2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/cleancache.h>
42 #include <asm/uaccess.h>
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
48 #include "ext4_extents.h" /* Needed for trace points definition */
49 #include "ext4_jbd2.h"
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
57 static struct proc_dir_entry
*ext4_proc_root
;
58 static struct kset
*ext4_kset
;
59 static struct ext4_lazy_init
*ext4_li_info
;
60 static struct mutex ext4_li_mtx
;
61 static struct ext4_features
*ext4_feat
;
62 static int ext4_mballoc_ready
;
64 static int ext4_load_journal(struct super_block
*, struct ext4_super_block
*,
65 unsigned long journal_devnum
);
66 static int ext4_show_options(struct seq_file
*seq
, struct dentry
*root
);
67 static int ext4_commit_super(struct super_block
*sb
, int sync
);
68 static void ext4_mark_recovery_complete(struct super_block
*sb
,
69 struct ext4_super_block
*es
);
70 static void ext4_clear_journal_err(struct super_block
*sb
,
71 struct ext4_super_block
*es
);
72 static int ext4_sync_fs(struct super_block
*sb
, int wait
);
73 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
);
74 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
75 static int ext4_unfreeze(struct super_block
*sb
);
76 static int ext4_freeze(struct super_block
*sb
);
77 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
78 const char *dev_name
, void *data
);
79 static inline int ext2_feature_set_ok(struct super_block
*sb
);
80 static inline int ext3_feature_set_ok(struct super_block
*sb
);
81 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
);
82 static void ext4_destroy_lazyinit_thread(void);
83 static void ext4_unregister_li_request(struct super_block
*sb
);
84 static void ext4_clear_request_list(void);
85 static int ext4_reserve_clusters(struct ext4_sb_info
*, ext4_fsblk_t
);
87 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88 static struct file_system_type ext2_fs_type
= {
92 .kill_sb
= kill_block_super
,
93 .fs_flags
= FS_REQUIRES_DEV
,
95 MODULE_ALIAS_FS("ext2");
97 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
99 #define IS_EXT2_SB(sb) (0)
103 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
104 static struct file_system_type ext3_fs_type
= {
105 .owner
= THIS_MODULE
,
108 .kill_sb
= kill_block_super
,
109 .fs_flags
= FS_REQUIRES_DEV
,
111 MODULE_ALIAS_FS("ext3");
112 MODULE_ALIAS("ext3");
113 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
115 #define IS_EXT3_SB(sb) (0)
118 static int ext4_verify_csum_type(struct super_block
*sb
,
119 struct ext4_super_block
*es
)
121 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
,
122 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
))
125 return es
->s_checksum_type
== EXT4_CRC32C_CHKSUM
;
128 static __le32
ext4_superblock_csum(struct super_block
*sb
,
129 struct ext4_super_block
*es
)
131 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
132 int offset
= offsetof(struct ext4_super_block
, s_checksum
);
135 csum
= ext4_chksum(sbi
, ~0, (char *)es
, offset
);
137 return cpu_to_le32(csum
);
140 static int ext4_superblock_csum_verify(struct super_block
*sb
,
141 struct ext4_super_block
*es
)
143 if (!ext4_has_metadata_csum(sb
))
146 return es
->s_checksum
== ext4_superblock_csum(sb
, es
);
149 void ext4_superblock_csum_set(struct super_block
*sb
)
151 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
153 if (!ext4_has_metadata_csum(sb
))
156 es
->s_checksum
= ext4_superblock_csum(sb
, es
);
159 void *ext4_kvmalloc(size_t size
, gfp_t flags
)
163 ret
= kmalloc(size
, flags
| __GFP_NOWARN
);
165 ret
= __vmalloc(size
, flags
, PAGE_KERNEL
);
169 void *ext4_kvzalloc(size_t size
, gfp_t flags
)
173 ret
= kzalloc(size
, flags
| __GFP_NOWARN
);
175 ret
= __vmalloc(size
, flags
| __GFP_ZERO
, PAGE_KERNEL
);
179 ext4_fsblk_t
ext4_block_bitmap(struct super_block
*sb
,
180 struct ext4_group_desc
*bg
)
182 return le32_to_cpu(bg
->bg_block_bitmap_lo
) |
183 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
184 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_block_bitmap_hi
) << 32 : 0);
187 ext4_fsblk_t
ext4_inode_bitmap(struct super_block
*sb
,
188 struct ext4_group_desc
*bg
)
190 return le32_to_cpu(bg
->bg_inode_bitmap_lo
) |
191 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
192 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_bitmap_hi
) << 32 : 0);
195 ext4_fsblk_t
ext4_inode_table(struct super_block
*sb
,
196 struct ext4_group_desc
*bg
)
198 return le32_to_cpu(bg
->bg_inode_table_lo
) |
199 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
200 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_table_hi
) << 32 : 0);
203 __u32
ext4_free_group_clusters(struct super_block
*sb
,
204 struct ext4_group_desc
*bg
)
206 return le16_to_cpu(bg
->bg_free_blocks_count_lo
) |
207 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
208 (__u32
)le16_to_cpu(bg
->bg_free_blocks_count_hi
) << 16 : 0);
211 __u32
ext4_free_inodes_count(struct super_block
*sb
,
212 struct ext4_group_desc
*bg
)
214 return le16_to_cpu(bg
->bg_free_inodes_count_lo
) |
215 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
216 (__u32
)le16_to_cpu(bg
->bg_free_inodes_count_hi
) << 16 : 0);
219 __u32
ext4_used_dirs_count(struct super_block
*sb
,
220 struct ext4_group_desc
*bg
)
222 return le16_to_cpu(bg
->bg_used_dirs_count_lo
) |
223 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
224 (__u32
)le16_to_cpu(bg
->bg_used_dirs_count_hi
) << 16 : 0);
227 __u32
ext4_itable_unused_count(struct super_block
*sb
,
228 struct ext4_group_desc
*bg
)
230 return le16_to_cpu(bg
->bg_itable_unused_lo
) |
231 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
232 (__u32
)le16_to_cpu(bg
->bg_itable_unused_hi
) << 16 : 0);
235 void ext4_block_bitmap_set(struct super_block
*sb
,
236 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
238 bg
->bg_block_bitmap_lo
= cpu_to_le32((u32
)blk
);
239 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
240 bg
->bg_block_bitmap_hi
= cpu_to_le32(blk
>> 32);
243 void ext4_inode_bitmap_set(struct super_block
*sb
,
244 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
246 bg
->bg_inode_bitmap_lo
= cpu_to_le32((u32
)blk
);
247 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
248 bg
->bg_inode_bitmap_hi
= cpu_to_le32(blk
>> 32);
251 void ext4_inode_table_set(struct super_block
*sb
,
252 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
254 bg
->bg_inode_table_lo
= cpu_to_le32((u32
)blk
);
255 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
256 bg
->bg_inode_table_hi
= cpu_to_le32(blk
>> 32);
259 void ext4_free_group_clusters_set(struct super_block
*sb
,
260 struct ext4_group_desc
*bg
, __u32 count
)
262 bg
->bg_free_blocks_count_lo
= cpu_to_le16((__u16
)count
);
263 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
264 bg
->bg_free_blocks_count_hi
= cpu_to_le16(count
>> 16);
267 void ext4_free_inodes_set(struct super_block
*sb
,
268 struct ext4_group_desc
*bg
, __u32 count
)
270 bg
->bg_free_inodes_count_lo
= cpu_to_le16((__u16
)count
);
271 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
272 bg
->bg_free_inodes_count_hi
= cpu_to_le16(count
>> 16);
275 void ext4_used_dirs_set(struct super_block
*sb
,
276 struct ext4_group_desc
*bg
, __u32 count
)
278 bg
->bg_used_dirs_count_lo
= cpu_to_le16((__u16
)count
);
279 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
280 bg
->bg_used_dirs_count_hi
= cpu_to_le16(count
>> 16);
283 void ext4_itable_unused_set(struct super_block
*sb
,
284 struct ext4_group_desc
*bg
, __u32 count
)
286 bg
->bg_itable_unused_lo
= cpu_to_le16((__u16
)count
);
287 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
288 bg
->bg_itable_unused_hi
= cpu_to_le16(count
>> 16);
292 static void __save_error_info(struct super_block
*sb
, const char *func
,
295 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
297 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
298 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
299 es
->s_last_error_time
= cpu_to_le32(get_seconds());
300 strncpy(es
->s_last_error_func
, func
, sizeof(es
->s_last_error_func
));
301 es
->s_last_error_line
= cpu_to_le32(line
);
302 if (!es
->s_first_error_time
) {
303 es
->s_first_error_time
= es
->s_last_error_time
;
304 strncpy(es
->s_first_error_func
, func
,
305 sizeof(es
->s_first_error_func
));
306 es
->s_first_error_line
= cpu_to_le32(line
);
307 es
->s_first_error_ino
= es
->s_last_error_ino
;
308 es
->s_first_error_block
= es
->s_last_error_block
;
311 * Start the daily error reporting function if it hasn't been
314 if (!es
->s_error_count
)
315 mod_timer(&EXT4_SB(sb
)->s_err_report
, jiffies
+ 24*60*60*HZ
);
316 le32_add_cpu(&es
->s_error_count
, 1);
319 static void save_error_info(struct super_block
*sb
, const char *func
,
322 __save_error_info(sb
, func
, line
);
323 ext4_commit_super(sb
, 1);
327 * The del_gendisk() function uninitializes the disk-specific data
328 * structures, including the bdi structure, without telling anyone
329 * else. Once this happens, any attempt to call mark_buffer_dirty()
330 * (for example, by ext4_commit_super), will cause a kernel OOPS.
331 * This is a kludge to prevent these oops until we can put in a proper
332 * hook in del_gendisk() to inform the VFS and file system layers.
334 static int block_device_ejected(struct super_block
*sb
)
336 struct inode
*bd_inode
= sb
->s_bdev
->bd_inode
;
337 struct backing_dev_info
*bdi
= inode_to_bdi(bd_inode
);
339 return bdi
->dev
== NULL
;
342 static void ext4_journal_commit_callback(journal_t
*journal
, transaction_t
*txn
)
344 struct super_block
*sb
= journal
->j_private
;
345 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
346 int error
= is_journal_aborted(journal
);
347 struct ext4_journal_cb_entry
*jce
;
349 BUG_ON(txn
->t_state
== T_FINISHED
);
350 spin_lock(&sbi
->s_md_lock
);
351 while (!list_empty(&txn
->t_private_list
)) {
352 jce
= list_entry(txn
->t_private_list
.next
,
353 struct ext4_journal_cb_entry
, jce_list
);
354 list_del_init(&jce
->jce_list
);
355 spin_unlock(&sbi
->s_md_lock
);
356 jce
->jce_func(sb
, jce
, error
);
357 spin_lock(&sbi
->s_md_lock
);
359 spin_unlock(&sbi
->s_md_lock
);
362 /* Deal with the reporting of failure conditions on a filesystem such as
363 * inconsistencies detected or read IO failures.
365 * On ext2, we can store the error state of the filesystem in the
366 * superblock. That is not possible on ext4, because we may have other
367 * write ordering constraints on the superblock which prevent us from
368 * writing it out straight away; and given that the journal is about to
369 * be aborted, we can't rely on the current, or future, transactions to
370 * write out the superblock safely.
372 * We'll just use the jbd2_journal_abort() error code to record an error in
373 * the journal instead. On recovery, the journal will complain about
374 * that error until we've noted it down and cleared it.
377 static void ext4_handle_error(struct super_block
*sb
)
379 if (sb
->s_flags
& MS_RDONLY
)
382 if (!test_opt(sb
, ERRORS_CONT
)) {
383 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
385 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
387 jbd2_journal_abort(journal
, -EIO
);
389 if (test_opt(sb
, ERRORS_RO
)) {
390 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
392 * Make sure updated value of ->s_mount_flags will be visible
393 * before ->s_flags update
396 sb
->s_flags
|= MS_RDONLY
;
398 if (test_opt(sb
, ERRORS_PANIC
))
399 panic("EXT4-fs (device %s): panic forced after error\n",
403 #define ext4_error_ratelimit(sb) \
404 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
407 void __ext4_error(struct super_block
*sb
, const char *function
,
408 unsigned int line
, const char *fmt
, ...)
410 struct va_format vaf
;
413 if (ext4_error_ratelimit(sb
)) {
418 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
419 sb
->s_id
, function
, line
, current
->comm
, &vaf
);
422 save_error_info(sb
, function
, line
);
423 ext4_handle_error(sb
);
426 void __ext4_error_inode(struct inode
*inode
, const char *function
,
427 unsigned int line
, ext4_fsblk_t block
,
428 const char *fmt
, ...)
431 struct va_format vaf
;
432 struct ext4_super_block
*es
= EXT4_SB(inode
->i_sb
)->s_es
;
434 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
435 es
->s_last_error_block
= cpu_to_le64(block
);
436 if (ext4_error_ratelimit(inode
->i_sb
)) {
441 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
442 "inode #%lu: block %llu: comm %s: %pV\n",
443 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
444 block
, current
->comm
, &vaf
);
446 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
447 "inode #%lu: comm %s: %pV\n",
448 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
449 current
->comm
, &vaf
);
452 save_error_info(inode
->i_sb
, function
, line
);
453 ext4_handle_error(inode
->i_sb
);
456 void __ext4_error_file(struct file
*file
, const char *function
,
457 unsigned int line
, ext4_fsblk_t block
,
458 const char *fmt
, ...)
461 struct va_format vaf
;
462 struct ext4_super_block
*es
;
463 struct inode
*inode
= file_inode(file
);
464 char pathname
[80], *path
;
466 es
= EXT4_SB(inode
->i_sb
)->s_es
;
467 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
468 if (ext4_error_ratelimit(inode
->i_sb
)) {
469 path
= d_path(&(file
->f_path
), pathname
, sizeof(pathname
));
477 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
478 "block %llu: comm %s: path %s: %pV\n",
479 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
480 block
, current
->comm
, path
, &vaf
);
483 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
484 "comm %s: path %s: %pV\n",
485 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
486 current
->comm
, path
, &vaf
);
489 save_error_info(inode
->i_sb
, function
, line
);
490 ext4_handle_error(inode
->i_sb
);
493 const char *ext4_decode_error(struct super_block
*sb
, int errno
,
500 errstr
= "IO failure";
503 errstr
= "Out of memory";
506 if (!sb
|| (EXT4_SB(sb
)->s_journal
&&
507 EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
))
508 errstr
= "Journal has aborted";
510 errstr
= "Readonly filesystem";
513 /* If the caller passed in an extra buffer for unknown
514 * errors, textualise them now. Else we just return
517 /* Check for truncated error codes... */
518 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
527 /* __ext4_std_error decodes expected errors from journaling functions
528 * automatically and invokes the appropriate error response. */
530 void __ext4_std_error(struct super_block
*sb
, const char *function
,
531 unsigned int line
, int errno
)
536 /* Special case: if the error is EROFS, and we're not already
537 * inside a transaction, then there's really no point in logging
539 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
540 (sb
->s_flags
& MS_RDONLY
))
543 if (ext4_error_ratelimit(sb
)) {
544 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
545 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s:%d: %s\n",
546 sb
->s_id
, function
, line
, errstr
);
549 save_error_info(sb
, function
, line
);
550 ext4_handle_error(sb
);
554 * ext4_abort is a much stronger failure handler than ext4_error. The
555 * abort function may be used to deal with unrecoverable failures such
556 * as journal IO errors or ENOMEM at a critical moment in log management.
558 * We unconditionally force the filesystem into an ABORT|READONLY state,
559 * unless the error response on the fs has been set to panic in which
560 * case we take the easy way out and panic immediately.
563 void __ext4_abort(struct super_block
*sb
, const char *function
,
564 unsigned int line
, const char *fmt
, ...)
568 save_error_info(sb
, function
, line
);
570 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: ", sb
->s_id
,
576 if ((sb
->s_flags
& MS_RDONLY
) == 0) {
577 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
578 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
580 * Make sure updated value of ->s_mount_flags will be visible
581 * before ->s_flags update
584 sb
->s_flags
|= MS_RDONLY
;
585 if (EXT4_SB(sb
)->s_journal
)
586 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
587 save_error_info(sb
, function
, line
);
589 if (test_opt(sb
, ERRORS_PANIC
))
590 panic("EXT4-fs panic from previous error\n");
593 void __ext4_msg(struct super_block
*sb
,
594 const char *prefix
, const char *fmt
, ...)
596 struct va_format vaf
;
599 if (!___ratelimit(&(EXT4_SB(sb
)->s_msg_ratelimit_state
), "EXT4-fs"))
605 printk("%sEXT4-fs (%s): %pV\n", prefix
, sb
->s_id
, &vaf
);
609 void __ext4_warning(struct super_block
*sb
, const char *function
,
610 unsigned int line
, const char *fmt
, ...)
612 struct va_format vaf
;
615 if (!___ratelimit(&(EXT4_SB(sb
)->s_warning_ratelimit_state
),
622 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: %pV\n",
623 sb
->s_id
, function
, line
, &vaf
);
627 void __ext4_grp_locked_error(const char *function
, unsigned int line
,
628 struct super_block
*sb
, ext4_group_t grp
,
629 unsigned long ino
, ext4_fsblk_t block
,
630 const char *fmt
, ...)
634 struct va_format vaf
;
636 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
638 es
->s_last_error_ino
= cpu_to_le32(ino
);
639 es
->s_last_error_block
= cpu_to_le64(block
);
640 __save_error_info(sb
, function
, line
);
642 if (ext4_error_ratelimit(sb
)) {
646 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: group %u, ",
647 sb
->s_id
, function
, line
, grp
);
649 printk(KERN_CONT
"inode %lu: ", ino
);
651 printk(KERN_CONT
"block %llu:",
652 (unsigned long long) block
);
653 printk(KERN_CONT
"%pV\n", &vaf
);
657 if (test_opt(sb
, ERRORS_CONT
)) {
658 ext4_commit_super(sb
, 0);
662 ext4_unlock_group(sb
, grp
);
663 ext4_handle_error(sb
);
665 * We only get here in the ERRORS_RO case; relocking the group
666 * may be dangerous, but nothing bad will happen since the
667 * filesystem will have already been marked read/only and the
668 * journal has been aborted. We return 1 as a hint to callers
669 * who might what to use the return value from
670 * ext4_grp_locked_error() to distinguish between the
671 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
672 * aggressively from the ext4 function in question, with a
673 * more appropriate error code.
675 ext4_lock_group(sb
, grp
);
679 void ext4_update_dynamic_rev(struct super_block
*sb
)
681 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
683 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
687 "updating to rev %d because of new feature flag, "
688 "running e2fsck is recommended",
691 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
692 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
693 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
694 /* leave es->s_feature_*compat flags alone */
695 /* es->s_uuid will be set by e2fsck if empty */
698 * The rest of the superblock fields should be zero, and if not it
699 * means they are likely already in use, so leave them alone. We
700 * can leave it up to e2fsck to clean up any inconsistencies there.
705 * Open the external journal device
707 static struct block_device
*ext4_blkdev_get(dev_t dev
, struct super_block
*sb
)
709 struct block_device
*bdev
;
710 char b
[BDEVNAME_SIZE
];
712 bdev
= blkdev_get_by_dev(dev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
, sb
);
718 ext4_msg(sb
, KERN_ERR
, "failed to open journal device %s: %ld",
719 __bdevname(dev
, b
), PTR_ERR(bdev
));
724 * Release the journal device
726 static void ext4_blkdev_put(struct block_device
*bdev
)
728 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
);
731 static void ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
733 struct block_device
*bdev
;
734 bdev
= sbi
->journal_bdev
;
736 ext4_blkdev_put(bdev
);
737 sbi
->journal_bdev
= NULL
;
741 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
743 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
746 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
750 ext4_msg(sb
, KERN_ERR
, "sb orphan head is %d",
751 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
753 printk(KERN_ERR
"sb_info orphan list:\n");
754 list_for_each(l
, &sbi
->s_orphan
) {
755 struct inode
*inode
= orphan_list_entry(l
);
757 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
758 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
759 inode
->i_mode
, inode
->i_nlink
,
764 static void ext4_put_super(struct super_block
*sb
)
766 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
767 struct ext4_super_block
*es
= sbi
->s_es
;
770 ext4_unregister_li_request(sb
);
771 dquot_disable(sb
, -1, DQUOT_USAGE_ENABLED
| DQUOT_LIMITS_ENABLED
);
773 flush_workqueue(sbi
->rsv_conversion_wq
);
774 destroy_workqueue(sbi
->rsv_conversion_wq
);
776 if (sbi
->s_journal
) {
777 err
= jbd2_journal_destroy(sbi
->s_journal
);
778 sbi
->s_journal
= NULL
;
780 ext4_abort(sb
, "Couldn't clean up the journal");
783 ext4_es_unregister_shrinker(sbi
);
784 del_timer_sync(&sbi
->s_err_report
);
785 ext4_release_system_zone(sb
);
787 ext4_ext_release(sb
);
788 ext4_xattr_put_super(sb
);
790 if (!(sb
->s_flags
& MS_RDONLY
)) {
791 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
792 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
794 if (!(sb
->s_flags
& MS_RDONLY
))
795 ext4_commit_super(sb
, 1);
798 remove_proc_entry("options", sbi
->s_proc
);
799 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
801 kobject_del(&sbi
->s_kobj
);
803 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
804 brelse(sbi
->s_group_desc
[i
]);
805 kvfree(sbi
->s_group_desc
);
806 kvfree(sbi
->s_flex_groups
);
807 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
808 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
809 percpu_counter_destroy(&sbi
->s_dirs_counter
);
810 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
813 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
814 kfree(sbi
->s_qf_names
[i
]);
817 /* Debugging code just in case the in-memory inode orphan list
818 * isn't empty. The on-disk one can be non-empty if we've
819 * detected an error and taken the fs readonly, but the
820 * in-memory list had better be clean by this point. */
821 if (!list_empty(&sbi
->s_orphan
))
822 dump_orphan_list(sb
, sbi
);
823 J_ASSERT(list_empty(&sbi
->s_orphan
));
825 invalidate_bdev(sb
->s_bdev
);
826 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
828 * Invalidate the journal device's buffers. We don't want them
829 * floating about in memory - the physical journal device may
830 * hotswapped, and it breaks the `ro-after' testing code.
832 sync_blockdev(sbi
->journal_bdev
);
833 invalidate_bdev(sbi
->journal_bdev
);
834 ext4_blkdev_remove(sbi
);
836 if (sbi
->s_mb_cache
) {
837 ext4_xattr_destroy_cache(sbi
->s_mb_cache
);
838 sbi
->s_mb_cache
= NULL
;
841 kthread_stop(sbi
->s_mmp_tsk
);
842 sb
->s_fs_info
= NULL
;
844 * Now that we are completely done shutting down the
845 * superblock, we need to actually destroy the kobject.
847 kobject_put(&sbi
->s_kobj
);
848 wait_for_completion(&sbi
->s_kobj_unregister
);
849 if (sbi
->s_chksum_driver
)
850 crypto_free_shash(sbi
->s_chksum_driver
);
851 kfree(sbi
->s_blockgroup_lock
);
855 static struct kmem_cache
*ext4_inode_cachep
;
858 * Called inside transaction, so use GFP_NOFS
860 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
862 struct ext4_inode_info
*ei
;
864 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
868 ei
->vfs_inode
.i_version
= 1;
869 spin_lock_init(&ei
->i_raw_lock
);
870 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
871 spin_lock_init(&ei
->i_prealloc_lock
);
872 ext4_es_init_tree(&ei
->i_es_tree
);
873 rwlock_init(&ei
->i_es_lock
);
874 INIT_LIST_HEAD(&ei
->i_es_list
);
877 ei
->i_es_shrink_lblk
= 0;
878 ei
->i_reserved_data_blocks
= 0;
879 ei
->i_reserved_meta_blocks
= 0;
880 ei
->i_allocated_meta_blocks
= 0;
881 ei
->i_da_metadata_calc_len
= 0;
882 ei
->i_da_metadata_calc_last_lblock
= 0;
883 spin_lock_init(&(ei
->i_block_reservation_lock
));
885 ei
->i_reserved_quota
= 0;
886 memset(&ei
->i_dquot
, 0, sizeof(ei
->i_dquot
));
889 INIT_LIST_HEAD(&ei
->i_rsv_conversion_list
);
890 spin_lock_init(&ei
->i_completed_io_lock
);
892 ei
->i_datasync_tid
= 0;
893 atomic_set(&ei
->i_ioend_count
, 0);
894 atomic_set(&ei
->i_unwritten
, 0);
895 INIT_WORK(&ei
->i_rsv_conversion_work
, ext4_end_io_rsv_work
);
897 return &ei
->vfs_inode
;
900 static int ext4_drop_inode(struct inode
*inode
)
902 int drop
= generic_drop_inode(inode
);
904 trace_ext4_drop_inode(inode
, drop
);
908 static void ext4_i_callback(struct rcu_head
*head
)
910 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
911 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
914 static void ext4_destroy_inode(struct inode
*inode
)
916 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
917 ext4_msg(inode
->i_sb
, KERN_ERR
,
918 "Inode %lu (%p): orphan list check failed!",
919 inode
->i_ino
, EXT4_I(inode
));
920 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
921 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
925 call_rcu(&inode
->i_rcu
, ext4_i_callback
);
928 static void init_once(void *foo
)
930 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
932 INIT_LIST_HEAD(&ei
->i_orphan
);
933 init_rwsem(&ei
->xattr_sem
);
934 init_rwsem(&ei
->i_data_sem
);
935 inode_init_once(&ei
->vfs_inode
);
938 static int __init
init_inodecache(void)
940 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
941 sizeof(struct ext4_inode_info
),
942 0, (SLAB_RECLAIM_ACCOUNT
|
945 if (ext4_inode_cachep
== NULL
)
950 static void destroy_inodecache(void)
953 * Make sure all delayed rcu free inodes are flushed before we
957 kmem_cache_destroy(ext4_inode_cachep
);
960 void ext4_clear_inode(struct inode
*inode
)
962 invalidate_inode_buffers(inode
);
965 ext4_discard_preallocations(inode
);
966 ext4_es_remove_extent(inode
, 0, EXT_MAX_BLOCKS
);
967 if (EXT4_I(inode
)->jinode
) {
968 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode
),
969 EXT4_I(inode
)->jinode
);
970 jbd2_free_inode(EXT4_I(inode
)->jinode
);
971 EXT4_I(inode
)->jinode
= NULL
;
975 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
976 u64 ino
, u32 generation
)
980 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
981 return ERR_PTR(-ESTALE
);
982 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
983 return ERR_PTR(-ESTALE
);
985 /* iget isn't really right if the inode is currently unallocated!!
987 * ext4_read_inode will return a bad_inode if the inode had been
988 * deleted, so we should be safe.
990 * Currently we don't know the generation for parent directory, so
991 * a generation of 0 means "accept any"
993 inode
= ext4_iget_normal(sb
, ino
);
995 return ERR_CAST(inode
);
996 if (generation
&& inode
->i_generation
!= generation
) {
998 return ERR_PTR(-ESTALE
);
1004 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1005 int fh_len
, int fh_type
)
1007 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
1008 ext4_nfs_get_inode
);
1011 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1012 int fh_len
, int fh_type
)
1014 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1015 ext4_nfs_get_inode
);
1019 * Try to release metadata pages (indirect blocks, directories) which are
1020 * mapped via the block device. Since these pages could have journal heads
1021 * which would prevent try_to_free_buffers() from freeing them, we must use
1022 * jbd2 layer's try_to_free_buffers() function to release them.
1024 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
1027 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
1029 WARN_ON(PageChecked(page
));
1030 if (!page_has_buffers(page
))
1033 return jbd2_journal_try_to_free_buffers(journal
, page
,
1034 wait
& ~__GFP_WAIT
);
1035 return try_to_free_buffers(page
);
1039 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1040 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1042 static int ext4_write_dquot(struct dquot
*dquot
);
1043 static int ext4_acquire_dquot(struct dquot
*dquot
);
1044 static int ext4_release_dquot(struct dquot
*dquot
);
1045 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
1046 static int ext4_write_info(struct super_block
*sb
, int type
);
1047 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
1049 static int ext4_quota_off(struct super_block
*sb
, int type
);
1050 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
1051 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
1052 size_t len
, loff_t off
);
1053 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
1054 const char *data
, size_t len
, loff_t off
);
1055 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
1056 unsigned int flags
);
1057 static int ext4_enable_quotas(struct super_block
*sb
);
1059 static struct dquot
**ext4_get_dquots(struct inode
*inode
)
1061 return EXT4_I(inode
)->i_dquot
;
1064 static const struct dquot_operations ext4_quota_operations
= {
1065 .get_reserved_space
= ext4_get_reserved_space
,
1066 .write_dquot
= ext4_write_dquot
,
1067 .acquire_dquot
= ext4_acquire_dquot
,
1068 .release_dquot
= ext4_release_dquot
,
1069 .mark_dirty
= ext4_mark_dquot_dirty
,
1070 .write_info
= ext4_write_info
,
1071 .alloc_dquot
= dquot_alloc
,
1072 .destroy_dquot
= dquot_destroy
,
1075 static const struct quotactl_ops ext4_qctl_operations
= {
1076 .quota_on
= ext4_quota_on
,
1077 .quota_off
= ext4_quota_off
,
1078 .quota_sync
= dquot_quota_sync
,
1079 .get_info
= dquot_get_dqinfo
,
1080 .set_info
= dquot_set_dqinfo
,
1081 .get_dqblk
= dquot_get_dqblk
,
1082 .set_dqblk
= dquot_set_dqblk
1086 static const struct super_operations ext4_sops
= {
1087 .alloc_inode
= ext4_alloc_inode
,
1088 .destroy_inode
= ext4_destroy_inode
,
1089 .write_inode
= ext4_write_inode
,
1090 .dirty_inode
= ext4_dirty_inode
,
1091 .drop_inode
= ext4_drop_inode
,
1092 .evict_inode
= ext4_evict_inode
,
1093 .put_super
= ext4_put_super
,
1094 .sync_fs
= ext4_sync_fs
,
1095 .freeze_fs
= ext4_freeze
,
1096 .unfreeze_fs
= ext4_unfreeze
,
1097 .statfs
= ext4_statfs
,
1098 .remount_fs
= ext4_remount
,
1099 .show_options
= ext4_show_options
,
1101 .quota_read
= ext4_quota_read
,
1102 .quota_write
= ext4_quota_write
,
1103 .get_dquots
= ext4_get_dquots
,
1105 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1108 static const struct export_operations ext4_export_ops
= {
1109 .fh_to_dentry
= ext4_fh_to_dentry
,
1110 .fh_to_parent
= ext4_fh_to_parent
,
1111 .get_parent
= ext4_get_parent
,
1115 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1116 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1117 Opt_nouid32
, Opt_debug
, Opt_removed
,
1118 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1119 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
,
1120 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
, Opt_journal_dev
,
1121 Opt_journal_path
, Opt_journal_checksum
, Opt_journal_async_commit
,
1122 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1123 Opt_data_err_abort
, Opt_data_err_ignore
,
1124 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1125 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
1126 Opt_noquota
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
1127 Opt_usrquota
, Opt_grpquota
, Opt_i_version
, Opt_dax
,
1128 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
, Opt_mblk_io_submit
,
1129 Opt_lazytime
, Opt_nolazytime
,
1130 Opt_nomblk_io_submit
, Opt_block_validity
, Opt_noblock_validity
,
1131 Opt_inode_readahead_blks
, Opt_journal_ioprio
,
1132 Opt_dioread_nolock
, Opt_dioread_lock
,
1133 Opt_discard
, Opt_nodiscard
, Opt_init_itable
, Opt_noinit_itable
,
1134 Opt_max_dir_size_kb
, Opt_nojournal_checksum
,
1137 static const match_table_t tokens
= {
1138 {Opt_bsd_df
, "bsddf"},
1139 {Opt_minix_df
, "minixdf"},
1140 {Opt_grpid
, "grpid"},
1141 {Opt_grpid
, "bsdgroups"},
1142 {Opt_nogrpid
, "nogrpid"},
1143 {Opt_nogrpid
, "sysvgroups"},
1144 {Opt_resgid
, "resgid=%u"},
1145 {Opt_resuid
, "resuid=%u"},
1147 {Opt_err_cont
, "errors=continue"},
1148 {Opt_err_panic
, "errors=panic"},
1149 {Opt_err_ro
, "errors=remount-ro"},
1150 {Opt_nouid32
, "nouid32"},
1151 {Opt_debug
, "debug"},
1152 {Opt_removed
, "oldalloc"},
1153 {Opt_removed
, "orlov"},
1154 {Opt_user_xattr
, "user_xattr"},
1155 {Opt_nouser_xattr
, "nouser_xattr"},
1157 {Opt_noacl
, "noacl"},
1158 {Opt_noload
, "norecovery"},
1159 {Opt_noload
, "noload"},
1160 {Opt_removed
, "nobh"},
1161 {Opt_removed
, "bh"},
1162 {Opt_commit
, "commit=%u"},
1163 {Opt_min_batch_time
, "min_batch_time=%u"},
1164 {Opt_max_batch_time
, "max_batch_time=%u"},
1165 {Opt_journal_dev
, "journal_dev=%u"},
1166 {Opt_journal_path
, "journal_path=%s"},
1167 {Opt_journal_checksum
, "journal_checksum"},
1168 {Opt_nojournal_checksum
, "nojournal_checksum"},
1169 {Opt_journal_async_commit
, "journal_async_commit"},
1170 {Opt_abort
, "abort"},
1171 {Opt_data_journal
, "data=journal"},
1172 {Opt_data_ordered
, "data=ordered"},
1173 {Opt_data_writeback
, "data=writeback"},
1174 {Opt_data_err_abort
, "data_err=abort"},
1175 {Opt_data_err_ignore
, "data_err=ignore"},
1176 {Opt_offusrjquota
, "usrjquota="},
1177 {Opt_usrjquota
, "usrjquota=%s"},
1178 {Opt_offgrpjquota
, "grpjquota="},
1179 {Opt_grpjquota
, "grpjquota=%s"},
1180 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1181 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1182 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
1183 {Opt_grpquota
, "grpquota"},
1184 {Opt_noquota
, "noquota"},
1185 {Opt_quota
, "quota"},
1186 {Opt_usrquota
, "usrquota"},
1187 {Opt_barrier
, "barrier=%u"},
1188 {Opt_barrier
, "barrier"},
1189 {Opt_nobarrier
, "nobarrier"},
1190 {Opt_i_version
, "i_version"},
1192 {Opt_stripe
, "stripe=%u"},
1193 {Opt_delalloc
, "delalloc"},
1194 {Opt_lazytime
, "lazytime"},
1195 {Opt_nolazytime
, "nolazytime"},
1196 {Opt_nodelalloc
, "nodelalloc"},
1197 {Opt_removed
, "mblk_io_submit"},
1198 {Opt_removed
, "nomblk_io_submit"},
1199 {Opt_block_validity
, "block_validity"},
1200 {Opt_noblock_validity
, "noblock_validity"},
1201 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1202 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1203 {Opt_auto_da_alloc
, "auto_da_alloc=%u"},
1204 {Opt_auto_da_alloc
, "auto_da_alloc"},
1205 {Opt_noauto_da_alloc
, "noauto_da_alloc"},
1206 {Opt_dioread_nolock
, "dioread_nolock"},
1207 {Opt_dioread_lock
, "dioread_lock"},
1208 {Opt_discard
, "discard"},
1209 {Opt_nodiscard
, "nodiscard"},
1210 {Opt_init_itable
, "init_itable=%u"},
1211 {Opt_init_itable
, "init_itable"},
1212 {Opt_noinit_itable
, "noinit_itable"},
1213 {Opt_max_dir_size_kb
, "max_dir_size_kb=%u"},
1214 {Opt_removed
, "check=none"}, /* mount option from ext2/3 */
1215 {Opt_removed
, "nocheck"}, /* mount option from ext2/3 */
1216 {Opt_removed
, "reservation"}, /* mount option from ext2/3 */
1217 {Opt_removed
, "noreservation"}, /* mount option from ext2/3 */
1218 {Opt_removed
, "journal=%u"}, /* mount option from ext2/3 */
1222 static ext4_fsblk_t
get_sb_block(void **data
)
1224 ext4_fsblk_t sb_block
;
1225 char *options
= (char *) *data
;
1227 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1228 return 1; /* Default location */
1231 /* TODO: use simple_strtoll with >32bit ext4 */
1232 sb_block
= simple_strtoul(options
, &options
, 0);
1233 if (*options
&& *options
!= ',') {
1234 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1238 if (*options
== ',')
1240 *data
= (void *) options
;
1245 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1246 static char deprecated_msg
[] = "Mount option \"%s\" will be removed by %s\n"
1247 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1250 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
1252 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1256 if (sb_any_quota_loaded(sb
) &&
1257 !sbi
->s_qf_names
[qtype
]) {
1258 ext4_msg(sb
, KERN_ERR
,
1259 "Cannot change journaled "
1260 "quota options when quota turned on");
1263 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
)) {
1264 ext4_msg(sb
, KERN_ERR
, "Cannot set journaled quota options "
1265 "when QUOTA feature is enabled");
1268 qname
= match_strdup(args
);
1270 ext4_msg(sb
, KERN_ERR
,
1271 "Not enough memory for storing quotafile name");
1274 if (sbi
->s_qf_names
[qtype
]) {
1275 if (strcmp(sbi
->s_qf_names
[qtype
], qname
) == 0)
1278 ext4_msg(sb
, KERN_ERR
,
1279 "%s quota file already specified",
1283 if (strchr(qname
, '/')) {
1284 ext4_msg(sb
, KERN_ERR
,
1285 "quotafile must be on filesystem root");
1288 sbi
->s_qf_names
[qtype
] = qname
;
1296 static int clear_qf_name(struct super_block
*sb
, int qtype
)
1299 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1301 if (sb_any_quota_loaded(sb
) &&
1302 sbi
->s_qf_names
[qtype
]) {
1303 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
1304 " when quota turned on");
1307 kfree(sbi
->s_qf_names
[qtype
]);
1308 sbi
->s_qf_names
[qtype
] = NULL
;
1313 #define MOPT_SET 0x0001
1314 #define MOPT_CLEAR 0x0002
1315 #define MOPT_NOSUPPORT 0x0004
1316 #define MOPT_EXPLICIT 0x0008
1317 #define MOPT_CLEAR_ERR 0x0010
1318 #define MOPT_GTE0 0x0020
1321 #define MOPT_QFMT 0x0040
1323 #define MOPT_Q MOPT_NOSUPPORT
1324 #define MOPT_QFMT MOPT_NOSUPPORT
1326 #define MOPT_DATAJ 0x0080
1327 #define MOPT_NO_EXT2 0x0100
1328 #define MOPT_NO_EXT3 0x0200
1329 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1330 #define MOPT_STRING 0x0400
1332 static const struct mount_opts
{
1336 } ext4_mount_opts
[] = {
1337 {Opt_minix_df
, EXT4_MOUNT_MINIX_DF
, MOPT_SET
},
1338 {Opt_bsd_df
, EXT4_MOUNT_MINIX_DF
, MOPT_CLEAR
},
1339 {Opt_grpid
, EXT4_MOUNT_GRPID
, MOPT_SET
},
1340 {Opt_nogrpid
, EXT4_MOUNT_GRPID
, MOPT_CLEAR
},
1341 {Opt_block_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_SET
},
1342 {Opt_noblock_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_CLEAR
},
1343 {Opt_dioread_nolock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1344 MOPT_EXT4_ONLY
| MOPT_SET
},
1345 {Opt_dioread_lock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1346 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1347 {Opt_discard
, EXT4_MOUNT_DISCARD
, MOPT_SET
},
1348 {Opt_nodiscard
, EXT4_MOUNT_DISCARD
, MOPT_CLEAR
},
1349 {Opt_delalloc
, EXT4_MOUNT_DELALLOC
,
1350 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1351 {Opt_nodelalloc
, EXT4_MOUNT_DELALLOC
,
1352 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1353 {Opt_nojournal_checksum
, EXT4_MOUNT_JOURNAL_CHECKSUM
,
1354 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1355 {Opt_journal_checksum
, EXT4_MOUNT_JOURNAL_CHECKSUM
,
1356 MOPT_EXT4_ONLY
| MOPT_SET
},
1357 {Opt_journal_async_commit
, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT
|
1358 EXT4_MOUNT_JOURNAL_CHECKSUM
),
1359 MOPT_EXT4_ONLY
| MOPT_SET
},
1360 {Opt_noload
, EXT4_MOUNT_NOLOAD
, MOPT_NO_EXT2
| MOPT_SET
},
1361 {Opt_err_panic
, EXT4_MOUNT_ERRORS_PANIC
, MOPT_SET
| MOPT_CLEAR_ERR
},
1362 {Opt_err_ro
, EXT4_MOUNT_ERRORS_RO
, MOPT_SET
| MOPT_CLEAR_ERR
},
1363 {Opt_err_cont
, EXT4_MOUNT_ERRORS_CONT
, MOPT_SET
| MOPT_CLEAR_ERR
},
1364 {Opt_data_err_abort
, EXT4_MOUNT_DATA_ERR_ABORT
,
1365 MOPT_NO_EXT2
| MOPT_SET
},
1366 {Opt_data_err_ignore
, EXT4_MOUNT_DATA_ERR_ABORT
,
1367 MOPT_NO_EXT2
| MOPT_CLEAR
},
1368 {Opt_barrier
, EXT4_MOUNT_BARRIER
, MOPT_SET
},
1369 {Opt_nobarrier
, EXT4_MOUNT_BARRIER
, MOPT_CLEAR
},
1370 {Opt_noauto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_SET
},
1371 {Opt_auto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_CLEAR
},
1372 {Opt_noinit_itable
, EXT4_MOUNT_INIT_INODE_TABLE
, MOPT_CLEAR
},
1373 {Opt_commit
, 0, MOPT_GTE0
},
1374 {Opt_max_batch_time
, 0, MOPT_GTE0
},
1375 {Opt_min_batch_time
, 0, MOPT_GTE0
},
1376 {Opt_inode_readahead_blks
, 0, MOPT_GTE0
},
1377 {Opt_init_itable
, 0, MOPT_GTE0
},
1378 {Opt_dax
, EXT4_MOUNT_DAX
, MOPT_SET
},
1379 {Opt_stripe
, 0, MOPT_GTE0
},
1380 {Opt_resuid
, 0, MOPT_GTE0
},
1381 {Opt_resgid
, 0, MOPT_GTE0
},
1382 {Opt_journal_dev
, 0, MOPT_GTE0
},
1383 {Opt_journal_path
, 0, MOPT_STRING
},
1384 {Opt_journal_ioprio
, 0, MOPT_GTE0
},
1385 {Opt_data_journal
, EXT4_MOUNT_JOURNAL_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1386 {Opt_data_ordered
, EXT4_MOUNT_ORDERED_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1387 {Opt_data_writeback
, EXT4_MOUNT_WRITEBACK_DATA
,
1388 MOPT_NO_EXT2
| MOPT_DATAJ
},
1389 {Opt_user_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_SET
},
1390 {Opt_nouser_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_CLEAR
},
1391 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1392 {Opt_acl
, EXT4_MOUNT_POSIX_ACL
, MOPT_SET
},
1393 {Opt_noacl
, EXT4_MOUNT_POSIX_ACL
, MOPT_CLEAR
},
1395 {Opt_acl
, 0, MOPT_NOSUPPORT
},
1396 {Opt_noacl
, 0, MOPT_NOSUPPORT
},
1398 {Opt_nouid32
, EXT4_MOUNT_NO_UID32
, MOPT_SET
},
1399 {Opt_debug
, EXT4_MOUNT_DEBUG
, MOPT_SET
},
1400 {Opt_quota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
, MOPT_SET
| MOPT_Q
},
1401 {Opt_usrquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
,
1403 {Opt_grpquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_GRPQUOTA
,
1405 {Opt_noquota
, (EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
|
1406 EXT4_MOUNT_GRPQUOTA
), MOPT_CLEAR
| MOPT_Q
},
1407 {Opt_usrjquota
, 0, MOPT_Q
},
1408 {Opt_grpjquota
, 0, MOPT_Q
},
1409 {Opt_offusrjquota
, 0, MOPT_Q
},
1410 {Opt_offgrpjquota
, 0, MOPT_Q
},
1411 {Opt_jqfmt_vfsold
, QFMT_VFS_OLD
, MOPT_QFMT
},
1412 {Opt_jqfmt_vfsv0
, QFMT_VFS_V0
, MOPT_QFMT
},
1413 {Opt_jqfmt_vfsv1
, QFMT_VFS_V1
, MOPT_QFMT
},
1414 {Opt_max_dir_size_kb
, 0, MOPT_GTE0
},
1418 static int handle_mount_opt(struct super_block
*sb
, char *opt
, int token
,
1419 substring_t
*args
, unsigned long *journal_devnum
,
1420 unsigned int *journal_ioprio
, int is_remount
)
1422 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1423 const struct mount_opts
*m
;
1429 if (token
== Opt_usrjquota
)
1430 return set_qf_name(sb
, USRQUOTA
, &args
[0]);
1431 else if (token
== Opt_grpjquota
)
1432 return set_qf_name(sb
, GRPQUOTA
, &args
[0]);
1433 else if (token
== Opt_offusrjquota
)
1434 return clear_qf_name(sb
, USRQUOTA
);
1435 else if (token
== Opt_offgrpjquota
)
1436 return clear_qf_name(sb
, GRPQUOTA
);
1440 case Opt_nouser_xattr
:
1441 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, opt
, "3.5");
1444 return 1; /* handled by get_sb_block() */
1446 ext4_msg(sb
, KERN_WARNING
, "Ignoring removed %s option", opt
);
1449 sbi
->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
1452 sb
->s_flags
|= MS_I_VERSION
;
1455 sb
->s_flags
|= MS_LAZYTIME
;
1457 case Opt_nolazytime
:
1458 sb
->s_flags
&= ~MS_LAZYTIME
;
1462 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++)
1463 if (token
== m
->token
)
1466 if (m
->token
== Opt_err
) {
1467 ext4_msg(sb
, KERN_ERR
, "Unrecognized mount option \"%s\" "
1468 "or missing value", opt
);
1472 if ((m
->flags
& MOPT_NO_EXT2
) && IS_EXT2_SB(sb
)) {
1473 ext4_msg(sb
, KERN_ERR
,
1474 "Mount option \"%s\" incompatible with ext2", opt
);
1477 if ((m
->flags
& MOPT_NO_EXT3
) && IS_EXT3_SB(sb
)) {
1478 ext4_msg(sb
, KERN_ERR
,
1479 "Mount option \"%s\" incompatible with ext3", opt
);
1483 if (args
->from
&& !(m
->flags
& MOPT_STRING
) && match_int(args
, &arg
))
1485 if (args
->from
&& (m
->flags
& MOPT_GTE0
) && (arg
< 0))
1487 if (m
->flags
& MOPT_EXPLICIT
)
1488 set_opt2(sb
, EXPLICIT_DELALLOC
);
1489 if (m
->flags
& MOPT_CLEAR_ERR
)
1490 clear_opt(sb
, ERRORS_MASK
);
1491 if (token
== Opt_noquota
&& sb_any_quota_loaded(sb
)) {
1492 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
1493 "options when quota turned on");
1497 if (m
->flags
& MOPT_NOSUPPORT
) {
1498 ext4_msg(sb
, KERN_ERR
, "%s option not supported", opt
);
1499 } else if (token
== Opt_commit
) {
1501 arg
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1502 sbi
->s_commit_interval
= HZ
* arg
;
1503 } else if (token
== Opt_max_batch_time
) {
1504 sbi
->s_max_batch_time
= arg
;
1505 } else if (token
== Opt_min_batch_time
) {
1506 sbi
->s_min_batch_time
= arg
;
1507 } else if (token
== Opt_inode_readahead_blks
) {
1508 if (arg
&& (arg
> (1 << 30) || !is_power_of_2(arg
))) {
1509 ext4_msg(sb
, KERN_ERR
,
1510 "EXT4-fs: inode_readahead_blks must be "
1511 "0 or a power of 2 smaller than 2^31");
1514 sbi
->s_inode_readahead_blks
= arg
;
1515 } else if (token
== Opt_init_itable
) {
1516 set_opt(sb
, INIT_INODE_TABLE
);
1518 arg
= EXT4_DEF_LI_WAIT_MULT
;
1519 sbi
->s_li_wait_mult
= arg
;
1520 } else if (token
== Opt_max_dir_size_kb
) {
1521 sbi
->s_max_dir_size_kb
= arg
;
1522 } else if (token
== Opt_stripe
) {
1523 sbi
->s_stripe
= arg
;
1524 } else if (token
== Opt_resuid
) {
1525 uid
= make_kuid(current_user_ns(), arg
);
1526 if (!uid_valid(uid
)) {
1527 ext4_msg(sb
, KERN_ERR
, "Invalid uid value %d", arg
);
1530 sbi
->s_resuid
= uid
;
1531 } else if (token
== Opt_resgid
) {
1532 gid
= make_kgid(current_user_ns(), arg
);
1533 if (!gid_valid(gid
)) {
1534 ext4_msg(sb
, KERN_ERR
, "Invalid gid value %d", arg
);
1537 sbi
->s_resgid
= gid
;
1538 } else if (token
== Opt_journal_dev
) {
1540 ext4_msg(sb
, KERN_ERR
,
1541 "Cannot specify journal on remount");
1544 *journal_devnum
= arg
;
1545 } else if (token
== Opt_journal_path
) {
1547 struct inode
*journal_inode
;
1552 ext4_msg(sb
, KERN_ERR
,
1553 "Cannot specify journal on remount");
1556 journal_path
= match_strdup(&args
[0]);
1557 if (!journal_path
) {
1558 ext4_msg(sb
, KERN_ERR
, "error: could not dup "
1559 "journal device string");
1563 error
= kern_path(journal_path
, LOOKUP_FOLLOW
, &path
);
1565 ext4_msg(sb
, KERN_ERR
, "error: could not find "
1566 "journal device path: error %d", error
);
1567 kfree(journal_path
);
1571 journal_inode
= path
.dentry
->d_inode
;
1572 if (!S_ISBLK(journal_inode
->i_mode
)) {
1573 ext4_msg(sb
, KERN_ERR
, "error: journal path %s "
1574 "is not a block device", journal_path
);
1576 kfree(journal_path
);
1580 *journal_devnum
= new_encode_dev(journal_inode
->i_rdev
);
1582 kfree(journal_path
);
1583 } else if (token
== Opt_journal_ioprio
) {
1585 ext4_msg(sb
, KERN_ERR
, "Invalid journal IO priority"
1590 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
, arg
);
1591 } else if (m
->flags
& MOPT_DATAJ
) {
1593 if (!sbi
->s_journal
)
1594 ext4_msg(sb
, KERN_WARNING
, "Remounting file system with no journal so ignoring journalled data option");
1595 else if (test_opt(sb
, DATA_FLAGS
) != m
->mount_opt
) {
1596 ext4_msg(sb
, KERN_ERR
,
1597 "Cannot change data mode on remount");
1601 clear_opt(sb
, DATA_FLAGS
);
1602 sbi
->s_mount_opt
|= m
->mount_opt
;
1605 } else if (m
->flags
& MOPT_QFMT
) {
1606 if (sb_any_quota_loaded(sb
) &&
1607 sbi
->s_jquota_fmt
!= m
->mount_opt
) {
1608 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled "
1609 "quota options when quota turned on");
1612 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
1613 EXT4_FEATURE_RO_COMPAT_QUOTA
)) {
1614 ext4_msg(sb
, KERN_ERR
,
1615 "Cannot set journaled quota options "
1616 "when QUOTA feature is enabled");
1619 sbi
->s_jquota_fmt
= m
->mount_opt
;
1621 #ifndef CONFIG_FS_DAX
1622 } else if (token
== Opt_dax
) {
1623 ext4_msg(sb
, KERN_INFO
, "dax option not supported");
1629 if (m
->flags
& MOPT_CLEAR
)
1631 else if (unlikely(!(m
->flags
& MOPT_SET
))) {
1632 ext4_msg(sb
, KERN_WARNING
,
1633 "buggy handling of option %s", opt
);
1638 sbi
->s_mount_opt
|= m
->mount_opt
;
1640 sbi
->s_mount_opt
&= ~m
->mount_opt
;
1645 static int parse_options(char *options
, struct super_block
*sb
,
1646 unsigned long *journal_devnum
,
1647 unsigned int *journal_ioprio
,
1650 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1652 substring_t args
[MAX_OPT_ARGS
];
1658 while ((p
= strsep(&options
, ",")) != NULL
) {
1662 * Initialize args struct so we know whether arg was
1663 * found; some options take optional arguments.
1665 args
[0].to
= args
[0].from
= NULL
;
1666 token
= match_token(p
, tokens
, args
);
1667 if (handle_mount_opt(sb
, p
, token
, args
, journal_devnum
,
1668 journal_ioprio
, is_remount
) < 0)
1672 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
) &&
1673 (test_opt(sb
, USRQUOTA
) || test_opt(sb
, GRPQUOTA
))) {
1674 ext4_msg(sb
, KERN_ERR
, "Cannot set quota options when QUOTA "
1675 "feature is enabled");
1678 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1679 if (test_opt(sb
, USRQUOTA
) && sbi
->s_qf_names
[USRQUOTA
])
1680 clear_opt(sb
, USRQUOTA
);
1682 if (test_opt(sb
, GRPQUOTA
) && sbi
->s_qf_names
[GRPQUOTA
])
1683 clear_opt(sb
, GRPQUOTA
);
1685 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
1686 ext4_msg(sb
, KERN_ERR
, "old and new quota "
1691 if (!sbi
->s_jquota_fmt
) {
1692 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1698 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
1700 BLOCK_SIZE
<< le32_to_cpu(sbi
->s_es
->s_log_block_size
);
1702 if (blocksize
< PAGE_CACHE_SIZE
) {
1703 ext4_msg(sb
, KERN_ERR
, "can't mount with "
1704 "dioread_nolock if block size != PAGE_SIZE");
1708 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
&&
1709 test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
1710 ext4_msg(sb
, KERN_ERR
, "can't mount with journal_async_commit "
1711 "in data=ordered mode");
1717 static inline void ext4_show_quota_options(struct seq_file
*seq
,
1718 struct super_block
*sb
)
1720 #if defined(CONFIG_QUOTA)
1721 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1723 if (sbi
->s_jquota_fmt
) {
1726 switch (sbi
->s_jquota_fmt
) {
1737 seq_printf(seq
, ",jqfmt=%s", fmtname
);
1740 if (sbi
->s_qf_names
[USRQUOTA
])
1741 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
1743 if (sbi
->s_qf_names
[GRPQUOTA
])
1744 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
1748 static const char *token2str(int token
)
1750 const struct match_token
*t
;
1752 for (t
= tokens
; t
->token
!= Opt_err
; t
++)
1753 if (t
->token
== token
&& !strchr(t
->pattern
, '='))
1760 * - it's set to a non-default value OR
1761 * - if the per-sb default is different from the global default
1763 static int _ext4_show_options(struct seq_file
*seq
, struct super_block
*sb
,
1766 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1767 struct ext4_super_block
*es
= sbi
->s_es
;
1768 int def_errors
, def_mount_opt
= nodefs
? 0 : sbi
->s_def_mount_opt
;
1769 const struct mount_opts
*m
;
1770 char sep
= nodefs
? '\n' : ',';
1772 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1773 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1775 if (sbi
->s_sb_block
!= 1)
1776 SEQ_OPTS_PRINT("sb=%llu", sbi
->s_sb_block
);
1778 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++) {
1779 int want_set
= m
->flags
& MOPT_SET
;
1780 if (((m
->flags
& (MOPT_SET
|MOPT_CLEAR
)) == 0) ||
1781 (m
->flags
& MOPT_CLEAR_ERR
))
1783 if (!(m
->mount_opt
& (sbi
->s_mount_opt
^ def_mount_opt
)))
1784 continue; /* skip if same as the default */
1786 (sbi
->s_mount_opt
& m
->mount_opt
) != m
->mount_opt
) ||
1787 (!want_set
&& (sbi
->s_mount_opt
& m
->mount_opt
)))
1788 continue; /* select Opt_noFoo vs Opt_Foo */
1789 SEQ_OPTS_PRINT("%s", token2str(m
->token
));
1792 if (nodefs
|| !uid_eq(sbi
->s_resuid
, make_kuid(&init_user_ns
, EXT4_DEF_RESUID
)) ||
1793 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
)
1794 SEQ_OPTS_PRINT("resuid=%u",
1795 from_kuid_munged(&init_user_ns
, sbi
->s_resuid
));
1796 if (nodefs
|| !gid_eq(sbi
->s_resgid
, make_kgid(&init_user_ns
, EXT4_DEF_RESGID
)) ||
1797 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
)
1798 SEQ_OPTS_PRINT("resgid=%u",
1799 from_kgid_munged(&init_user_ns
, sbi
->s_resgid
));
1800 def_errors
= nodefs
? -1 : le16_to_cpu(es
->s_errors
);
1801 if (test_opt(sb
, ERRORS_RO
) && def_errors
!= EXT4_ERRORS_RO
)
1802 SEQ_OPTS_PUTS("errors=remount-ro");
1803 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
1804 SEQ_OPTS_PUTS("errors=continue");
1805 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
1806 SEQ_OPTS_PUTS("errors=panic");
1807 if (nodefs
|| sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
)
1808 SEQ_OPTS_PRINT("commit=%lu", sbi
->s_commit_interval
/ HZ
);
1809 if (nodefs
|| sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
)
1810 SEQ_OPTS_PRINT("min_batch_time=%u", sbi
->s_min_batch_time
);
1811 if (nodefs
|| sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
)
1812 SEQ_OPTS_PRINT("max_batch_time=%u", sbi
->s_max_batch_time
);
1813 if (sb
->s_flags
& MS_I_VERSION
)
1814 SEQ_OPTS_PUTS("i_version");
1815 if (nodefs
|| sbi
->s_stripe
)
1816 SEQ_OPTS_PRINT("stripe=%lu", sbi
->s_stripe
);
1817 if (EXT4_MOUNT_DATA_FLAGS
& (sbi
->s_mount_opt
^ def_mount_opt
)) {
1818 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
1819 SEQ_OPTS_PUTS("data=journal");
1820 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
1821 SEQ_OPTS_PUTS("data=ordered");
1822 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
1823 SEQ_OPTS_PUTS("data=writeback");
1826 sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
1827 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1828 sbi
->s_inode_readahead_blks
);
1830 if (nodefs
|| (test_opt(sb
, INIT_INODE_TABLE
) &&
1831 (sbi
->s_li_wait_mult
!= EXT4_DEF_LI_WAIT_MULT
)))
1832 SEQ_OPTS_PRINT("init_itable=%u", sbi
->s_li_wait_mult
);
1833 if (nodefs
|| sbi
->s_max_dir_size_kb
)
1834 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi
->s_max_dir_size_kb
);
1836 ext4_show_quota_options(seq
, sb
);
1840 static int ext4_show_options(struct seq_file
*seq
, struct dentry
*root
)
1842 return _ext4_show_options(seq
, root
->d_sb
, 0);
1845 static int options_seq_show(struct seq_file
*seq
, void *offset
)
1847 struct super_block
*sb
= seq
->private;
1850 seq_puts(seq
, (sb
->s_flags
& MS_RDONLY
) ? "ro" : "rw");
1851 rc
= _ext4_show_options(seq
, sb
, 1);
1852 seq_puts(seq
, "\n");
1856 static int options_open_fs(struct inode
*inode
, struct file
*file
)
1858 return single_open(file
, options_seq_show
, PDE_DATA(inode
));
1861 static const struct file_operations ext4_seq_options_fops
= {
1862 .owner
= THIS_MODULE
,
1863 .open
= options_open_fs
,
1865 .llseek
= seq_lseek
,
1866 .release
= single_release
,
1869 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
1872 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1875 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
1876 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
1877 "forcing read-only mode");
1882 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
1883 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
1884 "running e2fsck is recommended");
1885 else if (sbi
->s_mount_state
& EXT4_ERROR_FS
)
1886 ext4_msg(sb
, KERN_WARNING
,
1887 "warning: mounting fs with errors, "
1888 "running e2fsck is recommended");
1889 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) > 0 &&
1890 le16_to_cpu(es
->s_mnt_count
) >=
1891 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1892 ext4_msg(sb
, KERN_WARNING
,
1893 "warning: maximal mount count reached, "
1894 "running e2fsck is recommended");
1895 else if (le32_to_cpu(es
->s_checkinterval
) &&
1896 (le32_to_cpu(es
->s_lastcheck
) +
1897 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1898 ext4_msg(sb
, KERN_WARNING
,
1899 "warning: checktime reached, "
1900 "running e2fsck is recommended");
1901 if (!sbi
->s_journal
)
1902 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
1903 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1904 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
1905 le16_add_cpu(&es
->s_mnt_count
, 1);
1906 es
->s_mtime
= cpu_to_le32(get_seconds());
1907 ext4_update_dynamic_rev(sb
);
1909 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
1911 ext4_commit_super(sb
, 1);
1913 if (test_opt(sb
, DEBUG
))
1914 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
1915 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1917 sbi
->s_groups_count
,
1918 EXT4_BLOCKS_PER_GROUP(sb
),
1919 EXT4_INODES_PER_GROUP(sb
),
1920 sbi
->s_mount_opt
, sbi
->s_mount_opt2
);
1922 cleancache_init_fs(sb
);
1926 int ext4_alloc_flex_bg_array(struct super_block
*sb
, ext4_group_t ngroup
)
1928 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1929 struct flex_groups
*new_groups
;
1932 if (!sbi
->s_log_groups_per_flex
)
1935 size
= ext4_flex_group(sbi
, ngroup
- 1) + 1;
1936 if (size
<= sbi
->s_flex_groups_allocated
)
1939 size
= roundup_pow_of_two(size
* sizeof(struct flex_groups
));
1940 new_groups
= ext4_kvzalloc(size
, GFP_KERNEL
);
1942 ext4_msg(sb
, KERN_ERR
, "not enough memory for %d flex groups",
1943 size
/ (int) sizeof(struct flex_groups
));
1947 if (sbi
->s_flex_groups
) {
1948 memcpy(new_groups
, sbi
->s_flex_groups
,
1949 (sbi
->s_flex_groups_allocated
*
1950 sizeof(struct flex_groups
)));
1951 kvfree(sbi
->s_flex_groups
);
1953 sbi
->s_flex_groups
= new_groups
;
1954 sbi
->s_flex_groups_allocated
= size
/ sizeof(struct flex_groups
);
1958 static int ext4_fill_flex_info(struct super_block
*sb
)
1960 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1961 struct ext4_group_desc
*gdp
= NULL
;
1962 ext4_group_t flex_group
;
1965 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
1966 if (sbi
->s_log_groups_per_flex
< 1 || sbi
->s_log_groups_per_flex
> 31) {
1967 sbi
->s_log_groups_per_flex
= 0;
1971 err
= ext4_alloc_flex_bg_array(sb
, sbi
->s_groups_count
);
1975 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1976 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1978 flex_group
= ext4_flex_group(sbi
, i
);
1979 atomic_add(ext4_free_inodes_count(sb
, gdp
),
1980 &sbi
->s_flex_groups
[flex_group
].free_inodes
);
1981 atomic64_add(ext4_free_group_clusters(sb
, gdp
),
1982 &sbi
->s_flex_groups
[flex_group
].free_clusters
);
1983 atomic_add(ext4_used_dirs_count(sb
, gdp
),
1984 &sbi
->s_flex_groups
[flex_group
].used_dirs
);
1992 static __le16
ext4_group_desc_csum(struct ext4_sb_info
*sbi
, __u32 block_group
,
1993 struct ext4_group_desc
*gdp
)
1997 __le32 le_group
= cpu_to_le32(block_group
);
1999 if (ext4_has_metadata_csum(sbi
->s_sb
)) {
2000 /* Use new metadata_csum algorithm */
2004 save_csum
= gdp
->bg_checksum
;
2005 gdp
->bg_checksum
= 0;
2006 csum32
= ext4_chksum(sbi
, sbi
->s_csum_seed
, (__u8
*)&le_group
,
2008 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)gdp
,
2010 gdp
->bg_checksum
= save_csum
;
2012 crc
= csum32
& 0xFFFF;
2016 /* old crc16 code */
2017 if (!(sbi
->s_es
->s_feature_ro_compat
&
2018 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)))
2021 offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
2023 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
2024 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
2025 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
2026 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
2027 /* for checksum of struct ext4_group_desc do the rest...*/
2028 if ((sbi
->s_es
->s_feature_incompat
&
2029 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT
)) &&
2030 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
2031 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
2032 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
2036 return cpu_to_le16(crc
);
2039 int ext4_group_desc_csum_verify(struct super_block
*sb
, __u32 block_group
,
2040 struct ext4_group_desc
*gdp
)
2042 if (ext4_has_group_desc_csum(sb
) &&
2043 (gdp
->bg_checksum
!= ext4_group_desc_csum(EXT4_SB(sb
),
2050 void ext4_group_desc_csum_set(struct super_block
*sb
, __u32 block_group
,
2051 struct ext4_group_desc
*gdp
)
2053 if (!ext4_has_group_desc_csum(sb
))
2055 gdp
->bg_checksum
= ext4_group_desc_csum(EXT4_SB(sb
), block_group
, gdp
);
2058 /* Called at mount-time, super-block is locked */
2059 static int ext4_check_descriptors(struct super_block
*sb
,
2060 ext4_group_t
*first_not_zeroed
)
2062 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2063 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
2064 ext4_fsblk_t last_block
;
2065 ext4_fsblk_t block_bitmap
;
2066 ext4_fsblk_t inode_bitmap
;
2067 ext4_fsblk_t inode_table
;
2068 int flexbg_flag
= 0;
2069 ext4_group_t i
, grp
= sbi
->s_groups_count
;
2071 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
2074 ext4_debug("Checking group descriptors");
2076 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2077 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2079 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
2080 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
2082 last_block
= first_block
+
2083 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2085 if ((grp
== sbi
->s_groups_count
) &&
2086 !(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2089 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
2090 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
2091 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2092 "Block bitmap for group %u not in group "
2093 "(block %llu)!", i
, block_bitmap
);
2096 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
2097 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
2098 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2099 "Inode bitmap for group %u not in group "
2100 "(block %llu)!", i
, inode_bitmap
);
2103 inode_table
= ext4_inode_table(sb
, gdp
);
2104 if (inode_table
< first_block
||
2105 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
2106 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2107 "Inode table for group %u not in group "
2108 "(block %llu)!", i
, inode_table
);
2111 ext4_lock_group(sb
, i
);
2112 if (!ext4_group_desc_csum_verify(sb
, i
, gdp
)) {
2113 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2114 "Checksum for group %u failed (%u!=%u)",
2115 i
, le16_to_cpu(ext4_group_desc_csum(sbi
, i
,
2116 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
2117 if (!(sb
->s_flags
& MS_RDONLY
)) {
2118 ext4_unlock_group(sb
, i
);
2122 ext4_unlock_group(sb
, i
);
2124 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
2126 if (NULL
!= first_not_zeroed
)
2127 *first_not_zeroed
= grp
;
2131 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2132 * the superblock) which were deleted from all directories, but held open by
2133 * a process at the time of a crash. We walk the list and try to delete these
2134 * inodes at recovery time (only with a read-write filesystem).
2136 * In order to keep the orphan inode chain consistent during traversal (in
2137 * case of crash during recovery), we link each inode into the superblock
2138 * orphan list_head and handle it the same way as an inode deletion during
2139 * normal operation (which journals the operations for us).
2141 * We only do an iget() and an iput() on each inode, which is very safe if we
2142 * accidentally point at an in-use or already deleted inode. The worst that
2143 * can happen in this case is that we get a "bit already cleared" message from
2144 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2145 * e2fsck was run on this filesystem, and it must have already done the orphan
2146 * inode cleanup for us, so we can safely abort without any further action.
2148 static void ext4_orphan_cleanup(struct super_block
*sb
,
2149 struct ext4_super_block
*es
)
2151 unsigned int s_flags
= sb
->s_flags
;
2152 int nr_orphans
= 0, nr_truncates
= 0;
2156 if (!es
->s_last_orphan
) {
2157 jbd_debug(4, "no orphan inodes to clean up\n");
2161 if (bdev_read_only(sb
->s_bdev
)) {
2162 ext4_msg(sb
, KERN_ERR
, "write access "
2163 "unavailable, skipping orphan cleanup");
2167 /* Check if feature set would not allow a r/w mount */
2168 if (!ext4_feature_set_ok(sb
, 0)) {
2169 ext4_msg(sb
, KERN_INFO
, "Skipping orphan cleanup due to "
2170 "unknown ROCOMPAT features");
2174 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2175 /* don't clear list on RO mount w/ errors */
2176 if (es
->s_last_orphan
&& !(s_flags
& MS_RDONLY
)) {
2177 ext4_msg(sb
, KERN_INFO
, "Errors on filesystem, "
2178 "clearing orphan list.\n");
2179 es
->s_last_orphan
= 0;
2181 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2185 if (s_flags
& MS_RDONLY
) {
2186 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
2187 sb
->s_flags
&= ~MS_RDONLY
;
2190 /* Needed for iput() to work correctly and not trash data */
2191 sb
->s_flags
|= MS_ACTIVE
;
2192 /* Turn on quotas so that they are updated correctly */
2193 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
2194 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
2195 int ret
= ext4_quota_on_mount(sb
, i
);
2197 ext4_msg(sb
, KERN_ERR
,
2198 "Cannot turn on journaled "
2199 "quota: error %d", ret
);
2204 while (es
->s_last_orphan
) {
2205 struct inode
*inode
;
2207 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
2208 if (IS_ERR(inode
)) {
2209 es
->s_last_orphan
= 0;
2213 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
2214 dquot_initialize(inode
);
2215 if (inode
->i_nlink
) {
2216 if (test_opt(sb
, DEBUG
))
2217 ext4_msg(sb
, KERN_DEBUG
,
2218 "%s: truncating inode %lu to %lld bytes",
2219 __func__
, inode
->i_ino
, inode
->i_size
);
2220 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2221 inode
->i_ino
, inode
->i_size
);
2222 mutex_lock(&inode
->i_mutex
);
2223 truncate_inode_pages(inode
->i_mapping
, inode
->i_size
);
2224 ext4_truncate(inode
);
2225 mutex_unlock(&inode
->i_mutex
);
2228 if (test_opt(sb
, DEBUG
))
2229 ext4_msg(sb
, KERN_DEBUG
,
2230 "%s: deleting unreferenced inode %lu",
2231 __func__
, inode
->i_ino
);
2232 jbd_debug(2, "deleting unreferenced inode %lu\n",
2236 iput(inode
); /* The delete magic happens here! */
2239 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2242 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
2243 PLURAL(nr_orphans
));
2245 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
2246 PLURAL(nr_truncates
));
2248 /* Turn quotas off */
2249 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
2250 if (sb_dqopt(sb
)->files
[i
])
2251 dquot_quota_off(sb
, i
);
2254 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
2258 * Maximal extent format file size.
2259 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2260 * extent format containers, within a sector_t, and within i_blocks
2261 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2262 * so that won't be a limiting factor.
2264 * However there is other limiting factor. We do store extents in the form
2265 * of starting block and length, hence the resulting length of the extent
2266 * covering maximum file size must fit into on-disk format containers as
2267 * well. Given that length is always by 1 unit bigger than max unit (because
2268 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2270 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2272 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
2275 loff_t upper_limit
= MAX_LFS_FILESIZE
;
2277 /* small i_blocks in vfs inode? */
2278 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2280 * CONFIG_LBDAF is not enabled implies the inode
2281 * i_block represent total blocks in 512 bytes
2282 * 32 == size of vfs inode i_blocks * 8
2284 upper_limit
= (1LL << 32) - 1;
2286 /* total blocks in file system block size */
2287 upper_limit
>>= (blkbits
- 9);
2288 upper_limit
<<= blkbits
;
2292 * 32-bit extent-start container, ee_block. We lower the maxbytes
2293 * by one fs block, so ee_len can cover the extent of maximum file
2296 res
= (1LL << 32) - 1;
2299 /* Sanity check against vm- & vfs- imposed limits */
2300 if (res
> upper_limit
)
2307 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2308 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2309 * We need to be 1 filesystem block less than the 2^48 sector limit.
2311 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
2313 loff_t res
= EXT4_NDIR_BLOCKS
;
2316 /* This is calculated to be the largest file size for a dense, block
2317 * mapped file such that the file's total number of 512-byte sectors,
2318 * including data and all indirect blocks, does not exceed (2^48 - 1).
2320 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2321 * number of 512-byte sectors of the file.
2324 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2326 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2327 * the inode i_block field represents total file blocks in
2328 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2330 upper_limit
= (1LL << 32) - 1;
2332 /* total blocks in file system block size */
2333 upper_limit
>>= (bits
- 9);
2337 * We use 48 bit ext4_inode i_blocks
2338 * With EXT4_HUGE_FILE_FL set the i_blocks
2339 * represent total number of blocks in
2340 * file system block size
2342 upper_limit
= (1LL << 48) - 1;
2346 /* indirect blocks */
2348 /* double indirect blocks */
2349 meta_blocks
+= 1 + (1LL << (bits
-2));
2350 /* tripple indirect blocks */
2351 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
2353 upper_limit
-= meta_blocks
;
2354 upper_limit
<<= bits
;
2356 res
+= 1LL << (bits
-2);
2357 res
+= 1LL << (2*(bits
-2));
2358 res
+= 1LL << (3*(bits
-2));
2360 if (res
> upper_limit
)
2363 if (res
> MAX_LFS_FILESIZE
)
2364 res
= MAX_LFS_FILESIZE
;
2369 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
2370 ext4_fsblk_t logical_sb_block
, int nr
)
2372 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2373 ext4_group_t bg
, first_meta_bg
;
2376 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
2378 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_META_BG
) ||
2380 return logical_sb_block
+ nr
+ 1;
2381 bg
= sbi
->s_desc_per_block
* nr
;
2382 if (ext4_bg_has_super(sb
, bg
))
2386 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2387 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2388 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2391 if (sb
->s_blocksize
== 1024 && nr
== 0 &&
2392 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_first_data_block
) == 0)
2395 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
2399 * ext4_get_stripe_size: Get the stripe size.
2400 * @sbi: In memory super block info
2402 * If we have specified it via mount option, then
2403 * use the mount option value. If the value specified at mount time is
2404 * greater than the blocks per group use the super block value.
2405 * If the super block value is greater than blocks per group return 0.
2406 * Allocator needs it be less than blocks per group.
2409 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
2411 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
2412 unsigned long stripe_width
=
2413 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
2416 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
2417 ret
= sbi
->s_stripe
;
2418 else if (stripe_width
<= sbi
->s_blocks_per_group
)
2420 else if (stride
<= sbi
->s_blocks_per_group
)
2426 * If the stripe width is 1, this makes no sense and
2427 * we set it to 0 to turn off stripe handling code.
2438 struct attribute attr
;
2439 ssize_t (*show
)(struct ext4_attr
*, struct ext4_sb_info
*, char *);
2440 ssize_t (*store
)(struct ext4_attr
*, struct ext4_sb_info
*,
2441 const char *, size_t);
2448 static int parse_strtoull(const char *buf
,
2449 unsigned long long max
, unsigned long long *value
)
2453 ret
= kstrtoull(skip_spaces(buf
), 0, value
);
2454 if (!ret
&& *value
> max
)
2459 static ssize_t
delayed_allocation_blocks_show(struct ext4_attr
*a
,
2460 struct ext4_sb_info
*sbi
,
2463 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2465 percpu_counter_sum(&sbi
->s_dirtyclusters_counter
)));
2468 static ssize_t
session_write_kbytes_show(struct ext4_attr
*a
,
2469 struct ext4_sb_info
*sbi
, char *buf
)
2471 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2473 if (!sb
->s_bdev
->bd_part
)
2474 return snprintf(buf
, PAGE_SIZE
, "0\n");
2475 return snprintf(buf
, PAGE_SIZE
, "%lu\n",
2476 (part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2477 sbi
->s_sectors_written_start
) >> 1);
2480 static ssize_t
lifetime_write_kbytes_show(struct ext4_attr
*a
,
2481 struct ext4_sb_info
*sbi
, char *buf
)
2483 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2485 if (!sb
->s_bdev
->bd_part
)
2486 return snprintf(buf
, PAGE_SIZE
, "0\n");
2487 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2488 (unsigned long long)(sbi
->s_kbytes_written
+
2489 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2490 EXT4_SB(sb
)->s_sectors_written_start
) >> 1)));
2493 static ssize_t
inode_readahead_blks_store(struct ext4_attr
*a
,
2494 struct ext4_sb_info
*sbi
,
2495 const char *buf
, size_t count
)
2500 ret
= kstrtoul(skip_spaces(buf
), 0, &t
);
2504 if (t
&& (!is_power_of_2(t
) || t
> 0x40000000))
2507 sbi
->s_inode_readahead_blks
= t
;
2511 static ssize_t
sbi_ui_show(struct ext4_attr
*a
,
2512 struct ext4_sb_info
*sbi
, char *buf
)
2514 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->u
.offset
);
2516 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
2519 static ssize_t
sbi_ui_store(struct ext4_attr
*a
,
2520 struct ext4_sb_info
*sbi
,
2521 const char *buf
, size_t count
)
2523 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->u
.offset
);
2527 ret
= kstrtoul(skip_spaces(buf
), 0, &t
);
2534 static ssize_t
es_ui_show(struct ext4_attr
*a
,
2535 struct ext4_sb_info
*sbi
, char *buf
)
2538 unsigned int *ui
= (unsigned int *) (((char *) sbi
->s_es
) +
2541 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
2544 static ssize_t
reserved_clusters_show(struct ext4_attr
*a
,
2545 struct ext4_sb_info
*sbi
, char *buf
)
2547 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2548 (unsigned long long) atomic64_read(&sbi
->s_resv_clusters
));
2551 static ssize_t
reserved_clusters_store(struct ext4_attr
*a
,
2552 struct ext4_sb_info
*sbi
,
2553 const char *buf
, size_t count
)
2555 unsigned long long val
;
2558 if (parse_strtoull(buf
, -1ULL, &val
))
2560 ret
= ext4_reserve_clusters(sbi
, val
);
2562 return ret
? ret
: count
;
2565 static ssize_t
trigger_test_error(struct ext4_attr
*a
,
2566 struct ext4_sb_info
*sbi
,
2567 const char *buf
, size_t count
)
2571 if (!capable(CAP_SYS_ADMIN
))
2574 if (len
&& buf
[len
-1] == '\n')
2578 ext4_error(sbi
->s_sb
, "%.*s", len
, buf
);
2582 static ssize_t
sbi_deprecated_show(struct ext4_attr
*a
,
2583 struct ext4_sb_info
*sbi
, char *buf
)
2585 return snprintf(buf
, PAGE_SIZE
, "%d\n", a
->u
.deprecated_val
);
2588 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2589 static struct ext4_attr ext4_attr_##_name = { \
2590 .attr = {.name = __stringify(_name), .mode = _mode }, \
2594 .offset = offsetof(struct ext4_sb_info, _elname),\
2598 #define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname) \
2599 static struct ext4_attr ext4_attr_##_name = { \
2600 .attr = {.name = __stringify(_name), .mode = _mode }, \
2604 .offset = offsetof(struct ext4_super_block, _elname), \
2608 #define EXT4_ATTR(name, mode, show, store) \
2609 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2611 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2612 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2613 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2615 #define EXT4_RO_ATTR_ES_UI(name, elname) \
2616 EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
2617 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2618 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2620 #define ATTR_LIST(name) &ext4_attr_##name.attr
2621 #define EXT4_DEPRECATED_ATTR(_name, _val) \
2622 static struct ext4_attr ext4_attr_##_name = { \
2623 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2624 .show = sbi_deprecated_show, \
2626 .deprecated_val = _val, \
2630 EXT4_RO_ATTR(delayed_allocation_blocks
);
2631 EXT4_RO_ATTR(session_write_kbytes
);
2632 EXT4_RO_ATTR(lifetime_write_kbytes
);
2633 EXT4_RW_ATTR(reserved_clusters
);
2634 EXT4_ATTR_OFFSET(inode_readahead_blks
, 0644, sbi_ui_show
,
2635 inode_readahead_blks_store
, s_inode_readahead_blks
);
2636 EXT4_RW_ATTR_SBI_UI(inode_goal
, s_inode_goal
);
2637 EXT4_RW_ATTR_SBI_UI(mb_stats
, s_mb_stats
);
2638 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan
, s_mb_max_to_scan
);
2639 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan
, s_mb_min_to_scan
);
2640 EXT4_RW_ATTR_SBI_UI(mb_order2_req
, s_mb_order2_reqs
);
2641 EXT4_RW_ATTR_SBI_UI(mb_stream_req
, s_mb_stream_request
);
2642 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc
, s_mb_group_prealloc
);
2643 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump
, 128);
2644 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb
, s_extent_max_zeroout_kb
);
2645 EXT4_ATTR(trigger_fs_error
, 0200, NULL
, trigger_test_error
);
2646 EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms
, s_err_ratelimit_state
.interval
);
2647 EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst
, s_err_ratelimit_state
.burst
);
2648 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms
, s_warning_ratelimit_state
.interval
);
2649 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst
, s_warning_ratelimit_state
.burst
);
2650 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms
, s_msg_ratelimit_state
.interval
);
2651 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst
, s_msg_ratelimit_state
.burst
);
2652 EXT4_RO_ATTR_ES_UI(errors_count
, s_error_count
);
2653 EXT4_RO_ATTR_ES_UI(first_error_time
, s_first_error_time
);
2654 EXT4_RO_ATTR_ES_UI(last_error_time
, s_last_error_time
);
2656 static struct attribute
*ext4_attrs
[] = {
2657 ATTR_LIST(delayed_allocation_blocks
),
2658 ATTR_LIST(session_write_kbytes
),
2659 ATTR_LIST(lifetime_write_kbytes
),
2660 ATTR_LIST(reserved_clusters
),
2661 ATTR_LIST(inode_readahead_blks
),
2662 ATTR_LIST(inode_goal
),
2663 ATTR_LIST(mb_stats
),
2664 ATTR_LIST(mb_max_to_scan
),
2665 ATTR_LIST(mb_min_to_scan
),
2666 ATTR_LIST(mb_order2_req
),
2667 ATTR_LIST(mb_stream_req
),
2668 ATTR_LIST(mb_group_prealloc
),
2669 ATTR_LIST(max_writeback_mb_bump
),
2670 ATTR_LIST(extent_max_zeroout_kb
),
2671 ATTR_LIST(trigger_fs_error
),
2672 ATTR_LIST(err_ratelimit_interval_ms
),
2673 ATTR_LIST(err_ratelimit_burst
),
2674 ATTR_LIST(warning_ratelimit_interval_ms
),
2675 ATTR_LIST(warning_ratelimit_burst
),
2676 ATTR_LIST(msg_ratelimit_interval_ms
),
2677 ATTR_LIST(msg_ratelimit_burst
),
2678 ATTR_LIST(errors_count
),
2679 ATTR_LIST(first_error_time
),
2680 ATTR_LIST(last_error_time
),
2684 /* Features this copy of ext4 supports */
2685 EXT4_INFO_ATTR(lazy_itable_init
);
2686 EXT4_INFO_ATTR(batched_discard
);
2687 EXT4_INFO_ATTR(meta_bg_resize
);
2689 static struct attribute
*ext4_feat_attrs
[] = {
2690 ATTR_LIST(lazy_itable_init
),
2691 ATTR_LIST(batched_discard
),
2692 ATTR_LIST(meta_bg_resize
),
2696 static ssize_t
ext4_attr_show(struct kobject
*kobj
,
2697 struct attribute
*attr
, char *buf
)
2699 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2701 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2703 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
2706 static ssize_t
ext4_attr_store(struct kobject
*kobj
,
2707 struct attribute
*attr
,
2708 const char *buf
, size_t len
)
2710 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2712 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2714 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
2717 static void ext4_sb_release(struct kobject
*kobj
)
2719 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2721 complete(&sbi
->s_kobj_unregister
);
2724 static const struct sysfs_ops ext4_attr_ops
= {
2725 .show
= ext4_attr_show
,
2726 .store
= ext4_attr_store
,
2729 static struct kobj_type ext4_ktype
= {
2730 .default_attrs
= ext4_attrs
,
2731 .sysfs_ops
= &ext4_attr_ops
,
2732 .release
= ext4_sb_release
,
2735 static void ext4_feat_release(struct kobject
*kobj
)
2737 complete(&ext4_feat
->f_kobj_unregister
);
2740 static ssize_t
ext4_feat_show(struct kobject
*kobj
,
2741 struct attribute
*attr
, char *buf
)
2743 return snprintf(buf
, PAGE_SIZE
, "supported\n");
2747 * We can not use ext4_attr_show/store because it relies on the kobject
2748 * being embedded in the ext4_sb_info structure which is definitely not
2749 * true in this case.
2751 static const struct sysfs_ops ext4_feat_ops
= {
2752 .show
= ext4_feat_show
,
2756 static struct kobj_type ext4_feat_ktype
= {
2757 .default_attrs
= ext4_feat_attrs
,
2758 .sysfs_ops
= &ext4_feat_ops
,
2759 .release
= ext4_feat_release
,
2763 * Check whether this filesystem can be mounted based on
2764 * the features present and the RDONLY/RDWR mount requested.
2765 * Returns 1 if this filesystem can be mounted as requested,
2766 * 0 if it cannot be.
2768 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
2770 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT4_FEATURE_INCOMPAT_SUPP
)) {
2771 ext4_msg(sb
, KERN_ERR
,
2772 "Couldn't mount because of "
2773 "unsupported optional features (%x)",
2774 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2775 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2782 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_READONLY
)) {
2783 ext4_msg(sb
, KERN_INFO
, "filesystem is read-only");
2784 sb
->s_flags
|= MS_RDONLY
;
2788 /* Check that feature set is OK for a read-write mount */
2789 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT4_FEATURE_RO_COMPAT_SUPP
)) {
2790 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
2791 "unsupported optional features (%x)",
2792 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2793 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2797 * Large file size enabled file system can only be mounted
2798 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2800 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_HUGE_FILE
)) {
2801 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
2802 ext4_msg(sb
, KERN_ERR
, "Filesystem with huge files "
2803 "cannot be mounted RDWR without "
2808 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_BIGALLOC
) &&
2809 !EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_EXTENTS
)) {
2810 ext4_msg(sb
, KERN_ERR
,
2811 "Can't support bigalloc feature without "
2812 "extents feature\n");
2816 #ifndef CONFIG_QUOTA
2817 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
) &&
2819 ext4_msg(sb
, KERN_ERR
,
2820 "Filesystem with quota feature cannot be mounted RDWR "
2821 "without CONFIG_QUOTA");
2824 #endif /* CONFIG_QUOTA */
2829 * This function is called once a day if we have errors logged
2830 * on the file system
2832 static void print_daily_error_info(unsigned long arg
)
2834 struct super_block
*sb
= (struct super_block
*) arg
;
2835 struct ext4_sb_info
*sbi
;
2836 struct ext4_super_block
*es
;
2841 if (es
->s_error_count
)
2842 /* fsck newer than v1.41.13 is needed to clean this condition. */
2843 ext4_msg(sb
, KERN_NOTICE
, "error count since last fsck: %u",
2844 le32_to_cpu(es
->s_error_count
));
2845 if (es
->s_first_error_time
) {
2846 printk(KERN_NOTICE
"EXT4-fs (%s): initial error at time %u: %.*s:%d",
2847 sb
->s_id
, le32_to_cpu(es
->s_first_error_time
),
2848 (int) sizeof(es
->s_first_error_func
),
2849 es
->s_first_error_func
,
2850 le32_to_cpu(es
->s_first_error_line
));
2851 if (es
->s_first_error_ino
)
2852 printk(": inode %u",
2853 le32_to_cpu(es
->s_first_error_ino
));
2854 if (es
->s_first_error_block
)
2855 printk(": block %llu", (unsigned long long)
2856 le64_to_cpu(es
->s_first_error_block
));
2859 if (es
->s_last_error_time
) {
2860 printk(KERN_NOTICE
"EXT4-fs (%s): last error at time %u: %.*s:%d",
2861 sb
->s_id
, le32_to_cpu(es
->s_last_error_time
),
2862 (int) sizeof(es
->s_last_error_func
),
2863 es
->s_last_error_func
,
2864 le32_to_cpu(es
->s_last_error_line
));
2865 if (es
->s_last_error_ino
)
2866 printk(": inode %u",
2867 le32_to_cpu(es
->s_last_error_ino
));
2868 if (es
->s_last_error_block
)
2869 printk(": block %llu", (unsigned long long)
2870 le64_to_cpu(es
->s_last_error_block
));
2873 mod_timer(&sbi
->s_err_report
, jiffies
+ 24*60*60*HZ
); /* Once a day */
2876 /* Find next suitable group and run ext4_init_inode_table */
2877 static int ext4_run_li_request(struct ext4_li_request
*elr
)
2879 struct ext4_group_desc
*gdp
= NULL
;
2880 ext4_group_t group
, ngroups
;
2881 struct super_block
*sb
;
2882 unsigned long timeout
= 0;
2886 ngroups
= EXT4_SB(sb
)->s_groups_count
;
2889 for (group
= elr
->lr_next_group
; group
< ngroups
; group
++) {
2890 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2896 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2900 if (group
>= ngroups
)
2905 ret
= ext4_init_inode_table(sb
, group
,
2906 elr
->lr_timeout
? 0 : 1);
2907 if (elr
->lr_timeout
== 0) {
2908 timeout
= (jiffies
- timeout
) *
2909 elr
->lr_sbi
->s_li_wait_mult
;
2910 elr
->lr_timeout
= timeout
;
2912 elr
->lr_next_sched
= jiffies
+ elr
->lr_timeout
;
2913 elr
->lr_next_group
= group
+ 1;
2921 * Remove lr_request from the list_request and free the
2922 * request structure. Should be called with li_list_mtx held
2924 static void ext4_remove_li_request(struct ext4_li_request
*elr
)
2926 struct ext4_sb_info
*sbi
;
2933 list_del(&elr
->lr_request
);
2934 sbi
->s_li_request
= NULL
;
2938 static void ext4_unregister_li_request(struct super_block
*sb
)
2940 mutex_lock(&ext4_li_mtx
);
2941 if (!ext4_li_info
) {
2942 mutex_unlock(&ext4_li_mtx
);
2946 mutex_lock(&ext4_li_info
->li_list_mtx
);
2947 ext4_remove_li_request(EXT4_SB(sb
)->s_li_request
);
2948 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2949 mutex_unlock(&ext4_li_mtx
);
2952 static struct task_struct
*ext4_lazyinit_task
;
2955 * This is the function where ext4lazyinit thread lives. It walks
2956 * through the request list searching for next scheduled filesystem.
2957 * When such a fs is found, run the lazy initialization request
2958 * (ext4_rn_li_request) and keep track of the time spend in this
2959 * function. Based on that time we compute next schedule time of
2960 * the request. When walking through the list is complete, compute
2961 * next waking time and put itself into sleep.
2963 static int ext4_lazyinit_thread(void *arg
)
2965 struct ext4_lazy_init
*eli
= (struct ext4_lazy_init
*)arg
;
2966 struct list_head
*pos
, *n
;
2967 struct ext4_li_request
*elr
;
2968 unsigned long next_wakeup
, cur
;
2970 BUG_ON(NULL
== eli
);
2974 next_wakeup
= MAX_JIFFY_OFFSET
;
2976 mutex_lock(&eli
->li_list_mtx
);
2977 if (list_empty(&eli
->li_request_list
)) {
2978 mutex_unlock(&eli
->li_list_mtx
);
2982 list_for_each_safe(pos
, n
, &eli
->li_request_list
) {
2983 elr
= list_entry(pos
, struct ext4_li_request
,
2986 if (time_after_eq(jiffies
, elr
->lr_next_sched
)) {
2987 if (ext4_run_li_request(elr
) != 0) {
2988 /* error, remove the lazy_init job */
2989 ext4_remove_li_request(elr
);
2994 if (time_before(elr
->lr_next_sched
, next_wakeup
))
2995 next_wakeup
= elr
->lr_next_sched
;
2997 mutex_unlock(&eli
->li_list_mtx
);
3002 if ((time_after_eq(cur
, next_wakeup
)) ||
3003 (MAX_JIFFY_OFFSET
== next_wakeup
)) {
3008 schedule_timeout_interruptible(next_wakeup
- cur
);
3010 if (kthread_should_stop()) {
3011 ext4_clear_request_list();
3018 * It looks like the request list is empty, but we need
3019 * to check it under the li_list_mtx lock, to prevent any
3020 * additions into it, and of course we should lock ext4_li_mtx
3021 * to atomically free the list and ext4_li_info, because at
3022 * this point another ext4 filesystem could be registering
3025 mutex_lock(&ext4_li_mtx
);
3026 mutex_lock(&eli
->li_list_mtx
);
3027 if (!list_empty(&eli
->li_request_list
)) {
3028 mutex_unlock(&eli
->li_list_mtx
);
3029 mutex_unlock(&ext4_li_mtx
);
3032 mutex_unlock(&eli
->li_list_mtx
);
3033 kfree(ext4_li_info
);
3034 ext4_li_info
= NULL
;
3035 mutex_unlock(&ext4_li_mtx
);
3040 static void ext4_clear_request_list(void)
3042 struct list_head
*pos
, *n
;
3043 struct ext4_li_request
*elr
;
3045 mutex_lock(&ext4_li_info
->li_list_mtx
);
3046 list_for_each_safe(pos
, n
, &ext4_li_info
->li_request_list
) {
3047 elr
= list_entry(pos
, struct ext4_li_request
,
3049 ext4_remove_li_request(elr
);
3051 mutex_unlock(&ext4_li_info
->li_list_mtx
);
3054 static int ext4_run_lazyinit_thread(void)
3056 ext4_lazyinit_task
= kthread_run(ext4_lazyinit_thread
,
3057 ext4_li_info
, "ext4lazyinit");
3058 if (IS_ERR(ext4_lazyinit_task
)) {
3059 int err
= PTR_ERR(ext4_lazyinit_task
);
3060 ext4_clear_request_list();
3061 kfree(ext4_li_info
);
3062 ext4_li_info
= NULL
;
3063 printk(KERN_CRIT
"EXT4-fs: error %d creating inode table "
3064 "initialization thread\n",
3068 ext4_li_info
->li_state
|= EXT4_LAZYINIT_RUNNING
;
3073 * Check whether it make sense to run itable init. thread or not.
3074 * If there is at least one uninitialized inode table, return
3075 * corresponding group number, else the loop goes through all
3076 * groups and return total number of groups.
3078 static ext4_group_t
ext4_has_uninit_itable(struct super_block
*sb
)
3080 ext4_group_t group
, ngroups
= EXT4_SB(sb
)->s_groups_count
;
3081 struct ext4_group_desc
*gdp
= NULL
;
3083 for (group
= 0; group
< ngroups
; group
++) {
3084 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
3088 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
3095 static int ext4_li_info_new(void)
3097 struct ext4_lazy_init
*eli
= NULL
;
3099 eli
= kzalloc(sizeof(*eli
), GFP_KERNEL
);
3103 INIT_LIST_HEAD(&eli
->li_request_list
);
3104 mutex_init(&eli
->li_list_mtx
);
3106 eli
->li_state
|= EXT4_LAZYINIT_QUIT
;
3113 static struct ext4_li_request
*ext4_li_request_new(struct super_block
*sb
,
3116 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3117 struct ext4_li_request
*elr
;
3119 elr
= kzalloc(sizeof(*elr
), GFP_KERNEL
);
3125 elr
->lr_next_group
= start
;
3128 * Randomize first schedule time of the request to
3129 * spread the inode table initialization requests
3132 elr
->lr_next_sched
= jiffies
+ (prandom_u32() %
3133 (EXT4_DEF_LI_MAX_START_DELAY
* HZ
));
3137 int ext4_register_li_request(struct super_block
*sb
,
3138 ext4_group_t first_not_zeroed
)
3140 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3141 struct ext4_li_request
*elr
= NULL
;
3142 ext4_group_t ngroups
= EXT4_SB(sb
)->s_groups_count
;
3145 mutex_lock(&ext4_li_mtx
);
3146 if (sbi
->s_li_request
!= NULL
) {
3148 * Reset timeout so it can be computed again, because
3149 * s_li_wait_mult might have changed.
3151 sbi
->s_li_request
->lr_timeout
= 0;
3155 if (first_not_zeroed
== ngroups
||
3156 (sb
->s_flags
& MS_RDONLY
) ||
3157 !test_opt(sb
, INIT_INODE_TABLE
))
3160 elr
= ext4_li_request_new(sb
, first_not_zeroed
);
3166 if (NULL
== ext4_li_info
) {
3167 ret
= ext4_li_info_new();
3172 mutex_lock(&ext4_li_info
->li_list_mtx
);
3173 list_add(&elr
->lr_request
, &ext4_li_info
->li_request_list
);
3174 mutex_unlock(&ext4_li_info
->li_list_mtx
);
3176 sbi
->s_li_request
= elr
;
3178 * set elr to NULL here since it has been inserted to
3179 * the request_list and the removal and free of it is
3180 * handled by ext4_clear_request_list from now on.
3184 if (!(ext4_li_info
->li_state
& EXT4_LAZYINIT_RUNNING
)) {
3185 ret
= ext4_run_lazyinit_thread();
3190 mutex_unlock(&ext4_li_mtx
);
3197 * We do not need to lock anything since this is called on
3200 static void ext4_destroy_lazyinit_thread(void)
3203 * If thread exited earlier
3204 * there's nothing to be done.
3206 if (!ext4_li_info
|| !ext4_lazyinit_task
)
3209 kthread_stop(ext4_lazyinit_task
);
3212 static int set_journal_csum_feature_set(struct super_block
*sb
)
3215 int compat
, incompat
;
3216 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3218 if (ext4_has_metadata_csum(sb
)) {
3219 /* journal checksum v3 */
3221 incompat
= JBD2_FEATURE_INCOMPAT_CSUM_V3
;
3223 /* journal checksum v1 */
3224 compat
= JBD2_FEATURE_COMPAT_CHECKSUM
;
3228 jbd2_journal_clear_features(sbi
->s_journal
,
3229 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3230 JBD2_FEATURE_INCOMPAT_CSUM_V3
|
3231 JBD2_FEATURE_INCOMPAT_CSUM_V2
);
3232 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
3233 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3235 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
|
3237 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
3238 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3241 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3242 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3244 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3245 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3252 * Note: calculating the overhead so we can be compatible with
3253 * historical BSD practice is quite difficult in the face of
3254 * clusters/bigalloc. This is because multiple metadata blocks from
3255 * different block group can end up in the same allocation cluster.
3256 * Calculating the exact overhead in the face of clustered allocation
3257 * requires either O(all block bitmaps) in memory or O(number of block
3258 * groups**2) in time. We will still calculate the superblock for
3259 * older file systems --- and if we come across with a bigalloc file
3260 * system with zero in s_overhead_clusters the estimate will be close to
3261 * correct especially for very large cluster sizes --- but for newer
3262 * file systems, it's better to calculate this figure once at mkfs
3263 * time, and store it in the superblock. If the superblock value is
3264 * present (even for non-bigalloc file systems), we will use it.
3266 static int count_overhead(struct super_block
*sb
, ext4_group_t grp
,
3269 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3270 struct ext4_group_desc
*gdp
;
3271 ext4_fsblk_t first_block
, last_block
, b
;
3272 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3273 int s
, j
, count
= 0;
3275 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_BIGALLOC
))
3276 return (ext4_bg_has_super(sb
, grp
) + ext4_bg_num_gdb(sb
, grp
) +
3277 sbi
->s_itb_per_group
+ 2);
3279 first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
) +
3280 (grp
* EXT4_BLOCKS_PER_GROUP(sb
));
3281 last_block
= first_block
+ EXT4_BLOCKS_PER_GROUP(sb
) - 1;
3282 for (i
= 0; i
< ngroups
; i
++) {
3283 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
3284 b
= ext4_block_bitmap(sb
, gdp
);
3285 if (b
>= first_block
&& b
<= last_block
) {
3286 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3289 b
= ext4_inode_bitmap(sb
, gdp
);
3290 if (b
>= first_block
&& b
<= last_block
) {
3291 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3294 b
= ext4_inode_table(sb
, gdp
);
3295 if (b
>= first_block
&& b
+ sbi
->s_itb_per_group
<= last_block
)
3296 for (j
= 0; j
< sbi
->s_itb_per_group
; j
++, b
++) {
3297 int c
= EXT4_B2C(sbi
, b
- first_block
);
3298 ext4_set_bit(c
, buf
);
3304 if (ext4_bg_has_super(sb
, grp
)) {
3305 ext4_set_bit(s
++, buf
);
3308 for (j
= ext4_bg_num_gdb(sb
, grp
); j
> 0; j
--) {
3309 ext4_set_bit(EXT4_B2C(sbi
, s
++), buf
);
3315 return EXT4_CLUSTERS_PER_GROUP(sb
) -
3316 ext4_count_free(buf
, EXT4_CLUSTERS_PER_GROUP(sb
) / 8);
3320 * Compute the overhead and stash it in sbi->s_overhead
3322 int ext4_calculate_overhead(struct super_block
*sb
)
3324 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3325 struct ext4_super_block
*es
= sbi
->s_es
;
3326 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3327 ext4_fsblk_t overhead
= 0;
3328 char *buf
= (char *) get_zeroed_page(GFP_NOFS
);
3334 * Compute the overhead (FS structures). This is constant
3335 * for a given filesystem unless the number of block groups
3336 * changes so we cache the previous value until it does.
3340 * All of the blocks before first_data_block are overhead
3342 overhead
= EXT4_B2C(sbi
, le32_to_cpu(es
->s_first_data_block
));
3345 * Add the overhead found in each block group
3347 for (i
= 0; i
< ngroups
; i
++) {
3350 blks
= count_overhead(sb
, i
, buf
);
3353 memset(buf
, 0, PAGE_SIZE
);
3356 /* Add the internal journal blocks as well */
3357 if (sbi
->s_journal
&& !sbi
->journal_bdev
)
3358 overhead
+= EXT4_NUM_B2C(sbi
, sbi
->s_journal
->j_maxlen
);
3360 sbi
->s_overhead
= overhead
;
3362 free_page((unsigned long) buf
);
3367 static ext4_fsblk_t
ext4_calculate_resv_clusters(struct super_block
*sb
)
3369 ext4_fsblk_t resv_clusters
;
3372 * There's no need to reserve anything when we aren't using extents.
3373 * The space estimates are exact, there are no unwritten extents,
3374 * hole punching doesn't need new metadata... This is needed especially
3375 * to keep ext2/3 backward compatibility.
3377 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_EXTENTS
))
3380 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3381 * This should cover the situations where we can not afford to run
3382 * out of space like for example punch hole, or converting
3383 * unwritten extents in delalloc path. In most cases such
3384 * allocation would require 1, or 2 blocks, higher numbers are
3387 resv_clusters
= ext4_blocks_count(EXT4_SB(sb
)->s_es
) >>
3388 EXT4_SB(sb
)->s_cluster_bits
;
3390 do_div(resv_clusters
, 50);
3391 resv_clusters
= min_t(ext4_fsblk_t
, resv_clusters
, 4096);
3393 return resv_clusters
;
3397 static int ext4_reserve_clusters(struct ext4_sb_info
*sbi
, ext4_fsblk_t count
)
3399 ext4_fsblk_t clusters
= ext4_blocks_count(sbi
->s_es
) >>
3400 sbi
->s_cluster_bits
;
3402 if (count
>= clusters
)
3405 atomic64_set(&sbi
->s_resv_clusters
, count
);
3409 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
3411 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
3412 struct buffer_head
*bh
;
3413 struct ext4_super_block
*es
= NULL
;
3414 struct ext4_sb_info
*sbi
;
3416 ext4_fsblk_t sb_block
= get_sb_block(&data
);
3417 ext4_fsblk_t logical_sb_block
;
3418 unsigned long offset
= 0;
3419 unsigned long journal_devnum
= 0;
3420 unsigned long def_mount_opts
;
3425 int blocksize
, clustersize
;
3426 unsigned int db_count
;
3428 int needs_recovery
, has_huge_files
, has_bigalloc
;
3431 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
3432 ext4_group_t first_not_zeroed
;
3434 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
3438 sbi
->s_blockgroup_lock
=
3439 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
3440 if (!sbi
->s_blockgroup_lock
) {
3444 sb
->s_fs_info
= sbi
;
3446 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
3447 sbi
->s_sb_block
= sb_block
;
3448 if (sb
->s_bdev
->bd_part
)
3449 sbi
->s_sectors_written_start
=
3450 part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]);
3452 /* Cleanup superblock name */
3453 for (cp
= sb
->s_id
; (cp
= strchr(cp
, '/'));)
3456 /* -EINVAL is default */
3458 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
3460 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
3465 * The ext4 superblock will not be buffer aligned for other than 1kB
3466 * block sizes. We need to calculate the offset from buffer start.
3468 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
3469 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3470 offset
= do_div(logical_sb_block
, blocksize
);
3472 logical_sb_block
= sb_block
;
3475 if (!(bh
= sb_bread_unmovable(sb
, logical_sb_block
))) {
3476 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
3480 * Note: s_es must be initialized as soon as possible because
3481 * some ext4 macro-instructions depend on its value
3483 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
3485 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
3486 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
3488 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
3490 /* Warn if metadata_csum and gdt_csum are both set. */
3491 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3492 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
) &&
3493 EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_GDT_CSUM
))
3494 ext4_warning(sb
, "metadata_csum and uninit_bg are "
3495 "redundant flags; please run fsck.");
3497 /* Check for a known checksum algorithm */
3498 if (!ext4_verify_csum_type(sb
, es
)) {
3499 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3500 "unknown checksum algorithm.");
3505 /* Load the checksum driver */
3506 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3507 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
)) {
3508 sbi
->s_chksum_driver
= crypto_alloc_shash("crc32c", 0, 0);
3509 if (IS_ERR(sbi
->s_chksum_driver
)) {
3510 ext4_msg(sb
, KERN_ERR
, "Cannot load crc32c driver.");
3511 ret
= PTR_ERR(sbi
->s_chksum_driver
);
3512 sbi
->s_chksum_driver
= NULL
;
3517 /* Check superblock checksum */
3518 if (!ext4_superblock_csum_verify(sb
, es
)) {
3519 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3520 "invalid superblock checksum. Run e2fsck?");
3525 /* Precompute checksum seed for all metadata */
3526 if (ext4_has_metadata_csum(sb
))
3527 sbi
->s_csum_seed
= ext4_chksum(sbi
, ~0, es
->s_uuid
,
3528 sizeof(es
->s_uuid
));
3530 /* Set defaults before we parse the mount options */
3531 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
3532 set_opt(sb
, INIT_INODE_TABLE
);
3533 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
3535 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
)
3537 if (def_mount_opts
& EXT4_DEFM_UID16
)
3538 set_opt(sb
, NO_UID32
);
3539 /* xattr user namespace & acls are now defaulted on */
3540 set_opt(sb
, XATTR_USER
);
3541 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3542 set_opt(sb
, POSIX_ACL
);
3544 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3545 if (ext4_has_metadata_csum(sb
))
3546 set_opt(sb
, JOURNAL_CHECKSUM
);
3548 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
3549 set_opt(sb
, JOURNAL_DATA
);
3550 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
3551 set_opt(sb
, ORDERED_DATA
);
3552 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
3553 set_opt(sb
, WRITEBACK_DATA
);
3555 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
3556 set_opt(sb
, ERRORS_PANIC
);
3557 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
3558 set_opt(sb
, ERRORS_CONT
);
3560 set_opt(sb
, ERRORS_RO
);
3561 /* block_validity enabled by default; disable with noblock_validity */
3562 set_opt(sb
, BLOCK_VALIDITY
);
3563 if (def_mount_opts
& EXT4_DEFM_DISCARD
)
3564 set_opt(sb
, DISCARD
);
3566 sbi
->s_resuid
= make_kuid(&init_user_ns
, le16_to_cpu(es
->s_def_resuid
));
3567 sbi
->s_resgid
= make_kgid(&init_user_ns
, le16_to_cpu(es
->s_def_resgid
));
3568 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
3569 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
3570 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
3572 if ((def_mount_opts
& EXT4_DEFM_NOBARRIER
) == 0)
3573 set_opt(sb
, BARRIER
);
3576 * enable delayed allocation by default
3577 * Use -o nodelalloc to turn it off
3579 if (!IS_EXT3_SB(sb
) && !IS_EXT2_SB(sb
) &&
3580 ((def_mount_opts
& EXT4_DEFM_NODELALLOC
) == 0))
3581 set_opt(sb
, DELALLOC
);
3584 * set default s_li_wait_mult for lazyinit, for the case there is
3585 * no mount option specified.
3587 sbi
->s_li_wait_mult
= EXT4_DEF_LI_WAIT_MULT
;
3589 if (!parse_options((char *) sbi
->s_es
->s_mount_opts
, sb
,
3590 &journal_devnum
, &journal_ioprio
, 0)) {
3591 ext4_msg(sb
, KERN_WARNING
,
3592 "failed to parse options in superblock: %s",
3593 sbi
->s_es
->s_mount_opts
);
3595 sbi
->s_def_mount_opt
= sbi
->s_mount_opt
;
3596 if (!parse_options((char *) data
, sb
, &journal_devnum
,
3597 &journal_ioprio
, 0))
3600 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
3601 printk_once(KERN_WARNING
"EXT4-fs: Warning: mounting "
3602 "with data=journal disables delayed "
3603 "allocation and O_DIRECT support!\n");
3604 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
3605 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3606 "both data=journal and delalloc");
3609 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
3610 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3611 "both data=journal and dioread_nolock");
3614 if (test_opt(sb
, DAX
)) {
3615 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3616 "both data=journal and dax");
3619 if (test_opt(sb
, DELALLOC
))
3620 clear_opt(sb
, DELALLOC
);
3623 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3624 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
3626 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
3627 (EXT4_HAS_COMPAT_FEATURE(sb
, ~0U) ||
3628 EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
3629 EXT4_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
3630 ext4_msg(sb
, KERN_WARNING
,
3631 "feature flags set on rev 0 fs, "
3632 "running e2fsck is recommended");
3634 if (es
->s_creator_os
== cpu_to_le32(EXT4_OS_HURD
)) {
3635 set_opt2(sb
, HURD_COMPAT
);
3636 if (EXT4_HAS_INCOMPAT_FEATURE(sb
,
3637 EXT4_FEATURE_INCOMPAT_64BIT
)) {
3638 ext4_msg(sb
, KERN_ERR
,
3639 "The Hurd can't support 64-bit file systems");
3644 if (IS_EXT2_SB(sb
)) {
3645 if (ext2_feature_set_ok(sb
))
3646 ext4_msg(sb
, KERN_INFO
, "mounting ext2 file system "
3647 "using the ext4 subsystem");
3649 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext2 due "
3650 "to feature incompatibilities");
3655 if (IS_EXT3_SB(sb
)) {
3656 if (ext3_feature_set_ok(sb
))
3657 ext4_msg(sb
, KERN_INFO
, "mounting ext3 file system "
3658 "using the ext4 subsystem");
3660 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext3 due "
3661 "to feature incompatibilities");
3667 * Check feature flags regardless of the revision level, since we
3668 * previously didn't change the revision level when setting the flags,
3669 * so there is a chance incompat flags are set on a rev 0 filesystem.
3671 if (!ext4_feature_set_ok(sb
, (sb
->s_flags
& MS_RDONLY
)))
3674 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
3675 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
3676 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
3677 ext4_msg(sb
, KERN_ERR
,
3678 "Unsupported filesystem blocksize %d", blocksize
);
3682 if (sbi
->s_mount_opt
& EXT4_MOUNT_DAX
) {
3683 if (blocksize
!= PAGE_SIZE
) {
3684 ext4_msg(sb
, KERN_ERR
,
3685 "error: unsupported blocksize for dax");
3688 if (!sb
->s_bdev
->bd_disk
->fops
->direct_access
) {
3689 ext4_msg(sb
, KERN_ERR
,
3690 "error: device does not support dax");
3695 if (sb
->s_blocksize
!= blocksize
) {
3696 /* Validate the filesystem blocksize */
3697 if (!sb_set_blocksize(sb
, blocksize
)) {
3698 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
3704 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3705 offset
= do_div(logical_sb_block
, blocksize
);
3706 bh
= sb_bread_unmovable(sb
, logical_sb_block
);
3708 ext4_msg(sb
, KERN_ERR
,
3709 "Can't read superblock on 2nd try");
3712 es
= (struct ext4_super_block
*)(bh
->b_data
+ offset
);
3714 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
3715 ext4_msg(sb
, KERN_ERR
,
3716 "Magic mismatch, very weird!");
3721 has_huge_files
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3722 EXT4_FEATURE_RO_COMPAT_HUGE_FILE
);
3723 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
3725 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
3727 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
3728 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
3729 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
3731 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
3732 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
3733 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
3734 (!is_power_of_2(sbi
->s_inode_size
)) ||
3735 (sbi
->s_inode_size
> blocksize
)) {
3736 ext4_msg(sb
, KERN_ERR
,
3737 "unsupported inode size: %d",
3741 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
3742 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
3745 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
3746 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
)) {
3747 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
3748 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
3749 !is_power_of_2(sbi
->s_desc_size
)) {
3750 ext4_msg(sb
, KERN_ERR
,
3751 "unsupported descriptor size %lu",
3756 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
3758 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
3759 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
3760 if (EXT4_INODE_SIZE(sb
) == 0 || EXT4_INODES_PER_GROUP(sb
) == 0)
3763 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
3764 if (sbi
->s_inodes_per_block
== 0)
3766 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
3767 sbi
->s_inodes_per_block
;
3768 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
3770 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3771 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
3772 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
3774 for (i
= 0; i
< 4; i
++)
3775 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
3776 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
3777 if (EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_DIR_INDEX
)) {
3778 i
= le32_to_cpu(es
->s_flags
);
3779 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
3780 sbi
->s_hash_unsigned
= 3;
3781 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
3782 #ifdef __CHAR_UNSIGNED__
3783 if (!(sb
->s_flags
& MS_RDONLY
))
3785 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
3786 sbi
->s_hash_unsigned
= 3;
3788 if (!(sb
->s_flags
& MS_RDONLY
))
3790 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
3795 /* Handle clustersize */
3796 clustersize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_cluster_size
);
3797 has_bigalloc
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3798 EXT4_FEATURE_RO_COMPAT_BIGALLOC
);
3800 if (clustersize
< blocksize
) {
3801 ext4_msg(sb
, KERN_ERR
,
3802 "cluster size (%d) smaller than "
3803 "block size (%d)", clustersize
, blocksize
);
3806 sbi
->s_cluster_bits
= le32_to_cpu(es
->s_log_cluster_size
) -
3807 le32_to_cpu(es
->s_log_block_size
);
3808 sbi
->s_clusters_per_group
=
3809 le32_to_cpu(es
->s_clusters_per_group
);
3810 if (sbi
->s_clusters_per_group
> blocksize
* 8) {
3811 ext4_msg(sb
, KERN_ERR
,
3812 "#clusters per group too big: %lu",
3813 sbi
->s_clusters_per_group
);
3816 if (sbi
->s_blocks_per_group
!=
3817 (sbi
->s_clusters_per_group
* (clustersize
/ blocksize
))) {
3818 ext4_msg(sb
, KERN_ERR
, "blocks per group (%lu) and "
3819 "clusters per group (%lu) inconsistent",
3820 sbi
->s_blocks_per_group
,
3821 sbi
->s_clusters_per_group
);
3825 if (clustersize
!= blocksize
) {
3826 ext4_warning(sb
, "fragment/cluster size (%d) != "
3827 "block size (%d)", clustersize
,
3829 clustersize
= blocksize
;
3831 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
3832 ext4_msg(sb
, KERN_ERR
,
3833 "#blocks per group too big: %lu",
3834 sbi
->s_blocks_per_group
);
3837 sbi
->s_clusters_per_group
= sbi
->s_blocks_per_group
;
3838 sbi
->s_cluster_bits
= 0;
3840 sbi
->s_cluster_ratio
= clustersize
/ blocksize
;
3842 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
3843 ext4_msg(sb
, KERN_ERR
,
3844 "#inodes per group too big: %lu",
3845 sbi
->s_inodes_per_group
);
3849 /* Do we have standard group size of clustersize * 8 blocks ? */
3850 if (sbi
->s_blocks_per_group
== clustersize
<< 3)
3851 set_opt2(sb
, STD_GROUP_SIZE
);
3854 * Test whether we have more sectors than will fit in sector_t,
3855 * and whether the max offset is addressable by the page cache.
3857 err
= generic_check_addressable(sb
->s_blocksize_bits
,
3858 ext4_blocks_count(es
));
3860 ext4_msg(sb
, KERN_ERR
, "filesystem"
3861 " too large to mount safely on this system");
3862 if (sizeof(sector_t
) < 8)
3863 ext4_msg(sb
, KERN_WARNING
, "CONFIG_LBDAF not enabled");
3867 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
3870 /* check blocks count against device size */
3871 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
3872 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
3873 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
3874 "exceeds size of device (%llu blocks)",
3875 ext4_blocks_count(es
), blocks_count
);
3880 * It makes no sense for the first data block to be beyond the end
3881 * of the filesystem.
3883 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
3884 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data "
3885 "block %u is beyond end of filesystem (%llu)",
3886 le32_to_cpu(es
->s_first_data_block
),
3887 ext4_blocks_count(es
));
3890 blocks_count
= (ext4_blocks_count(es
) -
3891 le32_to_cpu(es
->s_first_data_block
) +
3892 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
3893 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
3894 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
3895 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %u "
3896 "(block count %llu, first data block %u, "
3897 "blocks per group %lu)", sbi
->s_groups_count
,
3898 ext4_blocks_count(es
),
3899 le32_to_cpu(es
->s_first_data_block
),
3900 EXT4_BLOCKS_PER_GROUP(sb
));
3903 sbi
->s_groups_count
= blocks_count
;
3904 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
3905 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
3906 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
3907 EXT4_DESC_PER_BLOCK(sb
);
3908 sbi
->s_group_desc
= ext4_kvmalloc(db_count
*
3909 sizeof(struct buffer_head
*),
3911 if (sbi
->s_group_desc
== NULL
) {
3912 ext4_msg(sb
, KERN_ERR
, "not enough memory");
3918 sbi
->s_proc
= proc_mkdir(sb
->s_id
, ext4_proc_root
);
3921 proc_create_data("options", S_IRUGO
, sbi
->s_proc
,
3922 &ext4_seq_options_fops
, sb
);
3924 bgl_lock_init(sbi
->s_blockgroup_lock
);
3926 for (i
= 0; i
< db_count
; i
++) {
3927 block
= descriptor_loc(sb
, logical_sb_block
, i
);
3928 sbi
->s_group_desc
[i
] = sb_bread_unmovable(sb
, block
);
3929 if (!sbi
->s_group_desc
[i
]) {
3930 ext4_msg(sb
, KERN_ERR
,
3931 "can't read group descriptor %d", i
);
3936 if (!ext4_check_descriptors(sb
, &first_not_zeroed
)) {
3937 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
3941 sbi
->s_gdb_count
= db_count
;
3942 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
3943 spin_lock_init(&sbi
->s_next_gen_lock
);
3945 setup_timer(&sbi
->s_err_report
, print_daily_error_info
,
3946 (unsigned long) sb
);
3948 /* Register extent status tree shrinker */
3949 if (ext4_es_register_shrinker(sbi
))
3952 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
3953 sbi
->s_extent_max_zeroout_kb
= 32;
3956 * set up enough so that it can read an inode
3958 sb
->s_op
= &ext4_sops
;
3959 sb
->s_export_op
= &ext4_export_ops
;
3960 sb
->s_xattr
= ext4_xattr_handlers
;
3962 sb
->dq_op
= &ext4_quota_operations
;
3963 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
))
3964 sb
->s_qcop
= &dquot_quotactl_sysfile_ops
;
3966 sb
->s_qcop
= &ext4_qctl_operations
;
3967 sb
->s_quota_types
= QTYPE_MASK_USR
| QTYPE_MASK_GRP
;
3969 memcpy(sb
->s_uuid
, es
->s_uuid
, sizeof(es
->s_uuid
));
3971 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
3972 mutex_init(&sbi
->s_orphan_lock
);
3976 needs_recovery
= (es
->s_last_orphan
!= 0 ||
3977 EXT4_HAS_INCOMPAT_FEATURE(sb
,
3978 EXT4_FEATURE_INCOMPAT_RECOVER
));
3980 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_MMP
) &&
3981 !(sb
->s_flags
& MS_RDONLY
))
3982 if (ext4_multi_mount_protect(sb
, le64_to_cpu(es
->s_mmp_block
)))
3983 goto failed_mount3a
;
3986 * The first inode we look at is the journal inode. Don't try
3987 * root first: it may be modified in the journal!
3989 if (!test_opt(sb
, NOLOAD
) &&
3990 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
3991 if (ext4_load_journal(sb
, es
, journal_devnum
))
3992 goto failed_mount3a
;
3993 } else if (test_opt(sb
, NOLOAD
) && !(sb
->s_flags
& MS_RDONLY
) &&
3994 EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
3995 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
3996 "suppressed and not mounted read-only");
3997 goto failed_mount_wq
;
3999 clear_opt(sb
, DATA_FLAGS
);
4000 sbi
->s_journal
= NULL
;
4005 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
) &&
4006 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
4007 JBD2_FEATURE_INCOMPAT_64BIT
)) {
4008 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
4009 goto failed_mount_wq
;
4012 if (!set_journal_csum_feature_set(sb
)) {
4013 ext4_msg(sb
, KERN_ERR
, "Failed to set journal checksum "
4015 goto failed_mount_wq
;
4018 /* We have now updated the journal if required, so we can
4019 * validate the data journaling mode. */
4020 switch (test_opt(sb
, DATA_FLAGS
)) {
4022 /* No mode set, assume a default based on the journal
4023 * capabilities: ORDERED_DATA if the journal can
4024 * cope, else JOURNAL_DATA
4026 if (jbd2_journal_check_available_features
4027 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
4028 set_opt(sb
, ORDERED_DATA
);
4030 set_opt(sb
, JOURNAL_DATA
);
4033 case EXT4_MOUNT_ORDERED_DATA
:
4034 case EXT4_MOUNT_WRITEBACK_DATA
:
4035 if (!jbd2_journal_check_available_features
4036 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
4037 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
4038 "requested data journaling mode");
4039 goto failed_mount_wq
;
4044 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
4046 sbi
->s_journal
->j_commit_callback
= ext4_journal_commit_callback
;
4049 if (ext4_mballoc_ready
) {
4050 sbi
->s_mb_cache
= ext4_xattr_create_cache(sb
->s_id
);
4051 if (!sbi
->s_mb_cache
) {
4052 ext4_msg(sb
, KERN_ERR
, "Failed to create an mb_cache");
4053 goto failed_mount_wq
;
4058 * Get the # of file system overhead blocks from the
4059 * superblock if present.
4061 if (es
->s_overhead_clusters
)
4062 sbi
->s_overhead
= le32_to_cpu(es
->s_overhead_clusters
);
4064 err
= ext4_calculate_overhead(sb
);
4066 goto failed_mount_wq
;
4070 * The maximum number of concurrent works can be high and
4071 * concurrency isn't really necessary. Limit it to 1.
4073 EXT4_SB(sb
)->rsv_conversion_wq
=
4074 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM
| WQ_UNBOUND
, 1);
4075 if (!EXT4_SB(sb
)->rsv_conversion_wq
) {
4076 printk(KERN_ERR
"EXT4-fs: failed to create workqueue\n");
4082 * The jbd2_journal_load will have done any necessary log recovery,
4083 * so we can safely mount the rest of the filesystem now.
4086 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
4088 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
4089 ret
= PTR_ERR(root
);
4093 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
4094 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
4098 sb
->s_root
= d_make_root(root
);
4100 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
4105 if (ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
))
4106 sb
->s_flags
|= MS_RDONLY
;
4108 /* determine the minimum size of new large inodes, if present */
4109 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
4110 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
4111 EXT4_GOOD_OLD_INODE_SIZE
;
4112 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
4113 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE
)) {
4114 if (sbi
->s_want_extra_isize
<
4115 le16_to_cpu(es
->s_want_extra_isize
))
4116 sbi
->s_want_extra_isize
=
4117 le16_to_cpu(es
->s_want_extra_isize
);
4118 if (sbi
->s_want_extra_isize
<
4119 le16_to_cpu(es
->s_min_extra_isize
))
4120 sbi
->s_want_extra_isize
=
4121 le16_to_cpu(es
->s_min_extra_isize
);
4124 /* Check if enough inode space is available */
4125 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
4126 sbi
->s_inode_size
) {
4127 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
4128 EXT4_GOOD_OLD_INODE_SIZE
;
4129 ext4_msg(sb
, KERN_INFO
, "required extra inode space not"
4133 err
= ext4_reserve_clusters(sbi
, ext4_calculate_resv_clusters(sb
));
4135 ext4_msg(sb
, KERN_ERR
, "failed to reserve %llu clusters for "
4136 "reserved pool", ext4_calculate_resv_clusters(sb
));
4137 goto failed_mount4a
;
4140 err
= ext4_setup_system_zone(sb
);
4142 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
4144 goto failed_mount4a
;
4148 err
= ext4_mb_init(sb
);
4150 ext4_msg(sb
, KERN_ERR
, "failed to initialize mballoc (%d)",
4155 block
= ext4_count_free_clusters(sb
);
4156 ext4_free_blocks_count_set(sbi
->s_es
,
4157 EXT4_C2B(sbi
, block
));
4158 err
= percpu_counter_init(&sbi
->s_freeclusters_counter
, block
,
4161 unsigned long freei
= ext4_count_free_inodes(sb
);
4162 sbi
->s_es
->s_free_inodes_count
= cpu_to_le32(freei
);
4163 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
, freei
,
4167 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
4168 ext4_count_dirs(sb
), GFP_KERNEL
);
4170 err
= percpu_counter_init(&sbi
->s_dirtyclusters_counter
, 0,
4173 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
4177 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
4178 if (!ext4_fill_flex_info(sb
)) {
4179 ext4_msg(sb
, KERN_ERR
,
4180 "unable to initialize "
4181 "flex_bg meta info!");
4185 err
= ext4_register_li_request(sb
, first_not_zeroed
);
4189 sbi
->s_kobj
.kset
= ext4_kset
;
4190 init_completion(&sbi
->s_kobj_unregister
);
4191 err
= kobject_init_and_add(&sbi
->s_kobj
, &ext4_ktype
, NULL
,
4197 /* Enable quota usage during mount. */
4198 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
) &&
4199 !(sb
->s_flags
& MS_RDONLY
)) {
4200 err
= ext4_enable_quotas(sb
);
4204 #endif /* CONFIG_QUOTA */
4206 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
4207 ext4_orphan_cleanup(sb
, es
);
4208 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
4209 if (needs_recovery
) {
4210 ext4_msg(sb
, KERN_INFO
, "recovery complete");
4211 ext4_mark_recovery_complete(sb
, es
);
4213 if (EXT4_SB(sb
)->s_journal
) {
4214 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
4215 descr
= " journalled data mode";
4216 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
4217 descr
= " ordered data mode";
4219 descr
= " writeback data mode";
4221 descr
= "out journal";
4223 if (test_opt(sb
, DISCARD
)) {
4224 struct request_queue
*q
= bdev_get_queue(sb
->s_bdev
);
4225 if (!blk_queue_discard(q
))
4226 ext4_msg(sb
, KERN_WARNING
,
4227 "mounting with \"discard\" option, but "
4228 "the device does not support discard");
4231 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s. "
4232 "Opts: %s%s%s", descr
, sbi
->s_es
->s_mount_opts
,
4233 *sbi
->s_es
->s_mount_opts
? "; " : "", orig_data
);
4235 if (es
->s_error_count
)
4236 mod_timer(&sbi
->s_err_report
, jiffies
+ 300*HZ
); /* 5 minutes */
4238 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4239 ratelimit_state_init(&sbi
->s_err_ratelimit_state
, 5 * HZ
, 10);
4240 ratelimit_state_init(&sbi
->s_warning_ratelimit_state
, 5 * HZ
, 10);
4241 ratelimit_state_init(&sbi
->s_msg_ratelimit_state
, 5 * HZ
, 10);
4248 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
4253 kobject_del(&sbi
->s_kobj
);
4256 ext4_unregister_li_request(sb
);
4258 ext4_mb_release(sb
);
4259 if (sbi
->s_flex_groups
)
4260 kvfree(sbi
->s_flex_groups
);
4261 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
4262 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
4263 percpu_counter_destroy(&sbi
->s_dirs_counter
);
4264 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
4266 ext4_ext_release(sb
);
4267 ext4_release_system_zone(sb
);
4272 ext4_msg(sb
, KERN_ERR
, "mount failed");
4273 if (EXT4_SB(sb
)->rsv_conversion_wq
)
4274 destroy_workqueue(EXT4_SB(sb
)->rsv_conversion_wq
);
4276 if (sbi
->s_journal
) {
4277 jbd2_journal_destroy(sbi
->s_journal
);
4278 sbi
->s_journal
= NULL
;
4281 ext4_es_unregister_shrinker(sbi
);
4283 del_timer_sync(&sbi
->s_err_report
);
4285 kthread_stop(sbi
->s_mmp_tsk
);
4287 for (i
= 0; i
< db_count
; i
++)
4288 brelse(sbi
->s_group_desc
[i
]);
4289 kvfree(sbi
->s_group_desc
);
4291 if (sbi
->s_chksum_driver
)
4292 crypto_free_shash(sbi
->s_chksum_driver
);
4294 remove_proc_entry("options", sbi
->s_proc
);
4295 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
4298 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
4299 kfree(sbi
->s_qf_names
[i
]);
4301 ext4_blkdev_remove(sbi
);
4304 sb
->s_fs_info
= NULL
;
4305 kfree(sbi
->s_blockgroup_lock
);
4309 return err
? err
: ret
;
4313 * Setup any per-fs journal parameters now. We'll do this both on
4314 * initial mount, once the journal has been initialised but before we've
4315 * done any recovery; and again on any subsequent remount.
4317 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
4319 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4321 journal
->j_commit_interval
= sbi
->s_commit_interval
;
4322 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
4323 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
4325 write_lock(&journal
->j_state_lock
);
4326 if (test_opt(sb
, BARRIER
))
4327 journal
->j_flags
|= JBD2_BARRIER
;
4329 journal
->j_flags
&= ~JBD2_BARRIER
;
4330 if (test_opt(sb
, DATA_ERR_ABORT
))
4331 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
4333 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
4334 write_unlock(&journal
->j_state_lock
);
4337 static journal_t
*ext4_get_journal(struct super_block
*sb
,
4338 unsigned int journal_inum
)
4340 struct inode
*journal_inode
;
4343 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4345 /* First, test for the existence of a valid inode on disk. Bad
4346 * things happen if we iget() an unused inode, as the subsequent
4347 * iput() will try to delete it. */
4349 journal_inode
= ext4_iget(sb
, journal_inum
);
4350 if (IS_ERR(journal_inode
)) {
4351 ext4_msg(sb
, KERN_ERR
, "no journal found");
4354 if (!journal_inode
->i_nlink
) {
4355 make_bad_inode(journal_inode
);
4356 iput(journal_inode
);
4357 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
4361 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4362 journal_inode
, journal_inode
->i_size
);
4363 if (!S_ISREG(journal_inode
->i_mode
)) {
4364 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
4365 iput(journal_inode
);
4369 journal
= jbd2_journal_init_inode(journal_inode
);
4371 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
4372 iput(journal_inode
);
4375 journal
->j_private
= sb
;
4376 ext4_init_journal_params(sb
, journal
);
4380 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
4383 struct buffer_head
*bh
;
4387 int hblock
, blocksize
;
4388 ext4_fsblk_t sb_block
;
4389 unsigned long offset
;
4390 struct ext4_super_block
*es
;
4391 struct block_device
*bdev
;
4393 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4395 bdev
= ext4_blkdev_get(j_dev
, sb
);
4399 blocksize
= sb
->s_blocksize
;
4400 hblock
= bdev_logical_block_size(bdev
);
4401 if (blocksize
< hblock
) {
4402 ext4_msg(sb
, KERN_ERR
,
4403 "blocksize too small for journal device");
4407 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
4408 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
4409 set_blocksize(bdev
, blocksize
);
4410 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
4411 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
4412 "external journal");
4416 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
4417 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
4418 !(le32_to_cpu(es
->s_feature_incompat
) &
4419 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
4420 ext4_msg(sb
, KERN_ERR
, "external journal has "
4426 if ((le32_to_cpu(es
->s_feature_ro_compat
) &
4427 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
) &&
4428 es
->s_checksum
!= ext4_superblock_csum(sb
, es
)) {
4429 ext4_msg(sb
, KERN_ERR
, "external journal has "
4430 "corrupt superblock");
4435 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
4436 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
4441 len
= ext4_blocks_count(es
);
4442 start
= sb_block
+ 1;
4443 brelse(bh
); /* we're done with the superblock */
4445 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
4446 start
, len
, blocksize
);
4448 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
4451 journal
->j_private
= sb
;
4452 ll_rw_block(READ
| REQ_META
| REQ_PRIO
, 1, &journal
->j_sb_buffer
);
4453 wait_on_buffer(journal
->j_sb_buffer
);
4454 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
4455 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
4458 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
4459 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
4460 "user (unsupported) - %d",
4461 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
4464 EXT4_SB(sb
)->journal_bdev
= bdev
;
4465 ext4_init_journal_params(sb
, journal
);
4469 jbd2_journal_destroy(journal
);
4471 ext4_blkdev_put(bdev
);
4475 static int ext4_load_journal(struct super_block
*sb
,
4476 struct ext4_super_block
*es
,
4477 unsigned long journal_devnum
)
4480 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
4483 int really_read_only
;
4485 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4487 if (journal_devnum
&&
4488 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4489 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
4490 "numbers have changed");
4491 journal_dev
= new_decode_dev(journal_devnum
);
4493 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
4495 really_read_only
= bdev_read_only(sb
->s_bdev
);
4498 * Are we loading a blank journal or performing recovery after a
4499 * crash? For recovery, we need to check in advance whether we
4500 * can get read-write access to the device.
4502 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
4503 if (sb
->s_flags
& MS_RDONLY
) {
4504 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
4505 "required on readonly filesystem");
4506 if (really_read_only
) {
4507 ext4_msg(sb
, KERN_ERR
, "write access "
4508 "unavailable, cannot proceed");
4511 ext4_msg(sb
, KERN_INFO
, "write access will "
4512 "be enabled during recovery");
4516 if (journal_inum
&& journal_dev
) {
4517 ext4_msg(sb
, KERN_ERR
, "filesystem has both journal "
4518 "and inode journals!");
4523 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
4526 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
4530 if (!(journal
->j_flags
& JBD2_BARRIER
))
4531 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
4533 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
))
4534 err
= jbd2_journal_wipe(journal
, !really_read_only
);
4536 char *save
= kmalloc(EXT4_S_ERR_LEN
, GFP_KERNEL
);
4538 memcpy(save
, ((char *) es
) +
4539 EXT4_S_ERR_START
, EXT4_S_ERR_LEN
);
4540 err
= jbd2_journal_load(journal
);
4542 memcpy(((char *) es
) + EXT4_S_ERR_START
,
4543 save
, EXT4_S_ERR_LEN
);
4548 ext4_msg(sb
, KERN_ERR
, "error loading journal");
4549 jbd2_journal_destroy(journal
);
4553 EXT4_SB(sb
)->s_journal
= journal
;
4554 ext4_clear_journal_err(sb
, es
);
4556 if (!really_read_only
&& journal_devnum
&&
4557 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4558 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
4560 /* Make sure we flush the recovery flag to disk. */
4561 ext4_commit_super(sb
, 1);
4567 static int ext4_commit_super(struct super_block
*sb
, int sync
)
4569 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
4570 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
4573 if (!sbh
|| block_device_ejected(sb
))
4575 if (buffer_write_io_error(sbh
)) {
4577 * Oh, dear. A previous attempt to write the
4578 * superblock failed. This could happen because the
4579 * USB device was yanked out. Or it could happen to
4580 * be a transient write error and maybe the block will
4581 * be remapped. Nothing we can do but to retry the
4582 * write and hope for the best.
4584 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
4585 "superblock detected");
4586 clear_buffer_write_io_error(sbh
);
4587 set_buffer_uptodate(sbh
);
4590 * If the file system is mounted read-only, don't update the
4591 * superblock write time. This avoids updating the superblock
4592 * write time when we are mounting the root file system
4593 * read/only but we need to replay the journal; at that point,
4594 * for people who are east of GMT and who make their clock
4595 * tick in localtime for Windows bug-for-bug compatibility,
4596 * the clock is set in the future, and this will cause e2fsck
4597 * to complain and force a full file system check.
4599 if (!(sb
->s_flags
& MS_RDONLY
))
4600 es
->s_wtime
= cpu_to_le32(get_seconds());
4601 if (sb
->s_bdev
->bd_part
)
4602 es
->s_kbytes_written
=
4603 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
+
4604 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
4605 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
4607 es
->s_kbytes_written
=
4608 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
);
4609 if (percpu_counter_initialized(&EXT4_SB(sb
)->s_freeclusters_counter
))
4610 ext4_free_blocks_count_set(es
,
4611 EXT4_C2B(EXT4_SB(sb
), percpu_counter_sum_positive(
4612 &EXT4_SB(sb
)->s_freeclusters_counter
)));
4613 if (percpu_counter_initialized(&EXT4_SB(sb
)->s_freeinodes_counter
))
4614 es
->s_free_inodes_count
=
4615 cpu_to_le32(percpu_counter_sum_positive(
4616 &EXT4_SB(sb
)->s_freeinodes_counter
));
4617 BUFFER_TRACE(sbh
, "marking dirty");
4618 ext4_superblock_csum_set(sb
);
4619 mark_buffer_dirty(sbh
);
4621 error
= sync_dirty_buffer(sbh
);
4625 error
= buffer_write_io_error(sbh
);
4627 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
4629 clear_buffer_write_io_error(sbh
);
4630 set_buffer_uptodate(sbh
);
4637 * Have we just finished recovery? If so, and if we are mounting (or
4638 * remounting) the filesystem readonly, then we will end up with a
4639 * consistent fs on disk. Record that fact.
4641 static void ext4_mark_recovery_complete(struct super_block
*sb
,
4642 struct ext4_super_block
*es
)
4644 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
4646 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
4647 BUG_ON(journal
!= NULL
);
4650 jbd2_journal_lock_updates(journal
);
4651 if (jbd2_journal_flush(journal
) < 0)
4654 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
) &&
4655 sb
->s_flags
& MS_RDONLY
) {
4656 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4657 ext4_commit_super(sb
, 1);
4661 jbd2_journal_unlock_updates(journal
);
4665 * If we are mounting (or read-write remounting) a filesystem whose journal
4666 * has recorded an error from a previous lifetime, move that error to the
4667 * main filesystem now.
4669 static void ext4_clear_journal_err(struct super_block
*sb
,
4670 struct ext4_super_block
*es
)
4676 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4678 journal
= EXT4_SB(sb
)->s_journal
;
4681 * Now check for any error status which may have been recorded in the
4682 * journal by a prior ext4_error() or ext4_abort()
4685 j_errno
= jbd2_journal_errno(journal
);
4689 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
4690 ext4_warning(sb
, "Filesystem error recorded "
4691 "from previous mount: %s", errstr
);
4692 ext4_warning(sb
, "Marking fs in need of filesystem check.");
4694 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
4695 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
4696 ext4_commit_super(sb
, 1);
4698 jbd2_journal_clear_err(journal
);
4699 jbd2_journal_update_sb_errno(journal
);
4704 * Force the running and committing transactions to commit,
4705 * and wait on the commit.
4707 int ext4_force_commit(struct super_block
*sb
)
4711 if (sb
->s_flags
& MS_RDONLY
)
4714 journal
= EXT4_SB(sb
)->s_journal
;
4715 return ext4_journal_force_commit(journal
);
4718 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
4722 bool needs_barrier
= false;
4723 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4725 trace_ext4_sync_fs(sb
, wait
);
4726 flush_workqueue(sbi
->rsv_conversion_wq
);
4728 * Writeback quota in non-journalled quota case - journalled quota has
4731 dquot_writeback_dquots(sb
, -1);
4733 * Data writeback is possible w/o journal transaction, so barrier must
4734 * being sent at the end of the function. But we can skip it if
4735 * transaction_commit will do it for us.
4737 if (sbi
->s_journal
) {
4738 target
= jbd2_get_latest_transaction(sbi
->s_journal
);
4739 if (wait
&& sbi
->s_journal
->j_flags
& JBD2_BARRIER
&&
4740 !jbd2_trans_will_send_data_barrier(sbi
->s_journal
, target
))
4741 needs_barrier
= true;
4743 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
4745 ret
= jbd2_log_wait_commit(sbi
->s_journal
,
4748 } else if (wait
&& test_opt(sb
, BARRIER
))
4749 needs_barrier
= true;
4750 if (needs_barrier
) {
4752 err
= blkdev_issue_flush(sb
->s_bdev
, GFP_KERNEL
, NULL
);
4761 * LVM calls this function before a (read-only) snapshot is created. This
4762 * gives us a chance to flush the journal completely and mark the fs clean.
4764 * Note that only this function cannot bring a filesystem to be in a clean
4765 * state independently. It relies on upper layer to stop all data & metadata
4768 static int ext4_freeze(struct super_block
*sb
)
4773 if (sb
->s_flags
& MS_RDONLY
)
4776 journal
= EXT4_SB(sb
)->s_journal
;
4779 /* Now we set up the journal barrier. */
4780 jbd2_journal_lock_updates(journal
);
4783 * Don't clear the needs_recovery flag if we failed to
4784 * flush the journal.
4786 error
= jbd2_journal_flush(journal
);
4791 /* Journal blocked and flushed, clear needs_recovery flag. */
4792 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4793 error
= ext4_commit_super(sb
, 1);
4796 /* we rely on upper layer to stop further updates */
4797 jbd2_journal_unlock_updates(journal
);
4802 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4803 * flag here, even though the filesystem is not technically dirty yet.
4805 static int ext4_unfreeze(struct super_block
*sb
)
4807 if (sb
->s_flags
& MS_RDONLY
)
4810 /* Reset the needs_recovery flag before the fs is unlocked. */
4811 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4812 ext4_commit_super(sb
, 1);
4817 * Structure to save mount options for ext4_remount's benefit
4819 struct ext4_mount_options
{
4820 unsigned long s_mount_opt
;
4821 unsigned long s_mount_opt2
;
4824 unsigned long s_commit_interval
;
4825 u32 s_min_batch_time
, s_max_batch_time
;
4828 char *s_qf_names
[EXT4_MAXQUOTAS
];
4832 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
4834 struct ext4_super_block
*es
;
4835 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4836 unsigned long old_sb_flags
;
4837 struct ext4_mount_options old_opts
;
4838 int enable_quota
= 0;
4840 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
4845 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
4847 /* Store the original options */
4848 old_sb_flags
= sb
->s_flags
;
4849 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
4850 old_opts
.s_mount_opt2
= sbi
->s_mount_opt2
;
4851 old_opts
.s_resuid
= sbi
->s_resuid
;
4852 old_opts
.s_resgid
= sbi
->s_resgid
;
4853 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
4854 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
4855 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
4857 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
4858 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
4859 if (sbi
->s_qf_names
[i
]) {
4860 old_opts
.s_qf_names
[i
] = kstrdup(sbi
->s_qf_names
[i
],
4862 if (!old_opts
.s_qf_names
[i
]) {
4863 for (j
= 0; j
< i
; j
++)
4864 kfree(old_opts
.s_qf_names
[j
]);
4869 old_opts
.s_qf_names
[i
] = NULL
;
4871 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
4872 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
4874 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
, 1)) {
4879 if ((old_opts
.s_mount_opt
& EXT4_MOUNT_JOURNAL_CHECKSUM
) ^
4880 test_opt(sb
, JOURNAL_CHECKSUM
)) {
4881 ext4_msg(sb
, KERN_ERR
, "changing journal_checksum "
4882 "during remount not supported; ignoring");
4883 sbi
->s_mount_opt
^= EXT4_MOUNT_JOURNAL_CHECKSUM
;
4886 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
4887 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
4888 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4889 "both data=journal and delalloc");
4893 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
4894 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4895 "both data=journal and dioread_nolock");
4899 if (test_opt(sb
, DAX
)) {
4900 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4901 "both data=journal and dax");
4907 if ((sbi
->s_mount_opt
^ old_opts
.s_mount_opt
) & EXT4_MOUNT_DAX
) {
4908 ext4_msg(sb
, KERN_WARNING
, "warning: refusing change of "
4909 "dax flag with busy inodes while remounting");
4910 sbi
->s_mount_opt
^= EXT4_MOUNT_DAX
;
4913 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
)
4914 ext4_abort(sb
, "Abort forced by user");
4916 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
4917 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
4921 if (sbi
->s_journal
) {
4922 ext4_init_journal_params(sb
, sbi
->s_journal
);
4923 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
4926 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
)) {
4927 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
) {
4932 if (*flags
& MS_RDONLY
) {
4933 err
= sync_filesystem(sb
);
4936 err
= dquot_suspend(sb
, -1);
4941 * First of all, the unconditional stuff we have to do
4942 * to disable replay of the journal when we next remount
4944 sb
->s_flags
|= MS_RDONLY
;
4947 * OK, test if we are remounting a valid rw partition
4948 * readonly, and if so set the rdonly flag and then
4949 * mark the partition as valid again.
4951 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
4952 (sbi
->s_mount_state
& EXT4_VALID_FS
))
4953 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
4956 ext4_mark_recovery_complete(sb
, es
);
4958 /* Make sure we can mount this feature set readwrite */
4959 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
4960 EXT4_FEATURE_RO_COMPAT_READONLY
) ||
4961 !ext4_feature_set_ok(sb
, 0)) {
4966 * Make sure the group descriptor checksums
4967 * are sane. If they aren't, refuse to remount r/w.
4969 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
4970 struct ext4_group_desc
*gdp
=
4971 ext4_get_group_desc(sb
, g
, NULL
);
4973 if (!ext4_group_desc_csum_verify(sb
, g
, gdp
)) {
4974 ext4_msg(sb
, KERN_ERR
,
4975 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4976 g
, le16_to_cpu(ext4_group_desc_csum(sbi
, g
, gdp
)),
4977 le16_to_cpu(gdp
->bg_checksum
));
4984 * If we have an unprocessed orphan list hanging
4985 * around from a previously readonly bdev mount,
4986 * require a full umount/remount for now.
4988 if (es
->s_last_orphan
) {
4989 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
4990 "remount RDWR because of unprocessed "
4991 "orphan inode list. Please "
4992 "umount/remount instead");
4998 * Mounting a RDONLY partition read-write, so reread
4999 * and store the current valid flag. (It may have
5000 * been changed by e2fsck since we originally mounted
5004 ext4_clear_journal_err(sb
, es
);
5005 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
5006 if (!ext4_setup_super(sb
, es
, 0))
5007 sb
->s_flags
&= ~MS_RDONLY
;
5008 if (EXT4_HAS_INCOMPAT_FEATURE(sb
,
5009 EXT4_FEATURE_INCOMPAT_MMP
))
5010 if (ext4_multi_mount_protect(sb
,
5011 le64_to_cpu(es
->s_mmp_block
))) {
5020 * Reinitialize lazy itable initialization thread based on
5023 if ((sb
->s_flags
& MS_RDONLY
) || !test_opt(sb
, INIT_INODE_TABLE
))
5024 ext4_unregister_li_request(sb
);
5026 ext4_group_t first_not_zeroed
;
5027 first_not_zeroed
= ext4_has_uninit_itable(sb
);
5028 ext4_register_li_request(sb
, first_not_zeroed
);
5031 ext4_setup_system_zone(sb
);
5032 if (sbi
->s_journal
== NULL
&& !(old_sb_flags
& MS_RDONLY
))
5033 ext4_commit_super(sb
, 1);
5036 /* Release old quota file names */
5037 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
5038 kfree(old_opts
.s_qf_names
[i
]);
5040 if (sb_any_quota_suspended(sb
))
5041 dquot_resume(sb
, -1);
5042 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
5043 EXT4_FEATURE_RO_COMPAT_QUOTA
)) {
5044 err
= ext4_enable_quotas(sb
);
5051 *flags
= (*flags
& ~MS_LAZYTIME
) | (sb
->s_flags
& MS_LAZYTIME
);
5052 ext4_msg(sb
, KERN_INFO
, "re-mounted. Opts: %s", orig_data
);
5057 sb
->s_flags
= old_sb_flags
;
5058 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
5059 sbi
->s_mount_opt2
= old_opts
.s_mount_opt2
;
5060 sbi
->s_resuid
= old_opts
.s_resuid
;
5061 sbi
->s_resgid
= old_opts
.s_resgid
;
5062 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
5063 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
5064 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
5066 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
5067 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
5068 kfree(sbi
->s_qf_names
[i
]);
5069 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
5076 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
5078 struct super_block
*sb
= dentry
->d_sb
;
5079 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
5080 struct ext4_super_block
*es
= sbi
->s_es
;
5081 ext4_fsblk_t overhead
= 0, resv_blocks
;
5084 resv_blocks
= EXT4_C2B(sbi
, atomic64_read(&sbi
->s_resv_clusters
));
5086 if (!test_opt(sb
, MINIX_DF
))
5087 overhead
= sbi
->s_overhead
;
5089 buf
->f_type
= EXT4_SUPER_MAGIC
;
5090 buf
->f_bsize
= sb
->s_blocksize
;
5091 buf
->f_blocks
= ext4_blocks_count(es
) - EXT4_C2B(sbi
, overhead
);
5092 bfree
= percpu_counter_sum_positive(&sbi
->s_freeclusters_counter
) -
5093 percpu_counter_sum_positive(&sbi
->s_dirtyclusters_counter
);
5094 /* prevent underflow in case that few free space is available */
5095 buf
->f_bfree
= EXT4_C2B(sbi
, max_t(s64
, bfree
, 0));
5096 buf
->f_bavail
= buf
->f_bfree
-
5097 (ext4_r_blocks_count(es
) + resv_blocks
);
5098 if (buf
->f_bfree
< (ext4_r_blocks_count(es
) + resv_blocks
))
5100 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
5101 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
5102 buf
->f_namelen
= EXT4_NAME_LEN
;
5103 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
5104 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
5105 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
5106 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
5111 /* Helper function for writing quotas on sync - we need to start transaction
5112 * before quota file is locked for write. Otherwise the are possible deadlocks:
5113 * Process 1 Process 2
5114 * ext4_create() quota_sync()
5115 * jbd2_journal_start() write_dquot()
5116 * dquot_initialize() down(dqio_mutex)
5117 * down(dqio_mutex) jbd2_journal_start()
5123 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
5125 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_id
.type
];
5128 static int ext4_write_dquot(struct dquot
*dquot
)
5132 struct inode
*inode
;
5134 inode
= dquot_to_inode(dquot
);
5135 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
,
5136 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
5138 return PTR_ERR(handle
);
5139 ret
= dquot_commit(dquot
);
5140 err
= ext4_journal_stop(handle
);
5146 static int ext4_acquire_dquot(struct dquot
*dquot
)
5151 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
5152 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
5154 return PTR_ERR(handle
);
5155 ret
= dquot_acquire(dquot
);
5156 err
= ext4_journal_stop(handle
);
5162 static int ext4_release_dquot(struct dquot
*dquot
)
5167 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
5168 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
5169 if (IS_ERR(handle
)) {
5170 /* Release dquot anyway to avoid endless cycle in dqput() */
5171 dquot_release(dquot
);
5172 return PTR_ERR(handle
);
5174 ret
= dquot_release(dquot
);
5175 err
= ext4_journal_stop(handle
);
5181 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
5183 struct super_block
*sb
= dquot
->dq_sb
;
5184 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
5186 /* Are we journaling quotas? */
5187 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
) ||
5188 sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
5189 dquot_mark_dquot_dirty(dquot
);
5190 return ext4_write_dquot(dquot
);
5192 return dquot_mark_dquot_dirty(dquot
);
5196 static int ext4_write_info(struct super_block
*sb
, int type
)
5201 /* Data block + inode block */
5202 handle
= ext4_journal_start(sb
->s_root
->d_inode
, EXT4_HT_QUOTA
, 2);
5204 return PTR_ERR(handle
);
5205 ret
= dquot_commit_info(sb
, type
);
5206 err
= ext4_journal_stop(handle
);
5213 * Turn on quotas during mount time - we need to find
5214 * the quota file and such...
5216 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
5218 return dquot_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
5219 EXT4_SB(sb
)->s_jquota_fmt
, type
);
5223 * Standard function to be called on quota_on
5225 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
5230 if (!test_opt(sb
, QUOTA
))
5233 /* Quotafile not on the same filesystem? */
5234 if (path
->dentry
->d_sb
!= sb
)
5236 /* Journaling quota? */
5237 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
5238 /* Quotafile not in fs root? */
5239 if (path
->dentry
->d_parent
!= sb
->s_root
)
5240 ext4_msg(sb
, KERN_WARNING
,
5241 "Quota file not on filesystem root. "
5242 "Journaled quota will not work");
5246 * When we journal data on quota file, we have to flush journal to see
5247 * all updates to the file when we bypass pagecache...
5249 if (EXT4_SB(sb
)->s_journal
&&
5250 ext4_should_journal_data(path
->dentry
->d_inode
)) {
5252 * We don't need to lock updates but journal_flush() could
5253 * otherwise be livelocked...
5255 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
5256 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
5257 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
5262 return dquot_quota_on(sb
, type
, format_id
, path
);
5265 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
5269 struct inode
*qf_inode
;
5270 unsigned long qf_inums
[EXT4_MAXQUOTAS
] = {
5271 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
5272 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
)
5275 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
));
5277 if (!qf_inums
[type
])
5280 qf_inode
= ext4_iget(sb
, qf_inums
[type
]);
5281 if (IS_ERR(qf_inode
)) {
5282 ext4_error(sb
, "Bad quota inode # %lu", qf_inums
[type
]);
5283 return PTR_ERR(qf_inode
);
5286 /* Don't account quota for quota files to avoid recursion */
5287 qf_inode
->i_flags
|= S_NOQUOTA
;
5288 err
= dquot_enable(qf_inode
, type
, format_id
, flags
);
5294 /* Enable usage tracking for all quota types. */
5295 static int ext4_enable_quotas(struct super_block
*sb
)
5298 unsigned long qf_inums
[EXT4_MAXQUOTAS
] = {
5299 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
5300 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
)
5303 sb_dqopt(sb
)->flags
|= DQUOT_QUOTA_SYS_FILE
;
5304 for (type
= 0; type
< EXT4_MAXQUOTAS
; type
++) {
5305 if (qf_inums
[type
]) {
5306 err
= ext4_quota_enable(sb
, type
, QFMT_VFS_V1
,
5307 DQUOT_USAGE_ENABLED
);
5310 "Failed to enable quota tracking "
5311 "(type=%d, err=%d). Please run "
5312 "e2fsck to fix.", type
, err
);
5320 static int ext4_quota_off(struct super_block
*sb
, int type
)
5322 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5325 /* Force all delayed allocation blocks to be allocated.
5326 * Caller already holds s_umount sem */
5327 if (test_opt(sb
, DELALLOC
))
5328 sync_filesystem(sb
);
5333 /* Update modification times of quota files when userspace can
5334 * start looking at them */
5335 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
, 1);
5338 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
5339 ext4_mark_inode_dirty(handle
, inode
);
5340 ext4_journal_stop(handle
);
5343 return dquot_quota_off(sb
, type
);
5346 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5347 * acquiring the locks... As quota files are never truncated and quota code
5348 * itself serializes the operations (and no one else should touch the files)
5349 * we don't have to be afraid of races */
5350 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
5351 size_t len
, loff_t off
)
5353 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5354 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5355 int offset
= off
& (sb
->s_blocksize
- 1);
5358 struct buffer_head
*bh
;
5359 loff_t i_size
= i_size_read(inode
);
5363 if (off
+len
> i_size
)
5366 while (toread
> 0) {
5367 tocopy
= sb
->s_blocksize
- offset
< toread
?
5368 sb
->s_blocksize
- offset
: toread
;
5369 bh
= ext4_bread(NULL
, inode
, blk
, 0);
5372 if (!bh
) /* A hole? */
5373 memset(data
, 0, tocopy
);
5375 memcpy(data
, bh
->b_data
+offset
, tocopy
);
5385 /* Write to quotafile (we know the transaction is already started and has
5386 * enough credits) */
5387 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
5388 const char *data
, size_t len
, loff_t off
)
5390 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5391 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5392 int err
, offset
= off
& (sb
->s_blocksize
- 1);
5393 struct buffer_head
*bh
;
5394 handle_t
*handle
= journal_current_handle();
5396 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
5397 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5398 " cancelled because transaction is not started",
5399 (unsigned long long)off
, (unsigned long long)len
);
5403 * Since we account only one data block in transaction credits,
5404 * then it is impossible to cross a block boundary.
5406 if (sb
->s_blocksize
- offset
< len
) {
5407 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5408 " cancelled because not block aligned",
5409 (unsigned long long)off
, (unsigned long long)len
);
5413 bh
= ext4_bread(handle
, inode
, blk
, 1);
5418 BUFFER_TRACE(bh
, "get write access");
5419 err
= ext4_journal_get_write_access(handle
, bh
);
5425 memcpy(bh
->b_data
+offset
, data
, len
);
5426 flush_dcache_page(bh
->b_page
);
5428 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
5431 if (inode
->i_size
< off
+ len
) {
5432 i_size_write(inode
, off
+ len
);
5433 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
5434 ext4_mark_inode_dirty(handle
, inode
);
5441 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
5442 const char *dev_name
, void *data
)
5444 return mount_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
);
5447 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5448 static inline void register_as_ext2(void)
5450 int err
= register_filesystem(&ext2_fs_type
);
5453 "EXT4-fs: Unable to register as ext2 (%d)\n", err
);
5456 static inline void unregister_as_ext2(void)
5458 unregister_filesystem(&ext2_fs_type
);
5461 static inline int ext2_feature_set_ok(struct super_block
*sb
)
5463 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT2_FEATURE_INCOMPAT_SUPP
))
5465 if (sb
->s_flags
& MS_RDONLY
)
5467 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT2_FEATURE_RO_COMPAT_SUPP
))
5472 static inline void register_as_ext2(void) { }
5473 static inline void unregister_as_ext2(void) { }
5474 static inline int ext2_feature_set_ok(struct super_block
*sb
) { return 0; }
5477 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5478 static inline void register_as_ext3(void)
5480 int err
= register_filesystem(&ext3_fs_type
);
5483 "EXT4-fs: Unable to register as ext3 (%d)\n", err
);
5486 static inline void unregister_as_ext3(void)
5488 unregister_filesystem(&ext3_fs_type
);
5491 static inline int ext3_feature_set_ok(struct super_block
*sb
)
5493 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT3_FEATURE_INCOMPAT_SUPP
))
5495 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
5497 if (sb
->s_flags
& MS_RDONLY
)
5499 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT3_FEATURE_RO_COMPAT_SUPP
))
5504 static inline void register_as_ext3(void) { }
5505 static inline void unregister_as_ext3(void) { }
5506 static inline int ext3_feature_set_ok(struct super_block
*sb
) { return 0; }
5509 static struct file_system_type ext4_fs_type
= {
5510 .owner
= THIS_MODULE
,
5512 .mount
= ext4_mount
,
5513 .kill_sb
= kill_block_super
,
5514 .fs_flags
= FS_REQUIRES_DEV
,
5516 MODULE_ALIAS_FS("ext4");
5518 static int __init
ext4_init_feat_adverts(void)
5520 struct ext4_features
*ef
;
5523 ef
= kzalloc(sizeof(struct ext4_features
), GFP_KERNEL
);
5527 ef
->f_kobj
.kset
= ext4_kset
;
5528 init_completion(&ef
->f_kobj_unregister
);
5529 ret
= kobject_init_and_add(&ef
->f_kobj
, &ext4_feat_ktype
, NULL
,
5542 static void ext4_exit_feat_adverts(void)
5544 kobject_put(&ext4_feat
->f_kobj
);
5545 wait_for_completion(&ext4_feat
->f_kobj_unregister
);
5549 /* Shared across all ext4 file systems */
5550 wait_queue_head_t ext4__ioend_wq
[EXT4_WQ_HASH_SZ
];
5551 struct mutex ext4__aio_mutex
[EXT4_WQ_HASH_SZ
];
5553 static int __init
ext4_init_fs(void)
5557 ext4_li_info
= NULL
;
5558 mutex_init(&ext4_li_mtx
);
5560 /* Build-time check for flags consistency */
5561 ext4_check_flag_values();
5563 for (i
= 0; i
< EXT4_WQ_HASH_SZ
; i
++) {
5564 mutex_init(&ext4__aio_mutex
[i
]);
5565 init_waitqueue_head(&ext4__ioend_wq
[i
]);
5568 err
= ext4_init_es();
5572 err
= ext4_init_pageio();
5576 err
= ext4_init_system_zone();
5579 ext4_kset
= kset_create_and_add("ext4", NULL
, fs_kobj
);
5584 ext4_proc_root
= proc_mkdir("fs/ext4", NULL
);
5586 err
= ext4_init_feat_adverts();
5590 err
= ext4_init_mballoc();
5594 ext4_mballoc_ready
= 1;
5595 err
= init_inodecache();
5600 err
= register_filesystem(&ext4_fs_type
);
5606 unregister_as_ext2();
5607 unregister_as_ext3();
5608 destroy_inodecache();
5610 ext4_mballoc_ready
= 0;
5611 ext4_exit_mballoc();
5613 ext4_exit_feat_adverts();
5616 remove_proc_entry("fs/ext4", NULL
);
5617 kset_unregister(ext4_kset
);
5619 ext4_exit_system_zone();
5628 static void __exit
ext4_exit_fs(void)
5630 ext4_destroy_lazyinit_thread();
5631 unregister_as_ext2();
5632 unregister_as_ext3();
5633 unregister_filesystem(&ext4_fs_type
);
5634 destroy_inodecache();
5635 ext4_exit_mballoc();
5636 ext4_exit_feat_adverts();
5637 remove_proc_entry("fs/ext4", NULL
);
5638 kset_unregister(ext4_kset
);
5639 ext4_exit_system_zone();
5644 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5645 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5646 MODULE_LICENSE("GPL");
5647 module_init(ext4_init_fs
)
5648 module_exit(ext4_exit_fs
)