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 panic("EXT4-fs (device %s): panic forced after error\n",
408 void __ext4_error(struct super_block
*sb
, const char *function
,
409 unsigned int line
, const char *fmt
, ...)
411 struct va_format vaf
;
417 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
418 sb
->s_id
, function
, line
, current
->comm
, &vaf
);
420 save_error_info(sb
, function
, line
);
422 ext4_handle_error(sb
);
425 void ext4_error_inode(struct inode
*inode
, const char *function
,
426 unsigned int line
, ext4_fsblk_t block
,
427 const char *fmt
, ...)
430 struct va_format vaf
;
431 struct ext4_super_block
*es
= EXT4_SB(inode
->i_sb
)->s_es
;
433 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
434 es
->s_last_error_block
= cpu_to_le64(block
);
435 save_error_info(inode
->i_sb
, function
, line
);
440 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
441 "inode #%lu: block %llu: comm %s: %pV\n",
442 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
443 block
, current
->comm
, &vaf
);
445 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
446 "inode #%lu: comm %s: %pV\n",
447 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
448 current
->comm
, &vaf
);
451 ext4_handle_error(inode
->i_sb
);
454 void ext4_error_file(struct file
*file
, const char *function
,
455 unsigned int line
, ext4_fsblk_t block
,
456 const char *fmt
, ...)
459 struct va_format vaf
;
460 struct ext4_super_block
*es
;
461 struct inode
*inode
= file_inode(file
);
462 char pathname
[80], *path
;
464 es
= EXT4_SB(inode
->i_sb
)->s_es
;
465 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
466 save_error_info(inode
->i_sb
, function
, line
);
467 path
= d_path(&(file
->f_path
), pathname
, sizeof(pathname
));
475 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
476 "block %llu: comm %s: path %s: %pV\n",
477 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
478 block
, current
->comm
, path
, &vaf
);
481 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
482 "comm %s: path %s: %pV\n",
483 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
484 current
->comm
, path
, &vaf
);
487 ext4_handle_error(inode
->i_sb
);
490 const char *ext4_decode_error(struct super_block
*sb
, int errno
,
497 errstr
= "IO failure";
500 errstr
= "Out of memory";
503 if (!sb
|| (EXT4_SB(sb
)->s_journal
&&
504 EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
))
505 errstr
= "Journal has aborted";
507 errstr
= "Readonly filesystem";
510 /* If the caller passed in an extra buffer for unknown
511 * errors, textualise them now. Else we just return
514 /* Check for truncated error codes... */
515 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
524 /* __ext4_std_error decodes expected errors from journaling functions
525 * automatically and invokes the appropriate error response. */
527 void __ext4_std_error(struct super_block
*sb
, const char *function
,
528 unsigned int line
, int errno
)
533 /* Special case: if the error is EROFS, and we're not already
534 * inside a transaction, then there's really no point in logging
536 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
537 (sb
->s_flags
& MS_RDONLY
))
540 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
541 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s:%d: %s\n",
542 sb
->s_id
, function
, line
, errstr
);
543 save_error_info(sb
, function
, line
);
545 ext4_handle_error(sb
);
549 * ext4_abort is a much stronger failure handler than ext4_error. The
550 * abort function may be used to deal with unrecoverable failures such
551 * as journal IO errors or ENOMEM at a critical moment in log management.
553 * We unconditionally force the filesystem into an ABORT|READONLY state,
554 * unless the error response on the fs has been set to panic in which
555 * case we take the easy way out and panic immediately.
558 void __ext4_abort(struct super_block
*sb
, const char *function
,
559 unsigned int line
, const char *fmt
, ...)
563 save_error_info(sb
, function
, line
);
565 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: ", sb
->s_id
,
571 if ((sb
->s_flags
& MS_RDONLY
) == 0) {
572 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
573 sb
->s_flags
|= MS_RDONLY
;
574 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
575 if (EXT4_SB(sb
)->s_journal
)
576 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
577 save_error_info(sb
, function
, line
);
579 if (test_opt(sb
, ERRORS_PANIC
))
580 panic("EXT4-fs panic from previous error\n");
583 void ext4_msg(struct super_block
*sb
, const char *prefix
, const char *fmt
, ...)
585 struct va_format vaf
;
591 printk("%sEXT4-fs (%s): %pV\n", prefix
, sb
->s_id
, &vaf
);
595 void __ext4_warning(struct super_block
*sb
, const char *function
,
596 unsigned int line
, const char *fmt
, ...)
598 struct va_format vaf
;
604 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: %pV\n",
605 sb
->s_id
, function
, line
, &vaf
);
609 void __ext4_grp_locked_error(const char *function
, unsigned int line
,
610 struct super_block
*sb
, ext4_group_t grp
,
611 unsigned long ino
, ext4_fsblk_t block
,
612 const char *fmt
, ...)
616 struct va_format vaf
;
618 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
620 es
->s_last_error_ino
= cpu_to_le32(ino
);
621 es
->s_last_error_block
= cpu_to_le64(block
);
622 __save_error_info(sb
, function
, line
);
628 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: group %u, ",
629 sb
->s_id
, function
, line
, grp
);
631 printk(KERN_CONT
"inode %lu: ", ino
);
633 printk(KERN_CONT
"block %llu:", (unsigned long long) block
);
634 printk(KERN_CONT
"%pV\n", &vaf
);
637 if (test_opt(sb
, ERRORS_CONT
)) {
638 ext4_commit_super(sb
, 0);
642 ext4_unlock_group(sb
, grp
);
643 ext4_handle_error(sb
);
645 * We only get here in the ERRORS_RO case; relocking the group
646 * may be dangerous, but nothing bad will happen since the
647 * filesystem will have already been marked read/only and the
648 * journal has been aborted. We return 1 as a hint to callers
649 * who might what to use the return value from
650 * ext4_grp_locked_error() to distinguish between the
651 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
652 * aggressively from the ext4 function in question, with a
653 * more appropriate error code.
655 ext4_lock_group(sb
, grp
);
659 void ext4_update_dynamic_rev(struct super_block
*sb
)
661 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
663 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
667 "updating to rev %d because of new feature flag, "
668 "running e2fsck is recommended",
671 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
672 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
673 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
674 /* leave es->s_feature_*compat flags alone */
675 /* es->s_uuid will be set by e2fsck if empty */
678 * The rest of the superblock fields should be zero, and if not it
679 * means they are likely already in use, so leave them alone. We
680 * can leave it up to e2fsck to clean up any inconsistencies there.
685 * Open the external journal device
687 static struct block_device
*ext4_blkdev_get(dev_t dev
, struct super_block
*sb
)
689 struct block_device
*bdev
;
690 char b
[BDEVNAME_SIZE
];
692 bdev
= blkdev_get_by_dev(dev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
, sb
);
698 ext4_msg(sb
, KERN_ERR
, "failed to open journal device %s: %ld",
699 __bdevname(dev
, b
), PTR_ERR(bdev
));
704 * Release the journal device
706 static void ext4_blkdev_put(struct block_device
*bdev
)
708 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
);
711 static void ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
713 struct block_device
*bdev
;
714 bdev
= sbi
->journal_bdev
;
716 ext4_blkdev_put(bdev
);
717 sbi
->journal_bdev
= NULL
;
721 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
723 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
726 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
730 ext4_msg(sb
, KERN_ERR
, "sb orphan head is %d",
731 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
733 printk(KERN_ERR
"sb_info orphan list:\n");
734 list_for_each(l
, &sbi
->s_orphan
) {
735 struct inode
*inode
= orphan_list_entry(l
);
737 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
738 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
739 inode
->i_mode
, inode
->i_nlink
,
744 static void ext4_put_super(struct super_block
*sb
)
746 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
747 struct ext4_super_block
*es
= sbi
->s_es
;
750 ext4_unregister_li_request(sb
);
751 dquot_disable(sb
, -1, DQUOT_USAGE_ENABLED
| DQUOT_LIMITS_ENABLED
);
753 flush_workqueue(sbi
->dio_unwritten_wq
);
754 destroy_workqueue(sbi
->dio_unwritten_wq
);
756 if (sbi
->s_journal
) {
757 err
= jbd2_journal_destroy(sbi
->s_journal
);
758 sbi
->s_journal
= NULL
;
760 ext4_abort(sb
, "Couldn't clean up the journal");
763 ext4_es_unregister_shrinker(sb
);
764 del_timer(&sbi
->s_err_report
);
765 ext4_release_system_zone(sb
);
767 ext4_ext_release(sb
);
768 ext4_xattr_put_super(sb
);
770 if (!(sb
->s_flags
& MS_RDONLY
)) {
771 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
772 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
774 if (!(sb
->s_flags
& MS_RDONLY
))
775 ext4_commit_super(sb
, 1);
778 remove_proc_entry("options", sbi
->s_proc
);
779 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
781 kobject_del(&sbi
->s_kobj
);
783 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
784 brelse(sbi
->s_group_desc
[i
]);
785 ext4_kvfree(sbi
->s_group_desc
);
786 ext4_kvfree(sbi
->s_flex_groups
);
787 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
788 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
789 percpu_counter_destroy(&sbi
->s_dirs_counter
);
790 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
791 percpu_counter_destroy(&sbi
->s_extent_cache_cnt
);
794 for (i
= 0; i
< MAXQUOTAS
; i
++)
795 kfree(sbi
->s_qf_names
[i
]);
798 /* Debugging code just in case the in-memory inode orphan list
799 * isn't empty. The on-disk one can be non-empty if we've
800 * detected an error and taken the fs readonly, but the
801 * in-memory list had better be clean by this point. */
802 if (!list_empty(&sbi
->s_orphan
))
803 dump_orphan_list(sb
, sbi
);
804 J_ASSERT(list_empty(&sbi
->s_orphan
));
806 invalidate_bdev(sb
->s_bdev
);
807 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
809 * Invalidate the journal device's buffers. We don't want them
810 * floating about in memory - the physical journal device may
811 * hotswapped, and it breaks the `ro-after' testing code.
813 sync_blockdev(sbi
->journal_bdev
);
814 invalidate_bdev(sbi
->journal_bdev
);
815 ext4_blkdev_remove(sbi
);
818 kthread_stop(sbi
->s_mmp_tsk
);
819 sb
->s_fs_info
= NULL
;
821 * Now that we are completely done shutting down the
822 * superblock, we need to actually destroy the kobject.
824 kobject_put(&sbi
->s_kobj
);
825 wait_for_completion(&sbi
->s_kobj_unregister
);
826 if (sbi
->s_chksum_driver
)
827 crypto_free_shash(sbi
->s_chksum_driver
);
828 kfree(sbi
->s_blockgroup_lock
);
832 static struct kmem_cache
*ext4_inode_cachep
;
835 * Called inside transaction, so use GFP_NOFS
837 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
839 struct ext4_inode_info
*ei
;
841 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
845 ei
->vfs_inode
.i_version
= 1;
846 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
847 spin_lock_init(&ei
->i_prealloc_lock
);
848 ext4_es_init_tree(&ei
->i_es_tree
);
849 rwlock_init(&ei
->i_es_lock
);
850 INIT_LIST_HEAD(&ei
->i_es_lru
);
852 ei
->i_reserved_data_blocks
= 0;
853 ei
->i_reserved_meta_blocks
= 0;
854 ei
->i_allocated_meta_blocks
= 0;
855 ei
->i_da_metadata_calc_len
= 0;
856 ei
->i_da_metadata_calc_last_lblock
= 0;
857 spin_lock_init(&(ei
->i_block_reservation_lock
));
859 ei
->i_reserved_quota
= 0;
862 INIT_LIST_HEAD(&ei
->i_completed_io_list
);
863 spin_lock_init(&ei
->i_completed_io_lock
);
865 ei
->i_datasync_tid
= 0;
866 atomic_set(&ei
->i_ioend_count
, 0);
867 atomic_set(&ei
->i_unwritten
, 0);
868 INIT_WORK(&ei
->i_unwritten_work
, ext4_end_io_work
);
870 return &ei
->vfs_inode
;
873 static int ext4_drop_inode(struct inode
*inode
)
875 int drop
= generic_drop_inode(inode
);
877 trace_ext4_drop_inode(inode
, drop
);
881 static void ext4_i_callback(struct rcu_head
*head
)
883 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
884 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
887 static void ext4_destroy_inode(struct inode
*inode
)
889 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
890 ext4_msg(inode
->i_sb
, KERN_ERR
,
891 "Inode %lu (%p): orphan list check failed!",
892 inode
->i_ino
, EXT4_I(inode
));
893 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
894 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
898 call_rcu(&inode
->i_rcu
, ext4_i_callback
);
901 static void init_once(void *foo
)
903 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
905 INIT_LIST_HEAD(&ei
->i_orphan
);
906 init_rwsem(&ei
->xattr_sem
);
907 init_rwsem(&ei
->i_data_sem
);
908 inode_init_once(&ei
->vfs_inode
);
911 static int init_inodecache(void)
913 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
914 sizeof(struct ext4_inode_info
),
915 0, (SLAB_RECLAIM_ACCOUNT
|
918 if (ext4_inode_cachep
== NULL
)
923 static void destroy_inodecache(void)
926 * Make sure all delayed rcu free inodes are flushed before we
930 kmem_cache_destroy(ext4_inode_cachep
);
933 void ext4_clear_inode(struct inode
*inode
)
935 invalidate_inode_buffers(inode
);
938 ext4_discard_preallocations(inode
);
939 ext4_es_remove_extent(inode
, 0, EXT_MAX_BLOCKS
);
940 ext4_es_lru_del(inode
);
941 if (EXT4_I(inode
)->jinode
) {
942 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode
),
943 EXT4_I(inode
)->jinode
);
944 jbd2_free_inode(EXT4_I(inode
)->jinode
);
945 EXT4_I(inode
)->jinode
= NULL
;
949 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
950 u64 ino
, u32 generation
)
954 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
955 return ERR_PTR(-ESTALE
);
956 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
957 return ERR_PTR(-ESTALE
);
959 /* iget isn't really right if the inode is currently unallocated!!
961 * ext4_read_inode will return a bad_inode if the inode had been
962 * deleted, so we should be safe.
964 * Currently we don't know the generation for parent directory, so
965 * a generation of 0 means "accept any"
967 inode
= ext4_iget_normal(sb
, ino
);
969 return ERR_CAST(inode
);
970 if (generation
&& inode
->i_generation
!= generation
) {
972 return ERR_PTR(-ESTALE
);
978 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
979 int fh_len
, int fh_type
)
981 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
985 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
986 int fh_len
, int fh_type
)
988 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
993 * Try to release metadata pages (indirect blocks, directories) which are
994 * mapped via the block device. Since these pages could have journal heads
995 * which would prevent try_to_free_buffers() from freeing them, we must use
996 * jbd2 layer's try_to_free_buffers() function to release them.
998 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
1001 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
1003 WARN_ON(PageChecked(page
));
1004 if (!page_has_buffers(page
))
1007 return jbd2_journal_try_to_free_buffers(journal
, page
,
1008 wait
& ~__GFP_WAIT
);
1009 return try_to_free_buffers(page
);
1013 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1014 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1016 static int ext4_write_dquot(struct dquot
*dquot
);
1017 static int ext4_acquire_dquot(struct dquot
*dquot
);
1018 static int ext4_release_dquot(struct dquot
*dquot
);
1019 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
1020 static int ext4_write_info(struct super_block
*sb
, int type
);
1021 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
1023 static int ext4_quota_on_sysfile(struct super_block
*sb
, int type
,
1025 static int ext4_quota_off(struct super_block
*sb
, int type
);
1026 static int ext4_quota_off_sysfile(struct super_block
*sb
, int type
);
1027 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
1028 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
1029 size_t len
, loff_t off
);
1030 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
1031 const char *data
, size_t len
, loff_t off
);
1032 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
1033 unsigned int flags
);
1034 static int ext4_enable_quotas(struct super_block
*sb
);
1036 static const struct dquot_operations ext4_quota_operations
= {
1037 .get_reserved_space
= ext4_get_reserved_space
,
1038 .write_dquot
= ext4_write_dquot
,
1039 .acquire_dquot
= ext4_acquire_dquot
,
1040 .release_dquot
= ext4_release_dquot
,
1041 .mark_dirty
= ext4_mark_dquot_dirty
,
1042 .write_info
= ext4_write_info
,
1043 .alloc_dquot
= dquot_alloc
,
1044 .destroy_dquot
= dquot_destroy
,
1047 static const struct quotactl_ops ext4_qctl_operations
= {
1048 .quota_on
= ext4_quota_on
,
1049 .quota_off
= ext4_quota_off
,
1050 .quota_sync
= dquot_quota_sync
,
1051 .get_info
= dquot_get_dqinfo
,
1052 .set_info
= dquot_set_dqinfo
,
1053 .get_dqblk
= dquot_get_dqblk
,
1054 .set_dqblk
= dquot_set_dqblk
1057 static const struct quotactl_ops ext4_qctl_sysfile_operations
= {
1058 .quota_on_meta
= ext4_quota_on_sysfile
,
1059 .quota_off
= ext4_quota_off_sysfile
,
1060 .quota_sync
= dquot_quota_sync
,
1061 .get_info
= dquot_get_dqinfo
,
1062 .set_info
= dquot_set_dqinfo
,
1063 .get_dqblk
= dquot_get_dqblk
,
1064 .set_dqblk
= dquot_set_dqblk
1068 static const struct super_operations ext4_sops
= {
1069 .alloc_inode
= ext4_alloc_inode
,
1070 .destroy_inode
= ext4_destroy_inode
,
1071 .write_inode
= ext4_write_inode
,
1072 .dirty_inode
= ext4_dirty_inode
,
1073 .drop_inode
= ext4_drop_inode
,
1074 .evict_inode
= ext4_evict_inode
,
1075 .put_super
= ext4_put_super
,
1076 .sync_fs
= ext4_sync_fs
,
1077 .freeze_fs
= ext4_freeze
,
1078 .unfreeze_fs
= ext4_unfreeze
,
1079 .statfs
= ext4_statfs
,
1080 .remount_fs
= ext4_remount
,
1081 .show_options
= ext4_show_options
,
1083 .quota_read
= ext4_quota_read
,
1084 .quota_write
= ext4_quota_write
,
1086 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1089 static const struct super_operations ext4_nojournal_sops
= {
1090 .alloc_inode
= ext4_alloc_inode
,
1091 .destroy_inode
= ext4_destroy_inode
,
1092 .write_inode
= ext4_write_inode
,
1093 .dirty_inode
= ext4_dirty_inode
,
1094 .drop_inode
= ext4_drop_inode
,
1095 .evict_inode
= ext4_evict_inode
,
1096 .put_super
= ext4_put_super
,
1097 .statfs
= ext4_statfs
,
1098 .remount_fs
= ext4_remount
,
1099 .show_options
= ext4_show_options
,
1101 .quota_read
= ext4_quota_read
,
1102 .quota_write
= ext4_quota_write
,
1104 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1107 static const struct export_operations ext4_export_ops
= {
1108 .fh_to_dentry
= ext4_fh_to_dentry
,
1109 .fh_to_parent
= ext4_fh_to_parent
,
1110 .get_parent
= ext4_get_parent
,
1114 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1115 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1116 Opt_nouid32
, Opt_debug
, Opt_removed
,
1117 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1118 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
,
1119 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
,
1120 Opt_journal_dev
, Opt_journal_checksum
, Opt_journal_async_commit
,
1121 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1122 Opt_data_err_abort
, Opt_data_err_ignore
,
1123 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1124 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
1125 Opt_noquota
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
1126 Opt_usrquota
, Opt_grpquota
, Opt_i_version
,
1127 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
, Opt_mblk_io_submit
,
1128 Opt_nomblk_io_submit
, Opt_block_validity
, Opt_noblock_validity
,
1129 Opt_inode_readahead_blks
, Opt_journal_ioprio
,
1130 Opt_dioread_nolock
, Opt_dioread_lock
,
1131 Opt_discard
, Opt_nodiscard
, Opt_init_itable
, Opt_noinit_itable
,
1132 Opt_max_dir_size_kb
,
1135 static const match_table_t tokens
= {
1136 {Opt_bsd_df
, "bsddf"},
1137 {Opt_minix_df
, "minixdf"},
1138 {Opt_grpid
, "grpid"},
1139 {Opt_grpid
, "bsdgroups"},
1140 {Opt_nogrpid
, "nogrpid"},
1141 {Opt_nogrpid
, "sysvgroups"},
1142 {Opt_resgid
, "resgid=%u"},
1143 {Opt_resuid
, "resuid=%u"},
1145 {Opt_err_cont
, "errors=continue"},
1146 {Opt_err_panic
, "errors=panic"},
1147 {Opt_err_ro
, "errors=remount-ro"},
1148 {Opt_nouid32
, "nouid32"},
1149 {Opt_debug
, "debug"},
1150 {Opt_removed
, "oldalloc"},
1151 {Opt_removed
, "orlov"},
1152 {Opt_user_xattr
, "user_xattr"},
1153 {Opt_nouser_xattr
, "nouser_xattr"},
1155 {Opt_noacl
, "noacl"},
1156 {Opt_noload
, "norecovery"},
1157 {Opt_noload
, "noload"},
1158 {Opt_removed
, "nobh"},
1159 {Opt_removed
, "bh"},
1160 {Opt_commit
, "commit=%u"},
1161 {Opt_min_batch_time
, "min_batch_time=%u"},
1162 {Opt_max_batch_time
, "max_batch_time=%u"},
1163 {Opt_journal_dev
, "journal_dev=%u"},
1164 {Opt_journal_checksum
, "journal_checksum"},
1165 {Opt_journal_async_commit
, "journal_async_commit"},
1166 {Opt_abort
, "abort"},
1167 {Opt_data_journal
, "data=journal"},
1168 {Opt_data_ordered
, "data=ordered"},
1169 {Opt_data_writeback
, "data=writeback"},
1170 {Opt_data_err_abort
, "data_err=abort"},
1171 {Opt_data_err_ignore
, "data_err=ignore"},
1172 {Opt_offusrjquota
, "usrjquota="},
1173 {Opt_usrjquota
, "usrjquota=%s"},
1174 {Opt_offgrpjquota
, "grpjquota="},
1175 {Opt_grpjquota
, "grpjquota=%s"},
1176 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1177 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1178 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
1179 {Opt_grpquota
, "grpquota"},
1180 {Opt_noquota
, "noquota"},
1181 {Opt_quota
, "quota"},
1182 {Opt_usrquota
, "usrquota"},
1183 {Opt_barrier
, "barrier=%u"},
1184 {Opt_barrier
, "barrier"},
1185 {Opt_nobarrier
, "nobarrier"},
1186 {Opt_i_version
, "i_version"},
1187 {Opt_stripe
, "stripe=%u"},
1188 {Opt_delalloc
, "delalloc"},
1189 {Opt_nodelalloc
, "nodelalloc"},
1190 {Opt_removed
, "mblk_io_submit"},
1191 {Opt_removed
, "nomblk_io_submit"},
1192 {Opt_block_validity
, "block_validity"},
1193 {Opt_noblock_validity
, "noblock_validity"},
1194 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1195 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1196 {Opt_auto_da_alloc
, "auto_da_alloc=%u"},
1197 {Opt_auto_da_alloc
, "auto_da_alloc"},
1198 {Opt_noauto_da_alloc
, "noauto_da_alloc"},
1199 {Opt_dioread_nolock
, "dioread_nolock"},
1200 {Opt_dioread_lock
, "dioread_lock"},
1201 {Opt_discard
, "discard"},
1202 {Opt_nodiscard
, "nodiscard"},
1203 {Opt_init_itable
, "init_itable=%u"},
1204 {Opt_init_itable
, "init_itable"},
1205 {Opt_noinit_itable
, "noinit_itable"},
1206 {Opt_max_dir_size_kb
, "max_dir_size_kb=%u"},
1207 {Opt_removed
, "check=none"}, /* mount option from ext2/3 */
1208 {Opt_removed
, "nocheck"}, /* mount option from ext2/3 */
1209 {Opt_removed
, "reservation"}, /* mount option from ext2/3 */
1210 {Opt_removed
, "noreservation"}, /* mount option from ext2/3 */
1211 {Opt_removed
, "journal=%u"}, /* mount option from ext2/3 */
1215 static ext4_fsblk_t
get_sb_block(void **data
)
1217 ext4_fsblk_t sb_block
;
1218 char *options
= (char *) *data
;
1220 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1221 return 1; /* Default location */
1224 /* TODO: use simple_strtoll with >32bit ext4 */
1225 sb_block
= simple_strtoul(options
, &options
, 0);
1226 if (*options
&& *options
!= ',') {
1227 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1231 if (*options
== ',')
1233 *data
= (void *) options
;
1238 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1239 static char deprecated_msg
[] = "Mount option \"%s\" will be removed by %s\n"
1240 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1243 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
1245 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1249 if (sb_any_quota_loaded(sb
) &&
1250 !sbi
->s_qf_names
[qtype
]) {
1251 ext4_msg(sb
, KERN_ERR
,
1252 "Cannot change journaled "
1253 "quota options when quota turned on");
1256 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
)) {
1257 ext4_msg(sb
, KERN_ERR
, "Cannot set journaled quota options "
1258 "when QUOTA feature is enabled");
1261 qname
= match_strdup(args
);
1263 ext4_msg(sb
, KERN_ERR
,
1264 "Not enough memory for storing quotafile name");
1267 if (sbi
->s_qf_names
[qtype
]) {
1268 if (strcmp(sbi
->s_qf_names
[qtype
], qname
) == 0)
1271 ext4_msg(sb
, KERN_ERR
,
1272 "%s quota file already specified",
1276 if (strchr(qname
, '/')) {
1277 ext4_msg(sb
, KERN_ERR
,
1278 "quotafile must be on filesystem root");
1281 sbi
->s_qf_names
[qtype
] = qname
;
1289 static int clear_qf_name(struct super_block
*sb
, int qtype
)
1292 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1294 if (sb_any_quota_loaded(sb
) &&
1295 sbi
->s_qf_names
[qtype
]) {
1296 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
1297 " when quota turned on");
1300 kfree(sbi
->s_qf_names
[qtype
]);
1301 sbi
->s_qf_names
[qtype
] = NULL
;
1306 #define MOPT_SET 0x0001
1307 #define MOPT_CLEAR 0x0002
1308 #define MOPT_NOSUPPORT 0x0004
1309 #define MOPT_EXPLICIT 0x0008
1310 #define MOPT_CLEAR_ERR 0x0010
1311 #define MOPT_GTE0 0x0020
1314 #define MOPT_QFMT 0x0040
1316 #define MOPT_Q MOPT_NOSUPPORT
1317 #define MOPT_QFMT MOPT_NOSUPPORT
1319 #define MOPT_DATAJ 0x0080
1320 #define MOPT_NO_EXT2 0x0100
1321 #define MOPT_NO_EXT3 0x0200
1322 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1324 static const struct mount_opts
{
1328 } ext4_mount_opts
[] = {
1329 {Opt_minix_df
, EXT4_MOUNT_MINIX_DF
, MOPT_SET
},
1330 {Opt_bsd_df
, EXT4_MOUNT_MINIX_DF
, MOPT_CLEAR
},
1331 {Opt_grpid
, EXT4_MOUNT_GRPID
, MOPT_SET
},
1332 {Opt_nogrpid
, EXT4_MOUNT_GRPID
, MOPT_CLEAR
},
1333 {Opt_block_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_SET
},
1334 {Opt_noblock_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_CLEAR
},
1335 {Opt_dioread_nolock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1336 MOPT_EXT4_ONLY
| MOPT_SET
},
1337 {Opt_dioread_lock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1338 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1339 {Opt_discard
, EXT4_MOUNT_DISCARD
, MOPT_SET
},
1340 {Opt_nodiscard
, EXT4_MOUNT_DISCARD
, MOPT_CLEAR
},
1341 {Opt_delalloc
, EXT4_MOUNT_DELALLOC
,
1342 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1343 {Opt_nodelalloc
, EXT4_MOUNT_DELALLOC
,
1344 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1345 {Opt_journal_checksum
, EXT4_MOUNT_JOURNAL_CHECKSUM
,
1346 MOPT_EXT4_ONLY
| MOPT_SET
},
1347 {Opt_journal_async_commit
, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT
|
1348 EXT4_MOUNT_JOURNAL_CHECKSUM
),
1349 MOPT_EXT4_ONLY
| MOPT_SET
},
1350 {Opt_noload
, EXT4_MOUNT_NOLOAD
, MOPT_NO_EXT2
| MOPT_SET
},
1351 {Opt_err_panic
, EXT4_MOUNT_ERRORS_PANIC
, MOPT_SET
| MOPT_CLEAR_ERR
},
1352 {Opt_err_ro
, EXT4_MOUNT_ERRORS_RO
, MOPT_SET
| MOPT_CLEAR_ERR
},
1353 {Opt_err_cont
, EXT4_MOUNT_ERRORS_CONT
, MOPT_SET
| MOPT_CLEAR_ERR
},
1354 {Opt_data_err_abort
, EXT4_MOUNT_DATA_ERR_ABORT
,
1355 MOPT_NO_EXT2
| MOPT_SET
},
1356 {Opt_data_err_ignore
, EXT4_MOUNT_DATA_ERR_ABORT
,
1357 MOPT_NO_EXT2
| MOPT_CLEAR
},
1358 {Opt_barrier
, EXT4_MOUNT_BARRIER
, MOPT_SET
},
1359 {Opt_nobarrier
, EXT4_MOUNT_BARRIER
, MOPT_CLEAR
},
1360 {Opt_noauto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_SET
},
1361 {Opt_auto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_CLEAR
},
1362 {Opt_noinit_itable
, EXT4_MOUNT_INIT_INODE_TABLE
, MOPT_CLEAR
},
1363 {Opt_commit
, 0, MOPT_GTE0
},
1364 {Opt_max_batch_time
, 0, MOPT_GTE0
},
1365 {Opt_min_batch_time
, 0, MOPT_GTE0
},
1366 {Opt_inode_readahead_blks
, 0, MOPT_GTE0
},
1367 {Opt_init_itable
, 0, MOPT_GTE0
},
1368 {Opt_stripe
, 0, MOPT_GTE0
},
1369 {Opt_resuid
, 0, MOPT_GTE0
},
1370 {Opt_resgid
, 0, MOPT_GTE0
},
1371 {Opt_journal_dev
, 0, MOPT_GTE0
},
1372 {Opt_journal_ioprio
, 0, MOPT_GTE0
},
1373 {Opt_data_journal
, EXT4_MOUNT_JOURNAL_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1374 {Opt_data_ordered
, EXT4_MOUNT_ORDERED_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1375 {Opt_data_writeback
, EXT4_MOUNT_WRITEBACK_DATA
,
1376 MOPT_NO_EXT2
| MOPT_DATAJ
},
1377 {Opt_user_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_SET
},
1378 {Opt_nouser_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_CLEAR
},
1379 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1380 {Opt_acl
, EXT4_MOUNT_POSIX_ACL
, MOPT_SET
},
1381 {Opt_noacl
, EXT4_MOUNT_POSIX_ACL
, MOPT_CLEAR
},
1383 {Opt_acl
, 0, MOPT_NOSUPPORT
},
1384 {Opt_noacl
, 0, MOPT_NOSUPPORT
},
1386 {Opt_nouid32
, EXT4_MOUNT_NO_UID32
, MOPT_SET
},
1387 {Opt_debug
, EXT4_MOUNT_DEBUG
, MOPT_SET
},
1388 {Opt_quota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
, MOPT_SET
| MOPT_Q
},
1389 {Opt_usrquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
,
1391 {Opt_grpquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_GRPQUOTA
,
1393 {Opt_noquota
, (EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
|
1394 EXT4_MOUNT_GRPQUOTA
), MOPT_CLEAR
| MOPT_Q
},
1395 {Opt_usrjquota
, 0, MOPT_Q
},
1396 {Opt_grpjquota
, 0, MOPT_Q
},
1397 {Opt_offusrjquota
, 0, MOPT_Q
},
1398 {Opt_offgrpjquota
, 0, MOPT_Q
},
1399 {Opt_jqfmt_vfsold
, QFMT_VFS_OLD
, MOPT_QFMT
},
1400 {Opt_jqfmt_vfsv0
, QFMT_VFS_V0
, MOPT_QFMT
},
1401 {Opt_jqfmt_vfsv1
, QFMT_VFS_V1
, MOPT_QFMT
},
1402 {Opt_max_dir_size_kb
, 0, MOPT_GTE0
},
1406 static int handle_mount_opt(struct super_block
*sb
, char *opt
, int token
,
1407 substring_t
*args
, unsigned long *journal_devnum
,
1408 unsigned int *journal_ioprio
, int is_remount
)
1410 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1411 const struct mount_opts
*m
;
1417 if (token
== Opt_usrjquota
)
1418 return set_qf_name(sb
, USRQUOTA
, &args
[0]);
1419 else if (token
== Opt_grpjquota
)
1420 return set_qf_name(sb
, GRPQUOTA
, &args
[0]);
1421 else if (token
== Opt_offusrjquota
)
1422 return clear_qf_name(sb
, USRQUOTA
);
1423 else if (token
== Opt_offgrpjquota
)
1424 return clear_qf_name(sb
, GRPQUOTA
);
1428 case Opt_nouser_xattr
:
1429 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, opt
, "3.5");
1432 return 1; /* handled by get_sb_block() */
1434 ext4_msg(sb
, KERN_WARNING
, "Ignoring removed %s option", opt
);
1437 sbi
->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
1440 sb
->s_flags
|= MS_I_VERSION
;
1444 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++)
1445 if (token
== m
->token
)
1448 if (m
->token
== Opt_err
) {
1449 ext4_msg(sb
, KERN_ERR
, "Unrecognized mount option \"%s\" "
1450 "or missing value", opt
);
1454 if ((m
->flags
& MOPT_NO_EXT2
) && IS_EXT2_SB(sb
)) {
1455 ext4_msg(sb
, KERN_ERR
,
1456 "Mount option \"%s\" incompatible with ext2", opt
);
1459 if ((m
->flags
& MOPT_NO_EXT3
) && IS_EXT3_SB(sb
)) {
1460 ext4_msg(sb
, KERN_ERR
,
1461 "Mount option \"%s\" incompatible with ext3", opt
);
1465 if (args
->from
&& match_int(args
, &arg
))
1467 if (args
->from
&& (m
->flags
& MOPT_GTE0
) && (arg
< 0))
1469 if (m
->flags
& MOPT_EXPLICIT
)
1470 set_opt2(sb
, EXPLICIT_DELALLOC
);
1471 if (m
->flags
& MOPT_CLEAR_ERR
)
1472 clear_opt(sb
, ERRORS_MASK
);
1473 if (token
== Opt_noquota
&& sb_any_quota_loaded(sb
)) {
1474 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
1475 "options when quota turned on");
1479 if (m
->flags
& MOPT_NOSUPPORT
) {
1480 ext4_msg(sb
, KERN_ERR
, "%s option not supported", opt
);
1481 } else if (token
== Opt_commit
) {
1483 arg
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1484 sbi
->s_commit_interval
= HZ
* arg
;
1485 } else if (token
== Opt_max_batch_time
) {
1486 sbi
->s_max_batch_time
= arg
;
1487 } else if (token
== Opt_min_batch_time
) {
1488 sbi
->s_min_batch_time
= arg
;
1489 } else if (token
== Opt_inode_readahead_blks
) {
1490 if (arg
&& (arg
> (1 << 30) || !is_power_of_2(arg
))) {
1491 ext4_msg(sb
, KERN_ERR
,
1492 "EXT4-fs: inode_readahead_blks must be "
1493 "0 or a power of 2 smaller than 2^31");
1496 sbi
->s_inode_readahead_blks
= arg
;
1497 } else if (token
== Opt_init_itable
) {
1498 set_opt(sb
, INIT_INODE_TABLE
);
1500 arg
= EXT4_DEF_LI_WAIT_MULT
;
1501 sbi
->s_li_wait_mult
= arg
;
1502 } else if (token
== Opt_max_dir_size_kb
) {
1503 sbi
->s_max_dir_size_kb
= arg
;
1504 } else if (token
== Opt_stripe
) {
1505 sbi
->s_stripe
= arg
;
1506 } else if (token
== Opt_resuid
) {
1507 uid
= make_kuid(current_user_ns(), arg
);
1508 if (!uid_valid(uid
)) {
1509 ext4_msg(sb
, KERN_ERR
, "Invalid uid value %d", arg
);
1512 sbi
->s_resuid
= uid
;
1513 } else if (token
== Opt_resgid
) {
1514 gid
= make_kgid(current_user_ns(), arg
);
1515 if (!gid_valid(gid
)) {
1516 ext4_msg(sb
, KERN_ERR
, "Invalid gid value %d", arg
);
1519 sbi
->s_resgid
= gid
;
1520 } else if (token
== Opt_journal_dev
) {
1522 ext4_msg(sb
, KERN_ERR
,
1523 "Cannot specify journal on remount");
1526 *journal_devnum
= arg
;
1527 } else if (token
== Opt_journal_ioprio
) {
1529 ext4_msg(sb
, KERN_ERR
, "Invalid journal IO priority"
1534 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
, arg
);
1535 } else if (m
->flags
& MOPT_DATAJ
) {
1537 if (!sbi
->s_journal
)
1538 ext4_msg(sb
, KERN_WARNING
, "Remounting file system with no journal so ignoring journalled data option");
1539 else if (test_opt(sb
, DATA_FLAGS
) != m
->mount_opt
) {
1540 ext4_msg(sb
, KERN_ERR
,
1541 "Cannot change data mode on remount");
1545 clear_opt(sb
, DATA_FLAGS
);
1546 sbi
->s_mount_opt
|= m
->mount_opt
;
1549 } else if (m
->flags
& MOPT_QFMT
) {
1550 if (sb_any_quota_loaded(sb
) &&
1551 sbi
->s_jquota_fmt
!= m
->mount_opt
) {
1552 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled "
1553 "quota options when quota turned on");
1556 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
1557 EXT4_FEATURE_RO_COMPAT_QUOTA
)) {
1558 ext4_msg(sb
, KERN_ERR
,
1559 "Cannot set journaled quota options "
1560 "when QUOTA feature is enabled");
1563 sbi
->s_jquota_fmt
= m
->mount_opt
;
1568 if (m
->flags
& MOPT_CLEAR
)
1570 else if (unlikely(!(m
->flags
& MOPT_SET
))) {
1571 ext4_msg(sb
, KERN_WARNING
,
1572 "buggy handling of option %s", opt
);
1577 sbi
->s_mount_opt
|= m
->mount_opt
;
1579 sbi
->s_mount_opt
&= ~m
->mount_opt
;
1584 static int parse_options(char *options
, struct super_block
*sb
,
1585 unsigned long *journal_devnum
,
1586 unsigned int *journal_ioprio
,
1589 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1591 substring_t args
[MAX_OPT_ARGS
];
1597 while ((p
= strsep(&options
, ",")) != NULL
) {
1601 * Initialize args struct so we know whether arg was
1602 * found; some options take optional arguments.
1604 args
[0].to
= args
[0].from
= NULL
;
1605 token
= match_token(p
, tokens
, args
);
1606 if (handle_mount_opt(sb
, p
, token
, args
, journal_devnum
,
1607 journal_ioprio
, is_remount
) < 0)
1611 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
) &&
1612 (test_opt(sb
, USRQUOTA
) || test_opt(sb
, GRPQUOTA
))) {
1613 ext4_msg(sb
, KERN_ERR
, "Cannot set quota options when QUOTA "
1614 "feature is enabled");
1617 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1618 if (test_opt(sb
, USRQUOTA
) && sbi
->s_qf_names
[USRQUOTA
])
1619 clear_opt(sb
, USRQUOTA
);
1621 if (test_opt(sb
, GRPQUOTA
) && sbi
->s_qf_names
[GRPQUOTA
])
1622 clear_opt(sb
, GRPQUOTA
);
1624 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
1625 ext4_msg(sb
, KERN_ERR
, "old and new quota "
1630 if (!sbi
->s_jquota_fmt
) {
1631 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1637 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
1639 BLOCK_SIZE
<< le32_to_cpu(sbi
->s_es
->s_log_block_size
);
1641 if (blocksize
< PAGE_CACHE_SIZE
) {
1642 ext4_msg(sb
, KERN_ERR
, "can't mount with "
1643 "dioread_nolock if block size != PAGE_SIZE");
1650 static inline void ext4_show_quota_options(struct seq_file
*seq
,
1651 struct super_block
*sb
)
1653 #if defined(CONFIG_QUOTA)
1654 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1656 if (sbi
->s_jquota_fmt
) {
1659 switch (sbi
->s_jquota_fmt
) {
1670 seq_printf(seq
, ",jqfmt=%s", fmtname
);
1673 if (sbi
->s_qf_names
[USRQUOTA
])
1674 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
1676 if (sbi
->s_qf_names
[GRPQUOTA
])
1677 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
1681 static const char *token2str(int token
)
1683 const struct match_token
*t
;
1685 for (t
= tokens
; t
->token
!= Opt_err
; t
++)
1686 if (t
->token
== token
&& !strchr(t
->pattern
, '='))
1693 * - it's set to a non-default value OR
1694 * - if the per-sb default is different from the global default
1696 static int _ext4_show_options(struct seq_file
*seq
, struct super_block
*sb
,
1699 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1700 struct ext4_super_block
*es
= sbi
->s_es
;
1701 int def_errors
, def_mount_opt
= nodefs
? 0 : sbi
->s_def_mount_opt
;
1702 const struct mount_opts
*m
;
1703 char sep
= nodefs
? '\n' : ',';
1705 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1706 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1708 if (sbi
->s_sb_block
!= 1)
1709 SEQ_OPTS_PRINT("sb=%llu", sbi
->s_sb_block
);
1711 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++) {
1712 int want_set
= m
->flags
& MOPT_SET
;
1713 if (((m
->flags
& (MOPT_SET
|MOPT_CLEAR
)) == 0) ||
1714 (m
->flags
& MOPT_CLEAR_ERR
))
1716 if (!(m
->mount_opt
& (sbi
->s_mount_opt
^ def_mount_opt
)))
1717 continue; /* skip if same as the default */
1719 (sbi
->s_mount_opt
& m
->mount_opt
) != m
->mount_opt
) ||
1720 (!want_set
&& (sbi
->s_mount_opt
& m
->mount_opt
)))
1721 continue; /* select Opt_noFoo vs Opt_Foo */
1722 SEQ_OPTS_PRINT("%s", token2str(m
->token
));
1725 if (nodefs
|| !uid_eq(sbi
->s_resuid
, make_kuid(&init_user_ns
, EXT4_DEF_RESUID
)) ||
1726 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
)
1727 SEQ_OPTS_PRINT("resuid=%u",
1728 from_kuid_munged(&init_user_ns
, sbi
->s_resuid
));
1729 if (nodefs
|| !gid_eq(sbi
->s_resgid
, make_kgid(&init_user_ns
, EXT4_DEF_RESGID
)) ||
1730 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
)
1731 SEQ_OPTS_PRINT("resgid=%u",
1732 from_kgid_munged(&init_user_ns
, sbi
->s_resgid
));
1733 def_errors
= nodefs
? -1 : le16_to_cpu(es
->s_errors
);
1734 if (test_opt(sb
, ERRORS_RO
) && def_errors
!= EXT4_ERRORS_RO
)
1735 SEQ_OPTS_PUTS("errors=remount-ro");
1736 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
1737 SEQ_OPTS_PUTS("errors=continue");
1738 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
1739 SEQ_OPTS_PUTS("errors=panic");
1740 if (nodefs
|| sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
)
1741 SEQ_OPTS_PRINT("commit=%lu", sbi
->s_commit_interval
/ HZ
);
1742 if (nodefs
|| sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
)
1743 SEQ_OPTS_PRINT("min_batch_time=%u", sbi
->s_min_batch_time
);
1744 if (nodefs
|| sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
)
1745 SEQ_OPTS_PRINT("max_batch_time=%u", sbi
->s_max_batch_time
);
1746 if (sb
->s_flags
& MS_I_VERSION
)
1747 SEQ_OPTS_PUTS("i_version");
1748 if (nodefs
|| sbi
->s_stripe
)
1749 SEQ_OPTS_PRINT("stripe=%lu", sbi
->s_stripe
);
1750 if (EXT4_MOUNT_DATA_FLAGS
& (sbi
->s_mount_opt
^ def_mount_opt
)) {
1751 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
1752 SEQ_OPTS_PUTS("data=journal");
1753 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
1754 SEQ_OPTS_PUTS("data=ordered");
1755 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
1756 SEQ_OPTS_PUTS("data=writeback");
1759 sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
1760 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1761 sbi
->s_inode_readahead_blks
);
1763 if (nodefs
|| (test_opt(sb
, INIT_INODE_TABLE
) &&
1764 (sbi
->s_li_wait_mult
!= EXT4_DEF_LI_WAIT_MULT
)))
1765 SEQ_OPTS_PRINT("init_itable=%u", sbi
->s_li_wait_mult
);
1766 if (nodefs
|| sbi
->s_max_dir_size_kb
)
1767 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi
->s_max_dir_size_kb
);
1769 ext4_show_quota_options(seq
, sb
);
1773 static int ext4_show_options(struct seq_file
*seq
, struct dentry
*root
)
1775 return _ext4_show_options(seq
, root
->d_sb
, 0);
1778 static int options_seq_show(struct seq_file
*seq
, void *offset
)
1780 struct super_block
*sb
= seq
->private;
1783 seq_puts(seq
, (sb
->s_flags
& MS_RDONLY
) ? "ro" : "rw");
1784 rc
= _ext4_show_options(seq
, sb
, 1);
1785 seq_puts(seq
, "\n");
1789 static int options_open_fs(struct inode
*inode
, struct file
*file
)
1791 return single_open(file
, options_seq_show
, PDE_DATA(inode
));
1794 static const struct file_operations ext4_seq_options_fops
= {
1795 .owner
= THIS_MODULE
,
1796 .open
= options_open_fs
,
1798 .llseek
= seq_lseek
,
1799 .release
= single_release
,
1802 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
1805 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1808 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
1809 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
1810 "forcing read-only mode");
1815 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
1816 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
1817 "running e2fsck is recommended");
1818 else if ((sbi
->s_mount_state
& EXT4_ERROR_FS
))
1819 ext4_msg(sb
, KERN_WARNING
,
1820 "warning: mounting fs with errors, "
1821 "running e2fsck is recommended");
1822 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) > 0 &&
1823 le16_to_cpu(es
->s_mnt_count
) >=
1824 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1825 ext4_msg(sb
, KERN_WARNING
,
1826 "warning: maximal mount count reached, "
1827 "running e2fsck is recommended");
1828 else if (le32_to_cpu(es
->s_checkinterval
) &&
1829 (le32_to_cpu(es
->s_lastcheck
) +
1830 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1831 ext4_msg(sb
, KERN_WARNING
,
1832 "warning: checktime reached, "
1833 "running e2fsck is recommended");
1834 if (!sbi
->s_journal
)
1835 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
1836 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1837 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
1838 le16_add_cpu(&es
->s_mnt_count
, 1);
1839 es
->s_mtime
= cpu_to_le32(get_seconds());
1840 ext4_update_dynamic_rev(sb
);
1842 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
1844 ext4_commit_super(sb
, 1);
1846 if (test_opt(sb
, DEBUG
))
1847 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
1848 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1850 sbi
->s_groups_count
,
1851 EXT4_BLOCKS_PER_GROUP(sb
),
1852 EXT4_INODES_PER_GROUP(sb
),
1853 sbi
->s_mount_opt
, sbi
->s_mount_opt2
);
1855 cleancache_init_fs(sb
);
1859 int ext4_alloc_flex_bg_array(struct super_block
*sb
, ext4_group_t ngroup
)
1861 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1862 struct flex_groups
*new_groups
;
1865 if (!sbi
->s_log_groups_per_flex
)
1868 size
= ext4_flex_group(sbi
, ngroup
- 1) + 1;
1869 if (size
<= sbi
->s_flex_groups_allocated
)
1872 size
= roundup_pow_of_two(size
* sizeof(struct flex_groups
));
1873 new_groups
= ext4_kvzalloc(size
, GFP_KERNEL
);
1875 ext4_msg(sb
, KERN_ERR
, "not enough memory for %d flex groups",
1876 size
/ (int) sizeof(struct flex_groups
));
1880 if (sbi
->s_flex_groups
) {
1881 memcpy(new_groups
, sbi
->s_flex_groups
,
1882 (sbi
->s_flex_groups_allocated
*
1883 sizeof(struct flex_groups
)));
1884 ext4_kvfree(sbi
->s_flex_groups
);
1886 sbi
->s_flex_groups
= new_groups
;
1887 sbi
->s_flex_groups_allocated
= size
/ sizeof(struct flex_groups
);
1891 static int ext4_fill_flex_info(struct super_block
*sb
)
1893 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1894 struct ext4_group_desc
*gdp
= NULL
;
1895 ext4_group_t flex_group
;
1896 unsigned int groups_per_flex
= 0;
1899 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
1900 if (sbi
->s_log_groups_per_flex
< 1 || sbi
->s_log_groups_per_flex
> 31) {
1901 sbi
->s_log_groups_per_flex
= 0;
1904 groups_per_flex
= 1U << sbi
->s_log_groups_per_flex
;
1906 err
= ext4_alloc_flex_bg_array(sb
, sbi
->s_groups_count
);
1910 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1911 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1913 flex_group
= ext4_flex_group(sbi
, i
);
1914 atomic_add(ext4_free_inodes_count(sb
, gdp
),
1915 &sbi
->s_flex_groups
[flex_group
].free_inodes
);
1916 atomic64_add(ext4_free_group_clusters(sb
, gdp
),
1917 &sbi
->s_flex_groups
[flex_group
].free_clusters
);
1918 atomic_add(ext4_used_dirs_count(sb
, gdp
),
1919 &sbi
->s_flex_groups
[flex_group
].used_dirs
);
1927 static __le16
ext4_group_desc_csum(struct ext4_sb_info
*sbi
, __u32 block_group
,
1928 struct ext4_group_desc
*gdp
)
1932 __le32 le_group
= cpu_to_le32(block_group
);
1934 if ((sbi
->s_es
->s_feature_ro_compat
&
1935 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
))) {
1936 /* Use new metadata_csum algorithm */
1940 save_csum
= gdp
->bg_checksum
;
1941 gdp
->bg_checksum
= 0;
1942 csum32
= ext4_chksum(sbi
, sbi
->s_csum_seed
, (__u8
*)&le_group
,
1944 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)gdp
,
1946 gdp
->bg_checksum
= save_csum
;
1948 crc
= csum32
& 0xFFFF;
1952 /* old crc16 code */
1953 if (!(sbi
->s_es
->s_feature_ro_compat
&
1954 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)))
1957 offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
1959 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
1960 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
1961 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
1962 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
1963 /* for checksum of struct ext4_group_desc do the rest...*/
1964 if ((sbi
->s_es
->s_feature_incompat
&
1965 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT
)) &&
1966 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
1967 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
1968 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
1972 return cpu_to_le16(crc
);
1975 int ext4_group_desc_csum_verify(struct super_block
*sb
, __u32 block_group
,
1976 struct ext4_group_desc
*gdp
)
1978 if (ext4_has_group_desc_csum(sb
) &&
1979 (gdp
->bg_checksum
!= ext4_group_desc_csum(EXT4_SB(sb
),
1986 void ext4_group_desc_csum_set(struct super_block
*sb
, __u32 block_group
,
1987 struct ext4_group_desc
*gdp
)
1989 if (!ext4_has_group_desc_csum(sb
))
1991 gdp
->bg_checksum
= ext4_group_desc_csum(EXT4_SB(sb
), block_group
, gdp
);
1994 /* Called at mount-time, super-block is locked */
1995 static int ext4_check_descriptors(struct super_block
*sb
,
1996 ext4_group_t
*first_not_zeroed
)
1998 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1999 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
2000 ext4_fsblk_t last_block
;
2001 ext4_fsblk_t block_bitmap
;
2002 ext4_fsblk_t inode_bitmap
;
2003 ext4_fsblk_t inode_table
;
2004 int flexbg_flag
= 0;
2005 ext4_group_t i
, grp
= sbi
->s_groups_count
;
2007 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
2010 ext4_debug("Checking group descriptors");
2012 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2013 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2015 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
2016 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
2018 last_block
= first_block
+
2019 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2021 if ((grp
== sbi
->s_groups_count
) &&
2022 !(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2025 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
2026 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
2027 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2028 "Block bitmap for group %u not in group "
2029 "(block %llu)!", i
, block_bitmap
);
2032 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
2033 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
2034 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2035 "Inode bitmap for group %u not in group "
2036 "(block %llu)!", i
, inode_bitmap
);
2039 inode_table
= ext4_inode_table(sb
, gdp
);
2040 if (inode_table
< first_block
||
2041 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
2042 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2043 "Inode table for group %u not in group "
2044 "(block %llu)!", i
, inode_table
);
2047 ext4_lock_group(sb
, i
);
2048 if (!ext4_group_desc_csum_verify(sb
, i
, gdp
)) {
2049 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2050 "Checksum for group %u failed (%u!=%u)",
2051 i
, le16_to_cpu(ext4_group_desc_csum(sbi
, i
,
2052 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
2053 if (!(sb
->s_flags
& MS_RDONLY
)) {
2054 ext4_unlock_group(sb
, i
);
2058 ext4_unlock_group(sb
, i
);
2060 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
2062 if (NULL
!= first_not_zeroed
)
2063 *first_not_zeroed
= grp
;
2065 ext4_free_blocks_count_set(sbi
->s_es
,
2066 EXT4_C2B(sbi
, ext4_count_free_clusters(sb
)));
2067 sbi
->s_es
->s_free_inodes_count
=cpu_to_le32(ext4_count_free_inodes(sb
));
2071 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2072 * the superblock) which were deleted from all directories, but held open by
2073 * a process at the time of a crash. We walk the list and try to delete these
2074 * inodes at recovery time (only with a read-write filesystem).
2076 * In order to keep the orphan inode chain consistent during traversal (in
2077 * case of crash during recovery), we link each inode into the superblock
2078 * orphan list_head and handle it the same way as an inode deletion during
2079 * normal operation (which journals the operations for us).
2081 * We only do an iget() and an iput() on each inode, which is very safe if we
2082 * accidentally point at an in-use or already deleted inode. The worst that
2083 * can happen in this case is that we get a "bit already cleared" message from
2084 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2085 * e2fsck was run on this filesystem, and it must have already done the orphan
2086 * inode cleanup for us, so we can safely abort without any further action.
2088 static void ext4_orphan_cleanup(struct super_block
*sb
,
2089 struct ext4_super_block
*es
)
2091 unsigned int s_flags
= sb
->s_flags
;
2092 int nr_orphans
= 0, nr_truncates
= 0;
2096 if (!es
->s_last_orphan
) {
2097 jbd_debug(4, "no orphan inodes to clean up\n");
2101 if (bdev_read_only(sb
->s_bdev
)) {
2102 ext4_msg(sb
, KERN_ERR
, "write access "
2103 "unavailable, skipping orphan cleanup");
2107 /* Check if feature set would not allow a r/w mount */
2108 if (!ext4_feature_set_ok(sb
, 0)) {
2109 ext4_msg(sb
, KERN_INFO
, "Skipping orphan cleanup due to "
2110 "unknown ROCOMPAT features");
2114 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2115 /* don't clear list on RO mount w/ errors */
2116 if (es
->s_last_orphan
&& !(s_flags
& MS_RDONLY
)) {
2117 jbd_debug(1, "Errors on filesystem, "
2118 "clearing orphan list.\n");
2119 es
->s_last_orphan
= 0;
2121 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2125 if (s_flags
& MS_RDONLY
) {
2126 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
2127 sb
->s_flags
&= ~MS_RDONLY
;
2130 /* Needed for iput() to work correctly and not trash data */
2131 sb
->s_flags
|= MS_ACTIVE
;
2132 /* Turn on quotas so that they are updated correctly */
2133 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2134 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
2135 int ret
= ext4_quota_on_mount(sb
, i
);
2137 ext4_msg(sb
, KERN_ERR
,
2138 "Cannot turn on journaled "
2139 "quota: error %d", ret
);
2144 while (es
->s_last_orphan
) {
2145 struct inode
*inode
;
2147 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
2148 if (IS_ERR(inode
)) {
2149 es
->s_last_orphan
= 0;
2153 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
2154 dquot_initialize(inode
);
2155 if (inode
->i_nlink
) {
2156 ext4_msg(sb
, KERN_DEBUG
,
2157 "%s: truncating inode %lu to %lld bytes",
2158 __func__
, inode
->i_ino
, inode
->i_size
);
2159 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2160 inode
->i_ino
, inode
->i_size
);
2161 mutex_lock(&inode
->i_mutex
);
2162 ext4_truncate(inode
);
2163 mutex_unlock(&inode
->i_mutex
);
2166 ext4_msg(sb
, KERN_DEBUG
,
2167 "%s: deleting unreferenced inode %lu",
2168 __func__
, inode
->i_ino
);
2169 jbd_debug(2, "deleting unreferenced inode %lu\n",
2173 iput(inode
); /* The delete magic happens here! */
2176 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2179 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
2180 PLURAL(nr_orphans
));
2182 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
2183 PLURAL(nr_truncates
));
2185 /* Turn quotas off */
2186 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2187 if (sb_dqopt(sb
)->files
[i
])
2188 dquot_quota_off(sb
, i
);
2191 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
2195 * Maximal extent format file size.
2196 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2197 * extent format containers, within a sector_t, and within i_blocks
2198 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2199 * so that won't be a limiting factor.
2201 * However there is other limiting factor. We do store extents in the form
2202 * of starting block and length, hence the resulting length of the extent
2203 * covering maximum file size must fit into on-disk format containers as
2204 * well. Given that length is always by 1 unit bigger than max unit (because
2205 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2207 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2209 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
2212 loff_t upper_limit
= MAX_LFS_FILESIZE
;
2214 /* small i_blocks in vfs inode? */
2215 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2217 * CONFIG_LBDAF is not enabled implies the inode
2218 * i_block represent total blocks in 512 bytes
2219 * 32 == size of vfs inode i_blocks * 8
2221 upper_limit
= (1LL << 32) - 1;
2223 /* total blocks in file system block size */
2224 upper_limit
>>= (blkbits
- 9);
2225 upper_limit
<<= blkbits
;
2229 * 32-bit extent-start container, ee_block. We lower the maxbytes
2230 * by one fs block, so ee_len can cover the extent of maximum file
2233 res
= (1LL << 32) - 1;
2236 /* Sanity check against vm- & vfs- imposed limits */
2237 if (res
> upper_limit
)
2244 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2245 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2246 * We need to be 1 filesystem block less than the 2^48 sector limit.
2248 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
2250 loff_t res
= EXT4_NDIR_BLOCKS
;
2253 /* This is calculated to be the largest file size for a dense, block
2254 * mapped file such that the file's total number of 512-byte sectors,
2255 * including data and all indirect blocks, does not exceed (2^48 - 1).
2257 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2258 * number of 512-byte sectors of the file.
2261 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2263 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2264 * the inode i_block field represents total file blocks in
2265 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2267 upper_limit
= (1LL << 32) - 1;
2269 /* total blocks in file system block size */
2270 upper_limit
>>= (bits
- 9);
2274 * We use 48 bit ext4_inode i_blocks
2275 * With EXT4_HUGE_FILE_FL set the i_blocks
2276 * represent total number of blocks in
2277 * file system block size
2279 upper_limit
= (1LL << 48) - 1;
2283 /* indirect blocks */
2285 /* double indirect blocks */
2286 meta_blocks
+= 1 + (1LL << (bits
-2));
2287 /* tripple indirect blocks */
2288 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
2290 upper_limit
-= meta_blocks
;
2291 upper_limit
<<= bits
;
2293 res
+= 1LL << (bits
-2);
2294 res
+= 1LL << (2*(bits
-2));
2295 res
+= 1LL << (3*(bits
-2));
2297 if (res
> upper_limit
)
2300 if (res
> MAX_LFS_FILESIZE
)
2301 res
= MAX_LFS_FILESIZE
;
2306 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
2307 ext4_fsblk_t logical_sb_block
, int nr
)
2309 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2310 ext4_group_t bg
, first_meta_bg
;
2313 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
2315 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_META_BG
) ||
2317 return logical_sb_block
+ nr
+ 1;
2318 bg
= sbi
->s_desc_per_block
* nr
;
2319 if (ext4_bg_has_super(sb
, bg
))
2322 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
2326 * ext4_get_stripe_size: Get the stripe size.
2327 * @sbi: In memory super block info
2329 * If we have specified it via mount option, then
2330 * use the mount option value. If the value specified at mount time is
2331 * greater than the blocks per group use the super block value.
2332 * If the super block value is greater than blocks per group return 0.
2333 * Allocator needs it be less than blocks per group.
2336 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
2338 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
2339 unsigned long stripe_width
=
2340 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
2343 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
2344 ret
= sbi
->s_stripe
;
2345 else if (stripe_width
<= sbi
->s_blocks_per_group
)
2347 else if (stride
<= sbi
->s_blocks_per_group
)
2353 * If the stripe width is 1, this makes no sense and
2354 * we set it to 0 to turn off stripe handling code.
2365 struct attribute attr
;
2366 ssize_t (*show
)(struct ext4_attr
*, struct ext4_sb_info
*, char *);
2367 ssize_t (*store
)(struct ext4_attr
*, struct ext4_sb_info
*,
2368 const char *, size_t);
2372 static int parse_strtoull(const char *buf
,
2373 unsigned long long max
, unsigned long long *value
)
2377 ret
= kstrtoull(skip_spaces(buf
), 0, value
);
2378 if (!ret
&& *value
> max
)
2383 static ssize_t
delayed_allocation_blocks_show(struct ext4_attr
*a
,
2384 struct ext4_sb_info
*sbi
,
2387 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2389 percpu_counter_sum(&sbi
->s_dirtyclusters_counter
)));
2392 static ssize_t
session_write_kbytes_show(struct ext4_attr
*a
,
2393 struct ext4_sb_info
*sbi
, char *buf
)
2395 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2397 if (!sb
->s_bdev
->bd_part
)
2398 return snprintf(buf
, PAGE_SIZE
, "0\n");
2399 return snprintf(buf
, PAGE_SIZE
, "%lu\n",
2400 (part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2401 sbi
->s_sectors_written_start
) >> 1);
2404 static ssize_t
lifetime_write_kbytes_show(struct ext4_attr
*a
,
2405 struct ext4_sb_info
*sbi
, char *buf
)
2407 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2409 if (!sb
->s_bdev
->bd_part
)
2410 return snprintf(buf
, PAGE_SIZE
, "0\n");
2411 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2412 (unsigned long long)(sbi
->s_kbytes_written
+
2413 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2414 EXT4_SB(sb
)->s_sectors_written_start
) >> 1)));
2417 static ssize_t
inode_readahead_blks_store(struct ext4_attr
*a
,
2418 struct ext4_sb_info
*sbi
,
2419 const char *buf
, size_t count
)
2424 ret
= kstrtoul(skip_spaces(buf
), 0, &t
);
2428 if (t
&& (!is_power_of_2(t
) || t
> 0x40000000))
2431 sbi
->s_inode_readahead_blks
= t
;
2435 static ssize_t
sbi_ui_show(struct ext4_attr
*a
,
2436 struct ext4_sb_info
*sbi
, char *buf
)
2438 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2440 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
2443 static ssize_t
sbi_ui_store(struct ext4_attr
*a
,
2444 struct ext4_sb_info
*sbi
,
2445 const char *buf
, size_t count
)
2447 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2451 ret
= kstrtoul(skip_spaces(buf
), 0, &t
);
2458 static ssize_t
reserved_clusters_show(struct ext4_attr
*a
,
2459 struct ext4_sb_info
*sbi
, char *buf
)
2461 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2462 (unsigned long long) atomic64_read(&sbi
->s_resv_clusters
));
2465 static ssize_t
reserved_clusters_store(struct ext4_attr
*a
,
2466 struct ext4_sb_info
*sbi
,
2467 const char *buf
, size_t count
)
2469 unsigned long long val
;
2472 if (parse_strtoull(buf
, -1ULL, &val
))
2474 ret
= ext4_reserve_clusters(sbi
, val
);
2476 return ret
? ret
: count
;
2479 static ssize_t
trigger_test_error(struct ext4_attr
*a
,
2480 struct ext4_sb_info
*sbi
,
2481 const char *buf
, size_t count
)
2485 if (!capable(CAP_SYS_ADMIN
))
2488 if (len
&& buf
[len
-1] == '\n')
2492 ext4_error(sbi
->s_sb
, "%.*s", len
, buf
);
2496 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2497 static struct ext4_attr ext4_attr_##_name = { \
2498 .attr = {.name = __stringify(_name), .mode = _mode }, \
2501 .offset = offsetof(struct ext4_sb_info, _elname), \
2503 #define EXT4_ATTR(name, mode, show, store) \
2504 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2506 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2507 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2508 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2509 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2510 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2511 #define ATTR_LIST(name) &ext4_attr_##name.attr
2513 EXT4_RO_ATTR(delayed_allocation_blocks
);
2514 EXT4_RO_ATTR(session_write_kbytes
);
2515 EXT4_RO_ATTR(lifetime_write_kbytes
);
2516 EXT4_RW_ATTR(reserved_clusters
);
2517 EXT4_ATTR_OFFSET(inode_readahead_blks
, 0644, sbi_ui_show
,
2518 inode_readahead_blks_store
, s_inode_readahead_blks
);
2519 EXT4_RW_ATTR_SBI_UI(inode_goal
, s_inode_goal
);
2520 EXT4_RW_ATTR_SBI_UI(mb_stats
, s_mb_stats
);
2521 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan
, s_mb_max_to_scan
);
2522 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan
, s_mb_min_to_scan
);
2523 EXT4_RW_ATTR_SBI_UI(mb_order2_req
, s_mb_order2_reqs
);
2524 EXT4_RW_ATTR_SBI_UI(mb_stream_req
, s_mb_stream_request
);
2525 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc
, s_mb_group_prealloc
);
2526 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump
, s_max_writeback_mb_bump
);
2527 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb
, s_extent_max_zeroout_kb
);
2528 EXT4_ATTR(trigger_fs_error
, 0200, NULL
, trigger_test_error
);
2530 static struct attribute
*ext4_attrs
[] = {
2531 ATTR_LIST(delayed_allocation_blocks
),
2532 ATTR_LIST(session_write_kbytes
),
2533 ATTR_LIST(lifetime_write_kbytes
),
2534 ATTR_LIST(reserved_clusters
),
2535 ATTR_LIST(inode_readahead_blks
),
2536 ATTR_LIST(inode_goal
),
2537 ATTR_LIST(mb_stats
),
2538 ATTR_LIST(mb_max_to_scan
),
2539 ATTR_LIST(mb_min_to_scan
),
2540 ATTR_LIST(mb_order2_req
),
2541 ATTR_LIST(mb_stream_req
),
2542 ATTR_LIST(mb_group_prealloc
),
2543 ATTR_LIST(max_writeback_mb_bump
),
2544 ATTR_LIST(extent_max_zeroout_kb
),
2545 ATTR_LIST(trigger_fs_error
),
2549 /* Features this copy of ext4 supports */
2550 EXT4_INFO_ATTR(lazy_itable_init
);
2551 EXT4_INFO_ATTR(batched_discard
);
2552 EXT4_INFO_ATTR(meta_bg_resize
);
2554 static struct attribute
*ext4_feat_attrs
[] = {
2555 ATTR_LIST(lazy_itable_init
),
2556 ATTR_LIST(batched_discard
),
2557 ATTR_LIST(meta_bg_resize
),
2561 static ssize_t
ext4_attr_show(struct kobject
*kobj
,
2562 struct attribute
*attr
, char *buf
)
2564 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2566 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2568 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
2571 static ssize_t
ext4_attr_store(struct kobject
*kobj
,
2572 struct attribute
*attr
,
2573 const char *buf
, size_t len
)
2575 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2577 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2579 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
2582 static void ext4_sb_release(struct kobject
*kobj
)
2584 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2586 complete(&sbi
->s_kobj_unregister
);
2589 static const struct sysfs_ops ext4_attr_ops
= {
2590 .show
= ext4_attr_show
,
2591 .store
= ext4_attr_store
,
2594 static struct kobj_type ext4_ktype
= {
2595 .default_attrs
= ext4_attrs
,
2596 .sysfs_ops
= &ext4_attr_ops
,
2597 .release
= ext4_sb_release
,
2600 static void ext4_feat_release(struct kobject
*kobj
)
2602 complete(&ext4_feat
->f_kobj_unregister
);
2605 static struct kobj_type ext4_feat_ktype
= {
2606 .default_attrs
= ext4_feat_attrs
,
2607 .sysfs_ops
= &ext4_attr_ops
,
2608 .release
= ext4_feat_release
,
2612 * Check whether this filesystem can be mounted based on
2613 * the features present and the RDONLY/RDWR mount requested.
2614 * Returns 1 if this filesystem can be mounted as requested,
2615 * 0 if it cannot be.
2617 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
2619 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT4_FEATURE_INCOMPAT_SUPP
)) {
2620 ext4_msg(sb
, KERN_ERR
,
2621 "Couldn't mount because of "
2622 "unsupported optional features (%x)",
2623 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2624 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2631 /* Check that feature set is OK for a read-write mount */
2632 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT4_FEATURE_RO_COMPAT_SUPP
)) {
2633 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
2634 "unsupported optional features (%x)",
2635 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2636 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2640 * Large file size enabled file system can only be mounted
2641 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2643 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_HUGE_FILE
)) {
2644 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
2645 ext4_msg(sb
, KERN_ERR
, "Filesystem with huge files "
2646 "cannot be mounted RDWR without "
2651 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_BIGALLOC
) &&
2652 !EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_EXTENTS
)) {
2653 ext4_msg(sb
, KERN_ERR
,
2654 "Can't support bigalloc feature without "
2655 "extents feature\n");
2659 #ifndef CONFIG_QUOTA
2660 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
) &&
2662 ext4_msg(sb
, KERN_ERR
,
2663 "Filesystem with quota feature cannot be mounted RDWR "
2664 "without CONFIG_QUOTA");
2667 #endif /* CONFIG_QUOTA */
2672 * This function is called once a day if we have errors logged
2673 * on the file system
2675 static void print_daily_error_info(unsigned long arg
)
2677 struct super_block
*sb
= (struct super_block
*) arg
;
2678 struct ext4_sb_info
*sbi
;
2679 struct ext4_super_block
*es
;
2684 if (es
->s_error_count
)
2685 /* fsck newer than v1.41.13 is needed to clean this condition. */
2686 ext4_msg(sb
, KERN_NOTICE
, "error count since last fsck: %u",
2687 le32_to_cpu(es
->s_error_count
));
2688 if (es
->s_first_error_time
) {
2689 printk(KERN_NOTICE
"EXT4-fs (%s): initial error at time %u: %.*s:%d",
2690 sb
->s_id
, le32_to_cpu(es
->s_first_error_time
),
2691 (int) sizeof(es
->s_first_error_func
),
2692 es
->s_first_error_func
,
2693 le32_to_cpu(es
->s_first_error_line
));
2694 if (es
->s_first_error_ino
)
2695 printk(": inode %u",
2696 le32_to_cpu(es
->s_first_error_ino
));
2697 if (es
->s_first_error_block
)
2698 printk(": block %llu", (unsigned long long)
2699 le64_to_cpu(es
->s_first_error_block
));
2702 if (es
->s_last_error_time
) {
2703 printk(KERN_NOTICE
"EXT4-fs (%s): last error at time %u: %.*s:%d",
2704 sb
->s_id
, le32_to_cpu(es
->s_last_error_time
),
2705 (int) sizeof(es
->s_last_error_func
),
2706 es
->s_last_error_func
,
2707 le32_to_cpu(es
->s_last_error_line
));
2708 if (es
->s_last_error_ino
)
2709 printk(": inode %u",
2710 le32_to_cpu(es
->s_last_error_ino
));
2711 if (es
->s_last_error_block
)
2712 printk(": block %llu", (unsigned long long)
2713 le64_to_cpu(es
->s_last_error_block
));
2716 mod_timer(&sbi
->s_err_report
, jiffies
+ 24*60*60*HZ
); /* Once a day */
2719 /* Find next suitable group and run ext4_init_inode_table */
2720 static int ext4_run_li_request(struct ext4_li_request
*elr
)
2722 struct ext4_group_desc
*gdp
= NULL
;
2723 ext4_group_t group
, ngroups
;
2724 struct super_block
*sb
;
2725 unsigned long timeout
= 0;
2729 ngroups
= EXT4_SB(sb
)->s_groups_count
;
2732 for (group
= elr
->lr_next_group
; group
< ngroups
; group
++) {
2733 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2739 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2743 if (group
>= ngroups
)
2748 ret
= ext4_init_inode_table(sb
, group
,
2749 elr
->lr_timeout
? 0 : 1);
2750 if (elr
->lr_timeout
== 0) {
2751 timeout
= (jiffies
- timeout
) *
2752 elr
->lr_sbi
->s_li_wait_mult
;
2753 elr
->lr_timeout
= timeout
;
2755 elr
->lr_next_sched
= jiffies
+ elr
->lr_timeout
;
2756 elr
->lr_next_group
= group
+ 1;
2764 * Remove lr_request from the list_request and free the
2765 * request structure. Should be called with li_list_mtx held
2767 static void ext4_remove_li_request(struct ext4_li_request
*elr
)
2769 struct ext4_sb_info
*sbi
;
2776 list_del(&elr
->lr_request
);
2777 sbi
->s_li_request
= NULL
;
2781 static void ext4_unregister_li_request(struct super_block
*sb
)
2783 mutex_lock(&ext4_li_mtx
);
2784 if (!ext4_li_info
) {
2785 mutex_unlock(&ext4_li_mtx
);
2789 mutex_lock(&ext4_li_info
->li_list_mtx
);
2790 ext4_remove_li_request(EXT4_SB(sb
)->s_li_request
);
2791 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2792 mutex_unlock(&ext4_li_mtx
);
2795 static struct task_struct
*ext4_lazyinit_task
;
2798 * This is the function where ext4lazyinit thread lives. It walks
2799 * through the request list searching for next scheduled filesystem.
2800 * When such a fs is found, run the lazy initialization request
2801 * (ext4_rn_li_request) and keep track of the time spend in this
2802 * function. Based on that time we compute next schedule time of
2803 * the request. When walking through the list is complete, compute
2804 * next waking time and put itself into sleep.
2806 static int ext4_lazyinit_thread(void *arg
)
2808 struct ext4_lazy_init
*eli
= (struct ext4_lazy_init
*)arg
;
2809 struct list_head
*pos
, *n
;
2810 struct ext4_li_request
*elr
;
2811 unsigned long next_wakeup
, cur
;
2813 BUG_ON(NULL
== eli
);
2818 next_wakeup
= MAX_JIFFY_OFFSET
;
2820 mutex_lock(&eli
->li_list_mtx
);
2821 if (list_empty(&eli
->li_request_list
)) {
2822 mutex_unlock(&eli
->li_list_mtx
);
2826 list_for_each_safe(pos
, n
, &eli
->li_request_list
) {
2827 elr
= list_entry(pos
, struct ext4_li_request
,
2830 if (time_after_eq(jiffies
, elr
->lr_next_sched
)) {
2831 if (ext4_run_li_request(elr
) != 0) {
2832 /* error, remove the lazy_init job */
2833 ext4_remove_li_request(elr
);
2838 if (time_before(elr
->lr_next_sched
, next_wakeup
))
2839 next_wakeup
= elr
->lr_next_sched
;
2841 mutex_unlock(&eli
->li_list_mtx
);
2846 if ((time_after_eq(cur
, next_wakeup
)) ||
2847 (MAX_JIFFY_OFFSET
== next_wakeup
)) {
2852 schedule_timeout_interruptible(next_wakeup
- cur
);
2854 if (kthread_freezable_should_stop(NULL
)) {
2855 ext4_clear_request_list();
2862 * It looks like the request list is empty, but we need
2863 * to check it under the li_list_mtx lock, to prevent any
2864 * additions into it, and of course we should lock ext4_li_mtx
2865 * to atomically free the list and ext4_li_info, because at
2866 * this point another ext4 filesystem could be registering
2869 mutex_lock(&ext4_li_mtx
);
2870 mutex_lock(&eli
->li_list_mtx
);
2871 if (!list_empty(&eli
->li_request_list
)) {
2872 mutex_unlock(&eli
->li_list_mtx
);
2873 mutex_unlock(&ext4_li_mtx
);
2876 mutex_unlock(&eli
->li_list_mtx
);
2877 kfree(ext4_li_info
);
2878 ext4_li_info
= NULL
;
2879 mutex_unlock(&ext4_li_mtx
);
2884 static void ext4_clear_request_list(void)
2886 struct list_head
*pos
, *n
;
2887 struct ext4_li_request
*elr
;
2889 mutex_lock(&ext4_li_info
->li_list_mtx
);
2890 list_for_each_safe(pos
, n
, &ext4_li_info
->li_request_list
) {
2891 elr
= list_entry(pos
, struct ext4_li_request
,
2893 ext4_remove_li_request(elr
);
2895 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2898 static int ext4_run_lazyinit_thread(void)
2900 ext4_lazyinit_task
= kthread_run(ext4_lazyinit_thread
,
2901 ext4_li_info
, "ext4lazyinit");
2902 if (IS_ERR(ext4_lazyinit_task
)) {
2903 int err
= PTR_ERR(ext4_lazyinit_task
);
2904 ext4_clear_request_list();
2905 kfree(ext4_li_info
);
2906 ext4_li_info
= NULL
;
2907 printk(KERN_CRIT
"EXT4-fs: error %d creating inode table "
2908 "initialization thread\n",
2912 ext4_li_info
->li_state
|= EXT4_LAZYINIT_RUNNING
;
2917 * Check whether it make sense to run itable init. thread or not.
2918 * If there is at least one uninitialized inode table, return
2919 * corresponding group number, else the loop goes through all
2920 * groups and return total number of groups.
2922 static ext4_group_t
ext4_has_uninit_itable(struct super_block
*sb
)
2924 ext4_group_t group
, ngroups
= EXT4_SB(sb
)->s_groups_count
;
2925 struct ext4_group_desc
*gdp
= NULL
;
2927 for (group
= 0; group
< ngroups
; group
++) {
2928 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2932 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2939 static int ext4_li_info_new(void)
2941 struct ext4_lazy_init
*eli
= NULL
;
2943 eli
= kzalloc(sizeof(*eli
), GFP_KERNEL
);
2947 INIT_LIST_HEAD(&eli
->li_request_list
);
2948 mutex_init(&eli
->li_list_mtx
);
2950 eli
->li_state
|= EXT4_LAZYINIT_QUIT
;
2957 static struct ext4_li_request
*ext4_li_request_new(struct super_block
*sb
,
2960 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2961 struct ext4_li_request
*elr
;
2964 elr
= kzalloc(sizeof(*elr
), GFP_KERNEL
);
2970 elr
->lr_next_group
= start
;
2973 * Randomize first schedule time of the request to
2974 * spread the inode table initialization requests
2977 get_random_bytes(&rnd
, sizeof(rnd
));
2978 elr
->lr_next_sched
= jiffies
+ (unsigned long)rnd
%
2979 (EXT4_DEF_LI_MAX_START_DELAY
* HZ
);
2984 int ext4_register_li_request(struct super_block
*sb
,
2985 ext4_group_t first_not_zeroed
)
2987 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2988 struct ext4_li_request
*elr
= NULL
;
2989 ext4_group_t ngroups
= EXT4_SB(sb
)->s_groups_count
;
2992 mutex_lock(&ext4_li_mtx
);
2993 if (sbi
->s_li_request
!= NULL
) {
2995 * Reset timeout so it can be computed again, because
2996 * s_li_wait_mult might have changed.
2998 sbi
->s_li_request
->lr_timeout
= 0;
3002 if (first_not_zeroed
== ngroups
||
3003 (sb
->s_flags
& MS_RDONLY
) ||
3004 !test_opt(sb
, INIT_INODE_TABLE
))
3007 elr
= ext4_li_request_new(sb
, first_not_zeroed
);
3013 if (NULL
== ext4_li_info
) {
3014 ret
= ext4_li_info_new();
3019 mutex_lock(&ext4_li_info
->li_list_mtx
);
3020 list_add(&elr
->lr_request
, &ext4_li_info
->li_request_list
);
3021 mutex_unlock(&ext4_li_info
->li_list_mtx
);
3023 sbi
->s_li_request
= elr
;
3025 * set elr to NULL here since it has been inserted to
3026 * the request_list and the removal and free of it is
3027 * handled by ext4_clear_request_list from now on.
3031 if (!(ext4_li_info
->li_state
& EXT4_LAZYINIT_RUNNING
)) {
3032 ret
= ext4_run_lazyinit_thread();
3037 mutex_unlock(&ext4_li_mtx
);
3044 * We do not need to lock anything since this is called on
3047 static void ext4_destroy_lazyinit_thread(void)
3050 * If thread exited earlier
3051 * there's nothing to be done.
3053 if (!ext4_li_info
|| !ext4_lazyinit_task
)
3056 kthread_stop(ext4_lazyinit_task
);
3059 static int set_journal_csum_feature_set(struct super_block
*sb
)
3062 int compat
, incompat
;
3063 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3065 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3066 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
)) {
3067 /* journal checksum v2 */
3069 incompat
= JBD2_FEATURE_INCOMPAT_CSUM_V2
;
3071 /* journal checksum v1 */
3072 compat
= JBD2_FEATURE_COMPAT_CHECKSUM
;
3076 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
3077 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3079 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
|
3081 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
3082 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3085 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3086 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3088 jbd2_journal_clear_features(sbi
->s_journal
,
3089 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3090 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
|
3091 JBD2_FEATURE_INCOMPAT_CSUM_V2
);
3098 * Note: calculating the overhead so we can be compatible with
3099 * historical BSD practice is quite difficult in the face of
3100 * clusters/bigalloc. This is because multiple metadata blocks from
3101 * different block group can end up in the same allocation cluster.
3102 * Calculating the exact overhead in the face of clustered allocation
3103 * requires either O(all block bitmaps) in memory or O(number of block
3104 * groups**2) in time. We will still calculate the superblock for
3105 * older file systems --- and if we come across with a bigalloc file
3106 * system with zero in s_overhead_clusters the estimate will be close to
3107 * correct especially for very large cluster sizes --- but for newer
3108 * file systems, it's better to calculate this figure once at mkfs
3109 * time, and store it in the superblock. If the superblock value is
3110 * present (even for non-bigalloc file systems), we will use it.
3112 static int count_overhead(struct super_block
*sb
, ext4_group_t grp
,
3115 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3116 struct ext4_group_desc
*gdp
;
3117 ext4_fsblk_t first_block
, last_block
, b
;
3118 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3119 int s
, j
, count
= 0;
3121 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_BIGALLOC
))
3122 return (ext4_bg_has_super(sb
, grp
) + ext4_bg_num_gdb(sb
, grp
) +
3123 sbi
->s_itb_per_group
+ 2);
3125 first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
) +
3126 (grp
* EXT4_BLOCKS_PER_GROUP(sb
));
3127 last_block
= first_block
+ EXT4_BLOCKS_PER_GROUP(sb
) - 1;
3128 for (i
= 0; i
< ngroups
; i
++) {
3129 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
3130 b
= ext4_block_bitmap(sb
, gdp
);
3131 if (b
>= first_block
&& b
<= last_block
) {
3132 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3135 b
= ext4_inode_bitmap(sb
, gdp
);
3136 if (b
>= first_block
&& b
<= last_block
) {
3137 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3140 b
= ext4_inode_table(sb
, gdp
);
3141 if (b
>= first_block
&& b
+ sbi
->s_itb_per_group
<= last_block
)
3142 for (j
= 0; j
< sbi
->s_itb_per_group
; j
++, b
++) {
3143 int c
= EXT4_B2C(sbi
, b
- first_block
);
3144 ext4_set_bit(c
, buf
);
3150 if (ext4_bg_has_super(sb
, grp
)) {
3151 ext4_set_bit(s
++, buf
);
3154 for (j
= ext4_bg_num_gdb(sb
, grp
); j
> 0; j
--) {
3155 ext4_set_bit(EXT4_B2C(sbi
, s
++), buf
);
3161 return EXT4_CLUSTERS_PER_GROUP(sb
) -
3162 ext4_count_free(buf
, EXT4_CLUSTERS_PER_GROUP(sb
) / 8);
3166 * Compute the overhead and stash it in sbi->s_overhead
3168 int ext4_calculate_overhead(struct super_block
*sb
)
3170 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3171 struct ext4_super_block
*es
= sbi
->s_es
;
3172 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3173 ext4_fsblk_t overhead
= 0;
3174 char *buf
= (char *) get_zeroed_page(GFP_KERNEL
);
3180 * Compute the overhead (FS structures). This is constant
3181 * for a given filesystem unless the number of block groups
3182 * changes so we cache the previous value until it does.
3186 * All of the blocks before first_data_block are overhead
3188 overhead
= EXT4_B2C(sbi
, le32_to_cpu(es
->s_first_data_block
));
3191 * Add the overhead found in each block group
3193 for (i
= 0; i
< ngroups
; i
++) {
3196 blks
= count_overhead(sb
, i
, buf
);
3199 memset(buf
, 0, PAGE_SIZE
);
3202 /* Add the journal blocks as well */
3204 overhead
+= EXT4_NUM_B2C(sbi
, sbi
->s_journal
->j_maxlen
);
3206 sbi
->s_overhead
= overhead
;
3208 free_page((unsigned long) buf
);
3213 static ext4_fsblk_t
ext4_calculate_resv_clusters(struct super_block
*sb
)
3215 ext4_fsblk_t resv_clusters
;
3218 * There's no need to reserve anything when we aren't using extents.
3219 * The space estimates are exact, there are no unwritten extents,
3220 * hole punching doesn't need new metadata... This is needed especially
3221 * to keep ext2/3 backward compatibility.
3223 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_EXTENTS
))
3226 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3227 * This should cover the situations where we can not afford to run
3228 * out of space like for example punch hole, or converting
3229 * uninitialized extents in delalloc path. In most cases such
3230 * allocation would require 1, or 2 blocks, higher numbers are
3233 resv_clusters
= ext4_blocks_count(EXT4_SB(sb
)->s_es
) >>
3234 EXT4_SB(sb
)->s_cluster_bits
;
3236 do_div(resv_clusters
, 50);
3237 resv_clusters
= min_t(ext4_fsblk_t
, resv_clusters
, 4096);
3239 return resv_clusters
;
3243 static int ext4_reserve_clusters(struct ext4_sb_info
*sbi
, ext4_fsblk_t count
)
3245 ext4_fsblk_t clusters
= ext4_blocks_count(sbi
->s_es
) >>
3246 sbi
->s_cluster_bits
;
3248 if (count
>= clusters
)
3251 atomic64_set(&sbi
->s_resv_clusters
, count
);
3255 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
3257 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
3258 struct buffer_head
*bh
;
3259 struct ext4_super_block
*es
= NULL
;
3260 struct ext4_sb_info
*sbi
;
3262 ext4_fsblk_t sb_block
= get_sb_block(&data
);
3263 ext4_fsblk_t logical_sb_block
;
3264 unsigned long offset
= 0;
3265 unsigned long journal_devnum
= 0;
3266 unsigned long def_mount_opts
;
3271 int blocksize
, clustersize
;
3272 unsigned int db_count
;
3274 int needs_recovery
, has_huge_files
, has_bigalloc
;
3277 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
3278 ext4_group_t first_not_zeroed
;
3280 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
3284 sbi
->s_blockgroup_lock
=
3285 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
3286 if (!sbi
->s_blockgroup_lock
) {
3290 sb
->s_fs_info
= sbi
;
3292 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
3293 sbi
->s_sb_block
= sb_block
;
3294 if (sb
->s_bdev
->bd_part
)
3295 sbi
->s_sectors_written_start
=
3296 part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]);
3298 /* Cleanup superblock name */
3299 for (cp
= sb
->s_id
; (cp
= strchr(cp
, '/'));)
3302 /* -EINVAL is default */
3304 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
3306 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
3311 * The ext4 superblock will not be buffer aligned for other than 1kB
3312 * block sizes. We need to calculate the offset from buffer start.
3314 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
3315 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3316 offset
= do_div(logical_sb_block
, blocksize
);
3318 logical_sb_block
= sb_block
;
3321 if (!(bh
= sb_bread(sb
, logical_sb_block
))) {
3322 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
3326 * Note: s_es must be initialized as soon as possible because
3327 * some ext4 macro-instructions depend on its value
3329 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
3331 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
3332 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
3334 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
3336 /* Warn if metadata_csum and gdt_csum are both set. */
3337 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3338 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
) &&
3339 EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_GDT_CSUM
))
3340 ext4_warning(sb
, KERN_INFO
"metadata_csum and uninit_bg are "
3341 "redundant flags; please run fsck.");
3343 /* Check for a known checksum algorithm */
3344 if (!ext4_verify_csum_type(sb
, es
)) {
3345 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3346 "unknown checksum algorithm.");
3351 /* Load the checksum driver */
3352 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3353 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
)) {
3354 sbi
->s_chksum_driver
= crypto_alloc_shash("crc32c", 0, 0);
3355 if (IS_ERR(sbi
->s_chksum_driver
)) {
3356 ext4_msg(sb
, KERN_ERR
, "Cannot load crc32c driver.");
3357 ret
= PTR_ERR(sbi
->s_chksum_driver
);
3358 sbi
->s_chksum_driver
= NULL
;
3363 /* Check superblock checksum */
3364 if (!ext4_superblock_csum_verify(sb
, es
)) {
3365 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3366 "invalid superblock checksum. Run e2fsck?");
3371 /* Precompute checksum seed for all metadata */
3372 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3373 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
))
3374 sbi
->s_csum_seed
= ext4_chksum(sbi
, ~0, es
->s_uuid
,
3375 sizeof(es
->s_uuid
));
3377 /* Set defaults before we parse the mount options */
3378 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
3379 set_opt(sb
, INIT_INODE_TABLE
);
3380 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
3382 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
)
3384 if (def_mount_opts
& EXT4_DEFM_UID16
)
3385 set_opt(sb
, NO_UID32
);
3386 /* xattr user namespace & acls are now defaulted on */
3387 set_opt(sb
, XATTR_USER
);
3388 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3389 set_opt(sb
, POSIX_ACL
);
3391 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
3392 set_opt(sb
, JOURNAL_DATA
);
3393 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
3394 set_opt(sb
, ORDERED_DATA
);
3395 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
3396 set_opt(sb
, WRITEBACK_DATA
);
3398 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
3399 set_opt(sb
, ERRORS_PANIC
);
3400 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
3401 set_opt(sb
, ERRORS_CONT
);
3403 set_opt(sb
, ERRORS_RO
);
3404 if (def_mount_opts
& EXT4_DEFM_BLOCK_VALIDITY
)
3405 set_opt(sb
, BLOCK_VALIDITY
);
3406 if (def_mount_opts
& EXT4_DEFM_DISCARD
)
3407 set_opt(sb
, DISCARD
);
3409 sbi
->s_resuid
= make_kuid(&init_user_ns
, le16_to_cpu(es
->s_def_resuid
));
3410 sbi
->s_resgid
= make_kgid(&init_user_ns
, le16_to_cpu(es
->s_def_resgid
));
3411 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
3412 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
3413 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
3415 if ((def_mount_opts
& EXT4_DEFM_NOBARRIER
) == 0)
3416 set_opt(sb
, BARRIER
);
3419 * enable delayed allocation by default
3420 * Use -o nodelalloc to turn it off
3422 if (!IS_EXT3_SB(sb
) && !IS_EXT2_SB(sb
) &&
3423 ((def_mount_opts
& EXT4_DEFM_NODELALLOC
) == 0))
3424 set_opt(sb
, DELALLOC
);
3427 * set default s_li_wait_mult for lazyinit, for the case there is
3428 * no mount option specified.
3430 sbi
->s_li_wait_mult
= EXT4_DEF_LI_WAIT_MULT
;
3432 if (!parse_options((char *) sbi
->s_es
->s_mount_opts
, sb
,
3433 &journal_devnum
, &journal_ioprio
, 0)) {
3434 ext4_msg(sb
, KERN_WARNING
,
3435 "failed to parse options in superblock: %s",
3436 sbi
->s_es
->s_mount_opts
);
3438 sbi
->s_def_mount_opt
= sbi
->s_mount_opt
;
3439 if (!parse_options((char *) data
, sb
, &journal_devnum
,
3440 &journal_ioprio
, 0))
3443 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
3444 printk_once(KERN_WARNING
"EXT4-fs: Warning: mounting "
3445 "with data=journal disables delayed "
3446 "allocation and O_DIRECT support!\n");
3447 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
3448 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3449 "both data=journal and delalloc");
3452 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
3453 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3454 "both data=journal and dioread_nolock");
3457 if (test_opt(sb
, DELALLOC
))
3458 clear_opt(sb
, DELALLOC
);
3461 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3462 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
3464 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
3465 (EXT4_HAS_COMPAT_FEATURE(sb
, ~0U) ||
3466 EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
3467 EXT4_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
3468 ext4_msg(sb
, KERN_WARNING
,
3469 "feature flags set on rev 0 fs, "
3470 "running e2fsck is recommended");
3472 if (IS_EXT2_SB(sb
)) {
3473 if (ext2_feature_set_ok(sb
))
3474 ext4_msg(sb
, KERN_INFO
, "mounting ext2 file system "
3475 "using the ext4 subsystem");
3477 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext2 due "
3478 "to feature incompatibilities");
3483 if (IS_EXT3_SB(sb
)) {
3484 if (ext3_feature_set_ok(sb
))
3485 ext4_msg(sb
, KERN_INFO
, "mounting ext3 file system "
3486 "using the ext4 subsystem");
3488 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext3 due "
3489 "to feature incompatibilities");
3495 * Check feature flags regardless of the revision level, since we
3496 * previously didn't change the revision level when setting the flags,
3497 * so there is a chance incompat flags are set on a rev 0 filesystem.
3499 if (!ext4_feature_set_ok(sb
, (sb
->s_flags
& MS_RDONLY
)))
3502 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
3503 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
3504 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
3505 ext4_msg(sb
, KERN_ERR
,
3506 "Unsupported filesystem blocksize %d", blocksize
);
3510 if (sb
->s_blocksize
!= blocksize
) {
3511 /* Validate the filesystem blocksize */
3512 if (!sb_set_blocksize(sb
, blocksize
)) {
3513 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
3519 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3520 offset
= do_div(logical_sb_block
, blocksize
);
3521 bh
= sb_bread(sb
, logical_sb_block
);
3523 ext4_msg(sb
, KERN_ERR
,
3524 "Can't read superblock on 2nd try");
3527 es
= (struct ext4_super_block
*)(bh
->b_data
+ offset
);
3529 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
3530 ext4_msg(sb
, KERN_ERR
,
3531 "Magic mismatch, very weird!");
3536 has_huge_files
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3537 EXT4_FEATURE_RO_COMPAT_HUGE_FILE
);
3538 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
3540 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
3542 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
3543 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
3544 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
3546 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
3547 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
3548 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
3549 (!is_power_of_2(sbi
->s_inode_size
)) ||
3550 (sbi
->s_inode_size
> blocksize
)) {
3551 ext4_msg(sb
, KERN_ERR
,
3552 "unsupported inode size: %d",
3556 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
3557 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
3560 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
3561 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
)) {
3562 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
3563 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
3564 !is_power_of_2(sbi
->s_desc_size
)) {
3565 ext4_msg(sb
, KERN_ERR
,
3566 "unsupported descriptor size %lu",
3571 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
3573 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
3574 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
3575 if (EXT4_INODE_SIZE(sb
) == 0 || EXT4_INODES_PER_GROUP(sb
) == 0)
3578 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
3579 if (sbi
->s_inodes_per_block
== 0)
3581 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
3582 sbi
->s_inodes_per_block
;
3583 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
3585 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3586 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
3587 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
3589 for (i
= 0; i
< 4; i
++)
3590 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
3591 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
3592 if (EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_DIR_INDEX
)) {
3593 i
= le32_to_cpu(es
->s_flags
);
3594 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
3595 sbi
->s_hash_unsigned
= 3;
3596 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
3597 #ifdef __CHAR_UNSIGNED__
3598 if (!(sb
->s_flags
& MS_RDONLY
))
3600 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
3601 sbi
->s_hash_unsigned
= 3;
3603 if (!(sb
->s_flags
& MS_RDONLY
))
3605 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
3610 /* Handle clustersize */
3611 clustersize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_cluster_size
);
3612 has_bigalloc
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3613 EXT4_FEATURE_RO_COMPAT_BIGALLOC
);
3615 if (clustersize
< blocksize
) {
3616 ext4_msg(sb
, KERN_ERR
,
3617 "cluster size (%d) smaller than "
3618 "block size (%d)", clustersize
, blocksize
);
3621 sbi
->s_cluster_bits
= le32_to_cpu(es
->s_log_cluster_size
) -
3622 le32_to_cpu(es
->s_log_block_size
);
3623 sbi
->s_clusters_per_group
=
3624 le32_to_cpu(es
->s_clusters_per_group
);
3625 if (sbi
->s_clusters_per_group
> blocksize
* 8) {
3626 ext4_msg(sb
, KERN_ERR
,
3627 "#clusters per group too big: %lu",
3628 sbi
->s_clusters_per_group
);
3631 if (sbi
->s_blocks_per_group
!=
3632 (sbi
->s_clusters_per_group
* (clustersize
/ blocksize
))) {
3633 ext4_msg(sb
, KERN_ERR
, "blocks per group (%lu) and "
3634 "clusters per group (%lu) inconsistent",
3635 sbi
->s_blocks_per_group
,
3636 sbi
->s_clusters_per_group
);
3640 if (clustersize
!= blocksize
) {
3641 ext4_warning(sb
, "fragment/cluster size (%d) != "
3642 "block size (%d)", clustersize
,
3644 clustersize
= blocksize
;
3646 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
3647 ext4_msg(sb
, KERN_ERR
,
3648 "#blocks per group too big: %lu",
3649 sbi
->s_blocks_per_group
);
3652 sbi
->s_clusters_per_group
= sbi
->s_blocks_per_group
;
3653 sbi
->s_cluster_bits
= 0;
3655 sbi
->s_cluster_ratio
= clustersize
/ blocksize
;
3657 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
3658 ext4_msg(sb
, KERN_ERR
,
3659 "#inodes per group too big: %lu",
3660 sbi
->s_inodes_per_group
);
3664 /* Do we have standard group size of clustersize * 8 blocks ? */
3665 if (sbi
->s_blocks_per_group
== clustersize
<< 3)
3666 set_opt2(sb
, STD_GROUP_SIZE
);
3669 * Test whether we have more sectors than will fit in sector_t,
3670 * and whether the max offset is addressable by the page cache.
3672 err
= generic_check_addressable(sb
->s_blocksize_bits
,
3673 ext4_blocks_count(es
));
3675 ext4_msg(sb
, KERN_ERR
, "filesystem"
3676 " too large to mount safely on this system");
3677 if (sizeof(sector_t
) < 8)
3678 ext4_msg(sb
, KERN_WARNING
, "CONFIG_LBDAF not enabled");
3682 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
3685 /* check blocks count against device size */
3686 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
3687 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
3688 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
3689 "exceeds size of device (%llu blocks)",
3690 ext4_blocks_count(es
), blocks_count
);
3695 * It makes no sense for the first data block to be beyond the end
3696 * of the filesystem.
3698 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
3699 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data "
3700 "block %u is beyond end of filesystem (%llu)",
3701 le32_to_cpu(es
->s_first_data_block
),
3702 ext4_blocks_count(es
));
3705 blocks_count
= (ext4_blocks_count(es
) -
3706 le32_to_cpu(es
->s_first_data_block
) +
3707 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
3708 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
3709 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
3710 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %u "
3711 "(block count %llu, first data block %u, "
3712 "blocks per group %lu)", sbi
->s_groups_count
,
3713 ext4_blocks_count(es
),
3714 le32_to_cpu(es
->s_first_data_block
),
3715 EXT4_BLOCKS_PER_GROUP(sb
));
3718 sbi
->s_groups_count
= blocks_count
;
3719 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
3720 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
3721 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
3722 EXT4_DESC_PER_BLOCK(sb
);
3723 sbi
->s_group_desc
= ext4_kvmalloc(db_count
*
3724 sizeof(struct buffer_head
*),
3726 if (sbi
->s_group_desc
== NULL
) {
3727 ext4_msg(sb
, KERN_ERR
, "not enough memory");
3733 sbi
->s_proc
= proc_mkdir(sb
->s_id
, ext4_proc_root
);
3736 proc_create_data("options", S_IRUGO
, sbi
->s_proc
,
3737 &ext4_seq_options_fops
, sb
);
3739 bgl_lock_init(sbi
->s_blockgroup_lock
);
3741 for (i
= 0; i
< db_count
; i
++) {
3742 block
= descriptor_loc(sb
, logical_sb_block
, i
);
3743 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
3744 if (!sbi
->s_group_desc
[i
]) {
3745 ext4_msg(sb
, KERN_ERR
,
3746 "can't read group descriptor %d", i
);
3751 if (!ext4_check_descriptors(sb
, &first_not_zeroed
)) {
3752 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
3755 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
3756 if (!ext4_fill_flex_info(sb
)) {
3757 ext4_msg(sb
, KERN_ERR
,
3758 "unable to initialize "
3759 "flex_bg meta info!");
3763 sbi
->s_gdb_count
= db_count
;
3764 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
3765 spin_lock_init(&sbi
->s_next_gen_lock
);
3767 init_timer(&sbi
->s_err_report
);
3768 sbi
->s_err_report
.function
= print_daily_error_info
;
3769 sbi
->s_err_report
.data
= (unsigned long) sb
;
3771 /* Register extent status tree shrinker */
3772 ext4_es_register_shrinker(sb
);
3774 err
= percpu_counter_init(&sbi
->s_freeclusters_counter
,
3775 ext4_count_free_clusters(sb
));
3777 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
,
3778 ext4_count_free_inodes(sb
));
3781 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
3782 ext4_count_dirs(sb
));
3785 err
= percpu_counter_init(&sbi
->s_dirtyclusters_counter
, 0);
3788 err
= percpu_counter_init(&sbi
->s_extent_cache_cnt
, 0);
3791 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
3795 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
3796 sbi
->s_max_writeback_mb_bump
= 128;
3797 sbi
->s_extent_max_zeroout_kb
= 32;
3800 * set up enough so that it can read an inode
3802 if (!test_opt(sb
, NOLOAD
) &&
3803 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
3804 sb
->s_op
= &ext4_sops
;
3806 sb
->s_op
= &ext4_nojournal_sops
;
3807 sb
->s_export_op
= &ext4_export_ops
;
3808 sb
->s_xattr
= ext4_xattr_handlers
;
3810 sb
->dq_op
= &ext4_quota_operations
;
3811 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
))
3812 sb
->s_qcop
= &ext4_qctl_sysfile_operations
;
3814 sb
->s_qcop
= &ext4_qctl_operations
;
3816 memcpy(sb
->s_uuid
, es
->s_uuid
, sizeof(es
->s_uuid
));
3818 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
3819 mutex_init(&sbi
->s_orphan_lock
);
3823 needs_recovery
= (es
->s_last_orphan
!= 0 ||
3824 EXT4_HAS_INCOMPAT_FEATURE(sb
,
3825 EXT4_FEATURE_INCOMPAT_RECOVER
));
3827 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_MMP
) &&
3828 !(sb
->s_flags
& MS_RDONLY
))
3829 if (ext4_multi_mount_protect(sb
, le64_to_cpu(es
->s_mmp_block
)))
3833 * The first inode we look at is the journal inode. Don't try
3834 * root first: it may be modified in the journal!
3836 if (!test_opt(sb
, NOLOAD
) &&
3837 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
3838 if (ext4_load_journal(sb
, es
, journal_devnum
))
3840 } else if (test_opt(sb
, NOLOAD
) && !(sb
->s_flags
& MS_RDONLY
) &&
3841 EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
3842 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
3843 "suppressed and not mounted read-only");
3844 goto failed_mount_wq
;
3846 clear_opt(sb
, DATA_FLAGS
);
3847 sbi
->s_journal
= NULL
;
3852 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
) &&
3853 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
3854 JBD2_FEATURE_INCOMPAT_64BIT
)) {
3855 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
3856 goto failed_mount_wq
;
3859 if (!set_journal_csum_feature_set(sb
)) {
3860 ext4_msg(sb
, KERN_ERR
, "Failed to set journal checksum "
3862 goto failed_mount_wq
;
3865 /* We have now updated the journal if required, so we can
3866 * validate the data journaling mode. */
3867 switch (test_opt(sb
, DATA_FLAGS
)) {
3869 /* No mode set, assume a default based on the journal
3870 * capabilities: ORDERED_DATA if the journal can
3871 * cope, else JOURNAL_DATA
3873 if (jbd2_journal_check_available_features
3874 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
3875 set_opt(sb
, ORDERED_DATA
);
3877 set_opt(sb
, JOURNAL_DATA
);
3880 case EXT4_MOUNT_ORDERED_DATA
:
3881 case EXT4_MOUNT_WRITEBACK_DATA
:
3882 if (!jbd2_journal_check_available_features
3883 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
3884 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
3885 "requested data journaling mode");
3886 goto failed_mount_wq
;
3891 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
3893 sbi
->s_journal
->j_commit_callback
= ext4_journal_commit_callback
;
3896 * The journal may have updated the bg summary counts, so we
3897 * need to update the global counters.
3899 percpu_counter_set(&sbi
->s_freeclusters_counter
,
3900 ext4_count_free_clusters(sb
));
3901 percpu_counter_set(&sbi
->s_freeinodes_counter
,
3902 ext4_count_free_inodes(sb
));
3903 percpu_counter_set(&sbi
->s_dirs_counter
,
3904 ext4_count_dirs(sb
));
3905 percpu_counter_set(&sbi
->s_dirtyclusters_counter
, 0);
3909 * Get the # of file system overhead blocks from the
3910 * superblock if present.
3912 if (es
->s_overhead_clusters
)
3913 sbi
->s_overhead
= le32_to_cpu(es
->s_overhead_clusters
);
3915 err
= ext4_calculate_overhead(sb
);
3917 goto failed_mount_wq
;
3921 * The maximum number of concurrent works can be high and
3922 * concurrency isn't really necessary. Limit it to 1.
3924 EXT4_SB(sb
)->dio_unwritten_wq
=
3925 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM
| WQ_UNBOUND
, 1);
3926 if (!EXT4_SB(sb
)->dio_unwritten_wq
) {
3927 printk(KERN_ERR
"EXT4-fs: failed to create DIO workqueue\n");
3929 goto failed_mount_wq
;
3933 * The jbd2_journal_load will have done any necessary log recovery,
3934 * so we can safely mount the rest of the filesystem now.
3937 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
3939 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
3940 ret
= PTR_ERR(root
);
3944 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
3945 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
3949 sb
->s_root
= d_make_root(root
);
3951 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
3956 if (ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
))
3957 sb
->s_flags
|= MS_RDONLY
;
3959 /* determine the minimum size of new large inodes, if present */
3960 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
3961 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
3962 EXT4_GOOD_OLD_INODE_SIZE
;
3963 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3964 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE
)) {
3965 if (sbi
->s_want_extra_isize
<
3966 le16_to_cpu(es
->s_want_extra_isize
))
3967 sbi
->s_want_extra_isize
=
3968 le16_to_cpu(es
->s_want_extra_isize
);
3969 if (sbi
->s_want_extra_isize
<
3970 le16_to_cpu(es
->s_min_extra_isize
))
3971 sbi
->s_want_extra_isize
=
3972 le16_to_cpu(es
->s_min_extra_isize
);
3975 /* Check if enough inode space is available */
3976 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
3977 sbi
->s_inode_size
) {
3978 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
3979 EXT4_GOOD_OLD_INODE_SIZE
;
3980 ext4_msg(sb
, KERN_INFO
, "required extra inode space not"
3984 err
= ext4_reserve_clusters(sbi
, ext4_calculate_resv_clusters(sb
));
3986 ext4_msg(sb
, KERN_ERR
, "failed to reserve %llu clusters for "
3987 "reserved pool", ext4_calculate_resv_clusters(sb
));
3988 goto failed_mount4a
;
3991 err
= ext4_setup_system_zone(sb
);
3993 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
3995 goto failed_mount4a
;
3999 err
= ext4_mb_init(sb
);
4001 ext4_msg(sb
, KERN_ERR
, "failed to initialize mballoc (%d)",
4006 err
= ext4_register_li_request(sb
, first_not_zeroed
);
4010 sbi
->s_kobj
.kset
= ext4_kset
;
4011 init_completion(&sbi
->s_kobj_unregister
);
4012 err
= kobject_init_and_add(&sbi
->s_kobj
, &ext4_ktype
, NULL
,
4018 /* Enable quota usage during mount. */
4019 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
) &&
4020 !(sb
->s_flags
& MS_RDONLY
)) {
4021 err
= ext4_enable_quotas(sb
);
4025 #endif /* CONFIG_QUOTA */
4027 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
4028 ext4_orphan_cleanup(sb
, es
);
4029 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
4030 if (needs_recovery
) {
4031 ext4_msg(sb
, KERN_INFO
, "recovery complete");
4032 ext4_mark_recovery_complete(sb
, es
);
4034 if (EXT4_SB(sb
)->s_journal
) {
4035 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
4036 descr
= " journalled data mode";
4037 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
4038 descr
= " ordered data mode";
4040 descr
= " writeback data mode";
4042 descr
= "out journal";
4044 if (test_opt(sb
, DISCARD
)) {
4045 struct request_queue
*q
= bdev_get_queue(sb
->s_bdev
);
4046 if (!blk_queue_discard(q
))
4047 ext4_msg(sb
, KERN_WARNING
,
4048 "mounting with \"discard\" option, but "
4049 "the device does not support discard");
4052 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s. "
4053 "Opts: %s%s%s", descr
, sbi
->s_es
->s_mount_opts
,
4054 *sbi
->s_es
->s_mount_opts
? "; " : "", orig_data
);
4056 if (es
->s_error_count
)
4057 mod_timer(&sbi
->s_err_report
, jiffies
+ 300*HZ
); /* 5 minutes */
4064 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
4069 kobject_del(&sbi
->s_kobj
);
4072 ext4_unregister_li_request(sb
);
4074 ext4_mb_release(sb
);
4076 ext4_ext_release(sb
);
4077 ext4_release_system_zone(sb
);
4082 ext4_msg(sb
, KERN_ERR
, "mount failed");
4083 destroy_workqueue(EXT4_SB(sb
)->dio_unwritten_wq
);
4085 if (sbi
->s_journal
) {
4086 jbd2_journal_destroy(sbi
->s_journal
);
4087 sbi
->s_journal
= NULL
;
4090 ext4_es_unregister_shrinker(sb
);
4091 del_timer(&sbi
->s_err_report
);
4092 if (sbi
->s_flex_groups
)
4093 ext4_kvfree(sbi
->s_flex_groups
);
4094 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
4095 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
4096 percpu_counter_destroy(&sbi
->s_dirs_counter
);
4097 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
4098 percpu_counter_destroy(&sbi
->s_extent_cache_cnt
);
4100 kthread_stop(sbi
->s_mmp_tsk
);
4102 for (i
= 0; i
< db_count
; i
++)
4103 brelse(sbi
->s_group_desc
[i
]);
4104 ext4_kvfree(sbi
->s_group_desc
);
4106 if (sbi
->s_chksum_driver
)
4107 crypto_free_shash(sbi
->s_chksum_driver
);
4109 remove_proc_entry("options", sbi
->s_proc
);
4110 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
4113 for (i
= 0; i
< MAXQUOTAS
; i
++)
4114 kfree(sbi
->s_qf_names
[i
]);
4116 ext4_blkdev_remove(sbi
);
4119 sb
->s_fs_info
= NULL
;
4120 kfree(sbi
->s_blockgroup_lock
);
4124 return err
? err
: ret
;
4128 * Setup any per-fs journal parameters now. We'll do this both on
4129 * initial mount, once the journal has been initialised but before we've
4130 * done any recovery; and again on any subsequent remount.
4132 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
4134 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4136 journal
->j_commit_interval
= sbi
->s_commit_interval
;
4137 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
4138 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
4140 write_lock(&journal
->j_state_lock
);
4141 if (test_opt(sb
, BARRIER
))
4142 journal
->j_flags
|= JBD2_BARRIER
;
4144 journal
->j_flags
&= ~JBD2_BARRIER
;
4145 if (test_opt(sb
, DATA_ERR_ABORT
))
4146 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
4148 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
4149 write_unlock(&journal
->j_state_lock
);
4152 static journal_t
*ext4_get_journal(struct super_block
*sb
,
4153 unsigned int journal_inum
)
4155 struct inode
*journal_inode
;
4158 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4160 /* First, test for the existence of a valid inode on disk. Bad
4161 * things happen if we iget() an unused inode, as the subsequent
4162 * iput() will try to delete it. */
4164 journal_inode
= ext4_iget(sb
, journal_inum
);
4165 if (IS_ERR(journal_inode
)) {
4166 ext4_msg(sb
, KERN_ERR
, "no journal found");
4169 if (!journal_inode
->i_nlink
) {
4170 make_bad_inode(journal_inode
);
4171 iput(journal_inode
);
4172 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
4176 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4177 journal_inode
, journal_inode
->i_size
);
4178 if (!S_ISREG(journal_inode
->i_mode
)) {
4179 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
4180 iput(journal_inode
);
4184 journal
= jbd2_journal_init_inode(journal_inode
);
4186 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
4187 iput(journal_inode
);
4190 journal
->j_private
= sb
;
4191 ext4_init_journal_params(sb
, journal
);
4195 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
4198 struct buffer_head
*bh
;
4202 int hblock
, blocksize
;
4203 ext4_fsblk_t sb_block
;
4204 unsigned long offset
;
4205 struct ext4_super_block
*es
;
4206 struct block_device
*bdev
;
4208 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4210 bdev
= ext4_blkdev_get(j_dev
, sb
);
4214 blocksize
= sb
->s_blocksize
;
4215 hblock
= bdev_logical_block_size(bdev
);
4216 if (blocksize
< hblock
) {
4217 ext4_msg(sb
, KERN_ERR
,
4218 "blocksize too small for journal device");
4222 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
4223 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
4224 set_blocksize(bdev
, blocksize
);
4225 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
4226 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
4227 "external journal");
4231 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
4232 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
4233 !(le32_to_cpu(es
->s_feature_incompat
) &
4234 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
4235 ext4_msg(sb
, KERN_ERR
, "external journal has "
4241 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
4242 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
4247 len
= ext4_blocks_count(es
);
4248 start
= sb_block
+ 1;
4249 brelse(bh
); /* we're done with the superblock */
4251 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
4252 start
, len
, blocksize
);
4254 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
4257 journal
->j_private
= sb
;
4258 ll_rw_block(READ
| REQ_META
| REQ_PRIO
, 1, &journal
->j_sb_buffer
);
4259 wait_on_buffer(journal
->j_sb_buffer
);
4260 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
4261 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
4264 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
4265 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
4266 "user (unsupported) - %d",
4267 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
4270 EXT4_SB(sb
)->journal_bdev
= bdev
;
4271 ext4_init_journal_params(sb
, journal
);
4275 jbd2_journal_destroy(journal
);
4277 ext4_blkdev_put(bdev
);
4281 static int ext4_load_journal(struct super_block
*sb
,
4282 struct ext4_super_block
*es
,
4283 unsigned long journal_devnum
)
4286 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
4289 int really_read_only
;
4291 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4293 if (journal_devnum
&&
4294 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4295 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
4296 "numbers have changed");
4297 journal_dev
= new_decode_dev(journal_devnum
);
4299 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
4301 really_read_only
= bdev_read_only(sb
->s_bdev
);
4304 * Are we loading a blank journal or performing recovery after a
4305 * crash? For recovery, we need to check in advance whether we
4306 * can get read-write access to the device.
4308 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
4309 if (sb
->s_flags
& MS_RDONLY
) {
4310 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
4311 "required on readonly filesystem");
4312 if (really_read_only
) {
4313 ext4_msg(sb
, KERN_ERR
, "write access "
4314 "unavailable, cannot proceed");
4317 ext4_msg(sb
, KERN_INFO
, "write access will "
4318 "be enabled during recovery");
4322 if (journal_inum
&& journal_dev
) {
4323 ext4_msg(sb
, KERN_ERR
, "filesystem has both journal "
4324 "and inode journals!");
4329 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
4332 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
4336 if (!(journal
->j_flags
& JBD2_BARRIER
))
4337 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
4339 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
))
4340 err
= jbd2_journal_wipe(journal
, !really_read_only
);
4342 char *save
= kmalloc(EXT4_S_ERR_LEN
, GFP_KERNEL
);
4344 memcpy(save
, ((char *) es
) +
4345 EXT4_S_ERR_START
, EXT4_S_ERR_LEN
);
4346 err
= jbd2_journal_load(journal
);
4348 memcpy(((char *) es
) + EXT4_S_ERR_START
,
4349 save
, EXT4_S_ERR_LEN
);
4354 ext4_msg(sb
, KERN_ERR
, "error loading journal");
4355 jbd2_journal_destroy(journal
);
4359 EXT4_SB(sb
)->s_journal
= journal
;
4360 ext4_clear_journal_err(sb
, es
);
4362 if (!really_read_only
&& journal_devnum
&&
4363 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4364 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
4366 /* Make sure we flush the recovery flag to disk. */
4367 ext4_commit_super(sb
, 1);
4373 static int ext4_commit_super(struct super_block
*sb
, int sync
)
4375 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
4376 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
4379 if (!sbh
|| block_device_ejected(sb
))
4381 if (buffer_write_io_error(sbh
)) {
4383 * Oh, dear. A previous attempt to write the
4384 * superblock failed. This could happen because the
4385 * USB device was yanked out. Or it could happen to
4386 * be a transient write error and maybe the block will
4387 * be remapped. Nothing we can do but to retry the
4388 * write and hope for the best.
4390 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
4391 "superblock detected");
4392 clear_buffer_write_io_error(sbh
);
4393 set_buffer_uptodate(sbh
);
4396 * If the file system is mounted read-only, don't update the
4397 * superblock write time. This avoids updating the superblock
4398 * write time when we are mounting the root file system
4399 * read/only but we need to replay the journal; at that point,
4400 * for people who are east of GMT and who make their clock
4401 * tick in localtime for Windows bug-for-bug compatibility,
4402 * the clock is set in the future, and this will cause e2fsck
4403 * to complain and force a full file system check.
4405 if (!(sb
->s_flags
& MS_RDONLY
))
4406 es
->s_wtime
= cpu_to_le32(get_seconds());
4407 if (sb
->s_bdev
->bd_part
)
4408 es
->s_kbytes_written
=
4409 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
+
4410 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
4411 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
4413 es
->s_kbytes_written
=
4414 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
);
4415 ext4_free_blocks_count_set(es
,
4416 EXT4_C2B(EXT4_SB(sb
), percpu_counter_sum_positive(
4417 &EXT4_SB(sb
)->s_freeclusters_counter
)));
4418 es
->s_free_inodes_count
=
4419 cpu_to_le32(percpu_counter_sum_positive(
4420 &EXT4_SB(sb
)->s_freeinodes_counter
));
4421 BUFFER_TRACE(sbh
, "marking dirty");
4422 ext4_superblock_csum_set(sb
);
4423 mark_buffer_dirty(sbh
);
4425 error
= sync_dirty_buffer(sbh
);
4429 error
= buffer_write_io_error(sbh
);
4431 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
4433 clear_buffer_write_io_error(sbh
);
4434 set_buffer_uptodate(sbh
);
4441 * Have we just finished recovery? If so, and if we are mounting (or
4442 * remounting) the filesystem readonly, then we will end up with a
4443 * consistent fs on disk. Record that fact.
4445 static void ext4_mark_recovery_complete(struct super_block
*sb
,
4446 struct ext4_super_block
*es
)
4448 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
4450 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
4451 BUG_ON(journal
!= NULL
);
4454 jbd2_journal_lock_updates(journal
);
4455 if (jbd2_journal_flush(journal
) < 0)
4458 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
) &&
4459 sb
->s_flags
& MS_RDONLY
) {
4460 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4461 ext4_commit_super(sb
, 1);
4465 jbd2_journal_unlock_updates(journal
);
4469 * If we are mounting (or read-write remounting) a filesystem whose journal
4470 * has recorded an error from a previous lifetime, move that error to the
4471 * main filesystem now.
4473 static void ext4_clear_journal_err(struct super_block
*sb
,
4474 struct ext4_super_block
*es
)
4480 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4482 journal
= EXT4_SB(sb
)->s_journal
;
4485 * Now check for any error status which may have been recorded in the
4486 * journal by a prior ext4_error() or ext4_abort()
4489 j_errno
= jbd2_journal_errno(journal
);
4493 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
4494 ext4_warning(sb
, "Filesystem error recorded "
4495 "from previous mount: %s", errstr
);
4496 ext4_warning(sb
, "Marking fs in need of filesystem check.");
4498 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
4499 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
4500 ext4_commit_super(sb
, 1);
4502 jbd2_journal_clear_err(journal
);
4503 jbd2_journal_update_sb_errno(journal
);
4508 * Force the running and committing transactions to commit,
4509 * and wait on the commit.
4511 int ext4_force_commit(struct super_block
*sb
)
4515 if (sb
->s_flags
& MS_RDONLY
)
4518 journal
= EXT4_SB(sb
)->s_journal
;
4519 return ext4_journal_force_commit(journal
);
4522 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
4526 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4528 trace_ext4_sync_fs(sb
, wait
);
4529 flush_workqueue(sbi
->dio_unwritten_wq
);
4531 * Writeback quota in non-journalled quota case - journalled quota has
4534 dquot_writeback_dquots(sb
, -1);
4535 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
4537 jbd2_log_wait_commit(sbi
->s_journal
, target
);
4543 * LVM calls this function before a (read-only) snapshot is created. This
4544 * gives us a chance to flush the journal completely and mark the fs clean.
4546 * Note that only this function cannot bring a filesystem to be in a clean
4547 * state independently. It relies on upper layer to stop all data & metadata
4550 static int ext4_freeze(struct super_block
*sb
)
4555 if (sb
->s_flags
& MS_RDONLY
)
4558 journal
= EXT4_SB(sb
)->s_journal
;
4560 /* Now we set up the journal barrier. */
4561 jbd2_journal_lock_updates(journal
);
4564 * Don't clear the needs_recovery flag if we failed to flush
4567 error
= jbd2_journal_flush(journal
);
4571 /* Journal blocked and flushed, clear needs_recovery flag. */
4572 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4573 error
= ext4_commit_super(sb
, 1);
4575 /* we rely on upper layer to stop further updates */
4576 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
4581 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4582 * flag here, even though the filesystem is not technically dirty yet.
4584 static int ext4_unfreeze(struct super_block
*sb
)
4586 if (sb
->s_flags
& MS_RDONLY
)
4589 /* Reset the needs_recovery flag before the fs is unlocked. */
4590 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4591 ext4_commit_super(sb
, 1);
4596 * Structure to save mount options for ext4_remount's benefit
4598 struct ext4_mount_options
{
4599 unsigned long s_mount_opt
;
4600 unsigned long s_mount_opt2
;
4603 unsigned long s_commit_interval
;
4604 u32 s_min_batch_time
, s_max_batch_time
;
4607 char *s_qf_names
[MAXQUOTAS
];
4611 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
4613 struct ext4_super_block
*es
;
4614 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4615 unsigned long old_sb_flags
;
4616 struct ext4_mount_options old_opts
;
4617 int enable_quota
= 0;
4619 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
4624 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
4626 /* Store the original options */
4627 old_sb_flags
= sb
->s_flags
;
4628 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
4629 old_opts
.s_mount_opt2
= sbi
->s_mount_opt2
;
4630 old_opts
.s_resuid
= sbi
->s_resuid
;
4631 old_opts
.s_resgid
= sbi
->s_resgid
;
4632 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
4633 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
4634 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
4636 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
4637 for (i
= 0; i
< MAXQUOTAS
; i
++)
4638 if (sbi
->s_qf_names
[i
]) {
4639 old_opts
.s_qf_names
[i
] = kstrdup(sbi
->s_qf_names
[i
],
4641 if (!old_opts
.s_qf_names
[i
]) {
4642 for (j
= 0; j
< i
; j
++)
4643 kfree(old_opts
.s_qf_names
[j
]);
4648 old_opts
.s_qf_names
[i
] = NULL
;
4650 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
4651 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
4654 * Allow the "check" option to be passed as a remount option.
4656 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
, 1)) {
4661 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
4662 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
4663 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4664 "both data=journal and delalloc");
4668 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
4669 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4670 "both data=journal and dioread_nolock");
4676 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
)
4677 ext4_abort(sb
, "Abort forced by user");
4679 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
4680 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
4684 if (sbi
->s_journal
) {
4685 ext4_init_journal_params(sb
, sbi
->s_journal
);
4686 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
4689 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
)) {
4690 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
) {
4695 if (*flags
& MS_RDONLY
) {
4696 err
= dquot_suspend(sb
, -1);
4701 * First of all, the unconditional stuff we have to do
4702 * to disable replay of the journal when we next remount
4704 sb
->s_flags
|= MS_RDONLY
;
4707 * OK, test if we are remounting a valid rw partition
4708 * readonly, and if so set the rdonly flag and then
4709 * mark the partition as valid again.
4711 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
4712 (sbi
->s_mount_state
& EXT4_VALID_FS
))
4713 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
4716 ext4_mark_recovery_complete(sb
, es
);
4718 /* Make sure we can mount this feature set readwrite */
4719 if (!ext4_feature_set_ok(sb
, 0)) {
4724 * Make sure the group descriptor checksums
4725 * are sane. If they aren't, refuse to remount r/w.
4727 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
4728 struct ext4_group_desc
*gdp
=
4729 ext4_get_group_desc(sb
, g
, NULL
);
4731 if (!ext4_group_desc_csum_verify(sb
, g
, gdp
)) {
4732 ext4_msg(sb
, KERN_ERR
,
4733 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4734 g
, le16_to_cpu(ext4_group_desc_csum(sbi
, g
, gdp
)),
4735 le16_to_cpu(gdp
->bg_checksum
));
4742 * If we have an unprocessed orphan list hanging
4743 * around from a previously readonly bdev mount,
4744 * require a full umount/remount for now.
4746 if (es
->s_last_orphan
) {
4747 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
4748 "remount RDWR because of unprocessed "
4749 "orphan inode list. Please "
4750 "umount/remount instead");
4756 * Mounting a RDONLY partition read-write, so reread
4757 * and store the current valid flag. (It may have
4758 * been changed by e2fsck since we originally mounted
4762 ext4_clear_journal_err(sb
, es
);
4763 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
4764 if (!ext4_setup_super(sb
, es
, 0))
4765 sb
->s_flags
&= ~MS_RDONLY
;
4766 if (EXT4_HAS_INCOMPAT_FEATURE(sb
,
4767 EXT4_FEATURE_INCOMPAT_MMP
))
4768 if (ext4_multi_mount_protect(sb
,
4769 le64_to_cpu(es
->s_mmp_block
))) {
4778 * Reinitialize lazy itable initialization thread based on
4781 if ((sb
->s_flags
& MS_RDONLY
) || !test_opt(sb
, INIT_INODE_TABLE
))
4782 ext4_unregister_li_request(sb
);
4784 ext4_group_t first_not_zeroed
;
4785 first_not_zeroed
= ext4_has_uninit_itable(sb
);
4786 ext4_register_li_request(sb
, first_not_zeroed
);
4789 ext4_setup_system_zone(sb
);
4790 if (sbi
->s_journal
== NULL
&& !(old_sb_flags
& MS_RDONLY
))
4791 ext4_commit_super(sb
, 1);
4794 /* Release old quota file names */
4795 for (i
= 0; i
< MAXQUOTAS
; i
++)
4796 kfree(old_opts
.s_qf_names
[i
]);
4798 if (sb_any_quota_suspended(sb
))
4799 dquot_resume(sb
, -1);
4800 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
4801 EXT4_FEATURE_RO_COMPAT_QUOTA
)) {
4802 err
= ext4_enable_quotas(sb
);
4809 ext4_msg(sb
, KERN_INFO
, "re-mounted. Opts: %s", orig_data
);
4814 sb
->s_flags
= old_sb_flags
;
4815 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
4816 sbi
->s_mount_opt2
= old_opts
.s_mount_opt2
;
4817 sbi
->s_resuid
= old_opts
.s_resuid
;
4818 sbi
->s_resgid
= old_opts
.s_resgid
;
4819 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
4820 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
4821 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
4823 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
4824 for (i
= 0; i
< MAXQUOTAS
; i
++) {
4825 kfree(sbi
->s_qf_names
[i
]);
4826 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
4833 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
4835 struct super_block
*sb
= dentry
->d_sb
;
4836 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4837 struct ext4_super_block
*es
= sbi
->s_es
;
4838 ext4_fsblk_t overhead
= 0, resv_blocks
;
4841 resv_blocks
= EXT4_C2B(sbi
, atomic64_read(&sbi
->s_resv_clusters
));
4843 if (!test_opt(sb
, MINIX_DF
))
4844 overhead
= sbi
->s_overhead
;
4846 buf
->f_type
= EXT4_SUPER_MAGIC
;
4847 buf
->f_bsize
= sb
->s_blocksize
;
4848 buf
->f_blocks
= ext4_blocks_count(es
) - EXT4_C2B(sbi
, overhead
);
4849 bfree
= percpu_counter_sum_positive(&sbi
->s_freeclusters_counter
) -
4850 percpu_counter_sum_positive(&sbi
->s_dirtyclusters_counter
);
4851 /* prevent underflow in case that few free space is available */
4852 buf
->f_bfree
= EXT4_C2B(sbi
, max_t(s64
, bfree
, 0));
4853 buf
->f_bavail
= buf
->f_bfree
-
4854 (ext4_r_blocks_count(es
) + resv_blocks
);
4855 if (buf
->f_bfree
< (ext4_r_blocks_count(es
) + resv_blocks
))
4857 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
4858 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
4859 buf
->f_namelen
= EXT4_NAME_LEN
;
4860 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
4861 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
4862 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
4863 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
4868 /* Helper function for writing quotas on sync - we need to start transaction
4869 * before quota file is locked for write. Otherwise the are possible deadlocks:
4870 * Process 1 Process 2
4871 * ext4_create() quota_sync()
4872 * jbd2_journal_start() write_dquot()
4873 * dquot_initialize() down(dqio_mutex)
4874 * down(dqio_mutex) jbd2_journal_start()
4880 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
4882 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_id
.type
];
4885 static int ext4_write_dquot(struct dquot
*dquot
)
4889 struct inode
*inode
;
4891 inode
= dquot_to_inode(dquot
);
4892 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
,
4893 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
4895 return PTR_ERR(handle
);
4896 ret
= dquot_commit(dquot
);
4897 err
= ext4_journal_stop(handle
);
4903 static int ext4_acquire_dquot(struct dquot
*dquot
)
4908 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
4909 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
4911 return PTR_ERR(handle
);
4912 ret
= dquot_acquire(dquot
);
4913 err
= ext4_journal_stop(handle
);
4919 static int ext4_release_dquot(struct dquot
*dquot
)
4924 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
4925 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
4926 if (IS_ERR(handle
)) {
4927 /* Release dquot anyway to avoid endless cycle in dqput() */
4928 dquot_release(dquot
);
4929 return PTR_ERR(handle
);
4931 ret
= dquot_release(dquot
);
4932 err
= ext4_journal_stop(handle
);
4938 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
4940 struct super_block
*sb
= dquot
->dq_sb
;
4941 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4943 /* Are we journaling quotas? */
4944 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
) ||
4945 sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
4946 dquot_mark_dquot_dirty(dquot
);
4947 return ext4_write_dquot(dquot
);
4949 return dquot_mark_dquot_dirty(dquot
);
4953 static int ext4_write_info(struct super_block
*sb
, int type
)
4958 /* Data block + inode block */
4959 handle
= ext4_journal_start(sb
->s_root
->d_inode
, EXT4_HT_QUOTA
, 2);
4961 return PTR_ERR(handle
);
4962 ret
= dquot_commit_info(sb
, type
);
4963 err
= ext4_journal_stop(handle
);
4970 * Turn on quotas during mount time - we need to find
4971 * the quota file and such...
4973 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
4975 return dquot_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
4976 EXT4_SB(sb
)->s_jquota_fmt
, type
);
4980 * Standard function to be called on quota_on
4982 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
4987 if (!test_opt(sb
, QUOTA
))
4990 /* Quotafile not on the same filesystem? */
4991 if (path
->dentry
->d_sb
!= sb
)
4993 /* Journaling quota? */
4994 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
4995 /* Quotafile not in fs root? */
4996 if (path
->dentry
->d_parent
!= sb
->s_root
)
4997 ext4_msg(sb
, KERN_WARNING
,
4998 "Quota file not on filesystem root. "
4999 "Journaled quota will not work");
5003 * When we journal data on quota file, we have to flush journal to see
5004 * all updates to the file when we bypass pagecache...
5006 if (EXT4_SB(sb
)->s_journal
&&
5007 ext4_should_journal_data(path
->dentry
->d_inode
)) {
5009 * We don't need to lock updates but journal_flush() could
5010 * otherwise be livelocked...
5012 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
5013 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
5014 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
5019 return dquot_quota_on(sb
, type
, format_id
, path
);
5022 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
5026 struct inode
*qf_inode
;
5027 unsigned long qf_inums
[MAXQUOTAS
] = {
5028 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
5029 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
)
5032 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
));
5034 if (!qf_inums
[type
])
5037 qf_inode
= ext4_iget(sb
, qf_inums
[type
]);
5038 if (IS_ERR(qf_inode
)) {
5039 ext4_error(sb
, "Bad quota inode # %lu", qf_inums
[type
]);
5040 return PTR_ERR(qf_inode
);
5043 /* Don't account quota for quota files to avoid recursion */
5044 qf_inode
->i_flags
|= S_NOQUOTA
;
5045 err
= dquot_enable(qf_inode
, type
, format_id
, flags
);
5051 /* Enable usage tracking for all quota types. */
5052 static int ext4_enable_quotas(struct super_block
*sb
)
5055 unsigned long qf_inums
[MAXQUOTAS
] = {
5056 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
5057 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
)
5060 sb_dqopt(sb
)->flags
|= DQUOT_QUOTA_SYS_FILE
;
5061 for (type
= 0; type
< MAXQUOTAS
; type
++) {
5062 if (qf_inums
[type
]) {
5063 err
= ext4_quota_enable(sb
, type
, QFMT_VFS_V1
,
5064 DQUOT_USAGE_ENABLED
);
5067 "Failed to enable quota tracking "
5068 "(type=%d, err=%d). Please run "
5069 "e2fsck to fix.", type
, err
);
5078 * quota_on function that is used when QUOTA feature is set.
5080 static int ext4_quota_on_sysfile(struct super_block
*sb
, int type
,
5083 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
))
5087 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5089 return ext4_quota_enable(sb
, type
, format_id
, DQUOT_LIMITS_ENABLED
);
5092 static int ext4_quota_off(struct super_block
*sb
, int type
)
5094 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5097 /* Force all delayed allocation blocks to be allocated.
5098 * Caller already holds s_umount sem */
5099 if (test_opt(sb
, DELALLOC
))
5100 sync_filesystem(sb
);
5105 /* Update modification times of quota files when userspace can
5106 * start looking at them */
5107 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
, 1);
5110 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
5111 ext4_mark_inode_dirty(handle
, inode
);
5112 ext4_journal_stop(handle
);
5115 return dquot_quota_off(sb
, type
);
5119 * quota_off function that is used when QUOTA feature is set.
5121 static int ext4_quota_off_sysfile(struct super_block
*sb
, int type
)
5123 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_QUOTA
))
5126 /* Disable only the limits. */
5127 return dquot_disable(sb
, type
, DQUOT_LIMITS_ENABLED
);
5130 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5131 * acquiring the locks... As quota files are never truncated and quota code
5132 * itself serializes the operations (and no one else should touch the files)
5133 * we don't have to be afraid of races */
5134 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
5135 size_t len
, loff_t off
)
5137 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5138 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5140 int offset
= off
& (sb
->s_blocksize
- 1);
5143 struct buffer_head
*bh
;
5144 loff_t i_size
= i_size_read(inode
);
5148 if (off
+len
> i_size
)
5151 while (toread
> 0) {
5152 tocopy
= sb
->s_blocksize
- offset
< toread
?
5153 sb
->s_blocksize
- offset
: toread
;
5154 bh
= ext4_bread(NULL
, inode
, blk
, 0, &err
);
5157 if (!bh
) /* A hole? */
5158 memset(data
, 0, tocopy
);
5160 memcpy(data
, bh
->b_data
+offset
, tocopy
);
5170 /* Write to quotafile (we know the transaction is already started and has
5171 * enough credits) */
5172 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
5173 const char *data
, size_t len
, loff_t off
)
5175 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5176 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5178 int offset
= off
& (sb
->s_blocksize
- 1);
5179 struct buffer_head
*bh
;
5180 handle_t
*handle
= journal_current_handle();
5182 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
5183 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5184 " cancelled because transaction is not started",
5185 (unsigned long long)off
, (unsigned long long)len
);
5189 * Since we account only one data block in transaction credits,
5190 * then it is impossible to cross a block boundary.
5192 if (sb
->s_blocksize
- offset
< len
) {
5193 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5194 " cancelled because not block aligned",
5195 (unsigned long long)off
, (unsigned long long)len
);
5199 bh
= ext4_bread(handle
, inode
, blk
, 1, &err
);
5202 err
= ext4_journal_get_write_access(handle
, bh
);
5208 memcpy(bh
->b_data
+offset
, data
, len
);
5209 flush_dcache_page(bh
->b_page
);
5211 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
5216 if (inode
->i_size
< off
+ len
) {
5217 i_size_write(inode
, off
+ len
);
5218 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
5219 ext4_mark_inode_dirty(handle
, inode
);
5226 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
5227 const char *dev_name
, void *data
)
5229 return mount_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
);
5232 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5233 static inline void register_as_ext2(void)
5235 int err
= register_filesystem(&ext2_fs_type
);
5238 "EXT4-fs: Unable to register as ext2 (%d)\n", err
);
5241 static inline void unregister_as_ext2(void)
5243 unregister_filesystem(&ext2_fs_type
);
5246 static inline int ext2_feature_set_ok(struct super_block
*sb
)
5248 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT2_FEATURE_INCOMPAT_SUPP
))
5250 if (sb
->s_flags
& MS_RDONLY
)
5252 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT2_FEATURE_RO_COMPAT_SUPP
))
5257 static inline void register_as_ext2(void) { }
5258 static inline void unregister_as_ext2(void) { }
5259 static inline int ext2_feature_set_ok(struct super_block
*sb
) { return 0; }
5262 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5263 static inline void register_as_ext3(void)
5265 int err
= register_filesystem(&ext3_fs_type
);
5268 "EXT4-fs: Unable to register as ext3 (%d)\n", err
);
5271 static inline void unregister_as_ext3(void)
5273 unregister_filesystem(&ext3_fs_type
);
5276 static inline int ext3_feature_set_ok(struct super_block
*sb
)
5278 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT3_FEATURE_INCOMPAT_SUPP
))
5280 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
5282 if (sb
->s_flags
& MS_RDONLY
)
5284 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT3_FEATURE_RO_COMPAT_SUPP
))
5289 static inline void register_as_ext3(void) { }
5290 static inline void unregister_as_ext3(void) { }
5291 static inline int ext3_feature_set_ok(struct super_block
*sb
) { return 0; }
5294 static struct file_system_type ext4_fs_type
= {
5295 .owner
= THIS_MODULE
,
5297 .mount
= ext4_mount
,
5298 .kill_sb
= kill_block_super
,
5299 .fs_flags
= FS_REQUIRES_DEV
,
5301 MODULE_ALIAS_FS("ext4");
5303 static int __init
ext4_init_feat_adverts(void)
5305 struct ext4_features
*ef
;
5308 ef
= kzalloc(sizeof(struct ext4_features
), GFP_KERNEL
);
5312 ef
->f_kobj
.kset
= ext4_kset
;
5313 init_completion(&ef
->f_kobj_unregister
);
5314 ret
= kobject_init_and_add(&ef
->f_kobj
, &ext4_feat_ktype
, NULL
,
5327 static void ext4_exit_feat_adverts(void)
5329 kobject_put(&ext4_feat
->f_kobj
);
5330 wait_for_completion(&ext4_feat
->f_kobj_unregister
);
5334 /* Shared across all ext4 file systems */
5335 wait_queue_head_t ext4__ioend_wq
[EXT4_WQ_HASH_SZ
];
5336 struct mutex ext4__aio_mutex
[EXT4_WQ_HASH_SZ
];
5338 static int __init
ext4_init_fs(void)
5342 ext4_li_info
= NULL
;
5343 mutex_init(&ext4_li_mtx
);
5345 /* Build-time check for flags consistency */
5346 ext4_check_flag_values();
5348 for (i
= 0; i
< EXT4_WQ_HASH_SZ
; i
++) {
5349 mutex_init(&ext4__aio_mutex
[i
]);
5350 init_waitqueue_head(&ext4__ioend_wq
[i
]);
5353 err
= ext4_init_es();
5357 err
= ext4_init_pageio();
5361 err
= ext4_init_system_zone();
5364 ext4_kset
= kset_create_and_add("ext4", NULL
, fs_kobj
);
5369 ext4_proc_root
= proc_mkdir("fs/ext4", NULL
);
5371 err
= ext4_init_feat_adverts();
5375 err
= ext4_init_mballoc();
5379 err
= ext4_init_xattr();
5382 err
= init_inodecache();
5387 err
= register_filesystem(&ext4_fs_type
);
5393 unregister_as_ext2();
5394 unregister_as_ext3();
5395 destroy_inodecache();
5399 ext4_exit_mballoc();
5401 ext4_exit_feat_adverts();
5404 remove_proc_entry("fs/ext4", NULL
);
5405 kset_unregister(ext4_kset
);
5407 ext4_exit_system_zone();
5416 static void __exit
ext4_exit_fs(void)
5418 ext4_destroy_lazyinit_thread();
5419 unregister_as_ext2();
5420 unregister_as_ext3();
5421 unregister_filesystem(&ext4_fs_type
);
5422 destroy_inodecache();
5424 ext4_exit_mballoc();
5425 ext4_exit_feat_adverts();
5426 remove_proc_entry("fs/ext4", NULL
);
5427 kset_unregister(ext4_kset
);
5428 ext4_exit_system_zone();
5433 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5434 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5435 MODULE_LICENSE("GPL");
5436 module_init(ext4_init_fs
)
5437 module_exit(ext4_exit_fs
)