2292 * by one fs block, so ee_len can cover the extent of maximum file
2295 res
= (1LL << 32) - 1;
2298 /* Sanity check against vm- & vfs- imposed limits */
2299 if (res
> upper_limit
)
2306 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2307 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2308 * We need to be 1 filesystem block less than the 2^48 sector limit.
2310 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
2312 loff_t res
= EXT4_NDIR_BLOCKS
;
2315 /* This is calculated to be the largest file size for a dense, block
2316 * mapped file such that the file's total number of 512-byte sectors,
2317 * including data and all indirect blocks, does not exceed (2^48 - 1).
2319 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2320 * number of 512-byte sectors of the file.
2323 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2325 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2326 * the inode i_block field represents total file blocks in
2327 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2329 upper_limit
= (1LL << 32) - 1;
2331 /* total blocks in file system block size */
2332 upper_limit
>>= (bits
- 9);
2336 * We use 48 bit ext4_inode i_blocks
2337 * With EXT4_HUGE_FILE_FL set the i_blocks
2338 * represent total number of blocks in
2339 * file system block size
2341 upper_limit
= (1LL << 48) - 1;
2345 /* indirect blocks */
2347 /* double indirect blocks */
2348 meta_blocks
+= 1 + (1LL << (bits
-2));
2349 /* tripple indirect blocks */
2350 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
2352 upper_limit
-= meta_blocks
;
2353 upper_limit
<<= bits
;
2355 res
+= 1LL << (bits
-2);
2356 res
+= 1LL << (2*(bits
-2));
2357 res
+= 1LL << (3*(bits
-2));
2359 if (res
> upper_limit
)
2362 if (res
> MAX_LFS_FILESIZE
)
2363 res
= MAX_LFS_FILESIZE
;
2368 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
2369 ext4_fsblk_t logical_sb_block
, int nr
)
2371 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2372 ext4_group_t bg
, first_meta_bg
;
2375 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
2377 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_META_BG
) ||
2379 return logical_sb_block
+ nr
+ 1;
2380 bg
= sbi
->s_desc_per_block
* nr
;
2381 if (ext4_bg_has_super(sb
, bg
))
2385 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2386 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2387 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2390 if (sb
->s_blocksize
== 1024 && nr
== 0 &&
2391 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_first_data_block
) == 0)
2394 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
2398 * ext4_get_stripe_size: Get the stripe size.
2399 * @sbi: In memory super block info
2401 * If we have specified it via mount option, then
2402 * use the mount option value. If the value specified at mount time is
2403 * greater than the blocks per group use the super block value.
2404 * If the super block value is greater than blocks per group return 0.
2405 * Allocator needs it be less than blocks per group.
2408 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
2410 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
2411 unsigned long stripe_width
=
2412 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
2415 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
2416 ret
= sbi
->s_stripe
;
2417 else if (stripe_width
<= sbi
->s_blocks_per_group
)
2419 else if (stride
<= sbi
->s_blocks_per_group
)
2425 * If the stripe width is 1, this makes no sense and
2426 * we set it to 0 to turn off stripe handling code.
2437 struct attribute attr
;
2438 ssize_t (*show
)(struct ext4_attr
*, struct ext4_sb_info
*, char *);
2439 ssize_t (*store
)(struct ext4_attr
*, struct ext4_sb_info
*,
2440 const char *, size_t);
2447 static int parse_strtoull(const char *buf
,
2448 unsigned long long max
, unsigned long long *value
)
2452 ret
= kstrtoull(skip_spaces(buf
), 0, value
);
2453 if (!ret
&& *value
> max
)
2458 static ssize_t
delayed_allocation_blocks_show(struct ext4_attr
*a
,
2459 struct ext4_sb_info
*sbi
,
2462 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2464 percpu_counter_sum(&sbi
->s_dirtyclusters_counter
)));
2467 static ssize_t
session_write_kbytes_show(struct ext4_attr
*a
,
2468 struct ext4_sb_info
*sbi
, char *buf
)
2470 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2472 if (!sb
->s_bdev
->bd_part
)
2473 return snprintf(buf
, PAGE_SIZE
, "0\n");
2474 return snprintf(buf
, PAGE_SIZE
, "%lu\n",
2475 (part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2476 sbi
->s_sectors_written_start
) >> 1);
2479 static ssize_t
lifetime_write_kbytes_show(struct ext4_attr
*a
,
2480 struct ext4_sb_info
*sbi
, char *buf
)
2482 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2484 if (!sb
->s_bdev
->bd_part
)
2485 return snprintf(buf
, PAGE_SIZE
, "0\n");
2486 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2487 (unsigned long long)(sbi
->s_kbytes_written
+
2488 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2489 EXT4_SB(sb
)->s_sectors_written_start
) >> 1)));
2492 static ssize_t
inode_readahead_blks_store(struct ext4_attr
*a
,
2493 struct ext4_sb_info
*sbi
,
2494 const char *buf
, size_t count
)
2499 ret
= kstrtoul(skip_spaces(buf
), 0, &t
);
2503 if (t
&& (!is_power_of_2(t
) || t
> 0x40000000))
2506 sbi
->s_inode_readahead_blks
= t
;
2510 static ssize_t
sbi_ui_show(struct ext4_attr
*a
,
2511 struct ext4_sb_info
*sbi
, char *buf
)
2513 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->u
.offset
);
2515 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
2518 static ssize_t
sbi_ui_store(struct ext4_attr
*a
,
2519 struct ext4_sb_info
*sbi
,
2520 const char *buf
, size_t count
)
2522 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->u
.offset
);
2526 ret
= kstrtoul(skip_spaces(buf
), 0, &t
);
2533 static ssize_t
es_ui_show(struct ext4_attr
*a
,
2534 struct ext4_sb_info
*sbi
, char *buf
)
2537 unsigned int *ui
= (unsigned int *) (((char *) sbi
->s_es
) +
2540 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
2543 static ssize_t
reserved_clusters_show(struct ext4_attr
*a
,
2544 struct ext4_sb_info
*sbi
, char *buf
)
2546 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2547 (unsigned long long) atomic64_read(&sbi
->s_resv_clusters
));
2550 static ssize_t
reserved_clusters_store(struct ext4_attr
*a
,
2551 struct ext4_sb_info
*sbi
,
2552 const char *buf
, size_t count
)
2554 unsigned long long val
;
2557 if (parse_strtoull(buf
, -1ULL, &val
))
2559 ret
= ext4_reserve_clusters(sbi
, val
);
2561 return ret
? ret
: count
;
2564 static ssize_t
trigger_test_error(struct ext4_attr
*a
,
2565 struct ext4_sb_info
*sbi
,
2566 const char *buf
, size_t count
)
2570 if (!capable(CAP_SYS_ADMIN
))
2573 if (len
&& buf
[len
-1] == '\n')
2577 ext4_error(sbi
->s_sb
, "%.*s", len
, buf
);
2581 static ssize_t
sbi_deprecated_show(struct ext4_attr
*a
,
2582 struct ext4_sb_info
*sbi
, char *buf
)
2584 return snprintf(buf
, PAGE_SIZE
, "%d\n", a
->u
.deprecated_val
);
2587 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2588 static struct ext4_attr ext4_attr_##_name = { \
2589 .attr = {.name = __stringify(_name), .mode = _mode }, \
2593 .offset = offsetof(struct ext4_sb_info, _elname),\
2597 #define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname) \
2598 static struct ext4_attr ext4_attr_##_name = { \
2599 .attr = {.name = __stringify(_name), .mode = _mode }, \
2603 .offset = offsetof(struct ext4_super_block, _elname), \
2607 #define EXT4_ATTR(name, mode, show, store) \
2608 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2610 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2611 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2612 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2614 #define EXT4_RO_ATTR_ES_UI(name, elname) \
2615 EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
2616 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2617 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2619 #define ATTR_LIST(name) &ext4_attr_##name.attr
2620 #define EXT4_DEPRECATED_ATTR(_name, _val) \
2621 static struct ext4_attr ext4_attr_##_name = { \
2622 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2623 .show = sbi_deprecated_show, \
2625 .deprecated_val = _val, \
2629 EXT4_RO_ATTR(delayed_allocation_blocks
);
2630 EXT4_RO_ATTR(session_write_kbytes
);
2631 EXT4_RO_ATTR(lifetime_write_kbytes
);
2632 EXT4_RW_ATTR(reserved_clusters
);
2633 EXT4_ATTR_OFFSET(inode_readahead_blks
, 0644, sbi_ui_show
,
2634 inode_readahead_blks_store
, s_inode_readahead_blks
);
2635 EXT4_RW_ATTR_SBI_UI(inode_goal
, s_inode_goal
);
2636 EXT4_RW_ATTR_SBI_UI(mb_stats
, s_mb_stats
);
2637 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan
, s_mb_max_to_scan
);
2638 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan
, s_mb_min_to_scan
);
2639 EXT4_RW_ATTR_SBI_UI(mb_order2_req
, s_mb_order2_reqs
);
2640 EXT4_RW_ATTR_SBI_UI(mb_stream_req
, s_mb_stream_request
);
2641 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc
, s_mb_group_prealloc
);
2642 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump
, 128);
2643 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb
, s_extent_max_zeroout_kb
);
2644 EXT4_ATTR(trigger_fs_error
, 0200, NULL
, trigger_test_error
);
2645 EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms
, s_err_ratelimit_state
.interval
);
2646 EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst
, s_err_ratelimit_state
.burst
);
2647 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms
, s_warning_ratelimit_state
.interval
);
2648 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst
, s_warning_ratelimit_state
.burst
);
2649 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms
, s_msg_ratelimit_state
.interval
);
2650 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst
, s_msg_ratelimit_state
.burst
);
2651 EXT4_RO_ATTR_ES_UI(errors_count
, s_error_count
);
2652 EXT4_RO_ATTR_ES_UI(first_error_time
, s_first_error_time
);
2653 EXT4_RO_ATTR_ES_UI(last_error_time
, s_last_error_time
);
2655 static struct attribute
*ext4_attrs
[] = {
2656 ATTR_LIST(delayed_allocation_blocks
),
2657 ATTR_LIST(session_write_kbytes
),
2658 ATTR_LIST(lifetime_write_kbytes
),
2659 ATTR_LIST(reserved_clusters
),
2660 ATTR_LIST(inode_readahead_blks
),
2661 ATTR_LIST(inode_goal
),
2662 ATTR_LIST(mb_stats
),
2663 ATTR_LIST(mb_max_to_scan
),
2664 ATTR_LIST(mb_min_to_scan
),
2665 ATTR_LIST(mb_order2_req
),
2666 ATTR_LIST(mb_stream_req
),
2667 ATTR_LIST(mb_group_prealloc
),
2668 ATTR_LIST(max_writeback_mb_bump
),
2669 ATTR_LIST(extent_max_zeroout_kb
),
2670 ATTR_LIST(trigger_fs_error
),
2671 ATTR_LIST(err_ratelimit_interval_ms
),
2672 ATTR_LIST(err_ratelimit_burst
),
2673 ATTR_LIST(warning_ratelimit_interval_ms
),
2674 ATTR_LIST(warning_ratelimit_burst
),
2675 ATTR_LIST(msg_ratelimit_interval_ms
),
2676 ATTR_LIST(msg_ratelimit_burst
),
2677 ATTR_LIST(errors_count
),
2678 ATTR_LIST(first_error_time
),
2679 ATTR_LIST(last_error_time
),
2683 /* Features this copy of ext4 supports */
2684 EXT4_INFO_ATTR(lazy_itable_init
);
2685 EXT4_INFO_ATTR(batched_discard
);
2686 EXT4_INFO_ATTR(meta_bg_resize
);
2688 static struct attribute
*ext4_feat_attrs
[] = {
2689 ATTR_LIST(lazy_itable_init
),
2690 ATTR_LIST(batched_discard
),
2691 ATTR_LIST(meta_bg_resize
),
2695 static ssize_t
ext4_attr_show(struct kobject
*kobj
,
2696 struct attribute
*attr
, char *buf
)
2698 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2700 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2702 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
2705 static ssize_t
ext4_attr_store(struct kobject
*kobj
,
2706 struct attribute
*attr
,
2707 const char *buf
, size_t len
)
2709 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2711 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2713 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
2716 static void ext4_sb_release(struct kobject
*kobj
)
2718 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2720 complete(&sbi
->s_kobj_unregister
);
2723 static const struct sysfs_ops ext4_attr_ops
= {
2724 .show
= ext4_attr_show
,
2725 .store
= ext4_attr_store
,
2728 static struct kobj_type ext4_ktype
= {
2729 .default_attrs
= ext4_attrs
,
2730 .sysfs_ops
= &ext4_attr_ops
,
2731 .release
= ext4_sb_release
,
2734 static void ext4_feat_release(struct kobject
*kobj
)
2736 complete(&ext4_feat
->f_kobj_unregister
);
2739 static ssize_t
ext4_feat_show(struct kobject
*kobj
,
2740 struct attribute
*attr
, char *buf
)
2742 return snprintf(buf
, PAGE_SIZE
, "supported\n");
2746 * We can not use ext4_attr_show/store because it relies on the kobject
2747 * being embedded in the ext4_sb_info structure which is definitely not
2748 * true in this case.
2750 static const struct sysfs_ops ext4_feat_ops
= {
2751 .show
= ext4_feat_show
,
2755 static struct kobj_type ext4_feat_ktype
= {
2756 .default_attrs
= ext4_feat_attrs
,
2757 .sysfs_ops
= &ext4_feat_ops
,
2758 .release
= ext4_feat_release
,
2762 * Check whether this filesystem can be mounted based on
2763 * the features present and the RDONLY/RDWR mount requested.
2764 * Returns 1 if this filesystem can be mounted as requested,
2765 * 0 if it cannot be.
2767 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
2769 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT4_FEATURE_INCOMPAT_SUPP
)) {
2770 ext4_msg(sb
, KERN_ERR
,
2771 "Couldn't mount because of "
2772 "unsupported optional features (%x)",
2773 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2774 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2781 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_READONLY
)) {
2782 ext4_msg(sb
, KERN_INFO
, "filesystem is read-only");
2783 sb
->s_flags
|= MS_RDONLY
;
2787 /* Check that feature set is OK for a read-write mount */
2788 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT4_FEATURE_RO_COMPAT_SUPP
)) {
2789 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
2790 "unsupported optional features (%x)",
2791 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2792 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2796 * Large file size enabled file system can only be mounted
2797 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2799 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_HUGE_FILE
)) {
2800 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
2801 ext4_msg(sb
, KERN_ERR
, "Filesystem with huge files "
2802 "cannot be mounted RDWR without "
2807 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_BIGALLOC
) &&
2808 !EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_EXTENTS
)) {
2809 ext4_msg(sb
, KERN_ERR
,
2810 "Can't support bigalloc feature without "
2811 "extents feature\n");
2815 #ifndef CONFIG_QUOTA
2816 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
) &&
2818 ext4_msg(sb
, KERN_ERR
,
2819 "Filesystem with quota feature cannot be mounted RDWR "
2820 "without CONFIG_QUOTA");
2823 #endif /* CONFIG_QUOTA */
2828 * This function is called once a day if we have errors logged
2829 * on the file system
2831 static void print_daily_error_info(unsigned long arg
)
2833 struct super_block
*sb
= (struct super_block
*) arg
;
2834 struct ext4_sb_info
*sbi
;
2835 struct ext4_super_block
*es
;
2840 if (es
->s_error_count
)
2841 /* fsck newer than v1.41.13 is needed to clean this condition. */
2842 ext4_msg(sb
, KERN_NOTICE
, "error count since last fsck: %u",
2843 le32_to_cpu(es
->s_error_count
));
2844 if (es
->s_first_error_time
) {
2845 printk(KERN_NOTICE
"EXT4-fs (%s): initial error at time %u: %.*s:%d",
2846 sb
->s_id
, le32_to_cpu(es
->s_first_error_time
),
2847 (int) sizeof(es
->s_first_error_func
),
2848 es
->s_first_error_func
,
2849 le32_to_cpu(es
->s_first_error_line
));
2850 if (es
->s_first_error_ino
)
2851 printk(": inode %u",
2852 le32_to_cpu(es
->s_first_error_ino
));
2853 if (es
->s_first_error_block
)
2854 printk(": block %llu", (unsigned long long)
2855 le64_to_cpu(es
->s_first_error_block
));
2858 if (es
->s_last_error_time
) {
2859 printk(KERN_NOTICE
"EXT4-fs (%s): last error at time %u: %.*s:%d",
2860 sb
->s_id
, le32_to_cpu(es
->s_last_error_time
),
2861 (int) sizeof(es
->s_last_error_func
),
2862 es
->s_last_error_func
,
2863 le32_to_cpu(es
->s_last_error_line
));
2864 if (es
->s_last_error_ino
)
2865 printk(": inode %u",
2866 le32_to_cpu(es
->s_last_error_ino
));
2867 if (es
->s_last_error_block
)
2868 printk(": block %llu", (unsigned long long)
2869 le64_to_cpu(es
->s_last_error_block
));
2872 mod_timer(&sbi
->s_err_report
, jiffies
+ 24*60*60*HZ
); /* Once a day */
2875 /* Find next suitable group and run ext4_init_inode_table */
2876 static int ext4_run_li_request(struct ext4_li_request
*elr
)
2878 struct ext4_group_desc
*gdp
= NULL
;
2879 ext4_group_t group
, ngroups
;
2880 struct super_block
*sb
;
2881 unsigned long timeout
= 0;
2885 ngroups
= EXT4_SB(sb
)->s_groups_count
;
2888 for (group
= elr
->lr_next_group
; group
< ngroups
; group
++) {
2889 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2895 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2899 if (group
>= ngroups
)
2904 ret
= ext4_init_inode_table(sb
, group
,
2905 elr
->lr_timeout
? 0 : 1);
2906 if (elr
->lr_timeout
== 0) {
2907 timeout
= (jiffies
- timeout
) *
2908 elr
->lr_sbi
->s_li_wait_mult
;
2909 elr
->lr_timeout
= timeout
;
2911 elr
->lr_next_sched
= jiffies
+ elr
->lr_timeout
;
2912 elr
->lr_next_group
= group
+ 1;
2920 * Remove lr_request from the list_request and free the
2921 * request structure. Should be called with li_list_mtx held
2923 static void ext4_remove_li_request(struct ext4_li_request
*elr
)
2925 struct ext4_sb_info
*sbi
;
2932 list_del(&elr
->lr_request
);
2933 sbi
->s_li_request
= NULL
;
2937 static void ext4_unregister_li_request(struct super_block
*sb
)
2939 mutex_lock(&ext4_li_mtx
);
2940 if (!ext4_li_info
) {
2941 mutex_unlock(&ext4_li_mtx
);
2945 mutex_lock(&ext4_li_info
->li_list_mtx
);
2946 ext4_remove_li_request(EXT4_SB(sb
)->s_li_request
);
2947 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2948 mutex_unlock(&ext4_li_mtx
);
2951 static struct task_struct
*ext4_lazyinit_task
;
2954 * This is the function where ext4lazyinit thread lives. It walks
2955 * through the request list searching for next scheduled filesystem.
2956 * When such a fs is found, run the lazy initialization request
2957 * (ext4_rn_li_request) and keep track of the time spend in this
2958 * function. Based on that time we compute next schedule time of
2959 * the request. When walking through the list is complete, compute
2960 * next waking time and put itself into sleep.
2962 static int ext4_lazyinit_thread(void *arg
)
2964 struct ext4_lazy_init
*eli
= (struct ext4_lazy_init
*)arg
;
2965 struct list_head
*pos
, *n
;
2966 struct ext4_li_request
*elr
;
2967 unsigned long next_wakeup
, cur
;
2969 BUG_ON(NULL
== eli
);
2973 next_wakeup
= MAX_JIFFY_OFFSET
;
2975 mutex_lock(&eli
->li_list_mtx
);
2976 if (list_empty(&eli
->li_request_list
)) {
2977 mutex_unlock(&eli
->li_list_mtx
);
2981 list_for_each_safe(pos
, n
, &eli
->li_request_list
) {
2982 elr
= list_entry(pos
, struct ext4_li_request
,
2985 if (time_after_eq(jiffies
, elr
->lr_next_sched
)) {
2986 if (ext4_run_li_request(elr
) != 0) {
2987 /* error, remove the lazy_init job */
2988 ext4_remove_li_request(elr
);
2993 if (time_before(elr
->lr_next_sched
, next_wakeup
))
2994 next_wakeup
= elr
->lr_next_sched
;
2996 mutex_unlock(&eli
->li_list_mtx
);
3001 if ((time_after_eq(cur
, next_wakeup
)) ||
3002 (MAX_JIFFY_OFFSET
== next_wakeup
)) {
3007 schedule_timeout_interruptible(next_wakeup
- cur
);
3009 if (kthread_should_stop()) {
3010 ext4_clear_request_list();
3017 * It looks like the request list is empty, but we need
3018 * to check it under the li_list_mtx lock, to prevent any
3019 * additions into it, and of course we should lock ext4_li_mtx
3020 * to atomically free the list and ext4_li_info, because at
3021 * this point another ext4 filesystem could be registering
3024 mutex_lock(&ext4_li_mtx
);
3025 mutex_lock(&eli
->li_list_mtx
);
3026 if (!list_empty(&eli
->li_request_list
)) {
3027 mutex_unlock(&eli
->li_list_mtx
);
3028 mutex_unlock(&ext4_li_mtx
);
3031 mutex_unlock(&eli
->li_list_mtx
);
3032 kfree(ext4_li_info
);
3033 ext4_li_info
= NULL
;
3034 mutex_unlock(&ext4_li_mtx
);
3039 static void ext4_clear_request_list(void)
3041 struct list_head
*pos
, *n
;
3042 struct ext4_li_request
*elr
;
3044 mutex_lock(&ext4_li_info
->li_list_mtx
);
3045 list_for_each_safe(pos
, n
, &ext4_li_info
->li_request_list
) {
3046 elr
= list_entry(pos
, struct ext4_li_request
,
3048 ext4_remove_li_request(elr
);
3050 mutex_unlock(&ext4_li_info
->li_list_mtx
);
3053 static int ext4_run_lazyinit_thread(void)
3055 ext4_lazyinit_task
= kthread_run(ext4_lazyinit_thread
,
3056 ext4_li_info
, "ext4lazyinit");
3057 if (IS_ERR(ext4_lazyinit_task
)) {
3058 int err
= PTR_ERR(ext4_lazyinit_task
);
3059 ext4_clear_request_list();
3060 kfree(ext4_li_info
);
3061 ext4_li_info
= NULL
;
3062 printk(KERN_CRIT
"EXT4-fs: error %d creating inode table "
3063 "initialization thread\n",
3067 ext4_li_info
->li_state
|= EXT4_LAZYINIT_RUNNING
;
3072 * Check whether it make sense to run itable init. thread or not.
3073 * If there is at least one uninitialized inode table, return
3074 * corresponding group number, else the loop goes through all
3075 * groups and return total number of groups.
3077 static ext4_group_t
ext4_has_uninit_itable(struct super_block
*sb
)
3079 ext4_group_t group
, ngroups
= EXT4_SB(sb
)->s_groups_count
;
3080 struct ext4_group_desc
*gdp
= NULL
;
3082 for (group
= 0; group
< ngroups
; group
++) {
3083 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
3087 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
3094 static int ext4_li_info_new(void)
3096 struct ext4_lazy_init
*eli
= NULL
;
3098 eli
= kzalloc(sizeof(*eli
), GFP_KERNEL
);
3102 INIT_LIST_HEAD(&eli
->li_request_list
);
3103 mutex_init(&eli
->li_list_mtx
);
3105 eli
->li_state
|= EXT4_LAZYINIT_QUIT
;
3112 static struct ext4_li_request
*ext4_li_request_new(struct super_block
*sb
,
3115 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3116 struct ext4_li_request
*elr
;
3118 elr
= kzalloc(sizeof(*elr
), GFP_KERNEL
);
3124 elr
->lr_next_group
= start
;
3127 * Randomize first schedule time of the request to
3128 * spread the inode table initialization requests
3131 elr
->lr_next_sched
= jiffies
+ (prandom_u32() %
3132 (EXT4_DEF_LI_MAX_START_DELAY
* HZ
));
3136 int ext4_register_li_request(struct super_block
*sb
,
3137 ext4_group_t first_not_zeroed
)
3139 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3140 struct ext4_li_request
*elr
= NULL
;
3141 ext4_group_t ngroups
= EXT4_SB(sb
)->s_groups_count
;
3144 mutex_lock(&ext4_li_mtx
);
3145 if (sbi
->s_li_request
!= NULL
) {
3147 * Reset timeout so it can be computed again, because
3148 * s_li_wait_mult might have changed.
3150 sbi
->s_li_request
->lr_timeout
= 0;
3154 if (first_not_zeroed
== ngroups
||
3155 (sb
->s_flags
& MS_RDONLY
) ||
3156 !test_opt(sb
, INIT_INODE_TABLE
))
3159 elr
= ext4_li_request_new(sb
, first_not_zeroed
);
3165 if (NULL
== ext4_li_info
) {
3166 ret
= ext4_li_info_new();
3171 mutex_lock(&ext4_li_info
->li_list_mtx
);
3172 list_add(&elr
->lr_request
, &ext4_li_info
->li_request_list
);
3173 mutex_unlock(&ext4_li_info
->li_list_mtx
);
3175 sbi
->s_li_request
= elr
;
3177 * set elr to NULL here since it has been inserted to
3178 * the request_list and the removal and free of it is
3179 * handled by ext4_clear_request_list from now on.
3183 if (!(ext4_li_info
->li_state
& EXT4_LAZYINIT_RUNNING
)) {
3184 ret
= ext4_run_lazyinit_thread();
3189 mutex_unlock(&ext4_li_mtx
);
3196 * We do not need to lock anything since this is called on
3199 static void ext4_destroy_lazyinit_thread(void)
3202 * If thread exited earlier
3203 * there's nothing to be done.
3205 if (!ext4_li_info
|| !ext4_lazyinit_task
)
3208 kthread_stop(ext4_lazyinit_task
);
3211 static int set_journal_csum_feature_set(struct super_block
*sb
)
3214 int compat
, incompat
;
3215 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3217 if (ext4_has_metadata_csum(sb
)) {
3218 /* journal checksum v3 */
3220 incompat
= JBD2_FEATURE_INCOMPAT_CSUM_V3
;
3222 /* journal checksum v1 */
3223 compat
= JBD2_FEATURE_COMPAT_CHECKSUM
;
3227 jbd2_journal_clear_features(sbi
->s_journal
,
3228 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3229 JBD2_FEATURE_INCOMPAT_CSUM_V3
|
3230 JBD2_FEATURE_INCOMPAT_CSUM_V2
);
3231 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
3232 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3234 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
|
3236 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
3237 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3240 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3241 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3243 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3244 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3251 * Note: calculating the overhead so we can be compatible with
3252 * historical BSD practice is quite difficult in the face of
3253 * clusters/bigalloc. This is because multiple metadata blocks from
3254 * different block group can end up in the same allocation cluster.
3255 * Calculating the exact overhead in the face of clustered allocation
3256 * requires either O(all block bitmaps) in memory or O(number of block
3257 * groups**2) in time. We will still calculate the superblock for
3258 * older file systems --- and if we come across with a bigalloc file
3259 * system with zero in s_overhead_clusters the estimate will be close to
3260 * correct especially for very large cluster sizes --- but for newer
3261 * file systems, it's better to calculate this figure once at mkfs
3262 * time, and store it in the superblock. If the superblock value is
3263 * present (even for non-bigalloc file systems), we will use it.
3265 static int count_overhead(struct super_block
*sb
, ext4_group_t grp
,
3268 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3269 struct ext4_group_desc
*gdp
;
3270 ext4_fsblk_t first_block
, last_block
, b
;
3271 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3272 int s
, j
, count
= 0;
3274 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_BIGALLOC
))
3275 return (ext4_bg_has_super(sb
, grp
) + ext4_bg_num_gdb(sb
, grp
) +
3276 sbi
->s_itb_per_group
+ 2);
3278 first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
) +
3279 (grp
* EXT4_BLOCKS_PER_GROUP(sb
));
3280 last_block
= first_block
+ EXT4_BLOCKS_PER_GROUP(sb
) - 1;
3281 for (i
= 0; i
< ngroups
; i
++) {
3282 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
3283 b
= ext4_block_bitmap(sb
, gdp
);
3284 if (b
>= first_block
&& b
<= last_block
) {
3285 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3288 b
= ext4_inode_bitmap(sb
, gdp
);
3289 if (b
>= first_block
&& b
<= last_block
) {
3290 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3293 b
= ext4_inode_table(sb
, gdp
);
3294 if (b
>= first_block
&& b
+ sbi
->s_itb_per_group
<= last_block
)
3295 for (j
= 0; j
< sbi
->s_itb_per_group
; j
++, b
++) {
3296 int c
= EXT4_B2C(sbi
, b
- first_block
);
3297 ext4_set_bit(c
, buf
);
3303 if (ext4_bg_has_super(sb
, grp
)) {
3304 ext4_set_bit(s
++, buf
);
3307 for (j
= ext4_bg_num_gdb(sb
, grp
); j
> 0; j
--) {
3308 ext4_set_bit(EXT4_B2C(sbi
, s
++), buf
);
3314 return EXT4_CLUSTERS_PER_GROUP(sb
) -
3315 ext4_count_free(buf
, EXT4_CLUSTERS_PER_GROUP(sb
) / 8);
3319 * Compute the overhead and stash it in sbi->s_overhead
3321 int ext4_calculate_overhead(struct super_block
*sb
)
3323 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3324 struct ext4_super_block
*es
= sbi
->s_es
;
3325 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3326 ext4_fsblk_t overhead
= 0;
3327 char *buf
= (char *) get_zeroed_page(GFP_NOFS
);
3333 * Compute the overhead (FS structures). This is constant
3334 * for a given filesystem unless the number of block groups
3335 * changes so we cache the previous value until it does.
3339 * All of the blocks before first_data_block are overhead
3341 overhead
= EXT4_B2C(sbi
, le32_to_cpu(es
->s_first_data_block
));
3344 * Add the overhead found in each block group
3346 for (i
= 0; i
< ngroups
; i
++) {
3349 blks
= count_overhead(sb
, i
, buf
);
3352 memset(buf
, 0, PAGE_SIZE
);
3355 /* Add the internal journal blocks as well */
3356 if (sbi
->s_journal
&& !sbi
->journal_bdev
)
3357 overhead
+= EXT4_NUM_B2C(sbi
, sbi
->s_journal
->j_maxlen
);
3359 sbi
->s_overhead
= overhead
;
3361 free_page((unsigned long) buf
);
3366 static ext4_fsblk_t
ext4_calculate_resv_clusters(struct super_block
*sb
)
3368 ext4_fsblk_t resv_clusters
;
3371 * There's no need to reserve anything when we aren't using extents.
3372 * The space estimates are exact, there are no unwritten extents,
3373 * hole punching doesn't need new metadata... This is needed especially
3374 * to keep ext2/3 backward compatibility.
3376 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_EXTENTS
))
3379 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3380 * This should cover the situations where we can not afford to run
3381 * out of space like for example punch hole, or converting
3382 * unwritten extents in delalloc path. In most cases such
3383 * allocation would require 1, or 2 blocks, higher numbers are
3386 resv_clusters
= ext4_blocks_count(EXT4_SB(sb
)->s_es
) >>
3387 EXT4_SB(sb
)->s_cluster_bits
;
3389 do_div(resv_clusters
, 50);
3390 resv_clusters
= min_t(ext4_fsblk_t
, resv_clusters
, 4096);
3392 return resv_clusters
;
3396 static int ext4_reserve_clusters(struct ext4_sb_info
*sbi
, ext4_fsblk_t count
)
3398 ext4_fsblk_t clusters
= ext4_blocks_count(sbi
->s_es
) >>
3399 sbi
->s_cluster_bits
;
3401 if (count
>= clusters
)
3404 atomic64_set(&sbi
->s_resv_clusters
, count
);
3408 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
3410 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
3411 struct buffer_head
*bh
;
3412 struct ext4_super_block
*es
= NULL
;
3413 struct ext4_sb_info
*sbi
;
3415 ext4_fsblk_t sb_block
= get_sb_block(&data
);
3416 ext4_fsblk_t logical_sb_block
;
3417 unsigned long offset
= 0;
3418 unsigned long journal_devnum
= 0;
3419 unsigned long def_mount_opts
;
3424 int blocksize
, clustersize
;
3425 unsigned int db_count
;
3427 int needs_recovery
, has_huge_files
, has_bigalloc
;
3430 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
3431 ext4_group_t first_not_zeroed
;
3433 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
3437 sbi
->s_blockgroup_lock
=
3438 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
3439 if (!sbi
->s_blockgroup_lock
) {
3443 sb
->s_fs_info
= sbi
;
3445 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
3446 sbi
->s_sb_block
= sb_block
;
3447 if (sb
->s_bdev
->bd_part
)
3448 sbi
->s_sectors_written_start
=
3449 part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]);
3451 /* Cleanup superblock name */
3452 for (cp
= sb
->s_id
; (cp
= strchr(cp
, '/'));)
3455 /* -EINVAL is default */
3457 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
3459 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
3464 * The ext4 superblock will not be buffer aligned for other than 1kB
3465 * block sizes. We need to calculate the offset from buffer start.
3467 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
3468 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3469 offset
= do_div(logical_sb_block
, blocksize
);
3471 logical_sb_block
= sb_block
;
3474 if (!(bh
= sb_bread_unmovable(sb
, logical_sb_block
))) {
3475 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
3479 * Note: s_es must be initialized as soon as possible because
3480 * some ext4 macro-instructions depend on its value
3482 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
3484 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
3485 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
3487 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
3489 /* Warn if metadata_csum and gdt_csum are both set. */
3490 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3491 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
) &&
3492 EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_GDT_CSUM
))
3493 ext4_warning(sb
, "metadata_csum and uninit_bg are "
3494 "redundant flags; please run fsck.");
3496 /* Check for a known checksum algorithm */
3497 if (!ext4_verify_csum_type(sb
, es
)) {
3498 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3499 "unknown checksum algorithm.");
3504 /* Load the checksum driver */
3505 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3506 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
)) {
3507 sbi
->s_chksum_driver
= crypto_alloc_shash("crc32c", 0, 0);
3508 if (IS_ERR(sbi
->s_chksum_driver
)) {
3509 ext4_msg(sb
, KERN_ERR
, "Cannot load crc32c driver.");
3510 ret
= PTR_ERR(sbi
->s_chksum_driver
);
3511 sbi
->s_chksum_driver
= NULL
;
3516 /* Check superblock checksum */
3517 if (!ext4_superblock_csum_verify(sb
, es
)) {
3518 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3519 "invalid superblock checksum. Run e2fsck?");
3524 /* Precompute checksum seed for all metadata */
3525 if (ext4_has_metadata_csum(sb
))
3526 sbi
->s_csum_seed
= ext4_chksum(sbi
, ~0, es
->s_uuid
,
3527 sizeof(es
->s_uuid
));
3529 /* Set defaults before we parse the mount options */
3530 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
3531 set_opt(sb
, INIT_INODE_TABLE
);
3532 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
3534 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
)
3536 if (def_mount_opts
& EXT4_DEFM_UID16
)
3537 set_opt(sb
, NO_UID32
);
3538 /* xattr user namespace & acls are now defaulted on */
3539 set_opt(sb
, XATTR_USER
);
3540 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3541 set_opt(sb
, POSIX_ACL
);
3543 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3544 if (ext4_has_metadata_csum(sb
))
3545 set_opt(sb
, JOURNAL_CHECKSUM
);
3547 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
3548 set_opt(sb
, JOURNAL_DATA
);
3549 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
3550 set_opt(sb
, ORDERED_DATA
);
3551 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
3552 set_opt(sb
, WRITEBACK_DATA
);
3554 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
3555 set_opt(sb
, ERRORS_PANIC
);
3556 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
3557 set_opt(sb
, ERRORS_CONT
);
3559 set_opt(sb
, ERRORS_RO
);
3560 /* block_validity enabled by default; disable with noblock_validity */
3561 set_opt(sb
, BLOCK_VALIDITY
);
3562 if (def_mount_opts
& EXT4_DEFM_DISCARD
)
3563 set_opt(sb
, DISCARD
);
3565 sbi
->s_resuid
= make_kuid(&init_user_ns
, le16_to_cpu(es
->s_def_resuid
));
3566 sbi
->s_resgid
= make_kgid(&init_user_ns
, le16_to_cpu(es
->s_def_resgid
));
3567 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
3568 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
3569 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
3571 if ((def_mount_opts
& EXT4_DEFM_NOBARRIER
) == 0)
3572 set_opt(sb
, BARRIER
);
3575 * enable delayed allocation by default
3576 * Use -o nodelalloc to turn it off
3578 if (!IS_EXT3_SB(sb
) && !IS_EXT2_SB(sb
) &&
3579 ((def_mount_opts
& EXT4_DEFM_NODELALLOC
) == 0))
3580 set_opt(sb
, DELALLOC
);
3583 * set default s_li_wait_mult for lazyinit, for the case there is
3584 * no mount option specified.
3586 sbi
->s_li_wait_mult
= EXT4_DEF_LI_WAIT_MULT
;
3588 if (!parse_options((char *) sbi
->s_es
->s_mount_opts
, sb
,
3589 &journal_devnum
, &journal_ioprio
, 0)) {
3590 ext4_msg(sb
, KERN_WARNING
,
3591 "failed to parse options in superblock: %s",
3592 sbi
->s_es
->s_mount_opts
);
3594 sbi
->s_def_mount_opt
= sbi
->s_mount_opt
;
3595 if (!parse_options((char *) data
, sb
, &journal_devnum
,
3596 &journal_ioprio
, 0))
3599 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
3600 printk_once(KERN_WARNING
"EXT4-fs: Warning: mounting "
3601 "with data=journal disables delayed "
3602 "allocation and O_DIRECT support!\n");
3603 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
3604 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3605 "both data=journal and delalloc");
3608 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
3609 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3610 "both data=journal and dioread_nolock");
3613 if (test_opt(sb
, DAX
)) {
3614 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3615 "both data=journal and dax");
3618 if (test_opt(sb
, DELALLOC
))
3619 clear_opt(sb
, DELALLOC
);
3622 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3623 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
3625 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
3626 (EXT4_HAS_COMPAT_FEATURE(sb
, ~0U) ||
3627 EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
3628 EXT4_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
3629 ext4_msg(sb
, KERN_WARNING
,
3630 "feature flags set on rev 0 fs, "
3631 "running e2fsck is recommended");
3633 if (es
->s_creator_os
== cpu_to_le32(EXT4_OS_HURD
)) {
3634 set_opt2(sb
, HURD_COMPAT
);
3635 if (EXT4_HAS_INCOMPAT_FEATURE(sb
,
3636 EXT4_FEATURE_INCOMPAT_64BIT
)) {
3637 ext4_msg(sb
, KERN_ERR
,
3638 "The Hurd can't support 64-bit file systems");
3643 if (IS_EXT2_SB(sb
)) {
3644 if (ext2_feature_set_ok(sb
))
3645 ext4_msg(sb
, KERN_INFO
, "mounting ext2 file system "
3646 "using the ext4 subsystem");
3648 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext2 due "
3649 "to feature incompatibilities");
3654 if (IS_EXT3_SB(sb
)) {
3655 if (ext3_feature_set_ok(sb
))
3656 ext4_msg(sb
, KERN_INFO
, "mounting ext3 file system "
3657 "using the ext4 subsystem");
3659 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext3 due "
3660 "to feature incompatibilities");
3666 * Check feature flags regardless of the revision level, since we
3667 * previously didn't change the revision level when setting the flags,
3668 * so there is a chance incompat flags are set on a rev 0 filesystem.
3670 if (!ext4_feature_set_ok(sb
, (sb
->s_flags
& MS_RDONLY
)))
3673 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
3674 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
3675 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
3676 ext4_msg(sb
, KERN_ERR
,
3677 "Unsupported filesystem blocksize %d", blocksize
);
3681 if (sbi
->s_mount_opt
& EXT4_MOUNT_DAX
) {
3682 if (blocksize
!= PAGE_SIZE
) {
3683 ext4_msg(sb
, KERN_ERR
,
3684 "error: unsupported blocksize for dax");
3687 if (!sb
->s_bdev
->bd_disk
->fops
->direct_access
) {
3688 ext4_msg(sb
, KERN_ERR
,
3689 "error: device does not support dax");
3694 if (sb
->s_blocksize
!= blocksize
) {
3695 /* Validate the filesystem blocksize */
3696 if (!sb_set_blocksize(sb
, blocksize
)) {
3697 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
3703 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3704 offset
= do_div(logical_sb_block
, blocksize
);
3705 bh
= sb_bread_unmovable(sb
, logical_sb_block
);
3707 ext4_msg(sb
, KERN_ERR
,
3708 "Can't read superblock on 2nd try");
3711 es
= (struct ext4_super_block
*)(bh
->b_data
+ offset
);
3713 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
3714 ext4_msg(sb
, KERN_ERR
,
3715 "Magic mismatch, very weird!");
3720 has_huge_files
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3721 EXT4_FEATURE_RO_COMPAT_HUGE_FILE
);
3722 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
3724 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
3726 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
3727 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
3728 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
3730 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
3731 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
3732 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
3733 (!is_power_of_2(sbi
->s_inode_size
)) ||
3734 (sbi
->s_inode_size
> blocksize
)) {
3735 ext4_msg(sb
, KERN_ERR
,
3736 "unsupported inode size: %d",
3740 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
3741 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
3744 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
3745 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
)) {
3746 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
3747 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
3748 !is_power_of_2(sbi
->s_desc_size
)) {
3749 ext4_msg(sb
, KERN_ERR
,
3750 "unsupported descriptor size %lu",
3755 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
3757 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
3758 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
3759 if (EXT4_INODE_SIZE(sb
) == 0 || EXT4_INODES_PER_GROUP(sb
) == 0)
3762 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
3763 if (sbi
->s_inodes_per_block
== 0)
3765 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
3766 sbi
->s_inodes_per_block
;
3767 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
3769 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3770 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
3771 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
3773 for (i
= 0; i
< 4; i
++)
3774 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
3775 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
3776 if (EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_DIR_INDEX
)) {
3777 i
= le32_to_cpu(es
->s_flags
);
3778 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
3779 sbi
->s_hash_unsigned
= 3;
3780 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
3781 #ifdef __CHAR_UNSIGNED__
3782 if (!(sb
->s_flags
& MS_RDONLY
))
3784 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
3785 sbi
->s_hash_unsigned
= 3;
3787 if (!(sb
->s_flags
& MS_RDONLY
))
3789 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
3794 /* Handle clustersize */
3795 clustersize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_cluster_size
);
3796 has_bigalloc
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3797 EXT4_FEATURE_RO_COMPAT_BIGALLOC
);
3799 if (clustersize
< blocksize
) {
3800 ext4_msg(sb
, KERN_ERR
,
3801 "cluster size (%d) smaller than "
3802 "block size (%d)", clustersize
, blocksize
);
3805 sbi
->s_cluster_bits
= le32_to_cpu(es
->s_log_cluster_size
) -
3806 le32_to_cpu(es
->s_log_block_size
);
3807 sbi
->s_clusters_per_group
=
3808 le32_to_cpu(es
->s_clusters_per_group
);
3809 if (sbi
->s_clusters_per_group
> blocksize
* 8) {
3810 ext4_msg(sb
, KERN_ERR
,
3811 "#clusters per group too big: %lu",
3812 sbi
->s_clusters_per_group
);
3815 if (sbi
->s_blocks_per_group
!=
3816 (sbi
->s_clusters_per_group
* (clustersize
/ blocksize
))) {
3817 ext4_msg(sb
, KERN_ERR
, "blocks per group (%lu) and "
3818 "clusters per group (%lu) inconsistent",
3819 sbi
->s_blocks_per_group
,
3820 sbi
->s_clusters_per_group
);
3824 if (clustersize
!= blocksize
) {
3825 ext4_warning(sb
, "fragment/cluster size (%d) != "
3826 "block size (%d)", clustersize
,
3828 clustersize
= blocksize
;
3830 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
3831 ext4_msg(sb
, KERN_ERR
,
3832 "#blocks per group too big: %lu",
3833 sbi
->s_blocks_per_group
);
3836 sbi
->s_clusters_per_group
= sbi
->s_blocks_per_group
;
3837 sbi
->s_cluster_bits
= 0;
3839 sbi
->s_cluster_ratio
= clustersize
/ blocksize
;
3841 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
3842 ext4_msg(sb
, KERN_ERR
,
3843 "#inodes per group too big: %lu",
3844 sbi
->s_inodes_per_group
);
3848 /* Do we have standard group size of clustersize * 8 blocks ? */
3849 if (sbi
->s_blocks_per_group
== clustersize
<< 3)
3850 set_opt2(sb
, STD_GROUP_SIZE
);
3853 * Test whether we have more sectors than will fit in sector_t,
3854 * and whether the max offset is addressable by the page cache.
3856 err
= generic_check_addressable(sb
->s_blocksize_bits
,
3857 ext4_blocks_count(es
));
3859 ext4_msg(sb
, KERN_ERR
, "filesystem"
3860 " too large to mount safely on this system");
3861 if (sizeof(sector_t
) < 8)
3862 ext4_msg(sb
, KERN_WARNING
, "CONFIG_LBDAF not enabled");
3866 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
3869 /* check blocks count against device size */
3870 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
3871 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
3872 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
3873 "exceeds size of device (%llu blocks)",
3874 ext4_blocks_count(es
), blocks_count
);
3879 * It makes no sense for the first data block to be beyond the end
3880 * of the filesystem.
3882 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
3883 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data "
3884 "block %u is beyond end of filesystem (%llu)",
3885 le32_to_cpu(es
->s_first_data_block
),
3886 ext4_blocks_count(es
));
3889 blocks_count
= (ext4_blocks_count(es
) -
3890 le32_to_cpu(es
->s_first_data_block
) +
3891 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
3892 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
3893 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
3894 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %u "
3895 "(block count %llu, first data block %u, "
3896 "blocks per group %lu)", sbi
->s_groups_count
,
3897 ext4_blocks_count(es
),
3898 le32_to_cpu(es
->s_first_data_block
),
3899 EXT4_BLOCKS_PER_GROUP(sb
));
3902 sbi
->s_groups_count
= blocks_count
;
3903 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
3904 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
3905 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
3906 EXT4_DESC_PER_BLOCK(sb
);
3907 sbi
->s_group_desc
= ext4_kvmalloc(db_count
*
3908 sizeof(struct buffer_head
*),
3910 if (sbi
->s_group_desc
== NULL
) {
3911 ext4_msg(sb
, KERN_ERR
, "not enough memory");
3917 sbi
->s_proc
= proc_mkdir(sb
->s_id
, ext4_proc_root
);
3920 proc_create_data("options", S_IRUGO
, sbi
->s_proc
,
3921 &ext4_seq_options_fops
, sb
);
3923 bgl_lock_init(sbi
->s_blockgroup_lock
);
3925 for (i
= 0; i
< db_count
; i
++) {
3926 block
= descriptor_loc(sb
, logical_sb_block
, i
);
3927 sbi
->s_group_desc
[i
] = sb_bread_unmovable(sb
, block
);
3928 if (!sbi
->s_group_desc
[i
]) {
3929 ext4_msg(sb
, KERN_ERR
,
3930 "can't read group descriptor %d", i
);
3935 if (!ext4_check_descriptors(sb
, &first_not_zeroed
)) {
3936 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
3940 sbi
->s_gdb_count
= db_count
;
3941 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
3942 spin_lock_init(&sbi
->s_next_gen_lock
);
3944 setup_timer(&sbi
->s_err_report
, print_daily_error_info
,
3945 (unsigned long) sb
);
3947 /* Register extent status tree shrinker */
3948 if (ext4_es_register_shrinker(sbi
))
3951 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
3952 sbi
->s_extent_max_zeroout_kb
= 32;
3955 * set up enough so that it can read an inode
3957 sb
->s_op
= &ext4_sops
;
3958 sb
->s_export_op
= &ext4_export_ops
;
3959 sb
->s_xattr
= ext4_xattr_handlers
;
3961 sb
->dq_op
= &ext4_quota_operations
;
3962 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
))
3963 sb
->s_qcop
= &dquot_quotactl_sysfile_ops
;
3965 sb
->s_qcop
= &ext4_qctl_operations
;
3966 sb
->s_quota_types
= QTYPE_MASK_USR
| QTYPE_MASK_GRP
;
3968 memcpy(sb
->s_uuid
, es
->s_uuid
, sizeof(es
->s_uuid
));
3970 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
3971 mutex_init(&sbi
->s_orphan_lock
);
3975 needs_recovery
= (es
->s_last_orphan
!= 0 ||
3976 EXT4_HAS_INCOMPAT_FEATURE(sb
,
3977 EXT4_FEATURE_INCOMPAT_RECOVER
));
3979 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_MMP
) &&
3980 !(sb
->s_flags
& MS_RDONLY
))
3981 if (ext4_multi_mount_protect(sb
, le64_to_cpu(es
->s_mmp_block
)))
3982 goto failed_mount3a
;
3985 * The first inode we look at is the journal inode. Don't try
3986 * root first: it may be modified in the journal!
3988 if (!test_opt(sb
, NOLOAD
) &&
3989 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
3990 if (ext4_load_journal(sb
, es
, journal_devnum
))
3991 goto failed_mount3a
;
3992 } else if (test_opt(sb
, NOLOAD
) && !(sb
->s_flags
& MS_RDONLY
) &&
3993 EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
3994 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
3995 "suppressed and not mounted read-only");
3996 goto failed_mount_wq
;
3998 clear_opt(sb
, DATA_FLAGS
);
3999 sbi
->s_journal
= NULL
;
4004 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
) &&
4005 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
4006 JBD2_FEATURE_INCOMPAT_64BIT
)) {
4007 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
4008 goto failed_mount_wq
;
4011 if (!set_journal_csum_feature_set(sb
)) {
4012 ext4_msg(sb
, KERN_ERR
, "Failed to set journal checksum "
4014 goto failed_mount_wq
;
4017 /* We have now updated the journal if required, so we can
4018 * validate the data journaling mode. */
4019 switch (test_opt(sb
, DATA_FLAGS
)) {
4021 /* No mode set, assume a default based on the journal
4022 * capabilities: ORDERED_DATA if the journal can
4023 * cope, else JOURNAL_DATA
4025 if (jbd2_journal_check_available_features
4026 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
4027 set_opt(sb
, ORDERED_DATA
);
4029 set_opt(sb
, JOURNAL_DATA
);
4032 case EXT4_MOUNT_ORDERED_DATA
:
4033 case EXT4_MOUNT_WRITEBACK_DATA
:
4034 if (!jbd2_journal_check_available_features
4035 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
4036 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
4037 "requested data journaling mode");
4038 goto failed_mount_wq
;
4043 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
4045 sbi
->s_journal
->j_commit_callback
= ext4_journal_commit_callback
;
4048 if (ext4_mballoc_ready
) {
4049 sbi
->s_mb_cache
= ext4_xattr_create_cache(sb
->s_id
);
4050 if (!sbi
->s_mb_cache
) {
4051 ext4_msg(sb
, KERN_ERR
, "Failed to create an mb_cache");
4052 goto failed_mount_wq
;
4057 * Get the # of file system overhead blocks from the
4058 * superblock if present.
4060 if (es
->s_overhead_clusters
)
4061 sbi
->s_overhead
= le32_to_cpu(es
->s_overhead_clusters
);
4063 err
= ext4_calculate_overhead(sb
);
4065 goto failed_mount_wq
;
4069 * The maximum number of concurrent works can be high and
4070 * concurrency isn't really necessary. Limit it to 1.
4072 EXT4_SB(sb
)->rsv_conversion_wq
=
4073 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM
| WQ_UNBOUND
, 1);
4074 if (!EXT4_SB(sb
)->rsv_conversion_wq
) {
4075 printk(KERN_ERR
"EXT4-fs: failed to create workqueue\n");
4081 * The jbd2_journal_load will have done any necessary log recovery,
4082 * so we can safely mount the rest of the filesystem now.
4085 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
4087 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
4088 ret
= PTR_ERR(root
);
4092 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
4093 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
4097 sb
->s_root
= d_make_root(root
);
4099 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
4104 if (ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
))
4105 sb
->s_flags
|= MS_RDONLY
;
4107 /* determine the minimum size of new large inodes, if present */
4108 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
4109 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
4110 EXT4_GOOD_OLD_INODE_SIZE
;
4111 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
4112 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE
)) {
4113 if (sbi
->s_want_extra_isize
<
4114 le16_to_cpu(es
->s_want_extra_isize
))
4115 sbi
->s_want_extra_isize
=
4116 le16_to_cpu(es
->s_want_extra_isize
);
4117 if (sbi
->s_want_extra_isize
<
4118 le16_to_cpu(es
->s_min_extra_isize
))
4119 sbi
->s_want_extra_isize
=
4120 le16_to_cpu(es
->s_min_extra_isize
);
4123 /* Check if enough inode space is available */
4124 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
4125 sbi
->s_inode_size
) {
4126 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
4127 EXT4_GOOD_OLD_INODE_SIZE
;
4128 ext4_msg(sb
, KERN_INFO
, "required extra inode space not"
4132 err
= ext4_reserve_clusters(sbi
, ext4_calculate_resv_clusters(sb
));
4134 ext4_msg(sb
, KERN_ERR
, "failed to reserve %llu clusters for "
4135 "reserved pool", ext4_calculate_resv_clusters(sb
));
4136 goto failed_mount4a
;
4139 err
= ext4_setup_system_zone(sb
);
4141 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
4143 goto failed_mount4a
;
4147 err
= ext4_mb_init(sb
);
4149 ext4_msg(sb
, KERN_ERR
, "failed to initialize mballoc (%d)",
4154 block
= ext4_count_free_clusters(sb
);
4155 ext4_free_blocks_count_set(sbi
->s_es
,
4156 EXT4_C2B(sbi
, block
));
4157 err
= percpu_counter_init(&sbi
->s_freeclusters_counter
, block
,
4160 unsigned long freei
= ext4_count_free_inodes(sb
);
4161 sbi
->s_es
->s_free_inodes_count
= cpu_to_le32(freei
);
4162 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
, freei
,
4166 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
4167 ext4_count_dirs(sb
), GFP_KERNEL
);
4169 err
= percpu_counter_init(&sbi
->s_dirtyclusters_counter
, 0,
4172 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
4176 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
4177 if (!ext4_fill_flex_info(sb
)) {
4178 ext4_msg(sb
, KERN_ERR
,
4179 "unable to initialize "
4180 "flex_bg meta info!");
4184 err
= ext4_register_li_request(sb
, first_not_zeroed
);
4188 sbi
->s_kobj
.kset
= ext4_kset
;
4189 init_completion(&sbi
->s_kobj_unregister
);
4190 err
= kobject_init_and_add(&sbi
->s_kobj
, &ext4_ktype
, NULL
,
4196 /* Enable quota usage during mount. */
4197 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
) &&
4198 !(sb
->s_flags
& MS_RDONLY
)) {
4199 err
= ext4_enable_quotas(sb
);
4203 #endif /* CONFIG_QUOTA */
4205 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
4206 ext4_orphan_cleanup(sb
, es
);
4207 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
4208 if (needs_recovery
) {
4209 ext4_msg(sb
, KERN_INFO
, "recovery complete");
4210 ext4_mark_recovery_complete(sb
, es
);
4212 if (EXT4_SB(sb
)->s_journal
) {
4213 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
4214 descr
= " journalled data mode";
4215 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
4216 descr
= " ordered data mode";
4218 descr
= " writeback data mode";
4220 descr
= "out journal";
4222 if (test_opt(sb
, DISCARD
)) {
4223 struct request_queue
*q
= bdev_get_queue(sb
->s_bdev
);
4224 if (!blk_queue_discard(q
))
4225 ext4_msg(sb
, KERN_WARNING
,
4226 "mounting with \"discard\" option, but "
4227 "the device does not support discard");
4230 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s. "
4231 "Opts: %s%s%s", descr
, sbi
->s_es
->s_mount_opts
,
4232 *sbi
->s_es
->s_mount_opts
? "; " : "", orig_data
);
4234 if (es
->s_error_count
)
4235 mod_timer(&sbi
->s_err_report
, jiffies
+ 300*HZ
); /* 5 minutes */
4237 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4238 ratelimit_state_init(&sbi
->s_err_ratelimit_state
, 5 * HZ
, 10);
4239 ratelimit_state_init(&sbi
->s_warning_ratelimit_state
, 5 * HZ
, 10);
4240 ratelimit_state_init(&sbi
->s_msg_ratelimit_state
, 5 * HZ
, 10);
4247 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
4252 kobject_del(&sbi
->s_kobj
);
4255 ext4_unregister_li_request(sb
);
4257 ext4_mb_release(sb
);
4258 if (sbi
->s_flex_groups
)
4259 kvfree(sbi
->s_flex_groups
);
4260 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
4261 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
4262 percpu_counter_destroy(&sbi
->s_dirs_counter
);
4263 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
4265 ext4_ext_release(sb
);
4266 ext4_release_system_zone(sb
);
4271 ext4_msg(sb
, KERN_ERR
, "mount failed");
4272 if (EXT4_SB(sb
)->rsv_conversion_wq
)
4273 destroy_workqueue(EXT4_SB(sb
)->rsv_conversion_wq
);
4275 if (sbi
->s_journal
) {
4276 jbd2_journal_destroy(sbi
->s_journal
);
4277 sbi
->s_journal
= NULL
;
4280 ext4_es_unregister_shrinker(sbi
);
4282 del_timer_sync(&sbi
->s_err_report
);
4284 kthread_stop(sbi
->s_mmp_tsk
);
4286 for (i
= 0; i
< db_count
; i
++)
4287 brelse(sbi
->s_group_desc
[i
]);
4288 kvfree(sbi
->s_group_desc
);
4290 if (sbi
->s_chksum_driver
)
4291 crypto_free_shash(sbi
->s_chksum_driver
);
4293 remove_proc_entry("options", sbi
->s_proc
);
4294 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
4297 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
4298 kfree(sbi
->s_qf_names
[i
]);
4300 ext4_blkdev_remove(sbi
);
4303 sb
->s_fs_info
= NULL
;
4304 kfree(sbi
->s_blockgroup_lock
);
4308 return err
? err
: ret
;
4312 * Setup any per-fs journal parameters now. We'll do this both on
4313 * initial mount, once the journal has been initialised but before we've
4314 * done any recovery; and again on any subsequent remount.
4316 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
4318 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4320 journal
->j_commit_interval
= sbi
->s_commit_interval
;
4321 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
4322 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
4324 write_lock(&journal
->j_state_lock
);
4325 if (test_opt(sb
, BARRIER
))
4326 journal
->j_flags
|= JBD2_BARRIER
;
4328 journal
->j_flags
&= ~JBD2_BARRIER
;
4329 if (test_opt(sb
, DATA_ERR_ABORT
))
4330 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
4332 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
4333 write_unlock(&journal
->j_state_lock
);
4336 static journal_t
*ext4_get_journal(struct super_block
*sb
,
4337 unsigned int journal_inum
)
4339 struct inode
*journal_inode
;
4342 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4344 /* First, test for the existence of a valid inode on disk. Bad
4345 * things happen if we iget() an unused inode, as the subsequent
4346 * iput() will try to delete it. */
4348 journal_inode
= ext4_iget(sb
, journal_inum
);
4349 if (IS_ERR(journal_inode
)) {
4350 ext4_msg(sb
, KERN_ERR
, "no journal found");
4353 if (!journal_inode
->i_nlink
) {
4354 make_bad_inode(journal_inode
);
4355 iput(journal_inode
);
4356 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
4360 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4361 journal_inode
, journal_inode
->i_size
);
4362 if (!S_ISREG(journal_inode
->i_mode
)) {
4363 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
4364 iput(journal_inode
);
4368 journal
= jbd2_journal_init_inode(journal_inode
);
4370 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
4371 iput(journal_inode
);
4374 journal
->j_private
= sb
;
4375 ext4_init_journal_params(sb
, journal
);
4379 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
4382 struct buffer_head
*bh
;
4386 int hblock
, blocksize
;
4387 ext4_fsblk_t sb_block
;
4388 unsigned long offset
;
4389 struct ext4_super_block
*es
;
4390 struct block_device
*bdev
;
4392 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4394 bdev
= ext4_blkdev_get(j_dev
, sb
);
4398 blocksize
= sb
->s_blocksize
;
4399 hblock
= bdev_logical_block_size(bdev
);
4400 if (blocksize
< hblock
) {
4401 ext4_msg(sb
, KERN_ERR
,
4402 "blocksize too small for journal device");
4406 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
4407 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
4408 set_blocksize(bdev
, blocksize
);
4409 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
4410 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
4411 "external journal");
4415 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
4416 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
4417 !(le32_to_cpu(es
->s_feature_incompat
) &
4418 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
4419 ext4_msg(sb
, KERN_ERR
, "external journal has "
4425 if ((le32_to_cpu(es
->s_feature_ro_compat
) &
4426 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
) &&
4427 es
->s_checksum
!= ext4_superblock_csum(sb
, es
)) {
4428 ext4_msg(sb
, KERN_ERR
, "external journal has "
4429 "corrupt superblock");
4434 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
4435 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
4440 len
= ext4_blocks_count(es
);
4441 start
= sb_block
+ 1;
4442 brelse(bh
); /* we're done with the superblock */
4444 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
4445 start
, len
, blocksize
);
4447 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
4450 journal
->j_private
= sb
;
4451 ll_rw_block(READ
| REQ_META
| REQ_PRIO
, 1, &journal
->j_sb_buffer
);
4452 wait_on_buffer(journal
->j_sb_buffer
);
4453 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
4454 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
4457 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
4458 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
4459 "user (unsupported) - %d",
4460 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
4463 EXT4_SB(sb
)->journal_bdev
= bdev
;
4464 ext4_init_journal_params(sb
, journal
);
4468 jbd2_journal_destroy(journal
);
4470 ext4_blkdev_put(bdev
);
4474 static int ext4_load_journal(struct super_block
*sb
,
4475 struct ext4_super_block
*es
,
4476 unsigned long journal_devnum
)
4479 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
4482 int really_read_only
;
4484 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4486 if (journal_devnum
&&
4487 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4488 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
4489 "numbers have changed");
4490 journal_dev
= new_decode_dev(journal_devnum
);
4492 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
4494 really_read_only
= bdev_read_only(sb
->s_bdev
);
4497 * Are we loading a blank journal or performing recovery after a
4498 * crash? For recovery, we need to check in advance whether we
4499 * can get read-write access to the device.
4501 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
4502 if (sb
->s_flags
& MS_RDONLY
) {
4503 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
4504 "required on readonly filesystem");
4505 if (really_read_only
) {
4506 ext4_msg(sb
, KERN_ERR
, "write access "
4507 "unavailable, cannot proceed");
4510 ext4_msg(sb
, KERN_INFO
, "write access will "
4511 "be enabled during recovery");
4515 if (journal_inum
&& journal_dev
) {
4516 ext4_msg(sb
, KERN_ERR
, "filesystem has both journal "
4517 "and inode journals!");
4522 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
4525 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
4529 if (!(journal
->j_flags
& JBD2_BARRIER
))
4530 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
4532 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
))
4533 err
= jbd2_journal_wipe(journal
, !really_read_only
);
4535 char *save
= kmalloc(EXT4_S_ERR_LEN
, GFP_KERNEL
);
4537 memcpy(save
, ((char *) es
) +
4538 EXT4_S_ERR_START
, EXT4_S_ERR_LEN
);
4539 err
= jbd2_journal_load(journal
);
4541 memcpy(((char *) es
) + EXT4_S_ERR_START
,
4542 save
, EXT4_S_ERR_LEN
);
4547 ext4_msg(sb
, KERN_ERR
, "error loading journal");
4548 jbd2_journal_destroy(journal
);
4552 EXT4_SB(sb
)->s_journal
= journal
;
4553 ext4_clear_journal_err(sb
, es
);
4555 if (!really_read_only
&& journal_devnum
&&
4556 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4557 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
4559 /* Make sure we flush the recovery flag to disk. */
4560 ext4_commit_super(sb
, 1);
4566 static int ext4_commit_super(struct super_block
*sb
, int sync
)
4568 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
4569 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
4572 if (!sbh
|| block_device_ejected(sb
))
4574 if (buffer_write_io_error(sbh
)) {
4576 * Oh, dear. A previous attempt to write the
4577 * superblock failed. This could happen because the
4578 * USB device was yanked out. Or it could happen to
4579 * be a transient write error and maybe the block will
4580 * be remapped. Nothing we can do but to retry the
4581 * write and hope for the best.
4583 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
4584 "superblock detected");
4585 clear_buffer_write_io_error(sbh
);
4586 set_buffer_uptodate(sbh
);
4589 * If the file system is mounted read-only, don't update the
4590 * superblock write time. This avoids updating the superblock
4591 * write time when we are mounting the root file system
4592 * read/only but we need to replay the journal; at that point,
4593 * for people who are east of GMT and who make their clock
4594 * tick in localtime for Windows bug-for-bug compatibility,
4595 * the clock is set in the future, and this will cause e2fsck
4596 * to complain and force a full file system check.
4598 if (!(sb
->s_flags
& MS_RDONLY
))
4599 es
->s_wtime
= cpu_to_le32(get_seconds());
4600 if (sb
->s_bdev
->bd_part
)
4601 es
->s_kbytes_written
=
4602 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
+
4603 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
4604 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
4606 es
->s_kbytes_written
=
4607 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
);
4608 if (percpu_counter_initialized(&EXT4_SB(sb
)->s_freeclusters_counter
))
4609 ext4_free_blocks_count_set(es
,
4610 EXT4_C2B(EXT4_SB(sb
), percpu_counter_sum_positive(
4611 &EXT4_SB(sb
)->s_freeclusters_counter
)));
4612 if (percpu_counter_initialized(&EXT4_SB(sb
)->s_freeinodes_counter
))
4613 es
->s_free_inodes_count
=
4614 cpu_to_le32(percpu_counter_sum_positive(
4615 &EXT4_SB(sb
)->s_freeinodes_counter
));
4616 BUFFER_TRACE(sbh
, "marking dirty");
4617 ext4_superblock_csum_set(sb
);
4618 mark_buffer_dirty(sbh
);
4620 error
= sync_dirty_buffer(sbh
);
4624 error
= buffer_write_io_error(sbh
);
4626 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
4628 clear_buffer_write_io_error(sbh
);
4629 set_buffer_uptodate(sbh
);
4636 * Have we just finished recovery? If so, and if we are mounting (or
4637 * remounting) the filesystem readonly, then we will end up with a
4638 * consistent fs on disk. Record that fact.
4640 static void ext4_mark_recovery_complete(struct super_block
*sb
,
4641 struct ext4_super_block
*es
)
4643 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
4645 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
4646 BUG_ON(journal
!= NULL
);
4649 jbd2_journal_lock_updates(journal
);
4650 if (jbd2_journal_flush(journal
) < 0)
4653 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
) &&
4654 sb
->s_flags
& MS_RDONLY
) {
4655 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4656 ext4_commit_super(sb
, 1);
4660 jbd2_journal_unlock_updates(journal
);
4664 * If we are mounting (or read-write remounting) a filesystem whose journal
4665 * has recorded an error from a previous lifetime, move that error to the
4666 * main filesystem now.
4668 static void ext4_clear_journal_err(struct super_block
*sb
,
4669 struct ext4_super_block
*es
)
4675 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4677 journal
= EXT4_SB(sb
)->s_journal
;
4680 * Now check for any error status which may have been recorded in the
4681 * journal by a prior ext4_error() or ext4_abort()
4684 j_errno
= jbd2_journal_errno(journal
);
4688 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
4689 ext4_warning(sb
, "Filesystem error recorded "
4690 "from previous mount: %s", errstr
);
4691 ext4_warning(sb
, "Marking fs in need of filesystem check.");
4693 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
4694 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
4695 ext4_commit_super(sb
, 1);
4697 jbd2_journal_clear_err(journal
);
4698 jbd2_journal_update_sb_errno(journal
);
4703 * Force the running and committing transactions to commit,
4704 * and wait on the commit.
4706 int ext4_force_commit(struct super_block
*sb
)
4710 if (sb
->s_flags
& MS_RDONLY
)
4713 journal
= EXT4_SB(sb
)->s_journal
;
4714 return ext4_journal_force_commit(journal
);
4717 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
4721 bool needs_barrier
= false;
4722 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4724 trace_ext4_sync_fs(sb
, wait
);
4725 flush_workqueue(sbi
->rsv_conversion_wq
);
4727 * Writeback quota in non-journalled quota case - journalled quota has
4730 dquot_writeback_dquots(sb
, -1);
4732 * Data writeback is possible w/o journal transaction, so barrier must
4733 * being sent at the end of the function. But we can skip it if
4734 * transaction_commit will do it for us.
4736 if (sbi
->s_journal
) {
4737 target
= jbd2_get_latest_transaction(sbi
->s_journal
);
4738 if (wait
&& sbi
->s_journal
->j_flags
& JBD2_BARRIER
&&
4739 !jbd2_trans_will_send_data_barrier(sbi
->s_journal
, target
))
4740 needs_barrier
= true;
4742 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
4744 ret
= jbd2_log_wait_commit(sbi
->s_journal
,
4747 } else if (wait
&& test_opt(sb
, BARRIER
))
4748 needs_barrier
= true;
4749 if (needs_barrier
) {
4751 err
= blkdev_issue_flush(sb
->s_bdev
, GFP_KERNEL
, NULL
);
4760 * LVM calls this function before a (read-only) snapshot is created. This
4761 * gives us a chance to flush the journal completely and mark the fs clean.
4763 * Note that only this function cannot bring a filesystem to be in a clean
4764 * state independently. It relies on upper layer to stop all data & metadata
4767 static int ext4_freeze(struct super_block
*sb
)
4772 if (sb
->s_flags
& MS_RDONLY
)
4775 journal
= EXT4_SB(sb
)->s_journal
;
4778 /* Now we set up the journal barrier. */
4779 jbd2_journal_lock_updates(journal
);
4782 * Don't clear the needs_recovery flag if we failed to
4783 * flush the journal.
4785 error
= jbd2_journal_flush(journal
);
4790 /* Journal blocked and flushed, clear needs_recovery flag. */
4791 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4792 error
= ext4_commit_super(sb
, 1);
4795 /* we rely on upper layer to stop further updates */
4796 jbd2_journal_unlock_updates(journal
);
4801 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4802 * flag here, even though the filesystem is not technically dirty yet.
4804 static int ext4_unfreeze(struct super_block
*sb
)
4806 if (sb
->s_flags
& MS_RDONLY
)
4809 /* Reset the needs_recovery flag before the fs is unlocked. */
4810 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4811 ext4_commit_super(sb
, 1);
4816 * Structure to save mount options for ext4_remount's benefit
4818 struct ext4_mount_options
{
4819 unsigned long s_mount_opt
;
4820 unsigned long s_mount_opt2
;
4823 unsigned long s_commit_interval
;
4824 u32 s_min_batch_time
, s_max_batch_time
;
4827 char *s_qf_names
[EXT4_MAXQUOTAS
];
4831 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
4833 struct ext4_super_block
*es
;
4834 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4835 unsigned long old_sb_flags
;
4836 struct ext4_mount_options old_opts
;
4837 int enable_quota
= 0;
4839 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
4844 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
4846 /* Store the original options */
4847 old_sb_flags
= sb
->s_flags
;
4848 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
4849 old_opts
.s_mount_opt2
= sbi
->s_mount_opt2
;
4850 old_opts
.s_resuid
= sbi
->s_resuid
;
4851 old_opts
.s_resgid
= sbi
->s_resgid
;
4852 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
4853 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
4854 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
4856 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
4857 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
4858 if (sbi
->s_qf_names
[i
]) {
4859 old_opts
.s_qf_names
[i
] = kstrdup(sbi
->s_qf_names
[i
],
4861 if (!old_opts
.s_qf_names
[i
]) {
4862 for (j
= 0; j
< i
; j
++)
4863 kfree(old_opts
.s_qf_names
[j
]);
4868 old_opts
.s_qf_names
[i
] = NULL
;
4870 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
4871 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
4873 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
, 1)) {
4878 if ((old_opts
.s_mount_opt
& EXT4_MOUNT_JOURNAL_CHECKSUM
) ^
4879 test_opt(sb
, JOURNAL_CHECKSUM
)) {
4880 ext4_msg(sb
, KERN_ERR
, "changing journal_checksum "
4881 "during remount not supported; ignoring");
4882 sbi
->s_mount_opt
^= EXT4_MOUNT_JOURNAL_CHECKSUM
;
4885 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
4886 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
4887 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4888 "both data=journal and delalloc");
4892 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
4893 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4894 "both data=journal and dioread_nolock");
4898 if (test_opt(sb
, DAX
)) {
4899 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4900 "both data=journal and dax");
4906 if ((sbi
->s_mount_opt
^ old_opts
.s_mount_opt
) & EXT4_MOUNT_DAX
) {
4907 ext4_msg(sb
, KERN_WARNING
, "warning: refusing change of "
4908 "dax flag with busy inodes while remounting");
4909 sbi
->s_mount_opt
^= EXT4_MOUNT_DAX
;
4912 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
)
4913 ext4_abort(sb
, "Abort forced by user");
4915 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
4916 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
4920 if (sbi
->s_journal
) {
4921 ext4_init_journal_params(sb
, sbi
->s_journal
);
4922 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
4925 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
)) {
4926 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
) {
4931 if (*flags
& MS_RDONLY
) {
4932 err
= sync_filesystem(sb
);
4935 err
= dquot_suspend(sb
, -1);
4940 * First of all, the unconditional stuff we have to do
4941 * to disable replay of the journal when we next remount
4943 sb
->s_flags
|= MS_RDONLY
;
4946 * OK, test if we are remounting a valid rw partition
4947 * readonly, and if so set the rdonly flag and then
4948 * mark the partition as valid again.
4950 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
4951 (sbi
->s_mount_state
& EXT4_VALID_FS
))
4952 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
4955 ext4_mark_recovery_complete(sb
, es
);
4957 /* Make sure we can mount this feature set readwrite */
4958 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
4959 EXT4_FEATURE_RO_COMPAT_READONLY
) ||
4960 !ext4_feature_set_ok(sb
, 0)) {
4965 * Make sure the group descriptor checksums
4966 * are sane. If they aren't, refuse to remount r/w.
4968 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
4969 struct ext4_group_desc
*gdp
=
4970 ext4_get_group_desc(sb
, g
, NULL
);
4972 if (!ext4_group_desc_csum_verify(sb
, g
, gdp
)) {
4973 ext4_msg(sb
, KERN_ERR
,
4974 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4975 g
, le16_to_cpu(ext4_group_desc_csum(sbi
, g
, gdp
)),
4976 le16_to_cpu(gdp
->bg_checksum
));
4983 * If we have an unprocessed orphan list hanging
4984 * around from a previously readonly bdev mount,
4985 * require a full umount/remount for now.
4987 if (es
->s_last_orphan
) {
4988 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
4989 "remount RDWR because of unprocessed "
4990 "orphan inode list. Please "
4991 "umount/remount instead");
4997 * Mounting a RDONLY partition read-write, so reread
4998 * and store the current valid flag. (It may have
4999 * been changed by e2fsck since we originally mounted
5003 ext4_clear_journal_err(sb
, es
);
5004 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
5005 if (!ext4_setup_super(sb
, es
, 0))
5006 sb
->s_flags
&= ~MS_RDONLY
;
5007 if (EXT4_HAS_INCOMPAT_FEATURE(sb
,
5008 EXT4_FEATURE_INCOMPAT_MMP
))
5009 if (ext4_multi_mount_protect(sb
,
5010 le64_to_cpu(es
->s_mmp_block
))) {
5019 * Reinitialize lazy itable initialization thread based on
5022 if ((sb
->s_flags
& MS_RDONLY
) || !test_opt(sb
, INIT_INODE_TABLE
))
5023 ext4_unregister_li_request(sb
);
5025 ext4_group_t first_not_zeroed
;
5026 first_not_zeroed
= ext4_has_uninit_itable(sb
);
5027 ext4_register_li_request(sb
, first_not_zeroed
);
5030 ext4_setup_system_zone(sb
);
5031 if (sbi
->s_journal
== NULL
&& !(old_sb_flags
& MS_RDONLY
))
5032 ext4_commit_super(sb
, 1);
5035 /* Release old quota file names */
5036 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
5037 kfree(old_opts
.s_qf_names
[i
]);
5039 if (sb_any_quota_suspended(sb
))
5040 dquot_resume(sb
, -1);
5041 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
5042 EXT4_FEATURE_RO_COMPAT_QUOTA
)) {
5043 err
= ext4_enable_quotas(sb
);
5050 *flags
= (*flags
& ~MS_LAZYTIME
) | (sb
->s_flags
& MS_LAZYTIME
);
5051 ext4_msg(sb
, KERN_INFO
, "re-mounted. Opts: %s", orig_data
);
5056 sb
->s_flags
= old_sb_flags
;
5057 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
5058 sbi
->s_mount_opt2
= old_opts
.s_mount_opt2
;
5059 sbi
->s_resuid
= old_opts
.s_resuid
;
5060 sbi
->s_resgid
= old_opts
.s_resgid
;
5061 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
5062 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
5063 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
5065 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
5066 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
5067 kfree(sbi
->s_qf_names
[i
]);
5068 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
5075 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
5077 struct super_block
*sb
= dentry
->d_sb
;
5078 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
5079 struct ext4_super_block
*es
= sbi
->s_es
;
5080 ext4_fsblk_t overhead
= 0, resv_blocks
;
5083 resv_blocks
= EXT4_C2B(sbi
, atomic64_read(&sbi
->s_resv_clusters
));
5085 if (!test_opt(sb
, MINIX_DF
))
5086 overhead
= sbi
->s_overhead
;
5088 buf
->f_type
= EXT4_SUPER_MAGIC
;
5089 buf
->f_bsize
= sb
->s_blocksize
;
5090 buf
->f_blocks
= ext4_blocks_count(es
) - EXT4_C2B(sbi
, overhead
);
5091 bfree
= percpu_counter_sum_positive(&sbi
->s_freeclusters_counter
) -
5092 percpu_counter_sum_positive(&sbi
->s_dirtyclusters_counter
);
5093 /* prevent underflow in case that few free space is available */
5094 buf
->f_bfree
= EXT4_C2B(sbi
, max_t(s64
, bfree
, 0));
5095 buf
->f_bavail
= buf
->f_bfree
-
5096 (ext4_r_blocks_count(es
) + resv_blocks
);
5097 if (buf
->f_bfree
< (ext4_r_blocks_count(es
) + resv_blocks
))
5099 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
5100 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
5101 buf
->f_namelen
= EXT4_NAME_LEN
;
5102 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
5103 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
5104 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
5105 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
5110 /* Helper function for writing quotas on sync - we need to start transaction
5111 * before quota file is locked for write. Otherwise the are possible deadlocks:
5112 * Process 1 Process 2
5113 * ext4_create() quota_sync()
5114 * jbd2_journal_start() write_dquot()
5115 * dquot_initialize() down(dqio_mutex)
5116 * down(dqio_mutex) jbd2_journal_start()
5122 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
5124 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_id
.type
];
5127 static int ext4_write_dquot(struct dquot
*dquot
)
5131 struct inode
*inode
;
5133 inode
= dquot_to_inode(dquot
);
5134 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
,
5135 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
5137 return PTR_ERR(handle
);
5138 ret
= dquot_commit(dquot
);
5139 err
= ext4_journal_stop(handle
);
5145 static int ext4_acquire_dquot(struct dquot
*dquot
)
5150 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
5151 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
5153 return PTR_ERR(handle
);
5154 ret
= dquot_acquire(dquot
);
5155 err
= ext4_journal_stop(handle
);
5161 static int ext4_release_dquot(struct dquot
*dquot
)
5166 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
5167 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
5168 if (IS_ERR(handle
)) {
5169 /* Release dquot anyway to avoid endless cycle in dqput() */
5170 dquot_release(dquot
);
5171 return PTR_ERR(handle
);
5173 ret
= dquot_release(dquot
);
5174 err
= ext4_journal_stop(handle
);
5180 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
5182 struct super_block
*sb
= dquot
->dq_sb
;
5183 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
5185 /* Are we journaling quotas? */
5186 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
) ||
5187 sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
5188 dquot_mark_dquot_dirty(dquot
);
5189 return ext4_write_dquot(dquot
);
5191 return dquot_mark_dquot_dirty(dquot
);
5195 static int ext4_write_info(struct super_block
*sb
, int type
)
5200 /* Data block + inode block */
5201 handle
= ext4_journal_start(sb
->s_root
->d_inode
, EXT4_HT_QUOTA
, 2);
5203 return PTR_ERR(handle
);
5204 ret
= dquot_commit_info(sb
, type
);
5205 err
= ext4_journal_stop(handle
);
5212 * Turn on quotas during mount time - we need to find
5213 * the quota file and such...
5215 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
5217 return dquot_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
5218 EXT4_SB(sb
)->s_jquota_fmt
, type
);
5222 * Standard function to be called on quota_on
5224 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
5229 if (!test_opt(sb
, QUOTA
))
5232 /* Quotafile not on the same filesystem? */
5233 if (path
->dentry
->d_sb
!= sb
)
5235 /* Journaling quota? */
5236 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
5237 /* Quotafile not in fs root? */
5238 if (path
->dentry
->d_parent
!= sb
->s_root
)
5239 ext4_msg(sb
, KERN_WARNING
,
5240 "Quota file not on filesystem root. "
5241 "Journaled quota will not work");
5245 * When we journal data on quota file, we have to flush journal to see
5246 * all updates to the file when we bypass pagecache...
5248 if (EXT4_SB(sb
)->s_journal
&&
5249 ext4_should_journal_data(path
->dentry
->d_inode
)) {
5251 * We don't need to lock updates but journal_flush() could
5252 * otherwise be livelocked...
5254 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
5255 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
5256 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
5261 return dquot_quota_on(sb
, type
, format_id
, path
);
5264 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
5268 struct inode
*qf_inode
;
5269 unsigned long qf_inums
[EXT4_MAXQUOTAS
] = {
5270 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
5271 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
)
5274 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
));
5276 if (!qf_inums
[type
])
5279 qf_inode
= ext4_iget(sb
, qf_inums
[type
]);
5280 if (IS_ERR(qf_inode
)) {
5281 ext4_error(sb
, "Bad quota inode # %lu", qf_inums
[type
]);
5282 return PTR_ERR(qf_inode
);
5285 /* Don't account quota for quota files to avoid recursion */
5286 qf_inode
->i_flags
|= S_NOQUOTA
;
5287 err
= dquot_enable(qf_inode
, type
, format_id
, flags
);
5293 /* Enable usage tracking for all quota types. */
5294 static int ext4_enable_quotas(struct super_block
*sb
)
5297 unsigned long qf_inums
[EXT4_MAXQUOTAS
] = {
5298 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
5299 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
)
5302 sb_dqopt(sb
)->flags
|= DQUOT_QUOTA_SYS_FILE
;
5303 for (type
= 0; type
< EXT4_MAXQUOTAS
; type
++) {
5304 if (qf_inums
[type
]) {
5305 err
= ext4_quota_enable(sb
, type
, QFMT_VFS_V1
,
5306 DQUOT_USAGE_ENABLED
);
5309 "Failed to enable quota tracking "
5310 "(type=%d, err=%d). Please run "
5311 "e2fsck to fix.", type
, err
);
5319 static int ext4_quota_off(struct super_block
*sb
, int type
)
5321 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5324 /* Force all delayed allocation blocks to be allocated.
5325 * Caller already holds s_umount sem */
5326 if (test_opt(sb
, DELALLOC
))
5327 sync_filesystem(sb
);
5332 /* Update modification times of quota files when userspace can
5333 * start looking at them */
5334 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
, 1);
5337 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
5338 ext4_mark_inode_dirty(handle
, inode
);
5339 ext4_journal_stop(handle
);
5342 return dquot_quota_off(sb
, type
);
5345 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5346 * acquiring the locks... As quota files are never truncated and quota code
5347 * itself serializes the operations (and no one else should touch the files)
5348 * we don't have to be afraid of races */
5349 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
5350 size_t len
, loff_t off
)
5352 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5353 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5354 int offset
= off
& (sb
->s_blocksize
- 1);
5357 struct buffer_head
*bh
;
5358 loff_t i_size
= i_size_read(inode
);
5362 if (off
+len
> i_size
)
5365 while (toread
> 0) {
5366 tocopy
= sb
->s_blocksize
- offset
< toread
?
5367 sb
->s_blocksize
- offset
: toread
;
5368 bh
= ext4_bread(NULL
, inode
, blk
, 0);
5371 if (!bh
) /* A hole? */
5372 memset(data
, 0, tocopy
);
5374 memcpy(data
, bh
->b_data
+offset
, tocopy
);
5384 /* Write to quotafile (we know the transaction is already started and has
5385 * enough credits) */
5386 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
5387 const char *data
, size_t len
, loff_t off
)
5389 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5390 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5391 int err
, offset
= off
& (sb
->s_blocksize
- 1);
5392 struct buffer_head
*bh
;
5393 handle_t
*handle
= journal_current_handle();
5395 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
5396 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5397 " cancelled because transaction is not started",
5398 (unsigned long long)off
, (unsigned long long)len
);
5402 * Since we account only one data block in transaction credits,
5403 * then it is impossible to cross a block boundary.
5405 if (sb
->s_blocksize
- offset
< len
) {
5406 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5407 " cancelled because not block aligned",
5408 (unsigned long long)off
, (unsigned long long)len
);
5412 bh
= ext4_bread(handle
, inode
, blk
, 1);
5417 BUFFER_TRACE(bh
, "get write access");
5418 err
= ext4_journal_get_write_access(handle
, bh
);
5424 memcpy(bh
->b_data
+offset
, data
, len
);
5425 flush_dcache_page(bh
->b_page
);
5427 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
5430 if (inode
->i_size
< off
+ len
) {
5431 i_size_write(inode
, off
+ len
);
5432 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
5433 ext4_mark_inode_dirty(handle
, inode
);
5440 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
5441 const char *dev_name
, void *data
)
5443 return mount_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
);
5446 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5447 static inline void register_as_ext2(void)
5449 int err
= register_filesystem(&ext2_fs_type
);
5452 "EXT4-fs: Unable to register as ext2 (%d)\n", err
);
5455 static inline void unregister_as_ext2(void)
5457 unregister_filesystem(&ext2_fs_type
);
5460 static inline int ext2_feature_set_ok(struct super_block
*sb
)
5462 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT2_FEATURE_INCOMPAT_SUPP
))
5464 if (sb
->s_flags
& MS_RDONLY
)
5466 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT2_FEATURE_RO_COMPAT_SUPP
))
5471 static inline void register_as_ext2(void) { }
5472 static inline void unregister_as_ext2(void) { }
5473 static inline int ext2_feature_set_ok(struct super_block
*sb
) { return 0; }
5476 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5477 static inline void register_as_ext3(void)
5479 int err
= register_filesystem(&ext3_fs_type
);
5482 "EXT4-fs: Unable to register as ext3 (%d)\n", err
);
5485 static inline void unregister_as_ext3(void)
5487 unregister_filesystem(&ext3_fs_type
);
5490 static inline int ext3_feature_set_ok(struct super_block
*sb
)
5492 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT3_FEATURE_INCOMPAT_SUPP
))
5494 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
5496 if (sb
->s_flags
& MS_RDONLY
)
5498 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT3_FEATURE_RO_COMPAT_SUPP
))
5503 static inline void register_as_ext3(void) { }
5504 static inline void unregister_as_ext3(void) { }
5505 static inline int ext3_feature_set_ok(struct super_block
*sb
) { return 0; }
5508 static struct file_system_type ext4_fs_type
= {
5509 .owner
= THIS_MODULE
,
5511 .mount
= ext4_mount
,
5512 .kill_sb
= kill_block_super
,
5513 .fs_flags
= FS_REQUIRES_DEV
,
5515 MODULE_ALIAS_FS("ext4");
5517 static int __init
ext4_init_feat_adverts(void)
5519 struct ext4_features
*ef
;
5522 ef
= kzalloc(sizeof(struct ext4_features
), GFP_KERNEL
);
5526 ef
->f_kobj
.kset
= ext4_kset
;
5527 init_completion(&ef
->f_kobj_unregister
);
5528 ret
= kobject_init_and_add(&ef
->f_kobj
, &ext4_feat_ktype
, NULL
,
5541 static void ext4_exit_feat_adverts(void)
5543 kobject_put(&ext4_feat
->f_kobj
);
5544 wait_for_completion(&ext4_feat
->f_kobj_unregister
);
5548 /* Shared across all ext4 file systems */
5549 wait_queue_head_t ext4__ioend_wq
[EXT4_WQ_HASH_SZ
];
5550 struct mutex ext4__aio_mutex
[EXT4_WQ_HASH_SZ
];
5552 static int __init
ext4_init_fs(void)
5556 ext4_li_info
= NULL
;
5557 mutex_init(&ext4_li_mtx
);
5559 /* Build-time check for flags consistency */
5560 ext4_check_flag_values();
5562 for (i
= 0; i
< EXT4_WQ_HASH_SZ
; i
++) {
5563 mutex_init(&ext4__aio_mutex
[i
]);
5564 init_waitqueue_head(&ext4__ioend_wq
[i
]);
5567 err
= ext4_init_es();
5571 err
= ext4_init_pageio();
5575 err
= ext4_init_system_zone();
5578 ext4_kset
= kset_create_and_add("ext4", NULL
, fs_kobj
);
5583 ext4_proc_root
= proc_mkdir("fs/ext4", NULL
);
5585 err
= ext4_init_feat_adverts();
5589 err
= ext4_init_mballoc();
5593 ext4_mballoc_ready
= 1;
5594 err
= init_inodecache();
5599 err
= register_filesystem(&ext4_fs_type
);
5605 unregister_as_ext2();
5606 unregister_as_ext3();
5607 destroy_inodecache();
5609 ext4_mballoc_ready
= 0;
5610 ext4_exit_mballoc();
5612 ext4_exit_feat_adverts();
5615 remove_proc_entry("fs/ext4", NULL
);
5616 kset_unregister(ext4_kset
);
5618 ext4_exit_system_zone();
5627 static void __exit
ext4_exit_fs(void)
5629 ext4_destroy_lazyinit_thread();
5630 unregister_as_ext2();
5631 unregister_as_ext3();
5632 unregister_filesystem(&ext4_fs_type
);
5633 destroy_inodecache();
5634 ext4_exit_mballoc();
5635 ext4_exit_feat_adverts();
5636 remove_proc_entry("fs/ext4", NULL
);
5637 kset_unregister(ext4_kset
);
5638 ext4_exit_system_zone();
5643 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5644 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5645 MODULE_LICENSE("GPL");
5646 module_init(ext4_init_fs
)
5647 module_exit(ext4_exit_fs
)