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);
84 static int ext4_reserve_clusters(struct ext4_sb_info
*, ext4_fsblk_t
);
86 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
87 static struct file_system_type ext2_fs_type
= {
91 .kill_sb
= kill_block_super
,
92 .fs_flags
= FS_REQUIRES_DEV
,
94 MODULE_ALIAS_FS("ext2");
96 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
98 #define IS_EXT2_SB(sb) (0)
102 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
103 static struct file_system_type ext3_fs_type
= {
104 .owner
= THIS_MODULE
,
107 .kill_sb
= kill_block_super
,
108 .fs_flags
= FS_REQUIRES_DEV
,
110 MODULE_ALIAS_FS("ext3");
111 MODULE_ALIAS("ext3");
112 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
114 #define IS_EXT3_SB(sb) (0)
117 static int ext4_verify_csum_type(struct super_block
*sb
,
118 struct ext4_super_block
*es
)
120 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
,
121 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
))
124 return es
->s_checksum_type
== EXT4_CRC32C_CHKSUM
;
127 static __le32
ext4_superblock_csum(struct super_block
*sb
,
128 struct ext4_super_block
*es
)
130 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
131 int offset
= offsetof(struct ext4_super_block
, s_checksum
);
134 csum
= ext4_chksum(sbi
, ~0, (char *)es
, offset
);
136 return cpu_to_le32(csum
);
139 int ext4_superblock_csum_verify(struct super_block
*sb
,
140 struct ext4_super_block
*es
)
142 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
,
143 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
))
146 return es
->s_checksum
== ext4_superblock_csum(sb
, es
);
149 void ext4_superblock_csum_set(struct super_block
*sb
)
151 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
153 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
,
154 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
))
157 es
->s_checksum
= ext4_superblock_csum(sb
, es
);
160 void *ext4_kvmalloc(size_t size
, gfp_t flags
)
164 ret
= kmalloc(size
, flags
);
166 ret
= __vmalloc(size
, flags
, PAGE_KERNEL
);
170 void *ext4_kvzalloc(size_t size
, gfp_t flags
)
174 ret
= kzalloc(size
, flags
);
176 ret
= __vmalloc(size
, flags
| __GFP_ZERO
, PAGE_KERNEL
);
180 void ext4_kvfree(void *ptr
)
182 if (is_vmalloc_addr(ptr
))
189 ext4_fsblk_t
ext4_block_bitmap(struct super_block
*sb
,
190 struct ext4_group_desc
*bg
)
192 return le32_to_cpu(bg
->bg_block_bitmap_lo
) |
193 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
194 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_block_bitmap_hi
) << 32 : 0);
197 ext4_fsblk_t
ext4_inode_bitmap(struct super_block
*sb
,
198 struct ext4_group_desc
*bg
)
200 return le32_to_cpu(bg
->bg_inode_bitmap_lo
) |
201 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
202 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_bitmap_hi
) << 32 : 0);
205 ext4_fsblk_t
ext4_inode_table(struct super_block
*sb
,
206 struct ext4_group_desc
*bg
)
208 return le32_to_cpu(bg
->bg_inode_table_lo
) |
209 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
210 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_table_hi
) << 32 : 0);
213 __u32
ext4_free_group_clusters(struct super_block
*sb
,
214 struct ext4_group_desc
*bg
)
216 return le16_to_cpu(bg
->bg_free_blocks_count_lo
) |
217 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
218 (__u32
)le16_to_cpu(bg
->bg_free_blocks_count_hi
) << 16 : 0);
221 __u32
ext4_free_inodes_count(struct super_block
*sb
,
222 struct ext4_group_desc
*bg
)
224 return le16_to_cpu(bg
->bg_free_inodes_count_lo
) |
225 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
226 (__u32
)le16_to_cpu(bg
->bg_free_inodes_count_hi
) << 16 : 0);
229 __u32
ext4_used_dirs_count(struct super_block
*sb
,
230 struct ext4_group_desc
*bg
)
232 return le16_to_cpu(bg
->bg_used_dirs_count_lo
) |
233 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
234 (__u32
)le16_to_cpu(bg
->bg_used_dirs_count_hi
) << 16 : 0);
237 __u32
ext4_itable_unused_count(struct super_block
*sb
,
238 struct ext4_group_desc
*bg
)
240 return le16_to_cpu(bg
->bg_itable_unused_lo
) |
241 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
242 (__u32
)le16_to_cpu(bg
->bg_itable_unused_hi
) << 16 : 0);
245 void ext4_block_bitmap_set(struct super_block
*sb
,
246 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
248 bg
->bg_block_bitmap_lo
= cpu_to_le32((u32
)blk
);
249 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
250 bg
->bg_block_bitmap_hi
= cpu_to_le32(blk
>> 32);
253 void ext4_inode_bitmap_set(struct super_block
*sb
,
254 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
256 bg
->bg_inode_bitmap_lo
= cpu_to_le32((u32
)blk
);
257 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
258 bg
->bg_inode_bitmap_hi
= cpu_to_le32(blk
>> 32);
261 void ext4_inode_table_set(struct super_block
*sb
,
262 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
264 bg
->bg_inode_table_lo
= cpu_to_le32((u32
)blk
);
265 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
266 bg
->bg_inode_table_hi
= cpu_to_le32(blk
>> 32);
269 void ext4_free_group_clusters_set(struct super_block
*sb
,
270 struct ext4_group_desc
*bg
, __u32 count
)
272 bg
->bg_free_blocks_count_lo
= cpu_to_le16((__u16
)count
);
273 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
274 bg
->bg_free_blocks_count_hi
= cpu_to_le16(count
>> 16);
277 void ext4_free_inodes_set(struct super_block
*sb
,
278 struct ext4_group_desc
*bg
, __u32 count
)
280 bg
->bg_free_inodes_count_lo
= cpu_to_le16((__u16
)count
);
281 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
282 bg
->bg_free_inodes_count_hi
= cpu_to_le16(count
>> 16);
285 void ext4_used_dirs_set(struct super_block
*sb
,
286 struct ext4_group_desc
*bg
, __u32 count
)
288 bg
->bg_used_dirs_count_lo
= cpu_to_le16((__u16
)count
);
289 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
290 bg
->bg_used_dirs_count_hi
= cpu_to_le16(count
>> 16);
293 void ext4_itable_unused_set(struct super_block
*sb
,
294 struct ext4_group_desc
*bg
, __u32 count
)
296 bg
->bg_itable_unused_lo
= cpu_to_le16((__u16
)count
);
297 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
298 bg
->bg_itable_unused_hi
= cpu_to_le16(count
>> 16);
302 static void __save_error_info(struct super_block
*sb
, const char *func
,
305 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
307 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
308 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
309 es
->s_last_error_time
= cpu_to_le32(get_seconds());
310 strncpy(es
->s_last_error_func
, func
, sizeof(es
->s_last_error_func
));
311 es
->s_last_error_line
= cpu_to_le32(line
);
312 if (!es
->s_first_error_time
) {
313 es
->s_first_error_time
= es
->s_last_error_time
;
314 strncpy(es
->s_first_error_func
, func
,
315 sizeof(es
->s_first_error_func
));
316 es
->s_first_error_line
= cpu_to_le32(line
);
317 es
->s_first_error_ino
= es
->s_last_error_ino
;
318 es
->s_first_error_block
= es
->s_last_error_block
;
321 * Start the daily error reporting function if it hasn't been
324 if (!es
->s_error_count
)
325 mod_timer(&EXT4_SB(sb
)->s_err_report
, jiffies
+ 24*60*60*HZ
);
326 le32_add_cpu(&es
->s_error_count
, 1);
329 static void save_error_info(struct super_block
*sb
, const char *func
,
332 __save_error_info(sb
, func
, line
);
333 ext4_commit_super(sb
, 1);
337 * The del_gendisk() function uninitializes the disk-specific data
338 * structures, including the bdi structure, without telling anyone
339 * else. Once this happens, any attempt to call mark_buffer_dirty()
340 * (for example, by ext4_commit_super), will cause a kernel OOPS.
341 * This is a kludge to prevent these oops until we can put in a proper
342 * hook in del_gendisk() to inform the VFS and file system layers.
344 static int block_device_ejected(struct super_block
*sb
)
346 struct inode
*bd_inode
= sb
->s_bdev
->bd_inode
;
347 struct backing_dev_info
*bdi
= bd_inode
->i_mapping
->backing_dev_info
;
349 return bdi
->dev
== NULL
;
352 static void ext4_journal_commit_callback(journal_t
*journal
, transaction_t
*txn
)
354 struct super_block
*sb
= journal
->j_private
;
355 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
356 int error
= is_journal_aborted(journal
);
357 struct ext4_journal_cb_entry
*jce
;
359 BUG_ON(txn
->t_state
== T_FINISHED
);
360 spin_lock(&sbi
->s_md_lock
);
361 while (!list_empty(&txn
->t_private_list
)) {
362 jce
= list_entry(txn
->t_private_list
.next
,
363 struct ext4_journal_cb_entry
, jce_list
);
364 list_del_init(&jce
->jce_list
);
365 spin_unlock(&sbi
->s_md_lock
);
366 jce
->jce_func(sb
, jce
, error
);
367 spin_lock(&sbi
->s_md_lock
);
369 spin_unlock(&sbi
->s_md_lock
);
372 /* Deal with the reporting of failure conditions on a filesystem such as
373 * inconsistencies detected or read IO failures.
375 * On ext2, we can store the error state of the filesystem in the
376 * superblock. That is not possible on ext4, because we may have other
377 * write ordering constraints on the superblock which prevent us from
378 * writing it out straight away; and given that the journal is about to
379 * be aborted, we can't rely on the current, or future, transactions to
380 * write out the superblock safely.
382 * We'll just use the jbd2_journal_abort() error code to record an error in
383 * the journal instead. On recovery, the journal will complain about
384 * that error until we've noted it down and cleared it.
387 static void ext4_handle_error(struct super_block
*sb
)
389 if (sb
->s_flags
& MS_RDONLY
)
392 if (!test_opt(sb
, ERRORS_CONT
)) {
393 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
395 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
397 jbd2_journal_abort(journal
, -EIO
);
399 if (test_opt(sb
, ERRORS_RO
)) {
400 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
401 sb
->s_flags
|= MS_RDONLY
;
403 if (test_opt(sb
, ERRORS_PANIC
)) {
404 if (EXT4_SB(sb
)->s_journal
&&
405 !(EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_REC_ERR
))
407 panic("EXT4-fs (device %s): panic forced after error\n",
412 void __ext4_error(struct super_block
*sb
, const char *function
,
413 unsigned int line
, const char *fmt
, ...)
415 struct va_format vaf
;
421 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
422 sb
->s_id
, function
, line
, current
->comm
, &vaf
);
424 save_error_info(sb
, function
, line
);
426 ext4_handle_error(sb
);
429 void ext4_error_inode(struct inode
*inode
, const char *function
,
430 unsigned int line
, ext4_fsblk_t block
,
431 const char *fmt
, ...)
434 struct va_format vaf
;
435 struct ext4_super_block
*es
= EXT4_SB(inode
->i_sb
)->s_es
;
437 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
438 es
->s_last_error_block
= cpu_to_le64(block
);
439 save_error_info(inode
->i_sb
, function
, line
);
444 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
445 "inode #%lu: block %llu: comm %s: %pV\n",
446 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
447 block
, current
->comm
, &vaf
);
449 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
450 "inode #%lu: comm %s: %pV\n",
451 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
452 current
->comm
, &vaf
);
455 ext4_handle_error(inode
->i_sb
);
458 void ext4_error_file(struct file
*file
, const char *function
,
459 unsigned int line
, ext4_fsblk_t block
,
460 const char *fmt
, ...)
463 struct va_format vaf
;
464 struct ext4_super_block
*es
;
465 struct inode
*inode
= file_inode(file
);
466 char pathname
[80], *path
;
468 es
= EXT4_SB(inode
->i_sb
)->s_es
;
469 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
470 save_error_info(inode
->i_sb
, function
, line
);
471 path
= d_path(&(file
->f_path
), pathname
, sizeof(pathname
));
479 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
480 "block %llu: comm %s: path %s: %pV\n",
481 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
482 block
, current
->comm
, path
, &vaf
);
485 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
486 "comm %s: path %s: %pV\n",
487 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
488 current
->comm
, path
, &vaf
);
491 ext4_handle_error(inode
->i_sb
);
494 const char *ext4_decode_error(struct super_block
*sb
, int errno
,
501 errstr
= "IO failure";
504 errstr
= "Out of memory";
507 if (!sb
|| (EXT4_SB(sb
)->s_journal
&&
508 EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
))
509 errstr
= "Journal has aborted";
511 errstr
= "Readonly filesystem";
514 /* If the caller passed in an extra buffer for unknown
515 * errors, textualise them now. Else we just return
518 /* Check for truncated error codes... */
519 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
528 /* __ext4_std_error decodes expected errors from journaling functions
529 * automatically and invokes the appropriate error response. */
531 void __ext4_std_error(struct super_block
*sb
, const char *function
,
532 unsigned int line
, int errno
)
537 /* Special case: if the error is EROFS, and we're not already
538 * inside a transaction, then there's really no point in logging
540 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
541 (sb
->s_flags
& MS_RDONLY
))
544 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
545 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s:%d: %s\n",
546 sb
->s_id
, function
, line
, errstr
);
547 save_error_info(sb
, function
, line
);
549 ext4_handle_error(sb
);
553 * ext4_abort is a much stronger failure handler than ext4_error. The
554 * abort function may be used to deal with unrecoverable failures such
555 * as journal IO errors or ENOMEM at a critical moment in log management.
557 * We unconditionally force the filesystem into an ABORT|READONLY state,
558 * unless the error response on the fs has been set to panic in which
559 * case we take the easy way out and panic immediately.
562 void __ext4_abort(struct super_block
*sb
, const char *function
,
563 unsigned int line
, const char *fmt
, ...)
567 save_error_info(sb
, function
, line
);
569 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: ", sb
->s_id
,
575 if ((sb
->s_flags
& MS_RDONLY
) == 0) {
576 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
577 sb
->s_flags
|= MS_RDONLY
;
578 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
579 if (EXT4_SB(sb
)->s_journal
)
580 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
581 save_error_info(sb
, function
, line
);
583 if (test_opt(sb
, ERRORS_PANIC
)) {
584 if (EXT4_SB(sb
)->s_journal
&&
585 !(EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_REC_ERR
))
587 panic("EXT4-fs panic from previous error\n");
591 void ext4_msg(struct super_block
*sb
, const char *prefix
, const char *fmt
, ...)
593 struct va_format vaf
;
599 printk("%sEXT4-fs (%s): %pV\n", prefix
, sb
->s_id
, &vaf
);
603 void __ext4_warning(struct super_block
*sb
, const char *function
,
604 unsigned int line
, const char *fmt
, ...)
606 struct va_format vaf
;
612 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: %pV\n",
613 sb
->s_id
, function
, line
, &vaf
);
617 void __ext4_grp_locked_error(const char *function
, unsigned int line
,
618 struct super_block
*sb
, ext4_group_t grp
,
619 unsigned long ino
, ext4_fsblk_t block
,
620 const char *fmt
, ...)
624 struct va_format vaf
;
626 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
628 es
->s_last_error_ino
= cpu_to_le32(ino
);
629 es
->s_last_error_block
= cpu_to_le64(block
);
630 __save_error_info(sb
, function
, line
);
636 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: group %u, ",
637 sb
->s_id
, function
, line
, grp
);
639 printk(KERN_CONT
"inode %lu: ", ino
);
641 printk(KERN_CONT
"block %llu:", (unsigned long long) block
);
642 printk(KERN_CONT
"%pV\n", &vaf
);
645 if (test_opt(sb
, ERRORS_CONT
)) {
646 ext4_commit_super(sb
, 0);
650 ext4_unlock_group(sb
, grp
);
651 ext4_handle_error(sb
);
653 * We only get here in the ERRORS_RO case; relocking the group
654 * may be dangerous, but nothing bad will happen since the
655 * filesystem will have already been marked read/only and the
656 * journal has been aborted. We return 1 as a hint to callers
657 * who might what to use the return value from
658 * ext4_grp_locked_error() to distinguish between the
659 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
660 * aggressively from the ext4 function in question, with a
661 * more appropriate error code.
663 ext4_lock_group(sb
, grp
);
667 void ext4_update_dynamic_rev(struct super_block
*sb
)
669 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
671 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
675 "updating to rev %d because of new feature flag, "
676 "running e2fsck is recommended",
679 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
680 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
681 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
682 /* leave es->s_feature_*compat flags alone */
683 /* es->s_uuid will be set by e2fsck if empty */
686 * The rest of the superblock fields should be zero, and if not it
687 * means they are likely already in use, so leave them alone. We
688 * can leave it up to e2fsck to clean up any inconsistencies there.
693 * Open the external journal device
695 static struct block_device
*ext4_blkdev_get(dev_t dev
, struct super_block
*sb
)
697 struct block_device
*bdev
;
698 char b
[BDEVNAME_SIZE
];
700 bdev
= blkdev_get_by_dev(dev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
, sb
);
706 ext4_msg(sb
, KERN_ERR
, "failed to open journal device %s: %ld",
707 __bdevname(dev
, b
), PTR_ERR(bdev
));
712 * Release the journal device
714 static void ext4_blkdev_put(struct block_device
*bdev
)
716 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
);
719 static void ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
721 struct block_device
*bdev
;
722 bdev
= sbi
->journal_bdev
;
724 ext4_blkdev_put(bdev
);
725 sbi
->journal_bdev
= NULL
;
729 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
731 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
734 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
738 ext4_msg(sb
, KERN_ERR
, "sb orphan head is %d",
739 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
741 printk(KERN_ERR
"sb_info orphan list:\n");
742 list_for_each(l
, &sbi
->s_orphan
) {
743 struct inode
*inode
= orphan_list_entry(l
);
745 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
746 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
747 inode
->i_mode
, inode
->i_nlink
,
752 static void ext4_put_super(struct super_block
*sb
)
754 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
755 struct ext4_super_block
*es
= sbi
->s_es
;
758 ext4_unregister_li_request(sb
);
759 dquot_disable(sb
, -1, DQUOT_USAGE_ENABLED
| DQUOT_LIMITS_ENABLED
);
761 flush_workqueue(sbi
->dio_unwritten_wq
);
762 destroy_workqueue(sbi
->dio_unwritten_wq
);
764 if (sbi
->s_journal
) {
765 err
= jbd2_journal_destroy(sbi
->s_journal
);
766 sbi
->s_journal
= NULL
;
768 ext4_abort(sb
, "Couldn't clean up the journal");
771 ext4_es_unregister_shrinker(sb
);
772 del_timer(&sbi
->s_err_report
);
773 ext4_release_system_zone(sb
);
775 ext4_ext_release(sb
);
776 ext4_xattr_put_super(sb
);
778 if (!(sb
->s_flags
& MS_RDONLY
)) {
779 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
780 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
782 if (!(sb
->s_flags
& MS_RDONLY
))
783 ext4_commit_super(sb
, 1);
786 remove_proc_entry("options", sbi
->s_proc
);
787 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
789 kobject_del(&sbi
->s_kobj
);
791 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
792 brelse(sbi
->s_group_desc
[i
]);
793 ext4_kvfree(sbi
->s_group_desc
);
794 ext4_kvfree(sbi
->s_flex_groups
);
795 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
796 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
797 percpu_counter_destroy(&sbi
->s_dirs_counter
);
798 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
799 percpu_counter_destroy(&sbi
->s_extent_cache_cnt
);
802 for (i
= 0; i
< MAXQUOTAS
; i
++)
803 kfree(sbi
->s_qf_names
[i
]);
806 /* Debugging code just in case the in-memory inode orphan list
807 * isn't empty. The on-disk one can be non-empty if we've
808 * detected an error and taken the fs readonly, but the
809 * in-memory list had better be clean by this point. */
810 if (!list_empty(&sbi
->s_orphan
))
811 dump_orphan_list(sb
, sbi
);
812 J_ASSERT(list_empty(&sbi
->s_orphan
));
814 sync_blockdev(sb
->s_bdev
);
815 invalidate_bdev(sb
->s_bdev
);
816 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
818 * Invalidate the journal device's buffers. We don't want them
819 * floating about in memory - the physical journal device may
820 * hotswapped, and it breaks the `ro-after' testing code.
822 sync_blockdev(sbi
->journal_bdev
);
823 invalidate_bdev(sbi
->journal_bdev
);
824 ext4_blkdev_remove(sbi
);
827 kthread_stop(sbi
->s_mmp_tsk
);
828 sb
->s_fs_info
= NULL
;
830 * Now that we are completely done shutting down the
831 * superblock, we need to actually destroy the kobject.
833 kobject_put(&sbi
->s_kobj
);
834 wait_for_completion(&sbi
->s_kobj_unregister
);
835 if (sbi
->s_chksum_driver
)
836 crypto_free_shash(sbi
->s_chksum_driver
);
837 kfree(sbi
->s_blockgroup_lock
);
841 static struct kmem_cache
*ext4_inode_cachep
;
844 * Called inside transaction, so use GFP_NOFS
846 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
848 struct ext4_inode_info
*ei
;
850 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
854 ei
->vfs_inode
.i_version
= 1;
855 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
856 spin_lock_init(&ei
->i_prealloc_lock
);
857 ext4_es_init_tree(&ei
->i_es_tree
);
858 rwlock_init(&ei
->i_es_lock
);
859 INIT_LIST_HEAD(&ei
->i_es_lru
);
861 ei
->i_reserved_data_blocks
= 0;
862 ei
->i_reserved_meta_blocks
= 0;
863 ei
->i_allocated_meta_blocks
= 0;
864 ei
->i_da_metadata_calc_len
= 0;
865 ei
->i_da_metadata_calc_last_lblock
= 0;
866 spin_lock_init(&(ei
->i_block_reservation_lock
));
868 ei
->i_reserved_quota
= 0;
871 INIT_LIST_HEAD(&ei
->i_completed_io_list
);
872 spin_lock_init(&ei
->i_completed_io_lock
);
874 ei
->i_datasync_tid
= 0;
875 atomic_set(&ei
->i_ioend_count
, 0);
876 atomic_set(&ei
->i_unwritten
, 0);
877 INIT_WORK(&ei
->i_unwritten_work
, ext4_end_io_work
);
879 return &ei
->vfs_inode
;
882 static int ext4_drop_inode(struct inode
*inode
)
884 int drop
= generic_drop_inode(inode
);
886 trace_ext4_drop_inode(inode
, drop
);
890 static void ext4_i_callback(struct rcu_head
*head
)
892 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
893 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
896 static void ext4_destroy_inode(struct inode
*inode
)
898 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
899 ext4_msg(inode
->i_sb
, KERN_ERR
,
900 "Inode %lu (%p): orphan list check failed!",
901 inode
->i_ino
, EXT4_I(inode
));
902 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
903 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
907 call_rcu(&inode
->i_rcu
, ext4_i_callback
);
910 static void init_once(void *foo
)
912 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
914 INIT_LIST_HEAD(&ei
->i_orphan
);
915 init_rwsem(&ei
->xattr_sem
);
916 init_rwsem(&ei
->i_data_sem
);
917 inode_init_once(&ei
->vfs_inode
);
920 static int init_inodecache(void)
922 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
923 sizeof(struct ext4_inode_info
),
924 0, (SLAB_RECLAIM_ACCOUNT
|
927 if (ext4_inode_cachep
== NULL
)
932 static void destroy_inodecache(void)
935 * Make sure all delayed rcu free inodes are flushed before we
939 kmem_cache_destroy(ext4_inode_cachep
);
942 void ext4_clear_inode(struct inode
*inode
)
944 invalidate_inode_buffers(inode
);
947 ext4_discard_preallocations(inode
);
948 ext4_es_remove_extent(inode
, 0, EXT_MAX_BLOCKS
);
949 ext4_es_lru_del(inode
);
950 if (EXT4_I(inode
)->jinode
) {
951 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode
),
952 EXT4_I(inode
)->jinode
);
953 jbd2_free_inode(EXT4_I(inode
)->jinode
);
954 EXT4_I(inode
)->jinode
= NULL
;
958 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
959 u64 ino
, u32 generation
)
963 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
964 return ERR_PTR(-ESTALE
);
965 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
966 return ERR_PTR(-ESTALE
);
968 /* iget isn't really right if the inode is currently unallocated!!
970 * ext4_read_inode will return a bad_inode if the inode had been
971 * deleted, so we should be safe.
973 * Currently we don't know the generation for parent directory, so
974 * a generation of 0 means "accept any"
976 inode
= ext4_iget_normal(sb
, ino
);
978 return ERR_CAST(inode
);
979 if (generation
&& inode
->i_generation
!= generation
) {
981 return ERR_PTR(-ESTALE
);
987 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
988 int fh_len
, int fh_type
)
990 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
994 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
995 int fh_len
, int fh_type
)
997 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1002 * Try to release metadata pages (indirect blocks, directories) which are
1003 * mapped via the block device. Since these pages could have journal heads
1004 * which would prevent try_to_free_buffers() from freeing them, we must use
1005 * jbd2 layer's try_to_free_buffers() function to release them.
1007 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
1010 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
1012 WARN_ON(PageChecked(page
));
1013 if (!page_has_buffers(page
))
1016 return jbd2_journal_try_to_free_buffers(journal
, page
,
1017 wait
& ~__GFP_WAIT
);
1018 return try_to_free_buffers(page
);
1022 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1023 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1025 static int ext4_write_dquot(struct dquot
*dquot
);
1026 static int ext4_acquire_dquot(struct dquot
*dquot
);
1027 static int ext4_release_dquot(struct dquot
*dquot
);
1028 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
1029 static int ext4_write_info(struct super_block
*sb
, int type
);
1030 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
1032 static int ext4_quota_on_sysfile(struct super_block
*sb
, int type
,
1034 static int ext4_quota_off(struct super_block
*sb
, int type
);
1035 static int ext4_quota_off_sysfile(struct super_block
*sb
, int type
);
1036 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
1037 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
1038 size_t len
, loff_t off
);
1039 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
1040 const char *data
, size_t len
, loff_t off
);
1041 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
1042 unsigned int flags
);
1043 static int ext4_enable_quotas(struct super_block
*sb
);
1045 static const struct dquot_operations ext4_quota_operations
= {
1046 .get_reserved_space
= ext4_get_reserved_space
,
1047 .write_dquot
= ext4_write_dquot
,
1048 .acquire_dquot
= ext4_acquire_dquot
,
1049 .release_dquot
= ext4_release_dquot
,
1050 .mark_dirty
= ext4_mark_dquot_dirty
,
1051 .write_info
= ext4_write_info
,
1052 .alloc_dquot
= dquot_alloc
,
1053 .destroy_dquot
= dquot_destroy
,
1056 static const struct quotactl_ops ext4_qctl_operations
= {
1057 .quota_on
= ext4_quota_on
,
1058 .quota_off
= ext4_quota_off
,
1059 .quota_sync
= dquot_quota_sync
,
1060 .get_info
= dquot_get_dqinfo
,
1061 .set_info
= dquot_set_dqinfo
,
1062 .get_dqblk
= dquot_get_dqblk
,
1063 .set_dqblk
= dquot_set_dqblk
1066 static const struct quotactl_ops ext4_qctl_sysfile_operations
= {
1067 .quota_on_meta
= ext4_quota_on_sysfile
,
1068 .quota_off
= ext4_quota_off_sysfile
,
1069 .quota_sync
= dquot_quota_sync
,
1070 .get_info
= dquot_get_dqinfo
,
1071 .set_info
= dquot_set_dqinfo
,
1072 .get_dqblk
= dquot_get_dqblk
,
1073 .set_dqblk
= dquot_set_dqblk
1077 static const struct super_operations ext4_sops
= {
1078 .alloc_inode
= ext4_alloc_inode
,
1079 .destroy_inode
= ext4_destroy_inode
,
1080 .write_inode
= ext4_write_inode
,
1081 .dirty_inode
= ext4_dirty_inode
,
1082 .drop_inode
= ext4_drop_inode
,
1083 .evict_inode
= ext4_evict_inode
,
1084 .put_super
= ext4_put_super
,
1085 .sync_fs
= ext4_sync_fs
,
1086 .freeze_fs
= ext4_freeze
,
1087 .unfreeze_fs
= ext4_unfreeze
,
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 super_operations ext4_nojournal_sops
= {
1099 .alloc_inode
= ext4_alloc_inode
,
1100 .destroy_inode
= ext4_destroy_inode
,
1101 .write_inode
= ext4_write_inode
,
1102 .dirty_inode
= ext4_dirty_inode
,
1103 .drop_inode
= ext4_drop_inode
,
1104 .evict_inode
= ext4_evict_inode
,
1105 .put_super
= ext4_put_super
,
1106 .statfs
= ext4_statfs
,
1107 .remount_fs
= ext4_remount
,
1108 .show_options
= ext4_show_options
,
1110 .quota_read
= ext4_quota_read
,
1111 .quota_write
= ext4_quota_write
,
1113 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1116 static const struct export_operations ext4_export_ops
= {
1117 .fh_to_dentry
= ext4_fh_to_dentry
,
1118 .fh_to_parent
= ext4_fh_to_parent
,
1119 .get_parent
= ext4_get_parent
,
1123 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1124 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1125 Opt_nouid32
, Opt_debug
, Opt_removed
,
1126 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1127 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
,
1128 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
,
1129 Opt_journal_dev
, Opt_journal_checksum
, Opt_journal_async_commit
,
1130 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1131 Opt_data_err_abort
, Opt_data_err_ignore
,
1132 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1133 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
1134 Opt_noquota
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
1135 Opt_usrquota
, Opt_grpquota
, Opt_i_version
,
1136 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
, Opt_mblk_io_submit
,
1137 Opt_nomblk_io_submit
, Opt_block_validity
, Opt_noblock_validity
,
1138 Opt_inode_readahead_blks
, Opt_journal_ioprio
,
1139 Opt_dioread_nolock
, Opt_dioread_lock
,
1140 Opt_discard
, Opt_nodiscard
, Opt_init_itable
, Opt_noinit_itable
,
1141 Opt_max_dir_size_kb
,
1144 static const match_table_t tokens
= {
1145 {Opt_bsd_df
, "bsddf"},
1146 {Opt_minix_df
, "minixdf"},
1147 {Opt_grpid
, "grpid"},
1148 {Opt_grpid
, "bsdgroups"},
1149 {Opt_nogrpid
, "nogrpid"},
1150 {Opt_nogrpid
, "sysvgroups"},
1151 {Opt_resgid
, "resgid=%u"},
1152 {Opt_resuid
, "resuid=%u"},
1154 {Opt_err_cont
, "errors=continue"},
1155 {Opt_err_panic
, "errors=panic"},
1156 {Opt_err_ro
, "errors=remount-ro"},
1157 {Opt_nouid32
, "nouid32"},
1158 {Opt_debug
, "debug"},
1159 {Opt_removed
, "oldalloc"},
1160 {Opt_removed
, "orlov"},
1161 {Opt_user_xattr
, "user_xattr"},
1162 {Opt_nouser_xattr
, "nouser_xattr"},
1164 {Opt_noacl
, "noacl"},
1165 {Opt_noload
, "norecovery"},
1166 {Opt_noload
, "noload"},
1167 {Opt_removed
, "nobh"},
1168 {Opt_removed
, "bh"},
1169 {Opt_commit
, "commit=%u"},
1170 {Opt_min_batch_time
, "min_batch_time=%u"},
1171 {Opt_max_batch_time
, "max_batch_time=%u"},
1172 {Opt_journal_dev
, "journal_dev=%u"},
1173 {Opt_journal_checksum
, "journal_checksum"},
1174 {Opt_journal_async_commit
, "journal_async_commit"},
1175 {Opt_abort
, "abort"},
1176 {Opt_data_journal
, "data=journal"},
1177 {Opt_data_ordered
, "data=ordered"},
1178 {Opt_data_writeback
, "data=writeback"},
1179 {Opt_data_err_abort
, "data_err=abort"},
1180 {Opt_data_err_ignore
, "data_err=ignore"},
1181 {Opt_offusrjquota
, "usrjquota="},
1182 {Opt_usrjquota
, "usrjquota=%s"},
1183 {Opt_offgrpjquota
, "grpjquota="},
1184 {Opt_grpjquota
, "grpjquota=%s"},
1185 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1186 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1187 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
1188 {Opt_grpquota
, "grpquota"},
1189 {Opt_noquota
, "noquota"},
1190 {Opt_quota
, "quota"},
1191 {Opt_usrquota
, "usrquota"},
1192 {Opt_barrier
, "barrier=%u"},
1193 {Opt_barrier
, "barrier"},
1194 {Opt_nobarrier
, "nobarrier"},
1195 {Opt_i_version
, "i_version"},
1196 {Opt_stripe
, "stripe=%u"},
1197 {Opt_delalloc
, "delalloc"},
1198 {Opt_nodelalloc
, "nodelalloc"},
1199 {Opt_removed
, "mblk_io_submit"},
1200 {Opt_removed
, "nomblk_io_submit"},
1201 {Opt_block_validity
, "block_validity"},
1202 {Opt_noblock_validity
, "noblock_validity"},
1203 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1204 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1205 {Opt_auto_da_alloc
, "auto_da_alloc=%u"},
1206 {Opt_auto_da_alloc
, "auto_da_alloc"},
1207 {Opt_noauto_da_alloc
, "noauto_da_alloc"},
1208 {Opt_dioread_nolock
, "dioread_nolock"},
1209 {Opt_dioread_lock
, "dioread_lock"},
1210 {Opt_discard
, "discard"},
1211 {Opt_nodiscard
, "nodiscard"},
1212 {Opt_init_itable
, "init_itable=%u"},
1213 {Opt_init_itable
, "init_itable"},
1214 {Opt_noinit_itable
, "noinit_itable"},
1215 {Opt_max_dir_size_kb
, "max_dir_size_kb=%u"},
1216 {Opt_removed
, "check=none"}, /* mount option from ext2/3 */
1217 {Opt_removed
, "nocheck"}, /* mount option from ext2/3 */
1218 {Opt_removed
, "reservation"}, /* mount option from ext2/3 */
1219 {Opt_removed
, "noreservation"}, /* mount option from ext2/3 */
1220 {Opt_removed
, "journal=%u"}, /* mount option from ext2/3 */
1224 static ext4_fsblk_t
get_sb_block(void **data
)
1226 ext4_fsblk_t sb_block
;
1227 char *options
= (char *) *data
;
1229 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1230 return 1; /* Default location */
1233 /* TODO: use simple_strtoll with >32bit ext4 */
1234 sb_block
= simple_strtoul(options
, &options
, 0);
1235 if (*options
&& *options
!= ',') {
1236 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1240 if (*options
== ',')
1242 *data
= (void *) options
;
1247 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1248 static char deprecated_msg
[] = "Mount option \"%s\" will be removed by %s\n"
1249 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1252 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
1254 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1258 if (sb_any_quota_loaded(sb
) &&
1259 !sbi
->s_qf_names
[qtype
]) {
1260 ext4_msg(sb
, KERN_ERR
,
1261 "Cannot change journaled "
1262 "quota options when quota turned on");
1265 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
)) {
1266 ext4_msg(sb
, KERN_ERR
, "Cannot set journaled quota options "
1267 "when QUOTA feature is enabled");
1270 qname
= match_strdup(args
);
1272 ext4_msg(sb
, KERN_ERR
,
1273 "Not enough memory for storing quotafile name");
1276 if (sbi
->s_qf_names
[qtype
]) {
1277 if (strcmp(sbi
->s_qf_names
[qtype
], qname
) == 0)
1280 ext4_msg(sb
, KERN_ERR
,
1281 "%s quota file already specified",
1285 if (strchr(qname
, '/')) {
1286 ext4_msg(sb
, KERN_ERR
,
1287 "quotafile must be on filesystem root");
1290 sbi
->s_qf_names
[qtype
] = qname
;
1298 static int clear_qf_name(struct super_block
*sb
, int qtype
)
1301 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1303 if (sb_any_quota_loaded(sb
) &&
1304 sbi
->s_qf_names
[qtype
]) {
1305 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
1306 " when quota turned on");
1309 kfree(sbi
->s_qf_names
[qtype
]);
1310 sbi
->s_qf_names
[qtype
] = NULL
;
1315 #define MOPT_SET 0x0001
1316 #define MOPT_CLEAR 0x0002
1317 #define MOPT_NOSUPPORT 0x0004
1318 #define MOPT_EXPLICIT 0x0008
1319 #define MOPT_CLEAR_ERR 0x0010
1320 #define MOPT_GTE0 0x0020
1323 #define MOPT_QFMT 0x0040
1325 #define MOPT_Q MOPT_NOSUPPORT
1326 #define MOPT_QFMT MOPT_NOSUPPORT
1328 #define MOPT_DATAJ 0x0080
1329 #define MOPT_NO_EXT2 0x0100
1330 #define MOPT_NO_EXT3 0x0200
1331 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1333 static const struct mount_opts
{
1337 } ext4_mount_opts
[] = {
1338 {Opt_minix_df
, EXT4_MOUNT_MINIX_DF
, MOPT_SET
},
1339 {Opt_bsd_df
, EXT4_MOUNT_MINIX_DF
, MOPT_CLEAR
},
1340 {Opt_grpid
, EXT4_MOUNT_GRPID
, MOPT_SET
},
1341 {Opt_nogrpid
, EXT4_MOUNT_GRPID
, MOPT_CLEAR
},
1342 {Opt_block_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_SET
},
1343 {Opt_noblock_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_CLEAR
},
1344 {Opt_dioread_nolock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1345 MOPT_EXT4_ONLY
| MOPT_SET
},
1346 {Opt_dioread_lock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1347 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1348 {Opt_discard
, EXT4_MOUNT_DISCARD
, MOPT_SET
},
1349 {Opt_nodiscard
, EXT4_MOUNT_DISCARD
, MOPT_CLEAR
},
1350 {Opt_delalloc
, EXT4_MOUNT_DELALLOC
,
1351 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1352 {Opt_nodelalloc
, EXT4_MOUNT_DELALLOC
,
1353 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1354 {Opt_journal_checksum
, EXT4_MOUNT_JOURNAL_CHECKSUM
,
1355 MOPT_EXT4_ONLY
| MOPT_SET
},
1356 {Opt_journal_async_commit
, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT
|
1357 EXT4_MOUNT_JOURNAL_CHECKSUM
),
1358 MOPT_EXT4_ONLY
| MOPT_SET
},
1359 {Opt_noload
, EXT4_MOUNT_NOLOAD
, MOPT_NO_EXT2
| MOPT_SET
},
1360 {Opt_err_panic
, EXT4_MOUNT_ERRORS_PANIC
, MOPT_SET
| MOPT_CLEAR_ERR
},
1361 {Opt_err_ro
, EXT4_MOUNT_ERRORS_RO
, MOPT_SET
| MOPT_CLEAR_ERR
},
1362 {Opt_err_cont
, EXT4_MOUNT_ERRORS_CONT
, MOPT_SET
| MOPT_CLEAR_ERR
},
1363 {Opt_data_err_abort
, EXT4_MOUNT_DATA_ERR_ABORT
,
1364 MOPT_NO_EXT2
| MOPT_SET
},
1365 {Opt_data_err_ignore
, EXT4_MOUNT_DATA_ERR_ABORT
,
1366 MOPT_NO_EXT2
| MOPT_CLEAR
},
1367 {Opt_barrier
, EXT4_MOUNT_BARRIER
, MOPT_SET
},
1368 {Opt_nobarrier
, EXT4_MOUNT_BARRIER
, MOPT_CLEAR
},
1369 {Opt_noauto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_SET
},
1370 {Opt_auto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_CLEAR
},
1371 {Opt_noinit_itable
, EXT4_MOUNT_INIT_INODE_TABLE
, MOPT_CLEAR
},
1372 {Opt_commit
, 0, MOPT_GTE0
},
1373 {Opt_max_batch_time
, 0, MOPT_GTE0
},
1374 {Opt_min_batch_time
, 0, MOPT_GTE0
},
1375 {Opt_inode_readahead_blks
, 0, MOPT_GTE0
},
1376 {Opt_init_itable
, 0, MOPT_GTE0
},
1377 {Opt_stripe
, 0, MOPT_GTE0
},
1378 {Opt_resuid
, 0, MOPT_GTE0
},
1379 {Opt_resgid
, 0, MOPT_GTE0
},
1380 {Opt_journal_dev
, 0, MOPT_GTE0
},
1381 {Opt_journal_ioprio
, 0, MOPT_GTE0
},
1382 {Opt_data_journal
, EXT4_MOUNT_JOURNAL_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1383 {Opt_data_ordered
, EXT4_MOUNT_ORDERED_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1384 {Opt_data_writeback
, EXT4_MOUNT_WRITEBACK_DATA
,
1385 MOPT_NO_EXT2
| MOPT_DATAJ
},
1386 {Opt_user_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_SET
},
1387 {Opt_nouser_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_CLEAR
},
1388 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1389 {Opt_acl
, EXT4_MOUNT_POSIX_ACL
, MOPT_SET
},
1390 {Opt_noacl
, EXT4_MOUNT_POSIX_ACL
, MOPT_CLEAR
},
1392 {Opt_acl
, 0, MOPT_NOSUPPORT
},
1393 {Opt_noacl
, 0, MOPT_NOSUPPORT
},
1395 {Opt_nouid32
, EXT4_MOUNT_NO_UID32
, MOPT_SET
},
1396 {Opt_debug
, EXT4_MOUNT_DEBUG
, MOPT_SET
},
1397 {Opt_quota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
, MOPT_SET
| MOPT_Q
},
1398 {Opt_usrquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
,
1400 {Opt_grpquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_GRPQUOTA
,
1402 {Opt_noquota
, (EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
|
1403 EXT4_MOUNT_GRPQUOTA
), MOPT_CLEAR
| MOPT_Q
},
1404 {Opt_usrjquota
, 0, MOPT_Q
},
1405 {Opt_grpjquota
, 0, MOPT_Q
},
1406 {Opt_offusrjquota
, 0, MOPT_Q
},
1407 {Opt_offgrpjquota
, 0, MOPT_Q
},
1408 {Opt_jqfmt_vfsold
, QFMT_VFS_OLD
, MOPT_QFMT
},
1409 {Opt_jqfmt_vfsv0
, QFMT_VFS_V0
, MOPT_QFMT
},
1410 {Opt_jqfmt_vfsv1
, QFMT_VFS_V1
, MOPT_QFMT
},
1411 {Opt_max_dir_size_kb
, 0, MOPT_GTE0
},
1415 static int handle_mount_opt(struct super_block
*sb
, char *opt
, int token
,
1416 substring_t
*args
, unsigned long *journal_devnum
,
1417 unsigned int *journal_ioprio
, int is_remount
)
1419 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1420 const struct mount_opts
*m
;
1426 if (token
== Opt_usrjquota
)
1427 return set_qf_name(sb
, USRQUOTA
, &args
[0]);
1428 else if (token
== Opt_grpjquota
)
1429 return set_qf_name(sb
, GRPQUOTA
, &args
[0]);
1430 else if (token
== Opt_offusrjquota
)
1431 return clear_qf_name(sb
, USRQUOTA
);
1432 else if (token
== Opt_offgrpjquota
)
1433 return clear_qf_name(sb
, GRPQUOTA
);
1437 case Opt_nouser_xattr
:
1438 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, opt
, "3.5");
1441 return 1; /* handled by get_sb_block() */
1443 ext4_msg(sb
, KERN_WARNING
, "Ignoring removed %s option", opt
);
1446 sbi
->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
1449 sb
->s_flags
|= MS_I_VERSION
;
1453 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++)
1454 if (token
== m
->token
)
1457 if (m
->token
== Opt_err
) {
1458 ext4_msg(sb
, KERN_ERR
, "Unrecognized mount option \"%s\" "
1459 "or missing value", opt
);
1463 if ((m
->flags
& MOPT_NO_EXT2
) && IS_EXT2_SB(sb
)) {
1464 ext4_msg(sb
, KERN_ERR
,
1465 "Mount option \"%s\" incompatible with ext2", opt
);
1468 if ((m
->flags
& MOPT_NO_EXT3
) && IS_EXT3_SB(sb
)) {
1469 ext4_msg(sb
, KERN_ERR
,
1470 "Mount option \"%s\" incompatible with ext3", opt
);
1474 if (args
->from
&& match_int(args
, &arg
))
1476 if (args
->from
&& (m
->flags
& MOPT_GTE0
) && (arg
< 0))
1478 if (m
->flags
& MOPT_EXPLICIT
)
1479 set_opt2(sb
, EXPLICIT_DELALLOC
);
1480 if (m
->flags
& MOPT_CLEAR_ERR
)
1481 clear_opt(sb
, ERRORS_MASK
);
1482 if (token
== Opt_noquota
&& sb_any_quota_loaded(sb
)) {
1483 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
1484 "options when quota turned on");
1488 if (m
->flags
& MOPT_NOSUPPORT
) {
1489 ext4_msg(sb
, KERN_ERR
, "%s option not supported", opt
);
1490 } else if (token
== Opt_commit
) {
1492 arg
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1493 sbi
->s_commit_interval
= HZ
* arg
;
1494 } else if (token
== Opt_max_batch_time
) {
1495 sbi
->s_max_batch_time
= arg
;
1496 } else if (token
== Opt_min_batch_time
) {
1497 sbi
->s_min_batch_time
= arg
;
1498 } else if (token
== Opt_inode_readahead_blks
) {
1499 if (arg
&& (arg
> (1 << 30) || !is_power_of_2(arg
))) {
1500 ext4_msg(sb
, KERN_ERR
,
1501 "EXT4-fs: inode_readahead_blks must be "
1502 "0 or a power of 2 smaller than 2^31");
1505 sbi
->s_inode_readahead_blks
= arg
;
1506 } else if (token
== Opt_init_itable
) {
1507 set_opt(sb
, INIT_INODE_TABLE
);
1509 arg
= EXT4_DEF_LI_WAIT_MULT
;
1510 sbi
->s_li_wait_mult
= arg
;
1511 } else if (token
== Opt_max_dir_size_kb
) {
1512 sbi
->s_max_dir_size_kb
= arg
;
1513 } else if (token
== Opt_stripe
) {
1514 sbi
->s_stripe
= arg
;
1515 } else if (token
== Opt_resuid
) {
1516 uid
= make_kuid(current_user_ns(), arg
);
1517 if (!uid_valid(uid
)) {
1518 ext4_msg(sb
, KERN_ERR
, "Invalid uid value %d", arg
);
1521 sbi
->s_resuid
= uid
;
1522 } else if (token
== Opt_resgid
) {
1523 gid
= make_kgid(current_user_ns(), arg
);
1524 if (!gid_valid(gid
)) {
1525 ext4_msg(sb
, KERN_ERR
, "Invalid gid value %d", arg
);
1528 sbi
->s_resgid
= gid
;
1529 } else if (token
== Opt_journal_dev
) {
1531 ext4_msg(sb
, KERN_ERR
,
1532 "Cannot specify journal on remount");
1535 *journal_devnum
= arg
;
1536 } else if (token
== Opt_journal_ioprio
) {
1538 ext4_msg(sb
, KERN_ERR
, "Invalid journal IO priority"
1543 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
, arg
);
1544 } else if (m
->flags
& MOPT_DATAJ
) {
1546 if (!sbi
->s_journal
)
1547 ext4_msg(sb
, KERN_WARNING
, "Remounting file system with no journal so ignoring journalled data option");
1548 else if (test_opt(sb
, DATA_FLAGS
) != m
->mount_opt
) {
1549 ext4_msg(sb
, KERN_ERR
,
1550 "Cannot change data mode on remount");
1554 clear_opt(sb
, DATA_FLAGS
);
1555 sbi
->s_mount_opt
|= m
->mount_opt
;
1558 } else if (m
->flags
& MOPT_QFMT
) {
1559 if (sb_any_quota_loaded(sb
) &&
1560 sbi
->s_jquota_fmt
!= m
->mount_opt
) {
1561 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled "
1562 "quota options when quota turned on");
1565 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
1566 EXT4_FEATURE_RO_COMPAT_QUOTA
)) {
1567 ext4_msg(sb
, KERN_ERR
,
1568 "Cannot set journaled quota options "
1569 "when QUOTA feature is enabled");
1572 sbi
->s_jquota_fmt
= m
->mount_opt
;
1577 if (m
->flags
& MOPT_CLEAR
)
1579 else if (unlikely(!(m
->flags
& MOPT_SET
))) {
1580 ext4_msg(sb
, KERN_WARNING
,
1581 "buggy handling of option %s", opt
);
1586 sbi
->s_mount_opt
|= m
->mount_opt
;
1588 sbi
->s_mount_opt
&= ~m
->mount_opt
;
1593 static int parse_options(char *options
, struct super_block
*sb
,
1594 unsigned long *journal_devnum
,
1595 unsigned int *journal_ioprio
,
1598 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1600 substring_t args
[MAX_OPT_ARGS
];
1606 while ((p
= strsep(&options
, ",")) != NULL
) {
1610 * Initialize args struct so we know whether arg was
1611 * found; some options take optional arguments.
1613 args
[0].to
= args
[0].from
= NULL
;
1614 token
= match_token(p
, tokens
, args
);
1615 if (handle_mount_opt(sb
, p
, token
, args
, journal_devnum
,
1616 journal_ioprio
, is_remount
) < 0)
1620 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
) &&
1621 (test_opt(sb
, USRQUOTA
) || test_opt(sb
, GRPQUOTA
))) {
1622 ext4_msg(sb
, KERN_ERR
, "Cannot set quota options when QUOTA "
1623 "feature is enabled");
1626 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1627 if (test_opt(sb
, USRQUOTA
) && sbi
->s_qf_names
[USRQUOTA
])
1628 clear_opt(sb
, USRQUOTA
);
1630 if (test_opt(sb
, GRPQUOTA
) && sbi
->s_qf_names
[GRPQUOTA
])
1631 clear_opt(sb
, GRPQUOTA
);
1633 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
1634 ext4_msg(sb
, KERN_ERR
, "old and new quota "
1639 if (!sbi
->s_jquota_fmt
) {
1640 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1646 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
1648 BLOCK_SIZE
<< le32_to_cpu(sbi
->s_es
->s_log_block_size
);
1650 if (blocksize
< PAGE_CACHE_SIZE
) {
1651 ext4_msg(sb
, KERN_ERR
, "can't mount with "
1652 "dioread_nolock if block size != PAGE_SIZE");
1659 static inline void ext4_show_quota_options(struct seq_file
*seq
,
1660 struct super_block
*sb
)
1662 #if defined(CONFIG_QUOTA)
1663 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1665 if (sbi
->s_jquota_fmt
) {
1668 switch (sbi
->s_jquota_fmt
) {
1679 seq_printf(seq
, ",jqfmt=%s", fmtname
);
1682 if (sbi
->s_qf_names
[USRQUOTA
])
1683 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
1685 if (sbi
->s_qf_names
[GRPQUOTA
])
1686 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
1690 static const char *token2str(int token
)
1692 const struct match_token
*t
;
1694 for (t
= tokens
; t
->token
!= Opt_err
; t
++)
1695 if (t
->token
== token
&& !strchr(t
->pattern
, '='))
1702 * - it's set to a non-default value OR
1703 * - if the per-sb default is different from the global default
1705 static int _ext4_show_options(struct seq_file
*seq
, struct super_block
*sb
,
1708 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1709 struct ext4_super_block
*es
= sbi
->s_es
;
1710 int def_errors
, def_mount_opt
= nodefs
? 0 : sbi
->s_def_mount_opt
;
1711 const struct mount_opts
*m
;
1712 char sep
= nodefs
? '\n' : ',';
1714 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1715 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1717 if (sbi
->s_sb_block
!= 1)
1718 SEQ_OPTS_PRINT("sb=%llu", sbi
->s_sb_block
);
1720 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++) {
1721 int want_set
= m
->flags
& MOPT_SET
;
1722 if (((m
->flags
& (MOPT_SET
|MOPT_CLEAR
)) == 0) ||
1723 (m
->flags
& MOPT_CLEAR_ERR
))
1725 if (!(m
->mount_opt
& (sbi
->s_mount_opt
^ def_mount_opt
)))
1726 continue; /* skip if same as the default */
1728 (sbi
->s_mount_opt
& m
->mount_opt
) != m
->mount_opt
) ||
1729 (!want_set
&& (sbi
->s_mount_opt
& m
->mount_opt
)))
1730 continue; /* select Opt_noFoo vs Opt_Foo */
1731 SEQ_OPTS_PRINT("%s", token2str(m
->token
));
1734 if (nodefs
|| !uid_eq(sbi
->s_resuid
, make_kuid(&init_user_ns
, EXT4_DEF_RESUID
)) ||
1735 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
)
1736 SEQ_OPTS_PRINT("resuid=%u",
1737 from_kuid_munged(&init_user_ns
, sbi
->s_resuid
));
1738 if (nodefs
|| !gid_eq(sbi
->s_resgid
, make_kgid(&init_user_ns
, EXT4_DEF_RESGID
)) ||
1739 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
)
1740 SEQ_OPTS_PRINT("resgid=%u",
1741 from_kgid_munged(&init_user_ns
, sbi
->s_resgid
));
1742 def_errors
= nodefs
? -1 : le16_to_cpu(es
->s_errors
);
1743 if (test_opt(sb
, ERRORS_RO
) && def_errors
!= EXT4_ERRORS_RO
)
1744 SEQ_OPTS_PUTS("errors=remount-ro");
1745 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
1746 SEQ_OPTS_PUTS("errors=continue");
1747 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
1748 SEQ_OPTS_PUTS("errors=panic");
1749 if (nodefs
|| sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
)
1750 SEQ_OPTS_PRINT("commit=%lu", sbi
->s_commit_interval
/ HZ
);
1751 if (nodefs
|| sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
)
1752 SEQ_OPTS_PRINT("min_batch_time=%u", sbi
->s_min_batch_time
);
1753 if (nodefs
|| sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
)
1754 SEQ_OPTS_PRINT("max_batch_time=%u", sbi
->s_max_batch_time
);
1755 if (sb
->s_flags
& MS_I_VERSION
)
1756 SEQ_OPTS_PUTS("i_version");
1757 if (nodefs
|| sbi
->s_stripe
)
1758 SEQ_OPTS_PRINT("stripe=%lu", sbi
->s_stripe
);
1759 if (EXT4_MOUNT_DATA_FLAGS
& (sbi
->s_mount_opt
^ def_mount_opt
)) {
1760 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
1761 SEQ_OPTS_PUTS("data=journal");
1762 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
1763 SEQ_OPTS_PUTS("data=ordered");
1764 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
1765 SEQ_OPTS_PUTS("data=writeback");
1768 sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
1769 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1770 sbi
->s_inode_readahead_blks
);
1772 if (nodefs
|| (test_opt(sb
, INIT_INODE_TABLE
) &&
1773 (sbi
->s_li_wait_mult
!= EXT4_DEF_LI_WAIT_MULT
)))
1774 SEQ_OPTS_PRINT("init_itable=%u", sbi
->s_li_wait_mult
);
1775 if (nodefs
|| sbi
->s_max_dir_size_kb
)
1776 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi
->s_max_dir_size_kb
);
1778 ext4_show_quota_options(seq
, sb
);
1782 static int ext4_show_options(struct seq_file
*seq
, struct dentry
*root
)
1784 return _ext4_show_options(seq
, root
->d_sb
, 0);
1787 static int options_seq_show(struct seq_file
*seq
, void *offset
)
1789 struct super_block
*sb
= seq
->private;
1792 seq_puts(seq
, (sb
->s_flags
& MS_RDONLY
) ? "ro" : "rw");
1793 rc
= _ext4_show_options(seq
, sb
, 1);
1794 seq_puts(seq
, "\n");
1798 static int options_open_fs(struct inode
*inode
, struct file
*file
)
1800 return single_open(file
, options_seq_show
, PDE_DATA(inode
));
1803 static const struct file_operations ext4_seq_options_fops
= {
1804 .owner
= THIS_MODULE
,
1805 .open
= options_open_fs
,
1807 .llseek
= seq_lseek
,
1808 .release
= single_release
,
1811 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
1814 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1817 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
1818 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
1819 "forcing read-only mode");
1824 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
1825 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
1826 "running e2fsck is recommended");
1827 else if ((sbi
->s_mount_state
& EXT4_ERROR_FS
))
1828 ext4_msg(sb
, KERN_WARNING
,
1829 "warning: mounting fs with errors, "
1830 "running e2fsck is recommended");
1831 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) > 0 &&
1832 le16_to_cpu(es
->s_mnt_count
) >=
1833 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1834 ext4_msg(sb
, KERN_WARNING
,
1835 "warning: maximal mount count reached, "
1836 "running e2fsck is recommended");
1837 else if (le32_to_cpu(es
->s_checkinterval
) &&
1838 (le32_to_cpu(es
->s_lastcheck
) +
1839 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1840 ext4_msg(sb
, KERN_WARNING
,
1841 "warning: checktime reached, "
1842 "running e2fsck is recommended");
1843 if (!sbi
->s_journal
)
1844 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
1845 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1846 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
1847 le16_add_cpu(&es
->s_mnt_count
, 1);
1848 es
->s_mtime
= cpu_to_le32(get_seconds());
1849 ext4_update_dynamic_rev(sb
);
1851 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
1853 ext4_commit_super(sb
, 1);
1855 if (test_opt(sb
, DEBUG
))
1856 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
1857 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1859 sbi
->s_groups_count
,
1860 EXT4_BLOCKS_PER_GROUP(sb
),
1861 EXT4_INODES_PER_GROUP(sb
),
1862 sbi
->s_mount_opt
, sbi
->s_mount_opt2
);
1864 cleancache_init_fs(sb
);
1868 int ext4_alloc_flex_bg_array(struct super_block
*sb
, ext4_group_t ngroup
)
1870 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1871 struct flex_groups
*new_groups
;
1874 if (!sbi
->s_log_groups_per_flex
)
1877 size
= ext4_flex_group(sbi
, ngroup
- 1) + 1;
1878 if (size
<= sbi
->s_flex_groups_allocated
)
1881 size
= roundup_pow_of_two(size
* sizeof(struct flex_groups
));
1882 new_groups
= ext4_kvzalloc(size
, GFP_KERNEL
);
1884 ext4_msg(sb
, KERN_ERR
, "not enough memory for %d flex groups",
1885 size
/ (int) sizeof(struct flex_groups
));
1889 if (sbi
->s_flex_groups
) {
1890 memcpy(new_groups
, sbi
->s_flex_groups
,
1891 (sbi
->s_flex_groups_allocated
*
1892 sizeof(struct flex_groups
)));
1893 ext4_kvfree(sbi
->s_flex_groups
);
1895 sbi
->s_flex_groups
= new_groups
;
1896 sbi
->s_flex_groups_allocated
= size
/ sizeof(struct flex_groups
);
1900 static int ext4_fill_flex_info(struct super_block
*sb
)
1902 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1903 struct ext4_group_desc
*gdp
= NULL
;
1904 ext4_group_t flex_group
;
1905 unsigned int groups_per_flex
= 0;
1908 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
1909 if (sbi
->s_log_groups_per_flex
< 1 || sbi
->s_log_groups_per_flex
> 31) {
1910 sbi
->s_log_groups_per_flex
= 0;
1913 groups_per_flex
= 1U << sbi
->s_log_groups_per_flex
;
1915 err
= ext4_alloc_flex_bg_array(sb
, sbi
->s_groups_count
);
1919 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1920 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1922 flex_group
= ext4_flex_group(sbi
, i
);
1923 atomic_add(ext4_free_inodes_count(sb
, gdp
),
1924 &sbi
->s_flex_groups
[flex_group
].free_inodes
);
1925 atomic64_add(ext4_free_group_clusters(sb
, gdp
),
1926 &sbi
->s_flex_groups
[flex_group
].free_clusters
);
1927 atomic_add(ext4_used_dirs_count(sb
, gdp
),
1928 &sbi
->s_flex_groups
[flex_group
].used_dirs
);
1936 static __le16
ext4_group_desc_csum(struct ext4_sb_info
*sbi
, __u32 block_group
,
1937 struct ext4_group_desc
*gdp
)
1941 __le32 le_group
= cpu_to_le32(block_group
);
1943 if ((sbi
->s_es
->s_feature_ro_compat
&
1944 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
))) {
1945 /* Use new metadata_csum algorithm */
1949 save_csum
= gdp
->bg_checksum
;
1950 gdp
->bg_checksum
= 0;
1951 csum32
= ext4_chksum(sbi
, sbi
->s_csum_seed
, (__u8
*)&le_group
,
1953 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)gdp
,
1955 gdp
->bg_checksum
= save_csum
;
1957 crc
= csum32
& 0xFFFF;
1961 /* old crc16 code */
1962 if (!(sbi
->s_es
->s_feature_ro_compat
&
1963 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)))
1966 offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
1968 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
1969 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
1970 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
1971 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
1972 /* for checksum of struct ext4_group_desc do the rest...*/
1973 if ((sbi
->s_es
->s_feature_incompat
&
1974 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT
)) &&
1975 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
1976 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
1977 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
1981 return cpu_to_le16(crc
);
1984 int ext4_group_desc_csum_verify(struct super_block
*sb
, __u32 block_group
,
1985 struct ext4_group_desc
*gdp
)
1987 if (ext4_has_group_desc_csum(sb
) &&
1988 (gdp
->bg_checksum
!= ext4_group_desc_csum(EXT4_SB(sb
),
1995 void ext4_group_desc_csum_set(struct super_block
*sb
, __u32 block_group
,
1996 struct ext4_group_desc
*gdp
)
1998 if (!ext4_has_group_desc_csum(sb
))
2000 gdp
->bg_checksum
= ext4_group_desc_csum(EXT4_SB(sb
), block_group
, gdp
);
2003 /* Called at mount-time, super-block is locked */
2004 static int ext4_check_descriptors(struct super_block
*sb
,
2005 ext4_group_t
*first_not_zeroed
)
2007 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2008 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
2009 ext4_fsblk_t last_block
;
2010 ext4_fsblk_t block_bitmap
;
2011 ext4_fsblk_t inode_bitmap
;
2012 ext4_fsblk_t inode_table
;
2013 int flexbg_flag
= 0;
2014 ext4_group_t i
, grp
= sbi
->s_groups_count
;
2016 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
2019 ext4_debug("Checking group descriptors");
2021 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2022 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2024 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
2025 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
2027 last_block
= first_block
+
2028 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2030 if ((grp
== sbi
->s_groups_count
) &&
2031 !(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2034 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
2035 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
2036 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2037 "Block bitmap for group %u not in group "
2038 "(block %llu)!", i
, block_bitmap
);
2041 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
2042 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
2043 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2044 "Inode bitmap for group %u not in group "
2045 "(block %llu)!", i
, inode_bitmap
);
2048 inode_table
= ext4_inode_table(sb
, gdp
);
2049 if (inode_table
< first_block
||
2050 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
2051 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2052 "Inode table for group %u not in group "
2053 "(block %llu)!", i
, inode_table
);
2056 ext4_lock_group(sb
, i
);
2057 if (!ext4_group_desc_csum_verify(sb
, i
, gdp
)) {
2058 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2059 "Checksum for group %u failed (%u!=%u)",
2060 i
, le16_to_cpu(ext4_group_desc_csum(sbi
, i
,
2061 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
2062 if (!(sb
->s_flags
& MS_RDONLY
)) {
2063 ext4_unlock_group(sb
, i
);
2067 ext4_unlock_group(sb
, i
);
2069 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
2071 if (NULL
!= first_not_zeroed
)
2072 *first_not_zeroed
= grp
;
2074 ext4_free_blocks_count_set(sbi
->s_es
,
2075 EXT4_C2B(sbi
, ext4_count_free_clusters(sb
)));
2076 sbi
->s_es
->s_free_inodes_count
=cpu_to_le32(ext4_count_free_inodes(sb
));
2080 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2081 * the superblock) which were deleted from all directories, but held open by
2082 * a process at the time of a crash. We walk the list and try to delete these
2083 * inodes at recovery time (only with a read-write filesystem).
2085 * In order to keep the orphan inode chain consistent during traversal (in
2086 * case of crash during recovery), we link each inode into the superblock
2087 * orphan list_head and handle it the same way as an inode deletion during
2088 * normal operation (which journals the operations for us).
2090 * We only do an iget() and an iput() on each inode, which is very safe if we
2091 * accidentally point at an in-use or already deleted inode. The worst that
2092 * can happen in this case is that we get a "bit already cleared" message from
2093 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2094 * e2fsck was run on this filesystem, and it must have already done the orphan
2095 * inode cleanup for us, so we can safely abort without any further action.
2097 static void ext4_orphan_cleanup(struct super_block
*sb
,
2098 struct ext4_super_block
*es
)
2100 unsigned int s_flags
= sb
->s_flags
;
2101 int nr_orphans
= 0, nr_truncates
= 0;
2105 if (!es
->s_last_orphan
) {
2106 jbd_debug(4, "no orphan inodes to clean up\n");
2110 if (bdev_read_only(sb
->s_bdev
)) {
2111 ext4_msg(sb
, KERN_ERR
, "write access "
2112 "unavailable, skipping orphan cleanup");
2116 /* Check if feature set would not allow a r/w mount */
2117 if (!ext4_feature_set_ok(sb
, 0)) {
2118 ext4_msg(sb
, KERN_INFO
, "Skipping orphan cleanup due to "
2119 "unknown ROCOMPAT features");
2123 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2124 /* don't clear list on RO mount w/ errors */
2125 if (es
->s_last_orphan
&& !(s_flags
& MS_RDONLY
)) {
2126 jbd_debug(1, "Errors on filesystem, "
2127 "clearing orphan list.\n");
2128 es
->s_last_orphan
= 0;
2130 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2134 if (s_flags
& MS_RDONLY
) {
2135 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
2136 sb
->s_flags
&= ~MS_RDONLY
;
2139 /* Needed for iput() to work correctly and not trash data */
2140 sb
->s_flags
|= MS_ACTIVE
;
2141 /* Turn on quotas so that they are updated correctly */
2142 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2143 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
2144 int ret
= ext4_quota_on_mount(sb
, i
);
2146 ext4_msg(sb
, KERN_ERR
,
2147 "Cannot turn on journaled "
2148 "quota: error %d", ret
);
2153 while (es
->s_last_orphan
) {
2154 struct inode
*inode
;
2157 * We may have encountered an error during cleanup; if
2158 * so, skip the rest.
2160 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2161 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2162 es
->s_last_orphan
= 0;
2166 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
2167 if (IS_ERR(inode
)) {
2168 es
->s_last_orphan
= 0;
2172 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
2173 dquot_initialize(inode
);
2174 if (inode
->i_nlink
) {
2175 ext4_msg(sb
, KERN_DEBUG
,
2176 "%s: truncating inode %lu to %lld bytes",
2177 __func__
, inode
->i_ino
, inode
->i_size
);
2178 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2179 inode
->i_ino
, inode
->i_size
);
2180 mutex_lock(&inode
->i_mutex
);
2181 ext4_truncate(inode
);
2182 mutex_unlock(&inode
->i_mutex
);
2185 ext4_msg(sb
, KERN_DEBUG
,
2186 "%s: deleting unreferenced inode %lu",
2187 __func__
, inode
->i_ino
);
2188 jbd_debug(2, "deleting unreferenced inode %lu\n",
2192 iput(inode
); /* The delete magic happens here! */
2195 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2198 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
2199 PLURAL(nr_orphans
));
2201 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
2202 PLURAL(nr_truncates
));
2204 /* Turn quotas off */
2205 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2206 if (sb_dqopt(sb
)->files
[i
])
2207 dquot_quota_off(sb
, i
);
2210 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
2214 * Maximal extent format file size.
2215 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2216 * extent format containers, within a sector_t, and within i_blocks
2217 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2218 * so that won't be a limiting factor.
2220 * However there is other limiting factor. We do store extents in the form
2221 * of starting block and length, hence the resulting length of the extent
2222 * covering maximum file size must fit into on-disk format containers as
2223 * well. Given that length is always by 1 unit bigger than max unit (because
2224 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2226 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2228 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
2231 loff_t upper_limit
= MAX_LFS_FILESIZE
;
2233 /* small i_blocks in vfs inode? */
2234 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2236 * CONFIG_LBDAF is not enabled implies the inode
2237 * i_block represent total blocks in 512 bytes
2238 * 32 == size of vfs inode i_blocks * 8
2240 upper_limit
= (1LL << 32) - 1;
2242 /* total blocks in file system block size */
2243 upper_limit
>>= (blkbits
- 9);
2244 upper_limit
<<= blkbits
;
2248 * 32-bit extent-start container, ee_block. We lower the maxbytes
2249 * by one fs block, so ee_len can cover the extent of maximum file
2252 res
= (1LL << 32) - 1;
2255 /* Sanity check against vm- & vfs- imposed limits */
2256 if (res
> upper_limit
)
2263 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2264 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2265 * We need to be 1 filesystem block less than the 2^48 sector limit.
2267 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
2269 loff_t res
= EXT4_NDIR_BLOCKS
;
2272 /* This is calculated to be the largest file size for a dense, block
2273 * mapped file such that the file's total number of 512-byte sectors,
2274 * including data and all indirect blocks, does not exceed (2^48 - 1).
2276 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2277 * number of 512-byte sectors of the file.
2280 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2282 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2283 * the inode i_block field represents total file blocks in
2284 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2286 upper_limit
= (1LL << 32) - 1;
2288 /* total blocks in file system block size */
2289 upper_limit
>>= (bits
- 9);
2293 * We use 48 bit ext4_inode i_blocks
2294 * With EXT4_HUGE_FILE_FL set the i_blocks
2295 * represent total number of blocks in
2296 * file system block size
2298 upper_limit
= (1LL << 48) - 1;
2302 /* indirect blocks */
2304 /* double indirect blocks */
2305 meta_blocks
+= 1 + (1LL << (bits
-2));
2306 /* tripple indirect blocks */
2307 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
2309 upper_limit
-= meta_blocks
;
2310 upper_limit
<<= bits
;
2312 res
+= 1LL << (bits
-2);
2313 res
+= 1LL << (2*(bits
-2));
2314 res
+= 1LL << (3*(bits
-2));
2316 if (res
> upper_limit
)
2319 if (res
> MAX_LFS_FILESIZE
)
2320 res
= MAX_LFS_FILESIZE
;
2325 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
2326 ext4_fsblk_t logical_sb_block
, int nr
)
2328 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2329 ext4_group_t bg
, first_meta_bg
;
2332 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
2334 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_META_BG
) ||
2336 return logical_sb_block
+ nr
+ 1;
2337 bg
= sbi
->s_desc_per_block
* nr
;
2338 if (ext4_bg_has_super(sb
, bg
))
2341 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
2345 * ext4_get_stripe_size: Get the stripe size.
2346 * @sbi: In memory super block info
2348 * If we have specified it via mount option, then
2349 * use the mount option value. If the value specified at mount time is
2350 * greater than the blocks per group use the super block value.
2351 * If the super block value is greater than blocks per group return 0.
2352 * Allocator needs it be less than blocks per group.
2355 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
2357 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
2358 unsigned long stripe_width
=
2359 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
2362 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
2363 ret
= sbi
->s_stripe
;
2364 else if (stripe_width
<= sbi
->s_blocks_per_group
)
2366 else if (stride
<= sbi
->s_blocks_per_group
)
2372 * If the stripe width is 1, this makes no sense and
2373 * we set it to 0 to turn off stripe handling code.
2384 struct attribute attr
;
2385 ssize_t (*show
)(struct ext4_attr
*, struct ext4_sb_info
*, char *);
2386 ssize_t (*store
)(struct ext4_attr
*, struct ext4_sb_info
*,
2387 const char *, size_t);
2391 static int parse_strtoull(const char *buf
,
2392 unsigned long long max
, unsigned long long *value
)
2396 ret
= kstrtoull(skip_spaces(buf
), 0, value
);
2397 if (!ret
&& *value
> max
)
2402 static ssize_t
delayed_allocation_blocks_show(struct ext4_attr
*a
,
2403 struct ext4_sb_info
*sbi
,
2406 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2408 percpu_counter_sum(&sbi
->s_dirtyclusters_counter
)));
2411 static ssize_t
session_write_kbytes_show(struct ext4_attr
*a
,
2412 struct ext4_sb_info
*sbi
, char *buf
)
2414 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2416 if (!sb
->s_bdev
->bd_part
)
2417 return snprintf(buf
, PAGE_SIZE
, "0\n");
2418 return snprintf(buf
, PAGE_SIZE
, "%lu\n",
2419 (part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2420 sbi
->s_sectors_written_start
) >> 1);
2423 static ssize_t
lifetime_write_kbytes_show(struct ext4_attr
*a
,
2424 struct ext4_sb_info
*sbi
, char *buf
)
2426 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2428 if (!sb
->s_bdev
->bd_part
)
2429 return snprintf(buf
, PAGE_SIZE
, "0\n");
2430 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2431 (unsigned long long)(sbi
->s_kbytes_written
+
2432 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2433 EXT4_SB(sb
)->s_sectors_written_start
) >> 1)));
2436 static ssize_t
inode_readahead_blks_store(struct ext4_attr
*a
,
2437 struct ext4_sb_info
*sbi
,
2438 const char *buf
, size_t count
)
2443 ret
= kstrtoul(skip_spaces(buf
), 0, &t
);
2447 if (t
&& (!is_power_of_2(t
) || t
> 0x40000000))
2450 sbi
->s_inode_readahead_blks
= t
;
2454 static ssize_t
sbi_ui_show(struct ext4_attr
*a
,
2455 struct ext4_sb_info
*sbi
, char *buf
)
2457 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2459 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
2462 static ssize_t
sbi_ui_store(struct ext4_attr
*a
,
2463 struct ext4_sb_info
*sbi
,
2464 const char *buf
, size_t count
)
2466 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2470 ret
= kstrtoul(skip_spaces(buf
), 0, &t
);
2477 static ssize_t
reserved_clusters_show(struct ext4_attr
*a
,
2478 struct ext4_sb_info
*sbi
, char *buf
)
2480 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2481 (unsigned long long) atomic64_read(&sbi
->s_resv_clusters
));
2484 static ssize_t
reserved_clusters_store(struct ext4_attr
*a
,
2485 struct ext4_sb_info
*sbi
,
2486 const char *buf
, size_t count
)
2488 unsigned long long val
;
2491 if (parse_strtoull(buf
, -1ULL, &val
))
2493 ret
= ext4_reserve_clusters(sbi
, val
);
2495 return ret
? ret
: count
;
2498 static ssize_t
trigger_test_error(struct ext4_attr
*a
,
2499 struct ext4_sb_info
*sbi
,
2500 const char *buf
, size_t count
)
2504 if (!capable(CAP_SYS_ADMIN
))
2507 if (len
&& buf
[len
-1] == '\n')
2511 ext4_error(sbi
->s_sb
, "%.*s", len
, buf
);
2515 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2516 static struct ext4_attr ext4_attr_##_name = { \
2517 .attr = {.name = __stringify(_name), .mode = _mode }, \
2520 .offset = offsetof(struct ext4_sb_info, _elname), \
2522 #define EXT4_ATTR(name, mode, show, store) \
2523 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2525 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2526 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2527 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2528 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2529 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2530 #define ATTR_LIST(name) &ext4_attr_##name.attr
2532 EXT4_RO_ATTR(delayed_allocation_blocks
);
2533 EXT4_RO_ATTR(session_write_kbytes
);
2534 EXT4_RO_ATTR(lifetime_write_kbytes
);
2535 EXT4_RW_ATTR(reserved_clusters
);
2536 EXT4_ATTR_OFFSET(inode_readahead_blks
, 0644, sbi_ui_show
,
2537 inode_readahead_blks_store
, s_inode_readahead_blks
);
2538 EXT4_RW_ATTR_SBI_UI(inode_goal
, s_inode_goal
);
2539 EXT4_RW_ATTR_SBI_UI(mb_stats
, s_mb_stats
);
2540 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan
, s_mb_max_to_scan
);
2541 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan
, s_mb_min_to_scan
);
2542 EXT4_RW_ATTR_SBI_UI(mb_order2_req
, s_mb_order2_reqs
);
2543 EXT4_RW_ATTR_SBI_UI(mb_stream_req
, s_mb_stream_request
);
2544 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc
, s_mb_group_prealloc
);
2545 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump
, s_max_writeback_mb_bump
);
2546 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb
, s_extent_max_zeroout_kb
);
2547 EXT4_ATTR(trigger_fs_error
, 0200, NULL
, trigger_test_error
);
2549 static struct attribute
*ext4_attrs
[] = {
2550 ATTR_LIST(delayed_allocation_blocks
),
2551 ATTR_LIST(session_write_kbytes
),
2552 ATTR_LIST(lifetime_write_kbytes
),
2553 ATTR_LIST(reserved_clusters
),
2554 ATTR_LIST(inode_readahead_blks
),
2555 ATTR_LIST(inode_goal
),
2556 ATTR_LIST(mb_stats
),
2557 ATTR_LIST(mb_max_to_scan
),
2558 ATTR_LIST(mb_min_to_scan
),
2559 ATTR_LIST(mb_order2_req
),
2560 ATTR_LIST(mb_stream_req
),
2561 ATTR_LIST(mb_group_prealloc
),
2562 ATTR_LIST(max_writeback_mb_bump
),
2563 ATTR_LIST(extent_max_zeroout_kb
),
2564 ATTR_LIST(trigger_fs_error
),
2568 /* Features this copy of ext4 supports */
2569 EXT4_INFO_ATTR(lazy_itable_init
);
2570 EXT4_INFO_ATTR(batched_discard
);
2571 EXT4_INFO_ATTR(meta_bg_resize
);
2573 static struct attribute
*ext4_feat_attrs
[] = {
2574 ATTR_LIST(lazy_itable_init
),
2575 ATTR_LIST(batched_discard
),
2576 ATTR_LIST(meta_bg_resize
),
2580 static ssize_t
ext4_attr_show(struct kobject
*kobj
,
2581 struct attribute
*attr
, char *buf
)
2583 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2585 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2587 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
2590 static ssize_t
ext4_attr_store(struct kobject
*kobj
,
2591 struct attribute
*attr
,
2592 const char *buf
, size_t len
)
2594 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2596 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2598 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
2601 static void ext4_sb_release(struct kobject
*kobj
)
2603 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2605 complete(&sbi
->s_kobj_unregister
);
2608 static const struct sysfs_ops ext4_attr_ops
= {
2609 .show
= ext4_attr_show
,
2610 .store
= ext4_attr_store
,
2613 static struct kobj_type ext4_ktype
= {
2614 .default_attrs
= ext4_attrs
,
2615 .sysfs_ops
= &ext4_attr_ops
,
2616 .release
= ext4_sb_release
,
2619 static void ext4_feat_release(struct kobject
*kobj
)
2621 complete(&ext4_feat
->f_kobj_unregister
);
2624 static struct kobj_type ext4_feat_ktype
= {
2625 .default_attrs
= ext4_feat_attrs
,
2626 .sysfs_ops
= &ext4_attr_ops
,
2627 .release
= ext4_feat_release
,
2631 * Check whether this filesystem can be mounted based on
2632 * the features present and the RDONLY/RDWR mount requested.
2633 * Returns 1 if this filesystem can be mounted as requested,
2634 * 0 if it cannot be.
2636 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
2638 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT4_FEATURE_INCOMPAT_SUPP
)) {
2639 ext4_msg(sb
, KERN_ERR
,
2640 "Couldn't mount because of "
2641 "unsupported optional features (%x)",
2642 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2643 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2650 /* Check that feature set is OK for a read-write mount */
2651 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT4_FEATURE_RO_COMPAT_SUPP
)) {
2652 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
2653 "unsupported optional features (%x)",
2654 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2655 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2659 * Large file size enabled file system can only be mounted
2660 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2662 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_HUGE_FILE
)) {
2663 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
2664 ext4_msg(sb
, KERN_ERR
, "Filesystem with huge files "
2665 "cannot be mounted RDWR without "
2670 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_BIGALLOC
) &&
2671 !EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_EXTENTS
)) {
2672 ext4_msg(sb
, KERN_ERR
,
2673 "Can't support bigalloc feature without "
2674 "extents feature\n");
2678 #ifndef CONFIG_QUOTA
2679 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
) &&
2681 ext4_msg(sb
, KERN_ERR
,
2682 "Filesystem with quota feature cannot be mounted RDWR "
2683 "without CONFIG_QUOTA");
2686 #endif /* CONFIG_QUOTA */
2691 * This function is called once a day if we have errors logged
2692 * on the file system
2694 static void print_daily_error_info(unsigned long arg
)
2696 struct super_block
*sb
= (struct super_block
*) arg
;
2697 struct ext4_sb_info
*sbi
;
2698 struct ext4_super_block
*es
;
2703 if (es
->s_error_count
)
2704 /* fsck newer than v1.41.13 is needed to clean this condition. */
2705 ext4_msg(sb
, KERN_NOTICE
, "error count since last fsck: %u",
2706 le32_to_cpu(es
->s_error_count
));
2707 if (es
->s_first_error_time
) {
2708 printk(KERN_NOTICE
"EXT4-fs (%s): initial error at time %u: %.*s:%d",
2709 sb
->s_id
, le32_to_cpu(es
->s_first_error_time
),
2710 (int) sizeof(es
->s_first_error_func
),
2711 es
->s_first_error_func
,
2712 le32_to_cpu(es
->s_first_error_line
));
2713 if (es
->s_first_error_ino
)
2714 printk(": inode %u",
2715 le32_to_cpu(es
->s_first_error_ino
));
2716 if (es
->s_first_error_block
)
2717 printk(": block %llu", (unsigned long long)
2718 le64_to_cpu(es
->s_first_error_block
));
2721 if (es
->s_last_error_time
) {
2722 printk(KERN_NOTICE
"EXT4-fs (%s): last error at time %u: %.*s:%d",
2723 sb
->s_id
, le32_to_cpu(es
->s_last_error_time
),
2724 (int) sizeof(es
->s_last_error_func
),
2725 es
->s_last_error_func
,
2726 le32_to_cpu(es
->s_last_error_line
));
2727 if (es
->s_last_error_ino
)
2728 printk(": inode %u",
2729 le32_to_cpu(es
->s_last_error_ino
));
2730 if (es
->s_last_error_block
)
2731 printk(": block %llu", (unsigned long long)
2732 le64_to_cpu(es
->s_last_error_block
));
2735 mod_timer(&sbi
->s_err_report
, jiffies
+ 24*60*60*HZ
); /* Once a day */
2738 /* Find next suitable group and run ext4_init_inode_table */
2739 static int ext4_run_li_request(struct ext4_li_request
*elr
)
2741 struct ext4_group_desc
*gdp
= NULL
;
2742 ext4_group_t group
, ngroups
;
2743 struct super_block
*sb
;
2744 unsigned long timeout
= 0;
2748 ngroups
= EXT4_SB(sb
)->s_groups_count
;
2751 for (group
= elr
->lr_next_group
; group
< ngroups
; group
++) {
2752 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2758 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2762 if (group
>= ngroups
)
2767 ret
= ext4_init_inode_table(sb
, group
,
2768 elr
->lr_timeout
? 0 : 1);
2769 if (elr
->lr_timeout
== 0) {
2770 timeout
= (jiffies
- timeout
) *
2771 elr
->lr_sbi
->s_li_wait_mult
;
2772 elr
->lr_timeout
= timeout
;
2774 elr
->lr_next_sched
= jiffies
+ elr
->lr_timeout
;
2775 elr
->lr_next_group
= group
+ 1;
2783 * Remove lr_request from the list_request and free the
2784 * request structure. Should be called with li_list_mtx held
2786 static void ext4_remove_li_request(struct ext4_li_request
*elr
)
2788 struct ext4_sb_info
*sbi
;
2795 list_del(&elr
->lr_request
);
2796 sbi
->s_li_request
= NULL
;
2800 static void ext4_unregister_li_request(struct super_block
*sb
)
2802 mutex_lock(&ext4_li_mtx
);
2803 if (!ext4_li_info
) {
2804 mutex_unlock(&ext4_li_mtx
);
2808 mutex_lock(&ext4_li_info
->li_list_mtx
);
2809 ext4_remove_li_request(EXT4_SB(sb
)->s_li_request
);
2810 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2811 mutex_unlock(&ext4_li_mtx
);
2814 static struct task_struct
*ext4_lazyinit_task
;
2817 * This is the function where ext4lazyinit thread lives. It walks
2818 * through the request list searching for next scheduled filesystem.
2819 * When such a fs is found, run the lazy initialization request
2820 * (ext4_rn_li_request) and keep track of the time spend in this
2821 * function. Based on that time we compute next schedule time of
2822 * the request. When walking through the list is complete, compute
2823 * next waking time and put itself into sleep.
2825 static int ext4_lazyinit_thread(void *arg
)
2827 struct ext4_lazy_init
*eli
= (struct ext4_lazy_init
*)arg
;
2828 struct list_head
*pos
, *n
;
2829 struct ext4_li_request
*elr
;
2830 unsigned long next_wakeup
, cur
;
2832 BUG_ON(NULL
== eli
);
2837 next_wakeup
= MAX_JIFFY_OFFSET
;
2839 mutex_lock(&eli
->li_list_mtx
);
2840 if (list_empty(&eli
->li_request_list
)) {
2841 mutex_unlock(&eli
->li_list_mtx
);
2845 list_for_each_safe(pos
, n
, &eli
->li_request_list
) {
2846 elr
= list_entry(pos
, struct ext4_li_request
,
2849 if (time_after_eq(jiffies
, elr
->lr_next_sched
)) {
2850 if (ext4_run_li_request(elr
) != 0) {
2851 /* error, remove the lazy_init job */
2852 ext4_remove_li_request(elr
);
2857 if (time_before(elr
->lr_next_sched
, next_wakeup
))
2858 next_wakeup
= elr
->lr_next_sched
;
2860 mutex_unlock(&eli
->li_list_mtx
);
2865 if ((time_after_eq(cur
, next_wakeup
)) ||
2866 (MAX_JIFFY_OFFSET
== next_wakeup
)) {
2871 schedule_timeout_interruptible(next_wakeup
- cur
);
2873 if (kthread_freezable_should_stop(NULL
)) {
2874 ext4_clear_request_list();
2881 * It looks like the request list is empty, but we need
2882 * to check it under the li_list_mtx lock, to prevent any
2883 * additions into it, and of course we should lock ext4_li_mtx
2884 * to atomically free the list and ext4_li_info, because at
2885 * this point another ext4 filesystem could be registering
2888 mutex_lock(&ext4_li_mtx
);
2889 mutex_lock(&eli
->li_list_mtx
);
2890 if (!list_empty(&eli
->li_request_list
)) {
2891 mutex_unlock(&eli
->li_list_mtx
);
2892 mutex_unlock(&ext4_li_mtx
);
2895 mutex_unlock(&eli
->li_list_mtx
);
2896 kfree(ext4_li_info
);
2897 ext4_li_info
= NULL
;
2898 mutex_unlock(&ext4_li_mtx
);
2903 static void ext4_clear_request_list(void)
2905 struct list_head
*pos
, *n
;
2906 struct ext4_li_request
*elr
;
2908 mutex_lock(&ext4_li_info
->li_list_mtx
);
2909 list_for_each_safe(pos
, n
, &ext4_li_info
->li_request_list
) {
2910 elr
= list_entry(pos
, struct ext4_li_request
,
2912 ext4_remove_li_request(elr
);
2914 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2917 static int ext4_run_lazyinit_thread(void)
2919 ext4_lazyinit_task
= kthread_run(ext4_lazyinit_thread
,
2920 ext4_li_info
, "ext4lazyinit");
2921 if (IS_ERR(ext4_lazyinit_task
)) {
2922 int err
= PTR_ERR(ext4_lazyinit_task
);
2923 ext4_clear_request_list();
2924 kfree(ext4_li_info
);
2925 ext4_li_info
= NULL
;
2926 printk(KERN_CRIT
"EXT4-fs: error %d creating inode table "
2927 "initialization thread\n",
2931 ext4_li_info
->li_state
|= EXT4_LAZYINIT_RUNNING
;
2936 * Check whether it make sense to run itable init. thread or not.
2937 * If there is at least one uninitialized inode table, return
2938 * corresponding group number, else the loop goes through all
2939 * groups and return total number of groups.
2941 static ext4_group_t
ext4_has_uninit_itable(struct super_block
*sb
)
2943 ext4_group_t group
, ngroups
= EXT4_SB(sb
)->s_groups_count
;
2944 struct ext4_group_desc
*gdp
= NULL
;
2946 for (group
= 0; group
< ngroups
; group
++) {
2947 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2951 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2958 static int ext4_li_info_new(void)
2960 struct ext4_lazy_init
*eli
= NULL
;
2962 eli
= kzalloc(sizeof(*eli
), GFP_KERNEL
);
2966 INIT_LIST_HEAD(&eli
->li_request_list
);
2967 mutex_init(&eli
->li_list_mtx
);
2969 eli
->li_state
|= EXT4_LAZYINIT_QUIT
;
2976 static struct ext4_li_request
*ext4_li_request_new(struct super_block
*sb
,
2979 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2980 struct ext4_li_request
*elr
;
2983 elr
= kzalloc(sizeof(*elr
), GFP_KERNEL
);
2989 elr
->lr_next_group
= start
;
2992 * Randomize first schedule time of the request to
2993 * spread the inode table initialization requests
2996 get_random_bytes(&rnd
, sizeof(rnd
));
2997 elr
->lr_next_sched
= jiffies
+ (unsigned long)rnd
%
2998 (EXT4_DEF_LI_MAX_START_DELAY
* HZ
);
3003 int ext4_register_li_request(struct super_block
*sb
,
3004 ext4_group_t first_not_zeroed
)
3006 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3007 struct ext4_li_request
*elr
= NULL
;
3008 ext4_group_t ngroups
= EXT4_SB(sb
)->s_groups_count
;
3011 mutex_lock(&ext4_li_mtx
);
3012 if (sbi
->s_li_request
!= NULL
) {
3014 * Reset timeout so it can be computed again, because
3015 * s_li_wait_mult might have changed.
3017 sbi
->s_li_request
->lr_timeout
= 0;
3021 if (first_not_zeroed
== ngroups
||
3022 (sb
->s_flags
& MS_RDONLY
) ||
3023 !test_opt(sb
, INIT_INODE_TABLE
))
3026 elr
= ext4_li_request_new(sb
, first_not_zeroed
);
3032 if (NULL
== ext4_li_info
) {
3033 ret
= ext4_li_info_new();
3038 mutex_lock(&ext4_li_info
->li_list_mtx
);
3039 list_add(&elr
->lr_request
, &ext4_li_info
->li_request_list
);
3040 mutex_unlock(&ext4_li_info
->li_list_mtx
);
3042 sbi
->s_li_request
= elr
;
3044 * set elr to NULL here since it has been inserted to
3045 * the request_list and the removal and free of it is
3046 * handled by ext4_clear_request_list from now on.
3050 if (!(ext4_li_info
->li_state
& EXT4_LAZYINIT_RUNNING
)) {
3051 ret
= ext4_run_lazyinit_thread();
3056 mutex_unlock(&ext4_li_mtx
);
3063 * We do not need to lock anything since this is called on
3066 static void ext4_destroy_lazyinit_thread(void)
3069 * If thread exited earlier
3070 * there's nothing to be done.
3072 if (!ext4_li_info
|| !ext4_lazyinit_task
)
3075 kthread_stop(ext4_lazyinit_task
);
3078 static int set_journal_csum_feature_set(struct super_block
*sb
)
3081 int compat
, incompat
;
3082 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3084 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3085 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
)) {
3086 /* journal checksum v2 */
3088 incompat
= JBD2_FEATURE_INCOMPAT_CSUM_V2
;
3090 /* journal checksum v1 */
3091 compat
= JBD2_FEATURE_COMPAT_CHECKSUM
;
3095 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
3096 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3098 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
|
3100 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
3101 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3104 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3105 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3107 jbd2_journal_clear_features(sbi
->s_journal
,
3108 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3109 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
|
3110 JBD2_FEATURE_INCOMPAT_CSUM_V2
);
3117 * Note: calculating the overhead so we can be compatible with
3118 * historical BSD practice is quite difficult in the face of
3119 * clusters/bigalloc. This is because multiple metadata blocks from
3120 * different block group can end up in the same allocation cluster.
3121 * Calculating the exact overhead in the face of clustered allocation
3122 * requires either O(all block bitmaps) in memory or O(number of block
3123 * groups**2) in time. We will still calculate the superblock for
3124 * older file systems --- and if we come across with a bigalloc file
3125 * system with zero in s_overhead_clusters the estimate will be close to
3126 * correct especially for very large cluster sizes --- but for newer
3127 * file systems, it's better to calculate this figure once at mkfs
3128 * time, and store it in the superblock. If the superblock value is
3129 * present (even for non-bigalloc file systems), we will use it.
3131 static int count_overhead(struct super_block
*sb
, ext4_group_t grp
,
3134 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3135 struct ext4_group_desc
*gdp
;
3136 ext4_fsblk_t first_block
, last_block
, b
;
3137 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3138 int s
, j
, count
= 0;
3140 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_BIGALLOC
))
3141 return (ext4_bg_has_super(sb
, grp
) + ext4_bg_num_gdb(sb
, grp
) +
3142 sbi
->s_itb_per_group
+ 2);
3144 first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
) +
3145 (grp
* EXT4_BLOCKS_PER_GROUP(sb
));
3146 last_block
= first_block
+ EXT4_BLOCKS_PER_GROUP(sb
) - 1;
3147 for (i
= 0; i
< ngroups
; i
++) {
3148 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
3149 b
= ext4_block_bitmap(sb
, gdp
);
3150 if (b
>= first_block
&& b
<= last_block
) {
3151 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3154 b
= ext4_inode_bitmap(sb
, gdp
);
3155 if (b
>= first_block
&& b
<= last_block
) {
3156 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3159 b
= ext4_inode_table(sb
, gdp
);
3160 if (b
>= first_block
&& b
+ sbi
->s_itb_per_group
<= last_block
)
3161 for (j
= 0; j
< sbi
->s_itb_per_group
; j
++, b
++) {
3162 int c
= EXT4_B2C(sbi
, b
- first_block
);
3163 ext4_set_bit(c
, buf
);
3169 if (ext4_bg_has_super(sb
, grp
)) {
3170 ext4_set_bit(s
++, buf
);
3173 for (j
= ext4_bg_num_gdb(sb
, grp
); j
> 0; j
--) {
3174 ext4_set_bit(EXT4_B2C(sbi
, s
++), buf
);
3180 return EXT4_CLUSTERS_PER_GROUP(sb
) -
3181 ext4_count_free(buf
, EXT4_CLUSTERS_PER_GROUP(sb
) / 8);
3185 * Compute the overhead and stash it in sbi->s_overhead
3187 int ext4_calculate_overhead(struct super_block
*sb
)
3189 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3190 struct ext4_super_block
*es
= sbi
->s_es
;
3191 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3192 ext4_fsblk_t overhead
= 0;
3193 char *buf
= (char *) get_zeroed_page(GFP_KERNEL
);
3199 * Compute the overhead (FS structures). This is constant
3200 * for a given filesystem unless the number of block groups
3201 * changes so we cache the previous value until it does.
3205 * All of the blocks before first_data_block are overhead
3207 overhead
= EXT4_B2C(sbi
, le32_to_cpu(es
->s_first_data_block
));
3210 * Add the overhead found in each block group
3212 for (i
= 0; i
< ngroups
; i
++) {
3215 blks
= count_overhead(sb
, i
, buf
);
3218 memset(buf
, 0, PAGE_SIZE
);
3221 /* Add the journal blocks as well */
3223 overhead
+= EXT4_NUM_B2C(sbi
, sbi
->s_journal
->j_maxlen
);
3225 sbi
->s_overhead
= overhead
;
3227 free_page((unsigned long) buf
);
3232 static ext4_fsblk_t
ext4_calculate_resv_clusters(struct super_block
*sb
)
3234 ext4_fsblk_t resv_clusters
;
3237 * There's no need to reserve anything when we aren't using extents.
3238 * The space estimates are exact, there are no unwritten extents,
3239 * hole punching doesn't need new metadata... This is needed especially
3240 * to keep ext2/3 backward compatibility.
3242 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_EXTENTS
))
3245 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3246 * This should cover the situations where we can not afford to run
3247 * out of space like for example punch hole, or converting
3248 * uninitialized extents in delalloc path. In most cases such
3249 * allocation would require 1, or 2 blocks, higher numbers are
3252 resv_clusters
= ext4_blocks_count(EXT4_SB(sb
)->s_es
) >>
3253 EXT4_SB(sb
)->s_cluster_bits
;
3255 do_div(resv_clusters
, 50);
3256 resv_clusters
= min_t(ext4_fsblk_t
, resv_clusters
, 4096);
3258 return resv_clusters
;
3262 static int ext4_reserve_clusters(struct ext4_sb_info
*sbi
, ext4_fsblk_t count
)
3264 ext4_fsblk_t clusters
= ext4_blocks_count(sbi
->s_es
) >>
3265 sbi
->s_cluster_bits
;
3267 if (count
>= clusters
)
3270 atomic64_set(&sbi
->s_resv_clusters
, count
);
3274 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
3276 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
3277 struct buffer_head
*bh
;
3278 struct ext4_super_block
*es
= NULL
;
3279 struct ext4_sb_info
*sbi
;
3281 ext4_fsblk_t sb_block
= get_sb_block(&data
);
3282 ext4_fsblk_t logical_sb_block
;
3283 unsigned long offset
= 0;
3284 unsigned long journal_devnum
= 0;
3285 unsigned long def_mount_opts
;
3290 int blocksize
, clustersize
;
3291 unsigned int db_count
;
3293 int needs_recovery
, has_huge_files
, has_bigalloc
;
3296 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
3297 ext4_group_t first_not_zeroed
;
3299 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
3303 sbi
->s_blockgroup_lock
=
3304 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
3305 if (!sbi
->s_blockgroup_lock
) {
3309 sb
->s_fs_info
= sbi
;
3311 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
3312 sbi
->s_sb_block
= sb_block
;
3313 if (sb
->s_bdev
->bd_part
)
3314 sbi
->s_sectors_written_start
=
3315 part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]);
3317 /* Cleanup superblock name */
3318 for (cp
= sb
->s_id
; (cp
= strchr(cp
, '/'));)
3321 /* -EINVAL is default */
3323 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
3325 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
3330 * The ext4 superblock will not be buffer aligned for other than 1kB
3331 * block sizes. We need to calculate the offset from buffer start.
3333 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
3334 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3335 offset
= do_div(logical_sb_block
, blocksize
);
3337 logical_sb_block
= sb_block
;
3340 if (!(bh
= sb_bread(sb
, logical_sb_block
))) {
3341 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
3345 * Note: s_es must be initialized as soon as possible because
3346 * some ext4 macro-instructions depend on its value
3348 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
3350 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
3351 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
3353 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
3355 /* Warn if metadata_csum and gdt_csum are both set. */
3356 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3357 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
) &&
3358 EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_GDT_CSUM
))
3359 ext4_warning(sb
, KERN_INFO
"metadata_csum and uninit_bg are "
3360 "redundant flags; please run fsck.");
3362 /* Check for a known checksum algorithm */
3363 if (!ext4_verify_csum_type(sb
, es
)) {
3364 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3365 "unknown checksum algorithm.");
3370 /* Load the checksum driver */
3371 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3372 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
)) {
3373 sbi
->s_chksum_driver
= crypto_alloc_shash("crc32c", 0, 0);
3374 if (IS_ERR(sbi
->s_chksum_driver
)) {
3375 ext4_msg(sb
, KERN_ERR
, "Cannot load crc32c driver.");
3376 ret
= PTR_ERR(sbi
->s_chksum_driver
);
3377 sbi
->s_chksum_driver
= NULL
;
3382 /* Check superblock checksum */
3383 if (!ext4_superblock_csum_verify(sb
, es
)) {
3384 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3385 "invalid superblock checksum. Run e2fsck?");
3390 /* Precompute checksum seed for all metadata */
3391 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3392 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
))
3393 sbi
->s_csum_seed
= ext4_chksum(sbi
, ~0, es
->s_uuid
,
3394 sizeof(es
->s_uuid
));
3396 /* Set defaults before we parse the mount options */
3397 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
3398 set_opt(sb
, INIT_INODE_TABLE
);
3399 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
3401 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
)
3403 if (def_mount_opts
& EXT4_DEFM_UID16
)
3404 set_opt(sb
, NO_UID32
);
3405 /* xattr user namespace & acls are now defaulted on */
3406 set_opt(sb
, XATTR_USER
);
3407 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3408 set_opt(sb
, POSIX_ACL
);
3410 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
3411 set_opt(sb
, JOURNAL_DATA
);
3412 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
3413 set_opt(sb
, ORDERED_DATA
);
3414 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
3415 set_opt(sb
, WRITEBACK_DATA
);
3417 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
3418 set_opt(sb
, ERRORS_PANIC
);
3419 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
3420 set_opt(sb
, ERRORS_CONT
);
3422 set_opt(sb
, ERRORS_RO
);
3423 if (def_mount_opts
& EXT4_DEFM_BLOCK_VALIDITY
)
3424 set_opt(sb
, BLOCK_VALIDITY
);
3425 if (def_mount_opts
& EXT4_DEFM_DISCARD
)
3426 set_opt(sb
, DISCARD
);
3428 sbi
->s_resuid
= make_kuid(&init_user_ns
, le16_to_cpu(es
->s_def_resuid
));
3429 sbi
->s_resgid
= make_kgid(&init_user_ns
, le16_to_cpu(es
->s_def_resgid
));
3430 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
3431 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
3432 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
3434 if ((def_mount_opts
& EXT4_DEFM_NOBARRIER
) == 0)
3435 set_opt(sb
, BARRIER
);
3438 * enable delayed allocation by default
3439 * Use -o nodelalloc to turn it off
3441 if (!IS_EXT3_SB(sb
) && !IS_EXT2_SB(sb
) &&
3442 ((def_mount_opts
& EXT4_DEFM_NODELALLOC
) == 0))
3443 set_opt(sb
, DELALLOC
);
3446 * set default s_li_wait_mult for lazyinit, for the case there is
3447 * no mount option specified.
3449 sbi
->s_li_wait_mult
= EXT4_DEF_LI_WAIT_MULT
;
3451 if (!parse_options((char *) sbi
->s_es
->s_mount_opts
, sb
,
3452 &journal_devnum
, &journal_ioprio
, 0)) {
3453 ext4_msg(sb
, KERN_WARNING
,
3454 "failed to parse options in superblock: %s",
3455 sbi
->s_es
->s_mount_opts
);
3457 sbi
->s_def_mount_opt
= sbi
->s_mount_opt
;
3458 if (!parse_options((char *) data
, sb
, &journal_devnum
,
3459 &journal_ioprio
, 0))
3462 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
3463 printk_once(KERN_WARNING
"EXT4-fs: Warning: mounting "
3464 "with data=journal disables delayed "
3465 "allocation and O_DIRECT support!\n");
3466 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
3467 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3468 "both data=journal and delalloc");
3471 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
3472 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3473 "both data=journal and dioread_nolock");
3476 if (test_opt(sb
, DELALLOC
))
3477 clear_opt(sb
, DELALLOC
);
3480 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3481 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
3483 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
3484 (EXT4_HAS_COMPAT_FEATURE(sb
, ~0U) ||
3485 EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
3486 EXT4_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
3487 ext4_msg(sb
, KERN_WARNING
,
3488 "feature flags set on rev 0 fs, "
3489 "running e2fsck is recommended");
3491 if (IS_EXT2_SB(sb
)) {
3492 if (ext2_feature_set_ok(sb
))
3493 ext4_msg(sb
, KERN_INFO
, "mounting ext2 file system "
3494 "using the ext4 subsystem");
3496 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext2 due "
3497 "to feature incompatibilities");
3502 if (IS_EXT3_SB(sb
)) {
3503 if (ext3_feature_set_ok(sb
))
3504 ext4_msg(sb
, KERN_INFO
, "mounting ext3 file system "
3505 "using the ext4 subsystem");
3507 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext3 due "
3508 "to feature incompatibilities");
3514 * Check feature flags regardless of the revision level, since we
3515 * previously didn't change the revision level when setting the flags,
3516 * so there is a chance incompat flags are set on a rev 0 filesystem.
3518 if (!ext4_feature_set_ok(sb
, (sb
->s_flags
& MS_RDONLY
)))
3521 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
3522 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
3523 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
3524 ext4_msg(sb
, KERN_ERR
,
3525 "Unsupported filesystem blocksize %d", blocksize
);
3529 if (sb
->s_blocksize
!= blocksize
) {
3530 /* Validate the filesystem blocksize */
3531 if (!sb_set_blocksize(sb
, blocksize
)) {
3532 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
3538 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3539 offset
= do_div(logical_sb_block
, blocksize
);
3540 bh
= sb_bread(sb
, logical_sb_block
);
3542 ext4_msg(sb
, KERN_ERR
,
3543 "Can't read superblock on 2nd try");
3546 es
= (struct ext4_super_block
*)(bh
->b_data
+ offset
);
3548 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
3549 ext4_msg(sb
, KERN_ERR
,
3550 "Magic mismatch, very weird!");
3555 has_huge_files
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3556 EXT4_FEATURE_RO_COMPAT_HUGE_FILE
);
3557 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
3559 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
3561 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
3562 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
3563 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
3565 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
3566 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
3567 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
3568 (!is_power_of_2(sbi
->s_inode_size
)) ||
3569 (sbi
->s_inode_size
> blocksize
)) {
3570 ext4_msg(sb
, KERN_ERR
,
3571 "unsupported inode size: %d",
3575 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
3576 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
3579 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
3580 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
)) {
3581 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
3582 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
3583 !is_power_of_2(sbi
->s_desc_size
)) {
3584 ext4_msg(sb
, KERN_ERR
,
3585 "unsupported descriptor size %lu",
3590 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
3592 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
3593 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
3594 if (EXT4_INODE_SIZE(sb
) == 0 || EXT4_INODES_PER_GROUP(sb
) == 0)
3597 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
3598 if (sbi
->s_inodes_per_block
== 0)
3600 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
3601 sbi
->s_inodes_per_block
;
3602 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
3604 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3605 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
3606 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
3608 for (i
= 0; i
< 4; i
++)
3609 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
3610 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
3611 if (EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_DIR_INDEX
)) {
3612 i
= le32_to_cpu(es
->s_flags
);
3613 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
3614 sbi
->s_hash_unsigned
= 3;
3615 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
3616 #ifdef __CHAR_UNSIGNED__
3617 if (!(sb
->s_flags
& MS_RDONLY
))
3619 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
3620 sbi
->s_hash_unsigned
= 3;
3622 if (!(sb
->s_flags
& MS_RDONLY
))
3624 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
3629 /* Handle clustersize */
3630 clustersize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_cluster_size
);
3631 has_bigalloc
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3632 EXT4_FEATURE_RO_COMPAT_BIGALLOC
);
3634 if (clustersize
< blocksize
) {
3635 ext4_msg(sb
, KERN_ERR
,
3636 "cluster size (%d) smaller than "
3637 "block size (%d)", clustersize
, blocksize
);
3640 sbi
->s_cluster_bits
= le32_to_cpu(es
->s_log_cluster_size
) -
3641 le32_to_cpu(es
->s_log_block_size
);
3642 sbi
->s_clusters_per_group
=
3643 le32_to_cpu(es
->s_clusters_per_group
);
3644 if (sbi
->s_clusters_per_group
> blocksize
* 8) {
3645 ext4_msg(sb
, KERN_ERR
,
3646 "#clusters per group too big: %lu",
3647 sbi
->s_clusters_per_group
);
3650 if (sbi
->s_blocks_per_group
!=
3651 (sbi
->s_clusters_per_group
* (clustersize
/ blocksize
))) {
3652 ext4_msg(sb
, KERN_ERR
, "blocks per group (%lu) and "
3653 "clusters per group (%lu) inconsistent",
3654 sbi
->s_blocks_per_group
,
3655 sbi
->s_clusters_per_group
);
3659 if (clustersize
!= blocksize
) {
3660 ext4_warning(sb
, "fragment/cluster size (%d) != "
3661 "block size (%d)", clustersize
,
3663 clustersize
= blocksize
;
3665 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
3666 ext4_msg(sb
, KERN_ERR
,
3667 "#blocks per group too big: %lu",
3668 sbi
->s_blocks_per_group
);
3671 sbi
->s_clusters_per_group
= sbi
->s_blocks_per_group
;
3672 sbi
->s_cluster_bits
= 0;
3674 sbi
->s_cluster_ratio
= clustersize
/ blocksize
;
3676 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
3677 ext4_msg(sb
, KERN_ERR
,
3678 "#inodes per group too big: %lu",
3679 sbi
->s_inodes_per_group
);
3683 /* Do we have standard group size of clustersize * 8 blocks ? */
3684 if (sbi
->s_blocks_per_group
== clustersize
<< 3)
3685 set_opt2(sb
, STD_GROUP_SIZE
);
3688 * Test whether we have more sectors than will fit in sector_t,
3689 * and whether the max offset is addressable by the page cache.
3691 err
= generic_check_addressable(sb
->s_blocksize_bits
,
3692 ext4_blocks_count(es
));
3694 ext4_msg(sb
, KERN_ERR
, "filesystem"
3695 " too large to mount safely on this system");
3696 if (sizeof(sector_t
) < 8)
3697 ext4_msg(sb
, KERN_WARNING
, "CONFIG_LBDAF not enabled");
3701 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
3704 /* check blocks count against device size */
3705 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
3706 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
3707 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
3708 "exceeds size of device (%llu blocks)",
3709 ext4_blocks_count(es
), blocks_count
);
3714 * It makes no sense for the first data block to be beyond the end
3715 * of the filesystem.
3717 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
3718 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data "
3719 "block %u is beyond end of filesystem (%llu)",
3720 le32_to_cpu(es
->s_first_data_block
),
3721 ext4_blocks_count(es
));
3724 blocks_count
= (ext4_blocks_count(es
) -
3725 le32_to_cpu(es
->s_first_data_block
) +
3726 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
3727 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
3728 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
3729 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %u "
3730 "(block count %llu, first data block %u, "
3731 "blocks per group %lu)", sbi
->s_groups_count
,
3732 ext4_blocks_count(es
),
3733 le32_to_cpu(es
->s_first_data_block
),
3734 EXT4_BLOCKS_PER_GROUP(sb
));
3737 sbi
->s_groups_count
= blocks_count
;
3738 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
3739 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
3740 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
3741 EXT4_DESC_PER_BLOCK(sb
);
3742 sbi
->s_group_desc
= ext4_kvmalloc(db_count
*
3743 sizeof(struct buffer_head
*),
3745 if (sbi
->s_group_desc
== NULL
) {
3746 ext4_msg(sb
, KERN_ERR
, "not enough memory");
3752 sbi
->s_proc
= proc_mkdir(sb
->s_id
, ext4_proc_root
);
3755 proc_create_data("options", S_IRUGO
, sbi
->s_proc
,
3756 &ext4_seq_options_fops
, sb
);
3758 bgl_lock_init(sbi
->s_blockgroup_lock
);
3760 for (i
= 0; i
< db_count
; i
++) {
3761 block
= descriptor_loc(sb
, logical_sb_block
, i
);
3762 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
3763 if (!sbi
->s_group_desc
[i
]) {
3764 ext4_msg(sb
, KERN_ERR
,
3765 "can't read group descriptor %d", i
);
3770 if (!ext4_check_descriptors(sb
, &first_not_zeroed
)) {
3771 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
3774 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
3775 if (!ext4_fill_flex_info(sb
)) {
3776 ext4_msg(sb
, KERN_ERR
,
3777 "unable to initialize "
3778 "flex_bg meta info!");
3782 sbi
->s_gdb_count
= db_count
;
3783 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
3784 spin_lock_init(&sbi
->s_next_gen_lock
);
3786 init_timer(&sbi
->s_err_report
);
3787 sbi
->s_err_report
.function
= print_daily_error_info
;
3788 sbi
->s_err_report
.data
= (unsigned long) sb
;
3790 /* Register extent status tree shrinker */
3791 ext4_es_register_shrinker(sb
);
3793 err
= percpu_counter_init(&sbi
->s_freeclusters_counter
,
3794 ext4_count_free_clusters(sb
));
3796 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
,
3797 ext4_count_free_inodes(sb
));
3800 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
3801 ext4_count_dirs(sb
));
3804 err
= percpu_counter_init(&sbi
->s_dirtyclusters_counter
, 0);
3807 err
= percpu_counter_init(&sbi
->s_extent_cache_cnt
, 0);
3810 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
3814 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
3815 sbi
->s_max_writeback_mb_bump
= 128;
3816 sbi
->s_extent_max_zeroout_kb
= 32;
3819 * set up enough so that it can read an inode
3821 if (!test_opt(sb
, NOLOAD
) &&
3822 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
3823 sb
->s_op
= &ext4_sops
;
3825 sb
->s_op
= &ext4_nojournal_sops
;
3826 sb
->s_export_op
= &ext4_export_ops
;
3827 sb
->s_xattr
= ext4_xattr_handlers
;
3829 sb
->dq_op
= &ext4_quota_operations
;
3830 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
))
3831 sb
->s_qcop
= &ext4_qctl_sysfile_operations
;
3833 sb
->s_qcop
= &ext4_qctl_operations
;
3835 memcpy(sb
->s_uuid
, es
->s_uuid
, sizeof(es
->s_uuid
));
3837 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
3838 mutex_init(&sbi
->s_orphan_lock
);
3842 needs_recovery
= (es
->s_last_orphan
!= 0 ||
3843 EXT4_HAS_INCOMPAT_FEATURE(sb
,
3844 EXT4_FEATURE_INCOMPAT_RECOVER
));
3846 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_MMP
) &&
3847 !(sb
->s_flags
& MS_RDONLY
))
3848 if (ext4_multi_mount_protect(sb
, le64_to_cpu(es
->s_mmp_block
)))
3852 * The first inode we look at is the journal inode. Don't try
3853 * root first: it may be modified in the journal!
3855 if (!test_opt(sb
, NOLOAD
) &&
3856 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
3857 if (ext4_load_journal(sb
, es
, journal_devnum
))
3859 } else if (test_opt(sb
, NOLOAD
) && !(sb
->s_flags
& MS_RDONLY
) &&
3860 EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
3861 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
3862 "suppressed and not mounted read-only");
3863 goto failed_mount_wq
;
3865 clear_opt(sb
, DATA_FLAGS
);
3866 sbi
->s_journal
= NULL
;
3871 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
) &&
3872 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
3873 JBD2_FEATURE_INCOMPAT_64BIT
)) {
3874 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
3875 goto failed_mount_wq
;
3878 if (!set_journal_csum_feature_set(sb
)) {
3879 ext4_msg(sb
, KERN_ERR
, "Failed to set journal checksum "
3881 goto failed_mount_wq
;
3884 /* We have now updated the journal if required, so we can
3885 * validate the data journaling mode. */
3886 switch (test_opt(sb
, DATA_FLAGS
)) {
3888 /* No mode set, assume a default based on the journal
3889 * capabilities: ORDERED_DATA if the journal can
3890 * cope, else JOURNAL_DATA
3892 if (jbd2_journal_check_available_features
3893 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
3894 set_opt(sb
, ORDERED_DATA
);
3896 set_opt(sb
, JOURNAL_DATA
);
3899 case EXT4_MOUNT_ORDERED_DATA
:
3900 case EXT4_MOUNT_WRITEBACK_DATA
:
3901 if (!jbd2_journal_check_available_features
3902 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
3903 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
3904 "requested data journaling mode");
3905 goto failed_mount_wq
;
3910 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
3912 sbi
->s_journal
->j_commit_callback
= ext4_journal_commit_callback
;
3915 * The journal may have updated the bg summary counts, so we
3916 * need to update the global counters.
3918 percpu_counter_set(&sbi
->s_freeclusters_counter
,
3919 ext4_count_free_clusters(sb
));
3920 percpu_counter_set(&sbi
->s_freeinodes_counter
,
3921 ext4_count_free_inodes(sb
));
3922 percpu_counter_set(&sbi
->s_dirs_counter
,
3923 ext4_count_dirs(sb
));
3924 percpu_counter_set(&sbi
->s_dirtyclusters_counter
, 0);
3928 * Get the # of file system overhead blocks from the
3929 * superblock if present.
3931 if (es
->s_overhead_clusters
)
3932 sbi
->s_overhead
= le32_to_cpu(es
->s_overhead_clusters
);
3934 err
= ext4_calculate_overhead(sb
);
3936 goto failed_mount_wq
;
3940 * The maximum number of concurrent works can be high and
3941 * concurrency isn't really necessary. Limit it to 1.
3943 EXT4_SB(sb
)->dio_unwritten_wq
=
3944 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM
| WQ_UNBOUND
, 1);
3945 if (!EXT4_SB(sb
)->dio_unwritten_wq
) {
3946 printk(KERN_ERR
"EXT4-fs: failed to create DIO workqueue\n");
3948 goto failed_mount_wq
;
3952 * The jbd2_journal_load will have done any necessary log recovery,
3953 * so we can safely mount the rest of the filesystem now.
3956 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
3958 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
3959 ret
= PTR_ERR(root
);
3963 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
3964 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
3968 sb
->s_root
= d_make_root(root
);
3970 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
3975 if (ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
))
3976 sb
->s_flags
|= MS_RDONLY
;
3978 /* determine the minimum size of new large inodes, if present */
3979 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
3980 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
3981 EXT4_GOOD_OLD_INODE_SIZE
;
3982 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3983 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE
)) {
3984 if (sbi
->s_want_extra_isize
<
3985 le16_to_cpu(es
->s_want_extra_isize
))
3986 sbi
->s_want_extra_isize
=
3987 le16_to_cpu(es
->s_want_extra_isize
);
3988 if (sbi
->s_want_extra_isize
<
3989 le16_to_cpu(es
->s_min_extra_isize
))
3990 sbi
->s_want_extra_isize
=
3991 le16_to_cpu(es
->s_min_extra_isize
);
3994 /* Check if enough inode space is available */
3995 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
3996 sbi
->s_inode_size
) {
3997 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
3998 EXT4_GOOD_OLD_INODE_SIZE
;
3999 ext4_msg(sb
, KERN_INFO
, "required extra inode space not"
4003 err
= ext4_reserve_clusters(sbi
, ext4_calculate_resv_clusters(sb
));
4005 ext4_msg(sb
, KERN_ERR
, "failed to reserve %llu clusters for "
4006 "reserved pool", ext4_calculate_resv_clusters(sb
));
4007 goto failed_mount4a
;
4010 err
= ext4_setup_system_zone(sb
);
4012 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
4014 goto failed_mount4a
;
4018 err
= ext4_mb_init(sb
);
4020 ext4_msg(sb
, KERN_ERR
, "failed to initialize mballoc (%d)",
4025 err
= ext4_register_li_request(sb
, first_not_zeroed
);
4029 sbi
->s_kobj
.kset
= ext4_kset
;
4030 init_completion(&sbi
->s_kobj_unregister
);
4031 err
= kobject_init_and_add(&sbi
->s_kobj
, &ext4_ktype
, NULL
,
4037 /* Enable quota usage during mount. */
4038 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
) &&
4039 !(sb
->s_flags
& MS_RDONLY
)) {
4040 err
= ext4_enable_quotas(sb
);
4044 #endif /* CONFIG_QUOTA */
4046 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
4047 ext4_orphan_cleanup(sb
, es
);
4048 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
4049 if (needs_recovery
) {
4050 ext4_msg(sb
, KERN_INFO
, "recovery complete");
4051 ext4_mark_recovery_complete(sb
, es
);
4053 if (EXT4_SB(sb
)->s_journal
) {
4054 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
4055 descr
= " journalled data mode";
4056 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
4057 descr
= " ordered data mode";
4059 descr
= " writeback data mode";
4061 descr
= "out journal";
4063 if (test_opt(sb
, DISCARD
)) {
4064 struct request_queue
*q
= bdev_get_queue(sb
->s_bdev
);
4065 if (!blk_queue_discard(q
))
4066 ext4_msg(sb
, KERN_WARNING
,
4067 "mounting with \"discard\" option, but "
4068 "the device does not support discard");
4071 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s. "
4072 "Opts: %s%s%s", descr
, sbi
->s_es
->s_mount_opts
,
4073 *sbi
->s_es
->s_mount_opts
? "; " : "", orig_data
);
4075 if (es
->s_error_count
)
4076 mod_timer(&sbi
->s_err_report
, jiffies
+ 300*HZ
); /* 5 minutes */
4083 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
4088 kobject_del(&sbi
->s_kobj
);
4091 ext4_unregister_li_request(sb
);
4093 ext4_mb_release(sb
);
4095 ext4_ext_release(sb
);
4096 ext4_release_system_zone(sb
);
4101 ext4_msg(sb
, KERN_ERR
, "mount failed");
4102 destroy_workqueue(EXT4_SB(sb
)->dio_unwritten_wq
);
4104 if (sbi
->s_journal
) {
4105 jbd2_journal_destroy(sbi
->s_journal
);
4106 sbi
->s_journal
= NULL
;
4109 ext4_es_unregister_shrinker(sb
);
4110 del_timer(&sbi
->s_err_report
);
4111 if (sbi
->s_flex_groups
)
4112 ext4_kvfree(sbi
->s_flex_groups
);
4113 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
4114 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
4115 percpu_counter_destroy(&sbi
->s_dirs_counter
);
4116 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
4117 percpu_counter_destroy(&sbi
->s_extent_cache_cnt
);
4119 kthread_stop(sbi
->s_mmp_tsk
);
4121 for (i
= 0; i
< db_count
; i
++)
4122 brelse(sbi
->s_group_desc
[i
]);
4123 ext4_kvfree(sbi
->s_group_desc
);
4125 if (sbi
->s_chksum_driver
)
4126 crypto_free_shash(sbi
->s_chksum_driver
);
4128 remove_proc_entry("options", sbi
->s_proc
);
4129 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
4132 for (i
= 0; i
< MAXQUOTAS
; i
++)
4133 kfree(sbi
->s_qf_names
[i
]);
4135 ext4_blkdev_remove(sbi
);
4138 sb
->s_fs_info
= NULL
;
4139 kfree(sbi
->s_blockgroup_lock
);
4143 return err
? err
: ret
;
4147 * Setup any per-fs journal parameters now. We'll do this both on
4148 * initial mount, once the journal has been initialised but before we've
4149 * done any recovery; and again on any subsequent remount.
4151 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
4153 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4155 journal
->j_commit_interval
= sbi
->s_commit_interval
;
4156 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
4157 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
4159 write_lock(&journal
->j_state_lock
);
4160 if (test_opt(sb
, BARRIER
))
4161 journal
->j_flags
|= JBD2_BARRIER
;
4163 journal
->j_flags
&= ~JBD2_BARRIER
;
4164 if (test_opt(sb
, DATA_ERR_ABORT
))
4165 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
4167 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
4168 write_unlock(&journal
->j_state_lock
);
4171 static journal_t
*ext4_get_journal(struct super_block
*sb
,
4172 unsigned int journal_inum
)
4174 struct inode
*journal_inode
;
4177 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4179 /* First, test for the existence of a valid inode on disk. Bad
4180 * things happen if we iget() an unused inode, as the subsequent
4181 * iput() will try to delete it. */
4183 journal_inode
= ext4_iget(sb
, journal_inum
);
4184 if (IS_ERR(journal_inode
)) {
4185 ext4_msg(sb
, KERN_ERR
, "no journal found");
4188 if (!journal_inode
->i_nlink
) {
4189 make_bad_inode(journal_inode
);
4190 iput(journal_inode
);
4191 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
4195 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4196 journal_inode
, journal_inode
->i_size
);
4197 if (!S_ISREG(journal_inode
->i_mode
)) {
4198 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
4199 iput(journal_inode
);
4203 journal
= jbd2_journal_init_inode(journal_inode
);
4205 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
4206 iput(journal_inode
);
4209 journal
->j_private
= sb
;
4210 ext4_init_journal_params(sb
, journal
);
4214 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
4217 struct buffer_head
*bh
;
4221 int hblock
, blocksize
;
4222 ext4_fsblk_t sb_block
;
4223 unsigned long offset
;
4224 struct ext4_super_block
*es
;
4225 struct block_device
*bdev
;
4227 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4229 bdev
= ext4_blkdev_get(j_dev
, sb
);
4233 blocksize
= sb
->s_blocksize
;
4234 hblock
= bdev_logical_block_size(bdev
);
4235 if (blocksize
< hblock
) {
4236 ext4_msg(sb
, KERN_ERR
,
4237 "blocksize too small for journal device");
4241 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
4242 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
4243 set_blocksize(bdev
, blocksize
);
4244 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
4245 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
4246 "external journal");
4250 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
4251 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
4252 !(le32_to_cpu(es
->s_feature_incompat
) &
4253 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
4254 ext4_msg(sb
, KERN_ERR
, "external journal has "
4260 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
4261 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
4266 len
= ext4_blocks_count(es
);
4267 start
= sb_block
+ 1;
4268 brelse(bh
); /* we're done with the superblock */
4270 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
4271 start
, len
, blocksize
);
4273 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
4276 journal
->j_private
= sb
;
4277 ll_rw_block(READ
| REQ_META
| REQ_PRIO
, 1, &journal
->j_sb_buffer
);
4278 wait_on_buffer(journal
->j_sb_buffer
);
4279 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
4280 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
4283 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
4284 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
4285 "user (unsupported) - %d",
4286 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
4289 EXT4_SB(sb
)->journal_bdev
= bdev
;
4290 ext4_init_journal_params(sb
, journal
);
4294 jbd2_journal_destroy(journal
);
4296 ext4_blkdev_put(bdev
);
4300 static int ext4_load_journal(struct super_block
*sb
,
4301 struct ext4_super_block
*es
,
4302 unsigned long journal_devnum
)
4305 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
4308 int really_read_only
;
4310 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4312 if (journal_devnum
&&
4313 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4314 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
4315 "numbers have changed");
4316 journal_dev
= new_decode_dev(journal_devnum
);
4318 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
4320 really_read_only
= bdev_read_only(sb
->s_bdev
);
4323 * Are we loading a blank journal or performing recovery after a
4324 * crash? For recovery, we need to check in advance whether we
4325 * can get read-write access to the device.
4327 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
4328 if (sb
->s_flags
& MS_RDONLY
) {
4329 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
4330 "required on readonly filesystem");
4331 if (really_read_only
) {
4332 ext4_msg(sb
, KERN_ERR
, "write access "
4333 "unavailable, cannot proceed");
4336 ext4_msg(sb
, KERN_INFO
, "write access will "
4337 "be enabled during recovery");
4341 if (journal_inum
&& journal_dev
) {
4342 ext4_msg(sb
, KERN_ERR
, "filesystem has both journal "
4343 "and inode journals!");
4348 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
4351 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
4355 if (!(journal
->j_flags
& JBD2_BARRIER
))
4356 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
4358 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
))
4359 err
= jbd2_journal_wipe(journal
, !really_read_only
);
4361 char *save
= kmalloc(EXT4_S_ERR_LEN
, GFP_KERNEL
);
4363 memcpy(save
, ((char *) es
) +
4364 EXT4_S_ERR_START
, EXT4_S_ERR_LEN
);
4365 err
= jbd2_journal_load(journal
);
4367 memcpy(((char *) es
) + EXT4_S_ERR_START
,
4368 save
, EXT4_S_ERR_LEN
);
4373 ext4_msg(sb
, KERN_ERR
, "error loading journal");
4374 jbd2_journal_destroy(journal
);
4378 EXT4_SB(sb
)->s_journal
= journal
;
4379 ext4_clear_journal_err(sb
, es
);
4381 if (!really_read_only
&& journal_devnum
&&
4382 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4383 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
4385 /* Make sure we flush the recovery flag to disk. */
4386 ext4_commit_super(sb
, 1);
4392 static int ext4_commit_super(struct super_block
*sb
, int sync
)
4394 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
4395 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
4398 if (!sbh
|| block_device_ejected(sb
))
4400 if (buffer_write_io_error(sbh
)) {
4402 * Oh, dear. A previous attempt to write the
4403 * superblock failed. This could happen because the
4404 * USB device was yanked out. Or it could happen to
4405 * be a transient write error and maybe the block will
4406 * be remapped. Nothing we can do but to retry the
4407 * write and hope for the best.
4409 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
4410 "superblock detected");
4411 clear_buffer_write_io_error(sbh
);
4412 set_buffer_uptodate(sbh
);
4415 * If the file system is mounted read-only, don't update the
4416 * superblock write time. This avoids updating the superblock
4417 * write time when we are mounting the root file system
4418 * read/only but we need to replay the journal; at that point,
4419 * for people who are east of GMT and who make their clock
4420 * tick in localtime for Windows bug-for-bug compatibility,
4421 * the clock is set in the future, and this will cause e2fsck
4422 * to complain and force a full file system check.
4424 if (!(sb
->s_flags
& MS_RDONLY
))
4425 es
->s_wtime
= cpu_to_le32(get_seconds());
4426 if (sb
->s_bdev
->bd_part
)
4427 es
->s_kbytes_written
=
4428 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
+
4429 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
4430 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
4432 es
->s_kbytes_written
=
4433 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
);
4434 ext4_free_blocks_count_set(es
,
4435 EXT4_C2B(EXT4_SB(sb
), percpu_counter_sum_positive(
4436 &EXT4_SB(sb
)->s_freeclusters_counter
)));
4437 es
->s_free_inodes_count
=
4438 cpu_to_le32(percpu_counter_sum_positive(
4439 &EXT4_SB(sb
)->s_freeinodes_counter
));
4440 BUFFER_TRACE(sbh
, "marking dirty");
4441 ext4_superblock_csum_set(sb
);
4442 mark_buffer_dirty(sbh
);
4444 error
= sync_dirty_buffer(sbh
);
4448 error
= buffer_write_io_error(sbh
);
4450 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
4452 clear_buffer_write_io_error(sbh
);
4453 set_buffer_uptodate(sbh
);
4460 * Have we just finished recovery? If so, and if we are mounting (or
4461 * remounting) the filesystem readonly, then we will end up with a
4462 * consistent fs on disk. Record that fact.
4464 static void ext4_mark_recovery_complete(struct super_block
*sb
,
4465 struct ext4_super_block
*es
)
4467 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
4469 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
4470 BUG_ON(journal
!= NULL
);
4473 jbd2_journal_lock_updates(journal
);
4474 if (jbd2_journal_flush(journal
) < 0)
4477 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
) &&
4478 sb
->s_flags
& MS_RDONLY
) {
4479 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4480 ext4_commit_super(sb
, 1);
4484 jbd2_journal_unlock_updates(journal
);
4488 * If we are mounting (or read-write remounting) a filesystem whose journal
4489 * has recorded an error from a previous lifetime, move that error to the
4490 * main filesystem now.
4492 static void ext4_clear_journal_err(struct super_block
*sb
,
4493 struct ext4_super_block
*es
)
4499 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4501 journal
= EXT4_SB(sb
)->s_journal
;
4504 * Now check for any error status which may have been recorded in the
4505 * journal by a prior ext4_error() or ext4_abort()
4508 j_errno
= jbd2_journal_errno(journal
);
4512 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
4513 ext4_warning(sb
, "Filesystem error recorded "
4514 "from previous mount: %s", errstr
);
4515 ext4_warning(sb
, "Marking fs in need of filesystem check.");
4517 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
4518 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
4519 ext4_commit_super(sb
, 1);
4521 jbd2_journal_clear_err(journal
);
4522 jbd2_journal_update_sb_errno(journal
);
4527 * Force the running and committing transactions to commit,
4528 * and wait on the commit.
4530 int ext4_force_commit(struct super_block
*sb
)
4534 if (sb
->s_flags
& MS_RDONLY
)
4537 journal
= EXT4_SB(sb
)->s_journal
;
4538 return ext4_journal_force_commit(journal
);
4541 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
4545 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4547 trace_ext4_sync_fs(sb
, wait
);
4548 flush_workqueue(sbi
->dio_unwritten_wq
);
4550 * Writeback quota in non-journalled quota case - journalled quota has
4553 dquot_writeback_dquots(sb
, -1);
4554 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
4556 jbd2_log_wait_commit(sbi
->s_journal
, target
);
4562 * LVM calls this function before a (read-only) snapshot is created. This
4563 * gives us a chance to flush the journal completely and mark the fs clean.
4565 * Note that only this function cannot bring a filesystem to be in a clean
4566 * state independently. It relies on upper layer to stop all data & metadata
4569 static int ext4_freeze(struct super_block
*sb
)
4574 if (sb
->s_flags
& MS_RDONLY
)
4577 journal
= EXT4_SB(sb
)->s_journal
;
4579 /* Now we set up the journal barrier. */
4580 jbd2_journal_lock_updates(journal
);
4583 * Don't clear the needs_recovery flag if we failed to flush
4586 error
= jbd2_journal_flush(journal
);
4590 /* Journal blocked and flushed, clear needs_recovery flag. */
4591 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4592 error
= ext4_commit_super(sb
, 1);
4594 /* we rely on upper layer to stop further updates */
4595 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
4600 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4601 * flag here, even though the filesystem is not technically dirty yet.
4603 static int ext4_unfreeze(struct super_block
*sb
)
4605 if (sb
->s_flags
& MS_RDONLY
)
4608 /* Reset the needs_recovery flag before the fs is unlocked. */
4609 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4610 ext4_commit_super(sb
, 1);
4615 * Structure to save mount options for ext4_remount's benefit
4617 struct ext4_mount_options
{
4618 unsigned long s_mount_opt
;
4619 unsigned long s_mount_opt2
;
4622 unsigned long s_commit_interval
;
4623 u32 s_min_batch_time
, s_max_batch_time
;
4626 char *s_qf_names
[MAXQUOTAS
];
4630 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
4632 struct ext4_super_block
*es
;
4633 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4634 unsigned long old_sb_flags
;
4635 struct ext4_mount_options old_opts
;
4636 int enable_quota
= 0;
4638 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
4643 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
4645 /* Store the original options */
4646 old_sb_flags
= sb
->s_flags
;
4647 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
4648 old_opts
.s_mount_opt2
= sbi
->s_mount_opt2
;
4649 old_opts
.s_resuid
= sbi
->s_resuid
;
4650 old_opts
.s_resgid
= sbi
->s_resgid
;
4651 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
4652 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
4653 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
4655 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
4656 for (i
= 0; i
< MAXQUOTAS
; i
++)
4657 if (sbi
->s_qf_names
[i
]) {
4658 old_opts
.s_qf_names
[i
] = kstrdup(sbi
->s_qf_names
[i
],
4660 if (!old_opts
.s_qf_names
[i
]) {
4661 for (j
= 0; j
< i
; j
++)
4662 kfree(old_opts
.s_qf_names
[j
]);
4667 old_opts
.s_qf_names
[i
] = NULL
;
4669 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
4670 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
4673 * Allow the "check" option to be passed as a remount option.
4675 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
, 1)) {
4680 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
4681 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
4682 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4683 "both data=journal and delalloc");
4687 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
4688 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4689 "both data=journal and dioread_nolock");
4695 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
)
4696 ext4_abort(sb
, "Abort forced by user");
4698 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
4699 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
4703 if (sbi
->s_journal
) {
4704 ext4_init_journal_params(sb
, sbi
->s_journal
);
4705 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
4708 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
)) {
4709 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
) {
4714 if (*flags
& MS_RDONLY
) {
4715 err
= dquot_suspend(sb
, -1);
4720 * First of all, the unconditional stuff we have to do
4721 * to disable replay of the journal when we next remount
4723 sb
->s_flags
|= MS_RDONLY
;
4726 * OK, test if we are remounting a valid rw partition
4727 * readonly, and if so set the rdonly flag and then
4728 * mark the partition as valid again.
4730 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
4731 (sbi
->s_mount_state
& EXT4_VALID_FS
))
4732 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
4735 ext4_mark_recovery_complete(sb
, es
);
4737 /* Make sure we can mount this feature set readwrite */
4738 if (!ext4_feature_set_ok(sb
, 0)) {
4743 * Make sure the group descriptor checksums
4744 * are sane. If they aren't, refuse to remount r/w.
4746 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
4747 struct ext4_group_desc
*gdp
=
4748 ext4_get_group_desc(sb
, g
, NULL
);
4750 if (!ext4_group_desc_csum_verify(sb
, g
, gdp
)) {
4751 ext4_msg(sb
, KERN_ERR
,
4752 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4753 g
, le16_to_cpu(ext4_group_desc_csum(sbi
, g
, gdp
)),
4754 le16_to_cpu(gdp
->bg_checksum
));
4761 * If we have an unprocessed orphan list hanging
4762 * around from a previously readonly bdev mount,
4763 * require a full umount/remount for now.
4765 if (es
->s_last_orphan
) {
4766 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
4767 "remount RDWR because of unprocessed "
4768 "orphan inode list. Please "
4769 "umount/remount instead");
4775 * Mounting a RDONLY partition read-write, so reread
4776 * and store the current valid flag. (It may have
4777 * been changed by e2fsck since we originally mounted
4781 ext4_clear_journal_err(sb
, es
);
4782 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
4783 if (!ext4_setup_super(sb
, es
, 0))
4784 sb
->s_flags
&= ~MS_RDONLY
;
4785 if (EXT4_HAS_INCOMPAT_FEATURE(sb
,
4786 EXT4_FEATURE_INCOMPAT_MMP
))
4787 if (ext4_multi_mount_protect(sb
,
4788 le64_to_cpu(es
->s_mmp_block
))) {
4797 * Reinitialize lazy itable initialization thread based on
4800 if ((sb
->s_flags
& MS_RDONLY
) || !test_opt(sb
, INIT_INODE_TABLE
))
4801 ext4_unregister_li_request(sb
);
4803 ext4_group_t first_not_zeroed
;
4804 first_not_zeroed
= ext4_has_uninit_itable(sb
);
4805 ext4_register_li_request(sb
, first_not_zeroed
);
4808 ext4_setup_system_zone(sb
);
4809 if (sbi
->s_journal
== NULL
&& !(old_sb_flags
& MS_RDONLY
))
4810 ext4_commit_super(sb
, 1);
4813 /* Release old quota file names */
4814 for (i
= 0; i
< MAXQUOTAS
; i
++)
4815 kfree(old_opts
.s_qf_names
[i
]);
4817 if (sb_any_quota_suspended(sb
))
4818 dquot_resume(sb
, -1);
4819 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
4820 EXT4_FEATURE_RO_COMPAT_QUOTA
)) {
4821 err
= ext4_enable_quotas(sb
);
4828 ext4_msg(sb
, KERN_INFO
, "re-mounted. Opts: %s", orig_data
);
4833 sb
->s_flags
= old_sb_flags
;
4834 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
4835 sbi
->s_mount_opt2
= old_opts
.s_mount_opt2
;
4836 sbi
->s_resuid
= old_opts
.s_resuid
;
4837 sbi
->s_resgid
= old_opts
.s_resgid
;
4838 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
4839 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
4840 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
4842 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
4843 for (i
= 0; i
< MAXQUOTAS
; i
++) {
4844 kfree(sbi
->s_qf_names
[i
]);
4845 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
4852 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
4854 struct super_block
*sb
= dentry
->d_sb
;
4855 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4856 struct ext4_super_block
*es
= sbi
->s_es
;
4857 ext4_fsblk_t overhead
= 0, resv_blocks
;
4860 resv_blocks
= EXT4_C2B(sbi
, atomic64_read(&sbi
->s_resv_clusters
));
4862 if (!test_opt(sb
, MINIX_DF
))
4863 overhead
= sbi
->s_overhead
;
4865 buf
->f_type
= EXT4_SUPER_MAGIC
;
4866 buf
->f_bsize
= sb
->s_blocksize
;
4867 buf
->f_blocks
= ext4_blocks_count(es
) - EXT4_C2B(sbi
, overhead
);
4868 bfree
= percpu_counter_sum_positive(&sbi
->s_freeclusters_counter
) -
4869 percpu_counter_sum_positive(&sbi
->s_dirtyclusters_counter
);
4870 /* prevent underflow in case that few free space is available */
4871 buf
->f_bfree
= EXT4_C2B(sbi
, max_t(s64
, bfree
, 0));
4872 buf
->f_bavail
= buf
->f_bfree
-
4873 (ext4_r_blocks_count(es
) + resv_blocks
);
4874 if (buf
->f_bfree
< (ext4_r_blocks_count(es
) + resv_blocks
))
4876 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
4877 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
4878 buf
->f_namelen
= EXT4_NAME_LEN
;
4879 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
4880 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
4881 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
4882 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
4887 /* Helper function for writing quotas on sync - we need to start transaction
4888 * before quota file is locked for write. Otherwise the are possible deadlocks:
4889 * Process 1 Process 2
4890 * ext4_create() quota_sync()
4891 * jbd2_journal_start() write_dquot()
4892 * dquot_initialize() down(dqio_mutex)
4893 * down(dqio_mutex) jbd2_journal_start()
4899 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
4901 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_id
.type
];
4904 static int ext4_write_dquot(struct dquot
*dquot
)
4908 struct inode
*inode
;
4910 inode
= dquot_to_inode(dquot
);
4911 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
,
4912 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
4914 return PTR_ERR(handle
);
4915 ret
= dquot_commit(dquot
);
4916 err
= ext4_journal_stop(handle
);
4922 static int ext4_acquire_dquot(struct dquot
*dquot
)
4927 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
4928 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
4930 return PTR_ERR(handle
);
4931 ret
= dquot_acquire(dquot
);
4932 err
= ext4_journal_stop(handle
);
4938 static int ext4_release_dquot(struct dquot
*dquot
)
4943 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
4944 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
4945 if (IS_ERR(handle
)) {
4946 /* Release dquot anyway to avoid endless cycle in dqput() */
4947 dquot_release(dquot
);
4948 return PTR_ERR(handle
);
4950 ret
= dquot_release(dquot
);
4951 err
= ext4_journal_stop(handle
);
4957 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
4959 struct super_block
*sb
= dquot
->dq_sb
;
4960 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4962 /* Are we journaling quotas? */
4963 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
) ||
4964 sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
4965 dquot_mark_dquot_dirty(dquot
);
4966 return ext4_write_dquot(dquot
);
4968 return dquot_mark_dquot_dirty(dquot
);
4972 static int ext4_write_info(struct super_block
*sb
, int type
)
4977 /* Data block + inode block */
4978 handle
= ext4_journal_start(sb
->s_root
->d_inode
, EXT4_HT_QUOTA
, 2);
4980 return PTR_ERR(handle
);
4981 ret
= dquot_commit_info(sb
, type
);
4982 err
= ext4_journal_stop(handle
);
4989 * Turn on quotas during mount time - we need to find
4990 * the quota file and such...
4992 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
4994 return dquot_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
4995 EXT4_SB(sb
)->s_jquota_fmt
, type
);
4998 static void lockdep_set_quota_inode(struct inode
*inode
, int subclass
)
5000 struct ext4_inode_info
*ei
= EXT4_I(inode
);
5002 /* The first argument of lockdep_set_subclass has to be
5003 * *exactly* the same as the argument to init_rwsem() --- in
5004 * this case, in init_once() --- or lockdep gets unhappy
5005 * because the name of the lock is set using the
5006 * stringification of the argument to init_rwsem().
5008 (void) ei
; /* shut up clang warning if !CONFIG_LOCKDEP */
5009 lockdep_set_subclass(&ei
->i_data_sem
, subclass
);
5013 * Standard function to be called on quota_on
5015 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
5020 if (!test_opt(sb
, QUOTA
))
5023 /* Quotafile not on the same filesystem? */
5024 if (path
->dentry
->d_sb
!= sb
)
5026 /* Journaling quota? */
5027 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
5028 /* Quotafile not in fs root? */
5029 if (path
->dentry
->d_parent
!= sb
->s_root
)
5030 ext4_msg(sb
, KERN_WARNING
,
5031 "Quota file not on filesystem root. "
5032 "Journaled quota will not work");
5036 * When we journal data on quota file, we have to flush journal to see
5037 * all updates to the file when we bypass pagecache...
5039 if (EXT4_SB(sb
)->s_journal
&&
5040 ext4_should_journal_data(path
->dentry
->d_inode
)) {
5042 * We don't need to lock updates but journal_flush() could
5043 * otherwise be livelocked...
5045 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
5046 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
5047 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
5051 lockdep_set_quota_inode(path
->dentry
->d_inode
, I_DATA_SEM_QUOTA
);
5052 err
= dquot_quota_on(sb
, type
, format_id
, path
);
5054 lockdep_set_quota_inode(path
->dentry
->d_inode
,
5059 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
5063 struct inode
*qf_inode
;
5064 unsigned long qf_inums
[MAXQUOTAS
] = {
5065 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
5066 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
)
5069 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
));
5071 if (!qf_inums
[type
])
5074 qf_inode
= ext4_iget(sb
, qf_inums
[type
]);
5075 if (IS_ERR(qf_inode
)) {
5076 ext4_error(sb
, "Bad quota inode # %lu", qf_inums
[type
]);
5077 return PTR_ERR(qf_inode
);
5080 /* Don't account quota for quota files to avoid recursion */
5081 qf_inode
->i_flags
|= S_NOQUOTA
;
5082 lockdep_set_quota_inode(qf_inode
, I_DATA_SEM_QUOTA
);
5083 err
= dquot_enable(qf_inode
, type
, format_id
, flags
);
5086 lockdep_set_quota_inode(qf_inode
, I_DATA_SEM_NORMAL
);
5091 /* Enable usage tracking for all quota types. */
5092 static int ext4_enable_quotas(struct super_block
*sb
)
5095 unsigned long qf_inums
[MAXQUOTAS
] = {
5096 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
5097 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
)
5100 sb_dqopt(sb
)->flags
|= DQUOT_QUOTA_SYS_FILE
;
5101 for (type
= 0; type
< MAXQUOTAS
; type
++) {
5102 if (qf_inums
[type
]) {
5103 err
= ext4_quota_enable(sb
, type
, QFMT_VFS_V1
,
5104 DQUOT_USAGE_ENABLED
);
5107 "Failed to enable quota tracking "
5108 "(type=%d, err=%d). Please run "
5109 "e2fsck to fix.", type
, err
);
5118 * quota_on function that is used when QUOTA feature is set.
5120 static int ext4_quota_on_sysfile(struct super_block
*sb
, int type
,
5123 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
))
5127 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5129 return ext4_quota_enable(sb
, type
, format_id
, DQUOT_LIMITS_ENABLED
);
5132 static int ext4_quota_off(struct super_block
*sb
, int type
)
5134 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5137 /* Force all delayed allocation blocks to be allocated.
5138 * Caller already holds s_umount sem */
5139 if (test_opt(sb
, DELALLOC
))
5140 sync_filesystem(sb
);
5145 /* Update modification times of quota files when userspace can
5146 * start looking at them */
5147 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
, 1);
5150 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
5151 ext4_mark_inode_dirty(handle
, inode
);
5152 ext4_journal_stop(handle
);
5155 return dquot_quota_off(sb
, type
);
5159 * quota_off function that is used when QUOTA feature is set.
5161 static int ext4_quota_off_sysfile(struct super_block
*sb
, int type
)
5163 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
))
5166 /* Disable only the limits. */
5167 return dquot_disable(sb
, type
, DQUOT_LIMITS_ENABLED
);
5170 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5171 * acquiring the locks... As quota files are never truncated and quota code
5172 * itself serializes the operations (and no one else should touch the files)
5173 * we don't have to be afraid of races */
5174 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
5175 size_t len
, loff_t off
)
5177 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5178 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5180 int offset
= off
& (sb
->s_blocksize
- 1);
5183 struct buffer_head
*bh
;
5184 loff_t i_size
= i_size_read(inode
);
5188 if (off
+len
> i_size
)
5191 while (toread
> 0) {
5192 tocopy
= sb
->s_blocksize
- offset
< toread
?
5193 sb
->s_blocksize
- offset
: toread
;
5194 bh
= ext4_bread(NULL
, inode
, blk
, 0, &err
);
5197 if (!bh
) /* A hole? */
5198 memset(data
, 0, tocopy
);
5200 memcpy(data
, bh
->b_data
+offset
, tocopy
);
5210 /* Write to quotafile (we know the transaction is already started and has
5211 * enough credits) */
5212 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
5213 const char *data
, size_t len
, loff_t off
)
5215 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5216 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5218 int offset
= off
& (sb
->s_blocksize
- 1);
5219 struct buffer_head
*bh
;
5220 handle_t
*handle
= journal_current_handle();
5222 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
5223 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5224 " cancelled because transaction is not started",
5225 (unsigned long long)off
, (unsigned long long)len
);
5229 * Since we account only one data block in transaction credits,
5230 * then it is impossible to cross a block boundary.
5232 if (sb
->s_blocksize
- offset
< len
) {
5233 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5234 " cancelled because not block aligned",
5235 (unsigned long long)off
, (unsigned long long)len
);
5239 bh
= ext4_bread(handle
, inode
, blk
, 1, &err
);
5242 err
= ext4_journal_get_write_access(handle
, bh
);
5248 memcpy(bh
->b_data
+offset
, data
, len
);
5249 flush_dcache_page(bh
->b_page
);
5251 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
5256 if (inode
->i_size
< off
+ len
) {
5257 i_size_write(inode
, off
+ len
);
5258 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
5259 ext4_mark_inode_dirty(handle
, inode
);
5266 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
5267 const char *dev_name
, void *data
)
5269 return mount_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
);
5272 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5273 static inline void register_as_ext2(void)
5275 int err
= register_filesystem(&ext2_fs_type
);
5278 "EXT4-fs: Unable to register as ext2 (%d)\n", err
);
5281 static inline void unregister_as_ext2(void)
5283 unregister_filesystem(&ext2_fs_type
);
5286 static inline int ext2_feature_set_ok(struct super_block
*sb
)
5288 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT2_FEATURE_INCOMPAT_SUPP
))
5290 if (sb
->s_flags
& MS_RDONLY
)
5292 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT2_FEATURE_RO_COMPAT_SUPP
))
5297 static inline void register_as_ext2(void) { }
5298 static inline void unregister_as_ext2(void) { }
5299 static inline int ext2_feature_set_ok(struct super_block
*sb
) { return 0; }
5302 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5303 static inline void register_as_ext3(void)
5305 int err
= register_filesystem(&ext3_fs_type
);
5308 "EXT4-fs: Unable to register as ext3 (%d)\n", err
);
5311 static inline void unregister_as_ext3(void)
5313 unregister_filesystem(&ext3_fs_type
);
5316 static inline int ext3_feature_set_ok(struct super_block
*sb
)
5318 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT3_FEATURE_INCOMPAT_SUPP
))
5320 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
5322 if (sb
->s_flags
& MS_RDONLY
)
5324 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT3_FEATURE_RO_COMPAT_SUPP
))
5329 static inline void register_as_ext3(void) { }
5330 static inline void unregister_as_ext3(void) { }
5331 static inline int ext3_feature_set_ok(struct super_block
*sb
) { return 0; }
5334 static struct file_system_type ext4_fs_type
= {
5335 .owner
= THIS_MODULE
,
5337 .mount
= ext4_mount
,
5338 .kill_sb
= kill_block_super
,
5339 .fs_flags
= FS_REQUIRES_DEV
,
5341 MODULE_ALIAS_FS("ext4");
5343 static int __init
ext4_init_feat_adverts(void)
5345 struct ext4_features
*ef
;
5348 ef
= kzalloc(sizeof(struct ext4_features
), GFP_KERNEL
);
5352 ef
->f_kobj
.kset
= ext4_kset
;
5353 init_completion(&ef
->f_kobj_unregister
);
5354 ret
= kobject_init_and_add(&ef
->f_kobj
, &ext4_feat_ktype
, NULL
,
5367 static void ext4_exit_feat_adverts(void)
5369 kobject_put(&ext4_feat
->f_kobj
);
5370 wait_for_completion(&ext4_feat
->f_kobj_unregister
);
5374 /* Shared across all ext4 file systems */
5375 wait_queue_head_t ext4__ioend_wq
[EXT4_WQ_HASH_SZ
];
5376 struct mutex ext4__aio_mutex
[EXT4_WQ_HASH_SZ
];
5378 static int __init
ext4_init_fs(void)
5382 ext4_li_info
= NULL
;
5383 mutex_init(&ext4_li_mtx
);
5385 /* Build-time check for flags consistency */
5386 ext4_check_flag_values();
5388 for (i
= 0; i
< EXT4_WQ_HASH_SZ
; i
++) {
5389 mutex_init(&ext4__aio_mutex
[i
]);
5390 init_waitqueue_head(&ext4__ioend_wq
[i
]);
5393 err
= ext4_init_es();
5397 err
= ext4_init_pageio();
5401 err
= ext4_init_system_zone();
5404 ext4_kset
= kset_create_and_add("ext4", NULL
, fs_kobj
);
5409 ext4_proc_root
= proc_mkdir("fs/ext4", NULL
);
5411 err
= ext4_init_feat_adverts();
5415 err
= ext4_init_mballoc();
5419 err
= ext4_init_xattr();
5422 err
= init_inodecache();
5427 err
= register_filesystem(&ext4_fs_type
);
5433 unregister_as_ext2();
5434 unregister_as_ext3();
5435 destroy_inodecache();
5439 ext4_exit_mballoc();
5441 ext4_exit_feat_adverts();
5444 remove_proc_entry("fs/ext4", NULL
);
5445 kset_unregister(ext4_kset
);
5447 ext4_exit_system_zone();
5456 static void __exit
ext4_exit_fs(void)
5458 ext4_destroy_lazyinit_thread();
5459 unregister_as_ext2();
5460 unregister_as_ext3();
5461 unregister_filesystem(&ext4_fs_type
);
5462 destroy_inodecache();
5464 ext4_exit_mballoc();
5465 ext4_exit_feat_adverts();
5466 remove_proc_entry("fs/ext4", NULL
);
5467 kset_unregister(ext4_kset
);
5468 ext4_exit_system_zone();
5473 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5474 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5475 MODULE_LICENSE("GPL");
5476 module_init(ext4_init_fs
)
5477 module_exit(ext4_exit_fs
)