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
;
63 static int ext4_load_journal(struct super_block
*, struct ext4_super_block
*,
64 unsigned long journal_devnum
);
65 static int ext4_show_options(struct seq_file
*seq
, struct dentry
*root
);
66 static int ext4_commit_super(struct super_block
*sb
, int sync
);
67 static void ext4_mark_recovery_complete(struct super_block
*sb
,
68 struct ext4_super_block
*es
);
69 static void ext4_clear_journal_err(struct super_block
*sb
,
70 struct ext4_super_block
*es
);
71 static int ext4_sync_fs(struct super_block
*sb
, int wait
);
72 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
);
73 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
74 static int ext4_unfreeze(struct super_block
*sb
);
75 static int ext4_freeze(struct super_block
*sb
);
76 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
77 const char *dev_name
, void *data
);
78 static inline int ext2_feature_set_ok(struct super_block
*sb
);
79 static inline int ext3_feature_set_ok(struct super_block
*sb
);
80 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
);
81 static void ext4_destroy_lazyinit_thread(void);
82 static void ext4_unregister_li_request(struct super_block
*sb
);
83 static void ext4_clear_request_list(void);
85 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
86 static struct file_system_type ext2_fs_type
= {
90 .kill_sb
= kill_block_super
,
91 .fs_flags
= FS_REQUIRES_DEV
,
93 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
95 #define IS_EXT2_SB(sb) (0)
99 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
100 static struct file_system_type ext3_fs_type
= {
101 .owner
= THIS_MODULE
,
104 .kill_sb
= kill_block_super
,
105 .fs_flags
= FS_REQUIRES_DEV
,
107 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
109 #define IS_EXT3_SB(sb) (0)
112 static int ext4_verify_csum_type(struct super_block
*sb
,
113 struct ext4_super_block
*es
)
115 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
,
116 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
))
119 return es
->s_checksum_type
== EXT4_CRC32C_CHKSUM
;
122 static __le32
ext4_superblock_csum(struct super_block
*sb
,
123 struct ext4_super_block
*es
)
125 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
126 int offset
= offsetof(struct ext4_super_block
, s_checksum
);
129 csum
= ext4_chksum(sbi
, ~0, (char *)es
, offset
);
131 return cpu_to_le32(csum
);
134 int ext4_superblock_csum_verify(struct super_block
*sb
,
135 struct ext4_super_block
*es
)
137 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
,
138 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
))
141 return es
->s_checksum
== ext4_superblock_csum(sb
, es
);
144 void ext4_superblock_csum_set(struct super_block
*sb
)
146 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
148 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
,
149 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
))
152 es
->s_checksum
= ext4_superblock_csum(sb
, es
);
155 void *ext4_kvmalloc(size_t size
, gfp_t flags
)
159 ret
= kmalloc(size
, flags
);
161 ret
= __vmalloc(size
, flags
, PAGE_KERNEL
);
165 void *ext4_kvzalloc(size_t size
, gfp_t flags
)
169 ret
= kzalloc(size
, flags
);
171 ret
= __vmalloc(size
, flags
| __GFP_ZERO
, PAGE_KERNEL
);
175 void ext4_kvfree(void *ptr
)
177 if (is_vmalloc_addr(ptr
))
184 ext4_fsblk_t
ext4_block_bitmap(struct super_block
*sb
,
185 struct ext4_group_desc
*bg
)
187 return le32_to_cpu(bg
->bg_block_bitmap_lo
) |
188 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
189 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_block_bitmap_hi
) << 32 : 0);
192 ext4_fsblk_t
ext4_inode_bitmap(struct super_block
*sb
,
193 struct ext4_group_desc
*bg
)
195 return le32_to_cpu(bg
->bg_inode_bitmap_lo
) |
196 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
197 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_bitmap_hi
) << 32 : 0);
200 ext4_fsblk_t
ext4_inode_table(struct super_block
*sb
,
201 struct ext4_group_desc
*bg
)
203 return le32_to_cpu(bg
->bg_inode_table_lo
) |
204 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
205 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_table_hi
) << 32 : 0);
208 __u32
ext4_free_group_clusters(struct super_block
*sb
,
209 struct ext4_group_desc
*bg
)
211 return le16_to_cpu(bg
->bg_free_blocks_count_lo
) |
212 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
213 (__u32
)le16_to_cpu(bg
->bg_free_blocks_count_hi
) << 16 : 0);
216 __u32
ext4_free_inodes_count(struct super_block
*sb
,
217 struct ext4_group_desc
*bg
)
219 return le16_to_cpu(bg
->bg_free_inodes_count_lo
) |
220 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
221 (__u32
)le16_to_cpu(bg
->bg_free_inodes_count_hi
) << 16 : 0);
224 __u32
ext4_used_dirs_count(struct super_block
*sb
,
225 struct ext4_group_desc
*bg
)
227 return le16_to_cpu(bg
->bg_used_dirs_count_lo
) |
228 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
229 (__u32
)le16_to_cpu(bg
->bg_used_dirs_count_hi
) << 16 : 0);
232 __u32
ext4_itable_unused_count(struct super_block
*sb
,
233 struct ext4_group_desc
*bg
)
235 return le16_to_cpu(bg
->bg_itable_unused_lo
) |
236 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
237 (__u32
)le16_to_cpu(bg
->bg_itable_unused_hi
) << 16 : 0);
240 void ext4_block_bitmap_set(struct super_block
*sb
,
241 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
243 bg
->bg_block_bitmap_lo
= cpu_to_le32((u32
)blk
);
244 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
245 bg
->bg_block_bitmap_hi
= cpu_to_le32(blk
>> 32);
248 void ext4_inode_bitmap_set(struct super_block
*sb
,
249 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
251 bg
->bg_inode_bitmap_lo
= cpu_to_le32((u32
)blk
);
252 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
253 bg
->bg_inode_bitmap_hi
= cpu_to_le32(blk
>> 32);
256 void ext4_inode_table_set(struct super_block
*sb
,
257 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
259 bg
->bg_inode_table_lo
= cpu_to_le32((u32
)blk
);
260 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
261 bg
->bg_inode_table_hi
= cpu_to_le32(blk
>> 32);
264 void ext4_free_group_clusters_set(struct super_block
*sb
,
265 struct ext4_group_desc
*bg
, __u32 count
)
267 bg
->bg_free_blocks_count_lo
= cpu_to_le16((__u16
)count
);
268 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
269 bg
->bg_free_blocks_count_hi
= cpu_to_le16(count
>> 16);
272 void ext4_free_inodes_set(struct super_block
*sb
,
273 struct ext4_group_desc
*bg
, __u32 count
)
275 bg
->bg_free_inodes_count_lo
= cpu_to_le16((__u16
)count
);
276 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
277 bg
->bg_free_inodes_count_hi
= cpu_to_le16(count
>> 16);
280 void ext4_used_dirs_set(struct super_block
*sb
,
281 struct ext4_group_desc
*bg
, __u32 count
)
283 bg
->bg_used_dirs_count_lo
= cpu_to_le16((__u16
)count
);
284 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
285 bg
->bg_used_dirs_count_hi
= cpu_to_le16(count
>> 16);
288 void ext4_itable_unused_set(struct super_block
*sb
,
289 struct ext4_group_desc
*bg
, __u32 count
)
291 bg
->bg_itable_unused_lo
= cpu_to_le16((__u16
)count
);
292 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
293 bg
->bg_itable_unused_hi
= cpu_to_le16(count
>> 16);
297 static void __save_error_info(struct super_block
*sb
, const char *func
,
300 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
302 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
303 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
304 es
->s_last_error_time
= cpu_to_le32(get_seconds());
305 strncpy(es
->s_last_error_func
, func
, sizeof(es
->s_last_error_func
));
306 es
->s_last_error_line
= cpu_to_le32(line
);
307 if (!es
->s_first_error_time
) {
308 es
->s_first_error_time
= es
->s_last_error_time
;
309 strncpy(es
->s_first_error_func
, func
,
310 sizeof(es
->s_first_error_func
));
311 es
->s_first_error_line
= cpu_to_le32(line
);
312 es
->s_first_error_ino
= es
->s_last_error_ino
;
313 es
->s_first_error_block
= es
->s_last_error_block
;
316 * Start the daily error reporting function if it hasn't been
319 if (!es
->s_error_count
)
320 mod_timer(&EXT4_SB(sb
)->s_err_report
, jiffies
+ 24*60*60*HZ
);
321 le32_add_cpu(&es
->s_error_count
, 1);
324 static void save_error_info(struct super_block
*sb
, const char *func
,
327 __save_error_info(sb
, func
, line
);
328 ext4_commit_super(sb
, 1);
332 * The del_gendisk() function uninitializes the disk-specific data
333 * structures, including the bdi structure, without telling anyone
334 * else. Once this happens, any attempt to call mark_buffer_dirty()
335 * (for example, by ext4_commit_super), will cause a kernel OOPS.
336 * This is a kludge to prevent these oops until we can put in a proper
337 * hook in del_gendisk() to inform the VFS and file system layers.
339 static int block_device_ejected(struct super_block
*sb
)
341 struct inode
*bd_inode
= sb
->s_bdev
->bd_inode
;
342 struct backing_dev_info
*bdi
= bd_inode
->i_mapping
->backing_dev_info
;
344 return bdi
->dev
== NULL
;
347 static void ext4_journal_commit_callback(journal_t
*journal
, transaction_t
*txn
)
349 struct super_block
*sb
= journal
->j_private
;
350 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
351 int error
= is_journal_aborted(journal
);
352 struct ext4_journal_cb_entry
*jce
, *tmp
;
354 spin_lock(&sbi
->s_md_lock
);
355 list_for_each_entry_safe(jce
, tmp
, &txn
->t_private_list
, jce_list
) {
356 list_del_init(&jce
->jce_list
);
357 spin_unlock(&sbi
->s_md_lock
);
358 jce
->jce_func(sb
, jce
, error
);
359 spin_lock(&sbi
->s_md_lock
);
361 spin_unlock(&sbi
->s_md_lock
);
364 /* Deal with the reporting of failure conditions on a filesystem such as
365 * inconsistencies detected or read IO failures.
367 * On ext2, we can store the error state of the filesystem in the
368 * superblock. That is not possible on ext4, because we may have other
369 * write ordering constraints on the superblock which prevent us from
370 * writing it out straight away; and given that the journal is about to
371 * be aborted, we can't rely on the current, or future, transactions to
372 * write out the superblock safely.
374 * We'll just use the jbd2_journal_abort() error code to record an error in
375 * the journal instead. On recovery, the journal will complain about
376 * that error until we've noted it down and cleared it.
379 static void ext4_handle_error(struct super_block
*sb
)
381 if (sb
->s_flags
& MS_RDONLY
)
384 if (!test_opt(sb
, ERRORS_CONT
)) {
385 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
387 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
389 jbd2_journal_abort(journal
, -EIO
);
391 if (test_opt(sb
, ERRORS_RO
)) {
392 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
393 sb
->s_flags
|= MS_RDONLY
;
395 if (test_opt(sb
, ERRORS_PANIC
))
396 panic("EXT4-fs (device %s): panic forced after error\n",
400 void __ext4_error(struct super_block
*sb
, const char *function
,
401 unsigned int line
, const char *fmt
, ...)
403 struct va_format vaf
;
409 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
410 sb
->s_id
, function
, line
, current
->comm
, &vaf
);
412 save_error_info(sb
, function
, line
);
414 ext4_handle_error(sb
);
417 void ext4_error_inode(struct inode
*inode
, const char *function
,
418 unsigned int line
, ext4_fsblk_t block
,
419 const char *fmt
, ...)
422 struct va_format vaf
;
423 struct ext4_super_block
*es
= EXT4_SB(inode
->i_sb
)->s_es
;
425 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
426 es
->s_last_error_block
= cpu_to_le64(block
);
427 save_error_info(inode
->i_sb
, function
, line
);
432 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
433 "inode #%lu: block %llu: comm %s: %pV\n",
434 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
435 block
, current
->comm
, &vaf
);
437 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
438 "inode #%lu: comm %s: %pV\n",
439 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
440 current
->comm
, &vaf
);
443 ext4_handle_error(inode
->i_sb
);
446 void ext4_error_file(struct file
*file
, const char *function
,
447 unsigned int line
, ext4_fsblk_t block
,
448 const char *fmt
, ...)
451 struct va_format vaf
;
452 struct ext4_super_block
*es
;
453 struct inode
*inode
= file
->f_dentry
->d_inode
;
454 char pathname
[80], *path
;
456 es
= EXT4_SB(inode
->i_sb
)->s_es
;
457 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
458 save_error_info(inode
->i_sb
, function
, line
);
459 path
= d_path(&(file
->f_path
), pathname
, sizeof(pathname
));
467 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
468 "block %llu: comm %s: path %s: %pV\n",
469 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
470 block
, current
->comm
, path
, &vaf
);
473 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
474 "comm %s: path %s: %pV\n",
475 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
476 current
->comm
, path
, &vaf
);
479 ext4_handle_error(inode
->i_sb
);
482 const char *ext4_decode_error(struct super_block
*sb
, int errno
,
489 errstr
= "IO failure";
492 errstr
= "Out of memory";
495 if (!sb
|| (EXT4_SB(sb
)->s_journal
&&
496 EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
))
497 errstr
= "Journal has aborted";
499 errstr
= "Readonly filesystem";
502 /* If the caller passed in an extra buffer for unknown
503 * errors, textualise them now. Else we just return
506 /* Check for truncated error codes... */
507 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
516 /* __ext4_std_error decodes expected errors from journaling functions
517 * automatically and invokes the appropriate error response. */
519 void __ext4_std_error(struct super_block
*sb
, const char *function
,
520 unsigned int line
, int errno
)
525 /* Special case: if the error is EROFS, and we're not already
526 * inside a transaction, then there's really no point in logging
528 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
529 (sb
->s_flags
& MS_RDONLY
))
532 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
533 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s:%d: %s\n",
534 sb
->s_id
, function
, line
, errstr
);
535 save_error_info(sb
, function
, line
);
537 ext4_handle_error(sb
);
541 * ext4_abort is a much stronger failure handler than ext4_error. The
542 * abort function may be used to deal with unrecoverable failures such
543 * as journal IO errors or ENOMEM at a critical moment in log management.
545 * We unconditionally force the filesystem into an ABORT|READONLY state,
546 * unless the error response on the fs has been set to panic in which
547 * case we take the easy way out and panic immediately.
550 void __ext4_abort(struct super_block
*sb
, const char *function
,
551 unsigned int line
, const char *fmt
, ...)
555 save_error_info(sb
, function
, line
);
557 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: ", sb
->s_id
,
563 if ((sb
->s_flags
& MS_RDONLY
) == 0) {
564 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
565 sb
->s_flags
|= MS_RDONLY
;
566 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
567 if (EXT4_SB(sb
)->s_journal
)
568 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
569 save_error_info(sb
, function
, line
);
571 if (test_opt(sb
, ERRORS_PANIC
))
572 panic("EXT4-fs panic from previous error\n");
575 void ext4_msg(struct super_block
*sb
, const char *prefix
, const char *fmt
, ...)
577 struct va_format vaf
;
583 printk("%sEXT4-fs (%s): %pV\n", prefix
, sb
->s_id
, &vaf
);
587 void __ext4_warning(struct super_block
*sb
, const char *function
,
588 unsigned int line
, const char *fmt
, ...)
590 struct va_format vaf
;
596 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: %pV\n",
597 sb
->s_id
, function
, line
, &vaf
);
601 void __ext4_grp_locked_error(const char *function
, unsigned int line
,
602 struct super_block
*sb
, ext4_group_t grp
,
603 unsigned long ino
, ext4_fsblk_t block
,
604 const char *fmt
, ...)
608 struct va_format vaf
;
610 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
612 es
->s_last_error_ino
= cpu_to_le32(ino
);
613 es
->s_last_error_block
= cpu_to_le64(block
);
614 __save_error_info(sb
, function
, line
);
620 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: group %u, ",
621 sb
->s_id
, function
, line
, grp
);
623 printk(KERN_CONT
"inode %lu: ", ino
);
625 printk(KERN_CONT
"block %llu:", (unsigned long long) block
);
626 printk(KERN_CONT
"%pV\n", &vaf
);
629 if (test_opt(sb
, ERRORS_CONT
)) {
630 ext4_commit_super(sb
, 0);
634 ext4_unlock_group(sb
, grp
);
635 ext4_handle_error(sb
);
637 * We only get here in the ERRORS_RO case; relocking the group
638 * may be dangerous, but nothing bad will happen since the
639 * filesystem will have already been marked read/only and the
640 * journal has been aborted. We return 1 as a hint to callers
641 * who might what to use the return value from
642 * ext4_grp_locked_error() to distinguish between the
643 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
644 * aggressively from the ext4 function in question, with a
645 * more appropriate error code.
647 ext4_lock_group(sb
, grp
);
651 void ext4_update_dynamic_rev(struct super_block
*sb
)
653 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
655 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
659 "updating to rev %d because of new feature flag, "
660 "running e2fsck is recommended",
663 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
664 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
665 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
666 /* leave es->s_feature_*compat flags alone */
667 /* es->s_uuid will be set by e2fsck if empty */
670 * The rest of the superblock fields should be zero, and if not it
671 * means they are likely already in use, so leave them alone. We
672 * can leave it up to e2fsck to clean up any inconsistencies there.
677 * Open the external journal device
679 static struct block_device
*ext4_blkdev_get(dev_t dev
, struct super_block
*sb
)
681 struct block_device
*bdev
;
682 char b
[BDEVNAME_SIZE
];
684 bdev
= blkdev_get_by_dev(dev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
, sb
);
690 ext4_msg(sb
, KERN_ERR
, "failed to open journal device %s: %ld",
691 __bdevname(dev
, b
), PTR_ERR(bdev
));
696 * Release the journal device
698 static int ext4_blkdev_put(struct block_device
*bdev
)
700 return blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
);
703 static int ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
705 struct block_device
*bdev
;
708 bdev
= sbi
->journal_bdev
;
710 ret
= ext4_blkdev_put(bdev
);
711 sbi
->journal_bdev
= NULL
;
716 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
718 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
721 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
725 ext4_msg(sb
, KERN_ERR
, "sb orphan head is %d",
726 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
728 printk(KERN_ERR
"sb_info orphan list:\n");
729 list_for_each(l
, &sbi
->s_orphan
) {
730 struct inode
*inode
= orphan_list_entry(l
);
732 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
733 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
734 inode
->i_mode
, inode
->i_nlink
,
739 static void ext4_put_super(struct super_block
*sb
)
741 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
742 struct ext4_super_block
*es
= sbi
->s_es
;
745 ext4_unregister_li_request(sb
);
746 dquot_disable(sb
, -1, DQUOT_USAGE_ENABLED
| DQUOT_LIMITS_ENABLED
);
748 flush_workqueue(sbi
->dio_unwritten_wq
);
749 destroy_workqueue(sbi
->dio_unwritten_wq
);
751 if (sbi
->s_journal
) {
752 err
= jbd2_journal_destroy(sbi
->s_journal
);
753 sbi
->s_journal
= NULL
;
755 ext4_abort(sb
, "Couldn't clean up the journal");
758 ext4_es_unregister_shrinker(sb
);
759 del_timer(&sbi
->s_err_report
);
760 ext4_release_system_zone(sb
);
762 ext4_ext_release(sb
);
763 ext4_xattr_put_super(sb
);
765 if (!(sb
->s_flags
& MS_RDONLY
)) {
766 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
767 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
769 if (!(sb
->s_flags
& MS_RDONLY
))
770 ext4_commit_super(sb
, 1);
773 remove_proc_entry("options", sbi
->s_proc
);
774 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
776 kobject_del(&sbi
->s_kobj
);
778 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
779 brelse(sbi
->s_group_desc
[i
]);
780 ext4_kvfree(sbi
->s_group_desc
);
781 ext4_kvfree(sbi
->s_flex_groups
);
782 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
783 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
784 percpu_counter_destroy(&sbi
->s_dirs_counter
);
785 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
786 percpu_counter_destroy(&sbi
->s_extent_cache_cnt
);
789 for (i
= 0; i
< MAXQUOTAS
; i
++)
790 kfree(sbi
->s_qf_names
[i
]);
793 /* Debugging code just in case the in-memory inode orphan list
794 * isn't empty. The on-disk one can be non-empty if we've
795 * detected an error and taken the fs readonly, but the
796 * in-memory list had better be clean by this point. */
797 if (!list_empty(&sbi
->s_orphan
))
798 dump_orphan_list(sb
, sbi
);
799 J_ASSERT(list_empty(&sbi
->s_orphan
));
801 invalidate_bdev(sb
->s_bdev
);
802 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
804 * Invalidate the journal device's buffers. We don't want them
805 * floating about in memory - the physical journal device may
806 * hotswapped, and it breaks the `ro-after' testing code.
808 sync_blockdev(sbi
->journal_bdev
);
809 invalidate_bdev(sbi
->journal_bdev
);
810 ext4_blkdev_remove(sbi
);
813 kthread_stop(sbi
->s_mmp_tsk
);
814 sb
->s_fs_info
= NULL
;
816 * Now that we are completely done shutting down the
817 * superblock, we need to actually destroy the kobject.
819 kobject_put(&sbi
->s_kobj
);
820 wait_for_completion(&sbi
->s_kobj_unregister
);
821 if (sbi
->s_chksum_driver
)
822 crypto_free_shash(sbi
->s_chksum_driver
);
823 kfree(sbi
->s_blockgroup_lock
);
827 static struct kmem_cache
*ext4_inode_cachep
;
830 * Called inside transaction, so use GFP_NOFS
832 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
834 struct ext4_inode_info
*ei
;
836 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
840 ei
->vfs_inode
.i_version
= 1;
841 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
842 spin_lock_init(&ei
->i_prealloc_lock
);
843 ext4_es_init_tree(&ei
->i_es_tree
);
844 rwlock_init(&ei
->i_es_lock
);
845 INIT_LIST_HEAD(&ei
->i_es_lru
);
847 ei
->i_reserved_data_blocks
= 0;
848 ei
->i_reserved_meta_blocks
= 0;
849 ei
->i_allocated_meta_blocks
= 0;
850 ei
->i_da_metadata_calc_len
= 0;
851 ei
->i_da_metadata_calc_last_lblock
= 0;
852 spin_lock_init(&(ei
->i_block_reservation_lock
));
854 ei
->i_reserved_quota
= 0;
857 INIT_LIST_HEAD(&ei
->i_completed_io_list
);
858 spin_lock_init(&ei
->i_completed_io_lock
);
860 ei
->i_datasync_tid
= 0;
861 atomic_set(&ei
->i_ioend_count
, 0);
862 atomic_set(&ei
->i_unwritten
, 0);
863 INIT_WORK(&ei
->i_unwritten_work
, ext4_end_io_work
);
865 return &ei
->vfs_inode
;
868 static int ext4_drop_inode(struct inode
*inode
)
870 int drop
= generic_drop_inode(inode
);
872 trace_ext4_drop_inode(inode
, drop
);
876 static void ext4_i_callback(struct rcu_head
*head
)
878 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
879 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
882 static void ext4_destroy_inode(struct inode
*inode
)
884 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
885 ext4_msg(inode
->i_sb
, KERN_ERR
,
886 "Inode %lu (%p): orphan list check failed!",
887 inode
->i_ino
, EXT4_I(inode
));
888 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
889 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
893 call_rcu(&inode
->i_rcu
, ext4_i_callback
);
896 static void init_once(void *foo
)
898 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
900 INIT_LIST_HEAD(&ei
->i_orphan
);
901 init_rwsem(&ei
->xattr_sem
);
902 init_rwsem(&ei
->i_data_sem
);
903 inode_init_once(&ei
->vfs_inode
);
906 static int init_inodecache(void)
908 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
909 sizeof(struct ext4_inode_info
),
910 0, (SLAB_RECLAIM_ACCOUNT
|
913 if (ext4_inode_cachep
== NULL
)
918 static void destroy_inodecache(void)
921 * Make sure all delayed rcu free inodes are flushed before we
925 kmem_cache_destroy(ext4_inode_cachep
);
928 void ext4_clear_inode(struct inode
*inode
)
930 invalidate_inode_buffers(inode
);
933 ext4_discard_preallocations(inode
);
934 ext4_es_remove_extent(inode
, 0, EXT_MAX_BLOCKS
);
935 ext4_es_lru_del(inode
);
936 if (EXT4_I(inode
)->jinode
) {
937 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode
),
938 EXT4_I(inode
)->jinode
);
939 jbd2_free_inode(EXT4_I(inode
)->jinode
);
940 EXT4_I(inode
)->jinode
= NULL
;
944 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
945 u64 ino
, u32 generation
)
949 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
950 return ERR_PTR(-ESTALE
);
951 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
952 return ERR_PTR(-ESTALE
);
954 /* iget isn't really right if the inode is currently unallocated!!
956 * ext4_read_inode will return a bad_inode if the inode had been
957 * deleted, so we should be safe.
959 * Currently we don't know the generation for parent directory, so
960 * a generation of 0 means "accept any"
962 inode
= ext4_iget(sb
, ino
);
964 return ERR_CAST(inode
);
965 if (generation
&& inode
->i_generation
!= generation
) {
967 return ERR_PTR(-ESTALE
);
973 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
974 int fh_len
, int fh_type
)
976 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
980 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
981 int fh_len
, int fh_type
)
983 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
988 * Try to release metadata pages (indirect blocks, directories) which are
989 * mapped via the block device. Since these pages could have journal heads
990 * which would prevent try_to_free_buffers() from freeing them, we must use
991 * jbd2 layer's try_to_free_buffers() function to release them.
993 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
996 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
998 WARN_ON(PageChecked(page
));
999 if (!page_has_buffers(page
))
1002 return jbd2_journal_try_to_free_buffers(journal
, page
,
1003 wait
& ~__GFP_WAIT
);
1004 return try_to_free_buffers(page
);
1008 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1009 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1011 static int ext4_write_dquot(struct dquot
*dquot
);
1012 static int ext4_acquire_dquot(struct dquot
*dquot
);
1013 static int ext4_release_dquot(struct dquot
*dquot
);
1014 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
1015 static int ext4_write_info(struct super_block
*sb
, int type
);
1016 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
1018 static int ext4_quota_on_sysfile(struct super_block
*sb
, int type
,
1020 static int ext4_quota_off(struct super_block
*sb
, int type
);
1021 static int ext4_quota_off_sysfile(struct super_block
*sb
, int type
);
1022 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
1023 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
1024 size_t len
, loff_t off
);
1025 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
1026 const char *data
, size_t len
, loff_t off
);
1027 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
1028 unsigned int flags
);
1029 static int ext4_enable_quotas(struct super_block
*sb
);
1031 static const struct dquot_operations ext4_quota_operations
= {
1032 .get_reserved_space
= ext4_get_reserved_space
,
1033 .write_dquot
= ext4_write_dquot
,
1034 .acquire_dquot
= ext4_acquire_dquot
,
1035 .release_dquot
= ext4_release_dquot
,
1036 .mark_dirty
= ext4_mark_dquot_dirty
,
1037 .write_info
= ext4_write_info
,
1038 .alloc_dquot
= dquot_alloc
,
1039 .destroy_dquot
= dquot_destroy
,
1042 static const struct quotactl_ops ext4_qctl_operations
= {
1043 .quota_on
= ext4_quota_on
,
1044 .quota_off
= ext4_quota_off
,
1045 .quota_sync
= dquot_quota_sync
,
1046 .get_info
= dquot_get_dqinfo
,
1047 .set_info
= dquot_set_dqinfo
,
1048 .get_dqblk
= dquot_get_dqblk
,
1049 .set_dqblk
= dquot_set_dqblk
1052 static const struct quotactl_ops ext4_qctl_sysfile_operations
= {
1053 .quota_on_meta
= ext4_quota_on_sysfile
,
1054 .quota_off
= ext4_quota_off_sysfile
,
1055 .quota_sync
= dquot_quota_sync
,
1056 .get_info
= dquot_get_dqinfo
,
1057 .set_info
= dquot_set_dqinfo
,
1058 .get_dqblk
= dquot_get_dqblk
,
1059 .set_dqblk
= dquot_set_dqblk
1063 static const struct super_operations ext4_sops
= {
1064 .alloc_inode
= ext4_alloc_inode
,
1065 .destroy_inode
= ext4_destroy_inode
,
1066 .write_inode
= ext4_write_inode
,
1067 .dirty_inode
= ext4_dirty_inode
,
1068 .drop_inode
= ext4_drop_inode
,
1069 .evict_inode
= ext4_evict_inode
,
1070 .put_super
= ext4_put_super
,
1071 .sync_fs
= ext4_sync_fs
,
1072 .freeze_fs
= ext4_freeze
,
1073 .unfreeze_fs
= ext4_unfreeze
,
1074 .statfs
= ext4_statfs
,
1075 .remount_fs
= ext4_remount
,
1076 .show_options
= ext4_show_options
,
1078 .quota_read
= ext4_quota_read
,
1079 .quota_write
= ext4_quota_write
,
1081 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1084 static const struct super_operations ext4_nojournal_sops
= {
1085 .alloc_inode
= ext4_alloc_inode
,
1086 .destroy_inode
= ext4_destroy_inode
,
1087 .write_inode
= ext4_write_inode
,
1088 .dirty_inode
= ext4_dirty_inode
,
1089 .drop_inode
= ext4_drop_inode
,
1090 .evict_inode
= ext4_evict_inode
,
1091 .put_super
= ext4_put_super
,
1092 .statfs
= ext4_statfs
,
1093 .remount_fs
= ext4_remount
,
1094 .show_options
= ext4_show_options
,
1096 .quota_read
= ext4_quota_read
,
1097 .quota_write
= ext4_quota_write
,
1099 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1102 static const struct export_operations ext4_export_ops
= {
1103 .fh_to_dentry
= ext4_fh_to_dentry
,
1104 .fh_to_parent
= ext4_fh_to_parent
,
1105 .get_parent
= ext4_get_parent
,
1109 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1110 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1111 Opt_nouid32
, Opt_debug
, Opt_removed
,
1112 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1113 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
,
1114 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
,
1115 Opt_journal_dev
, Opt_journal_checksum
, Opt_journal_async_commit
,
1116 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1117 Opt_data_err_abort
, Opt_data_err_ignore
,
1118 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1119 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
1120 Opt_noquota
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
1121 Opt_usrquota
, Opt_grpquota
, Opt_i_version
,
1122 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
, Opt_mblk_io_submit
,
1123 Opt_nomblk_io_submit
, Opt_block_validity
, Opt_noblock_validity
,
1124 Opt_inode_readahead_blks
, Opt_journal_ioprio
,
1125 Opt_dioread_nolock
, Opt_dioread_lock
,
1126 Opt_discard
, Opt_nodiscard
, Opt_init_itable
, Opt_noinit_itable
,
1127 Opt_max_dir_size_kb
,
1130 static const match_table_t tokens
= {
1131 {Opt_bsd_df
, "bsddf"},
1132 {Opt_minix_df
, "minixdf"},
1133 {Opt_grpid
, "grpid"},
1134 {Opt_grpid
, "bsdgroups"},
1135 {Opt_nogrpid
, "nogrpid"},
1136 {Opt_nogrpid
, "sysvgroups"},
1137 {Opt_resgid
, "resgid=%u"},
1138 {Opt_resuid
, "resuid=%u"},
1140 {Opt_err_cont
, "errors=continue"},
1141 {Opt_err_panic
, "errors=panic"},
1142 {Opt_err_ro
, "errors=remount-ro"},
1143 {Opt_nouid32
, "nouid32"},
1144 {Opt_debug
, "debug"},
1145 {Opt_removed
, "oldalloc"},
1146 {Opt_removed
, "orlov"},
1147 {Opt_user_xattr
, "user_xattr"},
1148 {Opt_nouser_xattr
, "nouser_xattr"},
1150 {Opt_noacl
, "noacl"},
1151 {Opt_noload
, "norecovery"},
1152 {Opt_noload
, "noload"},
1153 {Opt_removed
, "nobh"},
1154 {Opt_removed
, "bh"},
1155 {Opt_commit
, "commit=%u"},
1156 {Opt_min_batch_time
, "min_batch_time=%u"},
1157 {Opt_max_batch_time
, "max_batch_time=%u"},
1158 {Opt_journal_dev
, "journal_dev=%u"},
1159 {Opt_journal_checksum
, "journal_checksum"},
1160 {Opt_journal_async_commit
, "journal_async_commit"},
1161 {Opt_abort
, "abort"},
1162 {Opt_data_journal
, "data=journal"},
1163 {Opt_data_ordered
, "data=ordered"},
1164 {Opt_data_writeback
, "data=writeback"},
1165 {Opt_data_err_abort
, "data_err=abort"},
1166 {Opt_data_err_ignore
, "data_err=ignore"},
1167 {Opt_offusrjquota
, "usrjquota="},
1168 {Opt_usrjquota
, "usrjquota=%s"},
1169 {Opt_offgrpjquota
, "grpjquota="},
1170 {Opt_grpjquota
, "grpjquota=%s"},
1171 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1172 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1173 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
1174 {Opt_grpquota
, "grpquota"},
1175 {Opt_noquota
, "noquota"},
1176 {Opt_quota
, "quota"},
1177 {Opt_usrquota
, "usrquota"},
1178 {Opt_barrier
, "barrier=%u"},
1179 {Opt_barrier
, "barrier"},
1180 {Opt_nobarrier
, "nobarrier"},
1181 {Opt_i_version
, "i_version"},
1182 {Opt_stripe
, "stripe=%u"},
1183 {Opt_delalloc
, "delalloc"},
1184 {Opt_nodelalloc
, "nodelalloc"},
1185 {Opt_removed
, "mblk_io_submit"},
1186 {Opt_removed
, "nomblk_io_submit"},
1187 {Opt_block_validity
, "block_validity"},
1188 {Opt_noblock_validity
, "noblock_validity"},
1189 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1190 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1191 {Opt_auto_da_alloc
, "auto_da_alloc=%u"},
1192 {Opt_auto_da_alloc
, "auto_da_alloc"},
1193 {Opt_noauto_da_alloc
, "noauto_da_alloc"},
1194 {Opt_dioread_nolock
, "dioread_nolock"},
1195 {Opt_dioread_lock
, "dioread_lock"},
1196 {Opt_discard
, "discard"},
1197 {Opt_nodiscard
, "nodiscard"},
1198 {Opt_init_itable
, "init_itable=%u"},
1199 {Opt_init_itable
, "init_itable"},
1200 {Opt_noinit_itable
, "noinit_itable"},
1201 {Opt_max_dir_size_kb
, "max_dir_size_kb=%u"},
1202 {Opt_removed
, "check=none"}, /* mount option from ext2/3 */
1203 {Opt_removed
, "nocheck"}, /* mount option from ext2/3 */
1204 {Opt_removed
, "reservation"}, /* mount option from ext2/3 */
1205 {Opt_removed
, "noreservation"}, /* mount option from ext2/3 */
1206 {Opt_removed
, "journal=%u"}, /* mount option from ext2/3 */
1210 static ext4_fsblk_t
get_sb_block(void **data
)
1212 ext4_fsblk_t sb_block
;
1213 char *options
= (char *) *data
;
1215 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1216 return 1; /* Default location */
1219 /* TODO: use simple_strtoll with >32bit ext4 */
1220 sb_block
= simple_strtoul(options
, &options
, 0);
1221 if (*options
&& *options
!= ',') {
1222 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1226 if (*options
== ',')
1228 *data
= (void *) options
;
1233 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1234 static char deprecated_msg
[] = "Mount option \"%s\" will be removed by %s\n"
1235 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1238 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
1240 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1244 if (sb_any_quota_loaded(sb
) &&
1245 !sbi
->s_qf_names
[qtype
]) {
1246 ext4_msg(sb
, KERN_ERR
,
1247 "Cannot change journaled "
1248 "quota options when quota turned on");
1251 qname
= match_strdup(args
);
1253 ext4_msg(sb
, KERN_ERR
,
1254 "Not enough memory for storing quotafile name");
1257 if (sbi
->s_qf_names
[qtype
]) {
1258 if (strcmp(sbi
->s_qf_names
[qtype
], qname
) == 0)
1261 ext4_msg(sb
, KERN_ERR
,
1262 "%s quota file already specified",
1266 if (strchr(qname
, '/')) {
1267 ext4_msg(sb
, KERN_ERR
,
1268 "quotafile must be on filesystem root");
1271 sbi
->s_qf_names
[qtype
] = qname
;
1279 static int clear_qf_name(struct super_block
*sb
, int qtype
)
1282 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1284 if (sb_any_quota_loaded(sb
) &&
1285 sbi
->s_qf_names
[qtype
]) {
1286 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
1287 " when quota turned on");
1290 kfree(sbi
->s_qf_names
[qtype
]);
1291 sbi
->s_qf_names
[qtype
] = NULL
;
1296 #define MOPT_SET 0x0001
1297 #define MOPT_CLEAR 0x0002
1298 #define MOPT_NOSUPPORT 0x0004
1299 #define MOPT_EXPLICIT 0x0008
1300 #define MOPT_CLEAR_ERR 0x0010
1301 #define MOPT_GTE0 0x0020
1304 #define MOPT_QFMT 0x0040
1306 #define MOPT_Q MOPT_NOSUPPORT
1307 #define MOPT_QFMT MOPT_NOSUPPORT
1309 #define MOPT_DATAJ 0x0080
1310 #define MOPT_NO_EXT2 0x0100
1311 #define MOPT_NO_EXT3 0x0200
1312 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1314 static const struct mount_opts
{
1318 } ext4_mount_opts
[] = {
1319 {Opt_minix_df
, EXT4_MOUNT_MINIX_DF
, MOPT_SET
},
1320 {Opt_bsd_df
, EXT4_MOUNT_MINIX_DF
, MOPT_CLEAR
},
1321 {Opt_grpid
, EXT4_MOUNT_GRPID
, MOPT_SET
},
1322 {Opt_nogrpid
, EXT4_MOUNT_GRPID
, MOPT_CLEAR
},
1323 {Opt_block_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_SET
},
1324 {Opt_noblock_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_CLEAR
},
1325 {Opt_dioread_nolock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1326 MOPT_EXT4_ONLY
| MOPT_SET
},
1327 {Opt_dioread_lock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1328 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1329 {Opt_discard
, EXT4_MOUNT_DISCARD
, MOPT_SET
},
1330 {Opt_nodiscard
, EXT4_MOUNT_DISCARD
, MOPT_CLEAR
},
1331 {Opt_delalloc
, EXT4_MOUNT_DELALLOC
,
1332 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1333 {Opt_nodelalloc
, EXT4_MOUNT_DELALLOC
,
1334 MOPT_EXT4_ONLY
| MOPT_CLEAR
| MOPT_EXPLICIT
},
1335 {Opt_journal_checksum
, EXT4_MOUNT_JOURNAL_CHECKSUM
,
1336 MOPT_EXT4_ONLY
| MOPT_SET
},
1337 {Opt_journal_async_commit
, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT
|
1338 EXT4_MOUNT_JOURNAL_CHECKSUM
),
1339 MOPT_EXT4_ONLY
| MOPT_SET
},
1340 {Opt_noload
, EXT4_MOUNT_NOLOAD
, MOPT_NO_EXT2
| MOPT_SET
},
1341 {Opt_err_panic
, EXT4_MOUNT_ERRORS_PANIC
, MOPT_SET
| MOPT_CLEAR_ERR
},
1342 {Opt_err_ro
, EXT4_MOUNT_ERRORS_RO
, MOPT_SET
| MOPT_CLEAR_ERR
},
1343 {Opt_err_cont
, EXT4_MOUNT_ERRORS_CONT
, MOPT_SET
| MOPT_CLEAR_ERR
},
1344 {Opt_data_err_abort
, EXT4_MOUNT_DATA_ERR_ABORT
,
1345 MOPT_NO_EXT2
| MOPT_SET
},
1346 {Opt_data_err_ignore
, EXT4_MOUNT_DATA_ERR_ABORT
,
1347 MOPT_NO_EXT2
| MOPT_CLEAR
},
1348 {Opt_barrier
, EXT4_MOUNT_BARRIER
, MOPT_SET
},
1349 {Opt_nobarrier
, EXT4_MOUNT_BARRIER
, MOPT_CLEAR
},
1350 {Opt_noauto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_SET
},
1351 {Opt_auto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_CLEAR
},
1352 {Opt_noinit_itable
, EXT4_MOUNT_INIT_INODE_TABLE
, MOPT_CLEAR
},
1353 {Opt_commit
, 0, MOPT_GTE0
},
1354 {Opt_max_batch_time
, 0, MOPT_GTE0
},
1355 {Opt_min_batch_time
, 0, MOPT_GTE0
},
1356 {Opt_inode_readahead_blks
, 0, MOPT_GTE0
},
1357 {Opt_init_itable
, 0, MOPT_GTE0
},
1358 {Opt_stripe
, 0, MOPT_GTE0
},
1359 {Opt_resuid
, 0, MOPT_GTE0
},
1360 {Opt_resgid
, 0, MOPT_GTE0
},
1361 {Opt_journal_dev
, 0, MOPT_GTE0
},
1362 {Opt_journal_ioprio
, 0, MOPT_GTE0
},
1363 {Opt_data_journal
, EXT4_MOUNT_JOURNAL_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1364 {Opt_data_ordered
, EXT4_MOUNT_ORDERED_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1365 {Opt_data_writeback
, EXT4_MOUNT_WRITEBACK_DATA
,
1366 MOPT_NO_EXT2
| MOPT_DATAJ
},
1367 {Opt_user_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_SET
},
1368 {Opt_nouser_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_CLEAR
},
1369 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1370 {Opt_acl
, EXT4_MOUNT_POSIX_ACL
, MOPT_SET
},
1371 {Opt_noacl
, EXT4_MOUNT_POSIX_ACL
, MOPT_CLEAR
},
1373 {Opt_acl
, 0, MOPT_NOSUPPORT
},
1374 {Opt_noacl
, 0, MOPT_NOSUPPORT
},
1376 {Opt_nouid32
, EXT4_MOUNT_NO_UID32
, MOPT_SET
},
1377 {Opt_debug
, EXT4_MOUNT_DEBUG
, MOPT_SET
},
1378 {Opt_quota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
, MOPT_SET
| MOPT_Q
},
1379 {Opt_usrquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
,
1381 {Opt_grpquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_GRPQUOTA
,
1383 {Opt_noquota
, (EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
|
1384 EXT4_MOUNT_GRPQUOTA
), MOPT_CLEAR
| MOPT_Q
},
1385 {Opt_usrjquota
, 0, MOPT_Q
},
1386 {Opt_grpjquota
, 0, MOPT_Q
},
1387 {Opt_offusrjquota
, 0, MOPT_Q
},
1388 {Opt_offgrpjquota
, 0, MOPT_Q
},
1389 {Opt_jqfmt_vfsold
, QFMT_VFS_OLD
, MOPT_QFMT
},
1390 {Opt_jqfmt_vfsv0
, QFMT_VFS_V0
, MOPT_QFMT
},
1391 {Opt_jqfmt_vfsv1
, QFMT_VFS_V1
, MOPT_QFMT
},
1392 {Opt_max_dir_size_kb
, 0, MOPT_GTE0
},
1396 static int handle_mount_opt(struct super_block
*sb
, char *opt
, int token
,
1397 substring_t
*args
, unsigned long *journal_devnum
,
1398 unsigned int *journal_ioprio
, int is_remount
)
1400 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1401 const struct mount_opts
*m
;
1407 if (token
== Opt_usrjquota
)
1408 return set_qf_name(sb
, USRQUOTA
, &args
[0]);
1409 else if (token
== Opt_grpjquota
)
1410 return set_qf_name(sb
, GRPQUOTA
, &args
[0]);
1411 else if (token
== Opt_offusrjquota
)
1412 return clear_qf_name(sb
, USRQUOTA
);
1413 else if (token
== Opt_offgrpjquota
)
1414 return clear_qf_name(sb
, GRPQUOTA
);
1418 case Opt_nouser_xattr
:
1419 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, opt
, "3.5");
1422 return 1; /* handled by get_sb_block() */
1424 ext4_msg(sb
, KERN_WARNING
, "Ignoring removed %s option", opt
);
1427 sbi
->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
1430 sb
->s_flags
|= MS_I_VERSION
;
1434 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++)
1435 if (token
== m
->token
)
1438 if (m
->token
== Opt_err
) {
1439 ext4_msg(sb
, KERN_ERR
, "Unrecognized mount option \"%s\" "
1440 "or missing value", opt
);
1444 if ((m
->flags
& MOPT_NO_EXT2
) && IS_EXT2_SB(sb
)) {
1445 ext4_msg(sb
, KERN_ERR
,
1446 "Mount option \"%s\" incompatible with ext2", opt
);
1449 if ((m
->flags
& MOPT_NO_EXT3
) && IS_EXT3_SB(sb
)) {
1450 ext4_msg(sb
, KERN_ERR
,
1451 "Mount option \"%s\" incompatible with ext3", opt
);
1455 if (args
->from
&& match_int(args
, &arg
))
1457 if (args
->from
&& (m
->flags
& MOPT_GTE0
) && (arg
< 0))
1459 if (m
->flags
& MOPT_EXPLICIT
)
1460 set_opt2(sb
, EXPLICIT_DELALLOC
);
1461 if (m
->flags
& MOPT_CLEAR_ERR
)
1462 clear_opt(sb
, ERRORS_MASK
);
1463 if (token
== Opt_noquota
&& sb_any_quota_loaded(sb
)) {
1464 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
1465 "options when quota turned on");
1469 if (m
->flags
& MOPT_NOSUPPORT
) {
1470 ext4_msg(sb
, KERN_ERR
, "%s option not supported", opt
);
1471 } else if (token
== Opt_commit
) {
1473 arg
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1474 sbi
->s_commit_interval
= HZ
* arg
;
1475 } else if (token
== Opt_max_batch_time
) {
1477 arg
= EXT4_DEF_MAX_BATCH_TIME
;
1478 sbi
->s_max_batch_time
= arg
;
1479 } else if (token
== Opt_min_batch_time
) {
1480 sbi
->s_min_batch_time
= arg
;
1481 } else if (token
== Opt_inode_readahead_blks
) {
1482 if (arg
&& (arg
> (1 << 30) || !is_power_of_2(arg
))) {
1483 ext4_msg(sb
, KERN_ERR
,
1484 "EXT4-fs: inode_readahead_blks must be "
1485 "0 or a power of 2 smaller than 2^31");
1488 sbi
->s_inode_readahead_blks
= arg
;
1489 } else if (token
== Opt_init_itable
) {
1490 set_opt(sb
, INIT_INODE_TABLE
);
1492 arg
= EXT4_DEF_LI_WAIT_MULT
;
1493 sbi
->s_li_wait_mult
= arg
;
1494 } else if (token
== Opt_max_dir_size_kb
) {
1495 sbi
->s_max_dir_size_kb
= arg
;
1496 } else if (token
== Opt_stripe
) {
1497 sbi
->s_stripe
= arg
;
1498 } else if (token
== Opt_resuid
) {
1499 uid
= make_kuid(current_user_ns(), arg
);
1500 if (!uid_valid(uid
)) {
1501 ext4_msg(sb
, KERN_ERR
, "Invalid uid value %d", arg
);
1504 sbi
->s_resuid
= uid
;
1505 } else if (token
== Opt_resgid
) {
1506 gid
= make_kgid(current_user_ns(), arg
);
1507 if (!gid_valid(gid
)) {
1508 ext4_msg(sb
, KERN_ERR
, "Invalid gid value %d", arg
);
1511 sbi
->s_resgid
= gid
;
1512 } else if (token
== Opt_journal_dev
) {
1514 ext4_msg(sb
, KERN_ERR
,
1515 "Cannot specify journal on remount");
1518 *journal_devnum
= arg
;
1519 } else if (token
== Opt_journal_ioprio
) {
1521 ext4_msg(sb
, KERN_ERR
, "Invalid journal IO priority"
1526 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
, arg
);
1527 } else if (m
->flags
& MOPT_DATAJ
) {
1529 if (!sbi
->s_journal
)
1530 ext4_msg(sb
, KERN_WARNING
, "Remounting file system with no journal so ignoring journalled data option");
1531 else if (test_opt(sb
, DATA_FLAGS
) != m
->mount_opt
) {
1532 ext4_msg(sb
, KERN_ERR
,
1533 "Cannot change data mode on remount");
1537 clear_opt(sb
, DATA_FLAGS
);
1538 sbi
->s_mount_opt
|= m
->mount_opt
;
1541 } else if (m
->flags
& MOPT_QFMT
) {
1542 if (sb_any_quota_loaded(sb
) &&
1543 sbi
->s_jquota_fmt
!= m
->mount_opt
) {
1544 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled "
1545 "quota options when quota turned on");
1548 sbi
->s_jquota_fmt
= m
->mount_opt
;
1553 if (m
->flags
& MOPT_CLEAR
)
1555 else if (unlikely(!(m
->flags
& MOPT_SET
))) {
1556 ext4_msg(sb
, KERN_WARNING
,
1557 "buggy handling of option %s", opt
);
1562 sbi
->s_mount_opt
|= m
->mount_opt
;
1564 sbi
->s_mount_opt
&= ~m
->mount_opt
;
1569 static int parse_options(char *options
, struct super_block
*sb
,
1570 unsigned long *journal_devnum
,
1571 unsigned int *journal_ioprio
,
1574 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1576 substring_t args
[MAX_OPT_ARGS
];
1582 while ((p
= strsep(&options
, ",")) != NULL
) {
1586 * Initialize args struct so we know whether arg was
1587 * found; some options take optional arguments.
1589 args
[0].to
= args
[0].from
= NULL
;
1590 token
= match_token(p
, tokens
, args
);
1591 if (handle_mount_opt(sb
, p
, token
, args
, journal_devnum
,
1592 journal_ioprio
, is_remount
) < 0)
1596 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1597 if (test_opt(sb
, USRQUOTA
) && sbi
->s_qf_names
[USRQUOTA
])
1598 clear_opt(sb
, USRQUOTA
);
1600 if (test_opt(sb
, GRPQUOTA
) && sbi
->s_qf_names
[GRPQUOTA
])
1601 clear_opt(sb
, GRPQUOTA
);
1603 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
1604 ext4_msg(sb
, KERN_ERR
, "old and new quota "
1609 if (!sbi
->s_jquota_fmt
) {
1610 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1615 if (sbi
->s_jquota_fmt
) {
1616 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1617 "specified with no journaling "
1623 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
1625 BLOCK_SIZE
<< le32_to_cpu(sbi
->s_es
->s_log_block_size
);
1627 if (blocksize
< PAGE_CACHE_SIZE
) {
1628 ext4_msg(sb
, KERN_ERR
, "can't mount with "
1629 "dioread_nolock if block size != PAGE_SIZE");
1636 static inline void ext4_show_quota_options(struct seq_file
*seq
,
1637 struct super_block
*sb
)
1639 #if defined(CONFIG_QUOTA)
1640 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1642 if (sbi
->s_jquota_fmt
) {
1645 switch (sbi
->s_jquota_fmt
) {
1656 seq_printf(seq
, ",jqfmt=%s", fmtname
);
1659 if (sbi
->s_qf_names
[USRQUOTA
])
1660 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
1662 if (sbi
->s_qf_names
[GRPQUOTA
])
1663 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
1665 if (test_opt(sb
, USRQUOTA
))
1666 seq_puts(seq
, ",usrquota");
1668 if (test_opt(sb
, GRPQUOTA
))
1669 seq_puts(seq
, ",grpquota");
1673 static const char *token2str(int token
)
1675 const struct match_token
*t
;
1677 for (t
= tokens
; t
->token
!= Opt_err
; t
++)
1678 if (t
->token
== token
&& !strchr(t
->pattern
, '='))
1685 * - it's set to a non-default value OR
1686 * - if the per-sb default is different from the global default
1688 static int _ext4_show_options(struct seq_file
*seq
, struct super_block
*sb
,
1691 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1692 struct ext4_super_block
*es
= sbi
->s_es
;
1693 int def_errors
, def_mount_opt
= nodefs
? 0 : sbi
->s_def_mount_opt
;
1694 const struct mount_opts
*m
;
1695 char sep
= nodefs
? '\n' : ',';
1697 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1698 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1700 if (sbi
->s_sb_block
!= 1)
1701 SEQ_OPTS_PRINT("sb=%llu", sbi
->s_sb_block
);
1703 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++) {
1704 int want_set
= m
->flags
& MOPT_SET
;
1705 if (((m
->flags
& (MOPT_SET
|MOPT_CLEAR
)) == 0) ||
1706 (m
->flags
& MOPT_CLEAR_ERR
))
1708 if (!(m
->mount_opt
& (sbi
->s_mount_opt
^ def_mount_opt
)))
1709 continue; /* skip if same as the default */
1711 (sbi
->s_mount_opt
& m
->mount_opt
) != m
->mount_opt
) ||
1712 (!want_set
&& (sbi
->s_mount_opt
& m
->mount_opt
)))
1713 continue; /* select Opt_noFoo vs Opt_Foo */
1714 SEQ_OPTS_PRINT("%s", token2str(m
->token
));
1717 if (nodefs
|| !uid_eq(sbi
->s_resuid
, make_kuid(&init_user_ns
, EXT4_DEF_RESUID
)) ||
1718 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
)
1719 SEQ_OPTS_PRINT("resuid=%u",
1720 from_kuid_munged(&init_user_ns
, sbi
->s_resuid
));
1721 if (nodefs
|| !gid_eq(sbi
->s_resgid
, make_kgid(&init_user_ns
, EXT4_DEF_RESGID
)) ||
1722 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
)
1723 SEQ_OPTS_PRINT("resgid=%u",
1724 from_kgid_munged(&init_user_ns
, sbi
->s_resgid
));
1725 def_errors
= nodefs
? -1 : le16_to_cpu(es
->s_errors
);
1726 if (test_opt(sb
, ERRORS_RO
) && def_errors
!= EXT4_ERRORS_RO
)
1727 SEQ_OPTS_PUTS("errors=remount-ro");
1728 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
1729 SEQ_OPTS_PUTS("errors=continue");
1730 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
1731 SEQ_OPTS_PUTS("errors=panic");
1732 if (nodefs
|| sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
)
1733 SEQ_OPTS_PRINT("commit=%lu", sbi
->s_commit_interval
/ HZ
);
1734 if (nodefs
|| sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
)
1735 SEQ_OPTS_PRINT("min_batch_time=%u", sbi
->s_min_batch_time
);
1736 if (nodefs
|| sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
)
1737 SEQ_OPTS_PRINT("max_batch_time=%u", sbi
->s_max_batch_time
);
1738 if (sb
->s_flags
& MS_I_VERSION
)
1739 SEQ_OPTS_PUTS("i_version");
1740 if (nodefs
|| sbi
->s_stripe
)
1741 SEQ_OPTS_PRINT("stripe=%lu", sbi
->s_stripe
);
1742 if (EXT4_MOUNT_DATA_FLAGS
& (sbi
->s_mount_opt
^ def_mount_opt
)) {
1743 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
1744 SEQ_OPTS_PUTS("data=journal");
1745 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
1746 SEQ_OPTS_PUTS("data=ordered");
1747 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
1748 SEQ_OPTS_PUTS("data=writeback");
1751 sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
1752 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1753 sbi
->s_inode_readahead_blks
);
1755 if (nodefs
|| (test_opt(sb
, INIT_INODE_TABLE
) &&
1756 (sbi
->s_li_wait_mult
!= EXT4_DEF_LI_WAIT_MULT
)))
1757 SEQ_OPTS_PRINT("init_itable=%u", sbi
->s_li_wait_mult
);
1758 if (nodefs
|| sbi
->s_max_dir_size_kb
)
1759 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi
->s_max_dir_size_kb
);
1761 ext4_show_quota_options(seq
, sb
);
1765 static int ext4_show_options(struct seq_file
*seq
, struct dentry
*root
)
1767 return _ext4_show_options(seq
, root
->d_sb
, 0);
1770 static int options_seq_show(struct seq_file
*seq
, void *offset
)
1772 struct super_block
*sb
= seq
->private;
1775 seq_puts(seq
, (sb
->s_flags
& MS_RDONLY
) ? "ro" : "rw");
1776 rc
= _ext4_show_options(seq
, sb
, 1);
1777 seq_puts(seq
, "\n");
1781 static int options_open_fs(struct inode
*inode
, struct file
*file
)
1783 return single_open(file
, options_seq_show
, PDE(inode
)->data
);
1786 static const struct file_operations ext4_seq_options_fops
= {
1787 .owner
= THIS_MODULE
,
1788 .open
= options_open_fs
,
1790 .llseek
= seq_lseek
,
1791 .release
= single_release
,
1794 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
1797 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1800 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
1801 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
1802 "forcing read-only mode");
1807 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
1808 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
1809 "running e2fsck is recommended");
1810 else if ((sbi
->s_mount_state
& EXT4_ERROR_FS
))
1811 ext4_msg(sb
, KERN_WARNING
,
1812 "warning: mounting fs with errors, "
1813 "running e2fsck is recommended");
1814 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) > 0 &&
1815 le16_to_cpu(es
->s_mnt_count
) >=
1816 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1817 ext4_msg(sb
, KERN_WARNING
,
1818 "warning: maximal mount count reached, "
1819 "running e2fsck is recommended");
1820 else if (le32_to_cpu(es
->s_checkinterval
) &&
1821 (le32_to_cpu(es
->s_lastcheck
) +
1822 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1823 ext4_msg(sb
, KERN_WARNING
,
1824 "warning: checktime reached, "
1825 "running e2fsck is recommended");
1826 if (!sbi
->s_journal
)
1827 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
1828 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1829 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
1830 le16_add_cpu(&es
->s_mnt_count
, 1);
1831 es
->s_mtime
= cpu_to_le32(get_seconds());
1832 ext4_update_dynamic_rev(sb
);
1834 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
1836 ext4_commit_super(sb
, 1);
1838 if (test_opt(sb
, DEBUG
))
1839 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
1840 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1842 sbi
->s_groups_count
,
1843 EXT4_BLOCKS_PER_GROUP(sb
),
1844 EXT4_INODES_PER_GROUP(sb
),
1845 sbi
->s_mount_opt
, sbi
->s_mount_opt2
);
1847 cleancache_init_fs(sb
);
1851 int ext4_alloc_flex_bg_array(struct super_block
*sb
, ext4_group_t ngroup
)
1853 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1854 struct flex_groups
*new_groups
;
1857 if (!sbi
->s_log_groups_per_flex
)
1860 size
= ext4_flex_group(sbi
, ngroup
- 1) + 1;
1861 if (size
<= sbi
->s_flex_groups_allocated
)
1864 size
= roundup_pow_of_two(size
* sizeof(struct flex_groups
));
1865 new_groups
= ext4_kvzalloc(size
, GFP_KERNEL
);
1867 ext4_msg(sb
, KERN_ERR
, "not enough memory for %d flex groups",
1868 size
/ (int) sizeof(struct flex_groups
));
1872 if (sbi
->s_flex_groups
) {
1873 memcpy(new_groups
, sbi
->s_flex_groups
,
1874 (sbi
->s_flex_groups_allocated
*
1875 sizeof(struct flex_groups
)));
1876 ext4_kvfree(sbi
->s_flex_groups
);
1878 sbi
->s_flex_groups
= new_groups
;
1879 sbi
->s_flex_groups_allocated
= size
/ sizeof(struct flex_groups
);
1883 static int ext4_fill_flex_info(struct super_block
*sb
)
1885 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1886 struct ext4_group_desc
*gdp
= NULL
;
1887 ext4_group_t flex_group
;
1888 unsigned int groups_per_flex
= 0;
1891 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
1892 if (sbi
->s_log_groups_per_flex
< 1 || sbi
->s_log_groups_per_flex
> 31) {
1893 sbi
->s_log_groups_per_flex
= 0;
1896 groups_per_flex
= 1U << sbi
->s_log_groups_per_flex
;
1898 err
= ext4_alloc_flex_bg_array(sb
, sbi
->s_groups_count
);
1902 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1903 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1905 flex_group
= ext4_flex_group(sbi
, i
);
1906 atomic_add(ext4_free_inodes_count(sb
, gdp
),
1907 &sbi
->s_flex_groups
[flex_group
].free_inodes
);
1908 atomic_add(ext4_free_group_clusters(sb
, gdp
),
1909 &sbi
->s_flex_groups
[flex_group
].free_clusters
);
1910 atomic_add(ext4_used_dirs_count(sb
, gdp
),
1911 &sbi
->s_flex_groups
[flex_group
].used_dirs
);
1919 static __le16
ext4_group_desc_csum(struct ext4_sb_info
*sbi
, __u32 block_group
,
1920 struct ext4_group_desc
*gdp
)
1924 __le32 le_group
= cpu_to_le32(block_group
);
1926 if ((sbi
->s_es
->s_feature_ro_compat
&
1927 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
))) {
1928 /* Use new metadata_csum algorithm */
1932 old_csum
= gdp
->bg_checksum
;
1933 gdp
->bg_checksum
= 0;
1934 csum32
= ext4_chksum(sbi
, sbi
->s_csum_seed
, (__u8
*)&le_group
,
1936 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)gdp
,
1938 gdp
->bg_checksum
= old_csum
;
1940 crc
= csum32
& 0xFFFF;
1944 /* old crc16 code */
1945 offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
1947 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
1948 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
1949 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
1950 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
1951 /* for checksum of struct ext4_group_desc do the rest...*/
1952 if ((sbi
->s_es
->s_feature_incompat
&
1953 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT
)) &&
1954 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
1955 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
1956 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
1960 return cpu_to_le16(crc
);
1963 int ext4_group_desc_csum_verify(struct super_block
*sb
, __u32 block_group
,
1964 struct ext4_group_desc
*gdp
)
1966 if (ext4_has_group_desc_csum(sb
) &&
1967 (gdp
->bg_checksum
!= ext4_group_desc_csum(EXT4_SB(sb
),
1974 void ext4_group_desc_csum_set(struct super_block
*sb
, __u32 block_group
,
1975 struct ext4_group_desc
*gdp
)
1977 if (!ext4_has_group_desc_csum(sb
))
1979 gdp
->bg_checksum
= ext4_group_desc_csum(EXT4_SB(sb
), block_group
, gdp
);
1982 /* Called at mount-time, super-block is locked */
1983 static int ext4_check_descriptors(struct super_block
*sb
,
1984 ext4_group_t
*first_not_zeroed
)
1986 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1987 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
1988 ext4_fsblk_t last_block
;
1989 ext4_fsblk_t block_bitmap
;
1990 ext4_fsblk_t inode_bitmap
;
1991 ext4_fsblk_t inode_table
;
1992 int flexbg_flag
= 0;
1993 ext4_group_t i
, grp
= sbi
->s_groups_count
;
1995 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
1998 ext4_debug("Checking group descriptors");
2000 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2001 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2003 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
2004 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
2006 last_block
= first_block
+
2007 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2009 if ((grp
== sbi
->s_groups_count
) &&
2010 !(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2013 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
2014 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
2015 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2016 "Block bitmap for group %u not in group "
2017 "(block %llu)!", i
, block_bitmap
);
2020 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
2021 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
2022 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2023 "Inode bitmap for group %u not in group "
2024 "(block %llu)!", i
, inode_bitmap
);
2027 inode_table
= ext4_inode_table(sb
, gdp
);
2028 if (inode_table
< first_block
||
2029 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
2030 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2031 "Inode table for group %u not in group "
2032 "(block %llu)!", i
, inode_table
);
2035 ext4_lock_group(sb
, i
);
2036 if (!ext4_group_desc_csum_verify(sb
, i
, gdp
)) {
2037 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2038 "Checksum for group %u failed (%u!=%u)",
2039 i
, le16_to_cpu(ext4_group_desc_csum(sbi
, i
,
2040 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
2041 if (!(sb
->s_flags
& MS_RDONLY
)) {
2042 ext4_unlock_group(sb
, i
);
2046 ext4_unlock_group(sb
, i
);
2048 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
2050 if (NULL
!= first_not_zeroed
)
2051 *first_not_zeroed
= grp
;
2053 ext4_free_blocks_count_set(sbi
->s_es
,
2054 EXT4_C2B(sbi
, ext4_count_free_clusters(sb
)));
2055 sbi
->s_es
->s_free_inodes_count
=cpu_to_le32(ext4_count_free_inodes(sb
));
2059 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2060 * the superblock) which were deleted from all directories, but held open by
2061 * a process at the time of a crash. We walk the list and try to delete these
2062 * inodes at recovery time (only with a read-write filesystem).
2064 * In order to keep the orphan inode chain consistent during traversal (in
2065 * case of crash during recovery), we link each inode into the superblock
2066 * orphan list_head and handle it the same way as an inode deletion during
2067 * normal operation (which journals the operations for us).
2069 * We only do an iget() and an iput() on each inode, which is very safe if we
2070 * accidentally point at an in-use or already deleted inode. The worst that
2071 * can happen in this case is that we get a "bit already cleared" message from
2072 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2073 * e2fsck was run on this filesystem, and it must have already done the orphan
2074 * inode cleanup for us, so we can safely abort without any further action.
2076 static void ext4_orphan_cleanup(struct super_block
*sb
,
2077 struct ext4_super_block
*es
)
2079 unsigned int s_flags
= sb
->s_flags
;
2080 int nr_orphans
= 0, nr_truncates
= 0;
2084 if (!es
->s_last_orphan
) {
2085 jbd_debug(4, "no orphan inodes to clean up\n");
2089 if (bdev_read_only(sb
->s_bdev
)) {
2090 ext4_msg(sb
, KERN_ERR
, "write access "
2091 "unavailable, skipping orphan cleanup");
2095 /* Check if feature set would not allow a r/w mount */
2096 if (!ext4_feature_set_ok(sb
, 0)) {
2097 ext4_msg(sb
, KERN_INFO
, "Skipping orphan cleanup due to "
2098 "unknown ROCOMPAT features");
2102 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2103 /* don't clear list on RO mount w/ errors */
2104 if (es
->s_last_orphan
&& !(s_flags
& MS_RDONLY
)) {
2105 jbd_debug(1, "Errors on filesystem, "
2106 "clearing orphan list.\n");
2107 es
->s_last_orphan
= 0;
2109 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2113 if (s_flags
& MS_RDONLY
) {
2114 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
2115 sb
->s_flags
&= ~MS_RDONLY
;
2118 /* Needed for iput() to work correctly and not trash data */
2119 sb
->s_flags
|= MS_ACTIVE
;
2120 /* Turn on quotas so that they are updated correctly */
2121 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2122 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
2123 int ret
= ext4_quota_on_mount(sb
, i
);
2125 ext4_msg(sb
, KERN_ERR
,
2126 "Cannot turn on journaled "
2127 "quota: error %d", ret
);
2132 while (es
->s_last_orphan
) {
2133 struct inode
*inode
;
2135 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
2136 if (IS_ERR(inode
)) {
2137 es
->s_last_orphan
= 0;
2141 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
2142 dquot_initialize(inode
);
2143 if (inode
->i_nlink
) {
2144 ext4_msg(sb
, KERN_DEBUG
,
2145 "%s: truncating inode %lu to %lld bytes",
2146 __func__
, inode
->i_ino
, inode
->i_size
);
2147 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2148 inode
->i_ino
, inode
->i_size
);
2149 mutex_lock(&inode
->i_mutex
);
2150 ext4_truncate(inode
);
2151 mutex_unlock(&inode
->i_mutex
);
2154 ext4_msg(sb
, KERN_DEBUG
,
2155 "%s: deleting unreferenced inode %lu",
2156 __func__
, inode
->i_ino
);
2157 jbd_debug(2, "deleting unreferenced inode %lu\n",
2161 iput(inode
); /* The delete magic happens here! */
2164 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2167 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
2168 PLURAL(nr_orphans
));
2170 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
2171 PLURAL(nr_truncates
));
2173 /* Turn quotas off */
2174 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2175 if (sb_dqopt(sb
)->files
[i
])
2176 dquot_quota_off(sb
, i
);
2179 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
2183 * Maximal extent format file size.
2184 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2185 * extent format containers, within a sector_t, and within i_blocks
2186 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2187 * so that won't be a limiting factor.
2189 * However there is other limiting factor. We do store extents in the form
2190 * of starting block and length, hence the resulting length of the extent
2191 * covering maximum file size must fit into on-disk format containers as
2192 * well. Given that length is always by 1 unit bigger than max unit (because
2193 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2195 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2197 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
2200 loff_t upper_limit
= MAX_LFS_FILESIZE
;
2202 /* small i_blocks in vfs inode? */
2203 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2205 * CONFIG_LBDAF is not enabled implies the inode
2206 * i_block represent total blocks in 512 bytes
2207 * 32 == size of vfs inode i_blocks * 8
2209 upper_limit
= (1LL << 32) - 1;
2211 /* total blocks in file system block size */
2212 upper_limit
>>= (blkbits
- 9);
2213 upper_limit
<<= blkbits
;
2217 * 32-bit extent-start container, ee_block. We lower the maxbytes
2218 * by one fs block, so ee_len can cover the extent of maximum file
2221 res
= (1LL << 32) - 1;
2224 /* Sanity check against vm- & vfs- imposed limits */
2225 if (res
> upper_limit
)
2232 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2233 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2234 * We need to be 1 filesystem block less than the 2^48 sector limit.
2236 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
2238 loff_t res
= EXT4_NDIR_BLOCKS
;
2241 /* This is calculated to be the largest file size for a dense, block
2242 * mapped file such that the file's total number of 512-byte sectors,
2243 * including data and all indirect blocks, does not exceed (2^48 - 1).
2245 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2246 * number of 512-byte sectors of the file.
2249 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2251 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2252 * the inode i_block field represents total file blocks in
2253 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2255 upper_limit
= (1LL << 32) - 1;
2257 /* total blocks in file system block size */
2258 upper_limit
>>= (bits
- 9);
2262 * We use 48 bit ext4_inode i_blocks
2263 * With EXT4_HUGE_FILE_FL set the i_blocks
2264 * represent total number of blocks in
2265 * file system block size
2267 upper_limit
= (1LL << 48) - 1;
2271 /* indirect blocks */
2273 /* double indirect blocks */
2274 meta_blocks
+= 1 + (1LL << (bits
-2));
2275 /* tripple indirect blocks */
2276 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
2278 upper_limit
-= meta_blocks
;
2279 upper_limit
<<= bits
;
2281 res
+= 1LL << (bits
-2);
2282 res
+= 1LL << (2*(bits
-2));
2283 res
+= 1LL << (3*(bits
-2));
2285 if (res
> upper_limit
)
2288 if (res
> MAX_LFS_FILESIZE
)
2289 res
= MAX_LFS_FILESIZE
;
2294 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
2295 ext4_fsblk_t logical_sb_block
, int nr
)
2297 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2298 ext4_group_t bg
, first_meta_bg
;
2301 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
2303 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_META_BG
) ||
2305 return logical_sb_block
+ nr
+ 1;
2306 bg
= sbi
->s_desc_per_block
* nr
;
2307 if (ext4_bg_has_super(sb
, bg
))
2310 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
2314 * ext4_get_stripe_size: Get the stripe size.
2315 * @sbi: In memory super block info
2317 * If we have specified it via mount option, then
2318 * use the mount option value. If the value specified at mount time is
2319 * greater than the blocks per group use the super block value.
2320 * If the super block value is greater than blocks per group return 0.
2321 * Allocator needs it be less than blocks per group.
2324 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
2326 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
2327 unsigned long stripe_width
=
2328 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
2331 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
2332 ret
= sbi
->s_stripe
;
2333 else if (stripe_width
<= sbi
->s_blocks_per_group
)
2335 else if (stride
<= sbi
->s_blocks_per_group
)
2341 * If the stripe width is 1, this makes no sense and
2342 * we set it to 0 to turn off stripe handling code.
2353 struct attribute attr
;
2354 ssize_t (*show
)(struct ext4_attr
*, struct ext4_sb_info
*, char *);
2355 ssize_t (*store
)(struct ext4_attr
*, struct ext4_sb_info
*,
2356 const char *, size_t);
2360 static int parse_strtoul(const char *buf
,
2361 unsigned long max
, unsigned long *value
)
2365 *value
= simple_strtoul(skip_spaces(buf
), &endp
, 0);
2366 endp
= skip_spaces(endp
);
2367 if (*endp
|| *value
> max
)
2373 static ssize_t
delayed_allocation_blocks_show(struct ext4_attr
*a
,
2374 struct ext4_sb_info
*sbi
,
2377 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2379 percpu_counter_sum(&sbi
->s_dirtyclusters_counter
)));
2382 static ssize_t
session_write_kbytes_show(struct ext4_attr
*a
,
2383 struct ext4_sb_info
*sbi
, char *buf
)
2385 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2387 if (!sb
->s_bdev
->bd_part
)
2388 return snprintf(buf
, PAGE_SIZE
, "0\n");
2389 return snprintf(buf
, PAGE_SIZE
, "%lu\n",
2390 (part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2391 sbi
->s_sectors_written_start
) >> 1);
2394 static ssize_t
lifetime_write_kbytes_show(struct ext4_attr
*a
,
2395 struct ext4_sb_info
*sbi
, char *buf
)
2397 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2399 if (!sb
->s_bdev
->bd_part
)
2400 return snprintf(buf
, PAGE_SIZE
, "0\n");
2401 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2402 (unsigned long long)(sbi
->s_kbytes_written
+
2403 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2404 EXT4_SB(sb
)->s_sectors_written_start
) >> 1)));
2407 static ssize_t
inode_readahead_blks_store(struct ext4_attr
*a
,
2408 struct ext4_sb_info
*sbi
,
2409 const char *buf
, size_t count
)
2413 if (parse_strtoul(buf
, 0x40000000, &t
))
2416 if (t
&& !is_power_of_2(t
))
2419 sbi
->s_inode_readahead_blks
= t
;
2423 static ssize_t
sbi_ui_show(struct ext4_attr
*a
,
2424 struct ext4_sb_info
*sbi
, char *buf
)
2426 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2428 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
2431 static ssize_t
sbi_ui_store(struct ext4_attr
*a
,
2432 struct ext4_sb_info
*sbi
,
2433 const char *buf
, size_t count
)
2435 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2438 if (parse_strtoul(buf
, 0xffffffff, &t
))
2444 static ssize_t
trigger_test_error(struct ext4_attr
*a
,
2445 struct ext4_sb_info
*sbi
,
2446 const char *buf
, size_t count
)
2450 if (!capable(CAP_SYS_ADMIN
))
2453 if (len
&& buf
[len
-1] == '\n')
2457 ext4_error(sbi
->s_sb
, "%.*s", len
, buf
);
2461 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2462 static struct ext4_attr ext4_attr_##_name = { \
2463 .attr = {.name = __stringify(_name), .mode = _mode }, \
2466 .offset = offsetof(struct ext4_sb_info, _elname), \
2468 #define EXT4_ATTR(name, mode, show, store) \
2469 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2471 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2472 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2473 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2474 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2475 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2476 #define ATTR_LIST(name) &ext4_attr_##name.attr
2478 EXT4_RO_ATTR(delayed_allocation_blocks
);
2479 EXT4_RO_ATTR(session_write_kbytes
);
2480 EXT4_RO_ATTR(lifetime_write_kbytes
);
2481 EXT4_ATTR_OFFSET(inode_readahead_blks
, 0644, sbi_ui_show
,
2482 inode_readahead_blks_store
, s_inode_readahead_blks
);
2483 EXT4_RW_ATTR_SBI_UI(inode_goal
, s_inode_goal
);
2484 EXT4_RW_ATTR_SBI_UI(mb_stats
, s_mb_stats
);
2485 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan
, s_mb_max_to_scan
);
2486 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan
, s_mb_min_to_scan
);
2487 EXT4_RW_ATTR_SBI_UI(mb_order2_req
, s_mb_order2_reqs
);
2488 EXT4_RW_ATTR_SBI_UI(mb_stream_req
, s_mb_stream_request
);
2489 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc
, s_mb_group_prealloc
);
2490 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump
, s_max_writeback_mb_bump
);
2491 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb
, s_extent_max_zeroout_kb
);
2492 EXT4_ATTR(trigger_fs_error
, 0200, NULL
, trigger_test_error
);
2494 static struct attribute
*ext4_attrs
[] = {
2495 ATTR_LIST(delayed_allocation_blocks
),
2496 ATTR_LIST(session_write_kbytes
),
2497 ATTR_LIST(lifetime_write_kbytes
),
2498 ATTR_LIST(inode_readahead_blks
),
2499 ATTR_LIST(inode_goal
),
2500 ATTR_LIST(mb_stats
),
2501 ATTR_LIST(mb_max_to_scan
),
2502 ATTR_LIST(mb_min_to_scan
),
2503 ATTR_LIST(mb_order2_req
),
2504 ATTR_LIST(mb_stream_req
),
2505 ATTR_LIST(mb_group_prealloc
),
2506 ATTR_LIST(max_writeback_mb_bump
),
2507 ATTR_LIST(extent_max_zeroout_kb
),
2508 ATTR_LIST(trigger_fs_error
),
2512 /* Features this copy of ext4 supports */
2513 EXT4_INFO_ATTR(lazy_itable_init
);
2514 EXT4_INFO_ATTR(batched_discard
);
2515 EXT4_INFO_ATTR(meta_bg_resize
);
2517 static struct attribute
*ext4_feat_attrs
[] = {
2518 ATTR_LIST(lazy_itable_init
),
2519 ATTR_LIST(batched_discard
),
2520 ATTR_LIST(meta_bg_resize
),
2524 static ssize_t
ext4_attr_show(struct kobject
*kobj
,
2525 struct attribute
*attr
, char *buf
)
2527 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2529 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2531 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
2534 static ssize_t
ext4_attr_store(struct kobject
*kobj
,
2535 struct attribute
*attr
,
2536 const char *buf
, size_t len
)
2538 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2540 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2542 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
2545 static void ext4_sb_release(struct kobject
*kobj
)
2547 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2549 complete(&sbi
->s_kobj_unregister
);
2552 static const struct sysfs_ops ext4_attr_ops
= {
2553 .show
= ext4_attr_show
,
2554 .store
= ext4_attr_store
,
2557 static struct kobj_type ext4_ktype
= {
2558 .default_attrs
= ext4_attrs
,
2559 .sysfs_ops
= &ext4_attr_ops
,
2560 .release
= ext4_sb_release
,
2563 static void ext4_feat_release(struct kobject
*kobj
)
2565 complete(&ext4_feat
->f_kobj_unregister
);
2568 static struct kobj_type ext4_feat_ktype
= {
2569 .default_attrs
= ext4_feat_attrs
,
2570 .sysfs_ops
= &ext4_attr_ops
,
2571 .release
= ext4_feat_release
,
2575 * Check whether this filesystem can be mounted based on
2576 * the features present and the RDONLY/RDWR mount requested.
2577 * Returns 1 if this filesystem can be mounted as requested,
2578 * 0 if it cannot be.
2580 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
2582 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT4_FEATURE_INCOMPAT_SUPP
)) {
2583 ext4_msg(sb
, KERN_ERR
,
2584 "Couldn't mount because of "
2585 "unsupported optional features (%x)",
2586 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2587 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2594 /* Check that feature set is OK for a read-write mount */
2595 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT4_FEATURE_RO_COMPAT_SUPP
)) {
2596 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
2597 "unsupported optional features (%x)",
2598 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2599 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2603 * Large file size enabled file system can only be mounted
2604 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2606 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_HUGE_FILE
)) {
2607 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
2608 ext4_msg(sb
, KERN_ERR
, "Filesystem with huge files "
2609 "cannot be mounted RDWR without "
2614 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_BIGALLOC
) &&
2615 !EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_EXTENTS
)) {
2616 ext4_msg(sb
, KERN_ERR
,
2617 "Can't support bigalloc feature without "
2618 "extents feature\n");
2622 #ifndef CONFIG_QUOTA
2623 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
) &&
2625 ext4_msg(sb
, KERN_ERR
,
2626 "Filesystem with quota feature cannot be mounted RDWR "
2627 "without CONFIG_QUOTA");
2630 #endif /* CONFIG_QUOTA */
2635 * This function is called once a day if we have errors logged
2636 * on the file system
2638 static void print_daily_error_info(unsigned long arg
)
2640 struct super_block
*sb
= (struct super_block
*) arg
;
2641 struct ext4_sb_info
*sbi
;
2642 struct ext4_super_block
*es
;
2647 if (es
->s_error_count
)
2648 ext4_msg(sb
, KERN_NOTICE
, "error count: %u",
2649 le32_to_cpu(es
->s_error_count
));
2650 if (es
->s_first_error_time
) {
2651 printk(KERN_NOTICE
"EXT4-fs (%s): initial error at %u: %.*s:%d",
2652 sb
->s_id
, le32_to_cpu(es
->s_first_error_time
),
2653 (int) sizeof(es
->s_first_error_func
),
2654 es
->s_first_error_func
,
2655 le32_to_cpu(es
->s_first_error_line
));
2656 if (es
->s_first_error_ino
)
2657 printk(": inode %u",
2658 le32_to_cpu(es
->s_first_error_ino
));
2659 if (es
->s_first_error_block
)
2660 printk(": block %llu", (unsigned long long)
2661 le64_to_cpu(es
->s_first_error_block
));
2664 if (es
->s_last_error_time
) {
2665 printk(KERN_NOTICE
"EXT4-fs (%s): last error at %u: %.*s:%d",
2666 sb
->s_id
, le32_to_cpu(es
->s_last_error_time
),
2667 (int) sizeof(es
->s_last_error_func
),
2668 es
->s_last_error_func
,
2669 le32_to_cpu(es
->s_last_error_line
));
2670 if (es
->s_last_error_ino
)
2671 printk(": inode %u",
2672 le32_to_cpu(es
->s_last_error_ino
));
2673 if (es
->s_last_error_block
)
2674 printk(": block %llu", (unsigned long long)
2675 le64_to_cpu(es
->s_last_error_block
));
2678 mod_timer(&sbi
->s_err_report
, jiffies
+ 24*60*60*HZ
); /* Once a day */
2681 /* Find next suitable group and run ext4_init_inode_table */
2682 static int ext4_run_li_request(struct ext4_li_request
*elr
)
2684 struct ext4_group_desc
*gdp
= NULL
;
2685 ext4_group_t group
, ngroups
;
2686 struct super_block
*sb
;
2687 unsigned long timeout
= 0;
2691 ngroups
= EXT4_SB(sb
)->s_groups_count
;
2694 for (group
= elr
->lr_next_group
; group
< ngroups
; group
++) {
2695 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2701 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2705 if (group
>= ngroups
)
2710 ret
= ext4_init_inode_table(sb
, group
,
2711 elr
->lr_timeout
? 0 : 1);
2712 if (elr
->lr_timeout
== 0) {
2713 timeout
= (jiffies
- timeout
) *
2714 elr
->lr_sbi
->s_li_wait_mult
;
2715 elr
->lr_timeout
= timeout
;
2717 elr
->lr_next_sched
= jiffies
+ elr
->lr_timeout
;
2718 elr
->lr_next_group
= group
+ 1;
2726 * Remove lr_request from the list_request and free the
2727 * request structure. Should be called with li_list_mtx held
2729 static void ext4_remove_li_request(struct ext4_li_request
*elr
)
2731 struct ext4_sb_info
*sbi
;
2738 list_del(&elr
->lr_request
);
2739 sbi
->s_li_request
= NULL
;
2743 static void ext4_unregister_li_request(struct super_block
*sb
)
2745 mutex_lock(&ext4_li_mtx
);
2746 if (!ext4_li_info
) {
2747 mutex_unlock(&ext4_li_mtx
);
2751 mutex_lock(&ext4_li_info
->li_list_mtx
);
2752 ext4_remove_li_request(EXT4_SB(sb
)->s_li_request
);
2753 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2754 mutex_unlock(&ext4_li_mtx
);
2757 static struct task_struct
*ext4_lazyinit_task
;
2760 * This is the function where ext4lazyinit thread lives. It walks
2761 * through the request list searching for next scheduled filesystem.
2762 * When such a fs is found, run the lazy initialization request
2763 * (ext4_rn_li_request) and keep track of the time spend in this
2764 * function. Based on that time we compute next schedule time of
2765 * the request. When walking through the list is complete, compute
2766 * next waking time and put itself into sleep.
2768 static int ext4_lazyinit_thread(void *arg
)
2770 struct ext4_lazy_init
*eli
= (struct ext4_lazy_init
*)arg
;
2771 struct list_head
*pos
, *n
;
2772 struct ext4_li_request
*elr
;
2773 unsigned long next_wakeup
, cur
;
2775 BUG_ON(NULL
== eli
);
2779 next_wakeup
= MAX_JIFFY_OFFSET
;
2781 mutex_lock(&eli
->li_list_mtx
);
2782 if (list_empty(&eli
->li_request_list
)) {
2783 mutex_unlock(&eli
->li_list_mtx
);
2787 list_for_each_safe(pos
, n
, &eli
->li_request_list
) {
2788 elr
= list_entry(pos
, struct ext4_li_request
,
2791 if (time_after_eq(jiffies
, elr
->lr_next_sched
)) {
2792 if (ext4_run_li_request(elr
) != 0) {
2793 /* error, remove the lazy_init job */
2794 ext4_remove_li_request(elr
);
2799 if (time_before(elr
->lr_next_sched
, next_wakeup
))
2800 next_wakeup
= elr
->lr_next_sched
;
2802 mutex_unlock(&eli
->li_list_mtx
);
2807 if ((time_after_eq(cur
, next_wakeup
)) ||
2808 (MAX_JIFFY_OFFSET
== next_wakeup
)) {
2813 schedule_timeout_interruptible(next_wakeup
- cur
);
2815 if (kthread_should_stop()) {
2816 ext4_clear_request_list();
2823 * It looks like the request list is empty, but we need
2824 * to check it under the li_list_mtx lock, to prevent any
2825 * additions into it, and of course we should lock ext4_li_mtx
2826 * to atomically free the list and ext4_li_info, because at
2827 * this point another ext4 filesystem could be registering
2830 mutex_lock(&ext4_li_mtx
);
2831 mutex_lock(&eli
->li_list_mtx
);
2832 if (!list_empty(&eli
->li_request_list
)) {
2833 mutex_unlock(&eli
->li_list_mtx
);
2834 mutex_unlock(&ext4_li_mtx
);
2837 mutex_unlock(&eli
->li_list_mtx
);
2838 kfree(ext4_li_info
);
2839 ext4_li_info
= NULL
;
2840 mutex_unlock(&ext4_li_mtx
);
2845 static void ext4_clear_request_list(void)
2847 struct list_head
*pos
, *n
;
2848 struct ext4_li_request
*elr
;
2850 mutex_lock(&ext4_li_info
->li_list_mtx
);
2851 list_for_each_safe(pos
, n
, &ext4_li_info
->li_request_list
) {
2852 elr
= list_entry(pos
, struct ext4_li_request
,
2854 ext4_remove_li_request(elr
);
2856 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2859 static int ext4_run_lazyinit_thread(void)
2861 ext4_lazyinit_task
= kthread_run(ext4_lazyinit_thread
,
2862 ext4_li_info
, "ext4lazyinit");
2863 if (IS_ERR(ext4_lazyinit_task
)) {
2864 int err
= PTR_ERR(ext4_lazyinit_task
);
2865 ext4_clear_request_list();
2866 kfree(ext4_li_info
);
2867 ext4_li_info
= NULL
;
2868 printk(KERN_CRIT
"EXT4-fs: error %d creating inode table "
2869 "initialization thread\n",
2873 ext4_li_info
->li_state
|= EXT4_LAZYINIT_RUNNING
;
2878 * Check whether it make sense to run itable init. thread or not.
2879 * If there is at least one uninitialized inode table, return
2880 * corresponding group number, else the loop goes through all
2881 * groups and return total number of groups.
2883 static ext4_group_t
ext4_has_uninit_itable(struct super_block
*sb
)
2885 ext4_group_t group
, ngroups
= EXT4_SB(sb
)->s_groups_count
;
2886 struct ext4_group_desc
*gdp
= NULL
;
2888 for (group
= 0; group
< ngroups
; group
++) {
2889 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2893 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2900 static int ext4_li_info_new(void)
2902 struct ext4_lazy_init
*eli
= NULL
;
2904 eli
= kzalloc(sizeof(*eli
), GFP_KERNEL
);
2908 INIT_LIST_HEAD(&eli
->li_request_list
);
2909 mutex_init(&eli
->li_list_mtx
);
2911 eli
->li_state
|= EXT4_LAZYINIT_QUIT
;
2918 static struct ext4_li_request
*ext4_li_request_new(struct super_block
*sb
,
2921 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2922 struct ext4_li_request
*elr
;
2925 elr
= kzalloc(sizeof(*elr
), GFP_KERNEL
);
2931 elr
->lr_next_group
= start
;
2934 * Randomize first schedule time of the request to
2935 * spread the inode table initialization requests
2938 get_random_bytes(&rnd
, sizeof(rnd
));
2939 elr
->lr_next_sched
= jiffies
+ (unsigned long)rnd
%
2940 (EXT4_DEF_LI_MAX_START_DELAY
* HZ
);
2945 int ext4_register_li_request(struct super_block
*sb
,
2946 ext4_group_t first_not_zeroed
)
2948 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2949 struct ext4_li_request
*elr
= NULL
;
2950 ext4_group_t ngroups
= EXT4_SB(sb
)->s_groups_count
;
2953 mutex_lock(&ext4_li_mtx
);
2954 if (sbi
->s_li_request
!= NULL
) {
2956 * Reset timeout so it can be computed again, because
2957 * s_li_wait_mult might have changed.
2959 sbi
->s_li_request
->lr_timeout
= 0;
2963 if (first_not_zeroed
== ngroups
||
2964 (sb
->s_flags
& MS_RDONLY
) ||
2965 !test_opt(sb
, INIT_INODE_TABLE
))
2968 elr
= ext4_li_request_new(sb
, first_not_zeroed
);
2974 if (NULL
== ext4_li_info
) {
2975 ret
= ext4_li_info_new();
2980 mutex_lock(&ext4_li_info
->li_list_mtx
);
2981 list_add(&elr
->lr_request
, &ext4_li_info
->li_request_list
);
2982 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2984 sbi
->s_li_request
= elr
;
2986 * set elr to NULL here since it has been inserted to
2987 * the request_list and the removal and free of it is
2988 * handled by ext4_clear_request_list from now on.
2992 if (!(ext4_li_info
->li_state
& EXT4_LAZYINIT_RUNNING
)) {
2993 ret
= ext4_run_lazyinit_thread();
2998 mutex_unlock(&ext4_li_mtx
);
3005 * We do not need to lock anything since this is called on
3008 static void ext4_destroy_lazyinit_thread(void)
3011 * If thread exited earlier
3012 * there's nothing to be done.
3014 if (!ext4_li_info
|| !ext4_lazyinit_task
)
3017 kthread_stop(ext4_lazyinit_task
);
3020 static int set_journal_csum_feature_set(struct super_block
*sb
)
3023 int compat
, incompat
;
3024 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3026 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3027 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
)) {
3028 /* journal checksum v2 */
3030 incompat
= JBD2_FEATURE_INCOMPAT_CSUM_V2
;
3032 /* journal checksum v1 */
3033 compat
= JBD2_FEATURE_COMPAT_CHECKSUM
;
3037 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
3038 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3040 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
|
3042 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
3043 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3046 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3047 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3049 jbd2_journal_clear_features(sbi
->s_journal
,
3050 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3051 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
|
3052 JBD2_FEATURE_INCOMPAT_CSUM_V2
);
3059 * Note: calculating the overhead so we can be compatible with
3060 * historical BSD practice is quite difficult in the face of
3061 * clusters/bigalloc. This is because multiple metadata blocks from
3062 * different block group can end up in the same allocation cluster.
3063 * Calculating the exact overhead in the face of clustered allocation
3064 * requires either O(all block bitmaps) in memory or O(number of block
3065 * groups**2) in time. We will still calculate the superblock for
3066 * older file systems --- and if we come across with a bigalloc file
3067 * system with zero in s_overhead_clusters the estimate will be close to
3068 * correct especially for very large cluster sizes --- but for newer
3069 * file systems, it's better to calculate this figure once at mkfs
3070 * time, and store it in the superblock. If the superblock value is
3071 * present (even for non-bigalloc file systems), we will use it.
3073 static int count_overhead(struct super_block
*sb
, ext4_group_t grp
,
3076 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3077 struct ext4_group_desc
*gdp
;
3078 ext4_fsblk_t first_block
, last_block
, b
;
3079 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3080 int s
, j
, count
= 0;
3082 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_BIGALLOC
))
3083 return (ext4_bg_has_super(sb
, grp
) + ext4_bg_num_gdb(sb
, grp
) +
3084 sbi
->s_itb_per_group
+ 2);
3086 first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
) +
3087 (grp
* EXT4_BLOCKS_PER_GROUP(sb
));
3088 last_block
= first_block
+ EXT4_BLOCKS_PER_GROUP(sb
) - 1;
3089 for (i
= 0; i
< ngroups
; i
++) {
3090 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
3091 b
= ext4_block_bitmap(sb
, gdp
);
3092 if (b
>= first_block
&& b
<= last_block
) {
3093 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3096 b
= ext4_inode_bitmap(sb
, gdp
);
3097 if (b
>= first_block
&& b
<= last_block
) {
3098 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3101 b
= ext4_inode_table(sb
, gdp
);
3102 if (b
>= first_block
&& b
+ sbi
->s_itb_per_group
<= last_block
)
3103 for (j
= 0; j
< sbi
->s_itb_per_group
; j
++, b
++) {
3104 int c
= EXT4_B2C(sbi
, b
- first_block
);
3105 ext4_set_bit(c
, buf
);
3111 if (ext4_bg_has_super(sb
, grp
)) {
3112 ext4_set_bit(s
++, buf
);
3115 for (j
= ext4_bg_num_gdb(sb
, grp
); j
> 0; j
--) {
3116 ext4_set_bit(EXT4_B2C(sbi
, s
++), buf
);
3122 return EXT4_CLUSTERS_PER_GROUP(sb
) -
3123 ext4_count_free(buf
, EXT4_CLUSTERS_PER_GROUP(sb
) / 8);
3127 * Compute the overhead and stash it in sbi->s_overhead
3129 int ext4_calculate_overhead(struct super_block
*sb
)
3131 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3132 struct ext4_super_block
*es
= sbi
->s_es
;
3133 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3134 ext4_fsblk_t overhead
= 0;
3135 char *buf
= (char *) get_zeroed_page(GFP_KERNEL
);
3141 * Compute the overhead (FS structures). This is constant
3142 * for a given filesystem unless the number of block groups
3143 * changes so we cache the previous value until it does.
3147 * All of the blocks before first_data_block are overhead
3149 overhead
= EXT4_B2C(sbi
, le32_to_cpu(es
->s_first_data_block
));
3152 * Add the overhead found in each block group
3154 for (i
= 0; i
< ngroups
; i
++) {
3157 blks
= count_overhead(sb
, i
, buf
);
3160 memset(buf
, 0, PAGE_SIZE
);
3163 /* Add the journal blocks as well */
3165 overhead
+= EXT4_B2C(sbi
, sbi
->s_journal
->j_maxlen
);
3167 sbi
->s_overhead
= overhead
;
3169 free_page((unsigned long) buf
);
3173 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
3175 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
3176 struct buffer_head
*bh
;
3177 struct ext4_super_block
*es
= NULL
;
3178 struct ext4_sb_info
*sbi
;
3180 ext4_fsblk_t sb_block
= get_sb_block(&data
);
3181 ext4_fsblk_t logical_sb_block
;
3182 unsigned long offset
= 0;
3183 unsigned long journal_devnum
= 0;
3184 unsigned long def_mount_opts
;
3189 int blocksize
, clustersize
;
3190 unsigned int db_count
;
3192 int needs_recovery
, has_huge_files
, has_bigalloc
;
3195 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
3196 ext4_group_t first_not_zeroed
;
3198 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
3202 sbi
->s_blockgroup_lock
=
3203 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
3204 if (!sbi
->s_blockgroup_lock
) {
3208 sb
->s_fs_info
= sbi
;
3210 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
3211 sbi
->s_sb_block
= sb_block
;
3212 if (sb
->s_bdev
->bd_part
)
3213 sbi
->s_sectors_written_start
=
3214 part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]);
3216 /* Cleanup superblock name */
3217 for (cp
= sb
->s_id
; (cp
= strchr(cp
, '/'));)
3220 /* -EINVAL is default */
3222 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
3224 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
3229 * The ext4 superblock will not be buffer aligned for other than 1kB
3230 * block sizes. We need to calculate the offset from buffer start.
3232 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
3233 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3234 offset
= do_div(logical_sb_block
, blocksize
);
3236 logical_sb_block
= sb_block
;
3239 if (!(bh
= sb_bread(sb
, logical_sb_block
))) {
3240 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
3244 * Note: s_es must be initialized as soon as possible because
3245 * some ext4 macro-instructions depend on its value
3247 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
3249 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
3250 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
3252 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
3254 /* Warn if metadata_csum and gdt_csum are both set. */
3255 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3256 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
) &&
3257 EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_GDT_CSUM
))
3258 ext4_warning(sb
, KERN_INFO
"metadata_csum and uninit_bg are "
3259 "redundant flags; please run fsck.");
3261 /* Check for a known checksum algorithm */
3262 if (!ext4_verify_csum_type(sb
, es
)) {
3263 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3264 "unknown checksum algorithm.");
3269 /* Load the checksum driver */
3270 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3271 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
)) {
3272 sbi
->s_chksum_driver
= crypto_alloc_shash("crc32c", 0, 0);
3273 if (IS_ERR(sbi
->s_chksum_driver
)) {
3274 ext4_msg(sb
, KERN_ERR
, "Cannot load crc32c driver.");
3275 ret
= PTR_ERR(sbi
->s_chksum_driver
);
3276 sbi
->s_chksum_driver
= NULL
;
3281 /* Check superblock checksum */
3282 if (!ext4_superblock_csum_verify(sb
, es
)) {
3283 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3284 "invalid superblock checksum. Run e2fsck?");
3289 /* Precompute checksum seed for all metadata */
3290 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3291 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
))
3292 sbi
->s_csum_seed
= ext4_chksum(sbi
, ~0, es
->s_uuid
,
3293 sizeof(es
->s_uuid
));
3295 /* Set defaults before we parse the mount options */
3296 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
3297 set_opt(sb
, INIT_INODE_TABLE
);
3298 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
3300 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
)
3302 if (def_mount_opts
& EXT4_DEFM_UID16
)
3303 set_opt(sb
, NO_UID32
);
3304 /* xattr user namespace & acls are now defaulted on */
3305 set_opt(sb
, XATTR_USER
);
3306 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3307 set_opt(sb
, POSIX_ACL
);
3309 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
3310 set_opt(sb
, JOURNAL_DATA
);
3311 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
3312 set_opt(sb
, ORDERED_DATA
);
3313 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
3314 set_opt(sb
, WRITEBACK_DATA
);
3316 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
3317 set_opt(sb
, ERRORS_PANIC
);
3318 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
3319 set_opt(sb
, ERRORS_CONT
);
3321 set_opt(sb
, ERRORS_RO
);
3322 if (def_mount_opts
& EXT4_DEFM_BLOCK_VALIDITY
)
3323 set_opt(sb
, BLOCK_VALIDITY
);
3324 if (def_mount_opts
& EXT4_DEFM_DISCARD
)
3325 set_opt(sb
, DISCARD
);
3327 sbi
->s_resuid
= make_kuid(&init_user_ns
, le16_to_cpu(es
->s_def_resuid
));
3328 sbi
->s_resgid
= make_kgid(&init_user_ns
, le16_to_cpu(es
->s_def_resgid
));
3329 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
3330 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
3331 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
3333 if ((def_mount_opts
& EXT4_DEFM_NOBARRIER
) == 0)
3334 set_opt(sb
, BARRIER
);
3337 * enable delayed allocation by default
3338 * Use -o nodelalloc to turn it off
3340 if (!IS_EXT3_SB(sb
) && !IS_EXT2_SB(sb
) &&
3341 ((def_mount_opts
& EXT4_DEFM_NODELALLOC
) == 0))
3342 set_opt(sb
, DELALLOC
);
3345 * set default s_li_wait_mult for lazyinit, for the case there is
3346 * no mount option specified.
3348 sbi
->s_li_wait_mult
= EXT4_DEF_LI_WAIT_MULT
;
3350 if (!parse_options((char *) sbi
->s_es
->s_mount_opts
, sb
,
3351 &journal_devnum
, &journal_ioprio
, 0)) {
3352 ext4_msg(sb
, KERN_WARNING
,
3353 "failed to parse options in superblock: %s",
3354 sbi
->s_es
->s_mount_opts
);
3356 sbi
->s_def_mount_opt
= sbi
->s_mount_opt
;
3357 if (!parse_options((char *) data
, sb
, &journal_devnum
,
3358 &journal_ioprio
, 0))
3361 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
3362 printk_once(KERN_WARNING
"EXT4-fs: Warning: mounting "
3363 "with data=journal disables delayed "
3364 "allocation and O_DIRECT support!\n");
3365 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
3366 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3367 "both data=journal and delalloc");
3370 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
3371 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3372 "both data=journal and delalloc");
3375 if (test_opt(sb
, DELALLOC
))
3376 clear_opt(sb
, DELALLOC
);
3379 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3380 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
3382 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
3383 (EXT4_HAS_COMPAT_FEATURE(sb
, ~0U) ||
3384 EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
3385 EXT4_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
3386 ext4_msg(sb
, KERN_WARNING
,
3387 "feature flags set on rev 0 fs, "
3388 "running e2fsck is recommended");
3390 if (IS_EXT2_SB(sb
)) {
3391 if (ext2_feature_set_ok(sb
))
3392 ext4_msg(sb
, KERN_INFO
, "mounting ext2 file system "
3393 "using the ext4 subsystem");
3395 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext2 due "
3396 "to feature incompatibilities");
3401 if (IS_EXT3_SB(sb
)) {
3402 if (ext3_feature_set_ok(sb
))
3403 ext4_msg(sb
, KERN_INFO
, "mounting ext3 file system "
3404 "using the ext4 subsystem");
3406 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext3 due "
3407 "to feature incompatibilities");
3413 * Check feature flags regardless of the revision level, since we
3414 * previously didn't change the revision level when setting the flags,
3415 * so there is a chance incompat flags are set on a rev 0 filesystem.
3417 if (!ext4_feature_set_ok(sb
, (sb
->s_flags
& MS_RDONLY
)))
3420 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
3421 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
3422 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
3423 ext4_msg(sb
, KERN_ERR
,
3424 "Unsupported filesystem blocksize %d", blocksize
);
3428 if (sb
->s_blocksize
!= blocksize
) {
3429 /* Validate the filesystem blocksize */
3430 if (!sb_set_blocksize(sb
, blocksize
)) {
3431 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
3437 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3438 offset
= do_div(logical_sb_block
, blocksize
);
3439 bh
= sb_bread(sb
, logical_sb_block
);
3441 ext4_msg(sb
, KERN_ERR
,
3442 "Can't read superblock on 2nd try");
3445 es
= (struct ext4_super_block
*)(bh
->b_data
+ offset
);
3447 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
3448 ext4_msg(sb
, KERN_ERR
,
3449 "Magic mismatch, very weird!");
3454 has_huge_files
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3455 EXT4_FEATURE_RO_COMPAT_HUGE_FILE
);
3456 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
3458 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
3460 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
3461 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
3462 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
3464 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
3465 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
3466 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
3467 (!is_power_of_2(sbi
->s_inode_size
)) ||
3468 (sbi
->s_inode_size
> blocksize
)) {
3469 ext4_msg(sb
, KERN_ERR
,
3470 "unsupported inode size: %d",
3474 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
3475 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
3478 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
3479 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
)) {
3480 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
3481 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
3482 !is_power_of_2(sbi
->s_desc_size
)) {
3483 ext4_msg(sb
, KERN_ERR
,
3484 "unsupported descriptor size %lu",
3489 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
3491 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
3492 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
3493 if (EXT4_INODE_SIZE(sb
) == 0 || EXT4_INODES_PER_GROUP(sb
) == 0)
3496 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
3497 if (sbi
->s_inodes_per_block
== 0)
3499 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
3500 sbi
->s_inodes_per_block
;
3501 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
3503 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3504 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
3505 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
3507 for (i
= 0; i
< 4; i
++)
3508 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
3509 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
3510 i
= le32_to_cpu(es
->s_flags
);
3511 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
3512 sbi
->s_hash_unsigned
= 3;
3513 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
3514 #ifdef __CHAR_UNSIGNED__
3515 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
3516 sbi
->s_hash_unsigned
= 3;
3518 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
3522 /* Handle clustersize */
3523 clustersize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_cluster_size
);
3524 has_bigalloc
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3525 EXT4_FEATURE_RO_COMPAT_BIGALLOC
);
3527 if (clustersize
< blocksize
) {
3528 ext4_msg(sb
, KERN_ERR
,
3529 "cluster size (%d) smaller than "
3530 "block size (%d)", clustersize
, blocksize
);
3533 sbi
->s_cluster_bits
= le32_to_cpu(es
->s_log_cluster_size
) -
3534 le32_to_cpu(es
->s_log_block_size
);
3535 sbi
->s_clusters_per_group
=
3536 le32_to_cpu(es
->s_clusters_per_group
);
3537 if (sbi
->s_clusters_per_group
> blocksize
* 8) {
3538 ext4_msg(sb
, KERN_ERR
,
3539 "#clusters per group too big: %lu",
3540 sbi
->s_clusters_per_group
);
3543 if (sbi
->s_blocks_per_group
!=
3544 (sbi
->s_clusters_per_group
* (clustersize
/ blocksize
))) {
3545 ext4_msg(sb
, KERN_ERR
, "blocks per group (%lu) and "
3546 "clusters per group (%lu) inconsistent",
3547 sbi
->s_blocks_per_group
,
3548 sbi
->s_clusters_per_group
);
3552 if (clustersize
!= blocksize
) {
3553 ext4_warning(sb
, "fragment/cluster size (%d) != "
3554 "block size (%d)", clustersize
,
3556 clustersize
= blocksize
;
3558 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
3559 ext4_msg(sb
, KERN_ERR
,
3560 "#blocks per group too big: %lu",
3561 sbi
->s_blocks_per_group
);
3564 sbi
->s_clusters_per_group
= sbi
->s_blocks_per_group
;
3565 sbi
->s_cluster_bits
= 0;
3567 sbi
->s_cluster_ratio
= clustersize
/ blocksize
;
3569 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
3570 ext4_msg(sb
, KERN_ERR
,
3571 "#inodes per group too big: %lu",
3572 sbi
->s_inodes_per_group
);
3577 * Test whether we have more sectors than will fit in sector_t,
3578 * and whether the max offset is addressable by the page cache.
3580 err
= generic_check_addressable(sb
->s_blocksize_bits
,
3581 ext4_blocks_count(es
));
3583 ext4_msg(sb
, KERN_ERR
, "filesystem"
3584 " too large to mount safely on this system");
3585 if (sizeof(sector_t
) < 8)
3586 ext4_msg(sb
, KERN_WARNING
, "CONFIG_LBDAF not enabled");
3590 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
3593 /* check blocks count against device size */
3594 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
3595 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
3596 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
3597 "exceeds size of device (%llu blocks)",
3598 ext4_blocks_count(es
), blocks_count
);
3603 * It makes no sense for the first data block to be beyond the end
3604 * of the filesystem.
3606 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
3607 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data "
3608 "block %u is beyond end of filesystem (%llu)",
3609 le32_to_cpu(es
->s_first_data_block
),
3610 ext4_blocks_count(es
));
3613 blocks_count
= (ext4_blocks_count(es
) -
3614 le32_to_cpu(es
->s_first_data_block
) +
3615 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
3616 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
3617 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
3618 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %u "
3619 "(block count %llu, first data block %u, "
3620 "blocks per group %lu)", sbi
->s_groups_count
,
3621 ext4_blocks_count(es
),
3622 le32_to_cpu(es
->s_first_data_block
),
3623 EXT4_BLOCKS_PER_GROUP(sb
));
3626 sbi
->s_groups_count
= blocks_count
;
3627 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
3628 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
3629 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
3630 EXT4_DESC_PER_BLOCK(sb
);
3631 sbi
->s_group_desc
= ext4_kvmalloc(db_count
*
3632 sizeof(struct buffer_head
*),
3634 if (sbi
->s_group_desc
== NULL
) {
3635 ext4_msg(sb
, KERN_ERR
, "not enough memory");
3641 sbi
->s_proc
= proc_mkdir(sb
->s_id
, ext4_proc_root
);
3644 proc_create_data("options", S_IRUGO
, sbi
->s_proc
,
3645 &ext4_seq_options_fops
, sb
);
3647 bgl_lock_init(sbi
->s_blockgroup_lock
);
3649 for (i
= 0; i
< db_count
; i
++) {
3650 block
= descriptor_loc(sb
, logical_sb_block
, i
);
3651 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
3652 if (!sbi
->s_group_desc
[i
]) {
3653 ext4_msg(sb
, KERN_ERR
,
3654 "can't read group descriptor %d", i
);
3659 if (!ext4_check_descriptors(sb
, &first_not_zeroed
)) {
3660 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
3663 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
3664 if (!ext4_fill_flex_info(sb
)) {
3665 ext4_msg(sb
, KERN_ERR
,
3666 "unable to initialize "
3667 "flex_bg meta info!");
3671 sbi
->s_gdb_count
= db_count
;
3672 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
3673 spin_lock_init(&sbi
->s_next_gen_lock
);
3675 init_timer(&sbi
->s_err_report
);
3676 sbi
->s_err_report
.function
= print_daily_error_info
;
3677 sbi
->s_err_report
.data
= (unsigned long) sb
;
3679 err
= percpu_counter_init(&sbi
->s_freeclusters_counter
,
3680 ext4_count_free_clusters(sb
));
3682 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
,
3683 ext4_count_free_inodes(sb
));
3686 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
3687 ext4_count_dirs(sb
));
3690 err
= percpu_counter_init(&sbi
->s_dirtyclusters_counter
, 0);
3693 err
= percpu_counter_init(&sbi
->s_extent_cache_cnt
, 0);
3696 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
3700 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
3701 sbi
->s_max_writeback_mb_bump
= 128;
3702 sbi
->s_extent_max_zeroout_kb
= 32;
3704 /* Register extent status tree shrinker */
3705 ext4_es_register_shrinker(sb
);
3708 * set up enough so that it can read an inode
3710 if (!test_opt(sb
, NOLOAD
) &&
3711 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
3712 sb
->s_op
= &ext4_sops
;
3714 sb
->s_op
= &ext4_nojournal_sops
;
3715 sb
->s_export_op
= &ext4_export_ops
;
3716 sb
->s_xattr
= ext4_xattr_handlers
;
3718 sb
->s_qcop
= &ext4_qctl_operations
;
3719 sb
->dq_op
= &ext4_quota_operations
;
3721 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
)) {
3722 /* Use qctl operations for hidden quota files. */
3723 sb
->s_qcop
= &ext4_qctl_sysfile_operations
;
3726 memcpy(sb
->s_uuid
, es
->s_uuid
, sizeof(es
->s_uuid
));
3728 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
3729 mutex_init(&sbi
->s_orphan_lock
);
3733 needs_recovery
= (es
->s_last_orphan
!= 0 ||
3734 EXT4_HAS_INCOMPAT_FEATURE(sb
,
3735 EXT4_FEATURE_INCOMPAT_RECOVER
));
3737 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_MMP
) &&
3738 !(sb
->s_flags
& MS_RDONLY
))
3739 if (ext4_multi_mount_protect(sb
, le64_to_cpu(es
->s_mmp_block
)))
3743 * The first inode we look at is the journal inode. Don't try
3744 * root first: it may be modified in the journal!
3746 if (!test_opt(sb
, NOLOAD
) &&
3747 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
3748 if (ext4_load_journal(sb
, es
, journal_devnum
))
3750 } else if (test_opt(sb
, NOLOAD
) && !(sb
->s_flags
& MS_RDONLY
) &&
3751 EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
3752 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
3753 "suppressed and not mounted read-only");
3754 goto failed_mount_wq
;
3756 clear_opt(sb
, DATA_FLAGS
);
3757 sbi
->s_journal
= NULL
;
3762 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
) &&
3763 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
3764 JBD2_FEATURE_INCOMPAT_64BIT
)) {
3765 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
3766 goto failed_mount_wq
;
3769 if (!set_journal_csum_feature_set(sb
)) {
3770 ext4_msg(sb
, KERN_ERR
, "Failed to set journal checksum "
3772 goto failed_mount_wq
;
3775 /* We have now updated the journal if required, so we can
3776 * validate the data journaling mode. */
3777 switch (test_opt(sb
, DATA_FLAGS
)) {
3779 /* No mode set, assume a default based on the journal
3780 * capabilities: ORDERED_DATA if the journal can
3781 * cope, else JOURNAL_DATA
3783 if (jbd2_journal_check_available_features
3784 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
3785 set_opt(sb
, ORDERED_DATA
);
3787 set_opt(sb
, JOURNAL_DATA
);
3790 case EXT4_MOUNT_ORDERED_DATA
:
3791 case EXT4_MOUNT_WRITEBACK_DATA
:
3792 if (!jbd2_journal_check_available_features
3793 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
3794 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
3795 "requested data journaling mode");
3796 goto failed_mount_wq
;
3801 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
3803 sbi
->s_journal
->j_commit_callback
= ext4_journal_commit_callback
;
3806 * The journal may have updated the bg summary counts, so we
3807 * need to update the global counters.
3809 percpu_counter_set(&sbi
->s_freeclusters_counter
,
3810 ext4_count_free_clusters(sb
));
3811 percpu_counter_set(&sbi
->s_freeinodes_counter
,
3812 ext4_count_free_inodes(sb
));
3813 percpu_counter_set(&sbi
->s_dirs_counter
,
3814 ext4_count_dirs(sb
));
3815 percpu_counter_set(&sbi
->s_dirtyclusters_counter
, 0);
3819 * Get the # of file system overhead blocks from the
3820 * superblock if present.
3822 if (es
->s_overhead_clusters
)
3823 sbi
->s_overhead
= le32_to_cpu(es
->s_overhead_clusters
);
3825 err
= ext4_calculate_overhead(sb
);
3827 goto failed_mount_wq
;
3831 * The maximum number of concurrent works can be high and
3832 * concurrency isn't really necessary. Limit it to 1.
3834 EXT4_SB(sb
)->dio_unwritten_wq
=
3835 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM
| WQ_UNBOUND
, 1);
3836 if (!EXT4_SB(sb
)->dio_unwritten_wq
) {
3837 printk(KERN_ERR
"EXT4-fs: failed to create DIO workqueue\n");
3839 goto failed_mount_wq
;
3843 * The jbd2_journal_load will have done any necessary log recovery,
3844 * so we can safely mount the rest of the filesystem now.
3847 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
3849 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
3850 ret
= PTR_ERR(root
);
3854 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
3855 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
3859 sb
->s_root
= d_make_root(root
);
3861 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
3866 if (ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
))
3867 sb
->s_flags
|= MS_RDONLY
;
3869 /* determine the minimum size of new large inodes, if present */
3870 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
3871 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
3872 EXT4_GOOD_OLD_INODE_SIZE
;
3873 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3874 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE
)) {
3875 if (sbi
->s_want_extra_isize
<
3876 le16_to_cpu(es
->s_want_extra_isize
))
3877 sbi
->s_want_extra_isize
=
3878 le16_to_cpu(es
->s_want_extra_isize
);
3879 if (sbi
->s_want_extra_isize
<
3880 le16_to_cpu(es
->s_min_extra_isize
))
3881 sbi
->s_want_extra_isize
=
3882 le16_to_cpu(es
->s_min_extra_isize
);
3885 /* Check if enough inode space is available */
3886 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
3887 sbi
->s_inode_size
) {
3888 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
3889 EXT4_GOOD_OLD_INODE_SIZE
;
3890 ext4_msg(sb
, KERN_INFO
, "required extra inode space not"
3894 err
= ext4_setup_system_zone(sb
);
3896 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
3898 goto failed_mount4a
;
3902 err
= ext4_mb_init(sb
);
3904 ext4_msg(sb
, KERN_ERR
, "failed to initialize mballoc (%d)",
3909 err
= ext4_register_li_request(sb
, first_not_zeroed
);
3913 sbi
->s_kobj
.kset
= ext4_kset
;
3914 init_completion(&sbi
->s_kobj_unregister
);
3915 err
= kobject_init_and_add(&sbi
->s_kobj
, &ext4_ktype
, NULL
,
3920 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
3921 ext4_orphan_cleanup(sb
, es
);
3922 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
3923 if (needs_recovery
) {
3924 ext4_msg(sb
, KERN_INFO
, "recovery complete");
3925 ext4_mark_recovery_complete(sb
, es
);
3927 if (EXT4_SB(sb
)->s_journal
) {
3928 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
3929 descr
= " journalled data mode";
3930 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
3931 descr
= " ordered data mode";
3933 descr
= " writeback data mode";
3935 descr
= "out journal";
3938 /* Enable quota usage during mount. */
3939 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
) &&
3940 !(sb
->s_flags
& MS_RDONLY
)) {
3941 err
= ext4_enable_quotas(sb
);
3945 #endif /* CONFIG_QUOTA */
3947 if (test_opt(sb
, DISCARD
)) {
3948 struct request_queue
*q
= bdev_get_queue(sb
->s_bdev
);
3949 if (!blk_queue_discard(q
))
3950 ext4_msg(sb
, KERN_WARNING
,
3951 "mounting with \"discard\" option, but "
3952 "the device does not support discard");
3955 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s. "
3956 "Opts: %s%s%s", descr
, sbi
->s_es
->s_mount_opts
,
3957 *sbi
->s_es
->s_mount_opts
? "; " : "", orig_data
);
3959 if (es
->s_error_count
)
3960 mod_timer(&sbi
->s_err_report
, jiffies
+ 300*HZ
); /* 5 minutes */
3967 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
3972 kobject_del(&sbi
->s_kobj
);
3975 ext4_unregister_li_request(sb
);
3977 ext4_mb_release(sb
);
3979 ext4_ext_release(sb
);
3980 ext4_release_system_zone(sb
);
3985 ext4_msg(sb
, KERN_ERR
, "mount failed");
3986 destroy_workqueue(EXT4_SB(sb
)->dio_unwritten_wq
);
3988 if (sbi
->s_journal
) {
3989 jbd2_journal_destroy(sbi
->s_journal
);
3990 sbi
->s_journal
= NULL
;
3993 del_timer(&sbi
->s_err_report
);
3994 if (sbi
->s_flex_groups
)
3995 ext4_kvfree(sbi
->s_flex_groups
);
3996 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
3997 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
3998 percpu_counter_destroy(&sbi
->s_dirs_counter
);
3999 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
4000 percpu_counter_destroy(&sbi
->s_extent_cache_cnt
);
4002 kthread_stop(sbi
->s_mmp_tsk
);
4004 for (i
= 0; i
< db_count
; i
++)
4005 brelse(sbi
->s_group_desc
[i
]);
4006 ext4_kvfree(sbi
->s_group_desc
);
4008 if (sbi
->s_chksum_driver
)
4009 crypto_free_shash(sbi
->s_chksum_driver
);
4011 remove_proc_entry("options", sbi
->s_proc
);
4012 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
4015 for (i
= 0; i
< MAXQUOTAS
; i
++)
4016 kfree(sbi
->s_qf_names
[i
]);
4018 ext4_blkdev_remove(sbi
);
4021 sb
->s_fs_info
= NULL
;
4022 kfree(sbi
->s_blockgroup_lock
);
4026 return err
? err
: ret
;
4030 * Setup any per-fs journal parameters now. We'll do this both on
4031 * initial mount, once the journal has been initialised but before we've
4032 * done any recovery; and again on any subsequent remount.
4034 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
4036 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4038 journal
->j_commit_interval
= sbi
->s_commit_interval
;
4039 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
4040 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
4042 write_lock(&journal
->j_state_lock
);
4043 if (test_opt(sb
, BARRIER
))
4044 journal
->j_flags
|= JBD2_BARRIER
;
4046 journal
->j_flags
&= ~JBD2_BARRIER
;
4047 if (test_opt(sb
, DATA_ERR_ABORT
))
4048 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
4050 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
4051 write_unlock(&journal
->j_state_lock
);
4054 static journal_t
*ext4_get_journal(struct super_block
*sb
,
4055 unsigned int journal_inum
)
4057 struct inode
*journal_inode
;
4060 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4062 /* First, test for the existence of a valid inode on disk. Bad
4063 * things happen if we iget() an unused inode, as the subsequent
4064 * iput() will try to delete it. */
4066 journal_inode
= ext4_iget(sb
, journal_inum
);
4067 if (IS_ERR(journal_inode
)) {
4068 ext4_msg(sb
, KERN_ERR
, "no journal found");
4071 if (!journal_inode
->i_nlink
) {
4072 make_bad_inode(journal_inode
);
4073 iput(journal_inode
);
4074 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
4078 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4079 journal_inode
, journal_inode
->i_size
);
4080 if (!S_ISREG(journal_inode
->i_mode
)) {
4081 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
4082 iput(journal_inode
);
4086 journal
= jbd2_journal_init_inode(journal_inode
);
4088 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
4089 iput(journal_inode
);
4092 journal
->j_private
= sb
;
4093 ext4_init_journal_params(sb
, journal
);
4097 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
4100 struct buffer_head
*bh
;
4104 int hblock
, blocksize
;
4105 ext4_fsblk_t sb_block
;
4106 unsigned long offset
;
4107 struct ext4_super_block
*es
;
4108 struct block_device
*bdev
;
4110 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4112 bdev
= ext4_blkdev_get(j_dev
, sb
);
4116 blocksize
= sb
->s_blocksize
;
4117 hblock
= bdev_logical_block_size(bdev
);
4118 if (blocksize
< hblock
) {
4119 ext4_msg(sb
, KERN_ERR
,
4120 "blocksize too small for journal device");
4124 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
4125 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
4126 set_blocksize(bdev
, blocksize
);
4127 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
4128 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
4129 "external journal");
4133 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
4134 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
4135 !(le32_to_cpu(es
->s_feature_incompat
) &
4136 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
4137 ext4_msg(sb
, KERN_ERR
, "external journal has "
4143 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
4144 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
4149 len
= ext4_blocks_count(es
);
4150 start
= sb_block
+ 1;
4151 brelse(bh
); /* we're done with the superblock */
4153 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
4154 start
, len
, blocksize
);
4156 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
4159 journal
->j_private
= sb
;
4160 ll_rw_block(READ
, 1, &journal
->j_sb_buffer
);
4161 wait_on_buffer(journal
->j_sb_buffer
);
4162 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
4163 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
4166 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
4167 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
4168 "user (unsupported) - %d",
4169 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
4172 EXT4_SB(sb
)->journal_bdev
= bdev
;
4173 ext4_init_journal_params(sb
, journal
);
4177 jbd2_journal_destroy(journal
);
4179 ext4_blkdev_put(bdev
);
4183 static int ext4_load_journal(struct super_block
*sb
,
4184 struct ext4_super_block
*es
,
4185 unsigned long journal_devnum
)
4188 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
4191 int really_read_only
;
4193 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4195 if (journal_devnum
&&
4196 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4197 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
4198 "numbers have changed");
4199 journal_dev
= new_decode_dev(journal_devnum
);
4201 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
4203 really_read_only
= bdev_read_only(sb
->s_bdev
);
4206 * Are we loading a blank journal or performing recovery after a
4207 * crash? For recovery, we need to check in advance whether we
4208 * can get read-write access to the device.
4210 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
4211 if (sb
->s_flags
& MS_RDONLY
) {
4212 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
4213 "required on readonly filesystem");
4214 if (really_read_only
) {
4215 ext4_msg(sb
, KERN_ERR
, "write access "
4216 "unavailable, cannot proceed");
4219 ext4_msg(sb
, KERN_INFO
, "write access will "
4220 "be enabled during recovery");
4224 if (journal_inum
&& journal_dev
) {
4225 ext4_msg(sb
, KERN_ERR
, "filesystem has both journal "
4226 "and inode journals!");
4231 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
4234 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
4238 if (!(journal
->j_flags
& JBD2_BARRIER
))
4239 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
4241 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
))
4242 err
= jbd2_journal_wipe(journal
, !really_read_only
);
4244 char *save
= kmalloc(EXT4_S_ERR_LEN
, GFP_KERNEL
);
4246 memcpy(save
, ((char *) es
) +
4247 EXT4_S_ERR_START
, EXT4_S_ERR_LEN
);
4248 err
= jbd2_journal_load(journal
);
4250 memcpy(((char *) es
) + EXT4_S_ERR_START
,
4251 save
, EXT4_S_ERR_LEN
);
4256 ext4_msg(sb
, KERN_ERR
, "error loading journal");
4257 jbd2_journal_destroy(journal
);
4261 EXT4_SB(sb
)->s_journal
= journal
;
4262 ext4_clear_journal_err(sb
, es
);
4264 if (!really_read_only
&& journal_devnum
&&
4265 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4266 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
4268 /* Make sure we flush the recovery flag to disk. */
4269 ext4_commit_super(sb
, 1);
4275 static int ext4_commit_super(struct super_block
*sb
, int sync
)
4277 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
4278 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
4281 if (!sbh
|| block_device_ejected(sb
))
4283 if (buffer_write_io_error(sbh
)) {
4285 * Oh, dear. A previous attempt to write the
4286 * superblock failed. This could happen because the
4287 * USB device was yanked out. Or it could happen to
4288 * be a transient write error and maybe the block will
4289 * be remapped. Nothing we can do but to retry the
4290 * write and hope for the best.
4292 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
4293 "superblock detected");
4294 clear_buffer_write_io_error(sbh
);
4295 set_buffer_uptodate(sbh
);
4298 * If the file system is mounted read-only, don't update the
4299 * superblock write time. This avoids updating the superblock
4300 * write time when we are mounting the root file system
4301 * read/only but we need to replay the journal; at that point,
4302 * for people who are east of GMT and who make their clock
4303 * tick in localtime for Windows bug-for-bug compatibility,
4304 * the clock is set in the future, and this will cause e2fsck
4305 * to complain and force a full file system check.
4307 if (!(sb
->s_flags
& MS_RDONLY
))
4308 es
->s_wtime
= cpu_to_le32(get_seconds());
4309 if (sb
->s_bdev
->bd_part
)
4310 es
->s_kbytes_written
=
4311 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
+
4312 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
4313 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
4315 es
->s_kbytes_written
=
4316 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
);
4317 ext4_free_blocks_count_set(es
,
4318 EXT4_C2B(EXT4_SB(sb
), percpu_counter_sum_positive(
4319 &EXT4_SB(sb
)->s_freeclusters_counter
)));
4320 es
->s_free_inodes_count
=
4321 cpu_to_le32(percpu_counter_sum_positive(
4322 &EXT4_SB(sb
)->s_freeinodes_counter
));
4323 BUFFER_TRACE(sbh
, "marking dirty");
4324 ext4_superblock_csum_set(sb
);
4325 mark_buffer_dirty(sbh
);
4327 error
= sync_dirty_buffer(sbh
);
4331 error
= buffer_write_io_error(sbh
);
4333 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
4335 clear_buffer_write_io_error(sbh
);
4336 set_buffer_uptodate(sbh
);
4343 * Have we just finished recovery? If so, and if we are mounting (or
4344 * remounting) the filesystem readonly, then we will end up with a
4345 * consistent fs on disk. Record that fact.
4347 static void ext4_mark_recovery_complete(struct super_block
*sb
,
4348 struct ext4_super_block
*es
)
4350 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
4352 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
4353 BUG_ON(journal
!= NULL
);
4356 jbd2_journal_lock_updates(journal
);
4357 if (jbd2_journal_flush(journal
) < 0)
4360 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
) &&
4361 sb
->s_flags
& MS_RDONLY
) {
4362 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4363 ext4_commit_super(sb
, 1);
4367 jbd2_journal_unlock_updates(journal
);
4371 * If we are mounting (or read-write remounting) a filesystem whose journal
4372 * has recorded an error from a previous lifetime, move that error to the
4373 * main filesystem now.
4375 static void ext4_clear_journal_err(struct super_block
*sb
,
4376 struct ext4_super_block
*es
)
4382 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4384 journal
= EXT4_SB(sb
)->s_journal
;
4387 * Now check for any error status which may have been recorded in the
4388 * journal by a prior ext4_error() or ext4_abort()
4391 j_errno
= jbd2_journal_errno(journal
);
4395 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
4396 ext4_warning(sb
, "Filesystem error recorded "
4397 "from previous mount: %s", errstr
);
4398 ext4_warning(sb
, "Marking fs in need of filesystem check.");
4400 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
4401 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
4402 ext4_commit_super(sb
, 1);
4404 jbd2_journal_clear_err(journal
);
4405 jbd2_journal_update_sb_errno(journal
);
4410 * Force the running and committing transactions to commit,
4411 * and wait on the commit.
4413 int ext4_force_commit(struct super_block
*sb
)
4417 if (sb
->s_flags
& MS_RDONLY
)
4420 journal
= EXT4_SB(sb
)->s_journal
;
4421 return ext4_journal_force_commit(journal
);
4424 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
4428 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4430 trace_ext4_sync_fs(sb
, wait
);
4431 flush_workqueue(sbi
->dio_unwritten_wq
);
4433 * Writeback quota in non-journalled quota case - journalled quota has
4436 dquot_writeback_dquots(sb
, -1);
4437 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
4439 jbd2_log_wait_commit(sbi
->s_journal
, target
);
4445 * LVM calls this function before a (read-only) snapshot is created. This
4446 * gives us a chance to flush the journal completely and mark the fs clean.
4448 * Note that only this function cannot bring a filesystem to be in a clean
4449 * state independently. It relies on upper layer to stop all data & metadata
4452 static int ext4_freeze(struct super_block
*sb
)
4457 if (sb
->s_flags
& MS_RDONLY
)
4460 journal
= EXT4_SB(sb
)->s_journal
;
4462 /* Now we set up the journal barrier. */
4463 jbd2_journal_lock_updates(journal
);
4466 * Don't clear the needs_recovery flag if we failed to flush
4469 error
= jbd2_journal_flush(journal
);
4473 /* Journal blocked and flushed, clear needs_recovery flag. */
4474 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4475 error
= ext4_commit_super(sb
, 1);
4477 /* we rely on upper layer to stop further updates */
4478 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
4483 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4484 * flag here, even though the filesystem is not technically dirty yet.
4486 static int ext4_unfreeze(struct super_block
*sb
)
4488 if (sb
->s_flags
& MS_RDONLY
)
4491 /* Reset the needs_recovery flag before the fs is unlocked. */
4492 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4493 ext4_commit_super(sb
, 1);
4498 * Structure to save mount options for ext4_remount's benefit
4500 struct ext4_mount_options
{
4501 unsigned long s_mount_opt
;
4502 unsigned long s_mount_opt2
;
4505 unsigned long s_commit_interval
;
4506 u32 s_min_batch_time
, s_max_batch_time
;
4509 char *s_qf_names
[MAXQUOTAS
];
4513 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
4515 struct ext4_super_block
*es
;
4516 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4517 unsigned long old_sb_flags
;
4518 struct ext4_mount_options old_opts
;
4519 int enable_quota
= 0;
4521 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
4526 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
4528 /* Store the original options */
4529 old_sb_flags
= sb
->s_flags
;
4530 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
4531 old_opts
.s_mount_opt2
= sbi
->s_mount_opt2
;
4532 old_opts
.s_resuid
= sbi
->s_resuid
;
4533 old_opts
.s_resgid
= sbi
->s_resgid
;
4534 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
4535 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
4536 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
4538 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
4539 for (i
= 0; i
< MAXQUOTAS
; i
++)
4540 if (sbi
->s_qf_names
[i
]) {
4541 old_opts
.s_qf_names
[i
] = kstrdup(sbi
->s_qf_names
[i
],
4543 if (!old_opts
.s_qf_names
[i
]) {
4544 for (j
= 0; j
< i
; j
++)
4545 kfree(old_opts
.s_qf_names
[j
]);
4549 old_opts
.s_qf_names
[i
] = NULL
;
4551 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
4552 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
4555 * Allow the "check" option to be passed as a remount option.
4557 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
, 1)) {
4562 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
)
4563 ext4_abort(sb
, "Abort forced by user");
4565 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
4566 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
4570 if (sbi
->s_journal
) {
4571 ext4_init_journal_params(sb
, sbi
->s_journal
);
4572 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
4575 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
)) {
4576 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
) {
4581 if (*flags
& MS_RDONLY
) {
4582 err
= dquot_suspend(sb
, -1);
4587 * First of all, the unconditional stuff we have to do
4588 * to disable replay of the journal when we next remount
4590 sb
->s_flags
|= MS_RDONLY
;
4593 * OK, test if we are remounting a valid rw partition
4594 * readonly, and if so set the rdonly flag and then
4595 * mark the partition as valid again.
4597 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
4598 (sbi
->s_mount_state
& EXT4_VALID_FS
))
4599 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
4602 ext4_mark_recovery_complete(sb
, es
);
4604 /* Make sure we can mount this feature set readwrite */
4605 if (!ext4_feature_set_ok(sb
, 0)) {
4610 * Make sure the group descriptor checksums
4611 * are sane. If they aren't, refuse to remount r/w.
4613 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
4614 struct ext4_group_desc
*gdp
=
4615 ext4_get_group_desc(sb
, g
, NULL
);
4617 if (!ext4_group_desc_csum_verify(sb
, g
, gdp
)) {
4618 ext4_msg(sb
, KERN_ERR
,
4619 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4620 g
, le16_to_cpu(ext4_group_desc_csum(sbi
, g
, gdp
)),
4621 le16_to_cpu(gdp
->bg_checksum
));
4628 * If we have an unprocessed orphan list hanging
4629 * around from a previously readonly bdev mount,
4630 * require a full umount/remount for now.
4632 if (es
->s_last_orphan
) {
4633 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
4634 "remount RDWR because of unprocessed "
4635 "orphan inode list. Please "
4636 "umount/remount instead");
4642 * Mounting a RDONLY partition read-write, so reread
4643 * and store the current valid flag. (It may have
4644 * been changed by e2fsck since we originally mounted
4648 ext4_clear_journal_err(sb
, es
);
4649 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
4650 if (!ext4_setup_super(sb
, es
, 0))
4651 sb
->s_flags
&= ~MS_RDONLY
;
4652 if (EXT4_HAS_INCOMPAT_FEATURE(sb
,
4653 EXT4_FEATURE_INCOMPAT_MMP
))
4654 if (ext4_multi_mount_protect(sb
,
4655 le64_to_cpu(es
->s_mmp_block
))) {
4664 * Reinitialize lazy itable initialization thread based on
4667 if ((sb
->s_flags
& MS_RDONLY
) || !test_opt(sb
, INIT_INODE_TABLE
))
4668 ext4_unregister_li_request(sb
);
4670 ext4_group_t first_not_zeroed
;
4671 first_not_zeroed
= ext4_has_uninit_itable(sb
);
4672 ext4_register_li_request(sb
, first_not_zeroed
);
4675 ext4_setup_system_zone(sb
);
4676 if (sbi
->s_journal
== NULL
&& !(old_sb_flags
& MS_RDONLY
))
4677 ext4_commit_super(sb
, 1);
4680 /* Release old quota file names */
4681 for (i
= 0; i
< MAXQUOTAS
; i
++)
4682 kfree(old_opts
.s_qf_names
[i
]);
4684 if (sb_any_quota_suspended(sb
))
4685 dquot_resume(sb
, -1);
4686 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
4687 EXT4_FEATURE_RO_COMPAT_QUOTA
)) {
4688 err
= ext4_enable_quotas(sb
);
4695 ext4_msg(sb
, KERN_INFO
, "re-mounted. Opts: %s", orig_data
);
4700 sb
->s_flags
= old_sb_flags
;
4701 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
4702 sbi
->s_mount_opt2
= old_opts
.s_mount_opt2
;
4703 sbi
->s_resuid
= old_opts
.s_resuid
;
4704 sbi
->s_resgid
= old_opts
.s_resgid
;
4705 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
4706 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
4707 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
4709 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
4710 for (i
= 0; i
< MAXQUOTAS
; i
++) {
4711 kfree(sbi
->s_qf_names
[i
]);
4712 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
4719 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
4721 struct super_block
*sb
= dentry
->d_sb
;
4722 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4723 struct ext4_super_block
*es
= sbi
->s_es
;
4724 ext4_fsblk_t overhead
= 0;
4728 if (!test_opt(sb
, MINIX_DF
))
4729 overhead
= sbi
->s_overhead
;
4731 buf
->f_type
= EXT4_SUPER_MAGIC
;
4732 buf
->f_bsize
= sb
->s_blocksize
;
4733 buf
->f_blocks
= ext4_blocks_count(es
) - EXT4_C2B(sbi
, overhead
);
4734 bfree
= percpu_counter_sum_positive(&sbi
->s_freeclusters_counter
) -
4735 percpu_counter_sum_positive(&sbi
->s_dirtyclusters_counter
);
4736 /* prevent underflow in case that few free space is available */
4737 buf
->f_bfree
= EXT4_C2B(sbi
, max_t(s64
, bfree
, 0));
4738 buf
->f_bavail
= buf
->f_bfree
- ext4_r_blocks_count(es
);
4739 if (buf
->f_bfree
< ext4_r_blocks_count(es
))
4741 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
4742 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
4743 buf
->f_namelen
= EXT4_NAME_LEN
;
4744 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
4745 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
4746 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
4747 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
4752 /* Helper function for writing quotas on sync - we need to start transaction
4753 * before quota file is locked for write. Otherwise the are possible deadlocks:
4754 * Process 1 Process 2
4755 * ext4_create() quota_sync()
4756 * jbd2_journal_start() write_dquot()
4757 * dquot_initialize() down(dqio_mutex)
4758 * down(dqio_mutex) jbd2_journal_start()
4764 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
4766 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_id
.type
];
4769 static int ext4_write_dquot(struct dquot
*dquot
)
4773 struct inode
*inode
;
4775 inode
= dquot_to_inode(dquot
);
4776 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
,
4777 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
4779 return PTR_ERR(handle
);
4780 ret
= dquot_commit(dquot
);
4781 err
= ext4_journal_stop(handle
);
4787 static int ext4_acquire_dquot(struct dquot
*dquot
)
4792 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
4793 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
4795 return PTR_ERR(handle
);
4796 ret
= dquot_acquire(dquot
);
4797 err
= ext4_journal_stop(handle
);
4803 static int ext4_release_dquot(struct dquot
*dquot
)
4808 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
4809 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
4810 if (IS_ERR(handle
)) {
4811 /* Release dquot anyway to avoid endless cycle in dqput() */
4812 dquot_release(dquot
);
4813 return PTR_ERR(handle
);
4815 ret
= dquot_release(dquot
);
4816 err
= ext4_journal_stop(handle
);
4822 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
4824 /* Are we journaling quotas? */
4825 if (EXT4_SB(dquot
->dq_sb
)->s_qf_names
[USRQUOTA
] ||
4826 EXT4_SB(dquot
->dq_sb
)->s_qf_names
[GRPQUOTA
]) {
4827 dquot_mark_dquot_dirty(dquot
);
4828 return ext4_write_dquot(dquot
);
4830 return dquot_mark_dquot_dirty(dquot
);
4834 static int ext4_write_info(struct super_block
*sb
, int type
)
4839 /* Data block + inode block */
4840 handle
= ext4_journal_start(sb
->s_root
->d_inode
, EXT4_HT_QUOTA
, 2);
4842 return PTR_ERR(handle
);
4843 ret
= dquot_commit_info(sb
, type
);
4844 err
= ext4_journal_stop(handle
);
4851 * Turn on quotas during mount time - we need to find
4852 * the quota file and such...
4854 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
4856 return dquot_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
4857 EXT4_SB(sb
)->s_jquota_fmt
, type
);
4861 * Standard function to be called on quota_on
4863 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
4868 if (!test_opt(sb
, QUOTA
))
4871 /* Quotafile not on the same filesystem? */
4872 if (path
->dentry
->d_sb
!= sb
)
4874 /* Journaling quota? */
4875 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
4876 /* Quotafile not in fs root? */
4877 if (path
->dentry
->d_parent
!= sb
->s_root
)
4878 ext4_msg(sb
, KERN_WARNING
,
4879 "Quota file not on filesystem root. "
4880 "Journaled quota will not work");
4884 * When we journal data on quota file, we have to flush journal to see
4885 * all updates to the file when we bypass pagecache...
4887 if (EXT4_SB(sb
)->s_journal
&&
4888 ext4_should_journal_data(path
->dentry
->d_inode
)) {
4890 * We don't need to lock updates but journal_flush() could
4891 * otherwise be livelocked...
4893 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
4894 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
4895 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
4900 return dquot_quota_on(sb
, type
, format_id
, path
);
4903 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
4907 struct inode
*qf_inode
;
4908 unsigned long qf_inums
[MAXQUOTAS
] = {
4909 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
4910 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
)
4913 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
));
4915 if (!qf_inums
[type
])
4918 qf_inode
= ext4_iget(sb
, qf_inums
[type
]);
4919 if (IS_ERR(qf_inode
)) {
4920 ext4_error(sb
, "Bad quota inode # %lu", qf_inums
[type
]);
4921 return PTR_ERR(qf_inode
);
4924 err
= dquot_enable(qf_inode
, type
, format_id
, flags
);
4930 /* Enable usage tracking for all quota types. */
4931 static int ext4_enable_quotas(struct super_block
*sb
)
4934 unsigned long qf_inums
[MAXQUOTAS
] = {
4935 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
4936 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
)
4939 sb_dqopt(sb
)->flags
|= DQUOT_QUOTA_SYS_FILE
;
4940 for (type
= 0; type
< MAXQUOTAS
; type
++) {
4941 if (qf_inums
[type
]) {
4942 err
= ext4_quota_enable(sb
, type
, QFMT_VFS_V1
,
4943 DQUOT_USAGE_ENABLED
);
4946 "Failed to enable quota tracking "
4947 "(type=%d, err=%d). Please run "
4948 "e2fsck to fix.", type
, err
);
4957 * quota_on function that is used when QUOTA feature is set.
4959 static int ext4_quota_on_sysfile(struct super_block
*sb
, int type
,
4962 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
))
4966 * USAGE was enabled at mount time. Only need to enable LIMITS now.
4968 return ext4_quota_enable(sb
, type
, format_id
, DQUOT_LIMITS_ENABLED
);
4971 static int ext4_quota_off(struct super_block
*sb
, int type
)
4973 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
4976 /* Force all delayed allocation blocks to be allocated.
4977 * Caller already holds s_umount sem */
4978 if (test_opt(sb
, DELALLOC
))
4979 sync_filesystem(sb
);
4984 /* Update modification times of quota files when userspace can
4985 * start looking at them */
4986 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
, 1);
4989 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
4990 ext4_mark_inode_dirty(handle
, inode
);
4991 ext4_journal_stop(handle
);
4994 return dquot_quota_off(sb
, type
);
4998 * quota_off function that is used when QUOTA feature is set.
5000 static int ext4_quota_off_sysfile(struct super_block
*sb
, int type
)
5002 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
))
5005 /* Disable only the limits. */
5006 return dquot_disable(sb
, type
, DQUOT_LIMITS_ENABLED
);
5009 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5010 * acquiring the locks... As quota files are never truncated and quota code
5011 * itself serializes the operations (and no one else should touch the files)
5012 * we don't have to be afraid of races */
5013 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
5014 size_t len
, loff_t off
)
5016 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5017 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5019 int offset
= off
& (sb
->s_blocksize
- 1);
5022 struct buffer_head
*bh
;
5023 loff_t i_size
= i_size_read(inode
);
5027 if (off
+len
> i_size
)
5030 while (toread
> 0) {
5031 tocopy
= sb
->s_blocksize
- offset
< toread
?
5032 sb
->s_blocksize
- offset
: toread
;
5033 bh
= ext4_bread(NULL
, inode
, blk
, 0, &err
);
5036 if (!bh
) /* A hole? */
5037 memset(data
, 0, tocopy
);
5039 memcpy(data
, bh
->b_data
+offset
, tocopy
);
5049 /* Write to quotafile (we know the transaction is already started and has
5050 * enough credits) */
5051 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
5052 const char *data
, size_t len
, loff_t off
)
5054 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5055 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5057 int offset
= off
& (sb
->s_blocksize
- 1);
5058 struct buffer_head
*bh
;
5059 handle_t
*handle
= journal_current_handle();
5061 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
5062 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5063 " cancelled because transaction is not started",
5064 (unsigned long long)off
, (unsigned long long)len
);
5068 * Since we account only one data block in transaction credits,
5069 * then it is impossible to cross a block boundary.
5071 if (sb
->s_blocksize
- offset
< len
) {
5072 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5073 " cancelled because not block aligned",
5074 (unsigned long long)off
, (unsigned long long)len
);
5078 bh
= ext4_bread(handle
, inode
, blk
, 1, &err
);
5081 err
= ext4_journal_get_write_access(handle
, bh
);
5087 memcpy(bh
->b_data
+offset
, data
, len
);
5088 flush_dcache_page(bh
->b_page
);
5090 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
5095 if (inode
->i_size
< off
+ len
) {
5096 i_size_write(inode
, off
+ len
);
5097 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
5098 ext4_mark_inode_dirty(handle
, inode
);
5105 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
5106 const char *dev_name
, void *data
)
5108 return mount_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
);
5111 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5112 static inline void register_as_ext2(void)
5114 int err
= register_filesystem(&ext2_fs_type
);
5117 "EXT4-fs: Unable to register as ext2 (%d)\n", err
);
5120 static inline void unregister_as_ext2(void)
5122 unregister_filesystem(&ext2_fs_type
);
5125 static inline int ext2_feature_set_ok(struct super_block
*sb
)
5127 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT2_FEATURE_INCOMPAT_SUPP
))
5129 if (sb
->s_flags
& MS_RDONLY
)
5131 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT2_FEATURE_RO_COMPAT_SUPP
))
5135 MODULE_ALIAS("ext2");
5137 static inline void register_as_ext2(void) { }
5138 static inline void unregister_as_ext2(void) { }
5139 static inline int ext2_feature_set_ok(struct super_block
*sb
) { return 0; }
5142 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5143 static inline void register_as_ext3(void)
5145 int err
= register_filesystem(&ext3_fs_type
);
5148 "EXT4-fs: Unable to register as ext3 (%d)\n", err
);
5151 static inline void unregister_as_ext3(void)
5153 unregister_filesystem(&ext3_fs_type
);
5156 static inline int ext3_feature_set_ok(struct super_block
*sb
)
5158 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT3_FEATURE_INCOMPAT_SUPP
))
5160 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
5162 if (sb
->s_flags
& MS_RDONLY
)
5164 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT3_FEATURE_RO_COMPAT_SUPP
))
5168 MODULE_ALIAS("ext3");
5170 static inline void register_as_ext3(void) { }
5171 static inline void unregister_as_ext3(void) { }
5172 static inline int ext3_feature_set_ok(struct super_block
*sb
) { return 0; }
5175 static struct file_system_type ext4_fs_type
= {
5176 .owner
= THIS_MODULE
,
5178 .mount
= ext4_mount
,
5179 .kill_sb
= kill_block_super
,
5180 .fs_flags
= FS_REQUIRES_DEV
,
5183 static int __init
ext4_init_feat_adverts(void)
5185 struct ext4_features
*ef
;
5188 ef
= kzalloc(sizeof(struct ext4_features
), GFP_KERNEL
);
5192 ef
->f_kobj
.kset
= ext4_kset
;
5193 init_completion(&ef
->f_kobj_unregister
);
5194 ret
= kobject_init_and_add(&ef
->f_kobj
, &ext4_feat_ktype
, NULL
,
5207 static void ext4_exit_feat_adverts(void)
5209 kobject_put(&ext4_feat
->f_kobj
);
5210 wait_for_completion(&ext4_feat
->f_kobj_unregister
);
5214 /* Shared across all ext4 file systems */
5215 wait_queue_head_t ext4__ioend_wq
[EXT4_WQ_HASH_SZ
];
5216 struct mutex ext4__aio_mutex
[EXT4_WQ_HASH_SZ
];
5218 static int __init
ext4_init_fs(void)
5222 ext4_li_info
= NULL
;
5223 mutex_init(&ext4_li_mtx
);
5225 /* Build-time check for flags consistency */
5226 ext4_check_flag_values();
5228 for (i
= 0; i
< EXT4_WQ_HASH_SZ
; i
++) {
5229 mutex_init(&ext4__aio_mutex
[i
]);
5230 init_waitqueue_head(&ext4__ioend_wq
[i
]);
5233 err
= ext4_init_es();
5237 err
= ext4_init_pageio();
5241 err
= ext4_init_system_zone();
5244 ext4_kset
= kset_create_and_add("ext4", NULL
, fs_kobj
);
5249 ext4_proc_root
= proc_mkdir("fs/ext4", NULL
);
5251 err
= ext4_init_feat_adverts();
5255 err
= ext4_init_mballoc();
5259 err
= ext4_init_xattr();
5262 err
= init_inodecache();
5267 err
= register_filesystem(&ext4_fs_type
);
5273 unregister_as_ext2();
5274 unregister_as_ext3();
5275 destroy_inodecache();
5279 ext4_exit_mballoc();
5281 ext4_exit_feat_adverts();
5284 remove_proc_entry("fs/ext4", NULL
);
5285 kset_unregister(ext4_kset
);
5287 ext4_exit_system_zone();
5296 static void __exit
ext4_exit_fs(void)
5298 ext4_destroy_lazyinit_thread();
5299 unregister_as_ext2();
5300 unregister_as_ext3();
5301 unregister_filesystem(&ext4_fs_type
);
5302 destroy_inodecache();
5304 ext4_exit_mballoc();
5305 ext4_exit_feat_adverts();
5306 remove_proc_entry("fs/ext4", NULL
);
5307 kset_unregister(ext4_kset
);
5308 ext4_exit_system_zone();
5312 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5313 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5314 MODULE_LICENSE("GPL");
5315 module_init(ext4_init_fs
)
5316 module_exit(ext4_exit_fs
)