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 del_timer(&sbi
->s_err_report
);
759 ext4_release_system_zone(sb
);
761 ext4_ext_release(sb
);
762 ext4_xattr_put_super(sb
);
764 if (!(sb
->s_flags
& MS_RDONLY
)) {
765 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
766 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
768 if (!(sb
->s_flags
& MS_RDONLY
))
769 ext4_commit_super(sb
, 1);
772 remove_proc_entry("options", sbi
->s_proc
);
773 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
775 kobject_del(&sbi
->s_kobj
);
777 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
778 brelse(sbi
->s_group_desc
[i
]);
779 ext4_kvfree(sbi
->s_group_desc
);
780 ext4_kvfree(sbi
->s_flex_groups
);
781 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
782 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
783 percpu_counter_destroy(&sbi
->s_dirs_counter
);
784 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
787 for (i
= 0; i
< MAXQUOTAS
; i
++)
788 kfree(sbi
->s_qf_names
[i
]);
791 /* Debugging code just in case the in-memory inode orphan list
792 * isn't empty. The on-disk one can be non-empty if we've
793 * detected an error and taken the fs readonly, but the
794 * in-memory list had better be clean by this point. */
795 if (!list_empty(&sbi
->s_orphan
))
796 dump_orphan_list(sb
, sbi
);
797 J_ASSERT(list_empty(&sbi
->s_orphan
));
799 invalidate_bdev(sb
->s_bdev
);
800 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
802 * Invalidate the journal device's buffers. We don't want them
803 * floating about in memory - the physical journal device may
804 * hotswapped, and it breaks the `ro-after' testing code.
806 sync_blockdev(sbi
->journal_bdev
);
807 invalidate_bdev(sbi
->journal_bdev
);
808 ext4_blkdev_remove(sbi
);
811 kthread_stop(sbi
->s_mmp_tsk
);
812 sb
->s_fs_info
= NULL
;
814 * Now that we are completely done shutting down the
815 * superblock, we need to actually destroy the kobject.
817 kobject_put(&sbi
->s_kobj
);
818 wait_for_completion(&sbi
->s_kobj_unregister
);
819 if (sbi
->s_chksum_driver
)
820 crypto_free_shash(sbi
->s_chksum_driver
);
821 kfree(sbi
->s_blockgroup_lock
);
825 static struct kmem_cache
*ext4_inode_cachep
;
828 * Called inside transaction, so use GFP_NOFS
830 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
832 struct ext4_inode_info
*ei
;
834 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
838 ei
->vfs_inode
.i_version
= 1;
839 memset(&ei
->i_cached_extent
, 0, sizeof(struct ext4_ext_cache
));
840 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
841 spin_lock_init(&ei
->i_prealloc_lock
);
842 ext4_es_init_tree(&ei
->i_es_tree
);
843 rwlock_init(&ei
->i_es_lock
);
844 ei
->i_reserved_data_blocks
= 0;
845 ei
->i_reserved_meta_blocks
= 0;
846 ei
->i_allocated_meta_blocks
= 0;
847 ei
->i_da_metadata_calc_len
= 0;
848 ei
->i_da_metadata_calc_last_lblock
= 0;
849 spin_lock_init(&(ei
->i_block_reservation_lock
));
851 ei
->i_reserved_quota
= 0;
854 INIT_LIST_HEAD(&ei
->i_completed_io_list
);
855 spin_lock_init(&ei
->i_completed_io_lock
);
857 ei
->i_datasync_tid
= 0;
858 atomic_set(&ei
->i_ioend_count
, 0);
859 atomic_set(&ei
->i_unwritten
, 0);
860 INIT_WORK(&ei
->i_unwritten_work
, ext4_end_io_work
);
862 return &ei
->vfs_inode
;
865 static int ext4_drop_inode(struct inode
*inode
)
867 int drop
= generic_drop_inode(inode
);
869 trace_ext4_drop_inode(inode
, drop
);
873 static void ext4_i_callback(struct rcu_head
*head
)
875 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
876 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
879 static void ext4_destroy_inode(struct inode
*inode
)
881 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
882 ext4_msg(inode
->i_sb
, KERN_ERR
,
883 "Inode %lu (%p): orphan list check failed!",
884 inode
->i_ino
, EXT4_I(inode
));
885 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
886 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
890 call_rcu(&inode
->i_rcu
, ext4_i_callback
);
893 static void init_once(void *foo
)
895 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
897 INIT_LIST_HEAD(&ei
->i_orphan
);
898 init_rwsem(&ei
->xattr_sem
);
899 init_rwsem(&ei
->i_data_sem
);
900 inode_init_once(&ei
->vfs_inode
);
903 static int init_inodecache(void)
905 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
906 sizeof(struct ext4_inode_info
),
907 0, (SLAB_RECLAIM_ACCOUNT
|
910 if (ext4_inode_cachep
== NULL
)
915 static void destroy_inodecache(void)
918 * Make sure all delayed rcu free inodes are flushed before we
922 kmem_cache_destroy(ext4_inode_cachep
);
925 void ext4_clear_inode(struct inode
*inode
)
927 invalidate_inode_buffers(inode
);
930 ext4_discard_preallocations(inode
);
931 ext4_es_remove_extent(inode
, 0, EXT_MAX_BLOCKS
);
932 if (EXT4_I(inode
)->jinode
) {
933 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode
),
934 EXT4_I(inode
)->jinode
);
935 jbd2_free_inode(EXT4_I(inode
)->jinode
);
936 EXT4_I(inode
)->jinode
= NULL
;
940 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
941 u64 ino
, u32 generation
)
945 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
946 return ERR_PTR(-ESTALE
);
947 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
948 return ERR_PTR(-ESTALE
);
950 /* iget isn't really right if the inode is currently unallocated!!
952 * ext4_read_inode will return a bad_inode if the inode had been
953 * deleted, so we should be safe.
955 * Currently we don't know the generation for parent directory, so
956 * a generation of 0 means "accept any"
958 inode
= ext4_iget(sb
, ino
);
960 return ERR_CAST(inode
);
961 if (generation
&& inode
->i_generation
!= generation
) {
963 return ERR_PTR(-ESTALE
);
969 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
970 int fh_len
, int fh_type
)
972 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
976 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
977 int fh_len
, int fh_type
)
979 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
984 * Try to release metadata pages (indirect blocks, directories) which are
985 * mapped via the block device. Since these pages could have journal heads
986 * which would prevent try_to_free_buffers() from freeing them, we must use
987 * jbd2 layer's try_to_free_buffers() function to release them.
989 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
992 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
994 WARN_ON(PageChecked(page
));
995 if (!page_has_buffers(page
))
998 return jbd2_journal_try_to_free_buffers(journal
, page
,
1000 return try_to_free_buffers(page
);
1004 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1005 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1007 static int ext4_write_dquot(struct dquot
*dquot
);
1008 static int ext4_acquire_dquot(struct dquot
*dquot
);
1009 static int ext4_release_dquot(struct dquot
*dquot
);
1010 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
1011 static int ext4_write_info(struct super_block
*sb
, int type
);
1012 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
1014 static int ext4_quota_on_sysfile(struct super_block
*sb
, int type
,
1016 static int ext4_quota_off(struct super_block
*sb
, int type
);
1017 static int ext4_quota_off_sysfile(struct super_block
*sb
, int type
);
1018 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
1019 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
1020 size_t len
, loff_t off
);
1021 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
1022 const char *data
, size_t len
, loff_t off
);
1023 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
1024 unsigned int flags
);
1025 static int ext4_enable_quotas(struct super_block
*sb
);
1027 static const struct dquot_operations ext4_quota_operations
= {
1028 .get_reserved_space
= ext4_get_reserved_space
,
1029 .write_dquot
= ext4_write_dquot
,
1030 .acquire_dquot
= ext4_acquire_dquot
,
1031 .release_dquot
= ext4_release_dquot
,
1032 .mark_dirty
= ext4_mark_dquot_dirty
,
1033 .write_info
= ext4_write_info
,
1034 .alloc_dquot
= dquot_alloc
,
1035 .destroy_dquot
= dquot_destroy
,
1038 static const struct quotactl_ops ext4_qctl_operations
= {
1039 .quota_on
= ext4_quota_on
,
1040 .quota_off
= ext4_quota_off
,
1041 .quota_sync
= dquot_quota_sync
,
1042 .get_info
= dquot_get_dqinfo
,
1043 .set_info
= dquot_set_dqinfo
,
1044 .get_dqblk
= dquot_get_dqblk
,
1045 .set_dqblk
= dquot_set_dqblk
1048 static const struct quotactl_ops ext4_qctl_sysfile_operations
= {
1049 .quota_on_meta
= ext4_quota_on_sysfile
,
1050 .quota_off
= ext4_quota_off_sysfile
,
1051 .quota_sync
= dquot_quota_sync
,
1052 .get_info
= dquot_get_dqinfo
,
1053 .set_info
= dquot_set_dqinfo
,
1054 .get_dqblk
= dquot_get_dqblk
,
1055 .set_dqblk
= dquot_set_dqblk
1059 static const struct super_operations ext4_sops
= {
1060 .alloc_inode
= ext4_alloc_inode
,
1061 .destroy_inode
= ext4_destroy_inode
,
1062 .write_inode
= ext4_write_inode
,
1063 .dirty_inode
= ext4_dirty_inode
,
1064 .drop_inode
= ext4_drop_inode
,
1065 .evict_inode
= ext4_evict_inode
,
1066 .put_super
= ext4_put_super
,
1067 .sync_fs
= ext4_sync_fs
,
1068 .freeze_fs
= ext4_freeze
,
1069 .unfreeze_fs
= ext4_unfreeze
,
1070 .statfs
= ext4_statfs
,
1071 .remount_fs
= ext4_remount
,
1072 .show_options
= ext4_show_options
,
1074 .quota_read
= ext4_quota_read
,
1075 .quota_write
= ext4_quota_write
,
1077 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1080 static const struct super_operations ext4_nojournal_sops
= {
1081 .alloc_inode
= ext4_alloc_inode
,
1082 .destroy_inode
= ext4_destroy_inode
,
1083 .write_inode
= ext4_write_inode
,
1084 .dirty_inode
= ext4_dirty_inode
,
1085 .drop_inode
= ext4_drop_inode
,
1086 .evict_inode
= ext4_evict_inode
,
1087 .put_super
= ext4_put_super
,
1088 .statfs
= ext4_statfs
,
1089 .remount_fs
= ext4_remount
,
1090 .show_options
= ext4_show_options
,
1092 .quota_read
= ext4_quota_read
,
1093 .quota_write
= ext4_quota_write
,
1095 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1098 static const struct export_operations ext4_export_ops
= {
1099 .fh_to_dentry
= ext4_fh_to_dentry
,
1100 .fh_to_parent
= ext4_fh_to_parent
,
1101 .get_parent
= ext4_get_parent
,
1105 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1106 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1107 Opt_nouid32
, Opt_debug
, Opt_removed
,
1108 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1109 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
,
1110 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
,
1111 Opt_journal_dev
, Opt_journal_checksum
, Opt_journal_async_commit
,
1112 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1113 Opt_data_err_abort
, Opt_data_err_ignore
,
1114 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1115 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
1116 Opt_noquota
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
1117 Opt_usrquota
, Opt_grpquota
, Opt_i_version
,
1118 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
, Opt_mblk_io_submit
,
1119 Opt_nomblk_io_submit
, Opt_block_validity
, Opt_noblock_validity
,
1120 Opt_inode_readahead_blks
, Opt_journal_ioprio
,
1121 Opt_dioread_nolock
, Opt_dioread_lock
,
1122 Opt_discard
, Opt_nodiscard
, Opt_init_itable
, Opt_noinit_itable
,
1123 Opt_max_dir_size_kb
,
1126 static const match_table_t tokens
= {
1127 {Opt_bsd_df
, "bsddf"},
1128 {Opt_minix_df
, "minixdf"},
1129 {Opt_grpid
, "grpid"},
1130 {Opt_grpid
, "bsdgroups"},
1131 {Opt_nogrpid
, "nogrpid"},
1132 {Opt_nogrpid
, "sysvgroups"},
1133 {Opt_resgid
, "resgid=%u"},
1134 {Opt_resuid
, "resuid=%u"},
1136 {Opt_err_cont
, "errors=continue"},
1137 {Opt_err_panic
, "errors=panic"},
1138 {Opt_err_ro
, "errors=remount-ro"},
1139 {Opt_nouid32
, "nouid32"},
1140 {Opt_debug
, "debug"},
1141 {Opt_removed
, "oldalloc"},
1142 {Opt_removed
, "orlov"},
1143 {Opt_user_xattr
, "user_xattr"},
1144 {Opt_nouser_xattr
, "nouser_xattr"},
1146 {Opt_noacl
, "noacl"},
1147 {Opt_noload
, "norecovery"},
1148 {Opt_noload
, "noload"},
1149 {Opt_removed
, "nobh"},
1150 {Opt_removed
, "bh"},
1151 {Opt_commit
, "commit=%u"},
1152 {Opt_min_batch_time
, "min_batch_time=%u"},
1153 {Opt_max_batch_time
, "max_batch_time=%u"},
1154 {Opt_journal_dev
, "journal_dev=%u"},
1155 {Opt_journal_checksum
, "journal_checksum"},
1156 {Opt_journal_async_commit
, "journal_async_commit"},
1157 {Opt_abort
, "abort"},
1158 {Opt_data_journal
, "data=journal"},
1159 {Opt_data_ordered
, "data=ordered"},
1160 {Opt_data_writeback
, "data=writeback"},
1161 {Opt_data_err_abort
, "data_err=abort"},
1162 {Opt_data_err_ignore
, "data_err=ignore"},
1163 {Opt_offusrjquota
, "usrjquota="},
1164 {Opt_usrjquota
, "usrjquota=%s"},
1165 {Opt_offgrpjquota
, "grpjquota="},
1166 {Opt_grpjquota
, "grpjquota=%s"},
1167 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1168 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1169 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
1170 {Opt_grpquota
, "grpquota"},
1171 {Opt_noquota
, "noquota"},
1172 {Opt_quota
, "quota"},
1173 {Opt_usrquota
, "usrquota"},
1174 {Opt_barrier
, "barrier=%u"},
1175 {Opt_barrier
, "barrier"},
1176 {Opt_nobarrier
, "nobarrier"},
1177 {Opt_i_version
, "i_version"},
1178 {Opt_stripe
, "stripe=%u"},
1179 {Opt_delalloc
, "delalloc"},
1180 {Opt_nodelalloc
, "nodelalloc"},
1181 {Opt_removed
, "mblk_io_submit"},
1182 {Opt_removed
, "nomblk_io_submit"},
1183 {Opt_block_validity
, "block_validity"},
1184 {Opt_noblock_validity
, "noblock_validity"},
1185 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1186 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1187 {Opt_auto_da_alloc
, "auto_da_alloc=%u"},
1188 {Opt_auto_da_alloc
, "auto_da_alloc"},
1189 {Opt_noauto_da_alloc
, "noauto_da_alloc"},
1190 {Opt_dioread_nolock
, "dioread_nolock"},
1191 {Opt_dioread_lock
, "dioread_lock"},
1192 {Opt_discard
, "discard"},
1193 {Opt_nodiscard
, "nodiscard"},
1194 {Opt_init_itable
, "init_itable=%u"},
1195 {Opt_init_itable
, "init_itable"},
1196 {Opt_noinit_itable
, "noinit_itable"},
1197 {Opt_max_dir_size_kb
, "max_dir_size_kb=%u"},
1198 {Opt_removed
, "check=none"}, /* mount option from ext2/3 */
1199 {Opt_removed
, "nocheck"}, /* mount option from ext2/3 */
1200 {Opt_removed
, "reservation"}, /* mount option from ext2/3 */
1201 {Opt_removed
, "noreservation"}, /* mount option from ext2/3 */
1202 {Opt_removed
, "journal=%u"}, /* mount option from ext2/3 */
1206 static ext4_fsblk_t
get_sb_block(void **data
)
1208 ext4_fsblk_t sb_block
;
1209 char *options
= (char *) *data
;
1211 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1212 return 1; /* Default location */
1215 /* TODO: use simple_strtoll with >32bit ext4 */
1216 sb_block
= simple_strtoul(options
, &options
, 0);
1217 if (*options
&& *options
!= ',') {
1218 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1222 if (*options
== ',')
1224 *data
= (void *) options
;
1229 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1230 static char deprecated_msg
[] = "Mount option \"%s\" will be removed by %s\n"
1231 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1234 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
1236 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1240 if (sb_any_quota_loaded(sb
) &&
1241 !sbi
->s_qf_names
[qtype
]) {
1242 ext4_msg(sb
, KERN_ERR
,
1243 "Cannot change journaled "
1244 "quota options when quota turned on");
1247 qname
= match_strdup(args
);
1249 ext4_msg(sb
, KERN_ERR
,
1250 "Not enough memory for storing quotafile name");
1253 if (sbi
->s_qf_names
[qtype
]) {
1254 if (strcmp(sbi
->s_qf_names
[qtype
], qname
) == 0)
1257 ext4_msg(sb
, KERN_ERR
,
1258 "%s quota file already specified",
1262 if (strchr(qname
, '/')) {
1263 ext4_msg(sb
, KERN_ERR
,
1264 "quotafile must be on filesystem root");
1267 sbi
->s_qf_names
[qtype
] = qname
;
1275 static int clear_qf_name(struct super_block
*sb
, int qtype
)
1278 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1280 if (sb_any_quota_loaded(sb
) &&
1281 sbi
->s_qf_names
[qtype
]) {
1282 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
1283 " when quota turned on");
1286 kfree(sbi
->s_qf_names
[qtype
]);
1287 sbi
->s_qf_names
[qtype
] = NULL
;
1292 #define MOPT_SET 0x0001
1293 #define MOPT_CLEAR 0x0002
1294 #define MOPT_NOSUPPORT 0x0004
1295 #define MOPT_EXPLICIT 0x0008
1296 #define MOPT_CLEAR_ERR 0x0010
1297 #define MOPT_GTE0 0x0020
1300 #define MOPT_QFMT 0x0040
1302 #define MOPT_Q MOPT_NOSUPPORT
1303 #define MOPT_QFMT MOPT_NOSUPPORT
1305 #define MOPT_DATAJ 0x0080
1306 #define MOPT_NO_EXT2 0x0100
1307 #define MOPT_NO_EXT3 0x0200
1308 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1310 static const struct mount_opts
{
1314 } ext4_mount_opts
[] = {
1315 {Opt_minix_df
, EXT4_MOUNT_MINIX_DF
, MOPT_SET
},
1316 {Opt_bsd_df
, EXT4_MOUNT_MINIX_DF
, MOPT_CLEAR
},
1317 {Opt_grpid
, EXT4_MOUNT_GRPID
, MOPT_SET
},
1318 {Opt_nogrpid
, EXT4_MOUNT_GRPID
, MOPT_CLEAR
},
1319 {Opt_block_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_SET
},
1320 {Opt_noblock_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_CLEAR
},
1321 {Opt_dioread_nolock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1322 MOPT_EXT4_ONLY
| MOPT_SET
},
1323 {Opt_dioread_lock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1324 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1325 {Opt_discard
, EXT4_MOUNT_DISCARD
, MOPT_SET
},
1326 {Opt_nodiscard
, EXT4_MOUNT_DISCARD
, MOPT_CLEAR
},
1327 {Opt_delalloc
, EXT4_MOUNT_DELALLOC
,
1328 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1329 {Opt_nodelalloc
, EXT4_MOUNT_DELALLOC
,
1330 MOPT_EXT4_ONLY
| MOPT_CLEAR
| MOPT_EXPLICIT
},
1331 {Opt_journal_checksum
, EXT4_MOUNT_JOURNAL_CHECKSUM
,
1332 MOPT_EXT4_ONLY
| MOPT_SET
},
1333 {Opt_journal_async_commit
, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT
|
1334 EXT4_MOUNT_JOURNAL_CHECKSUM
),
1335 MOPT_EXT4_ONLY
| MOPT_SET
},
1336 {Opt_noload
, EXT4_MOUNT_NOLOAD
, MOPT_NO_EXT2
| MOPT_SET
},
1337 {Opt_err_panic
, EXT4_MOUNT_ERRORS_PANIC
, MOPT_SET
| MOPT_CLEAR_ERR
},
1338 {Opt_err_ro
, EXT4_MOUNT_ERRORS_RO
, MOPT_SET
| MOPT_CLEAR_ERR
},
1339 {Opt_err_cont
, EXT4_MOUNT_ERRORS_CONT
, MOPT_SET
| MOPT_CLEAR_ERR
},
1340 {Opt_data_err_abort
, EXT4_MOUNT_DATA_ERR_ABORT
,
1341 MOPT_NO_EXT2
| MOPT_SET
},
1342 {Opt_data_err_ignore
, EXT4_MOUNT_DATA_ERR_ABORT
,
1343 MOPT_NO_EXT2
| MOPT_CLEAR
},
1344 {Opt_barrier
, EXT4_MOUNT_BARRIER
, MOPT_SET
},
1345 {Opt_nobarrier
, EXT4_MOUNT_BARRIER
, MOPT_CLEAR
},
1346 {Opt_noauto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_SET
},
1347 {Opt_auto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_CLEAR
},
1348 {Opt_noinit_itable
, EXT4_MOUNT_INIT_INODE_TABLE
, MOPT_CLEAR
},
1349 {Opt_commit
, 0, MOPT_GTE0
},
1350 {Opt_max_batch_time
, 0, MOPT_GTE0
},
1351 {Opt_min_batch_time
, 0, MOPT_GTE0
},
1352 {Opt_inode_readahead_blks
, 0, MOPT_GTE0
},
1353 {Opt_init_itable
, 0, MOPT_GTE0
},
1354 {Opt_stripe
, 0, MOPT_GTE0
},
1355 {Opt_resuid
, 0, MOPT_GTE0
},
1356 {Opt_resgid
, 0, MOPT_GTE0
},
1357 {Opt_journal_dev
, 0, MOPT_GTE0
},
1358 {Opt_journal_ioprio
, 0, MOPT_GTE0
},
1359 {Opt_data_journal
, EXT4_MOUNT_JOURNAL_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1360 {Opt_data_ordered
, EXT4_MOUNT_ORDERED_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1361 {Opt_data_writeback
, EXT4_MOUNT_WRITEBACK_DATA
,
1362 MOPT_NO_EXT2
| MOPT_DATAJ
},
1363 {Opt_user_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_SET
},
1364 {Opt_nouser_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_CLEAR
},
1365 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1366 {Opt_acl
, EXT4_MOUNT_POSIX_ACL
, MOPT_SET
},
1367 {Opt_noacl
, EXT4_MOUNT_POSIX_ACL
, MOPT_CLEAR
},
1369 {Opt_acl
, 0, MOPT_NOSUPPORT
},
1370 {Opt_noacl
, 0, MOPT_NOSUPPORT
},
1372 {Opt_nouid32
, EXT4_MOUNT_NO_UID32
, MOPT_SET
},
1373 {Opt_debug
, EXT4_MOUNT_DEBUG
, MOPT_SET
},
1374 {Opt_quota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
, MOPT_SET
| MOPT_Q
},
1375 {Opt_usrquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
,
1377 {Opt_grpquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_GRPQUOTA
,
1379 {Opt_noquota
, (EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
|
1380 EXT4_MOUNT_GRPQUOTA
), MOPT_CLEAR
| MOPT_Q
},
1381 {Opt_usrjquota
, 0, MOPT_Q
},
1382 {Opt_grpjquota
, 0, MOPT_Q
},
1383 {Opt_offusrjquota
, 0, MOPT_Q
},
1384 {Opt_offgrpjquota
, 0, MOPT_Q
},
1385 {Opt_jqfmt_vfsold
, QFMT_VFS_OLD
, MOPT_QFMT
},
1386 {Opt_jqfmt_vfsv0
, QFMT_VFS_V0
, MOPT_QFMT
},
1387 {Opt_jqfmt_vfsv1
, QFMT_VFS_V1
, MOPT_QFMT
},
1388 {Opt_max_dir_size_kb
, 0, MOPT_GTE0
},
1392 static int handle_mount_opt(struct super_block
*sb
, char *opt
, int token
,
1393 substring_t
*args
, unsigned long *journal_devnum
,
1394 unsigned int *journal_ioprio
, int is_remount
)
1396 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1397 const struct mount_opts
*m
;
1403 if (token
== Opt_usrjquota
)
1404 return set_qf_name(sb
, USRQUOTA
, &args
[0]);
1405 else if (token
== Opt_grpjquota
)
1406 return set_qf_name(sb
, GRPQUOTA
, &args
[0]);
1407 else if (token
== Opt_offusrjquota
)
1408 return clear_qf_name(sb
, USRQUOTA
);
1409 else if (token
== Opt_offgrpjquota
)
1410 return clear_qf_name(sb
, GRPQUOTA
);
1414 case Opt_nouser_xattr
:
1415 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, opt
, "3.5");
1418 return 1; /* handled by get_sb_block() */
1420 ext4_msg(sb
, KERN_WARNING
, "Ignoring removed %s option", opt
);
1423 sbi
->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
1426 sb
->s_flags
|= MS_I_VERSION
;
1430 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++)
1431 if (token
== m
->token
)
1434 if (m
->token
== Opt_err
) {
1435 ext4_msg(sb
, KERN_ERR
, "Unrecognized mount option \"%s\" "
1436 "or missing value", opt
);
1440 if ((m
->flags
& MOPT_NO_EXT2
) && IS_EXT2_SB(sb
)) {
1441 ext4_msg(sb
, KERN_ERR
,
1442 "Mount option \"%s\" incompatible with ext2", opt
);
1445 if ((m
->flags
& MOPT_NO_EXT3
) && IS_EXT3_SB(sb
)) {
1446 ext4_msg(sb
, KERN_ERR
,
1447 "Mount option \"%s\" incompatible with ext3", opt
);
1451 if (args
->from
&& match_int(args
, &arg
))
1453 if (args
->from
&& (m
->flags
& MOPT_GTE0
) && (arg
< 0))
1455 if (m
->flags
& MOPT_EXPLICIT
)
1456 set_opt2(sb
, EXPLICIT_DELALLOC
);
1457 if (m
->flags
& MOPT_CLEAR_ERR
)
1458 clear_opt(sb
, ERRORS_MASK
);
1459 if (token
== Opt_noquota
&& sb_any_quota_loaded(sb
)) {
1460 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
1461 "options when quota turned on");
1465 if (m
->flags
& MOPT_NOSUPPORT
) {
1466 ext4_msg(sb
, KERN_ERR
, "%s option not supported", opt
);
1467 } else if (token
== Opt_commit
) {
1469 arg
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1470 sbi
->s_commit_interval
= HZ
* arg
;
1471 } else if (token
== Opt_max_batch_time
) {
1473 arg
= EXT4_DEF_MAX_BATCH_TIME
;
1474 sbi
->s_max_batch_time
= arg
;
1475 } else if (token
== Opt_min_batch_time
) {
1476 sbi
->s_min_batch_time
= arg
;
1477 } else if (token
== Opt_inode_readahead_blks
) {
1478 if (arg
&& (arg
> (1 << 30) || !is_power_of_2(arg
))) {
1479 ext4_msg(sb
, KERN_ERR
,
1480 "EXT4-fs: inode_readahead_blks must be "
1481 "0 or a power of 2 smaller than 2^31");
1484 sbi
->s_inode_readahead_blks
= arg
;
1485 } else if (token
== Opt_init_itable
) {
1486 set_opt(sb
, INIT_INODE_TABLE
);
1488 arg
= EXT4_DEF_LI_WAIT_MULT
;
1489 sbi
->s_li_wait_mult
= arg
;
1490 } else if (token
== Opt_max_dir_size_kb
) {
1491 sbi
->s_max_dir_size_kb
= arg
;
1492 } else if (token
== Opt_stripe
) {
1493 sbi
->s_stripe
= arg
;
1494 } else if (token
== Opt_resuid
) {
1495 uid
= make_kuid(current_user_ns(), arg
);
1496 if (!uid_valid(uid
)) {
1497 ext4_msg(sb
, KERN_ERR
, "Invalid uid value %d", arg
);
1500 sbi
->s_resuid
= uid
;
1501 } else if (token
== Opt_resgid
) {
1502 gid
= make_kgid(current_user_ns(), arg
);
1503 if (!gid_valid(gid
)) {
1504 ext4_msg(sb
, KERN_ERR
, "Invalid gid value %d", arg
);
1507 sbi
->s_resgid
= gid
;
1508 } else if (token
== Opt_journal_dev
) {
1510 ext4_msg(sb
, KERN_ERR
,
1511 "Cannot specify journal on remount");
1514 *journal_devnum
= arg
;
1515 } else if (token
== Opt_journal_ioprio
) {
1517 ext4_msg(sb
, KERN_ERR
, "Invalid journal IO priority"
1522 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
, arg
);
1523 } else if (m
->flags
& MOPT_DATAJ
) {
1525 if (!sbi
->s_journal
)
1526 ext4_msg(sb
, KERN_WARNING
, "Remounting file system with no journal so ignoring journalled data option");
1527 else if (test_opt(sb
, DATA_FLAGS
) != m
->mount_opt
) {
1528 ext4_msg(sb
, KERN_ERR
,
1529 "Cannot change data mode on remount");
1533 clear_opt(sb
, DATA_FLAGS
);
1534 sbi
->s_mount_opt
|= m
->mount_opt
;
1537 } else if (m
->flags
& MOPT_QFMT
) {
1538 if (sb_any_quota_loaded(sb
) &&
1539 sbi
->s_jquota_fmt
!= m
->mount_opt
) {
1540 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled "
1541 "quota options when quota turned on");
1544 sbi
->s_jquota_fmt
= m
->mount_opt
;
1549 if (m
->flags
& MOPT_CLEAR
)
1551 else if (unlikely(!(m
->flags
& MOPT_SET
))) {
1552 ext4_msg(sb
, KERN_WARNING
,
1553 "buggy handling of option %s", opt
);
1558 sbi
->s_mount_opt
|= m
->mount_opt
;
1560 sbi
->s_mount_opt
&= ~m
->mount_opt
;
1565 static int parse_options(char *options
, struct super_block
*sb
,
1566 unsigned long *journal_devnum
,
1567 unsigned int *journal_ioprio
,
1570 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1572 substring_t args
[MAX_OPT_ARGS
];
1578 while ((p
= strsep(&options
, ",")) != NULL
) {
1582 * Initialize args struct so we know whether arg was
1583 * found; some options take optional arguments.
1585 args
[0].to
= args
[0].from
= NULL
;
1586 token
= match_token(p
, tokens
, args
);
1587 if (handle_mount_opt(sb
, p
, token
, args
, journal_devnum
,
1588 journal_ioprio
, is_remount
) < 0)
1592 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1593 if (test_opt(sb
, USRQUOTA
) && sbi
->s_qf_names
[USRQUOTA
])
1594 clear_opt(sb
, USRQUOTA
);
1596 if (test_opt(sb
, GRPQUOTA
) && sbi
->s_qf_names
[GRPQUOTA
])
1597 clear_opt(sb
, GRPQUOTA
);
1599 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
1600 ext4_msg(sb
, KERN_ERR
, "old and new quota "
1605 if (!sbi
->s_jquota_fmt
) {
1606 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1611 if (sbi
->s_jquota_fmt
) {
1612 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1613 "specified with no journaling "
1619 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
1621 BLOCK_SIZE
<< le32_to_cpu(sbi
->s_es
->s_log_block_size
);
1623 if (blocksize
< PAGE_CACHE_SIZE
) {
1624 ext4_msg(sb
, KERN_ERR
, "can't mount with "
1625 "dioread_nolock if block size != PAGE_SIZE");
1632 static inline void ext4_show_quota_options(struct seq_file
*seq
,
1633 struct super_block
*sb
)
1635 #if defined(CONFIG_QUOTA)
1636 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1638 if (sbi
->s_jquota_fmt
) {
1641 switch (sbi
->s_jquota_fmt
) {
1652 seq_printf(seq
, ",jqfmt=%s", fmtname
);
1655 if (sbi
->s_qf_names
[USRQUOTA
])
1656 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
1658 if (sbi
->s_qf_names
[GRPQUOTA
])
1659 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
1661 if (test_opt(sb
, USRQUOTA
))
1662 seq_puts(seq
, ",usrquota");
1664 if (test_opt(sb
, GRPQUOTA
))
1665 seq_puts(seq
, ",grpquota");
1669 static const char *token2str(int token
)
1671 const struct match_token
*t
;
1673 for (t
= tokens
; t
->token
!= Opt_err
; t
++)
1674 if (t
->token
== token
&& !strchr(t
->pattern
, '='))
1681 * - it's set to a non-default value OR
1682 * - if the per-sb default is different from the global default
1684 static int _ext4_show_options(struct seq_file
*seq
, struct super_block
*sb
,
1687 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1688 struct ext4_super_block
*es
= sbi
->s_es
;
1689 int def_errors
, def_mount_opt
= nodefs
? 0 : sbi
->s_def_mount_opt
;
1690 const struct mount_opts
*m
;
1691 char sep
= nodefs
? '\n' : ',';
1693 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1694 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1696 if (sbi
->s_sb_block
!= 1)
1697 SEQ_OPTS_PRINT("sb=%llu", sbi
->s_sb_block
);
1699 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++) {
1700 int want_set
= m
->flags
& MOPT_SET
;
1701 if (((m
->flags
& (MOPT_SET
|MOPT_CLEAR
)) == 0) ||
1702 (m
->flags
& MOPT_CLEAR_ERR
))
1704 if (!(m
->mount_opt
& (sbi
->s_mount_opt
^ def_mount_opt
)))
1705 continue; /* skip if same as the default */
1707 (sbi
->s_mount_opt
& m
->mount_opt
) != m
->mount_opt
) ||
1708 (!want_set
&& (sbi
->s_mount_opt
& m
->mount_opt
)))
1709 continue; /* select Opt_noFoo vs Opt_Foo */
1710 SEQ_OPTS_PRINT("%s", token2str(m
->token
));
1713 if (nodefs
|| !uid_eq(sbi
->s_resuid
, make_kuid(&init_user_ns
, EXT4_DEF_RESUID
)) ||
1714 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
)
1715 SEQ_OPTS_PRINT("resuid=%u",
1716 from_kuid_munged(&init_user_ns
, sbi
->s_resuid
));
1717 if (nodefs
|| !gid_eq(sbi
->s_resgid
, make_kgid(&init_user_ns
, EXT4_DEF_RESGID
)) ||
1718 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
)
1719 SEQ_OPTS_PRINT("resgid=%u",
1720 from_kgid_munged(&init_user_ns
, sbi
->s_resgid
));
1721 def_errors
= nodefs
? -1 : le16_to_cpu(es
->s_errors
);
1722 if (test_opt(sb
, ERRORS_RO
) && def_errors
!= EXT4_ERRORS_RO
)
1723 SEQ_OPTS_PUTS("errors=remount-ro");
1724 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
1725 SEQ_OPTS_PUTS("errors=continue");
1726 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
1727 SEQ_OPTS_PUTS("errors=panic");
1728 if (nodefs
|| sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
)
1729 SEQ_OPTS_PRINT("commit=%lu", sbi
->s_commit_interval
/ HZ
);
1730 if (nodefs
|| sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
)
1731 SEQ_OPTS_PRINT("min_batch_time=%u", sbi
->s_min_batch_time
);
1732 if (nodefs
|| sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
)
1733 SEQ_OPTS_PRINT("max_batch_time=%u", sbi
->s_max_batch_time
);
1734 if (sb
->s_flags
& MS_I_VERSION
)
1735 SEQ_OPTS_PUTS("i_version");
1736 if (nodefs
|| sbi
->s_stripe
)
1737 SEQ_OPTS_PRINT("stripe=%lu", sbi
->s_stripe
);
1738 if (EXT4_MOUNT_DATA_FLAGS
& (sbi
->s_mount_opt
^ def_mount_opt
)) {
1739 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
1740 SEQ_OPTS_PUTS("data=journal");
1741 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
1742 SEQ_OPTS_PUTS("data=ordered");
1743 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
1744 SEQ_OPTS_PUTS("data=writeback");
1747 sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
1748 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1749 sbi
->s_inode_readahead_blks
);
1751 if (nodefs
|| (test_opt(sb
, INIT_INODE_TABLE
) &&
1752 (sbi
->s_li_wait_mult
!= EXT4_DEF_LI_WAIT_MULT
)))
1753 SEQ_OPTS_PRINT("init_itable=%u", sbi
->s_li_wait_mult
);
1754 if (nodefs
|| sbi
->s_max_dir_size_kb
)
1755 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi
->s_max_dir_size_kb
);
1757 ext4_show_quota_options(seq
, sb
);
1761 static int ext4_show_options(struct seq_file
*seq
, struct dentry
*root
)
1763 return _ext4_show_options(seq
, root
->d_sb
, 0);
1766 static int options_seq_show(struct seq_file
*seq
, void *offset
)
1768 struct super_block
*sb
= seq
->private;
1771 seq_puts(seq
, (sb
->s_flags
& MS_RDONLY
) ? "ro" : "rw");
1772 rc
= _ext4_show_options(seq
, sb
, 1);
1773 seq_puts(seq
, "\n");
1777 static int options_open_fs(struct inode
*inode
, struct file
*file
)
1779 return single_open(file
, options_seq_show
, PDE(inode
)->data
);
1782 static const struct file_operations ext4_seq_options_fops
= {
1783 .owner
= THIS_MODULE
,
1784 .open
= options_open_fs
,
1786 .llseek
= seq_lseek
,
1787 .release
= single_release
,
1790 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
1793 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1796 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
1797 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
1798 "forcing read-only mode");
1803 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
1804 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
1805 "running e2fsck is recommended");
1806 else if ((sbi
->s_mount_state
& EXT4_ERROR_FS
))
1807 ext4_msg(sb
, KERN_WARNING
,
1808 "warning: mounting fs with errors, "
1809 "running e2fsck is recommended");
1810 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) > 0 &&
1811 le16_to_cpu(es
->s_mnt_count
) >=
1812 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1813 ext4_msg(sb
, KERN_WARNING
,
1814 "warning: maximal mount count reached, "
1815 "running e2fsck is recommended");
1816 else if (le32_to_cpu(es
->s_checkinterval
) &&
1817 (le32_to_cpu(es
->s_lastcheck
) +
1818 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1819 ext4_msg(sb
, KERN_WARNING
,
1820 "warning: checktime reached, "
1821 "running e2fsck is recommended");
1822 if (!sbi
->s_journal
)
1823 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
1824 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1825 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
1826 le16_add_cpu(&es
->s_mnt_count
, 1);
1827 es
->s_mtime
= cpu_to_le32(get_seconds());
1828 ext4_update_dynamic_rev(sb
);
1830 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
1832 ext4_commit_super(sb
, 1);
1834 if (test_opt(sb
, DEBUG
))
1835 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
1836 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1838 sbi
->s_groups_count
,
1839 EXT4_BLOCKS_PER_GROUP(sb
),
1840 EXT4_INODES_PER_GROUP(sb
),
1841 sbi
->s_mount_opt
, sbi
->s_mount_opt2
);
1843 cleancache_init_fs(sb
);
1847 int ext4_alloc_flex_bg_array(struct super_block
*sb
, ext4_group_t ngroup
)
1849 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1850 struct flex_groups
*new_groups
;
1853 if (!sbi
->s_log_groups_per_flex
)
1856 size
= ext4_flex_group(sbi
, ngroup
- 1) + 1;
1857 if (size
<= sbi
->s_flex_groups_allocated
)
1860 size
= roundup_pow_of_two(size
* sizeof(struct flex_groups
));
1861 new_groups
= ext4_kvzalloc(size
, GFP_KERNEL
);
1863 ext4_msg(sb
, KERN_ERR
, "not enough memory for %d flex groups",
1864 size
/ (int) sizeof(struct flex_groups
));
1868 if (sbi
->s_flex_groups
) {
1869 memcpy(new_groups
, sbi
->s_flex_groups
,
1870 (sbi
->s_flex_groups_allocated
*
1871 sizeof(struct flex_groups
)));
1872 ext4_kvfree(sbi
->s_flex_groups
);
1874 sbi
->s_flex_groups
= new_groups
;
1875 sbi
->s_flex_groups_allocated
= size
/ sizeof(struct flex_groups
);
1879 static int ext4_fill_flex_info(struct super_block
*sb
)
1881 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1882 struct ext4_group_desc
*gdp
= NULL
;
1883 ext4_group_t flex_group
;
1884 unsigned int groups_per_flex
= 0;
1887 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
1888 if (sbi
->s_log_groups_per_flex
< 1 || sbi
->s_log_groups_per_flex
> 31) {
1889 sbi
->s_log_groups_per_flex
= 0;
1892 groups_per_flex
= 1U << sbi
->s_log_groups_per_flex
;
1894 err
= ext4_alloc_flex_bg_array(sb
, sbi
->s_groups_count
);
1898 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1899 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1901 flex_group
= ext4_flex_group(sbi
, i
);
1902 atomic_add(ext4_free_inodes_count(sb
, gdp
),
1903 &sbi
->s_flex_groups
[flex_group
].free_inodes
);
1904 atomic_add(ext4_free_group_clusters(sb
, gdp
),
1905 &sbi
->s_flex_groups
[flex_group
].free_clusters
);
1906 atomic_add(ext4_used_dirs_count(sb
, gdp
),
1907 &sbi
->s_flex_groups
[flex_group
].used_dirs
);
1915 static __le16
ext4_group_desc_csum(struct ext4_sb_info
*sbi
, __u32 block_group
,
1916 struct ext4_group_desc
*gdp
)
1920 __le32 le_group
= cpu_to_le32(block_group
);
1922 if ((sbi
->s_es
->s_feature_ro_compat
&
1923 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
))) {
1924 /* Use new metadata_csum algorithm */
1928 old_csum
= gdp
->bg_checksum
;
1929 gdp
->bg_checksum
= 0;
1930 csum32
= ext4_chksum(sbi
, sbi
->s_csum_seed
, (__u8
*)&le_group
,
1932 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)gdp
,
1934 gdp
->bg_checksum
= old_csum
;
1936 crc
= csum32
& 0xFFFF;
1940 /* old crc16 code */
1941 offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
1943 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
1944 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
1945 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
1946 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
1947 /* for checksum of struct ext4_group_desc do the rest...*/
1948 if ((sbi
->s_es
->s_feature_incompat
&
1949 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT
)) &&
1950 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
1951 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
1952 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
1956 return cpu_to_le16(crc
);
1959 int ext4_group_desc_csum_verify(struct super_block
*sb
, __u32 block_group
,
1960 struct ext4_group_desc
*gdp
)
1962 if (ext4_has_group_desc_csum(sb
) &&
1963 (gdp
->bg_checksum
!= ext4_group_desc_csum(EXT4_SB(sb
),
1970 void ext4_group_desc_csum_set(struct super_block
*sb
, __u32 block_group
,
1971 struct ext4_group_desc
*gdp
)
1973 if (!ext4_has_group_desc_csum(sb
))
1975 gdp
->bg_checksum
= ext4_group_desc_csum(EXT4_SB(sb
), block_group
, gdp
);
1978 /* Called at mount-time, super-block is locked */
1979 static int ext4_check_descriptors(struct super_block
*sb
,
1980 ext4_group_t
*first_not_zeroed
)
1982 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1983 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
1984 ext4_fsblk_t last_block
;
1985 ext4_fsblk_t block_bitmap
;
1986 ext4_fsblk_t inode_bitmap
;
1987 ext4_fsblk_t inode_table
;
1988 int flexbg_flag
= 0;
1989 ext4_group_t i
, grp
= sbi
->s_groups_count
;
1991 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
1994 ext4_debug("Checking group descriptors");
1996 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1997 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1999 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
2000 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
2002 last_block
= first_block
+
2003 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2005 if ((grp
== sbi
->s_groups_count
) &&
2006 !(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2009 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
2010 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
2011 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2012 "Block bitmap for group %u not in group "
2013 "(block %llu)!", i
, block_bitmap
);
2016 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
2017 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
2018 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2019 "Inode bitmap for group %u not in group "
2020 "(block %llu)!", i
, inode_bitmap
);
2023 inode_table
= ext4_inode_table(sb
, gdp
);
2024 if (inode_table
< first_block
||
2025 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
2026 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2027 "Inode table for group %u not in group "
2028 "(block %llu)!", i
, inode_table
);
2031 ext4_lock_group(sb
, i
);
2032 if (!ext4_group_desc_csum_verify(sb
, i
, gdp
)) {
2033 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2034 "Checksum for group %u failed (%u!=%u)",
2035 i
, le16_to_cpu(ext4_group_desc_csum(sbi
, i
,
2036 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
2037 if (!(sb
->s_flags
& MS_RDONLY
)) {
2038 ext4_unlock_group(sb
, i
);
2042 ext4_unlock_group(sb
, i
);
2044 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
2046 if (NULL
!= first_not_zeroed
)
2047 *first_not_zeroed
= grp
;
2049 ext4_free_blocks_count_set(sbi
->s_es
,
2050 EXT4_C2B(sbi
, ext4_count_free_clusters(sb
)));
2051 sbi
->s_es
->s_free_inodes_count
=cpu_to_le32(ext4_count_free_inodes(sb
));
2055 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2056 * the superblock) which were deleted from all directories, but held open by
2057 * a process at the time of a crash. We walk the list and try to delete these
2058 * inodes at recovery time (only with a read-write filesystem).
2060 * In order to keep the orphan inode chain consistent during traversal (in
2061 * case of crash during recovery), we link each inode into the superblock
2062 * orphan list_head and handle it the same way as an inode deletion during
2063 * normal operation (which journals the operations for us).
2065 * We only do an iget() and an iput() on each inode, which is very safe if we
2066 * accidentally point at an in-use or already deleted inode. The worst that
2067 * can happen in this case is that we get a "bit already cleared" message from
2068 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2069 * e2fsck was run on this filesystem, and it must have already done the orphan
2070 * inode cleanup for us, so we can safely abort without any further action.
2072 static void ext4_orphan_cleanup(struct super_block
*sb
,
2073 struct ext4_super_block
*es
)
2075 unsigned int s_flags
= sb
->s_flags
;
2076 int nr_orphans
= 0, nr_truncates
= 0;
2080 if (!es
->s_last_orphan
) {
2081 jbd_debug(4, "no orphan inodes to clean up\n");
2085 if (bdev_read_only(sb
->s_bdev
)) {
2086 ext4_msg(sb
, KERN_ERR
, "write access "
2087 "unavailable, skipping orphan cleanup");
2091 /* Check if feature set would not allow a r/w mount */
2092 if (!ext4_feature_set_ok(sb
, 0)) {
2093 ext4_msg(sb
, KERN_INFO
, "Skipping orphan cleanup due to "
2094 "unknown ROCOMPAT features");
2098 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2099 /* don't clear list on RO mount w/ errors */
2100 if (es
->s_last_orphan
&& !(s_flags
& MS_RDONLY
)) {
2101 jbd_debug(1, "Errors on filesystem, "
2102 "clearing orphan list.\n");
2103 es
->s_last_orphan
= 0;
2105 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2109 if (s_flags
& MS_RDONLY
) {
2110 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
2111 sb
->s_flags
&= ~MS_RDONLY
;
2114 /* Needed for iput() to work correctly and not trash data */
2115 sb
->s_flags
|= MS_ACTIVE
;
2116 /* Turn on quotas so that they are updated correctly */
2117 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2118 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
2119 int ret
= ext4_quota_on_mount(sb
, i
);
2121 ext4_msg(sb
, KERN_ERR
,
2122 "Cannot turn on journaled "
2123 "quota: error %d", ret
);
2128 while (es
->s_last_orphan
) {
2129 struct inode
*inode
;
2131 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
2132 if (IS_ERR(inode
)) {
2133 es
->s_last_orphan
= 0;
2137 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
2138 dquot_initialize(inode
);
2139 if (inode
->i_nlink
) {
2140 ext4_msg(sb
, KERN_DEBUG
,
2141 "%s: truncating inode %lu to %lld bytes",
2142 __func__
, inode
->i_ino
, inode
->i_size
);
2143 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2144 inode
->i_ino
, inode
->i_size
);
2145 mutex_lock(&inode
->i_mutex
);
2146 ext4_truncate(inode
);
2147 mutex_unlock(&inode
->i_mutex
);
2150 ext4_msg(sb
, KERN_DEBUG
,
2151 "%s: deleting unreferenced inode %lu",
2152 __func__
, inode
->i_ino
);
2153 jbd_debug(2, "deleting unreferenced inode %lu\n",
2157 iput(inode
); /* The delete magic happens here! */
2160 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2163 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
2164 PLURAL(nr_orphans
));
2166 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
2167 PLURAL(nr_truncates
));
2169 /* Turn quotas off */
2170 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2171 if (sb_dqopt(sb
)->files
[i
])
2172 dquot_quota_off(sb
, i
);
2175 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
2179 * Maximal extent format file size.
2180 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2181 * extent format containers, within a sector_t, and within i_blocks
2182 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2183 * so that won't be a limiting factor.
2185 * However there is other limiting factor. We do store extents in the form
2186 * of starting block and length, hence the resulting length of the extent
2187 * covering maximum file size must fit into on-disk format containers as
2188 * well. Given that length is always by 1 unit bigger than max unit (because
2189 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2191 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2193 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
2196 loff_t upper_limit
= MAX_LFS_FILESIZE
;
2198 /* small i_blocks in vfs inode? */
2199 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2201 * CONFIG_LBDAF is not enabled implies the inode
2202 * i_block represent total blocks in 512 bytes
2203 * 32 == size of vfs inode i_blocks * 8
2205 upper_limit
= (1LL << 32) - 1;
2207 /* total blocks in file system block size */
2208 upper_limit
>>= (blkbits
- 9);
2209 upper_limit
<<= blkbits
;
2213 * 32-bit extent-start container, ee_block. We lower the maxbytes
2214 * by one fs block, so ee_len can cover the extent of maximum file
2217 res
= (1LL << 32) - 1;
2220 /* Sanity check against vm- & vfs- imposed limits */
2221 if (res
> upper_limit
)
2228 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2229 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2230 * We need to be 1 filesystem block less than the 2^48 sector limit.
2232 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
2234 loff_t res
= EXT4_NDIR_BLOCKS
;
2237 /* This is calculated to be the largest file size for a dense, block
2238 * mapped file such that the file's total number of 512-byte sectors,
2239 * including data and all indirect blocks, does not exceed (2^48 - 1).
2241 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2242 * number of 512-byte sectors of the file.
2245 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2247 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2248 * the inode i_block field represents total file blocks in
2249 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2251 upper_limit
= (1LL << 32) - 1;
2253 /* total blocks in file system block size */
2254 upper_limit
>>= (bits
- 9);
2258 * We use 48 bit ext4_inode i_blocks
2259 * With EXT4_HUGE_FILE_FL set the i_blocks
2260 * represent total number of blocks in
2261 * file system block size
2263 upper_limit
= (1LL << 48) - 1;
2267 /* indirect blocks */
2269 /* double indirect blocks */
2270 meta_blocks
+= 1 + (1LL << (bits
-2));
2271 /* tripple indirect blocks */
2272 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
2274 upper_limit
-= meta_blocks
;
2275 upper_limit
<<= bits
;
2277 res
+= 1LL << (bits
-2);
2278 res
+= 1LL << (2*(bits
-2));
2279 res
+= 1LL << (3*(bits
-2));
2281 if (res
> upper_limit
)
2284 if (res
> MAX_LFS_FILESIZE
)
2285 res
= MAX_LFS_FILESIZE
;
2290 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
2291 ext4_fsblk_t logical_sb_block
, int nr
)
2293 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2294 ext4_group_t bg
, first_meta_bg
;
2297 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
2299 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_META_BG
) ||
2301 return logical_sb_block
+ nr
+ 1;
2302 bg
= sbi
->s_desc_per_block
* nr
;
2303 if (ext4_bg_has_super(sb
, bg
))
2306 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
2310 * ext4_get_stripe_size: Get the stripe size.
2311 * @sbi: In memory super block info
2313 * If we have specified it via mount option, then
2314 * use the mount option value. If the value specified at mount time is
2315 * greater than the blocks per group use the super block value.
2316 * If the super block value is greater than blocks per group return 0.
2317 * Allocator needs it be less than blocks per group.
2320 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
2322 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
2323 unsigned long stripe_width
=
2324 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
2327 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
2328 ret
= sbi
->s_stripe
;
2329 else if (stripe_width
<= sbi
->s_blocks_per_group
)
2331 else if (stride
<= sbi
->s_blocks_per_group
)
2337 * If the stripe width is 1, this makes no sense and
2338 * we set it to 0 to turn off stripe handling code.
2349 struct attribute attr
;
2350 ssize_t (*show
)(struct ext4_attr
*, struct ext4_sb_info
*, char *);
2351 ssize_t (*store
)(struct ext4_attr
*, struct ext4_sb_info
*,
2352 const char *, size_t);
2356 static int parse_strtoul(const char *buf
,
2357 unsigned long max
, unsigned long *value
)
2361 *value
= simple_strtoul(skip_spaces(buf
), &endp
, 0);
2362 endp
= skip_spaces(endp
);
2363 if (*endp
|| *value
> max
)
2369 static ssize_t
delayed_allocation_blocks_show(struct ext4_attr
*a
,
2370 struct ext4_sb_info
*sbi
,
2373 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2375 percpu_counter_sum(&sbi
->s_dirtyclusters_counter
)));
2378 static ssize_t
session_write_kbytes_show(struct ext4_attr
*a
,
2379 struct ext4_sb_info
*sbi
, char *buf
)
2381 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2383 if (!sb
->s_bdev
->bd_part
)
2384 return snprintf(buf
, PAGE_SIZE
, "0\n");
2385 return snprintf(buf
, PAGE_SIZE
, "%lu\n",
2386 (part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2387 sbi
->s_sectors_written_start
) >> 1);
2390 static ssize_t
lifetime_write_kbytes_show(struct ext4_attr
*a
,
2391 struct ext4_sb_info
*sbi
, char *buf
)
2393 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2395 if (!sb
->s_bdev
->bd_part
)
2396 return snprintf(buf
, PAGE_SIZE
, "0\n");
2397 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2398 (unsigned long long)(sbi
->s_kbytes_written
+
2399 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2400 EXT4_SB(sb
)->s_sectors_written_start
) >> 1)));
2403 static ssize_t
inode_readahead_blks_store(struct ext4_attr
*a
,
2404 struct ext4_sb_info
*sbi
,
2405 const char *buf
, size_t count
)
2409 if (parse_strtoul(buf
, 0x40000000, &t
))
2412 if (t
&& !is_power_of_2(t
))
2415 sbi
->s_inode_readahead_blks
= t
;
2419 static ssize_t
sbi_ui_show(struct ext4_attr
*a
,
2420 struct ext4_sb_info
*sbi
, char *buf
)
2422 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2424 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
2427 static ssize_t
sbi_ui_store(struct ext4_attr
*a
,
2428 struct ext4_sb_info
*sbi
,
2429 const char *buf
, size_t count
)
2431 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2434 if (parse_strtoul(buf
, 0xffffffff, &t
))
2440 static ssize_t
trigger_test_error(struct ext4_attr
*a
,
2441 struct ext4_sb_info
*sbi
,
2442 const char *buf
, size_t count
)
2446 if (!capable(CAP_SYS_ADMIN
))
2449 if (len
&& buf
[len
-1] == '\n')
2453 ext4_error(sbi
->s_sb
, "%.*s", len
, buf
);
2457 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2458 static struct ext4_attr ext4_attr_##_name = { \
2459 .attr = {.name = __stringify(_name), .mode = _mode }, \
2462 .offset = offsetof(struct ext4_sb_info, _elname), \
2464 #define EXT4_ATTR(name, mode, show, store) \
2465 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2467 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2468 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2469 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2470 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2471 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2472 #define ATTR_LIST(name) &ext4_attr_##name.attr
2474 EXT4_RO_ATTR(delayed_allocation_blocks
);
2475 EXT4_RO_ATTR(session_write_kbytes
);
2476 EXT4_RO_ATTR(lifetime_write_kbytes
);
2477 EXT4_ATTR_OFFSET(inode_readahead_blks
, 0644, sbi_ui_show
,
2478 inode_readahead_blks_store
, s_inode_readahead_blks
);
2479 EXT4_RW_ATTR_SBI_UI(inode_goal
, s_inode_goal
);
2480 EXT4_RW_ATTR_SBI_UI(mb_stats
, s_mb_stats
);
2481 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan
, s_mb_max_to_scan
);
2482 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan
, s_mb_min_to_scan
);
2483 EXT4_RW_ATTR_SBI_UI(mb_order2_req
, s_mb_order2_reqs
);
2484 EXT4_RW_ATTR_SBI_UI(mb_stream_req
, s_mb_stream_request
);
2485 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc
, s_mb_group_prealloc
);
2486 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump
, s_max_writeback_mb_bump
);
2487 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb
, s_extent_max_zeroout_kb
);
2488 EXT4_ATTR(trigger_fs_error
, 0200, NULL
, trigger_test_error
);
2490 static struct attribute
*ext4_attrs
[] = {
2491 ATTR_LIST(delayed_allocation_blocks
),
2492 ATTR_LIST(session_write_kbytes
),
2493 ATTR_LIST(lifetime_write_kbytes
),
2494 ATTR_LIST(inode_readahead_blks
),
2495 ATTR_LIST(inode_goal
),
2496 ATTR_LIST(mb_stats
),
2497 ATTR_LIST(mb_max_to_scan
),
2498 ATTR_LIST(mb_min_to_scan
),
2499 ATTR_LIST(mb_order2_req
),
2500 ATTR_LIST(mb_stream_req
),
2501 ATTR_LIST(mb_group_prealloc
),
2502 ATTR_LIST(max_writeback_mb_bump
),
2503 ATTR_LIST(extent_max_zeroout_kb
),
2504 ATTR_LIST(trigger_fs_error
),
2508 /* Features this copy of ext4 supports */
2509 EXT4_INFO_ATTR(lazy_itable_init
);
2510 EXT4_INFO_ATTR(batched_discard
);
2511 EXT4_INFO_ATTR(meta_bg_resize
);
2513 static struct attribute
*ext4_feat_attrs
[] = {
2514 ATTR_LIST(lazy_itable_init
),
2515 ATTR_LIST(batched_discard
),
2516 ATTR_LIST(meta_bg_resize
),
2520 static ssize_t
ext4_attr_show(struct kobject
*kobj
,
2521 struct attribute
*attr
, char *buf
)
2523 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2525 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2527 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
2530 static ssize_t
ext4_attr_store(struct kobject
*kobj
,
2531 struct attribute
*attr
,
2532 const char *buf
, size_t len
)
2534 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2536 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2538 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
2541 static void ext4_sb_release(struct kobject
*kobj
)
2543 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2545 complete(&sbi
->s_kobj_unregister
);
2548 static const struct sysfs_ops ext4_attr_ops
= {
2549 .show
= ext4_attr_show
,
2550 .store
= ext4_attr_store
,
2553 static struct kobj_type ext4_ktype
= {
2554 .default_attrs
= ext4_attrs
,
2555 .sysfs_ops
= &ext4_attr_ops
,
2556 .release
= ext4_sb_release
,
2559 static void ext4_feat_release(struct kobject
*kobj
)
2561 complete(&ext4_feat
->f_kobj_unregister
);
2564 static struct kobj_type ext4_feat_ktype
= {
2565 .default_attrs
= ext4_feat_attrs
,
2566 .sysfs_ops
= &ext4_attr_ops
,
2567 .release
= ext4_feat_release
,
2571 * Check whether this filesystem can be mounted based on
2572 * the features present and the RDONLY/RDWR mount requested.
2573 * Returns 1 if this filesystem can be mounted as requested,
2574 * 0 if it cannot be.
2576 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
2578 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT4_FEATURE_INCOMPAT_SUPP
)) {
2579 ext4_msg(sb
, KERN_ERR
,
2580 "Couldn't mount because of "
2581 "unsupported optional features (%x)",
2582 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2583 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2590 /* Check that feature set is OK for a read-write mount */
2591 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT4_FEATURE_RO_COMPAT_SUPP
)) {
2592 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
2593 "unsupported optional features (%x)",
2594 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2595 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2599 * Large file size enabled file system can only be mounted
2600 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2602 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_HUGE_FILE
)) {
2603 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
2604 ext4_msg(sb
, KERN_ERR
, "Filesystem with huge files "
2605 "cannot be mounted RDWR without "
2610 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_BIGALLOC
) &&
2611 !EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_EXTENTS
)) {
2612 ext4_msg(sb
, KERN_ERR
,
2613 "Can't support bigalloc feature without "
2614 "extents feature\n");
2618 #ifndef CONFIG_QUOTA
2619 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
) &&
2621 ext4_msg(sb
, KERN_ERR
,
2622 "Filesystem with quota feature cannot be mounted RDWR "
2623 "without CONFIG_QUOTA");
2626 #endif /* CONFIG_QUOTA */
2631 * This function is called once a day if we have errors logged
2632 * on the file system
2634 static void print_daily_error_info(unsigned long arg
)
2636 struct super_block
*sb
= (struct super_block
*) arg
;
2637 struct ext4_sb_info
*sbi
;
2638 struct ext4_super_block
*es
;
2643 if (es
->s_error_count
)
2644 ext4_msg(sb
, KERN_NOTICE
, "error count: %u",
2645 le32_to_cpu(es
->s_error_count
));
2646 if (es
->s_first_error_time
) {
2647 printk(KERN_NOTICE
"EXT4-fs (%s): initial error at %u: %.*s:%d",
2648 sb
->s_id
, le32_to_cpu(es
->s_first_error_time
),
2649 (int) sizeof(es
->s_first_error_func
),
2650 es
->s_first_error_func
,
2651 le32_to_cpu(es
->s_first_error_line
));
2652 if (es
->s_first_error_ino
)
2653 printk(": inode %u",
2654 le32_to_cpu(es
->s_first_error_ino
));
2655 if (es
->s_first_error_block
)
2656 printk(": block %llu", (unsigned long long)
2657 le64_to_cpu(es
->s_first_error_block
));
2660 if (es
->s_last_error_time
) {
2661 printk(KERN_NOTICE
"EXT4-fs (%s): last error at %u: %.*s:%d",
2662 sb
->s_id
, le32_to_cpu(es
->s_last_error_time
),
2663 (int) sizeof(es
->s_last_error_func
),
2664 es
->s_last_error_func
,
2665 le32_to_cpu(es
->s_last_error_line
));
2666 if (es
->s_last_error_ino
)
2667 printk(": inode %u",
2668 le32_to_cpu(es
->s_last_error_ino
));
2669 if (es
->s_last_error_block
)
2670 printk(": block %llu", (unsigned long long)
2671 le64_to_cpu(es
->s_last_error_block
));
2674 mod_timer(&sbi
->s_err_report
, jiffies
+ 24*60*60*HZ
); /* Once a day */
2677 /* Find next suitable group and run ext4_init_inode_table */
2678 static int ext4_run_li_request(struct ext4_li_request
*elr
)
2680 struct ext4_group_desc
*gdp
= NULL
;
2681 ext4_group_t group
, ngroups
;
2682 struct super_block
*sb
;
2683 unsigned long timeout
= 0;
2687 ngroups
= EXT4_SB(sb
)->s_groups_count
;
2690 for (group
= elr
->lr_next_group
; group
< ngroups
; group
++) {
2691 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2697 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2701 if (group
>= ngroups
)
2706 ret
= ext4_init_inode_table(sb
, group
,
2707 elr
->lr_timeout
? 0 : 1);
2708 if (elr
->lr_timeout
== 0) {
2709 timeout
= (jiffies
- timeout
) *
2710 elr
->lr_sbi
->s_li_wait_mult
;
2711 elr
->lr_timeout
= timeout
;
2713 elr
->lr_next_sched
= jiffies
+ elr
->lr_timeout
;
2714 elr
->lr_next_group
= group
+ 1;
2722 * Remove lr_request from the list_request and free the
2723 * request structure. Should be called with li_list_mtx held
2725 static void ext4_remove_li_request(struct ext4_li_request
*elr
)
2727 struct ext4_sb_info
*sbi
;
2734 list_del(&elr
->lr_request
);
2735 sbi
->s_li_request
= NULL
;
2739 static void ext4_unregister_li_request(struct super_block
*sb
)
2741 mutex_lock(&ext4_li_mtx
);
2742 if (!ext4_li_info
) {
2743 mutex_unlock(&ext4_li_mtx
);
2747 mutex_lock(&ext4_li_info
->li_list_mtx
);
2748 ext4_remove_li_request(EXT4_SB(sb
)->s_li_request
);
2749 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2750 mutex_unlock(&ext4_li_mtx
);
2753 static struct task_struct
*ext4_lazyinit_task
;
2756 * This is the function where ext4lazyinit thread lives. It walks
2757 * through the request list searching for next scheduled filesystem.
2758 * When such a fs is found, run the lazy initialization request
2759 * (ext4_rn_li_request) and keep track of the time spend in this
2760 * function. Based on that time we compute next schedule time of
2761 * the request. When walking through the list is complete, compute
2762 * next waking time and put itself into sleep.
2764 static int ext4_lazyinit_thread(void *arg
)
2766 struct ext4_lazy_init
*eli
= (struct ext4_lazy_init
*)arg
;
2767 struct list_head
*pos
, *n
;
2768 struct ext4_li_request
*elr
;
2769 unsigned long next_wakeup
, cur
;
2771 BUG_ON(NULL
== eli
);
2775 next_wakeup
= MAX_JIFFY_OFFSET
;
2777 mutex_lock(&eli
->li_list_mtx
);
2778 if (list_empty(&eli
->li_request_list
)) {
2779 mutex_unlock(&eli
->li_list_mtx
);
2783 list_for_each_safe(pos
, n
, &eli
->li_request_list
) {
2784 elr
= list_entry(pos
, struct ext4_li_request
,
2787 if (time_after_eq(jiffies
, elr
->lr_next_sched
)) {
2788 if (ext4_run_li_request(elr
) != 0) {
2789 /* error, remove the lazy_init job */
2790 ext4_remove_li_request(elr
);
2795 if (time_before(elr
->lr_next_sched
, next_wakeup
))
2796 next_wakeup
= elr
->lr_next_sched
;
2798 mutex_unlock(&eli
->li_list_mtx
);
2803 if ((time_after_eq(cur
, next_wakeup
)) ||
2804 (MAX_JIFFY_OFFSET
== next_wakeup
)) {
2809 schedule_timeout_interruptible(next_wakeup
- cur
);
2811 if (kthread_should_stop()) {
2812 ext4_clear_request_list();
2819 * It looks like the request list is empty, but we need
2820 * to check it under the li_list_mtx lock, to prevent any
2821 * additions into it, and of course we should lock ext4_li_mtx
2822 * to atomically free the list and ext4_li_info, because at
2823 * this point another ext4 filesystem could be registering
2826 mutex_lock(&ext4_li_mtx
);
2827 mutex_lock(&eli
->li_list_mtx
);
2828 if (!list_empty(&eli
->li_request_list
)) {
2829 mutex_unlock(&eli
->li_list_mtx
);
2830 mutex_unlock(&ext4_li_mtx
);
2833 mutex_unlock(&eli
->li_list_mtx
);
2834 kfree(ext4_li_info
);
2835 ext4_li_info
= NULL
;
2836 mutex_unlock(&ext4_li_mtx
);
2841 static void ext4_clear_request_list(void)
2843 struct list_head
*pos
, *n
;
2844 struct ext4_li_request
*elr
;
2846 mutex_lock(&ext4_li_info
->li_list_mtx
);
2847 list_for_each_safe(pos
, n
, &ext4_li_info
->li_request_list
) {
2848 elr
= list_entry(pos
, struct ext4_li_request
,
2850 ext4_remove_li_request(elr
);
2852 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2855 static int ext4_run_lazyinit_thread(void)
2857 ext4_lazyinit_task
= kthread_run(ext4_lazyinit_thread
,
2858 ext4_li_info
, "ext4lazyinit");
2859 if (IS_ERR(ext4_lazyinit_task
)) {
2860 int err
= PTR_ERR(ext4_lazyinit_task
);
2861 ext4_clear_request_list();
2862 kfree(ext4_li_info
);
2863 ext4_li_info
= NULL
;
2864 printk(KERN_CRIT
"EXT4-fs: error %d creating inode table "
2865 "initialization thread\n",
2869 ext4_li_info
->li_state
|= EXT4_LAZYINIT_RUNNING
;
2874 * Check whether it make sense to run itable init. thread or not.
2875 * If there is at least one uninitialized inode table, return
2876 * corresponding group number, else the loop goes through all
2877 * groups and return total number of groups.
2879 static ext4_group_t
ext4_has_uninit_itable(struct super_block
*sb
)
2881 ext4_group_t group
, ngroups
= EXT4_SB(sb
)->s_groups_count
;
2882 struct ext4_group_desc
*gdp
= NULL
;
2884 for (group
= 0; group
< ngroups
; group
++) {
2885 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2889 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2896 static int ext4_li_info_new(void)
2898 struct ext4_lazy_init
*eli
= NULL
;
2900 eli
= kzalloc(sizeof(*eli
), GFP_KERNEL
);
2904 INIT_LIST_HEAD(&eli
->li_request_list
);
2905 mutex_init(&eli
->li_list_mtx
);
2907 eli
->li_state
|= EXT4_LAZYINIT_QUIT
;
2914 static struct ext4_li_request
*ext4_li_request_new(struct super_block
*sb
,
2917 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2918 struct ext4_li_request
*elr
;
2921 elr
= kzalloc(sizeof(*elr
), GFP_KERNEL
);
2927 elr
->lr_next_group
= start
;
2930 * Randomize first schedule time of the request to
2931 * spread the inode table initialization requests
2934 get_random_bytes(&rnd
, sizeof(rnd
));
2935 elr
->lr_next_sched
= jiffies
+ (unsigned long)rnd
%
2936 (EXT4_DEF_LI_MAX_START_DELAY
* HZ
);
2941 int ext4_register_li_request(struct super_block
*sb
,
2942 ext4_group_t first_not_zeroed
)
2944 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2945 struct ext4_li_request
*elr
= NULL
;
2946 ext4_group_t ngroups
= EXT4_SB(sb
)->s_groups_count
;
2949 mutex_lock(&ext4_li_mtx
);
2950 if (sbi
->s_li_request
!= NULL
) {
2952 * Reset timeout so it can be computed again, because
2953 * s_li_wait_mult might have changed.
2955 sbi
->s_li_request
->lr_timeout
= 0;
2959 if (first_not_zeroed
== ngroups
||
2960 (sb
->s_flags
& MS_RDONLY
) ||
2961 !test_opt(sb
, INIT_INODE_TABLE
))
2964 elr
= ext4_li_request_new(sb
, first_not_zeroed
);
2970 if (NULL
== ext4_li_info
) {
2971 ret
= ext4_li_info_new();
2976 mutex_lock(&ext4_li_info
->li_list_mtx
);
2977 list_add(&elr
->lr_request
, &ext4_li_info
->li_request_list
);
2978 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2980 sbi
->s_li_request
= elr
;
2982 * set elr to NULL here since it has been inserted to
2983 * the request_list and the removal and free of it is
2984 * handled by ext4_clear_request_list from now on.
2988 if (!(ext4_li_info
->li_state
& EXT4_LAZYINIT_RUNNING
)) {
2989 ret
= ext4_run_lazyinit_thread();
2994 mutex_unlock(&ext4_li_mtx
);
3001 * We do not need to lock anything since this is called on
3004 static void ext4_destroy_lazyinit_thread(void)
3007 * If thread exited earlier
3008 * there's nothing to be done.
3010 if (!ext4_li_info
|| !ext4_lazyinit_task
)
3013 kthread_stop(ext4_lazyinit_task
);
3016 static int set_journal_csum_feature_set(struct super_block
*sb
)
3019 int compat
, incompat
;
3020 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3022 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3023 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
)) {
3024 /* journal checksum v2 */
3026 incompat
= JBD2_FEATURE_INCOMPAT_CSUM_V2
;
3028 /* journal checksum v1 */
3029 compat
= JBD2_FEATURE_COMPAT_CHECKSUM
;
3033 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
3034 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3036 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
|
3038 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
3039 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3042 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3043 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3045 jbd2_journal_clear_features(sbi
->s_journal
,
3046 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3047 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
|
3048 JBD2_FEATURE_INCOMPAT_CSUM_V2
);
3055 * Note: calculating the overhead so we can be compatible with
3056 * historical BSD practice is quite difficult in the face of
3057 * clusters/bigalloc. This is because multiple metadata blocks from
3058 * different block group can end up in the same allocation cluster.
3059 * Calculating the exact overhead in the face of clustered allocation
3060 * requires either O(all block bitmaps) in memory or O(number of block
3061 * groups**2) in time. We will still calculate the superblock for
3062 * older file systems --- and if we come across with a bigalloc file
3063 * system with zero in s_overhead_clusters the estimate will be close to
3064 * correct especially for very large cluster sizes --- but for newer
3065 * file systems, it's better to calculate this figure once at mkfs
3066 * time, and store it in the superblock. If the superblock value is
3067 * present (even for non-bigalloc file systems), we will use it.
3069 static int count_overhead(struct super_block
*sb
, ext4_group_t grp
,
3072 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3073 struct ext4_group_desc
*gdp
;
3074 ext4_fsblk_t first_block
, last_block
, b
;
3075 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3076 int s
, j
, count
= 0;
3078 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_BIGALLOC
))
3079 return (ext4_bg_has_super(sb
, grp
) + ext4_bg_num_gdb(sb
, grp
) +
3080 sbi
->s_itb_per_group
+ 2);
3082 first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
) +
3083 (grp
* EXT4_BLOCKS_PER_GROUP(sb
));
3084 last_block
= first_block
+ EXT4_BLOCKS_PER_GROUP(sb
) - 1;
3085 for (i
= 0; i
< ngroups
; i
++) {
3086 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
3087 b
= ext4_block_bitmap(sb
, gdp
);
3088 if (b
>= first_block
&& b
<= last_block
) {
3089 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3092 b
= ext4_inode_bitmap(sb
, gdp
);
3093 if (b
>= first_block
&& b
<= last_block
) {
3094 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3097 b
= ext4_inode_table(sb
, gdp
);
3098 if (b
>= first_block
&& b
+ sbi
->s_itb_per_group
<= last_block
)
3099 for (j
= 0; j
< sbi
->s_itb_per_group
; j
++, b
++) {
3100 int c
= EXT4_B2C(sbi
, b
- first_block
);
3101 ext4_set_bit(c
, buf
);
3107 if (ext4_bg_has_super(sb
, grp
)) {
3108 ext4_set_bit(s
++, buf
);
3111 for (j
= ext4_bg_num_gdb(sb
, grp
); j
> 0; j
--) {
3112 ext4_set_bit(EXT4_B2C(sbi
, s
++), buf
);
3118 return EXT4_CLUSTERS_PER_GROUP(sb
) -
3119 ext4_count_free(buf
, EXT4_CLUSTERS_PER_GROUP(sb
) / 8);
3123 * Compute the overhead and stash it in sbi->s_overhead
3125 int ext4_calculate_overhead(struct super_block
*sb
)
3127 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3128 struct ext4_super_block
*es
= sbi
->s_es
;
3129 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3130 ext4_fsblk_t overhead
= 0;
3131 char *buf
= (char *) get_zeroed_page(GFP_KERNEL
);
3137 * Compute the overhead (FS structures). This is constant
3138 * for a given filesystem unless the number of block groups
3139 * changes so we cache the previous value until it does.
3143 * All of the blocks before first_data_block are overhead
3145 overhead
= EXT4_B2C(sbi
, le32_to_cpu(es
->s_first_data_block
));
3148 * Add the overhead found in each block group
3150 for (i
= 0; i
< ngroups
; i
++) {
3153 blks
= count_overhead(sb
, i
, buf
);
3156 memset(buf
, 0, PAGE_SIZE
);
3159 /* Add the journal blocks as well */
3161 overhead
+= EXT4_B2C(sbi
, sbi
->s_journal
->j_maxlen
);
3163 sbi
->s_overhead
= overhead
;
3165 free_page((unsigned long) buf
);
3169 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
3171 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
3172 struct buffer_head
*bh
;
3173 struct ext4_super_block
*es
= NULL
;
3174 struct ext4_sb_info
*sbi
;
3176 ext4_fsblk_t sb_block
= get_sb_block(&data
);
3177 ext4_fsblk_t logical_sb_block
;
3178 unsigned long offset
= 0;
3179 unsigned long journal_devnum
= 0;
3180 unsigned long def_mount_opts
;
3185 int blocksize
, clustersize
;
3186 unsigned int db_count
;
3188 int needs_recovery
, has_huge_files
, has_bigalloc
;
3191 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
3192 ext4_group_t first_not_zeroed
;
3194 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
3198 sbi
->s_blockgroup_lock
=
3199 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
3200 if (!sbi
->s_blockgroup_lock
) {
3204 sb
->s_fs_info
= sbi
;
3206 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
3207 sbi
->s_sb_block
= sb_block
;
3208 if (sb
->s_bdev
->bd_part
)
3209 sbi
->s_sectors_written_start
=
3210 part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]);
3212 /* Cleanup superblock name */
3213 for (cp
= sb
->s_id
; (cp
= strchr(cp
, '/'));)
3216 /* -EINVAL is default */
3218 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
3220 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
3225 * The ext4 superblock will not be buffer aligned for other than 1kB
3226 * block sizes. We need to calculate the offset from buffer start.
3228 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
3229 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3230 offset
= do_div(logical_sb_block
, blocksize
);
3232 logical_sb_block
= sb_block
;
3235 if (!(bh
= sb_bread(sb
, logical_sb_block
))) {
3236 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
3240 * Note: s_es must be initialized as soon as possible because
3241 * some ext4 macro-instructions depend on its value
3243 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
3245 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
3246 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
3248 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
3250 /* Warn if metadata_csum and gdt_csum are both set. */
3251 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3252 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
) &&
3253 EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_GDT_CSUM
))
3254 ext4_warning(sb
, KERN_INFO
"metadata_csum and uninit_bg are "
3255 "redundant flags; please run fsck.");
3257 /* Check for a known checksum algorithm */
3258 if (!ext4_verify_csum_type(sb
, es
)) {
3259 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3260 "unknown checksum algorithm.");
3265 /* Load the checksum driver */
3266 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3267 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
)) {
3268 sbi
->s_chksum_driver
= crypto_alloc_shash("crc32c", 0, 0);
3269 if (IS_ERR(sbi
->s_chksum_driver
)) {
3270 ext4_msg(sb
, KERN_ERR
, "Cannot load crc32c driver.");
3271 ret
= PTR_ERR(sbi
->s_chksum_driver
);
3272 sbi
->s_chksum_driver
= NULL
;
3277 /* Check superblock checksum */
3278 if (!ext4_superblock_csum_verify(sb
, es
)) {
3279 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3280 "invalid superblock checksum. Run e2fsck?");
3285 /* Precompute checksum seed for all metadata */
3286 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3287 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
))
3288 sbi
->s_csum_seed
= ext4_chksum(sbi
, ~0, es
->s_uuid
,
3289 sizeof(es
->s_uuid
));
3291 /* Set defaults before we parse the mount options */
3292 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
3293 set_opt(sb
, INIT_INODE_TABLE
);
3294 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
3296 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
)
3298 if (def_mount_opts
& EXT4_DEFM_UID16
)
3299 set_opt(sb
, NO_UID32
);
3300 /* xattr user namespace & acls are now defaulted on */
3301 set_opt(sb
, XATTR_USER
);
3302 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3303 set_opt(sb
, POSIX_ACL
);
3305 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
3306 set_opt(sb
, JOURNAL_DATA
);
3307 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
3308 set_opt(sb
, ORDERED_DATA
);
3309 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
3310 set_opt(sb
, WRITEBACK_DATA
);
3312 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
3313 set_opt(sb
, ERRORS_PANIC
);
3314 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
3315 set_opt(sb
, ERRORS_CONT
);
3317 set_opt(sb
, ERRORS_RO
);
3318 if (def_mount_opts
& EXT4_DEFM_BLOCK_VALIDITY
)
3319 set_opt(sb
, BLOCK_VALIDITY
);
3320 if (def_mount_opts
& EXT4_DEFM_DISCARD
)
3321 set_opt(sb
, DISCARD
);
3323 sbi
->s_resuid
= make_kuid(&init_user_ns
, le16_to_cpu(es
->s_def_resuid
));
3324 sbi
->s_resgid
= make_kgid(&init_user_ns
, le16_to_cpu(es
->s_def_resgid
));
3325 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
3326 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
3327 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
3329 if ((def_mount_opts
& EXT4_DEFM_NOBARRIER
) == 0)
3330 set_opt(sb
, BARRIER
);
3333 * enable delayed allocation by default
3334 * Use -o nodelalloc to turn it off
3336 if (!IS_EXT3_SB(sb
) && !IS_EXT2_SB(sb
) &&
3337 ((def_mount_opts
& EXT4_DEFM_NODELALLOC
) == 0))
3338 set_opt(sb
, DELALLOC
);
3341 * set default s_li_wait_mult for lazyinit, for the case there is
3342 * no mount option specified.
3344 sbi
->s_li_wait_mult
= EXT4_DEF_LI_WAIT_MULT
;
3346 if (!parse_options((char *) sbi
->s_es
->s_mount_opts
, sb
,
3347 &journal_devnum
, &journal_ioprio
, 0)) {
3348 ext4_msg(sb
, KERN_WARNING
,
3349 "failed to parse options in superblock: %s",
3350 sbi
->s_es
->s_mount_opts
);
3352 sbi
->s_def_mount_opt
= sbi
->s_mount_opt
;
3353 if (!parse_options((char *) data
, sb
, &journal_devnum
,
3354 &journal_ioprio
, 0))
3357 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
3358 printk_once(KERN_WARNING
"EXT4-fs: Warning: mounting "
3359 "with data=journal disables delayed "
3360 "allocation and O_DIRECT support!\n");
3361 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
3362 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3363 "both data=journal and delalloc");
3366 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
3367 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3368 "both data=journal and delalloc");
3371 if (test_opt(sb
, DELALLOC
))
3372 clear_opt(sb
, DELALLOC
);
3375 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3376 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
3378 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
3379 (EXT4_HAS_COMPAT_FEATURE(sb
, ~0U) ||
3380 EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
3381 EXT4_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
3382 ext4_msg(sb
, KERN_WARNING
,
3383 "feature flags set on rev 0 fs, "
3384 "running e2fsck is recommended");
3386 if (IS_EXT2_SB(sb
)) {
3387 if (ext2_feature_set_ok(sb
))
3388 ext4_msg(sb
, KERN_INFO
, "mounting ext2 file system "
3389 "using the ext4 subsystem");
3391 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext2 due "
3392 "to feature incompatibilities");
3397 if (IS_EXT3_SB(sb
)) {
3398 if (ext3_feature_set_ok(sb
))
3399 ext4_msg(sb
, KERN_INFO
, "mounting ext3 file system "
3400 "using the ext4 subsystem");
3402 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext3 due "
3403 "to feature incompatibilities");
3409 * Check feature flags regardless of the revision level, since we
3410 * previously didn't change the revision level when setting the flags,
3411 * so there is a chance incompat flags are set on a rev 0 filesystem.
3413 if (!ext4_feature_set_ok(sb
, (sb
->s_flags
& MS_RDONLY
)))
3416 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
3417 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
3418 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
3419 ext4_msg(sb
, KERN_ERR
,
3420 "Unsupported filesystem blocksize %d", blocksize
);
3424 if (sb
->s_blocksize
!= blocksize
) {
3425 /* Validate the filesystem blocksize */
3426 if (!sb_set_blocksize(sb
, blocksize
)) {
3427 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
3433 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3434 offset
= do_div(logical_sb_block
, blocksize
);
3435 bh
= sb_bread(sb
, logical_sb_block
);
3437 ext4_msg(sb
, KERN_ERR
,
3438 "Can't read superblock on 2nd try");
3441 es
= (struct ext4_super_block
*)(bh
->b_data
+ offset
);
3443 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
3444 ext4_msg(sb
, KERN_ERR
,
3445 "Magic mismatch, very weird!");
3450 has_huge_files
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3451 EXT4_FEATURE_RO_COMPAT_HUGE_FILE
);
3452 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
3454 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
3456 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
3457 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
3458 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
3460 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
3461 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
3462 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
3463 (!is_power_of_2(sbi
->s_inode_size
)) ||
3464 (sbi
->s_inode_size
> blocksize
)) {
3465 ext4_msg(sb
, KERN_ERR
,
3466 "unsupported inode size: %d",
3470 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
3471 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
3474 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
3475 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
)) {
3476 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
3477 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
3478 !is_power_of_2(sbi
->s_desc_size
)) {
3479 ext4_msg(sb
, KERN_ERR
,
3480 "unsupported descriptor size %lu",
3485 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
3487 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
3488 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
3489 if (EXT4_INODE_SIZE(sb
) == 0 || EXT4_INODES_PER_GROUP(sb
) == 0)
3492 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
3493 if (sbi
->s_inodes_per_block
== 0)
3495 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
3496 sbi
->s_inodes_per_block
;
3497 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
3499 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3500 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
3501 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
3503 for (i
= 0; i
< 4; i
++)
3504 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
3505 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
3506 i
= le32_to_cpu(es
->s_flags
);
3507 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
3508 sbi
->s_hash_unsigned
= 3;
3509 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
3510 #ifdef __CHAR_UNSIGNED__
3511 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
3512 sbi
->s_hash_unsigned
= 3;
3514 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
3518 /* Handle clustersize */
3519 clustersize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_cluster_size
);
3520 has_bigalloc
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3521 EXT4_FEATURE_RO_COMPAT_BIGALLOC
);
3523 if (clustersize
< blocksize
) {
3524 ext4_msg(sb
, KERN_ERR
,
3525 "cluster size (%d) smaller than "
3526 "block size (%d)", clustersize
, blocksize
);
3529 sbi
->s_cluster_bits
= le32_to_cpu(es
->s_log_cluster_size
) -
3530 le32_to_cpu(es
->s_log_block_size
);
3531 sbi
->s_clusters_per_group
=
3532 le32_to_cpu(es
->s_clusters_per_group
);
3533 if (sbi
->s_clusters_per_group
> blocksize
* 8) {
3534 ext4_msg(sb
, KERN_ERR
,
3535 "#clusters per group too big: %lu",
3536 sbi
->s_clusters_per_group
);
3539 if (sbi
->s_blocks_per_group
!=
3540 (sbi
->s_clusters_per_group
* (clustersize
/ blocksize
))) {
3541 ext4_msg(sb
, KERN_ERR
, "blocks per group (%lu) and "
3542 "clusters per group (%lu) inconsistent",
3543 sbi
->s_blocks_per_group
,
3544 sbi
->s_clusters_per_group
);
3548 if (clustersize
!= blocksize
) {
3549 ext4_warning(sb
, "fragment/cluster size (%d) != "
3550 "block size (%d)", clustersize
,
3552 clustersize
= blocksize
;
3554 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
3555 ext4_msg(sb
, KERN_ERR
,
3556 "#blocks per group too big: %lu",
3557 sbi
->s_blocks_per_group
);
3560 sbi
->s_clusters_per_group
= sbi
->s_blocks_per_group
;
3561 sbi
->s_cluster_bits
= 0;
3563 sbi
->s_cluster_ratio
= clustersize
/ blocksize
;
3565 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
3566 ext4_msg(sb
, KERN_ERR
,
3567 "#inodes per group too big: %lu",
3568 sbi
->s_inodes_per_group
);
3573 * Test whether we have more sectors than will fit in sector_t,
3574 * and whether the max offset is addressable by the page cache.
3576 err
= generic_check_addressable(sb
->s_blocksize_bits
,
3577 ext4_blocks_count(es
));
3579 ext4_msg(sb
, KERN_ERR
, "filesystem"
3580 " too large to mount safely on this system");
3581 if (sizeof(sector_t
) < 8)
3582 ext4_msg(sb
, KERN_WARNING
, "CONFIG_LBDAF not enabled");
3586 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
3589 /* check blocks count against device size */
3590 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
3591 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
3592 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
3593 "exceeds size of device (%llu blocks)",
3594 ext4_blocks_count(es
), blocks_count
);
3599 * It makes no sense for the first data block to be beyond the end
3600 * of the filesystem.
3602 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
3603 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data "
3604 "block %u is beyond end of filesystem (%llu)",
3605 le32_to_cpu(es
->s_first_data_block
),
3606 ext4_blocks_count(es
));
3609 blocks_count
= (ext4_blocks_count(es
) -
3610 le32_to_cpu(es
->s_first_data_block
) +
3611 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
3612 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
3613 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
3614 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %u "
3615 "(block count %llu, first data block %u, "
3616 "blocks per group %lu)", sbi
->s_groups_count
,
3617 ext4_blocks_count(es
),
3618 le32_to_cpu(es
->s_first_data_block
),
3619 EXT4_BLOCKS_PER_GROUP(sb
));
3622 sbi
->s_groups_count
= blocks_count
;
3623 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
3624 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
3625 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
3626 EXT4_DESC_PER_BLOCK(sb
);
3627 sbi
->s_group_desc
= ext4_kvmalloc(db_count
*
3628 sizeof(struct buffer_head
*),
3630 if (sbi
->s_group_desc
== NULL
) {
3631 ext4_msg(sb
, KERN_ERR
, "not enough memory");
3637 sbi
->s_proc
= proc_mkdir(sb
->s_id
, ext4_proc_root
);
3640 proc_create_data("options", S_IRUGO
, sbi
->s_proc
,
3641 &ext4_seq_options_fops
, sb
);
3643 bgl_lock_init(sbi
->s_blockgroup_lock
);
3645 for (i
= 0; i
< db_count
; i
++) {
3646 block
= descriptor_loc(sb
, logical_sb_block
, i
);
3647 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
3648 if (!sbi
->s_group_desc
[i
]) {
3649 ext4_msg(sb
, KERN_ERR
,
3650 "can't read group descriptor %d", i
);
3655 if (!ext4_check_descriptors(sb
, &first_not_zeroed
)) {
3656 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
3659 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
3660 if (!ext4_fill_flex_info(sb
)) {
3661 ext4_msg(sb
, KERN_ERR
,
3662 "unable to initialize "
3663 "flex_bg meta info!");
3667 sbi
->s_gdb_count
= db_count
;
3668 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
3669 spin_lock_init(&sbi
->s_next_gen_lock
);
3671 init_timer(&sbi
->s_err_report
);
3672 sbi
->s_err_report
.function
= print_daily_error_info
;
3673 sbi
->s_err_report
.data
= (unsigned long) sb
;
3675 err
= percpu_counter_init(&sbi
->s_freeclusters_counter
,
3676 ext4_count_free_clusters(sb
));
3678 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
,
3679 ext4_count_free_inodes(sb
));
3682 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
3683 ext4_count_dirs(sb
));
3686 err
= percpu_counter_init(&sbi
->s_dirtyclusters_counter
, 0);
3689 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
3693 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
3694 sbi
->s_max_writeback_mb_bump
= 128;
3695 sbi
->s_extent_max_zeroout_kb
= 32;
3698 * set up enough so that it can read an inode
3700 if (!test_opt(sb
, NOLOAD
) &&
3701 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
3702 sb
->s_op
= &ext4_sops
;
3704 sb
->s_op
= &ext4_nojournal_sops
;
3705 sb
->s_export_op
= &ext4_export_ops
;
3706 sb
->s_xattr
= ext4_xattr_handlers
;
3708 sb
->s_qcop
= &ext4_qctl_operations
;
3709 sb
->dq_op
= &ext4_quota_operations
;
3711 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
)) {
3712 /* Use qctl operations for hidden quota files. */
3713 sb
->s_qcop
= &ext4_qctl_sysfile_operations
;
3716 memcpy(sb
->s_uuid
, es
->s_uuid
, sizeof(es
->s_uuid
));
3718 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
3719 mutex_init(&sbi
->s_orphan_lock
);
3723 needs_recovery
= (es
->s_last_orphan
!= 0 ||
3724 EXT4_HAS_INCOMPAT_FEATURE(sb
,
3725 EXT4_FEATURE_INCOMPAT_RECOVER
));
3727 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_MMP
) &&
3728 !(sb
->s_flags
& MS_RDONLY
))
3729 if (ext4_multi_mount_protect(sb
, le64_to_cpu(es
->s_mmp_block
)))
3733 * The first inode we look at is the journal inode. Don't try
3734 * root first: it may be modified in the journal!
3736 if (!test_opt(sb
, NOLOAD
) &&
3737 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
3738 if (ext4_load_journal(sb
, es
, journal_devnum
))
3740 } else if (test_opt(sb
, NOLOAD
) && !(sb
->s_flags
& MS_RDONLY
) &&
3741 EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
3742 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
3743 "suppressed and not mounted read-only");
3744 goto failed_mount_wq
;
3746 clear_opt(sb
, DATA_FLAGS
);
3747 sbi
->s_journal
= NULL
;
3752 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
) &&
3753 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
3754 JBD2_FEATURE_INCOMPAT_64BIT
)) {
3755 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
3756 goto failed_mount_wq
;
3759 if (!set_journal_csum_feature_set(sb
)) {
3760 ext4_msg(sb
, KERN_ERR
, "Failed to set journal checksum "
3762 goto failed_mount_wq
;
3765 /* We have now updated the journal if required, so we can
3766 * validate the data journaling mode. */
3767 switch (test_opt(sb
, DATA_FLAGS
)) {
3769 /* No mode set, assume a default based on the journal
3770 * capabilities: ORDERED_DATA if the journal can
3771 * cope, else JOURNAL_DATA
3773 if (jbd2_journal_check_available_features
3774 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
3775 set_opt(sb
, ORDERED_DATA
);
3777 set_opt(sb
, JOURNAL_DATA
);
3780 case EXT4_MOUNT_ORDERED_DATA
:
3781 case EXT4_MOUNT_WRITEBACK_DATA
:
3782 if (!jbd2_journal_check_available_features
3783 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
3784 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
3785 "requested data journaling mode");
3786 goto failed_mount_wq
;
3791 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
3793 sbi
->s_journal
->j_commit_callback
= ext4_journal_commit_callback
;
3796 * The journal may have updated the bg summary counts, so we
3797 * need to update the global counters.
3799 percpu_counter_set(&sbi
->s_freeclusters_counter
,
3800 ext4_count_free_clusters(sb
));
3801 percpu_counter_set(&sbi
->s_freeinodes_counter
,
3802 ext4_count_free_inodes(sb
));
3803 percpu_counter_set(&sbi
->s_dirs_counter
,
3804 ext4_count_dirs(sb
));
3805 percpu_counter_set(&sbi
->s_dirtyclusters_counter
, 0);
3809 * Get the # of file system overhead blocks from the
3810 * superblock if present.
3812 if (es
->s_overhead_clusters
)
3813 sbi
->s_overhead
= le32_to_cpu(es
->s_overhead_clusters
);
3815 err
= ext4_calculate_overhead(sb
);
3817 goto failed_mount_wq
;
3821 * The maximum number of concurrent works can be high and
3822 * concurrency isn't really necessary. Limit it to 1.
3824 EXT4_SB(sb
)->dio_unwritten_wq
=
3825 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM
| WQ_UNBOUND
, 1);
3826 if (!EXT4_SB(sb
)->dio_unwritten_wq
) {
3827 printk(KERN_ERR
"EXT4-fs: failed to create DIO workqueue\n");
3829 goto failed_mount_wq
;
3833 * The jbd2_journal_load will have done any necessary log recovery,
3834 * so we can safely mount the rest of the filesystem now.
3837 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
3839 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
3840 ret
= PTR_ERR(root
);
3844 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
3845 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
3849 sb
->s_root
= d_make_root(root
);
3851 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
3856 if (ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
))
3857 sb
->s_flags
|= MS_RDONLY
;
3859 /* determine the minimum size of new large inodes, if present */
3860 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
3861 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
3862 EXT4_GOOD_OLD_INODE_SIZE
;
3863 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3864 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE
)) {
3865 if (sbi
->s_want_extra_isize
<
3866 le16_to_cpu(es
->s_want_extra_isize
))
3867 sbi
->s_want_extra_isize
=
3868 le16_to_cpu(es
->s_want_extra_isize
);
3869 if (sbi
->s_want_extra_isize
<
3870 le16_to_cpu(es
->s_min_extra_isize
))
3871 sbi
->s_want_extra_isize
=
3872 le16_to_cpu(es
->s_min_extra_isize
);
3875 /* Check if enough inode space is available */
3876 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
3877 sbi
->s_inode_size
) {
3878 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
3879 EXT4_GOOD_OLD_INODE_SIZE
;
3880 ext4_msg(sb
, KERN_INFO
, "required extra inode space not"
3884 err
= ext4_setup_system_zone(sb
);
3886 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
3888 goto failed_mount4a
;
3892 err
= ext4_mb_init(sb
);
3894 ext4_msg(sb
, KERN_ERR
, "failed to initialize mballoc (%d)",
3899 err
= ext4_register_li_request(sb
, first_not_zeroed
);
3903 sbi
->s_kobj
.kset
= ext4_kset
;
3904 init_completion(&sbi
->s_kobj_unregister
);
3905 err
= kobject_init_and_add(&sbi
->s_kobj
, &ext4_ktype
, NULL
,
3910 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
3911 ext4_orphan_cleanup(sb
, es
);
3912 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
3913 if (needs_recovery
) {
3914 ext4_msg(sb
, KERN_INFO
, "recovery complete");
3915 ext4_mark_recovery_complete(sb
, es
);
3917 if (EXT4_SB(sb
)->s_journal
) {
3918 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
3919 descr
= " journalled data mode";
3920 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
3921 descr
= " ordered data mode";
3923 descr
= " writeback data mode";
3925 descr
= "out journal";
3928 /* Enable quota usage during mount. */
3929 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
) &&
3930 !(sb
->s_flags
& MS_RDONLY
)) {
3931 err
= ext4_enable_quotas(sb
);
3935 #endif /* CONFIG_QUOTA */
3937 if (test_opt(sb
, DISCARD
)) {
3938 struct request_queue
*q
= bdev_get_queue(sb
->s_bdev
);
3939 if (!blk_queue_discard(q
))
3940 ext4_msg(sb
, KERN_WARNING
,
3941 "mounting with \"discard\" option, but "
3942 "the device does not support discard");
3945 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s. "
3946 "Opts: %s%s%s", descr
, sbi
->s_es
->s_mount_opts
,
3947 *sbi
->s_es
->s_mount_opts
? "; " : "", orig_data
);
3949 if (es
->s_error_count
)
3950 mod_timer(&sbi
->s_err_report
, jiffies
+ 300*HZ
); /* 5 minutes */
3957 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
3962 kobject_del(&sbi
->s_kobj
);
3965 ext4_unregister_li_request(sb
);
3967 ext4_mb_release(sb
);
3969 ext4_ext_release(sb
);
3970 ext4_release_system_zone(sb
);
3975 ext4_msg(sb
, KERN_ERR
, "mount failed");
3976 destroy_workqueue(EXT4_SB(sb
)->dio_unwritten_wq
);
3978 if (sbi
->s_journal
) {
3979 jbd2_journal_destroy(sbi
->s_journal
);
3980 sbi
->s_journal
= NULL
;
3983 del_timer(&sbi
->s_err_report
);
3984 if (sbi
->s_flex_groups
)
3985 ext4_kvfree(sbi
->s_flex_groups
);
3986 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
3987 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
3988 percpu_counter_destroy(&sbi
->s_dirs_counter
);
3989 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
3991 kthread_stop(sbi
->s_mmp_tsk
);
3993 for (i
= 0; i
< db_count
; i
++)
3994 brelse(sbi
->s_group_desc
[i
]);
3995 ext4_kvfree(sbi
->s_group_desc
);
3997 if (sbi
->s_chksum_driver
)
3998 crypto_free_shash(sbi
->s_chksum_driver
);
4000 remove_proc_entry("options", sbi
->s_proc
);
4001 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
4004 for (i
= 0; i
< MAXQUOTAS
; i
++)
4005 kfree(sbi
->s_qf_names
[i
]);
4007 ext4_blkdev_remove(sbi
);
4010 sb
->s_fs_info
= NULL
;
4011 kfree(sbi
->s_blockgroup_lock
);
4015 return err
? err
: ret
;
4019 * Setup any per-fs journal parameters now. We'll do this both on
4020 * initial mount, once the journal has been initialised but before we've
4021 * done any recovery; and again on any subsequent remount.
4023 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
4025 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4027 journal
->j_commit_interval
= sbi
->s_commit_interval
;
4028 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
4029 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
4031 write_lock(&journal
->j_state_lock
);
4032 if (test_opt(sb
, BARRIER
))
4033 journal
->j_flags
|= JBD2_BARRIER
;
4035 journal
->j_flags
&= ~JBD2_BARRIER
;
4036 if (test_opt(sb
, DATA_ERR_ABORT
))
4037 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
4039 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
4040 write_unlock(&journal
->j_state_lock
);
4043 static journal_t
*ext4_get_journal(struct super_block
*sb
,
4044 unsigned int journal_inum
)
4046 struct inode
*journal_inode
;
4049 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4051 /* First, test for the existence of a valid inode on disk. Bad
4052 * things happen if we iget() an unused inode, as the subsequent
4053 * iput() will try to delete it. */
4055 journal_inode
= ext4_iget(sb
, journal_inum
);
4056 if (IS_ERR(journal_inode
)) {
4057 ext4_msg(sb
, KERN_ERR
, "no journal found");
4060 if (!journal_inode
->i_nlink
) {
4061 make_bad_inode(journal_inode
);
4062 iput(journal_inode
);
4063 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
4067 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4068 journal_inode
, journal_inode
->i_size
);
4069 if (!S_ISREG(journal_inode
->i_mode
)) {
4070 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
4071 iput(journal_inode
);
4075 journal
= jbd2_journal_init_inode(journal_inode
);
4077 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
4078 iput(journal_inode
);
4081 journal
->j_private
= sb
;
4082 ext4_init_journal_params(sb
, journal
);
4086 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
4089 struct buffer_head
*bh
;
4093 int hblock
, blocksize
;
4094 ext4_fsblk_t sb_block
;
4095 unsigned long offset
;
4096 struct ext4_super_block
*es
;
4097 struct block_device
*bdev
;
4099 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4101 bdev
= ext4_blkdev_get(j_dev
, sb
);
4105 blocksize
= sb
->s_blocksize
;
4106 hblock
= bdev_logical_block_size(bdev
);
4107 if (blocksize
< hblock
) {
4108 ext4_msg(sb
, KERN_ERR
,
4109 "blocksize too small for journal device");
4113 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
4114 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
4115 set_blocksize(bdev
, blocksize
);
4116 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
4117 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
4118 "external journal");
4122 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
4123 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
4124 !(le32_to_cpu(es
->s_feature_incompat
) &
4125 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
4126 ext4_msg(sb
, KERN_ERR
, "external journal has "
4132 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
4133 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
4138 len
= ext4_blocks_count(es
);
4139 start
= sb_block
+ 1;
4140 brelse(bh
); /* we're done with the superblock */
4142 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
4143 start
, len
, blocksize
);
4145 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
4148 journal
->j_private
= sb
;
4149 ll_rw_block(READ
, 1, &journal
->j_sb_buffer
);
4150 wait_on_buffer(journal
->j_sb_buffer
);
4151 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
4152 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
4155 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
4156 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
4157 "user (unsupported) - %d",
4158 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
4161 EXT4_SB(sb
)->journal_bdev
= bdev
;
4162 ext4_init_journal_params(sb
, journal
);
4166 jbd2_journal_destroy(journal
);
4168 ext4_blkdev_put(bdev
);
4172 static int ext4_load_journal(struct super_block
*sb
,
4173 struct ext4_super_block
*es
,
4174 unsigned long journal_devnum
)
4177 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
4180 int really_read_only
;
4182 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4184 if (journal_devnum
&&
4185 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4186 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
4187 "numbers have changed");
4188 journal_dev
= new_decode_dev(journal_devnum
);
4190 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
4192 really_read_only
= bdev_read_only(sb
->s_bdev
);
4195 * Are we loading a blank journal or performing recovery after a
4196 * crash? For recovery, we need to check in advance whether we
4197 * can get read-write access to the device.
4199 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
4200 if (sb
->s_flags
& MS_RDONLY
) {
4201 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
4202 "required on readonly filesystem");
4203 if (really_read_only
) {
4204 ext4_msg(sb
, KERN_ERR
, "write access "
4205 "unavailable, cannot proceed");
4208 ext4_msg(sb
, KERN_INFO
, "write access will "
4209 "be enabled during recovery");
4213 if (journal_inum
&& journal_dev
) {
4214 ext4_msg(sb
, KERN_ERR
, "filesystem has both journal "
4215 "and inode journals!");
4220 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
4223 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
4227 if (!(journal
->j_flags
& JBD2_BARRIER
))
4228 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
4230 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
))
4231 err
= jbd2_journal_wipe(journal
, !really_read_only
);
4233 char *save
= kmalloc(EXT4_S_ERR_LEN
, GFP_KERNEL
);
4235 memcpy(save
, ((char *) es
) +
4236 EXT4_S_ERR_START
, EXT4_S_ERR_LEN
);
4237 err
= jbd2_journal_load(journal
);
4239 memcpy(((char *) es
) + EXT4_S_ERR_START
,
4240 save
, EXT4_S_ERR_LEN
);
4245 ext4_msg(sb
, KERN_ERR
, "error loading journal");
4246 jbd2_journal_destroy(journal
);
4250 EXT4_SB(sb
)->s_journal
= journal
;
4251 ext4_clear_journal_err(sb
, es
);
4253 if (!really_read_only
&& journal_devnum
&&
4254 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4255 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
4257 /* Make sure we flush the recovery flag to disk. */
4258 ext4_commit_super(sb
, 1);
4264 static int ext4_commit_super(struct super_block
*sb
, int sync
)
4266 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
4267 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
4270 if (!sbh
|| block_device_ejected(sb
))
4272 if (buffer_write_io_error(sbh
)) {
4274 * Oh, dear. A previous attempt to write the
4275 * superblock failed. This could happen because the
4276 * USB device was yanked out. Or it could happen to
4277 * be a transient write error and maybe the block will
4278 * be remapped. Nothing we can do but to retry the
4279 * write and hope for the best.
4281 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
4282 "superblock detected");
4283 clear_buffer_write_io_error(sbh
);
4284 set_buffer_uptodate(sbh
);
4287 * If the file system is mounted read-only, don't update the
4288 * superblock write time. This avoids updating the superblock
4289 * write time when we are mounting the root file system
4290 * read/only but we need to replay the journal; at that point,
4291 * for people who are east of GMT and who make their clock
4292 * tick in localtime for Windows bug-for-bug compatibility,
4293 * the clock is set in the future, and this will cause e2fsck
4294 * to complain and force a full file system check.
4296 if (!(sb
->s_flags
& MS_RDONLY
))
4297 es
->s_wtime
= cpu_to_le32(get_seconds());
4298 if (sb
->s_bdev
->bd_part
)
4299 es
->s_kbytes_written
=
4300 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
+
4301 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
4302 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
4304 es
->s_kbytes_written
=
4305 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
);
4306 ext4_free_blocks_count_set(es
,
4307 EXT4_C2B(EXT4_SB(sb
), percpu_counter_sum_positive(
4308 &EXT4_SB(sb
)->s_freeclusters_counter
)));
4309 es
->s_free_inodes_count
=
4310 cpu_to_le32(percpu_counter_sum_positive(
4311 &EXT4_SB(sb
)->s_freeinodes_counter
));
4312 BUFFER_TRACE(sbh
, "marking dirty");
4313 ext4_superblock_csum_set(sb
);
4314 mark_buffer_dirty(sbh
);
4316 error
= sync_dirty_buffer(sbh
);
4320 error
= buffer_write_io_error(sbh
);
4322 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
4324 clear_buffer_write_io_error(sbh
);
4325 set_buffer_uptodate(sbh
);
4332 * Have we just finished recovery? If so, and if we are mounting (or
4333 * remounting) the filesystem readonly, then we will end up with a
4334 * consistent fs on disk. Record that fact.
4336 static void ext4_mark_recovery_complete(struct super_block
*sb
,
4337 struct ext4_super_block
*es
)
4339 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
4341 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
4342 BUG_ON(journal
!= NULL
);
4345 jbd2_journal_lock_updates(journal
);
4346 if (jbd2_journal_flush(journal
) < 0)
4349 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
) &&
4350 sb
->s_flags
& MS_RDONLY
) {
4351 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4352 ext4_commit_super(sb
, 1);
4356 jbd2_journal_unlock_updates(journal
);
4360 * If we are mounting (or read-write remounting) a filesystem whose journal
4361 * has recorded an error from a previous lifetime, move that error to the
4362 * main filesystem now.
4364 static void ext4_clear_journal_err(struct super_block
*sb
,
4365 struct ext4_super_block
*es
)
4371 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4373 journal
= EXT4_SB(sb
)->s_journal
;
4376 * Now check for any error status which may have been recorded in the
4377 * journal by a prior ext4_error() or ext4_abort()
4380 j_errno
= jbd2_journal_errno(journal
);
4384 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
4385 ext4_warning(sb
, "Filesystem error recorded "
4386 "from previous mount: %s", errstr
);
4387 ext4_warning(sb
, "Marking fs in need of filesystem check.");
4389 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
4390 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
4391 ext4_commit_super(sb
, 1);
4393 jbd2_journal_clear_err(journal
);
4394 jbd2_journal_update_sb_errno(journal
);
4399 * Force the running and committing transactions to commit,
4400 * and wait on the commit.
4402 int ext4_force_commit(struct super_block
*sb
)
4406 if (sb
->s_flags
& MS_RDONLY
)
4409 journal
= EXT4_SB(sb
)->s_journal
;
4410 return ext4_journal_force_commit(journal
);
4413 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
4417 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4419 trace_ext4_sync_fs(sb
, wait
);
4420 flush_workqueue(sbi
->dio_unwritten_wq
);
4422 * Writeback quota in non-journalled quota case - journalled quota has
4425 dquot_writeback_dquots(sb
, -1);
4426 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
4428 jbd2_log_wait_commit(sbi
->s_journal
, target
);
4434 * LVM calls this function before a (read-only) snapshot is created. This
4435 * gives us a chance to flush the journal completely and mark the fs clean.
4437 * Note that only this function cannot bring a filesystem to be in a clean
4438 * state independently. It relies on upper layer to stop all data & metadata
4441 static int ext4_freeze(struct super_block
*sb
)
4446 if (sb
->s_flags
& MS_RDONLY
)
4449 journal
= EXT4_SB(sb
)->s_journal
;
4451 /* Now we set up the journal barrier. */
4452 jbd2_journal_lock_updates(journal
);
4455 * Don't clear the needs_recovery flag if we failed to flush
4458 error
= jbd2_journal_flush(journal
);
4462 /* Journal blocked and flushed, clear needs_recovery flag. */
4463 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4464 error
= ext4_commit_super(sb
, 1);
4466 /* we rely on upper layer to stop further updates */
4467 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
4472 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4473 * flag here, even though the filesystem is not technically dirty yet.
4475 static int ext4_unfreeze(struct super_block
*sb
)
4477 if (sb
->s_flags
& MS_RDONLY
)
4480 /* Reset the needs_recovery flag before the fs is unlocked. */
4481 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4482 ext4_commit_super(sb
, 1);
4487 * Structure to save mount options for ext4_remount's benefit
4489 struct ext4_mount_options
{
4490 unsigned long s_mount_opt
;
4491 unsigned long s_mount_opt2
;
4494 unsigned long s_commit_interval
;
4495 u32 s_min_batch_time
, s_max_batch_time
;
4498 char *s_qf_names
[MAXQUOTAS
];
4502 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
4504 struct ext4_super_block
*es
;
4505 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4506 unsigned long old_sb_flags
;
4507 struct ext4_mount_options old_opts
;
4508 int enable_quota
= 0;
4510 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
4515 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
4517 /* Store the original options */
4518 old_sb_flags
= sb
->s_flags
;
4519 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
4520 old_opts
.s_mount_opt2
= sbi
->s_mount_opt2
;
4521 old_opts
.s_resuid
= sbi
->s_resuid
;
4522 old_opts
.s_resgid
= sbi
->s_resgid
;
4523 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
4524 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
4525 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
4527 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
4528 for (i
= 0; i
< MAXQUOTAS
; i
++)
4529 if (sbi
->s_qf_names
[i
]) {
4530 old_opts
.s_qf_names
[i
] = kstrdup(sbi
->s_qf_names
[i
],
4532 if (!old_opts
.s_qf_names
[i
]) {
4533 for (j
= 0; j
< i
; j
++)
4534 kfree(old_opts
.s_qf_names
[j
]);
4538 old_opts
.s_qf_names
[i
] = NULL
;
4540 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
4541 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
4544 * Allow the "check" option to be passed as a remount option.
4546 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
, 1)) {
4551 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
)
4552 ext4_abort(sb
, "Abort forced by user");
4554 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
4555 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
4559 if (sbi
->s_journal
) {
4560 ext4_init_journal_params(sb
, sbi
->s_journal
);
4561 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
4564 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
)) {
4565 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
) {
4570 if (*flags
& MS_RDONLY
) {
4571 err
= dquot_suspend(sb
, -1);
4576 * First of all, the unconditional stuff we have to do
4577 * to disable replay of the journal when we next remount
4579 sb
->s_flags
|= MS_RDONLY
;
4582 * OK, test if we are remounting a valid rw partition
4583 * readonly, and if so set the rdonly flag and then
4584 * mark the partition as valid again.
4586 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
4587 (sbi
->s_mount_state
& EXT4_VALID_FS
))
4588 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
4591 ext4_mark_recovery_complete(sb
, es
);
4593 /* Make sure we can mount this feature set readwrite */
4594 if (!ext4_feature_set_ok(sb
, 0)) {
4599 * Make sure the group descriptor checksums
4600 * are sane. If they aren't, refuse to remount r/w.
4602 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
4603 struct ext4_group_desc
*gdp
=
4604 ext4_get_group_desc(sb
, g
, NULL
);
4606 if (!ext4_group_desc_csum_verify(sb
, g
, gdp
)) {
4607 ext4_msg(sb
, KERN_ERR
,
4608 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4609 g
, le16_to_cpu(ext4_group_desc_csum(sbi
, g
, gdp
)),
4610 le16_to_cpu(gdp
->bg_checksum
));
4617 * If we have an unprocessed orphan list hanging
4618 * around from a previously readonly bdev mount,
4619 * require a full umount/remount for now.
4621 if (es
->s_last_orphan
) {
4622 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
4623 "remount RDWR because of unprocessed "
4624 "orphan inode list. Please "
4625 "umount/remount instead");
4631 * Mounting a RDONLY partition read-write, so reread
4632 * and store the current valid flag. (It may have
4633 * been changed by e2fsck since we originally mounted
4637 ext4_clear_journal_err(sb
, es
);
4638 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
4639 if (!ext4_setup_super(sb
, es
, 0))
4640 sb
->s_flags
&= ~MS_RDONLY
;
4641 if (EXT4_HAS_INCOMPAT_FEATURE(sb
,
4642 EXT4_FEATURE_INCOMPAT_MMP
))
4643 if (ext4_multi_mount_protect(sb
,
4644 le64_to_cpu(es
->s_mmp_block
))) {
4653 * Reinitialize lazy itable initialization thread based on
4656 if ((sb
->s_flags
& MS_RDONLY
) || !test_opt(sb
, INIT_INODE_TABLE
))
4657 ext4_unregister_li_request(sb
);
4659 ext4_group_t first_not_zeroed
;
4660 first_not_zeroed
= ext4_has_uninit_itable(sb
);
4661 ext4_register_li_request(sb
, first_not_zeroed
);
4664 ext4_setup_system_zone(sb
);
4665 if (sbi
->s_journal
== NULL
&& !(old_sb_flags
& MS_RDONLY
))
4666 ext4_commit_super(sb
, 1);
4669 /* Release old quota file names */
4670 for (i
= 0; i
< MAXQUOTAS
; i
++)
4671 kfree(old_opts
.s_qf_names
[i
]);
4673 if (sb_any_quota_suspended(sb
))
4674 dquot_resume(sb
, -1);
4675 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
4676 EXT4_FEATURE_RO_COMPAT_QUOTA
)) {
4677 err
= ext4_enable_quotas(sb
);
4684 ext4_msg(sb
, KERN_INFO
, "re-mounted. Opts: %s", orig_data
);
4689 sb
->s_flags
= old_sb_flags
;
4690 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
4691 sbi
->s_mount_opt2
= old_opts
.s_mount_opt2
;
4692 sbi
->s_resuid
= old_opts
.s_resuid
;
4693 sbi
->s_resgid
= old_opts
.s_resgid
;
4694 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
4695 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
4696 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
4698 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
4699 for (i
= 0; i
< MAXQUOTAS
; i
++) {
4700 kfree(sbi
->s_qf_names
[i
]);
4701 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
4708 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
4710 struct super_block
*sb
= dentry
->d_sb
;
4711 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4712 struct ext4_super_block
*es
= sbi
->s_es
;
4713 ext4_fsblk_t overhead
= 0;
4717 if (!test_opt(sb
, MINIX_DF
))
4718 overhead
= sbi
->s_overhead
;
4720 buf
->f_type
= EXT4_SUPER_MAGIC
;
4721 buf
->f_bsize
= sb
->s_blocksize
;
4722 buf
->f_blocks
= ext4_blocks_count(es
) - EXT4_C2B(sbi
, overhead
);
4723 bfree
= percpu_counter_sum_positive(&sbi
->s_freeclusters_counter
) -
4724 percpu_counter_sum_positive(&sbi
->s_dirtyclusters_counter
);
4725 /* prevent underflow in case that few free space is available */
4726 buf
->f_bfree
= EXT4_C2B(sbi
, max_t(s64
, bfree
, 0));
4727 buf
->f_bavail
= buf
->f_bfree
- ext4_r_blocks_count(es
);
4728 if (buf
->f_bfree
< ext4_r_blocks_count(es
))
4730 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
4731 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
4732 buf
->f_namelen
= EXT4_NAME_LEN
;
4733 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
4734 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
4735 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
4736 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
4741 /* Helper function for writing quotas on sync - we need to start transaction
4742 * before quota file is locked for write. Otherwise the are possible deadlocks:
4743 * Process 1 Process 2
4744 * ext4_create() quota_sync()
4745 * jbd2_journal_start() write_dquot()
4746 * dquot_initialize() down(dqio_mutex)
4747 * down(dqio_mutex) jbd2_journal_start()
4753 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
4755 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_id
.type
];
4758 static int ext4_write_dquot(struct dquot
*dquot
)
4762 struct inode
*inode
;
4764 inode
= dquot_to_inode(dquot
);
4765 handle
= ext4_journal_start(inode
,
4766 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
4768 return PTR_ERR(handle
);
4769 ret
= dquot_commit(dquot
);
4770 err
= ext4_journal_stop(handle
);
4776 static int ext4_acquire_dquot(struct dquot
*dquot
)
4781 handle
= ext4_journal_start(dquot_to_inode(dquot
),
4782 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
4784 return PTR_ERR(handle
);
4785 ret
= dquot_acquire(dquot
);
4786 err
= ext4_journal_stop(handle
);
4792 static int ext4_release_dquot(struct dquot
*dquot
)
4797 handle
= ext4_journal_start(dquot_to_inode(dquot
),
4798 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
4799 if (IS_ERR(handle
)) {
4800 /* Release dquot anyway to avoid endless cycle in dqput() */
4801 dquot_release(dquot
);
4802 return PTR_ERR(handle
);
4804 ret
= dquot_release(dquot
);
4805 err
= ext4_journal_stop(handle
);
4811 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
4813 /* Are we journaling quotas? */
4814 if (EXT4_SB(dquot
->dq_sb
)->s_qf_names
[USRQUOTA
] ||
4815 EXT4_SB(dquot
->dq_sb
)->s_qf_names
[GRPQUOTA
]) {
4816 dquot_mark_dquot_dirty(dquot
);
4817 return ext4_write_dquot(dquot
);
4819 return dquot_mark_dquot_dirty(dquot
);
4823 static int ext4_write_info(struct super_block
*sb
, int type
)
4828 /* Data block + inode block */
4829 handle
= ext4_journal_start(sb
->s_root
->d_inode
, 2);
4831 return PTR_ERR(handle
);
4832 ret
= dquot_commit_info(sb
, type
);
4833 err
= ext4_journal_stop(handle
);
4840 * Turn on quotas during mount time - we need to find
4841 * the quota file and such...
4843 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
4845 return dquot_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
4846 EXT4_SB(sb
)->s_jquota_fmt
, type
);
4850 * Standard function to be called on quota_on
4852 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
4857 if (!test_opt(sb
, QUOTA
))
4860 /* Quotafile not on the same filesystem? */
4861 if (path
->dentry
->d_sb
!= sb
)
4863 /* Journaling quota? */
4864 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
4865 /* Quotafile not in fs root? */
4866 if (path
->dentry
->d_parent
!= sb
->s_root
)
4867 ext4_msg(sb
, KERN_WARNING
,
4868 "Quota file not on filesystem root. "
4869 "Journaled quota will not work");
4873 * When we journal data on quota file, we have to flush journal to see
4874 * all updates to the file when we bypass pagecache...
4876 if (EXT4_SB(sb
)->s_journal
&&
4877 ext4_should_journal_data(path
->dentry
->d_inode
)) {
4879 * We don't need to lock updates but journal_flush() could
4880 * otherwise be livelocked...
4882 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
4883 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
4884 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
4889 return dquot_quota_on(sb
, type
, format_id
, path
);
4892 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
4896 struct inode
*qf_inode
;
4897 unsigned long qf_inums
[MAXQUOTAS
] = {
4898 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
4899 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
)
4902 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
));
4904 if (!qf_inums
[type
])
4907 qf_inode
= ext4_iget(sb
, qf_inums
[type
]);
4908 if (IS_ERR(qf_inode
)) {
4909 ext4_error(sb
, "Bad quota inode # %lu", qf_inums
[type
]);
4910 return PTR_ERR(qf_inode
);
4913 err
= dquot_enable(qf_inode
, type
, format_id
, flags
);
4919 /* Enable usage tracking for all quota types. */
4920 static int ext4_enable_quotas(struct super_block
*sb
)
4923 unsigned long qf_inums
[MAXQUOTAS
] = {
4924 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
4925 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
)
4928 sb_dqopt(sb
)->flags
|= DQUOT_QUOTA_SYS_FILE
;
4929 for (type
= 0; type
< MAXQUOTAS
; type
++) {
4930 if (qf_inums
[type
]) {
4931 err
= ext4_quota_enable(sb
, type
, QFMT_VFS_V1
,
4932 DQUOT_USAGE_ENABLED
);
4935 "Failed to enable quota tracking "
4936 "(type=%d, err=%d). Please run "
4937 "e2fsck to fix.", type
, err
);
4946 * quota_on function that is used when QUOTA feature is set.
4948 static int ext4_quota_on_sysfile(struct super_block
*sb
, int type
,
4951 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
))
4955 * USAGE was enabled at mount time. Only need to enable LIMITS now.
4957 return ext4_quota_enable(sb
, type
, format_id
, DQUOT_LIMITS_ENABLED
);
4960 static int ext4_quota_off(struct super_block
*sb
, int type
)
4962 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
4965 /* Force all delayed allocation blocks to be allocated.
4966 * Caller already holds s_umount sem */
4967 if (test_opt(sb
, DELALLOC
))
4968 sync_filesystem(sb
);
4973 /* Update modification times of quota files when userspace can
4974 * start looking at them */
4975 handle
= ext4_journal_start(inode
, 1);
4978 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
4979 ext4_mark_inode_dirty(handle
, inode
);
4980 ext4_journal_stop(handle
);
4983 return dquot_quota_off(sb
, type
);
4987 * quota_off function that is used when QUOTA feature is set.
4989 static int ext4_quota_off_sysfile(struct super_block
*sb
, int type
)
4991 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
))
4994 /* Disable only the limits. */
4995 return dquot_disable(sb
, type
, DQUOT_LIMITS_ENABLED
);
4998 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4999 * acquiring the locks... As quota files are never truncated and quota code
5000 * itself serializes the operations (and no one else should touch the files)
5001 * we don't have to be afraid of races */
5002 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
5003 size_t len
, loff_t off
)
5005 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5006 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5008 int offset
= off
& (sb
->s_blocksize
- 1);
5011 struct buffer_head
*bh
;
5012 loff_t i_size
= i_size_read(inode
);
5016 if (off
+len
> i_size
)
5019 while (toread
> 0) {
5020 tocopy
= sb
->s_blocksize
- offset
< toread
?
5021 sb
->s_blocksize
- offset
: toread
;
5022 bh
= ext4_bread(NULL
, inode
, blk
, 0, &err
);
5025 if (!bh
) /* A hole? */
5026 memset(data
, 0, tocopy
);
5028 memcpy(data
, bh
->b_data
+offset
, tocopy
);
5038 /* Write to quotafile (we know the transaction is already started and has
5039 * enough credits) */
5040 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
5041 const char *data
, size_t len
, loff_t off
)
5043 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5044 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5046 int offset
= off
& (sb
->s_blocksize
- 1);
5047 struct buffer_head
*bh
;
5048 handle_t
*handle
= journal_current_handle();
5050 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
5051 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5052 " cancelled because transaction is not started",
5053 (unsigned long long)off
, (unsigned long long)len
);
5057 * Since we account only one data block in transaction credits,
5058 * then it is impossible to cross a block boundary.
5060 if (sb
->s_blocksize
- offset
< len
) {
5061 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5062 " cancelled because not block aligned",
5063 (unsigned long long)off
, (unsigned long long)len
);
5067 bh
= ext4_bread(handle
, inode
, blk
, 1, &err
);
5070 err
= ext4_journal_get_write_access(handle
, bh
);
5076 memcpy(bh
->b_data
+offset
, data
, len
);
5077 flush_dcache_page(bh
->b_page
);
5079 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
5084 if (inode
->i_size
< off
+ len
) {
5085 i_size_write(inode
, off
+ len
);
5086 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
5087 ext4_mark_inode_dirty(handle
, inode
);
5094 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
5095 const char *dev_name
, void *data
)
5097 return mount_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
);
5100 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5101 static inline void register_as_ext2(void)
5103 int err
= register_filesystem(&ext2_fs_type
);
5106 "EXT4-fs: Unable to register as ext2 (%d)\n", err
);
5109 static inline void unregister_as_ext2(void)
5111 unregister_filesystem(&ext2_fs_type
);
5114 static inline int ext2_feature_set_ok(struct super_block
*sb
)
5116 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT2_FEATURE_INCOMPAT_SUPP
))
5118 if (sb
->s_flags
& MS_RDONLY
)
5120 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT2_FEATURE_RO_COMPAT_SUPP
))
5124 MODULE_ALIAS("ext2");
5126 static inline void register_as_ext2(void) { }
5127 static inline void unregister_as_ext2(void) { }
5128 static inline int ext2_feature_set_ok(struct super_block
*sb
) { return 0; }
5131 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5132 static inline void register_as_ext3(void)
5134 int err
= register_filesystem(&ext3_fs_type
);
5137 "EXT4-fs: Unable to register as ext3 (%d)\n", err
);
5140 static inline void unregister_as_ext3(void)
5142 unregister_filesystem(&ext3_fs_type
);
5145 static inline int ext3_feature_set_ok(struct super_block
*sb
)
5147 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT3_FEATURE_INCOMPAT_SUPP
))
5149 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
5151 if (sb
->s_flags
& MS_RDONLY
)
5153 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT3_FEATURE_RO_COMPAT_SUPP
))
5157 MODULE_ALIAS("ext3");
5159 static inline void register_as_ext3(void) { }
5160 static inline void unregister_as_ext3(void) { }
5161 static inline int ext3_feature_set_ok(struct super_block
*sb
) { return 0; }
5164 static struct file_system_type ext4_fs_type
= {
5165 .owner
= THIS_MODULE
,
5167 .mount
= ext4_mount
,
5168 .kill_sb
= kill_block_super
,
5169 .fs_flags
= FS_REQUIRES_DEV
,
5172 static int __init
ext4_init_feat_adverts(void)
5174 struct ext4_features
*ef
;
5177 ef
= kzalloc(sizeof(struct ext4_features
), GFP_KERNEL
);
5181 ef
->f_kobj
.kset
= ext4_kset
;
5182 init_completion(&ef
->f_kobj_unregister
);
5183 ret
= kobject_init_and_add(&ef
->f_kobj
, &ext4_feat_ktype
, NULL
,
5196 static void ext4_exit_feat_adverts(void)
5198 kobject_put(&ext4_feat
->f_kobj
);
5199 wait_for_completion(&ext4_feat
->f_kobj_unregister
);
5203 /* Shared across all ext4 file systems */
5204 wait_queue_head_t ext4__ioend_wq
[EXT4_WQ_HASH_SZ
];
5205 struct mutex ext4__aio_mutex
[EXT4_WQ_HASH_SZ
];
5207 static int __init
ext4_init_fs(void)
5211 ext4_li_info
= NULL
;
5212 mutex_init(&ext4_li_mtx
);
5214 /* Build-time check for flags consistency */
5215 ext4_check_flag_values();
5217 for (i
= 0; i
< EXT4_WQ_HASH_SZ
; i
++) {
5218 mutex_init(&ext4__aio_mutex
[i
]);
5219 init_waitqueue_head(&ext4__ioend_wq
[i
]);
5222 err
= ext4_init_es();
5226 err
= ext4_init_pageio();
5230 err
= ext4_init_system_zone();
5233 ext4_kset
= kset_create_and_add("ext4", NULL
, fs_kobj
);
5238 ext4_proc_root
= proc_mkdir("fs/ext4", NULL
);
5240 err
= ext4_init_feat_adverts();
5244 err
= ext4_init_mballoc();
5248 err
= ext4_init_xattr();
5251 err
= init_inodecache();
5256 err
= register_filesystem(&ext4_fs_type
);
5262 unregister_as_ext2();
5263 unregister_as_ext3();
5264 destroy_inodecache();
5268 ext4_exit_mballoc();
5270 ext4_exit_feat_adverts();
5273 remove_proc_entry("fs/ext4", NULL
);
5274 kset_unregister(ext4_kset
);
5276 ext4_exit_system_zone();
5285 static void __exit
ext4_exit_fs(void)
5287 ext4_destroy_lazyinit_thread();
5288 unregister_as_ext2();
5289 unregister_as_ext3();
5290 unregister_filesystem(&ext4_fs_type
);
5291 destroy_inodecache();
5293 ext4_exit_mballoc();
5294 ext4_exit_feat_adverts();
5295 remove_proc_entry("fs/ext4", NULL
);
5296 kset_unregister(ext4_kset
);
5297 ext4_exit_system_zone();
5301 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5302 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5303 MODULE_LICENSE("GPL");
5304 module_init(ext4_init_fs
)
5305 module_exit(ext4_exit_fs
)