1 // SPDX-License-Identifier: GPL-2.0
3 * linux/fs/ext4/super.c
5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI)
12 * linux/fs/minix/inode.c
14 * Copyright (C) 1991, 1992 Linus Torvalds
16 * Big-endian to little-endian byte-swapping/bitmaps by
17 * David S. Miller (davem@caip.rutgers.edu), 1995
20 #include <linux/module.h>
21 #include <linux/string.h>
23 #include <linux/time.h>
24 #include <linux/vmalloc.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/backing-dev.h>
29 #include <linux/parser.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/dax.h>
42 #include <linux/cleancache.h>
43 #include <linux/uaccess.h>
44 #include <linux/iversion.h>
46 #include <linux/kthread.h>
47 #include <linux/freezer.h>
50 #include "ext4_extents.h" /* Needed for trace points definition */
51 #include "ext4_jbd2.h"
57 #define CREATE_TRACE_POINTS
58 #include <trace/events/ext4.h>
60 static struct ext4_lazy_init
*ext4_li_info
;
61 static struct mutex ext4_li_mtx
;
62 static struct ratelimit_state ext4_mount_msg_ratelimit
;
64 static int ext4_load_journal(struct super_block
*, struct ext4_super_block
*,
65 unsigned long journal_devnum
);
66 static int ext4_show_options(struct seq_file
*seq
, struct dentry
*root
);
67 static int ext4_commit_super(struct super_block
*sb
, int sync
);
68 static void ext4_mark_recovery_complete(struct super_block
*sb
,
69 struct ext4_super_block
*es
);
70 static void ext4_clear_journal_err(struct super_block
*sb
,
71 struct ext4_super_block
*es
);
72 static int ext4_sync_fs(struct super_block
*sb
, int wait
);
73 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
);
74 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
75 static int ext4_unfreeze(struct super_block
*sb
);
76 static int ext4_freeze(struct super_block
*sb
);
77 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
78 const char *dev_name
, void *data
);
79 static inline int ext2_feature_set_ok(struct super_block
*sb
);
80 static inline int ext3_feature_set_ok(struct super_block
*sb
);
81 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
);
82 static void ext4_destroy_lazyinit_thread(void);
83 static void ext4_unregister_li_request(struct super_block
*sb
);
84 static void ext4_clear_request_list(void);
85 static struct inode
*ext4_get_journal_inode(struct super_block
*sb
,
86 unsigned int journal_inum
);
91 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
92 * i_mmap_rwsem (inode->i_mmap_rwsem)!
95 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
96 * page lock -> i_data_sem (rw)
98 * buffered write path:
99 * sb_start_write -> i_mutex -> mmap_sem
100 * sb_start_write -> i_mutex -> transaction start -> page lock ->
104 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock
105 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start ->
109 * sb_start_write -> i_mutex -> mmap_sem
110 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
113 * transaction start -> page lock(s) -> i_data_sem (rw)
116 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
117 static struct file_system_type ext2_fs_type
= {
118 .owner
= THIS_MODULE
,
121 .kill_sb
= kill_block_super
,
122 .fs_flags
= FS_REQUIRES_DEV
,
124 MODULE_ALIAS_FS("ext2");
125 MODULE_ALIAS("ext2");
126 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
128 #define IS_EXT2_SB(sb) (0)
132 static struct file_system_type ext3_fs_type
= {
133 .owner
= THIS_MODULE
,
136 .kill_sb
= kill_block_super
,
137 .fs_flags
= FS_REQUIRES_DEV
,
139 MODULE_ALIAS_FS("ext3");
140 MODULE_ALIAS("ext3");
141 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
144 * This works like sb_bread() except it uses ERR_PTR for error
145 * returns. Currently with sb_bread it's impossible to distinguish
146 * between ENOMEM and EIO situations (since both result in a NULL
150 ext4_sb_bread(struct super_block
*sb
, sector_t block
, int op_flags
)
152 struct buffer_head
*bh
= sb_getblk(sb
, block
);
155 return ERR_PTR(-ENOMEM
);
156 if (buffer_uptodate(bh
))
158 ll_rw_block(REQ_OP_READ
, REQ_META
| op_flags
, 1, &bh
);
160 if (buffer_uptodate(bh
))
163 return ERR_PTR(-EIO
);
166 static int ext4_verify_csum_type(struct super_block
*sb
,
167 struct ext4_super_block
*es
)
169 if (!ext4_has_feature_metadata_csum(sb
))
172 return es
->s_checksum_type
== EXT4_CRC32C_CHKSUM
;
175 static __le32
ext4_superblock_csum(struct super_block
*sb
,
176 struct ext4_super_block
*es
)
178 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
179 int offset
= offsetof(struct ext4_super_block
, s_checksum
);
182 csum
= ext4_chksum(sbi
, ~0, (char *)es
, offset
);
184 return cpu_to_le32(csum
);
187 static int ext4_superblock_csum_verify(struct super_block
*sb
,
188 struct ext4_super_block
*es
)
190 if (!ext4_has_metadata_csum(sb
))
193 return es
->s_checksum
== ext4_superblock_csum(sb
, es
);
196 void ext4_superblock_csum_set(struct super_block
*sb
)
198 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
200 if (!ext4_has_metadata_csum(sb
))
203 es
->s_checksum
= ext4_superblock_csum(sb
, es
);
206 void *ext4_kvmalloc(size_t size
, gfp_t flags
)
210 ret
= kmalloc(size
, flags
| __GFP_NOWARN
);
212 ret
= __vmalloc(size
, flags
, PAGE_KERNEL
);
216 void *ext4_kvzalloc(size_t size
, gfp_t flags
)
220 ret
= kzalloc(size
, flags
| __GFP_NOWARN
);
222 ret
= __vmalloc(size
, flags
| __GFP_ZERO
, PAGE_KERNEL
);
226 ext4_fsblk_t
ext4_block_bitmap(struct super_block
*sb
,
227 struct ext4_group_desc
*bg
)
229 return le32_to_cpu(bg
->bg_block_bitmap_lo
) |
230 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
231 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_block_bitmap_hi
) << 32 : 0);
234 ext4_fsblk_t
ext4_inode_bitmap(struct super_block
*sb
,
235 struct ext4_group_desc
*bg
)
237 return le32_to_cpu(bg
->bg_inode_bitmap_lo
) |
238 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
239 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_bitmap_hi
) << 32 : 0);
242 ext4_fsblk_t
ext4_inode_table(struct super_block
*sb
,
243 struct ext4_group_desc
*bg
)
245 return le32_to_cpu(bg
->bg_inode_table_lo
) |
246 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
247 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_table_hi
) << 32 : 0);
250 __u32
ext4_free_group_clusters(struct super_block
*sb
,
251 struct ext4_group_desc
*bg
)
253 return le16_to_cpu(bg
->bg_free_blocks_count_lo
) |
254 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
255 (__u32
)le16_to_cpu(bg
->bg_free_blocks_count_hi
) << 16 : 0);
258 __u32
ext4_free_inodes_count(struct super_block
*sb
,
259 struct ext4_group_desc
*bg
)
261 return le16_to_cpu(bg
->bg_free_inodes_count_lo
) |
262 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
263 (__u32
)le16_to_cpu(bg
->bg_free_inodes_count_hi
) << 16 : 0);
266 __u32
ext4_used_dirs_count(struct super_block
*sb
,
267 struct ext4_group_desc
*bg
)
269 return le16_to_cpu(bg
->bg_used_dirs_count_lo
) |
270 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
271 (__u32
)le16_to_cpu(bg
->bg_used_dirs_count_hi
) << 16 : 0);
274 __u32
ext4_itable_unused_count(struct super_block
*sb
,
275 struct ext4_group_desc
*bg
)
277 return le16_to_cpu(bg
->bg_itable_unused_lo
) |
278 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
279 (__u32
)le16_to_cpu(bg
->bg_itable_unused_hi
) << 16 : 0);
282 void ext4_block_bitmap_set(struct super_block
*sb
,
283 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
285 bg
->bg_block_bitmap_lo
= cpu_to_le32((u32
)blk
);
286 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
287 bg
->bg_block_bitmap_hi
= cpu_to_le32(blk
>> 32);
290 void ext4_inode_bitmap_set(struct super_block
*sb
,
291 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
293 bg
->bg_inode_bitmap_lo
= cpu_to_le32((u32
)blk
);
294 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
295 bg
->bg_inode_bitmap_hi
= cpu_to_le32(blk
>> 32);
298 void ext4_inode_table_set(struct super_block
*sb
,
299 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
301 bg
->bg_inode_table_lo
= cpu_to_le32((u32
)blk
);
302 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
303 bg
->bg_inode_table_hi
= cpu_to_le32(blk
>> 32);
306 void ext4_free_group_clusters_set(struct super_block
*sb
,
307 struct ext4_group_desc
*bg
, __u32 count
)
309 bg
->bg_free_blocks_count_lo
= cpu_to_le16((__u16
)count
);
310 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
311 bg
->bg_free_blocks_count_hi
= cpu_to_le16(count
>> 16);
314 void ext4_free_inodes_set(struct super_block
*sb
,
315 struct ext4_group_desc
*bg
, __u32 count
)
317 bg
->bg_free_inodes_count_lo
= cpu_to_le16((__u16
)count
);
318 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
319 bg
->bg_free_inodes_count_hi
= cpu_to_le16(count
>> 16);
322 void ext4_used_dirs_set(struct super_block
*sb
,
323 struct ext4_group_desc
*bg
, __u32 count
)
325 bg
->bg_used_dirs_count_lo
= cpu_to_le16((__u16
)count
);
326 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
327 bg
->bg_used_dirs_count_hi
= cpu_to_le16(count
>> 16);
330 void ext4_itable_unused_set(struct super_block
*sb
,
331 struct ext4_group_desc
*bg
, __u32 count
)
333 bg
->bg_itable_unused_lo
= cpu_to_le16((__u16
)count
);
334 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
335 bg
->bg_itable_unused_hi
= cpu_to_le16(count
>> 16);
338 static void __ext4_update_tstamp(__le32
*lo
, __u8
*hi
)
340 time64_t now
= ktime_get_real_seconds();
342 now
= clamp_val(now
, 0, (1ull << 40) - 1);
344 *lo
= cpu_to_le32(lower_32_bits(now
));
345 *hi
= upper_32_bits(now
);
348 static time64_t
__ext4_get_tstamp(__le32
*lo
, __u8
*hi
)
350 return ((time64_t
)(*hi
) << 32) + le32_to_cpu(*lo
);
352 #define ext4_update_tstamp(es, tstamp) \
353 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
354 #define ext4_get_tstamp(es, tstamp) \
355 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
357 static void __save_error_info(struct super_block
*sb
, const char *func
,
360 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
362 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
363 if (bdev_read_only(sb
->s_bdev
))
365 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
366 ext4_update_tstamp(es
, s_last_error_time
);
367 strncpy(es
->s_last_error_func
, func
, sizeof(es
->s_last_error_func
));
368 es
->s_last_error_line
= cpu_to_le32(line
);
369 if (!es
->s_first_error_time
) {
370 es
->s_first_error_time
= es
->s_last_error_time
;
371 es
->s_first_error_time_hi
= es
->s_last_error_time_hi
;
372 strncpy(es
->s_first_error_func
, func
,
373 sizeof(es
->s_first_error_func
));
374 es
->s_first_error_line
= cpu_to_le32(line
);
375 es
->s_first_error_ino
= es
->s_last_error_ino
;
376 es
->s_first_error_block
= es
->s_last_error_block
;
379 * Start the daily error reporting function if it hasn't been
382 if (!es
->s_error_count
)
383 mod_timer(&EXT4_SB(sb
)->s_err_report
, jiffies
+ 24*60*60*HZ
);
384 le32_add_cpu(&es
->s_error_count
, 1);
387 static void save_error_info(struct super_block
*sb
, const char *func
,
390 __save_error_info(sb
, func
, line
);
391 ext4_commit_super(sb
, 1);
395 * The del_gendisk() function uninitializes the disk-specific data
396 * structures, including the bdi structure, without telling anyone
397 * else. Once this happens, any attempt to call mark_buffer_dirty()
398 * (for example, by ext4_commit_super), will cause a kernel OOPS.
399 * This is a kludge to prevent these oops until we can put in a proper
400 * hook in del_gendisk() to inform the VFS and file system layers.
402 static int block_device_ejected(struct super_block
*sb
)
404 struct inode
*bd_inode
= sb
->s_bdev
->bd_inode
;
405 struct backing_dev_info
*bdi
= inode_to_bdi(bd_inode
);
407 return bdi
->dev
== NULL
;
410 static void ext4_journal_commit_callback(journal_t
*journal
, transaction_t
*txn
)
412 struct super_block
*sb
= journal
->j_private
;
413 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
414 int error
= is_journal_aborted(journal
);
415 struct ext4_journal_cb_entry
*jce
;
417 BUG_ON(txn
->t_state
== T_FINISHED
);
419 ext4_process_freed_data(sb
, txn
->t_tid
);
421 spin_lock(&sbi
->s_md_lock
);
422 while (!list_empty(&txn
->t_private_list
)) {
423 jce
= list_entry(txn
->t_private_list
.next
,
424 struct ext4_journal_cb_entry
, jce_list
);
425 list_del_init(&jce
->jce_list
);
426 spin_unlock(&sbi
->s_md_lock
);
427 jce
->jce_func(sb
, jce
, error
);
428 spin_lock(&sbi
->s_md_lock
);
430 spin_unlock(&sbi
->s_md_lock
);
433 static bool system_going_down(void)
435 return system_state
== SYSTEM_HALT
|| system_state
== SYSTEM_POWER_OFF
436 || system_state
== SYSTEM_RESTART
;
439 /* Deal with the reporting of failure conditions on a filesystem such as
440 * inconsistencies detected or read IO failures.
442 * On ext2, we can store the error state of the filesystem in the
443 * superblock. That is not possible on ext4, because we may have other
444 * write ordering constraints on the superblock which prevent us from
445 * writing it out straight away; and given that the journal is about to
446 * be aborted, we can't rely on the current, or future, transactions to
447 * write out the superblock safely.
449 * We'll just use the jbd2_journal_abort() error code to record an error in
450 * the journal instead. On recovery, the journal will complain about
451 * that error until we've noted it down and cleared it.
454 static void ext4_handle_error(struct super_block
*sb
)
456 if (test_opt(sb
, WARN_ON_ERROR
))
462 if (!test_opt(sb
, ERRORS_CONT
)) {
463 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
465 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
467 jbd2_journal_abort(journal
, -EIO
);
470 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
471 * could panic during 'reboot -f' as the underlying device got already
474 if (test_opt(sb
, ERRORS_RO
) || system_going_down()) {
475 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
477 * Make sure updated value of ->s_mount_flags will be visible
478 * before ->s_flags update
481 sb
->s_flags
|= SB_RDONLY
;
482 } else if (test_opt(sb
, ERRORS_PANIC
)) {
483 if (EXT4_SB(sb
)->s_journal
&&
484 !(EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_REC_ERR
))
486 panic("EXT4-fs (device %s): panic forced after error\n",
491 #define ext4_error_ratelimit(sb) \
492 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
495 void __ext4_error(struct super_block
*sb
, const char *function
,
496 unsigned int line
, const char *fmt
, ...)
498 struct va_format vaf
;
501 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb
))))
504 trace_ext4_error(sb
, function
, line
);
505 if (ext4_error_ratelimit(sb
)) {
510 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
511 sb
->s_id
, function
, line
, current
->comm
, &vaf
);
514 save_error_info(sb
, function
, line
);
515 ext4_handle_error(sb
);
518 void __ext4_error_inode(struct inode
*inode
, const char *function
,
519 unsigned int line
, ext4_fsblk_t block
,
520 const char *fmt
, ...)
523 struct va_format vaf
;
524 struct ext4_super_block
*es
= EXT4_SB(inode
->i_sb
)->s_es
;
526 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode
->i_sb
))))
529 trace_ext4_error(inode
->i_sb
, function
, line
);
530 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
531 es
->s_last_error_block
= cpu_to_le64(block
);
532 if (ext4_error_ratelimit(inode
->i_sb
)) {
537 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
538 "inode #%lu: block %llu: comm %s: %pV\n",
539 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
540 block
, current
->comm
, &vaf
);
542 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
543 "inode #%lu: comm %s: %pV\n",
544 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
545 current
->comm
, &vaf
);
548 save_error_info(inode
->i_sb
, function
, line
);
549 ext4_handle_error(inode
->i_sb
);
552 void __ext4_error_file(struct file
*file
, const char *function
,
553 unsigned int line
, ext4_fsblk_t block
,
554 const char *fmt
, ...)
557 struct va_format vaf
;
558 struct ext4_super_block
*es
;
559 struct inode
*inode
= file_inode(file
);
560 char pathname
[80], *path
;
562 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode
->i_sb
))))
565 trace_ext4_error(inode
->i_sb
, function
, line
);
566 es
= EXT4_SB(inode
->i_sb
)->s_es
;
567 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
568 if (ext4_error_ratelimit(inode
->i_sb
)) {
569 path
= file_path(file
, pathname
, sizeof(pathname
));
577 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
578 "block %llu: comm %s: path %s: %pV\n",
579 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
580 block
, current
->comm
, path
, &vaf
);
583 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
584 "comm %s: path %s: %pV\n",
585 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
586 current
->comm
, path
, &vaf
);
589 save_error_info(inode
->i_sb
, function
, line
);
590 ext4_handle_error(inode
->i_sb
);
593 const char *ext4_decode_error(struct super_block
*sb
, int errno
,
600 errstr
= "Corrupt filesystem";
603 errstr
= "Filesystem failed CRC";
606 errstr
= "IO failure";
609 errstr
= "Out of memory";
612 if (!sb
|| (EXT4_SB(sb
)->s_journal
&&
613 EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
))
614 errstr
= "Journal has aborted";
616 errstr
= "Readonly filesystem";
619 /* If the caller passed in an extra buffer for unknown
620 * errors, textualise them now. Else we just return
623 /* Check for truncated error codes... */
624 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
633 /* __ext4_std_error decodes expected errors from journaling functions
634 * automatically and invokes the appropriate error response. */
636 void __ext4_std_error(struct super_block
*sb
, const char *function
,
637 unsigned int line
, int errno
)
642 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb
))))
645 /* Special case: if the error is EROFS, and we're not already
646 * inside a transaction, then there's really no point in logging
648 if (errno
== -EROFS
&& journal_current_handle() == NULL
&& sb_rdonly(sb
))
651 if (ext4_error_ratelimit(sb
)) {
652 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
653 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s:%d: %s\n",
654 sb
->s_id
, function
, line
, errstr
);
657 save_error_info(sb
, function
, line
);
658 ext4_handle_error(sb
);
662 * ext4_abort is a much stronger failure handler than ext4_error. The
663 * abort function may be used to deal with unrecoverable failures such
664 * as journal IO errors or ENOMEM at a critical moment in log management.
666 * We unconditionally force the filesystem into an ABORT|READONLY state,
667 * unless the error response on the fs has been set to panic in which
668 * case we take the easy way out and panic immediately.
671 void __ext4_abort(struct super_block
*sb
, const char *function
,
672 unsigned int line
, const char *fmt
, ...)
674 struct va_format vaf
;
677 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb
))))
680 save_error_info(sb
, function
, line
);
684 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: %pV\n",
685 sb
->s_id
, function
, line
, &vaf
);
688 if (sb_rdonly(sb
) == 0) {
689 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
690 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
692 * Make sure updated value of ->s_mount_flags will be visible
693 * before ->s_flags update
696 sb
->s_flags
|= SB_RDONLY
;
697 if (EXT4_SB(sb
)->s_journal
)
698 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
699 save_error_info(sb
, function
, line
);
701 if (test_opt(sb
, ERRORS_PANIC
)) {
702 if (EXT4_SB(sb
)->s_journal
&&
703 !(EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_REC_ERR
))
705 panic("EXT4-fs panic from previous error\n");
709 void __ext4_msg(struct super_block
*sb
,
710 const char *prefix
, const char *fmt
, ...)
712 struct va_format vaf
;
715 if (!___ratelimit(&(EXT4_SB(sb
)->s_msg_ratelimit_state
), "EXT4-fs"))
721 printk("%sEXT4-fs (%s): %pV\n", prefix
, sb
->s_id
, &vaf
);
725 #define ext4_warning_ratelimit(sb) \
726 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
729 void __ext4_warning(struct super_block
*sb
, const char *function
,
730 unsigned int line
, const char *fmt
, ...)
732 struct va_format vaf
;
735 if (!ext4_warning_ratelimit(sb
))
741 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: %pV\n",
742 sb
->s_id
, function
, line
, &vaf
);
746 void __ext4_warning_inode(const struct inode
*inode
, const char *function
,
747 unsigned int line
, const char *fmt
, ...)
749 struct va_format vaf
;
752 if (!ext4_warning_ratelimit(inode
->i_sb
))
758 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: "
759 "inode #%lu: comm %s: %pV\n", inode
->i_sb
->s_id
,
760 function
, line
, inode
->i_ino
, current
->comm
, &vaf
);
764 void __ext4_grp_locked_error(const char *function
, unsigned int line
,
765 struct super_block
*sb
, ext4_group_t grp
,
766 unsigned long ino
, ext4_fsblk_t block
,
767 const char *fmt
, ...)
771 struct va_format vaf
;
773 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
775 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb
))))
778 trace_ext4_error(sb
, function
, line
);
779 es
->s_last_error_ino
= cpu_to_le32(ino
);
780 es
->s_last_error_block
= cpu_to_le64(block
);
781 __save_error_info(sb
, function
, line
);
783 if (ext4_error_ratelimit(sb
)) {
787 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: group %u, ",
788 sb
->s_id
, function
, line
, grp
);
790 printk(KERN_CONT
"inode %lu: ", ino
);
792 printk(KERN_CONT
"block %llu:",
793 (unsigned long long) block
);
794 printk(KERN_CONT
"%pV\n", &vaf
);
798 if (test_opt(sb
, WARN_ON_ERROR
))
801 if (test_opt(sb
, ERRORS_CONT
)) {
802 ext4_commit_super(sb
, 0);
806 ext4_unlock_group(sb
, grp
);
807 ext4_commit_super(sb
, 1);
808 ext4_handle_error(sb
);
810 * We only get here in the ERRORS_RO case; relocking the group
811 * may be dangerous, but nothing bad will happen since the
812 * filesystem will have already been marked read/only and the
813 * journal has been aborted. We return 1 as a hint to callers
814 * who might what to use the return value from
815 * ext4_grp_locked_error() to distinguish between the
816 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
817 * aggressively from the ext4 function in question, with a
818 * more appropriate error code.
820 ext4_lock_group(sb
, grp
);
824 void ext4_mark_group_bitmap_corrupted(struct super_block
*sb
,
828 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
829 struct ext4_group_info
*grp
= ext4_get_group_info(sb
, group
);
830 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, group
, NULL
);
833 if (flags
& EXT4_GROUP_INFO_BBITMAP_CORRUPT
) {
834 ret
= ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT
,
837 percpu_counter_sub(&sbi
->s_freeclusters_counter
,
841 if (flags
& EXT4_GROUP_INFO_IBITMAP_CORRUPT
) {
842 ret
= ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT
,
847 count
= ext4_free_inodes_count(sb
, gdp
);
848 percpu_counter_sub(&sbi
->s_freeinodes_counter
,
854 void ext4_update_dynamic_rev(struct super_block
*sb
)
856 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
858 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
862 "updating to rev %d because of new feature flag, "
863 "running e2fsck is recommended",
866 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
867 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
868 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
869 /* leave es->s_feature_*compat flags alone */
870 /* es->s_uuid will be set by e2fsck if empty */
873 * The rest of the superblock fields should be zero, and if not it
874 * means they are likely already in use, so leave them alone. We
875 * can leave it up to e2fsck to clean up any inconsistencies there.
880 * Open the external journal device
882 static struct block_device
*ext4_blkdev_get(dev_t dev
, struct super_block
*sb
)
884 struct block_device
*bdev
;
885 char b
[BDEVNAME_SIZE
];
887 bdev
= blkdev_get_by_dev(dev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
, sb
);
893 ext4_msg(sb
, KERN_ERR
, "failed to open journal device %s: %ld",
894 __bdevname(dev
, b
), PTR_ERR(bdev
));
899 * Release the journal device
901 static void ext4_blkdev_put(struct block_device
*bdev
)
903 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
);
906 static void ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
908 struct block_device
*bdev
;
909 bdev
= sbi
->journal_bdev
;
911 ext4_blkdev_put(bdev
);
912 sbi
->journal_bdev
= NULL
;
916 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
918 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
921 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
925 ext4_msg(sb
, KERN_ERR
, "sb orphan head is %d",
926 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
928 printk(KERN_ERR
"sb_info orphan list:\n");
929 list_for_each(l
, &sbi
->s_orphan
) {
930 struct inode
*inode
= orphan_list_entry(l
);
932 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
933 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
934 inode
->i_mode
, inode
->i_nlink
,
940 static int ext4_quota_off(struct super_block
*sb
, int type
);
942 static inline void ext4_quota_off_umount(struct super_block
*sb
)
946 /* Use our quota_off function to clear inode flags etc. */
947 for (type
= 0; type
< EXT4_MAXQUOTAS
; type
++)
948 ext4_quota_off(sb
, type
);
952 * This is a helper function which is used in the mount/remount
953 * codepaths (which holds s_umount) to fetch the quota file name.
955 static inline char *get_qf_name(struct super_block
*sb
,
956 struct ext4_sb_info
*sbi
,
959 return rcu_dereference_protected(sbi
->s_qf_names
[type
],
960 lockdep_is_held(&sb
->s_umount
));
963 static inline void ext4_quota_off_umount(struct super_block
*sb
)
968 static void ext4_put_super(struct super_block
*sb
)
970 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
971 struct ext4_super_block
*es
= sbi
->s_es
;
975 ext4_unregister_li_request(sb
);
976 ext4_quota_off_umount(sb
);
978 destroy_workqueue(sbi
->rsv_conversion_wq
);
980 if (sbi
->s_journal
) {
981 aborted
= is_journal_aborted(sbi
->s_journal
);
982 err
= jbd2_journal_destroy(sbi
->s_journal
);
983 sbi
->s_journal
= NULL
;
984 if ((err
< 0) && !aborted
)
985 ext4_abort(sb
, "Couldn't clean up the journal");
988 ext4_unregister_sysfs(sb
);
989 ext4_es_unregister_shrinker(sbi
);
990 del_timer_sync(&sbi
->s_err_report
);
991 ext4_release_system_zone(sb
);
993 ext4_ext_release(sb
);
995 if (!sb_rdonly(sb
) && !aborted
) {
996 ext4_clear_feature_journal_needs_recovery(sb
);
997 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
1000 ext4_commit_super(sb
, 1);
1002 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
1003 brelse(sbi
->s_group_desc
[i
]);
1004 kvfree(sbi
->s_group_desc
);
1005 kvfree(sbi
->s_flex_groups
);
1006 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
1007 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
1008 percpu_counter_destroy(&sbi
->s_dirs_counter
);
1009 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
1010 percpu_free_rwsem(&sbi
->s_journal_flag_rwsem
);
1012 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
1013 kfree(get_qf_name(sb
, sbi
, i
));
1016 /* Debugging code just in case the in-memory inode orphan list
1017 * isn't empty. The on-disk one can be non-empty if we've
1018 * detected an error and taken the fs readonly, but the
1019 * in-memory list had better be clean by this point. */
1020 if (!list_empty(&sbi
->s_orphan
))
1021 dump_orphan_list(sb
, sbi
);
1022 J_ASSERT(list_empty(&sbi
->s_orphan
));
1024 sync_blockdev(sb
->s_bdev
);
1025 invalidate_bdev(sb
->s_bdev
);
1026 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
1028 * Invalidate the journal device's buffers. We don't want them
1029 * floating about in memory - the physical journal device may
1030 * hotswapped, and it breaks the `ro-after' testing code.
1032 sync_blockdev(sbi
->journal_bdev
);
1033 invalidate_bdev(sbi
->journal_bdev
);
1034 ext4_blkdev_remove(sbi
);
1037 ext4_xattr_destroy_cache(sbi
->s_ea_inode_cache
);
1038 sbi
->s_ea_inode_cache
= NULL
;
1040 ext4_xattr_destroy_cache(sbi
->s_ea_block_cache
);
1041 sbi
->s_ea_block_cache
= NULL
;
1044 kthread_stop(sbi
->s_mmp_tsk
);
1046 sb
->s_fs_info
= NULL
;
1048 * Now that we are completely done shutting down the
1049 * superblock, we need to actually destroy the kobject.
1051 kobject_put(&sbi
->s_kobj
);
1052 wait_for_completion(&sbi
->s_kobj_unregister
);
1053 if (sbi
->s_chksum_driver
)
1054 crypto_free_shash(sbi
->s_chksum_driver
);
1055 kfree(sbi
->s_blockgroup_lock
);
1056 fs_put_dax(sbi
->s_daxdev
);
1060 static struct kmem_cache
*ext4_inode_cachep
;
1063 * Called inside transaction, so use GFP_NOFS
1065 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
1067 struct ext4_inode_info
*ei
;
1069 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
1073 inode_set_iversion(&ei
->vfs_inode
, 1);
1074 spin_lock_init(&ei
->i_raw_lock
);
1075 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
1076 spin_lock_init(&ei
->i_prealloc_lock
);
1077 ext4_es_init_tree(&ei
->i_es_tree
);
1078 rwlock_init(&ei
->i_es_lock
);
1079 INIT_LIST_HEAD(&ei
->i_es_list
);
1080 ei
->i_es_all_nr
= 0;
1081 ei
->i_es_shk_nr
= 0;
1082 ei
->i_es_shrink_lblk
= 0;
1083 ei
->i_reserved_data_blocks
= 0;
1084 ei
->i_da_metadata_calc_len
= 0;
1085 ei
->i_da_metadata_calc_last_lblock
= 0;
1086 spin_lock_init(&(ei
->i_block_reservation_lock
));
1087 ext4_init_pending_tree(&ei
->i_pending_tree
);
1089 ei
->i_reserved_quota
= 0;
1090 memset(&ei
->i_dquot
, 0, sizeof(ei
->i_dquot
));
1093 INIT_LIST_HEAD(&ei
->i_rsv_conversion_list
);
1094 spin_lock_init(&ei
->i_completed_io_lock
);
1096 ei
->i_datasync_tid
= 0;
1097 atomic_set(&ei
->i_unwritten
, 0);
1098 INIT_WORK(&ei
->i_rsv_conversion_work
, ext4_end_io_rsv_work
);
1099 return &ei
->vfs_inode
;
1102 static int ext4_drop_inode(struct inode
*inode
)
1104 int drop
= generic_drop_inode(inode
);
1106 trace_ext4_drop_inode(inode
, drop
);
1110 static void ext4_i_callback(struct rcu_head
*head
)
1112 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
1113 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
1116 static void ext4_destroy_inode(struct inode
*inode
)
1118 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
1119 ext4_msg(inode
->i_sb
, KERN_ERR
,
1120 "Inode %lu (%p): orphan list check failed!",
1121 inode
->i_ino
, EXT4_I(inode
));
1122 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
1123 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
1127 call_rcu(&inode
->i_rcu
, ext4_i_callback
);
1130 static void init_once(void *foo
)
1132 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
1134 INIT_LIST_HEAD(&ei
->i_orphan
);
1135 init_rwsem(&ei
->xattr_sem
);
1136 init_rwsem(&ei
->i_data_sem
);
1137 init_rwsem(&ei
->i_mmap_sem
);
1138 inode_init_once(&ei
->vfs_inode
);
1141 static int __init
init_inodecache(void)
1143 ext4_inode_cachep
= kmem_cache_create_usercopy("ext4_inode_cache",
1144 sizeof(struct ext4_inode_info
), 0,
1145 (SLAB_RECLAIM_ACCOUNT
|SLAB_MEM_SPREAD
|
1147 offsetof(struct ext4_inode_info
, i_data
),
1148 sizeof_field(struct ext4_inode_info
, i_data
),
1150 if (ext4_inode_cachep
== NULL
)
1155 static void destroy_inodecache(void)
1158 * Make sure all delayed rcu free inodes are flushed before we
1162 kmem_cache_destroy(ext4_inode_cachep
);
1165 void ext4_clear_inode(struct inode
*inode
)
1167 invalidate_inode_buffers(inode
);
1170 ext4_discard_preallocations(inode
);
1171 ext4_es_remove_extent(inode
, 0, EXT_MAX_BLOCKS
);
1172 if (EXT4_I(inode
)->jinode
) {
1173 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode
),
1174 EXT4_I(inode
)->jinode
);
1175 jbd2_free_inode(EXT4_I(inode
)->jinode
);
1176 EXT4_I(inode
)->jinode
= NULL
;
1178 fscrypt_put_encryption_info(inode
);
1181 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
1182 u64 ino
, u32 generation
)
1184 struct inode
*inode
;
1187 * Currently we don't know the generation for parent directory, so
1188 * a generation of 0 means "accept any"
1190 inode
= ext4_iget(sb
, ino
, EXT4_IGET_HANDLE
);
1192 return ERR_CAST(inode
);
1193 if (generation
&& inode
->i_generation
!= generation
) {
1195 return ERR_PTR(-ESTALE
);
1201 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1202 int fh_len
, int fh_type
)
1204 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
1205 ext4_nfs_get_inode
);
1208 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1209 int fh_len
, int fh_type
)
1211 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1212 ext4_nfs_get_inode
);
1215 static int ext4_nfs_commit_metadata(struct inode
*inode
)
1217 struct writeback_control wbc
= {
1218 .sync_mode
= WB_SYNC_ALL
1221 trace_ext4_nfs_commit_metadata(inode
);
1222 return ext4_write_inode(inode
, &wbc
);
1226 * Try to release metadata pages (indirect blocks, directories) which are
1227 * mapped via the block device. Since these pages could have journal heads
1228 * which would prevent try_to_free_buffers() from freeing them, we must use
1229 * jbd2 layer's try_to_free_buffers() function to release them.
1231 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
1234 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
1236 WARN_ON(PageChecked(page
));
1237 if (!page_has_buffers(page
))
1240 return jbd2_journal_try_to_free_buffers(journal
, page
,
1241 wait
& ~__GFP_DIRECT_RECLAIM
);
1242 return try_to_free_buffers(page
);
1245 #ifdef CONFIG_FS_ENCRYPTION
1246 static int ext4_get_context(struct inode
*inode
, void *ctx
, size_t len
)
1248 return ext4_xattr_get(inode
, EXT4_XATTR_INDEX_ENCRYPTION
,
1249 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT
, ctx
, len
);
1252 static int ext4_set_context(struct inode
*inode
, const void *ctx
, size_t len
,
1255 handle_t
*handle
= fs_data
;
1256 int res
, res2
, credits
, retries
= 0;
1259 * Encrypting the root directory is not allowed because e2fsck expects
1260 * lost+found to exist and be unencrypted, and encrypting the root
1261 * directory would imply encrypting the lost+found directory as well as
1262 * the filename "lost+found" itself.
1264 if (inode
->i_ino
== EXT4_ROOT_INO
)
1267 if (WARN_ON_ONCE(IS_DAX(inode
) && i_size_read(inode
)))
1270 res
= ext4_convert_inline_data(inode
);
1275 * If a journal handle was specified, then the encryption context is
1276 * being set on a new inode via inheritance and is part of a larger
1277 * transaction to create the inode. Otherwise the encryption context is
1278 * being set on an existing inode in its own transaction. Only in the
1279 * latter case should the "retry on ENOSPC" logic be used.
1283 res
= ext4_xattr_set_handle(handle
, inode
,
1284 EXT4_XATTR_INDEX_ENCRYPTION
,
1285 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT
,
1288 ext4_set_inode_flag(inode
, EXT4_INODE_ENCRYPT
);
1289 ext4_clear_inode_state(inode
,
1290 EXT4_STATE_MAY_INLINE_DATA
);
1292 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1293 * S_DAX may be disabled
1295 ext4_set_inode_flags(inode
);
1300 res
= dquot_initialize(inode
);
1304 res
= ext4_xattr_set_credits(inode
, len
, false /* is_create */,
1309 handle
= ext4_journal_start(inode
, EXT4_HT_MISC
, credits
);
1311 return PTR_ERR(handle
);
1313 res
= ext4_xattr_set_handle(handle
, inode
, EXT4_XATTR_INDEX_ENCRYPTION
,
1314 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT
,
1317 ext4_set_inode_flag(inode
, EXT4_INODE_ENCRYPT
);
1319 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1320 * S_DAX may be disabled
1322 ext4_set_inode_flags(inode
);
1323 res
= ext4_mark_inode_dirty(handle
, inode
);
1325 EXT4_ERROR_INODE(inode
, "Failed to mark inode dirty");
1327 res2
= ext4_journal_stop(handle
);
1329 if (res
== -ENOSPC
&& ext4_should_retry_alloc(inode
->i_sb
, &retries
))
1336 static bool ext4_dummy_context(struct inode
*inode
)
1338 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode
->i_sb
));
1341 static const struct fscrypt_operations ext4_cryptops
= {
1342 .key_prefix
= "ext4:",
1343 .get_context
= ext4_get_context
,
1344 .set_context
= ext4_set_context
,
1345 .dummy_context
= ext4_dummy_context
,
1346 .empty_dir
= ext4_empty_dir
,
1347 .max_namelen
= EXT4_NAME_LEN
,
1352 static const char * const quotatypes
[] = INITQFNAMES
;
1353 #define QTYPE2NAME(t) (quotatypes[t])
1355 static int ext4_write_dquot(struct dquot
*dquot
);
1356 static int ext4_acquire_dquot(struct dquot
*dquot
);
1357 static int ext4_release_dquot(struct dquot
*dquot
);
1358 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
1359 static int ext4_write_info(struct super_block
*sb
, int type
);
1360 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
1361 const struct path
*path
);
1362 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
1363 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
1364 size_t len
, loff_t off
);
1365 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
1366 const char *data
, size_t len
, loff_t off
);
1367 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
1368 unsigned int flags
);
1369 static int ext4_enable_quotas(struct super_block
*sb
);
1370 static int ext4_get_next_id(struct super_block
*sb
, struct kqid
*qid
);
1372 static struct dquot
**ext4_get_dquots(struct inode
*inode
)
1374 return EXT4_I(inode
)->i_dquot
;
1377 static const struct dquot_operations ext4_quota_operations
= {
1378 .get_reserved_space
= ext4_get_reserved_space
,
1379 .write_dquot
= ext4_write_dquot
,
1380 .acquire_dquot
= ext4_acquire_dquot
,
1381 .release_dquot
= ext4_release_dquot
,
1382 .mark_dirty
= ext4_mark_dquot_dirty
,
1383 .write_info
= ext4_write_info
,
1384 .alloc_dquot
= dquot_alloc
,
1385 .destroy_dquot
= dquot_destroy
,
1386 .get_projid
= ext4_get_projid
,
1387 .get_inode_usage
= ext4_get_inode_usage
,
1388 .get_next_id
= ext4_get_next_id
,
1391 static const struct quotactl_ops ext4_qctl_operations
= {
1392 .quota_on
= ext4_quota_on
,
1393 .quota_off
= ext4_quota_off
,
1394 .quota_sync
= dquot_quota_sync
,
1395 .get_state
= dquot_get_state
,
1396 .set_info
= dquot_set_dqinfo
,
1397 .get_dqblk
= dquot_get_dqblk
,
1398 .set_dqblk
= dquot_set_dqblk
,
1399 .get_nextdqblk
= dquot_get_next_dqblk
,
1403 static const struct super_operations ext4_sops
= {
1404 .alloc_inode
= ext4_alloc_inode
,
1405 .destroy_inode
= ext4_destroy_inode
,
1406 .write_inode
= ext4_write_inode
,
1407 .dirty_inode
= ext4_dirty_inode
,
1408 .drop_inode
= ext4_drop_inode
,
1409 .evict_inode
= ext4_evict_inode
,
1410 .put_super
= ext4_put_super
,
1411 .sync_fs
= ext4_sync_fs
,
1412 .freeze_fs
= ext4_freeze
,
1413 .unfreeze_fs
= ext4_unfreeze
,
1414 .statfs
= ext4_statfs
,
1415 .remount_fs
= ext4_remount
,
1416 .show_options
= ext4_show_options
,
1418 .quota_read
= ext4_quota_read
,
1419 .quota_write
= ext4_quota_write
,
1420 .get_dquots
= ext4_get_dquots
,
1422 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1425 static const struct export_operations ext4_export_ops
= {
1426 .fh_to_dentry
= ext4_fh_to_dentry
,
1427 .fh_to_parent
= ext4_fh_to_parent
,
1428 .get_parent
= ext4_get_parent
,
1429 .commit_metadata
= ext4_nfs_commit_metadata
,
1433 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1434 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1435 Opt_nouid32
, Opt_debug
, Opt_removed
,
1436 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1437 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
,
1438 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
, Opt_journal_dev
,
1439 Opt_journal_path
, Opt_journal_checksum
, Opt_journal_async_commit
,
1440 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1441 Opt_data_err_abort
, Opt_data_err_ignore
, Opt_test_dummy_encryption
,
1442 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1443 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
1444 Opt_noquota
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
1445 Opt_usrquota
, Opt_grpquota
, Opt_prjquota
, Opt_i_version
, Opt_dax
,
1446 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
, Opt_warn_on_error
,
1447 Opt_nowarn_on_error
, Opt_mblk_io_submit
,
1448 Opt_lazytime
, Opt_nolazytime
, Opt_debug_want_extra_isize
,
1449 Opt_nomblk_io_submit
, Opt_block_validity
, Opt_noblock_validity
,
1450 Opt_inode_readahead_blks
, Opt_journal_ioprio
,
1451 Opt_dioread_nolock
, Opt_dioread_lock
,
1452 Opt_discard
, Opt_nodiscard
, Opt_init_itable
, Opt_noinit_itable
,
1453 Opt_max_dir_size_kb
, Opt_nojournal_checksum
, Opt_nombcache
,
1456 static const match_table_t tokens
= {
1457 {Opt_bsd_df
, "bsddf"},
1458 {Opt_minix_df
, "minixdf"},
1459 {Opt_grpid
, "grpid"},
1460 {Opt_grpid
, "bsdgroups"},
1461 {Opt_nogrpid
, "nogrpid"},
1462 {Opt_nogrpid
, "sysvgroups"},
1463 {Opt_resgid
, "resgid=%u"},
1464 {Opt_resuid
, "resuid=%u"},
1466 {Opt_err_cont
, "errors=continue"},
1467 {Opt_err_panic
, "errors=panic"},
1468 {Opt_err_ro
, "errors=remount-ro"},
1469 {Opt_nouid32
, "nouid32"},
1470 {Opt_debug
, "debug"},
1471 {Opt_removed
, "oldalloc"},
1472 {Opt_removed
, "orlov"},
1473 {Opt_user_xattr
, "user_xattr"},
1474 {Opt_nouser_xattr
, "nouser_xattr"},
1476 {Opt_noacl
, "noacl"},
1477 {Opt_noload
, "norecovery"},
1478 {Opt_noload
, "noload"},
1479 {Opt_removed
, "nobh"},
1480 {Opt_removed
, "bh"},
1481 {Opt_commit
, "commit=%u"},
1482 {Opt_min_batch_time
, "min_batch_time=%u"},
1483 {Opt_max_batch_time
, "max_batch_time=%u"},
1484 {Opt_journal_dev
, "journal_dev=%u"},
1485 {Opt_journal_path
, "journal_path=%s"},
1486 {Opt_journal_checksum
, "journal_checksum"},
1487 {Opt_nojournal_checksum
, "nojournal_checksum"},
1488 {Opt_journal_async_commit
, "journal_async_commit"},
1489 {Opt_abort
, "abort"},
1490 {Opt_data_journal
, "data=journal"},
1491 {Opt_data_ordered
, "data=ordered"},
1492 {Opt_data_writeback
, "data=writeback"},
1493 {Opt_data_err_abort
, "data_err=abort"},
1494 {Opt_data_err_ignore
, "data_err=ignore"},
1495 {Opt_offusrjquota
, "usrjquota="},
1496 {Opt_usrjquota
, "usrjquota=%s"},
1497 {Opt_offgrpjquota
, "grpjquota="},
1498 {Opt_grpjquota
, "grpjquota=%s"},
1499 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1500 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1501 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
1502 {Opt_grpquota
, "grpquota"},
1503 {Opt_noquota
, "noquota"},
1504 {Opt_quota
, "quota"},
1505 {Opt_usrquota
, "usrquota"},
1506 {Opt_prjquota
, "prjquota"},
1507 {Opt_barrier
, "barrier=%u"},
1508 {Opt_barrier
, "barrier"},
1509 {Opt_nobarrier
, "nobarrier"},
1510 {Opt_i_version
, "i_version"},
1512 {Opt_stripe
, "stripe=%u"},
1513 {Opt_delalloc
, "delalloc"},
1514 {Opt_warn_on_error
, "warn_on_error"},
1515 {Opt_nowarn_on_error
, "nowarn_on_error"},
1516 {Opt_lazytime
, "lazytime"},
1517 {Opt_nolazytime
, "nolazytime"},
1518 {Opt_debug_want_extra_isize
, "debug_want_extra_isize=%u"},
1519 {Opt_nodelalloc
, "nodelalloc"},
1520 {Opt_removed
, "mblk_io_submit"},
1521 {Opt_removed
, "nomblk_io_submit"},
1522 {Opt_block_validity
, "block_validity"},
1523 {Opt_noblock_validity
, "noblock_validity"},
1524 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1525 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1526 {Opt_auto_da_alloc
, "auto_da_alloc=%u"},
1527 {Opt_auto_da_alloc
, "auto_da_alloc"},
1528 {Opt_noauto_da_alloc
, "noauto_da_alloc"},
1529 {Opt_dioread_nolock
, "dioread_nolock"},
1530 {Opt_dioread_lock
, "dioread_lock"},
1531 {Opt_discard
, "discard"},
1532 {Opt_nodiscard
, "nodiscard"},
1533 {Opt_init_itable
, "init_itable=%u"},
1534 {Opt_init_itable
, "init_itable"},
1535 {Opt_noinit_itable
, "noinit_itable"},
1536 {Opt_max_dir_size_kb
, "max_dir_size_kb=%u"},
1537 {Opt_test_dummy_encryption
, "test_dummy_encryption"},
1538 {Opt_nombcache
, "nombcache"},
1539 {Opt_nombcache
, "no_mbcache"}, /* for backward compatibility */
1540 {Opt_removed
, "check=none"}, /* mount option from ext2/3 */
1541 {Opt_removed
, "nocheck"}, /* mount option from ext2/3 */
1542 {Opt_removed
, "reservation"}, /* mount option from ext2/3 */
1543 {Opt_removed
, "noreservation"}, /* mount option from ext2/3 */
1544 {Opt_removed
, "journal=%u"}, /* mount option from ext2/3 */
1548 static ext4_fsblk_t
get_sb_block(void **data
)
1550 ext4_fsblk_t sb_block
;
1551 char *options
= (char *) *data
;
1553 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1554 return 1; /* Default location */
1557 /* TODO: use simple_strtoll with >32bit ext4 */
1558 sb_block
= simple_strtoul(options
, &options
, 0);
1559 if (*options
&& *options
!= ',') {
1560 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1564 if (*options
== ',')
1566 *data
= (void *) options
;
1571 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1572 static const char deprecated_msg
[] =
1573 "Mount option \"%s\" will be removed by %s\n"
1574 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1577 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
1579 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1580 char *qname
, *old_qname
= get_qf_name(sb
, sbi
, qtype
);
1583 if (sb_any_quota_loaded(sb
) && !old_qname
) {
1584 ext4_msg(sb
, KERN_ERR
,
1585 "Cannot change journaled "
1586 "quota options when quota turned on");
1589 if (ext4_has_feature_quota(sb
)) {
1590 ext4_msg(sb
, KERN_INFO
, "Journaled quota options "
1591 "ignored when QUOTA feature is enabled");
1594 qname
= match_strdup(args
);
1596 ext4_msg(sb
, KERN_ERR
,
1597 "Not enough memory for storing quotafile name");
1601 if (strcmp(old_qname
, qname
) == 0)
1604 ext4_msg(sb
, KERN_ERR
,
1605 "%s quota file already specified",
1609 if (strchr(qname
, '/')) {
1610 ext4_msg(sb
, KERN_ERR
,
1611 "quotafile must be on filesystem root");
1614 rcu_assign_pointer(sbi
->s_qf_names
[qtype
], qname
);
1622 static int clear_qf_name(struct super_block
*sb
, int qtype
)
1625 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1626 char *old_qname
= get_qf_name(sb
, sbi
, qtype
);
1628 if (sb_any_quota_loaded(sb
) && old_qname
) {
1629 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
1630 " when quota turned on");
1633 rcu_assign_pointer(sbi
->s_qf_names
[qtype
], NULL
);
1640 #define MOPT_SET 0x0001
1641 #define MOPT_CLEAR 0x0002
1642 #define MOPT_NOSUPPORT 0x0004
1643 #define MOPT_EXPLICIT 0x0008
1644 #define MOPT_CLEAR_ERR 0x0010
1645 #define MOPT_GTE0 0x0020
1648 #define MOPT_QFMT 0x0040
1650 #define MOPT_Q MOPT_NOSUPPORT
1651 #define MOPT_QFMT MOPT_NOSUPPORT
1653 #define MOPT_DATAJ 0x0080
1654 #define MOPT_NO_EXT2 0x0100
1655 #define MOPT_NO_EXT3 0x0200
1656 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1657 #define MOPT_STRING 0x0400
1659 static const struct mount_opts
{
1663 } ext4_mount_opts
[] = {
1664 {Opt_minix_df
, EXT4_MOUNT_MINIX_DF
, MOPT_SET
},
1665 {Opt_bsd_df
, EXT4_MOUNT_MINIX_DF
, MOPT_CLEAR
},
1666 {Opt_grpid
, EXT4_MOUNT_GRPID
, MOPT_SET
},
1667 {Opt_nogrpid
, EXT4_MOUNT_GRPID
, MOPT_CLEAR
},
1668 {Opt_block_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_SET
},
1669 {Opt_noblock_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_CLEAR
},
1670 {Opt_dioread_nolock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1671 MOPT_EXT4_ONLY
| MOPT_SET
},
1672 {Opt_dioread_lock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1673 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1674 {Opt_discard
, EXT4_MOUNT_DISCARD
, MOPT_SET
},
1675 {Opt_nodiscard
, EXT4_MOUNT_DISCARD
, MOPT_CLEAR
},
1676 {Opt_delalloc
, EXT4_MOUNT_DELALLOC
,
1677 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1678 {Opt_nodelalloc
, EXT4_MOUNT_DELALLOC
,
1679 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1680 {Opt_warn_on_error
, EXT4_MOUNT_WARN_ON_ERROR
, MOPT_SET
},
1681 {Opt_nowarn_on_error
, EXT4_MOUNT_WARN_ON_ERROR
, MOPT_CLEAR
},
1682 {Opt_nojournal_checksum
, EXT4_MOUNT_JOURNAL_CHECKSUM
,
1683 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1684 {Opt_journal_checksum
, EXT4_MOUNT_JOURNAL_CHECKSUM
,
1685 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1686 {Opt_journal_async_commit
, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT
|
1687 EXT4_MOUNT_JOURNAL_CHECKSUM
),
1688 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1689 {Opt_noload
, EXT4_MOUNT_NOLOAD
, MOPT_NO_EXT2
| MOPT_SET
},
1690 {Opt_err_panic
, EXT4_MOUNT_ERRORS_PANIC
, MOPT_SET
| MOPT_CLEAR_ERR
},
1691 {Opt_err_ro
, EXT4_MOUNT_ERRORS_RO
, MOPT_SET
| MOPT_CLEAR_ERR
},
1692 {Opt_err_cont
, EXT4_MOUNT_ERRORS_CONT
, MOPT_SET
| MOPT_CLEAR_ERR
},
1693 {Opt_data_err_abort
, EXT4_MOUNT_DATA_ERR_ABORT
,
1695 {Opt_data_err_ignore
, EXT4_MOUNT_DATA_ERR_ABORT
,
1697 {Opt_barrier
, EXT4_MOUNT_BARRIER
, MOPT_SET
},
1698 {Opt_nobarrier
, EXT4_MOUNT_BARRIER
, MOPT_CLEAR
},
1699 {Opt_noauto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_SET
},
1700 {Opt_auto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_CLEAR
},
1701 {Opt_noinit_itable
, EXT4_MOUNT_INIT_INODE_TABLE
, MOPT_CLEAR
},
1702 {Opt_commit
, 0, MOPT_GTE0
},
1703 {Opt_max_batch_time
, 0, MOPT_GTE0
},
1704 {Opt_min_batch_time
, 0, MOPT_GTE0
},
1705 {Opt_inode_readahead_blks
, 0, MOPT_GTE0
},
1706 {Opt_init_itable
, 0, MOPT_GTE0
},
1707 {Opt_dax
, EXT4_MOUNT_DAX
, MOPT_SET
},
1708 {Opt_stripe
, 0, MOPT_GTE0
},
1709 {Opt_resuid
, 0, MOPT_GTE0
},
1710 {Opt_resgid
, 0, MOPT_GTE0
},
1711 {Opt_journal_dev
, 0, MOPT_NO_EXT2
| MOPT_GTE0
},
1712 {Opt_journal_path
, 0, MOPT_NO_EXT2
| MOPT_STRING
},
1713 {Opt_journal_ioprio
, 0, MOPT_NO_EXT2
| MOPT_GTE0
},
1714 {Opt_data_journal
, EXT4_MOUNT_JOURNAL_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1715 {Opt_data_ordered
, EXT4_MOUNT_ORDERED_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1716 {Opt_data_writeback
, EXT4_MOUNT_WRITEBACK_DATA
,
1717 MOPT_NO_EXT2
| MOPT_DATAJ
},
1718 {Opt_user_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_SET
},
1719 {Opt_nouser_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_CLEAR
},
1720 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1721 {Opt_acl
, EXT4_MOUNT_POSIX_ACL
, MOPT_SET
},
1722 {Opt_noacl
, EXT4_MOUNT_POSIX_ACL
, MOPT_CLEAR
},
1724 {Opt_acl
, 0, MOPT_NOSUPPORT
},
1725 {Opt_noacl
, 0, MOPT_NOSUPPORT
},
1727 {Opt_nouid32
, EXT4_MOUNT_NO_UID32
, MOPT_SET
},
1728 {Opt_debug
, EXT4_MOUNT_DEBUG
, MOPT_SET
},
1729 {Opt_debug_want_extra_isize
, 0, MOPT_GTE0
},
1730 {Opt_quota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
, MOPT_SET
| MOPT_Q
},
1731 {Opt_usrquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
,
1733 {Opt_grpquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_GRPQUOTA
,
1735 {Opt_prjquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_PRJQUOTA
,
1737 {Opt_noquota
, (EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
|
1738 EXT4_MOUNT_GRPQUOTA
| EXT4_MOUNT_PRJQUOTA
),
1739 MOPT_CLEAR
| MOPT_Q
},
1740 {Opt_usrjquota
, 0, MOPT_Q
},
1741 {Opt_grpjquota
, 0, MOPT_Q
},
1742 {Opt_offusrjquota
, 0, MOPT_Q
},
1743 {Opt_offgrpjquota
, 0, MOPT_Q
},
1744 {Opt_jqfmt_vfsold
, QFMT_VFS_OLD
, MOPT_QFMT
},
1745 {Opt_jqfmt_vfsv0
, QFMT_VFS_V0
, MOPT_QFMT
},
1746 {Opt_jqfmt_vfsv1
, QFMT_VFS_V1
, MOPT_QFMT
},
1747 {Opt_max_dir_size_kb
, 0, MOPT_GTE0
},
1748 {Opt_test_dummy_encryption
, 0, MOPT_GTE0
},
1749 {Opt_nombcache
, EXT4_MOUNT_NO_MBCACHE
, MOPT_SET
},
1753 static int handle_mount_opt(struct super_block
*sb
, char *opt
, int token
,
1754 substring_t
*args
, unsigned long *journal_devnum
,
1755 unsigned int *journal_ioprio
, int is_remount
)
1757 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1758 const struct mount_opts
*m
;
1764 if (token
== Opt_usrjquota
)
1765 return set_qf_name(sb
, USRQUOTA
, &args
[0]);
1766 else if (token
== Opt_grpjquota
)
1767 return set_qf_name(sb
, GRPQUOTA
, &args
[0]);
1768 else if (token
== Opt_offusrjquota
)
1769 return clear_qf_name(sb
, USRQUOTA
);
1770 else if (token
== Opt_offgrpjquota
)
1771 return clear_qf_name(sb
, GRPQUOTA
);
1775 case Opt_nouser_xattr
:
1776 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, opt
, "3.5");
1779 return 1; /* handled by get_sb_block() */
1781 ext4_msg(sb
, KERN_WARNING
, "Ignoring removed %s option", opt
);
1784 sbi
->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
1787 sb
->s_flags
|= SB_I_VERSION
;
1790 sb
->s_flags
|= SB_LAZYTIME
;
1792 case Opt_nolazytime
:
1793 sb
->s_flags
&= ~SB_LAZYTIME
;
1797 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++)
1798 if (token
== m
->token
)
1801 if (m
->token
== Opt_err
) {
1802 ext4_msg(sb
, KERN_ERR
, "Unrecognized mount option \"%s\" "
1803 "or missing value", opt
);
1807 if ((m
->flags
& MOPT_NO_EXT2
) && IS_EXT2_SB(sb
)) {
1808 ext4_msg(sb
, KERN_ERR
,
1809 "Mount option \"%s\" incompatible with ext2", opt
);
1812 if ((m
->flags
& MOPT_NO_EXT3
) && IS_EXT3_SB(sb
)) {
1813 ext4_msg(sb
, KERN_ERR
,
1814 "Mount option \"%s\" incompatible with ext3", opt
);
1818 if (args
->from
&& !(m
->flags
& MOPT_STRING
) && match_int(args
, &arg
))
1820 if (args
->from
&& (m
->flags
& MOPT_GTE0
) && (arg
< 0))
1822 if (m
->flags
& MOPT_EXPLICIT
) {
1823 if (m
->mount_opt
& EXT4_MOUNT_DELALLOC
) {
1824 set_opt2(sb
, EXPLICIT_DELALLOC
);
1825 } else if (m
->mount_opt
& EXT4_MOUNT_JOURNAL_CHECKSUM
) {
1826 set_opt2(sb
, EXPLICIT_JOURNAL_CHECKSUM
);
1830 if (m
->flags
& MOPT_CLEAR_ERR
)
1831 clear_opt(sb
, ERRORS_MASK
);
1832 if (token
== Opt_noquota
&& sb_any_quota_loaded(sb
)) {
1833 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
1834 "options when quota turned on");
1838 if (m
->flags
& MOPT_NOSUPPORT
) {
1839 ext4_msg(sb
, KERN_ERR
, "%s option not supported", opt
);
1840 } else if (token
== Opt_commit
) {
1842 arg
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1843 sbi
->s_commit_interval
= HZ
* arg
;
1844 } else if (token
== Opt_debug_want_extra_isize
) {
1845 sbi
->s_want_extra_isize
= arg
;
1846 } else if (token
== Opt_max_batch_time
) {
1847 sbi
->s_max_batch_time
= arg
;
1848 } else if (token
== Opt_min_batch_time
) {
1849 sbi
->s_min_batch_time
= arg
;
1850 } else if (token
== Opt_inode_readahead_blks
) {
1851 if (arg
&& (arg
> (1 << 30) || !is_power_of_2(arg
))) {
1852 ext4_msg(sb
, KERN_ERR
,
1853 "EXT4-fs: inode_readahead_blks must be "
1854 "0 or a power of 2 smaller than 2^31");
1857 sbi
->s_inode_readahead_blks
= arg
;
1858 } else if (token
== Opt_init_itable
) {
1859 set_opt(sb
, INIT_INODE_TABLE
);
1861 arg
= EXT4_DEF_LI_WAIT_MULT
;
1862 sbi
->s_li_wait_mult
= arg
;
1863 } else if (token
== Opt_max_dir_size_kb
) {
1864 sbi
->s_max_dir_size_kb
= arg
;
1865 } else if (token
== Opt_stripe
) {
1866 sbi
->s_stripe
= arg
;
1867 } else if (token
== Opt_resuid
) {
1868 uid
= make_kuid(current_user_ns(), arg
);
1869 if (!uid_valid(uid
)) {
1870 ext4_msg(sb
, KERN_ERR
, "Invalid uid value %d", arg
);
1873 sbi
->s_resuid
= uid
;
1874 } else if (token
== Opt_resgid
) {
1875 gid
= make_kgid(current_user_ns(), arg
);
1876 if (!gid_valid(gid
)) {
1877 ext4_msg(sb
, KERN_ERR
, "Invalid gid value %d", arg
);
1880 sbi
->s_resgid
= gid
;
1881 } else if (token
== Opt_journal_dev
) {
1883 ext4_msg(sb
, KERN_ERR
,
1884 "Cannot specify journal on remount");
1887 *journal_devnum
= arg
;
1888 } else if (token
== Opt_journal_path
) {
1890 struct inode
*journal_inode
;
1895 ext4_msg(sb
, KERN_ERR
,
1896 "Cannot specify journal on remount");
1899 journal_path
= match_strdup(&args
[0]);
1900 if (!journal_path
) {
1901 ext4_msg(sb
, KERN_ERR
, "error: could not dup "
1902 "journal device string");
1906 error
= kern_path(journal_path
, LOOKUP_FOLLOW
, &path
);
1908 ext4_msg(sb
, KERN_ERR
, "error: could not find "
1909 "journal device path: error %d", error
);
1910 kfree(journal_path
);
1914 journal_inode
= d_inode(path
.dentry
);
1915 if (!S_ISBLK(journal_inode
->i_mode
)) {
1916 ext4_msg(sb
, KERN_ERR
, "error: journal path %s "
1917 "is not a block device", journal_path
);
1919 kfree(journal_path
);
1923 *journal_devnum
= new_encode_dev(journal_inode
->i_rdev
);
1925 kfree(journal_path
);
1926 } else if (token
== Opt_journal_ioprio
) {
1928 ext4_msg(sb
, KERN_ERR
, "Invalid journal IO priority"
1933 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
, arg
);
1934 } else if (token
== Opt_test_dummy_encryption
) {
1935 #ifdef CONFIG_FS_ENCRYPTION
1936 sbi
->s_mount_flags
|= EXT4_MF_TEST_DUMMY_ENCRYPTION
;
1937 ext4_msg(sb
, KERN_WARNING
,
1938 "Test dummy encryption mode enabled");
1940 ext4_msg(sb
, KERN_WARNING
,
1941 "Test dummy encryption mount option ignored");
1943 } else if (m
->flags
& MOPT_DATAJ
) {
1945 if (!sbi
->s_journal
)
1946 ext4_msg(sb
, KERN_WARNING
, "Remounting file system with no journal so ignoring journalled data option");
1947 else if (test_opt(sb
, DATA_FLAGS
) != m
->mount_opt
) {
1948 ext4_msg(sb
, KERN_ERR
,
1949 "Cannot change data mode on remount");
1953 clear_opt(sb
, DATA_FLAGS
);
1954 sbi
->s_mount_opt
|= m
->mount_opt
;
1957 } else if (m
->flags
& MOPT_QFMT
) {
1958 if (sb_any_quota_loaded(sb
) &&
1959 sbi
->s_jquota_fmt
!= m
->mount_opt
) {
1960 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled "
1961 "quota options when quota turned on");
1964 if (ext4_has_feature_quota(sb
)) {
1965 ext4_msg(sb
, KERN_INFO
,
1966 "Quota format mount options ignored "
1967 "when QUOTA feature is enabled");
1970 sbi
->s_jquota_fmt
= m
->mount_opt
;
1972 } else if (token
== Opt_dax
) {
1973 #ifdef CONFIG_FS_DAX
1974 ext4_msg(sb
, KERN_WARNING
,
1975 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1976 sbi
->s_mount_opt
|= m
->mount_opt
;
1978 ext4_msg(sb
, KERN_INFO
, "dax option not supported");
1981 } else if (token
== Opt_data_err_abort
) {
1982 sbi
->s_mount_opt
|= m
->mount_opt
;
1983 } else if (token
== Opt_data_err_ignore
) {
1984 sbi
->s_mount_opt
&= ~m
->mount_opt
;
1988 if (m
->flags
& MOPT_CLEAR
)
1990 else if (unlikely(!(m
->flags
& MOPT_SET
))) {
1991 ext4_msg(sb
, KERN_WARNING
,
1992 "buggy handling of option %s", opt
);
1997 sbi
->s_mount_opt
|= m
->mount_opt
;
1999 sbi
->s_mount_opt
&= ~m
->mount_opt
;
2004 static int parse_options(char *options
, struct super_block
*sb
,
2005 unsigned long *journal_devnum
,
2006 unsigned int *journal_ioprio
,
2009 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2010 char *p
, __maybe_unused
*usr_qf_name
, __maybe_unused
*grp_qf_name
;
2011 substring_t args
[MAX_OPT_ARGS
];
2017 while ((p
= strsep(&options
, ",")) != NULL
) {
2021 * Initialize args struct so we know whether arg was
2022 * found; some options take optional arguments.
2024 args
[0].to
= args
[0].from
= NULL
;
2025 token
= match_token(p
, tokens
, args
);
2026 if (handle_mount_opt(sb
, p
, token
, args
, journal_devnum
,
2027 journal_ioprio
, is_remount
) < 0)
2032 * We do the test below only for project quotas. 'usrquota' and
2033 * 'grpquota' mount options are allowed even without quota feature
2034 * to support legacy quotas in quota files.
2036 if (test_opt(sb
, PRJQUOTA
) && !ext4_has_feature_project(sb
)) {
2037 ext4_msg(sb
, KERN_ERR
, "Project quota feature not enabled. "
2038 "Cannot enable project quota enforcement.");
2041 usr_qf_name
= get_qf_name(sb
, sbi
, USRQUOTA
);
2042 grp_qf_name
= get_qf_name(sb
, sbi
, GRPQUOTA
);
2043 if (usr_qf_name
|| grp_qf_name
) {
2044 if (test_opt(sb
, USRQUOTA
) && usr_qf_name
)
2045 clear_opt(sb
, USRQUOTA
);
2047 if (test_opt(sb
, GRPQUOTA
) && grp_qf_name
)
2048 clear_opt(sb
, GRPQUOTA
);
2050 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
2051 ext4_msg(sb
, KERN_ERR
, "old and new quota "
2056 if (!sbi
->s_jquota_fmt
) {
2057 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
2063 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
2065 BLOCK_SIZE
<< le32_to_cpu(sbi
->s_es
->s_log_block_size
);
2067 if (blocksize
< PAGE_SIZE
) {
2068 ext4_msg(sb
, KERN_ERR
, "can't mount with "
2069 "dioread_nolock if block size != PAGE_SIZE");
2076 static inline void ext4_show_quota_options(struct seq_file
*seq
,
2077 struct super_block
*sb
)
2079 #if defined(CONFIG_QUOTA)
2080 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2081 char *usr_qf_name
, *grp_qf_name
;
2083 if (sbi
->s_jquota_fmt
) {
2086 switch (sbi
->s_jquota_fmt
) {
2097 seq_printf(seq
, ",jqfmt=%s", fmtname
);
2101 usr_qf_name
= rcu_dereference(sbi
->s_qf_names
[USRQUOTA
]);
2102 grp_qf_name
= rcu_dereference(sbi
->s_qf_names
[GRPQUOTA
]);
2104 seq_show_option(seq
, "usrjquota", usr_qf_name
);
2106 seq_show_option(seq
, "grpjquota", grp_qf_name
);
2111 static const char *token2str(int token
)
2113 const struct match_token
*t
;
2115 for (t
= tokens
; t
->token
!= Opt_err
; t
++)
2116 if (t
->token
== token
&& !strchr(t
->pattern
, '='))
2123 * - it's set to a non-default value OR
2124 * - if the per-sb default is different from the global default
2126 static int _ext4_show_options(struct seq_file
*seq
, struct super_block
*sb
,
2129 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2130 struct ext4_super_block
*es
= sbi
->s_es
;
2131 int def_errors
, def_mount_opt
= sbi
->s_def_mount_opt
;
2132 const struct mount_opts
*m
;
2133 char sep
= nodefs
? '\n' : ',';
2135 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2136 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2138 if (sbi
->s_sb_block
!= 1)
2139 SEQ_OPTS_PRINT("sb=%llu", sbi
->s_sb_block
);
2141 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++) {
2142 int want_set
= m
->flags
& MOPT_SET
;
2143 if (((m
->flags
& (MOPT_SET
|MOPT_CLEAR
)) == 0) ||
2144 (m
->flags
& MOPT_CLEAR_ERR
))
2146 if (!nodefs
&& !(m
->mount_opt
& (sbi
->s_mount_opt
^ def_mount_opt
)))
2147 continue; /* skip if same as the default */
2149 (sbi
->s_mount_opt
& m
->mount_opt
) != m
->mount_opt
) ||
2150 (!want_set
&& (sbi
->s_mount_opt
& m
->mount_opt
)))
2151 continue; /* select Opt_noFoo vs Opt_Foo */
2152 SEQ_OPTS_PRINT("%s", token2str(m
->token
));
2155 if (nodefs
|| !uid_eq(sbi
->s_resuid
, make_kuid(&init_user_ns
, EXT4_DEF_RESUID
)) ||
2156 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
)
2157 SEQ_OPTS_PRINT("resuid=%u",
2158 from_kuid_munged(&init_user_ns
, sbi
->s_resuid
));
2159 if (nodefs
|| !gid_eq(sbi
->s_resgid
, make_kgid(&init_user_ns
, EXT4_DEF_RESGID
)) ||
2160 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
)
2161 SEQ_OPTS_PRINT("resgid=%u",
2162 from_kgid_munged(&init_user_ns
, sbi
->s_resgid
));
2163 def_errors
= nodefs
? -1 : le16_to_cpu(es
->s_errors
);
2164 if (test_opt(sb
, ERRORS_RO
) && def_errors
!= EXT4_ERRORS_RO
)
2165 SEQ_OPTS_PUTS("errors=remount-ro");
2166 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
2167 SEQ_OPTS_PUTS("errors=continue");
2168 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
2169 SEQ_OPTS_PUTS("errors=panic");
2170 if (nodefs
|| sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
)
2171 SEQ_OPTS_PRINT("commit=%lu", sbi
->s_commit_interval
/ HZ
);
2172 if (nodefs
|| sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
)
2173 SEQ_OPTS_PRINT("min_batch_time=%u", sbi
->s_min_batch_time
);
2174 if (nodefs
|| sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
)
2175 SEQ_OPTS_PRINT("max_batch_time=%u", sbi
->s_max_batch_time
);
2176 if (sb
->s_flags
& SB_I_VERSION
)
2177 SEQ_OPTS_PUTS("i_version");
2178 if (nodefs
|| sbi
->s_stripe
)
2179 SEQ_OPTS_PRINT("stripe=%lu", sbi
->s_stripe
);
2180 if (nodefs
|| EXT4_MOUNT_DATA_FLAGS
&
2181 (sbi
->s_mount_opt
^ def_mount_opt
)) {
2182 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
2183 SEQ_OPTS_PUTS("data=journal");
2184 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
2185 SEQ_OPTS_PUTS("data=ordered");
2186 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
2187 SEQ_OPTS_PUTS("data=writeback");
2190 sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
2191 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2192 sbi
->s_inode_readahead_blks
);
2194 if (test_opt(sb
, INIT_INODE_TABLE
) && (nodefs
||
2195 (sbi
->s_li_wait_mult
!= EXT4_DEF_LI_WAIT_MULT
)))
2196 SEQ_OPTS_PRINT("init_itable=%u", sbi
->s_li_wait_mult
);
2197 if (nodefs
|| sbi
->s_max_dir_size_kb
)
2198 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi
->s_max_dir_size_kb
);
2199 if (test_opt(sb
, DATA_ERR_ABORT
))
2200 SEQ_OPTS_PUTS("data_err=abort");
2201 if (DUMMY_ENCRYPTION_ENABLED(sbi
))
2202 SEQ_OPTS_PUTS("test_dummy_encryption");
2204 ext4_show_quota_options(seq
, sb
);
2208 static int ext4_show_options(struct seq_file
*seq
, struct dentry
*root
)
2210 return _ext4_show_options(seq
, root
->d_sb
, 0);
2213 int ext4_seq_options_show(struct seq_file
*seq
, void *offset
)
2215 struct super_block
*sb
= seq
->private;
2218 seq_puts(seq
, sb_rdonly(sb
) ? "ro" : "rw");
2219 rc
= _ext4_show_options(seq
, sb
, 1);
2220 seq_puts(seq
, "\n");
2224 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
2227 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2230 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
2231 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
2232 "forcing read-only mode");
2237 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
2238 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
2239 "running e2fsck is recommended");
2240 else if (sbi
->s_mount_state
& EXT4_ERROR_FS
)
2241 ext4_msg(sb
, KERN_WARNING
,
2242 "warning: mounting fs with errors, "
2243 "running e2fsck is recommended");
2244 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) > 0 &&
2245 le16_to_cpu(es
->s_mnt_count
) >=
2246 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
2247 ext4_msg(sb
, KERN_WARNING
,
2248 "warning: maximal mount count reached, "
2249 "running e2fsck is recommended");
2250 else if (le32_to_cpu(es
->s_checkinterval
) &&
2251 (ext4_get_tstamp(es
, s_lastcheck
) +
2252 le32_to_cpu(es
->s_checkinterval
) <= ktime_get_real_seconds()))
2253 ext4_msg(sb
, KERN_WARNING
,
2254 "warning: checktime reached, "
2255 "running e2fsck is recommended");
2256 if (!sbi
->s_journal
)
2257 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
2258 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
2259 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
2260 le16_add_cpu(&es
->s_mnt_count
, 1);
2261 ext4_update_tstamp(es
, s_mtime
);
2263 ext4_set_feature_journal_needs_recovery(sb
);
2265 err
= ext4_commit_super(sb
, 1);
2267 if (test_opt(sb
, DEBUG
))
2268 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
2269 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2271 sbi
->s_groups_count
,
2272 EXT4_BLOCKS_PER_GROUP(sb
),
2273 EXT4_INODES_PER_GROUP(sb
),
2274 sbi
->s_mount_opt
, sbi
->s_mount_opt2
);
2276 cleancache_init_fs(sb
);
2280 int ext4_alloc_flex_bg_array(struct super_block
*sb
, ext4_group_t ngroup
)
2282 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2283 struct flex_groups
*new_groups
;
2286 if (!sbi
->s_log_groups_per_flex
)
2289 size
= ext4_flex_group(sbi
, ngroup
- 1) + 1;
2290 if (size
<= sbi
->s_flex_groups_allocated
)
2293 size
= roundup_pow_of_two(size
* sizeof(struct flex_groups
));
2294 new_groups
= kvzalloc(size
, GFP_KERNEL
);
2296 ext4_msg(sb
, KERN_ERR
, "not enough memory for %d flex groups",
2297 size
/ (int) sizeof(struct flex_groups
));
2301 if (sbi
->s_flex_groups
) {
2302 memcpy(new_groups
, sbi
->s_flex_groups
,
2303 (sbi
->s_flex_groups_allocated
*
2304 sizeof(struct flex_groups
)));
2305 kvfree(sbi
->s_flex_groups
);
2307 sbi
->s_flex_groups
= new_groups
;
2308 sbi
->s_flex_groups_allocated
= size
/ sizeof(struct flex_groups
);
2312 static int ext4_fill_flex_info(struct super_block
*sb
)
2314 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2315 struct ext4_group_desc
*gdp
= NULL
;
2316 ext4_group_t flex_group
;
2319 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
2320 if (sbi
->s_log_groups_per_flex
< 1 || sbi
->s_log_groups_per_flex
> 31) {
2321 sbi
->s_log_groups_per_flex
= 0;
2325 err
= ext4_alloc_flex_bg_array(sb
, sbi
->s_groups_count
);
2329 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2330 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2332 flex_group
= ext4_flex_group(sbi
, i
);
2333 atomic_add(ext4_free_inodes_count(sb
, gdp
),
2334 &sbi
->s_flex_groups
[flex_group
].free_inodes
);
2335 atomic64_add(ext4_free_group_clusters(sb
, gdp
),
2336 &sbi
->s_flex_groups
[flex_group
].free_clusters
);
2337 atomic_add(ext4_used_dirs_count(sb
, gdp
),
2338 &sbi
->s_flex_groups
[flex_group
].used_dirs
);
2346 static __le16
ext4_group_desc_csum(struct super_block
*sb
, __u32 block_group
,
2347 struct ext4_group_desc
*gdp
)
2349 int offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
2351 __le32 le_group
= cpu_to_le32(block_group
);
2352 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2354 if (ext4_has_metadata_csum(sbi
->s_sb
)) {
2355 /* Use new metadata_csum algorithm */
2357 __u16 dummy_csum
= 0;
2359 csum32
= ext4_chksum(sbi
, sbi
->s_csum_seed
, (__u8
*)&le_group
,
2361 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)gdp
, offset
);
2362 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)&dummy_csum
,
2363 sizeof(dummy_csum
));
2364 offset
+= sizeof(dummy_csum
);
2365 if (offset
< sbi
->s_desc_size
)
2366 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)gdp
+ offset
,
2367 sbi
->s_desc_size
- offset
);
2369 crc
= csum32
& 0xFFFF;
2373 /* old crc16 code */
2374 if (!ext4_has_feature_gdt_csum(sb
))
2377 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
2378 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
2379 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
2380 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
2381 /* for checksum of struct ext4_group_desc do the rest...*/
2382 if (ext4_has_feature_64bit(sb
) &&
2383 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
2384 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
2385 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
2389 return cpu_to_le16(crc
);
2392 int ext4_group_desc_csum_verify(struct super_block
*sb
, __u32 block_group
,
2393 struct ext4_group_desc
*gdp
)
2395 if (ext4_has_group_desc_csum(sb
) &&
2396 (gdp
->bg_checksum
!= ext4_group_desc_csum(sb
, block_group
, gdp
)))
2402 void ext4_group_desc_csum_set(struct super_block
*sb
, __u32 block_group
,
2403 struct ext4_group_desc
*gdp
)
2405 if (!ext4_has_group_desc_csum(sb
))
2407 gdp
->bg_checksum
= ext4_group_desc_csum(sb
, block_group
, gdp
);
2410 /* Called at mount-time, super-block is locked */
2411 static int ext4_check_descriptors(struct super_block
*sb
,
2412 ext4_fsblk_t sb_block
,
2413 ext4_group_t
*first_not_zeroed
)
2415 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2416 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
2417 ext4_fsblk_t last_block
;
2418 ext4_fsblk_t last_bg_block
= sb_block
+ ext4_bg_num_gdb(sb
, 0);
2419 ext4_fsblk_t block_bitmap
;
2420 ext4_fsblk_t inode_bitmap
;
2421 ext4_fsblk_t inode_table
;
2422 int flexbg_flag
= 0;
2423 ext4_group_t i
, grp
= sbi
->s_groups_count
;
2425 if (ext4_has_feature_flex_bg(sb
))
2428 ext4_debug("Checking group descriptors");
2430 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2431 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2433 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
2434 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
2436 last_block
= first_block
+
2437 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2439 if ((grp
== sbi
->s_groups_count
) &&
2440 !(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2443 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
2444 if (block_bitmap
== sb_block
) {
2445 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2446 "Block bitmap for group %u overlaps "
2451 if (block_bitmap
>= sb_block
+ 1 &&
2452 block_bitmap
<= last_bg_block
) {
2453 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2454 "Block bitmap for group %u overlaps "
2455 "block group descriptors", i
);
2459 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
2460 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2461 "Block bitmap for group %u not in group "
2462 "(block %llu)!", i
, block_bitmap
);
2465 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
2466 if (inode_bitmap
== sb_block
) {
2467 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2468 "Inode bitmap for group %u overlaps "
2473 if (inode_bitmap
>= sb_block
+ 1 &&
2474 inode_bitmap
<= last_bg_block
) {
2475 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2476 "Inode bitmap for group %u overlaps "
2477 "block group descriptors", i
);
2481 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
2482 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2483 "Inode bitmap for group %u not in group "
2484 "(block %llu)!", i
, inode_bitmap
);
2487 inode_table
= ext4_inode_table(sb
, gdp
);
2488 if (inode_table
== sb_block
) {
2489 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2490 "Inode table for group %u overlaps "
2495 if (inode_table
>= sb_block
+ 1 &&
2496 inode_table
<= last_bg_block
) {
2497 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2498 "Inode table for group %u overlaps "
2499 "block group descriptors", i
);
2503 if (inode_table
< first_block
||
2504 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
2505 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2506 "Inode table for group %u not in group "
2507 "(block %llu)!", i
, inode_table
);
2510 ext4_lock_group(sb
, i
);
2511 if (!ext4_group_desc_csum_verify(sb
, i
, gdp
)) {
2512 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2513 "Checksum for group %u failed (%u!=%u)",
2514 i
, le16_to_cpu(ext4_group_desc_csum(sb
, i
,
2515 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
2516 if (!sb_rdonly(sb
)) {
2517 ext4_unlock_group(sb
, i
);
2521 ext4_unlock_group(sb
, i
);
2523 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
2525 if (NULL
!= first_not_zeroed
)
2526 *first_not_zeroed
= grp
;
2530 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2531 * the superblock) which were deleted from all directories, but held open by
2532 * a process at the time of a crash. We walk the list and try to delete these
2533 * inodes at recovery time (only with a read-write filesystem).
2535 * In order to keep the orphan inode chain consistent during traversal (in
2536 * case of crash during recovery), we link each inode into the superblock
2537 * orphan list_head and handle it the same way as an inode deletion during
2538 * normal operation (which journals the operations for us).
2540 * We only do an iget() and an iput() on each inode, which is very safe if we
2541 * accidentally point at an in-use or already deleted inode. The worst that
2542 * can happen in this case is that we get a "bit already cleared" message from
2543 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2544 * e2fsck was run on this filesystem, and it must have already done the orphan
2545 * inode cleanup for us, so we can safely abort without any further action.
2547 static void ext4_orphan_cleanup(struct super_block
*sb
,
2548 struct ext4_super_block
*es
)
2550 unsigned int s_flags
= sb
->s_flags
;
2551 int ret
, nr_orphans
= 0, nr_truncates
= 0;
2553 int quota_update
= 0;
2556 if (!es
->s_last_orphan
) {
2557 jbd_debug(4, "no orphan inodes to clean up\n");
2561 if (bdev_read_only(sb
->s_bdev
)) {
2562 ext4_msg(sb
, KERN_ERR
, "write access "
2563 "unavailable, skipping orphan cleanup");
2567 /* Check if feature set would not allow a r/w mount */
2568 if (!ext4_feature_set_ok(sb
, 0)) {
2569 ext4_msg(sb
, KERN_INFO
, "Skipping orphan cleanup due to "
2570 "unknown ROCOMPAT features");
2574 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2575 /* don't clear list on RO mount w/ errors */
2576 if (es
->s_last_orphan
&& !(s_flags
& SB_RDONLY
)) {
2577 ext4_msg(sb
, KERN_INFO
, "Errors on filesystem, "
2578 "clearing orphan list.\n");
2579 es
->s_last_orphan
= 0;
2581 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2585 if (s_flags
& SB_RDONLY
) {
2586 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
2587 sb
->s_flags
&= ~SB_RDONLY
;
2590 /* Needed for iput() to work correctly and not trash data */
2591 sb
->s_flags
|= SB_ACTIVE
;
2594 * Turn on quotas which were not enabled for read-only mounts if
2595 * filesystem has quota feature, so that they are updated correctly.
2597 if (ext4_has_feature_quota(sb
) && (s_flags
& SB_RDONLY
)) {
2598 int ret
= ext4_enable_quotas(sb
);
2603 ext4_msg(sb
, KERN_ERR
,
2604 "Cannot turn on quotas: error %d", ret
);
2607 /* Turn on journaled quotas used for old sytle */
2608 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
2609 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
2610 int ret
= ext4_quota_on_mount(sb
, i
);
2615 ext4_msg(sb
, KERN_ERR
,
2616 "Cannot turn on journaled "
2617 "quota: type %d: error %d", i
, ret
);
2622 while (es
->s_last_orphan
) {
2623 struct inode
*inode
;
2626 * We may have encountered an error during cleanup; if
2627 * so, skip the rest.
2629 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2630 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2631 es
->s_last_orphan
= 0;
2635 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
2636 if (IS_ERR(inode
)) {
2637 es
->s_last_orphan
= 0;
2641 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
2642 dquot_initialize(inode
);
2643 if (inode
->i_nlink
) {
2644 if (test_opt(sb
, DEBUG
))
2645 ext4_msg(sb
, KERN_DEBUG
,
2646 "%s: truncating inode %lu to %lld bytes",
2647 __func__
, inode
->i_ino
, inode
->i_size
);
2648 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2649 inode
->i_ino
, inode
->i_size
);
2651 truncate_inode_pages(inode
->i_mapping
, inode
->i_size
);
2652 ret
= ext4_truncate(inode
);
2654 ext4_std_error(inode
->i_sb
, ret
);
2655 inode_unlock(inode
);
2658 if (test_opt(sb
, DEBUG
))
2659 ext4_msg(sb
, KERN_DEBUG
,
2660 "%s: deleting unreferenced inode %lu",
2661 __func__
, inode
->i_ino
);
2662 jbd_debug(2, "deleting unreferenced inode %lu\n",
2666 iput(inode
); /* The delete magic happens here! */
2669 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2672 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
2673 PLURAL(nr_orphans
));
2675 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
2676 PLURAL(nr_truncates
));
2678 /* Turn off quotas if they were enabled for orphan cleanup */
2680 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
2681 if (sb_dqopt(sb
)->files
[i
])
2682 dquot_quota_off(sb
, i
);
2686 sb
->s_flags
= s_flags
; /* Restore SB_RDONLY status */
2690 * Maximal extent format file size.
2691 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2692 * extent format containers, within a sector_t, and within i_blocks
2693 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2694 * so that won't be a limiting factor.
2696 * However there is other limiting factor. We do store extents in the form
2697 * of starting block and length, hence the resulting length of the extent
2698 * covering maximum file size must fit into on-disk format containers as
2699 * well. Given that length is always by 1 unit bigger than max unit (because
2700 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2702 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2704 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
2707 loff_t upper_limit
= MAX_LFS_FILESIZE
;
2709 /* small i_blocks in vfs inode? */
2710 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2712 * CONFIG_LBDAF is not enabled implies the inode
2713 * i_block represent total blocks in 512 bytes
2714 * 32 == size of vfs inode i_blocks * 8
2716 upper_limit
= (1LL << 32) - 1;
2718 /* total blocks in file system block size */
2719 upper_limit
>>= (blkbits
- 9);
2720 upper_limit
<<= blkbits
;
2724 * 32-bit extent-start container, ee_block. We lower the maxbytes
2725 * by one fs block, so ee_len can cover the extent of maximum file
2728 res
= (1LL << 32) - 1;
2731 /* Sanity check against vm- & vfs- imposed limits */
2732 if (res
> upper_limit
)
2739 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2740 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2741 * We need to be 1 filesystem block less than the 2^48 sector limit.
2743 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
2745 loff_t res
= EXT4_NDIR_BLOCKS
;
2748 /* This is calculated to be the largest file size for a dense, block
2749 * mapped file such that the file's total number of 512-byte sectors,
2750 * including data and all indirect blocks, does not exceed (2^48 - 1).
2752 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2753 * number of 512-byte sectors of the file.
2756 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2758 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2759 * the inode i_block field represents total file blocks in
2760 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2762 upper_limit
= (1LL << 32) - 1;
2764 /* total blocks in file system block size */
2765 upper_limit
>>= (bits
- 9);
2769 * We use 48 bit ext4_inode i_blocks
2770 * With EXT4_HUGE_FILE_FL set the i_blocks
2771 * represent total number of blocks in
2772 * file system block size
2774 upper_limit
= (1LL << 48) - 1;
2778 /* indirect blocks */
2780 /* double indirect blocks */
2781 meta_blocks
+= 1 + (1LL << (bits
-2));
2782 /* tripple indirect blocks */
2783 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
2785 upper_limit
-= meta_blocks
;
2786 upper_limit
<<= bits
;
2788 res
+= 1LL << (bits
-2);
2789 res
+= 1LL << (2*(bits
-2));
2790 res
+= 1LL << (3*(bits
-2));
2792 if (res
> upper_limit
)
2795 if (res
> MAX_LFS_FILESIZE
)
2796 res
= MAX_LFS_FILESIZE
;
2801 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
2802 ext4_fsblk_t logical_sb_block
, int nr
)
2804 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2805 ext4_group_t bg
, first_meta_bg
;
2808 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
2810 if (!ext4_has_feature_meta_bg(sb
) || nr
< first_meta_bg
)
2811 return logical_sb_block
+ nr
+ 1;
2812 bg
= sbi
->s_desc_per_block
* nr
;
2813 if (ext4_bg_has_super(sb
, bg
))
2817 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2818 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2819 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2822 if (sb
->s_blocksize
== 1024 && nr
== 0 &&
2823 le32_to_cpu(sbi
->s_es
->s_first_data_block
) == 0)
2826 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
2830 * ext4_get_stripe_size: Get the stripe size.
2831 * @sbi: In memory super block info
2833 * If we have specified it via mount option, then
2834 * use the mount option value. If the value specified at mount time is
2835 * greater than the blocks per group use the super block value.
2836 * If the super block value is greater than blocks per group return 0.
2837 * Allocator needs it be less than blocks per group.
2840 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
2842 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
2843 unsigned long stripe_width
=
2844 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
2847 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
2848 ret
= sbi
->s_stripe
;
2849 else if (stripe_width
&& stripe_width
<= sbi
->s_blocks_per_group
)
2851 else if (stride
&& stride
<= sbi
->s_blocks_per_group
)
2857 * If the stripe width is 1, this makes no sense and
2858 * we set it to 0 to turn off stripe handling code.
2867 * Check whether this filesystem can be mounted based on
2868 * the features present and the RDONLY/RDWR mount requested.
2869 * Returns 1 if this filesystem can be mounted as requested,
2870 * 0 if it cannot be.
2872 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
2874 if (ext4_has_unknown_ext4_incompat_features(sb
)) {
2875 ext4_msg(sb
, KERN_ERR
,
2876 "Couldn't mount because of "
2877 "unsupported optional features (%x)",
2878 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2879 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2886 if (ext4_has_feature_readonly(sb
)) {
2887 ext4_msg(sb
, KERN_INFO
, "filesystem is read-only");
2888 sb
->s_flags
|= SB_RDONLY
;
2892 /* Check that feature set is OK for a read-write mount */
2893 if (ext4_has_unknown_ext4_ro_compat_features(sb
)) {
2894 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
2895 "unsupported optional features (%x)",
2896 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2897 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2901 * Large file size enabled file system can only be mounted
2902 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2904 if (ext4_has_feature_huge_file(sb
)) {
2905 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
2906 ext4_msg(sb
, KERN_ERR
, "Filesystem with huge files "
2907 "cannot be mounted RDWR without "
2912 if (ext4_has_feature_bigalloc(sb
) && !ext4_has_feature_extents(sb
)) {
2913 ext4_msg(sb
, KERN_ERR
,
2914 "Can't support bigalloc feature without "
2915 "extents feature\n");
2919 #ifndef CONFIG_QUOTA
2920 if (ext4_has_feature_quota(sb
) && !readonly
) {
2921 ext4_msg(sb
, KERN_ERR
,
2922 "Filesystem with quota feature cannot be mounted RDWR "
2923 "without CONFIG_QUOTA");
2926 if (ext4_has_feature_project(sb
) && !readonly
) {
2927 ext4_msg(sb
, KERN_ERR
,
2928 "Filesystem with project quota feature cannot be mounted RDWR "
2929 "without CONFIG_QUOTA");
2932 #endif /* CONFIG_QUOTA */
2937 * This function is called once a day if we have errors logged
2938 * on the file system
2940 static void print_daily_error_info(struct timer_list
*t
)
2942 struct ext4_sb_info
*sbi
= from_timer(sbi
, t
, s_err_report
);
2943 struct super_block
*sb
= sbi
->s_sb
;
2944 struct ext4_super_block
*es
= sbi
->s_es
;
2946 if (es
->s_error_count
)
2947 /* fsck newer than v1.41.13 is needed to clean this condition. */
2948 ext4_msg(sb
, KERN_NOTICE
, "error count since last fsck: %u",
2949 le32_to_cpu(es
->s_error_count
));
2950 if (es
->s_first_error_time
) {
2951 printk(KERN_NOTICE
"EXT4-fs (%s): initial error at time %llu: %.*s:%d",
2953 ext4_get_tstamp(es
, s_first_error_time
),
2954 (int) sizeof(es
->s_first_error_func
),
2955 es
->s_first_error_func
,
2956 le32_to_cpu(es
->s_first_error_line
));
2957 if (es
->s_first_error_ino
)
2958 printk(KERN_CONT
": inode %u",
2959 le32_to_cpu(es
->s_first_error_ino
));
2960 if (es
->s_first_error_block
)
2961 printk(KERN_CONT
": block %llu", (unsigned long long)
2962 le64_to_cpu(es
->s_first_error_block
));
2963 printk(KERN_CONT
"\n");
2965 if (es
->s_last_error_time
) {
2966 printk(KERN_NOTICE
"EXT4-fs (%s): last error at time %llu: %.*s:%d",
2968 ext4_get_tstamp(es
, s_last_error_time
),
2969 (int) sizeof(es
->s_last_error_func
),
2970 es
->s_last_error_func
,
2971 le32_to_cpu(es
->s_last_error_line
));
2972 if (es
->s_last_error_ino
)
2973 printk(KERN_CONT
": inode %u",
2974 le32_to_cpu(es
->s_last_error_ino
));
2975 if (es
->s_last_error_block
)
2976 printk(KERN_CONT
": block %llu", (unsigned long long)
2977 le64_to_cpu(es
->s_last_error_block
));
2978 printk(KERN_CONT
"\n");
2980 mod_timer(&sbi
->s_err_report
, jiffies
+ 24*60*60*HZ
); /* Once a day */
2983 /* Find next suitable group and run ext4_init_inode_table */
2984 static int ext4_run_li_request(struct ext4_li_request
*elr
)
2986 struct ext4_group_desc
*gdp
= NULL
;
2987 ext4_group_t group
, ngroups
;
2988 struct super_block
*sb
;
2989 unsigned long timeout
= 0;
2993 ngroups
= EXT4_SB(sb
)->s_groups_count
;
2995 for (group
= elr
->lr_next_group
; group
< ngroups
; group
++) {
2996 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
3002 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
3006 if (group
>= ngroups
)
3011 ret
= ext4_init_inode_table(sb
, group
,
3012 elr
->lr_timeout
? 0 : 1);
3013 if (elr
->lr_timeout
== 0) {
3014 timeout
= (jiffies
- timeout
) *
3015 elr
->lr_sbi
->s_li_wait_mult
;
3016 elr
->lr_timeout
= timeout
;
3018 elr
->lr_next_sched
= jiffies
+ elr
->lr_timeout
;
3019 elr
->lr_next_group
= group
+ 1;
3025 * Remove lr_request from the list_request and free the
3026 * request structure. Should be called with li_list_mtx held
3028 static void ext4_remove_li_request(struct ext4_li_request
*elr
)
3030 struct ext4_sb_info
*sbi
;
3037 list_del(&elr
->lr_request
);
3038 sbi
->s_li_request
= NULL
;
3042 static void ext4_unregister_li_request(struct super_block
*sb
)
3044 mutex_lock(&ext4_li_mtx
);
3045 if (!ext4_li_info
) {
3046 mutex_unlock(&ext4_li_mtx
);
3050 mutex_lock(&ext4_li_info
->li_list_mtx
);
3051 ext4_remove_li_request(EXT4_SB(sb
)->s_li_request
);
3052 mutex_unlock(&ext4_li_info
->li_list_mtx
);
3053 mutex_unlock(&ext4_li_mtx
);
3056 static struct task_struct
*ext4_lazyinit_task
;
3059 * This is the function where ext4lazyinit thread lives. It walks
3060 * through the request list searching for next scheduled filesystem.
3061 * When such a fs is found, run the lazy initialization request
3062 * (ext4_rn_li_request) and keep track of the time spend in this
3063 * function. Based on that time we compute next schedule time of
3064 * the request. When walking through the list is complete, compute
3065 * next waking time and put itself into sleep.
3067 static int ext4_lazyinit_thread(void *arg
)
3069 struct ext4_lazy_init
*eli
= (struct ext4_lazy_init
*)arg
;
3070 struct list_head
*pos
, *n
;
3071 struct ext4_li_request
*elr
;
3072 unsigned long next_wakeup
, cur
;
3074 BUG_ON(NULL
== eli
);
3078 next_wakeup
= MAX_JIFFY_OFFSET
;
3080 mutex_lock(&eli
->li_list_mtx
);
3081 if (list_empty(&eli
->li_request_list
)) {
3082 mutex_unlock(&eli
->li_list_mtx
);
3085 list_for_each_safe(pos
, n
, &eli
->li_request_list
) {
3088 elr
= list_entry(pos
, struct ext4_li_request
,
3091 if (time_before(jiffies
, elr
->lr_next_sched
)) {
3092 if (time_before(elr
->lr_next_sched
, next_wakeup
))
3093 next_wakeup
= elr
->lr_next_sched
;
3096 if (down_read_trylock(&elr
->lr_super
->s_umount
)) {
3097 if (sb_start_write_trylock(elr
->lr_super
)) {
3100 * We hold sb->s_umount, sb can not
3101 * be removed from the list, it is
3102 * now safe to drop li_list_mtx
3104 mutex_unlock(&eli
->li_list_mtx
);
3105 err
= ext4_run_li_request(elr
);
3106 sb_end_write(elr
->lr_super
);
3107 mutex_lock(&eli
->li_list_mtx
);
3110 up_read((&elr
->lr_super
->s_umount
));
3112 /* error, remove the lazy_init job */
3114 ext4_remove_li_request(elr
);
3118 elr
->lr_next_sched
= jiffies
+
3120 % (EXT4_DEF_LI_MAX_START_DELAY
* HZ
));
3122 if (time_before(elr
->lr_next_sched
, next_wakeup
))
3123 next_wakeup
= elr
->lr_next_sched
;
3125 mutex_unlock(&eli
->li_list_mtx
);
3130 if ((time_after_eq(cur
, next_wakeup
)) ||
3131 (MAX_JIFFY_OFFSET
== next_wakeup
)) {
3136 schedule_timeout_interruptible(next_wakeup
- cur
);
3138 if (kthread_should_stop()) {
3139 ext4_clear_request_list();
3146 * It looks like the request list is empty, but we need
3147 * to check it under the li_list_mtx lock, to prevent any
3148 * additions into it, and of course we should lock ext4_li_mtx
3149 * to atomically free the list and ext4_li_info, because at
3150 * this point another ext4 filesystem could be registering
3153 mutex_lock(&ext4_li_mtx
);
3154 mutex_lock(&eli
->li_list_mtx
);
3155 if (!list_empty(&eli
->li_request_list
)) {
3156 mutex_unlock(&eli
->li_list_mtx
);
3157 mutex_unlock(&ext4_li_mtx
);
3160 mutex_unlock(&eli
->li_list_mtx
);
3161 kfree(ext4_li_info
);
3162 ext4_li_info
= NULL
;
3163 mutex_unlock(&ext4_li_mtx
);
3168 static void ext4_clear_request_list(void)
3170 struct list_head
*pos
, *n
;
3171 struct ext4_li_request
*elr
;
3173 mutex_lock(&ext4_li_info
->li_list_mtx
);
3174 list_for_each_safe(pos
, n
, &ext4_li_info
->li_request_list
) {
3175 elr
= list_entry(pos
, struct ext4_li_request
,
3177 ext4_remove_li_request(elr
);
3179 mutex_unlock(&ext4_li_info
->li_list_mtx
);
3182 static int ext4_run_lazyinit_thread(void)
3184 ext4_lazyinit_task
= kthread_run(ext4_lazyinit_thread
,
3185 ext4_li_info
, "ext4lazyinit");
3186 if (IS_ERR(ext4_lazyinit_task
)) {
3187 int err
= PTR_ERR(ext4_lazyinit_task
);
3188 ext4_clear_request_list();
3189 kfree(ext4_li_info
);
3190 ext4_li_info
= NULL
;
3191 printk(KERN_CRIT
"EXT4-fs: error %d creating inode table "
3192 "initialization thread\n",
3196 ext4_li_info
->li_state
|= EXT4_LAZYINIT_RUNNING
;
3201 * Check whether it make sense to run itable init. thread or not.
3202 * If there is at least one uninitialized inode table, return
3203 * corresponding group number, else the loop goes through all
3204 * groups and return total number of groups.
3206 static ext4_group_t
ext4_has_uninit_itable(struct super_block
*sb
)
3208 ext4_group_t group
, ngroups
= EXT4_SB(sb
)->s_groups_count
;
3209 struct ext4_group_desc
*gdp
= NULL
;
3211 if (!ext4_has_group_desc_csum(sb
))
3214 for (group
= 0; group
< ngroups
; group
++) {
3215 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
3219 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
3226 static int ext4_li_info_new(void)
3228 struct ext4_lazy_init
*eli
= NULL
;
3230 eli
= kzalloc(sizeof(*eli
), GFP_KERNEL
);
3234 INIT_LIST_HEAD(&eli
->li_request_list
);
3235 mutex_init(&eli
->li_list_mtx
);
3237 eli
->li_state
|= EXT4_LAZYINIT_QUIT
;
3244 static struct ext4_li_request
*ext4_li_request_new(struct super_block
*sb
,
3247 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3248 struct ext4_li_request
*elr
;
3250 elr
= kzalloc(sizeof(*elr
), GFP_KERNEL
);
3256 elr
->lr_next_group
= start
;
3259 * Randomize first schedule time of the request to
3260 * spread the inode table initialization requests
3263 elr
->lr_next_sched
= jiffies
+ (prandom_u32() %
3264 (EXT4_DEF_LI_MAX_START_DELAY
* HZ
));
3268 int ext4_register_li_request(struct super_block
*sb
,
3269 ext4_group_t first_not_zeroed
)
3271 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3272 struct ext4_li_request
*elr
= NULL
;
3273 ext4_group_t ngroups
= sbi
->s_groups_count
;
3276 mutex_lock(&ext4_li_mtx
);
3277 if (sbi
->s_li_request
!= NULL
) {
3279 * Reset timeout so it can be computed again, because
3280 * s_li_wait_mult might have changed.
3282 sbi
->s_li_request
->lr_timeout
= 0;
3286 if (first_not_zeroed
== ngroups
|| sb_rdonly(sb
) ||
3287 !test_opt(sb
, INIT_INODE_TABLE
))
3290 elr
= ext4_li_request_new(sb
, first_not_zeroed
);
3296 if (NULL
== ext4_li_info
) {
3297 ret
= ext4_li_info_new();
3302 mutex_lock(&ext4_li_info
->li_list_mtx
);
3303 list_add(&elr
->lr_request
, &ext4_li_info
->li_request_list
);
3304 mutex_unlock(&ext4_li_info
->li_list_mtx
);
3306 sbi
->s_li_request
= elr
;
3308 * set elr to NULL here since it has been inserted to
3309 * the request_list and the removal and free of it is
3310 * handled by ext4_clear_request_list from now on.
3314 if (!(ext4_li_info
->li_state
& EXT4_LAZYINIT_RUNNING
)) {
3315 ret
= ext4_run_lazyinit_thread();
3320 mutex_unlock(&ext4_li_mtx
);
3327 * We do not need to lock anything since this is called on
3330 static void ext4_destroy_lazyinit_thread(void)
3333 * If thread exited earlier
3334 * there's nothing to be done.
3336 if (!ext4_li_info
|| !ext4_lazyinit_task
)
3339 kthread_stop(ext4_lazyinit_task
);
3342 static int set_journal_csum_feature_set(struct super_block
*sb
)
3345 int compat
, incompat
;
3346 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3348 if (ext4_has_metadata_csum(sb
)) {
3349 /* journal checksum v3 */
3351 incompat
= JBD2_FEATURE_INCOMPAT_CSUM_V3
;
3353 /* journal checksum v1 */
3354 compat
= JBD2_FEATURE_COMPAT_CHECKSUM
;
3358 jbd2_journal_clear_features(sbi
->s_journal
,
3359 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3360 JBD2_FEATURE_INCOMPAT_CSUM_V3
|
3361 JBD2_FEATURE_INCOMPAT_CSUM_V2
);
3362 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
3363 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3365 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
|
3367 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
3368 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3371 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3372 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3374 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3375 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3382 * Note: calculating the overhead so we can be compatible with
3383 * historical BSD practice is quite difficult in the face of
3384 * clusters/bigalloc. This is because multiple metadata blocks from
3385 * different block group can end up in the same allocation cluster.
3386 * Calculating the exact overhead in the face of clustered allocation
3387 * requires either O(all block bitmaps) in memory or O(number of block
3388 * groups**2) in time. We will still calculate the superblock for
3389 * older file systems --- and if we come across with a bigalloc file
3390 * system with zero in s_overhead_clusters the estimate will be close to
3391 * correct especially for very large cluster sizes --- but for newer
3392 * file systems, it's better to calculate this figure once at mkfs
3393 * time, and store it in the superblock. If the superblock value is
3394 * present (even for non-bigalloc file systems), we will use it.
3396 static int count_overhead(struct super_block
*sb
, ext4_group_t grp
,
3399 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3400 struct ext4_group_desc
*gdp
;
3401 ext4_fsblk_t first_block
, last_block
, b
;
3402 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3403 int s
, j
, count
= 0;
3405 if (!ext4_has_feature_bigalloc(sb
))
3406 return (ext4_bg_has_super(sb
, grp
) + ext4_bg_num_gdb(sb
, grp
) +
3407 sbi
->s_itb_per_group
+ 2);
3409 first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
) +
3410 (grp
* EXT4_BLOCKS_PER_GROUP(sb
));
3411 last_block
= first_block
+ EXT4_BLOCKS_PER_GROUP(sb
) - 1;
3412 for (i
= 0; i
< ngroups
; i
++) {
3413 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
3414 b
= ext4_block_bitmap(sb
, gdp
);
3415 if (b
>= first_block
&& b
<= last_block
) {
3416 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3419 b
= ext4_inode_bitmap(sb
, gdp
);
3420 if (b
>= first_block
&& b
<= last_block
) {
3421 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3424 b
= ext4_inode_table(sb
, gdp
);
3425 if (b
>= first_block
&& b
+ sbi
->s_itb_per_group
<= last_block
)
3426 for (j
= 0; j
< sbi
->s_itb_per_group
; j
++, b
++) {
3427 int c
= EXT4_B2C(sbi
, b
- first_block
);
3428 ext4_set_bit(c
, buf
);
3434 if (ext4_bg_has_super(sb
, grp
)) {
3435 ext4_set_bit(s
++, buf
);
3438 j
= ext4_bg_num_gdb(sb
, grp
);
3439 if (s
+ j
> EXT4_BLOCKS_PER_GROUP(sb
)) {
3440 ext4_error(sb
, "Invalid number of block group "
3441 "descriptor blocks: %d", j
);
3442 j
= EXT4_BLOCKS_PER_GROUP(sb
) - s
;
3446 ext4_set_bit(EXT4_B2C(sbi
, s
++), buf
);
3450 return EXT4_CLUSTERS_PER_GROUP(sb
) -
3451 ext4_count_free(buf
, EXT4_CLUSTERS_PER_GROUP(sb
) / 8);
3455 * Compute the overhead and stash it in sbi->s_overhead
3457 int ext4_calculate_overhead(struct super_block
*sb
)
3459 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3460 struct ext4_super_block
*es
= sbi
->s_es
;
3461 struct inode
*j_inode
;
3462 unsigned int j_blocks
, j_inum
= le32_to_cpu(es
->s_journal_inum
);
3463 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3464 ext4_fsblk_t overhead
= 0;
3465 char *buf
= (char *) get_zeroed_page(GFP_NOFS
);
3471 * Compute the overhead (FS structures). This is constant
3472 * for a given filesystem unless the number of block groups
3473 * changes so we cache the previous value until it does.
3477 * All of the blocks before first_data_block are overhead
3479 overhead
= EXT4_B2C(sbi
, le32_to_cpu(es
->s_first_data_block
));
3482 * Add the overhead found in each block group
3484 for (i
= 0; i
< ngroups
; i
++) {
3487 blks
= count_overhead(sb
, i
, buf
);
3490 memset(buf
, 0, PAGE_SIZE
);
3495 * Add the internal journal blocks whether the journal has been
3498 if (sbi
->s_journal
&& !sbi
->journal_bdev
)
3499 overhead
+= EXT4_NUM_B2C(sbi
, sbi
->s_journal
->j_maxlen
);
3500 else if (ext4_has_feature_journal(sb
) && !sbi
->s_journal
) {
3501 j_inode
= ext4_get_journal_inode(sb
, j_inum
);
3503 j_blocks
= j_inode
->i_size
>> sb
->s_blocksize_bits
;
3504 overhead
+= EXT4_NUM_B2C(sbi
, j_blocks
);
3507 ext4_msg(sb
, KERN_ERR
, "can't get journal size");
3510 sbi
->s_overhead
= overhead
;
3512 free_page((unsigned long) buf
);
3516 static void ext4_set_resv_clusters(struct super_block
*sb
)
3518 ext4_fsblk_t resv_clusters
;
3519 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3522 * There's no need to reserve anything when we aren't using extents.
3523 * The space estimates are exact, there are no unwritten extents,
3524 * hole punching doesn't need new metadata... This is needed especially
3525 * to keep ext2/3 backward compatibility.
3527 if (!ext4_has_feature_extents(sb
))
3530 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3531 * This should cover the situations where we can not afford to run
3532 * out of space like for example punch hole, or converting
3533 * unwritten extents in delalloc path. In most cases such
3534 * allocation would require 1, or 2 blocks, higher numbers are
3537 resv_clusters
= (ext4_blocks_count(sbi
->s_es
) >>
3538 sbi
->s_cluster_bits
);
3540 do_div(resv_clusters
, 50);
3541 resv_clusters
= min_t(ext4_fsblk_t
, resv_clusters
, 4096);
3543 atomic64_set(&sbi
->s_resv_clusters
, resv_clusters
);
3546 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
3548 struct dax_device
*dax_dev
= fs_dax_get_by_bdev(sb
->s_bdev
);
3549 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
3550 struct buffer_head
*bh
;
3551 struct ext4_super_block
*es
= NULL
;
3552 struct ext4_sb_info
*sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
3554 ext4_fsblk_t sb_block
= get_sb_block(&data
);
3555 ext4_fsblk_t logical_sb_block
;
3556 unsigned long offset
= 0;
3557 unsigned long journal_devnum
= 0;
3558 unsigned long def_mount_opts
;
3562 int blocksize
, clustersize
;
3563 unsigned int db_count
;
3565 int needs_recovery
, has_huge_files
, has_bigalloc
;
3568 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
3569 ext4_group_t first_not_zeroed
;
3571 if ((data
&& !orig_data
) || !sbi
)
3574 sbi
->s_daxdev
= dax_dev
;
3575 sbi
->s_blockgroup_lock
=
3576 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
3577 if (!sbi
->s_blockgroup_lock
)
3580 sb
->s_fs_info
= sbi
;
3582 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
3583 sbi
->s_sb_block
= sb_block
;
3584 if (sb
->s_bdev
->bd_part
)
3585 sbi
->s_sectors_written_start
=
3586 part_stat_read(sb
->s_bdev
->bd_part
, sectors
[STAT_WRITE
]);
3588 /* Cleanup superblock name */
3589 strreplace(sb
->s_id
, '/', '!');
3591 /* -EINVAL is default */
3593 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
3595 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
3600 * The ext4 superblock will not be buffer aligned for other than 1kB
3601 * block sizes. We need to calculate the offset from buffer start.
3603 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
3604 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3605 offset
= do_div(logical_sb_block
, blocksize
);
3607 logical_sb_block
= sb_block
;
3610 if (!(bh
= sb_bread_unmovable(sb
, logical_sb_block
))) {
3611 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
3615 * Note: s_es must be initialized as soon as possible because
3616 * some ext4 macro-instructions depend on its value
3618 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
3620 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
3621 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
3623 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
3625 /* Warn if metadata_csum and gdt_csum are both set. */
3626 if (ext4_has_feature_metadata_csum(sb
) &&
3627 ext4_has_feature_gdt_csum(sb
))
3628 ext4_warning(sb
, "metadata_csum and uninit_bg are "
3629 "redundant flags; please run fsck.");
3631 /* Check for a known checksum algorithm */
3632 if (!ext4_verify_csum_type(sb
, es
)) {
3633 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3634 "unknown checksum algorithm.");
3639 /* Load the checksum driver */
3640 sbi
->s_chksum_driver
= crypto_alloc_shash("crc32c", 0, 0);
3641 if (IS_ERR(sbi
->s_chksum_driver
)) {
3642 ext4_msg(sb
, KERN_ERR
, "Cannot load crc32c driver.");
3643 ret
= PTR_ERR(sbi
->s_chksum_driver
);
3644 sbi
->s_chksum_driver
= NULL
;
3648 /* Check superblock checksum */
3649 if (!ext4_superblock_csum_verify(sb
, es
)) {
3650 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3651 "invalid superblock checksum. Run e2fsck?");
3657 /* Precompute checksum seed for all metadata */
3658 if (ext4_has_feature_csum_seed(sb
))
3659 sbi
->s_csum_seed
= le32_to_cpu(es
->s_checksum_seed
);
3660 else if (ext4_has_metadata_csum(sb
) || ext4_has_feature_ea_inode(sb
))
3661 sbi
->s_csum_seed
= ext4_chksum(sbi
, ~0, es
->s_uuid
,
3662 sizeof(es
->s_uuid
));
3664 /* Set defaults before we parse the mount options */
3665 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
3666 set_opt(sb
, INIT_INODE_TABLE
);
3667 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
3669 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
)
3671 if (def_mount_opts
& EXT4_DEFM_UID16
)
3672 set_opt(sb
, NO_UID32
);
3673 /* xattr user namespace & acls are now defaulted on */
3674 set_opt(sb
, XATTR_USER
);
3675 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3676 set_opt(sb
, POSIX_ACL
);
3678 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3679 if (ext4_has_metadata_csum(sb
))
3680 set_opt(sb
, JOURNAL_CHECKSUM
);
3682 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
3683 set_opt(sb
, JOURNAL_DATA
);
3684 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
3685 set_opt(sb
, ORDERED_DATA
);
3686 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
3687 set_opt(sb
, WRITEBACK_DATA
);
3689 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
3690 set_opt(sb
, ERRORS_PANIC
);
3691 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
3692 set_opt(sb
, ERRORS_CONT
);
3694 set_opt(sb
, ERRORS_RO
);
3695 /* block_validity enabled by default; disable with noblock_validity */
3696 set_opt(sb
, BLOCK_VALIDITY
);
3697 if (def_mount_opts
& EXT4_DEFM_DISCARD
)
3698 set_opt(sb
, DISCARD
);
3700 sbi
->s_resuid
= make_kuid(&init_user_ns
, le16_to_cpu(es
->s_def_resuid
));
3701 sbi
->s_resgid
= make_kgid(&init_user_ns
, le16_to_cpu(es
->s_def_resgid
));
3702 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
3703 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
3704 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
3706 if ((def_mount_opts
& EXT4_DEFM_NOBARRIER
) == 0)
3707 set_opt(sb
, BARRIER
);
3710 * enable delayed allocation by default
3711 * Use -o nodelalloc to turn it off
3713 if (!IS_EXT3_SB(sb
) && !IS_EXT2_SB(sb
) &&
3714 ((def_mount_opts
& EXT4_DEFM_NODELALLOC
) == 0))
3715 set_opt(sb
, DELALLOC
);
3718 * set default s_li_wait_mult for lazyinit, for the case there is
3719 * no mount option specified.
3721 sbi
->s_li_wait_mult
= EXT4_DEF_LI_WAIT_MULT
;
3723 if (sbi
->s_es
->s_mount_opts
[0]) {
3724 char *s_mount_opts
= kstrndup(sbi
->s_es
->s_mount_opts
,
3725 sizeof(sbi
->s_es
->s_mount_opts
),
3729 if (!parse_options(s_mount_opts
, sb
, &journal_devnum
,
3730 &journal_ioprio
, 0)) {
3731 ext4_msg(sb
, KERN_WARNING
,
3732 "failed to parse options in superblock: %s",
3735 kfree(s_mount_opts
);
3737 sbi
->s_def_mount_opt
= sbi
->s_mount_opt
;
3738 if (!parse_options((char *) data
, sb
, &journal_devnum
,
3739 &journal_ioprio
, 0))
3742 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
3743 printk_once(KERN_WARNING
"EXT4-fs: Warning: mounting "
3744 "with data=journal disables delayed "
3745 "allocation and O_DIRECT support!\n");
3746 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
3747 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3748 "both data=journal and delalloc");
3751 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
3752 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3753 "both data=journal and dioread_nolock");
3756 if (test_opt(sb
, DAX
)) {
3757 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3758 "both data=journal and dax");
3761 if (ext4_has_feature_encrypt(sb
)) {
3762 ext4_msg(sb
, KERN_WARNING
,
3763 "encrypted files will use data=ordered "
3764 "instead of data journaling mode");
3766 if (test_opt(sb
, DELALLOC
))
3767 clear_opt(sb
, DELALLOC
);
3769 sb
->s_iflags
|= SB_I_CGROUPWB
;
3772 sb
->s_flags
= (sb
->s_flags
& ~SB_POSIXACL
) |
3773 (test_opt(sb
, POSIX_ACL
) ? SB_POSIXACL
: 0);
3775 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
3776 (ext4_has_compat_features(sb
) ||
3777 ext4_has_ro_compat_features(sb
) ||
3778 ext4_has_incompat_features(sb
)))
3779 ext4_msg(sb
, KERN_WARNING
,
3780 "feature flags set on rev 0 fs, "
3781 "running e2fsck is recommended");
3783 if (es
->s_creator_os
== cpu_to_le32(EXT4_OS_HURD
)) {
3784 set_opt2(sb
, HURD_COMPAT
);
3785 if (ext4_has_feature_64bit(sb
)) {
3786 ext4_msg(sb
, KERN_ERR
,
3787 "The Hurd can't support 64-bit file systems");
3792 * ea_inode feature uses l_i_version field which is not
3793 * available in HURD_COMPAT mode.
3795 if (ext4_has_feature_ea_inode(sb
)) {
3796 ext4_msg(sb
, KERN_ERR
,
3797 "ea_inode feature is not supported for Hurd");
3802 if (IS_EXT2_SB(sb
)) {
3803 if (ext2_feature_set_ok(sb
))
3804 ext4_msg(sb
, KERN_INFO
, "mounting ext2 file system "
3805 "using the ext4 subsystem");
3808 * If we're probing be silent, if this looks like
3809 * it's actually an ext[34] filesystem.
3811 if (silent
&& ext4_feature_set_ok(sb
, sb_rdonly(sb
)))
3813 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext2 due "
3814 "to feature incompatibilities");
3819 if (IS_EXT3_SB(sb
)) {
3820 if (ext3_feature_set_ok(sb
))
3821 ext4_msg(sb
, KERN_INFO
, "mounting ext3 file system "
3822 "using the ext4 subsystem");
3825 * If we're probing be silent, if this looks like
3826 * it's actually an ext4 filesystem.
3828 if (silent
&& ext4_feature_set_ok(sb
, sb_rdonly(sb
)))
3830 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext3 due "
3831 "to feature incompatibilities");
3837 * Check feature flags regardless of the revision level, since we
3838 * previously didn't change the revision level when setting the flags,
3839 * so there is a chance incompat flags are set on a rev 0 filesystem.
3841 if (!ext4_feature_set_ok(sb
, (sb_rdonly(sb
))))
3844 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
3845 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
3846 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
3847 ext4_msg(sb
, KERN_ERR
,
3848 "Unsupported filesystem blocksize %d (%d log_block_size)",
3849 blocksize
, le32_to_cpu(es
->s_log_block_size
));
3852 if (le32_to_cpu(es
->s_log_block_size
) >
3853 (EXT4_MAX_BLOCK_LOG_SIZE
- EXT4_MIN_BLOCK_LOG_SIZE
)) {
3854 ext4_msg(sb
, KERN_ERR
,
3855 "Invalid log block size: %u",
3856 le32_to_cpu(es
->s_log_block_size
));
3859 if (le32_to_cpu(es
->s_log_cluster_size
) >
3860 (EXT4_MAX_CLUSTER_LOG_SIZE
- EXT4_MIN_BLOCK_LOG_SIZE
)) {
3861 ext4_msg(sb
, KERN_ERR
,
3862 "Invalid log cluster size: %u",
3863 le32_to_cpu(es
->s_log_cluster_size
));
3867 if (le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
) > (blocksize
/ 4)) {
3868 ext4_msg(sb
, KERN_ERR
,
3869 "Number of reserved GDT blocks insanely large: %d",
3870 le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
));
3874 if (sbi
->s_mount_opt
& EXT4_MOUNT_DAX
) {
3875 if (ext4_has_feature_inline_data(sb
)) {
3876 ext4_msg(sb
, KERN_ERR
, "Cannot use DAX on a filesystem"
3877 " that may contain inline data");
3880 if (!bdev_dax_supported(sb
->s_bdev
, blocksize
)) {
3881 ext4_msg(sb
, KERN_ERR
,
3882 "DAX unsupported by block device.");
3887 if (ext4_has_feature_encrypt(sb
) && es
->s_encryption_level
) {
3888 ext4_msg(sb
, KERN_ERR
, "Unsupported encryption level %d",
3889 es
->s_encryption_level
);
3893 if (sb
->s_blocksize
!= blocksize
) {
3894 /* Validate the filesystem blocksize */
3895 if (!sb_set_blocksize(sb
, blocksize
)) {
3896 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
3902 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3903 offset
= do_div(logical_sb_block
, blocksize
);
3904 bh
= sb_bread_unmovable(sb
, logical_sb_block
);
3906 ext4_msg(sb
, KERN_ERR
,
3907 "Can't read superblock on 2nd try");
3910 es
= (struct ext4_super_block
*)(bh
->b_data
+ offset
);
3912 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
3913 ext4_msg(sb
, KERN_ERR
,
3914 "Magic mismatch, very weird!");
3919 has_huge_files
= ext4_has_feature_huge_file(sb
);
3920 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
3922 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
3924 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
3925 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
3926 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
3928 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
3929 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
3930 if (sbi
->s_first_ino
< EXT4_GOOD_OLD_FIRST_INO
) {
3931 ext4_msg(sb
, KERN_ERR
, "invalid first ino: %u",
3935 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
3936 (!is_power_of_2(sbi
->s_inode_size
)) ||
3937 (sbi
->s_inode_size
> blocksize
)) {
3938 ext4_msg(sb
, KERN_ERR
,
3939 "unsupported inode size: %d",
3943 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
3944 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
3947 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
3948 if (ext4_has_feature_64bit(sb
)) {
3949 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
3950 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
3951 !is_power_of_2(sbi
->s_desc_size
)) {
3952 ext4_msg(sb
, KERN_ERR
,
3953 "unsupported descriptor size %lu",
3958 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
3960 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
3961 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
3963 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
3964 if (sbi
->s_inodes_per_block
== 0)
3966 if (sbi
->s_inodes_per_group
< sbi
->s_inodes_per_block
||
3967 sbi
->s_inodes_per_group
> blocksize
* 8) {
3968 ext4_msg(sb
, KERN_ERR
, "invalid inodes per group: %lu\n",
3969 sbi
->s_blocks_per_group
);
3972 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
3973 sbi
->s_inodes_per_block
;
3974 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
3976 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3977 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
3978 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
3980 for (i
= 0; i
< 4; i
++)
3981 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
3982 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
3983 if (ext4_has_feature_dir_index(sb
)) {
3984 i
= le32_to_cpu(es
->s_flags
);
3985 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
3986 sbi
->s_hash_unsigned
= 3;
3987 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
3988 #ifdef __CHAR_UNSIGNED__
3991 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
3992 sbi
->s_hash_unsigned
= 3;
3996 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
4001 /* Handle clustersize */
4002 clustersize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_cluster_size
);
4003 has_bigalloc
= ext4_has_feature_bigalloc(sb
);
4005 if (clustersize
< blocksize
) {
4006 ext4_msg(sb
, KERN_ERR
,
4007 "cluster size (%d) smaller than "
4008 "block size (%d)", clustersize
, blocksize
);
4011 sbi
->s_cluster_bits
= le32_to_cpu(es
->s_log_cluster_size
) -
4012 le32_to_cpu(es
->s_log_block_size
);
4013 sbi
->s_clusters_per_group
=
4014 le32_to_cpu(es
->s_clusters_per_group
);
4015 if (sbi
->s_clusters_per_group
> blocksize
* 8) {
4016 ext4_msg(sb
, KERN_ERR
,
4017 "#clusters per group too big: %lu",
4018 sbi
->s_clusters_per_group
);
4021 if (sbi
->s_blocks_per_group
!=
4022 (sbi
->s_clusters_per_group
* (clustersize
/ blocksize
))) {
4023 ext4_msg(sb
, KERN_ERR
, "blocks per group (%lu) and "
4024 "clusters per group (%lu) inconsistent",
4025 sbi
->s_blocks_per_group
,
4026 sbi
->s_clusters_per_group
);
4030 if (clustersize
!= blocksize
) {
4031 ext4_msg(sb
, KERN_ERR
,
4032 "fragment/cluster size (%d) != "
4033 "block size (%d)", clustersize
, blocksize
);
4036 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
4037 ext4_msg(sb
, KERN_ERR
,
4038 "#blocks per group too big: %lu",
4039 sbi
->s_blocks_per_group
);
4042 sbi
->s_clusters_per_group
= sbi
->s_blocks_per_group
;
4043 sbi
->s_cluster_bits
= 0;
4045 sbi
->s_cluster_ratio
= clustersize
/ blocksize
;
4047 /* Do we have standard group size of clustersize * 8 blocks ? */
4048 if (sbi
->s_blocks_per_group
== clustersize
<< 3)
4049 set_opt2(sb
, STD_GROUP_SIZE
);
4052 * Test whether we have more sectors than will fit in sector_t,
4053 * and whether the max offset is addressable by the page cache.
4055 err
= generic_check_addressable(sb
->s_blocksize_bits
,
4056 ext4_blocks_count(es
));
4058 ext4_msg(sb
, KERN_ERR
, "filesystem"
4059 " too large to mount safely on this system");
4060 if (sizeof(sector_t
) < 8)
4061 ext4_msg(sb
, KERN_WARNING
, "CONFIG_LBDAF not enabled");
4065 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
4068 /* check blocks count against device size */
4069 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
4070 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
4071 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
4072 "exceeds size of device (%llu blocks)",
4073 ext4_blocks_count(es
), blocks_count
);
4078 * It makes no sense for the first data block to be beyond the end
4079 * of the filesystem.
4081 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
4082 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data "
4083 "block %u is beyond end of filesystem (%llu)",
4084 le32_to_cpu(es
->s_first_data_block
),
4085 ext4_blocks_count(es
));
4088 if ((es
->s_first_data_block
== 0) && (es
->s_log_block_size
== 0) &&
4089 (sbi
->s_cluster_ratio
== 1)) {
4090 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data "
4091 "block is 0 with a 1k block and cluster size");
4095 blocks_count
= (ext4_blocks_count(es
) -
4096 le32_to_cpu(es
->s_first_data_block
) +
4097 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
4098 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
4099 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
4100 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %u "
4101 "(block count %llu, first data block %u, "
4102 "blocks per group %lu)", sbi
->s_groups_count
,
4103 ext4_blocks_count(es
),
4104 le32_to_cpu(es
->s_first_data_block
),
4105 EXT4_BLOCKS_PER_GROUP(sb
));
4108 sbi
->s_groups_count
= blocks_count
;
4109 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
4110 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
4111 if (((u64
)sbi
->s_groups_count
* sbi
->s_inodes_per_group
) !=
4112 le32_to_cpu(es
->s_inodes_count
)) {
4113 ext4_msg(sb
, KERN_ERR
, "inodes count not valid: %u vs %llu",
4114 le32_to_cpu(es
->s_inodes_count
),
4115 ((u64
)sbi
->s_groups_count
* sbi
->s_inodes_per_group
));
4119 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
4120 EXT4_DESC_PER_BLOCK(sb
);
4121 if (ext4_has_feature_meta_bg(sb
)) {
4122 if (le32_to_cpu(es
->s_first_meta_bg
) > db_count
) {
4123 ext4_msg(sb
, KERN_WARNING
,
4124 "first meta block group too large: %u "
4125 "(group descriptor block count %u)",
4126 le32_to_cpu(es
->s_first_meta_bg
), db_count
);
4130 sbi
->s_group_desc
= kvmalloc_array(db_count
,
4131 sizeof(struct buffer_head
*),
4133 if (sbi
->s_group_desc
== NULL
) {
4134 ext4_msg(sb
, KERN_ERR
, "not enough memory");
4139 bgl_lock_init(sbi
->s_blockgroup_lock
);
4141 /* Pre-read the descriptors into the buffer cache */
4142 for (i
= 0; i
< db_count
; i
++) {
4143 block
= descriptor_loc(sb
, logical_sb_block
, i
);
4144 sb_breadahead(sb
, block
);
4147 for (i
= 0; i
< db_count
; i
++) {
4148 block
= descriptor_loc(sb
, logical_sb_block
, i
);
4149 sbi
->s_group_desc
[i
] = sb_bread_unmovable(sb
, block
);
4150 if (!sbi
->s_group_desc
[i
]) {
4151 ext4_msg(sb
, KERN_ERR
,
4152 "can't read group descriptor %d", i
);
4157 sbi
->s_gdb_count
= db_count
;
4158 if (!ext4_check_descriptors(sb
, logical_sb_block
, &first_not_zeroed
)) {
4159 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
4160 ret
= -EFSCORRUPTED
;
4164 timer_setup(&sbi
->s_err_report
, print_daily_error_info
, 0);
4166 /* Register extent status tree shrinker */
4167 if (ext4_es_register_shrinker(sbi
))
4170 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
4171 sbi
->s_extent_max_zeroout_kb
= 32;
4174 * set up enough so that it can read an inode
4176 sb
->s_op
= &ext4_sops
;
4177 sb
->s_export_op
= &ext4_export_ops
;
4178 sb
->s_xattr
= ext4_xattr_handlers
;
4179 #ifdef CONFIG_FS_ENCRYPTION
4180 sb
->s_cop
= &ext4_cryptops
;
4183 sb
->dq_op
= &ext4_quota_operations
;
4184 if (ext4_has_feature_quota(sb
))
4185 sb
->s_qcop
= &dquot_quotactl_sysfile_ops
;
4187 sb
->s_qcop
= &ext4_qctl_operations
;
4188 sb
->s_quota_types
= QTYPE_MASK_USR
| QTYPE_MASK_GRP
| QTYPE_MASK_PRJ
;
4190 memcpy(&sb
->s_uuid
, es
->s_uuid
, sizeof(es
->s_uuid
));
4192 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
4193 mutex_init(&sbi
->s_orphan_lock
);
4197 needs_recovery
= (es
->s_last_orphan
!= 0 ||
4198 ext4_has_feature_journal_needs_recovery(sb
));
4200 if (ext4_has_feature_mmp(sb
) && !sb_rdonly(sb
))
4201 if (ext4_multi_mount_protect(sb
, le64_to_cpu(es
->s_mmp_block
)))
4202 goto failed_mount3a
;
4205 * The first inode we look at is the journal inode. Don't try
4206 * root first: it may be modified in the journal!
4208 if (!test_opt(sb
, NOLOAD
) && ext4_has_feature_journal(sb
)) {
4209 err
= ext4_load_journal(sb
, es
, journal_devnum
);
4211 goto failed_mount3a
;
4212 } else if (test_opt(sb
, NOLOAD
) && !sb_rdonly(sb
) &&
4213 ext4_has_feature_journal_needs_recovery(sb
)) {
4214 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
4215 "suppressed and not mounted read-only");
4216 goto failed_mount_wq
;
4218 /* Nojournal mode, all journal mount options are illegal */
4219 if (test_opt2(sb
, EXPLICIT_JOURNAL_CHECKSUM
)) {
4220 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4221 "journal_checksum, fs mounted w/o journal");
4222 goto failed_mount_wq
;
4224 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
4225 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4226 "journal_async_commit, fs mounted w/o journal");
4227 goto failed_mount_wq
;
4229 if (sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
) {
4230 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4231 "commit=%lu, fs mounted w/o journal",
4232 sbi
->s_commit_interval
/ HZ
);
4233 goto failed_mount_wq
;
4235 if (EXT4_MOUNT_DATA_FLAGS
&
4236 (sbi
->s_mount_opt
^ sbi
->s_def_mount_opt
)) {
4237 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4238 "data=, fs mounted w/o journal");
4239 goto failed_mount_wq
;
4241 sbi
->s_def_mount_opt
&= EXT4_MOUNT_JOURNAL_CHECKSUM
;
4242 clear_opt(sb
, JOURNAL_CHECKSUM
);
4243 clear_opt(sb
, DATA_FLAGS
);
4244 sbi
->s_journal
= NULL
;
4249 if (ext4_has_feature_64bit(sb
) &&
4250 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
4251 JBD2_FEATURE_INCOMPAT_64BIT
)) {
4252 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
4253 goto failed_mount_wq
;
4256 if (!set_journal_csum_feature_set(sb
)) {
4257 ext4_msg(sb
, KERN_ERR
, "Failed to set journal checksum "
4259 goto failed_mount_wq
;
4262 /* We have now updated the journal if required, so we can
4263 * validate the data journaling mode. */
4264 switch (test_opt(sb
, DATA_FLAGS
)) {
4266 /* No mode set, assume a default based on the journal
4267 * capabilities: ORDERED_DATA if the journal can
4268 * cope, else JOURNAL_DATA
4270 if (jbd2_journal_check_available_features
4271 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
4272 set_opt(sb
, ORDERED_DATA
);
4273 sbi
->s_def_mount_opt
|= EXT4_MOUNT_ORDERED_DATA
;
4275 set_opt(sb
, JOURNAL_DATA
);
4276 sbi
->s_def_mount_opt
|= EXT4_MOUNT_JOURNAL_DATA
;
4280 case EXT4_MOUNT_ORDERED_DATA
:
4281 case EXT4_MOUNT_WRITEBACK_DATA
:
4282 if (!jbd2_journal_check_available_features
4283 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
4284 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
4285 "requested data journaling mode");
4286 goto failed_mount_wq
;
4292 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
&&
4293 test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
4294 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4295 "journal_async_commit in data=ordered mode");
4296 goto failed_mount_wq
;
4299 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
4301 sbi
->s_journal
->j_commit_callback
= ext4_journal_commit_callback
;
4304 if (!test_opt(sb
, NO_MBCACHE
)) {
4305 sbi
->s_ea_block_cache
= ext4_xattr_create_cache();
4306 if (!sbi
->s_ea_block_cache
) {
4307 ext4_msg(sb
, KERN_ERR
,
4308 "Failed to create ea_block_cache");
4309 goto failed_mount_wq
;
4312 if (ext4_has_feature_ea_inode(sb
)) {
4313 sbi
->s_ea_inode_cache
= ext4_xattr_create_cache();
4314 if (!sbi
->s_ea_inode_cache
) {
4315 ext4_msg(sb
, KERN_ERR
,
4316 "Failed to create ea_inode_cache");
4317 goto failed_mount_wq
;
4322 if ((DUMMY_ENCRYPTION_ENABLED(sbi
) || ext4_has_feature_encrypt(sb
)) &&
4323 (blocksize
!= PAGE_SIZE
)) {
4324 ext4_msg(sb
, KERN_ERR
,
4325 "Unsupported blocksize for fs encryption");
4326 goto failed_mount_wq
;
4329 if (DUMMY_ENCRYPTION_ENABLED(sbi
) && !sb_rdonly(sb
) &&
4330 !ext4_has_feature_encrypt(sb
)) {
4331 ext4_set_feature_encrypt(sb
);
4332 ext4_commit_super(sb
, 1);
4336 * Get the # of file system overhead blocks from the
4337 * superblock if present.
4339 if (es
->s_overhead_clusters
)
4340 sbi
->s_overhead
= le32_to_cpu(es
->s_overhead_clusters
);
4342 err
= ext4_calculate_overhead(sb
);
4344 goto failed_mount_wq
;
4348 * The maximum number of concurrent works can be high and
4349 * concurrency isn't really necessary. Limit it to 1.
4351 EXT4_SB(sb
)->rsv_conversion_wq
=
4352 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM
| WQ_UNBOUND
, 1);
4353 if (!EXT4_SB(sb
)->rsv_conversion_wq
) {
4354 printk(KERN_ERR
"EXT4-fs: failed to create workqueue\n");
4360 * The jbd2_journal_load will have done any necessary log recovery,
4361 * so we can safely mount the rest of the filesystem now.
4364 root
= ext4_iget(sb
, EXT4_ROOT_INO
, EXT4_IGET_SPECIAL
);
4366 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
4367 ret
= PTR_ERR(root
);
4371 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
4372 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
4376 sb
->s_root
= d_make_root(root
);
4378 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
4383 ret
= ext4_setup_super(sb
, es
, sb_rdonly(sb
));
4384 if (ret
== -EROFS
) {
4385 sb
->s_flags
|= SB_RDONLY
;
4388 goto failed_mount4a
;
4390 /* determine the minimum size of new large inodes, if present */
4391 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
&&
4392 sbi
->s_want_extra_isize
== 0) {
4393 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
4394 EXT4_GOOD_OLD_INODE_SIZE
;
4395 if (ext4_has_feature_extra_isize(sb
)) {
4396 if (sbi
->s_want_extra_isize
<
4397 le16_to_cpu(es
->s_want_extra_isize
))
4398 sbi
->s_want_extra_isize
=
4399 le16_to_cpu(es
->s_want_extra_isize
);
4400 if (sbi
->s_want_extra_isize
<
4401 le16_to_cpu(es
->s_min_extra_isize
))
4402 sbi
->s_want_extra_isize
=
4403 le16_to_cpu(es
->s_min_extra_isize
);
4406 /* Check if enough inode space is available */
4407 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
4408 sbi
->s_inode_size
) {
4409 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
4410 EXT4_GOOD_OLD_INODE_SIZE
;
4411 ext4_msg(sb
, KERN_INFO
, "required extra inode space not"
4415 ext4_set_resv_clusters(sb
);
4417 err
= ext4_setup_system_zone(sb
);
4419 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
4421 goto failed_mount4a
;
4425 err
= ext4_mb_init(sb
);
4427 ext4_msg(sb
, KERN_ERR
, "failed to initialize mballoc (%d)",
4432 block
= ext4_count_free_clusters(sb
);
4433 ext4_free_blocks_count_set(sbi
->s_es
,
4434 EXT4_C2B(sbi
, block
));
4435 ext4_superblock_csum_set(sb
);
4436 err
= percpu_counter_init(&sbi
->s_freeclusters_counter
, block
,
4439 unsigned long freei
= ext4_count_free_inodes(sb
);
4440 sbi
->s_es
->s_free_inodes_count
= cpu_to_le32(freei
);
4441 ext4_superblock_csum_set(sb
);
4442 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
, freei
,
4446 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
4447 ext4_count_dirs(sb
), GFP_KERNEL
);
4449 err
= percpu_counter_init(&sbi
->s_dirtyclusters_counter
, 0,
4452 err
= percpu_init_rwsem(&sbi
->s_journal_flag_rwsem
);
4455 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
4459 if (ext4_has_feature_flex_bg(sb
))
4460 if (!ext4_fill_flex_info(sb
)) {
4461 ext4_msg(sb
, KERN_ERR
,
4462 "unable to initialize "
4463 "flex_bg meta info!");
4467 err
= ext4_register_li_request(sb
, first_not_zeroed
);
4471 err
= ext4_register_sysfs(sb
);
4476 /* Enable quota usage during mount. */
4477 if (ext4_has_feature_quota(sb
) && !sb_rdonly(sb
)) {
4478 err
= ext4_enable_quotas(sb
);
4482 #endif /* CONFIG_QUOTA */
4484 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
4485 ext4_orphan_cleanup(sb
, es
);
4486 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
4487 if (needs_recovery
) {
4488 ext4_msg(sb
, KERN_INFO
, "recovery complete");
4489 ext4_mark_recovery_complete(sb
, es
);
4491 if (EXT4_SB(sb
)->s_journal
) {
4492 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
4493 descr
= " journalled data mode";
4494 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
4495 descr
= " ordered data mode";
4497 descr
= " writeback data mode";
4499 descr
= "out journal";
4501 if (test_opt(sb
, DISCARD
)) {
4502 struct request_queue
*q
= bdev_get_queue(sb
->s_bdev
);
4503 if (!blk_queue_discard(q
))
4504 ext4_msg(sb
, KERN_WARNING
,
4505 "mounting with \"discard\" option, but "
4506 "the device does not support discard");
4509 if (___ratelimit(&ext4_mount_msg_ratelimit
, "EXT4-fs mount"))
4510 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s. "
4511 "Opts: %.*s%s%s", descr
,
4512 (int) sizeof(sbi
->s_es
->s_mount_opts
),
4513 sbi
->s_es
->s_mount_opts
,
4514 *sbi
->s_es
->s_mount_opts
? "; " : "", orig_data
);
4516 if (es
->s_error_count
)
4517 mod_timer(&sbi
->s_err_report
, jiffies
+ 300*HZ
); /* 5 minutes */
4519 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4520 ratelimit_state_init(&sbi
->s_err_ratelimit_state
, 5 * HZ
, 10);
4521 ratelimit_state_init(&sbi
->s_warning_ratelimit_state
, 5 * HZ
, 10);
4522 ratelimit_state_init(&sbi
->s_msg_ratelimit_state
, 5 * HZ
, 10);
4529 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
4534 ext4_unregister_sysfs(sb
);
4537 ext4_unregister_li_request(sb
);
4539 ext4_mb_release(sb
);
4540 if (sbi
->s_flex_groups
)
4541 kvfree(sbi
->s_flex_groups
);
4542 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
4543 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
4544 percpu_counter_destroy(&sbi
->s_dirs_counter
);
4545 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
4546 percpu_free_rwsem(&sbi
->s_journal_flag_rwsem
);
4548 ext4_ext_release(sb
);
4549 ext4_release_system_zone(sb
);
4554 ext4_msg(sb
, KERN_ERR
, "mount failed");
4555 if (EXT4_SB(sb
)->rsv_conversion_wq
)
4556 destroy_workqueue(EXT4_SB(sb
)->rsv_conversion_wq
);
4558 ext4_xattr_destroy_cache(sbi
->s_ea_inode_cache
);
4559 sbi
->s_ea_inode_cache
= NULL
;
4561 ext4_xattr_destroy_cache(sbi
->s_ea_block_cache
);
4562 sbi
->s_ea_block_cache
= NULL
;
4564 if (sbi
->s_journal
) {
4565 jbd2_journal_destroy(sbi
->s_journal
);
4566 sbi
->s_journal
= NULL
;
4569 ext4_es_unregister_shrinker(sbi
);
4571 del_timer_sync(&sbi
->s_err_report
);
4573 kthread_stop(sbi
->s_mmp_tsk
);
4575 for (i
= 0; i
< db_count
; i
++)
4576 brelse(sbi
->s_group_desc
[i
]);
4577 kvfree(sbi
->s_group_desc
);
4579 if (sbi
->s_chksum_driver
)
4580 crypto_free_shash(sbi
->s_chksum_driver
);
4582 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
4583 kfree(sbi
->s_qf_names
[i
]);
4585 ext4_blkdev_remove(sbi
);
4588 sb
->s_fs_info
= NULL
;
4589 kfree(sbi
->s_blockgroup_lock
);
4593 fs_put_dax(dax_dev
);
4594 return err
? err
: ret
;
4598 * Setup any per-fs journal parameters now. We'll do this both on
4599 * initial mount, once the journal has been initialised but before we've
4600 * done any recovery; and again on any subsequent remount.
4602 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
4604 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4606 journal
->j_commit_interval
= sbi
->s_commit_interval
;
4607 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
4608 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
4610 write_lock(&journal
->j_state_lock
);
4611 if (test_opt(sb
, BARRIER
))
4612 journal
->j_flags
|= JBD2_BARRIER
;
4614 journal
->j_flags
&= ~JBD2_BARRIER
;
4615 if (test_opt(sb
, DATA_ERR_ABORT
))
4616 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
4618 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
4619 write_unlock(&journal
->j_state_lock
);
4622 static struct inode
*ext4_get_journal_inode(struct super_block
*sb
,
4623 unsigned int journal_inum
)
4625 struct inode
*journal_inode
;
4628 * Test for the existence of a valid inode on disk. Bad things
4629 * happen if we iget() an unused inode, as the subsequent iput()
4630 * will try to delete it.
4632 journal_inode
= ext4_iget(sb
, journal_inum
, EXT4_IGET_SPECIAL
);
4633 if (IS_ERR(journal_inode
)) {
4634 ext4_msg(sb
, KERN_ERR
, "no journal found");
4637 if (!journal_inode
->i_nlink
) {
4638 make_bad_inode(journal_inode
);
4639 iput(journal_inode
);
4640 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
4644 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4645 journal_inode
, journal_inode
->i_size
);
4646 if (!S_ISREG(journal_inode
->i_mode
)) {
4647 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
4648 iput(journal_inode
);
4651 return journal_inode
;
4654 static journal_t
*ext4_get_journal(struct super_block
*sb
,
4655 unsigned int journal_inum
)
4657 struct inode
*journal_inode
;
4660 BUG_ON(!ext4_has_feature_journal(sb
));
4662 journal_inode
= ext4_get_journal_inode(sb
, journal_inum
);
4666 journal
= jbd2_journal_init_inode(journal_inode
);
4668 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
4669 iput(journal_inode
);
4672 journal
->j_private
= sb
;
4673 ext4_init_journal_params(sb
, journal
);
4677 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
4680 struct buffer_head
*bh
;
4684 int hblock
, blocksize
;
4685 ext4_fsblk_t sb_block
;
4686 unsigned long offset
;
4687 struct ext4_super_block
*es
;
4688 struct block_device
*bdev
;
4690 BUG_ON(!ext4_has_feature_journal(sb
));
4692 bdev
= ext4_blkdev_get(j_dev
, sb
);
4696 blocksize
= sb
->s_blocksize
;
4697 hblock
= bdev_logical_block_size(bdev
);
4698 if (blocksize
< hblock
) {
4699 ext4_msg(sb
, KERN_ERR
,
4700 "blocksize too small for journal device");
4704 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
4705 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
4706 set_blocksize(bdev
, blocksize
);
4707 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
4708 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
4709 "external journal");
4713 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
4714 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
4715 !(le32_to_cpu(es
->s_feature_incompat
) &
4716 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
4717 ext4_msg(sb
, KERN_ERR
, "external journal has "
4723 if ((le32_to_cpu(es
->s_feature_ro_compat
) &
4724 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
) &&
4725 es
->s_checksum
!= ext4_superblock_csum(sb
, es
)) {
4726 ext4_msg(sb
, KERN_ERR
, "external journal has "
4727 "corrupt superblock");
4732 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
4733 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
4738 len
= ext4_blocks_count(es
);
4739 start
= sb_block
+ 1;
4740 brelse(bh
); /* we're done with the superblock */
4742 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
4743 start
, len
, blocksize
);
4745 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
4748 journal
->j_private
= sb
;
4749 ll_rw_block(REQ_OP_READ
, REQ_META
| REQ_PRIO
, 1, &journal
->j_sb_buffer
);
4750 wait_on_buffer(journal
->j_sb_buffer
);
4751 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
4752 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
4755 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
4756 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
4757 "user (unsupported) - %d",
4758 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
4761 EXT4_SB(sb
)->journal_bdev
= bdev
;
4762 ext4_init_journal_params(sb
, journal
);
4766 jbd2_journal_destroy(journal
);
4768 ext4_blkdev_put(bdev
);
4772 static int ext4_load_journal(struct super_block
*sb
,
4773 struct ext4_super_block
*es
,
4774 unsigned long journal_devnum
)
4777 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
4780 int really_read_only
;
4782 BUG_ON(!ext4_has_feature_journal(sb
));
4784 if (journal_devnum
&&
4785 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4786 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
4787 "numbers have changed");
4788 journal_dev
= new_decode_dev(journal_devnum
);
4790 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
4792 really_read_only
= bdev_read_only(sb
->s_bdev
);
4795 * Are we loading a blank journal or performing recovery after a
4796 * crash? For recovery, we need to check in advance whether we
4797 * can get read-write access to the device.
4799 if (ext4_has_feature_journal_needs_recovery(sb
)) {
4800 if (sb_rdonly(sb
)) {
4801 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
4802 "required on readonly filesystem");
4803 if (really_read_only
) {
4804 ext4_msg(sb
, KERN_ERR
, "write access "
4805 "unavailable, cannot proceed "
4806 "(try mounting with noload)");
4809 ext4_msg(sb
, KERN_INFO
, "write access will "
4810 "be enabled during recovery");
4814 if (journal_inum
&& journal_dev
) {
4815 ext4_msg(sb
, KERN_ERR
, "filesystem has both journal "
4816 "and inode journals!");
4821 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
4824 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
4828 if (!(journal
->j_flags
& JBD2_BARRIER
))
4829 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
4831 if (!ext4_has_feature_journal_needs_recovery(sb
))
4832 err
= jbd2_journal_wipe(journal
, !really_read_only
);
4834 char *save
= kmalloc(EXT4_S_ERR_LEN
, GFP_KERNEL
);
4836 memcpy(save
, ((char *) es
) +
4837 EXT4_S_ERR_START
, EXT4_S_ERR_LEN
);
4838 err
= jbd2_journal_load(journal
);
4840 memcpy(((char *) es
) + EXT4_S_ERR_START
,
4841 save
, EXT4_S_ERR_LEN
);
4846 ext4_msg(sb
, KERN_ERR
, "error loading journal");
4847 jbd2_journal_destroy(journal
);
4851 EXT4_SB(sb
)->s_journal
= journal
;
4852 ext4_clear_journal_err(sb
, es
);
4854 if (!really_read_only
&& journal_devnum
&&
4855 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4856 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
4858 /* Make sure we flush the recovery flag to disk. */
4859 ext4_commit_super(sb
, 1);
4865 static int ext4_commit_super(struct super_block
*sb
, int sync
)
4867 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
4868 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
4871 if (!sbh
|| block_device_ejected(sb
))
4875 * The superblock bh should be mapped, but it might not be if the
4876 * device was hot-removed. Not much we can do but fail the I/O.
4878 if (!buffer_mapped(sbh
))
4882 * If the file system is mounted read-only, don't update the
4883 * superblock write time. This avoids updating the superblock
4884 * write time when we are mounting the root file system
4885 * read/only but we need to replay the journal; at that point,
4886 * for people who are east of GMT and who make their clock
4887 * tick in localtime for Windows bug-for-bug compatibility,
4888 * the clock is set in the future, and this will cause e2fsck
4889 * to complain and force a full file system check.
4891 if (!(sb
->s_flags
& SB_RDONLY
))
4892 ext4_update_tstamp(es
, s_wtime
);
4893 if (sb
->s_bdev
->bd_part
)
4894 es
->s_kbytes_written
=
4895 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
+
4896 ((part_stat_read(sb
->s_bdev
->bd_part
,
4897 sectors
[STAT_WRITE
]) -
4898 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
4900 es
->s_kbytes_written
=
4901 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
);
4902 if (percpu_counter_initialized(&EXT4_SB(sb
)->s_freeclusters_counter
))
4903 ext4_free_blocks_count_set(es
,
4904 EXT4_C2B(EXT4_SB(sb
), percpu_counter_sum_positive(
4905 &EXT4_SB(sb
)->s_freeclusters_counter
)));
4906 if (percpu_counter_initialized(&EXT4_SB(sb
)->s_freeinodes_counter
))
4907 es
->s_free_inodes_count
=
4908 cpu_to_le32(percpu_counter_sum_positive(
4909 &EXT4_SB(sb
)->s_freeinodes_counter
));
4910 BUFFER_TRACE(sbh
, "marking dirty");
4911 ext4_superblock_csum_set(sb
);
4914 if (buffer_write_io_error(sbh
) || !buffer_uptodate(sbh
)) {
4916 * Oh, dear. A previous attempt to write the
4917 * superblock failed. This could happen because the
4918 * USB device was yanked out. Or it could happen to
4919 * be a transient write error and maybe the block will
4920 * be remapped. Nothing we can do but to retry the
4921 * write and hope for the best.
4923 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
4924 "superblock detected");
4925 clear_buffer_write_io_error(sbh
);
4926 set_buffer_uptodate(sbh
);
4928 mark_buffer_dirty(sbh
);
4931 error
= __sync_dirty_buffer(sbh
,
4932 REQ_SYNC
| (test_opt(sb
, BARRIER
) ? REQ_FUA
: 0));
4933 if (buffer_write_io_error(sbh
)) {
4934 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
4936 clear_buffer_write_io_error(sbh
);
4937 set_buffer_uptodate(sbh
);
4944 * Have we just finished recovery? If so, and if we are mounting (or
4945 * remounting) the filesystem readonly, then we will end up with a
4946 * consistent fs on disk. Record that fact.
4948 static void ext4_mark_recovery_complete(struct super_block
*sb
,
4949 struct ext4_super_block
*es
)
4951 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
4953 if (!ext4_has_feature_journal(sb
)) {
4954 BUG_ON(journal
!= NULL
);
4957 jbd2_journal_lock_updates(journal
);
4958 if (jbd2_journal_flush(journal
) < 0)
4961 if (ext4_has_feature_journal_needs_recovery(sb
) && sb_rdonly(sb
)) {
4962 ext4_clear_feature_journal_needs_recovery(sb
);
4963 ext4_commit_super(sb
, 1);
4967 jbd2_journal_unlock_updates(journal
);
4971 * If we are mounting (or read-write remounting) a filesystem whose journal
4972 * has recorded an error from a previous lifetime, move that error to the
4973 * main filesystem now.
4975 static void ext4_clear_journal_err(struct super_block
*sb
,
4976 struct ext4_super_block
*es
)
4982 BUG_ON(!ext4_has_feature_journal(sb
));
4984 journal
= EXT4_SB(sb
)->s_journal
;
4987 * Now check for any error status which may have been recorded in the
4988 * journal by a prior ext4_error() or ext4_abort()
4991 j_errno
= jbd2_journal_errno(journal
);
4995 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
4996 ext4_warning(sb
, "Filesystem error recorded "
4997 "from previous mount: %s", errstr
);
4998 ext4_warning(sb
, "Marking fs in need of filesystem check.");
5000 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
5001 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
5002 ext4_commit_super(sb
, 1);
5004 jbd2_journal_clear_err(journal
);
5005 jbd2_journal_update_sb_errno(journal
);
5010 * Force the running and committing transactions to commit,
5011 * and wait on the commit.
5013 int ext4_force_commit(struct super_block
*sb
)
5020 journal
= EXT4_SB(sb
)->s_journal
;
5021 return ext4_journal_force_commit(journal
);
5024 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
5028 bool needs_barrier
= false;
5029 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
5031 if (unlikely(ext4_forced_shutdown(sbi
)))
5034 trace_ext4_sync_fs(sb
, wait
);
5035 flush_workqueue(sbi
->rsv_conversion_wq
);
5037 * Writeback quota in non-journalled quota case - journalled quota has
5040 dquot_writeback_dquots(sb
, -1);
5042 * Data writeback is possible w/o journal transaction, so barrier must
5043 * being sent at the end of the function. But we can skip it if
5044 * transaction_commit will do it for us.
5046 if (sbi
->s_journal
) {
5047 target
= jbd2_get_latest_transaction(sbi
->s_journal
);
5048 if (wait
&& sbi
->s_journal
->j_flags
& JBD2_BARRIER
&&
5049 !jbd2_trans_will_send_data_barrier(sbi
->s_journal
, target
))
5050 needs_barrier
= true;
5052 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
5054 ret
= jbd2_log_wait_commit(sbi
->s_journal
,
5057 } else if (wait
&& test_opt(sb
, BARRIER
))
5058 needs_barrier
= true;
5059 if (needs_barrier
) {
5061 err
= blkdev_issue_flush(sb
->s_bdev
, GFP_KERNEL
, NULL
);
5070 * LVM calls this function before a (read-only) snapshot is created. This
5071 * gives us a chance to flush the journal completely and mark the fs clean.
5073 * Note that only this function cannot bring a filesystem to be in a clean
5074 * state independently. It relies on upper layer to stop all data & metadata
5077 static int ext4_freeze(struct super_block
*sb
)
5085 journal
= EXT4_SB(sb
)->s_journal
;
5088 /* Now we set up the journal barrier. */
5089 jbd2_journal_lock_updates(journal
);
5092 * Don't clear the needs_recovery flag if we failed to
5093 * flush the journal.
5095 error
= jbd2_journal_flush(journal
);
5099 /* Journal blocked and flushed, clear needs_recovery flag. */
5100 ext4_clear_feature_journal_needs_recovery(sb
);
5103 error
= ext4_commit_super(sb
, 1);
5106 /* we rely on upper layer to stop further updates */
5107 jbd2_journal_unlock_updates(journal
);
5112 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5113 * flag here, even though the filesystem is not technically dirty yet.
5115 static int ext4_unfreeze(struct super_block
*sb
)
5117 if (sb_rdonly(sb
) || ext4_forced_shutdown(EXT4_SB(sb
)))
5120 if (EXT4_SB(sb
)->s_journal
) {
5121 /* Reset the needs_recovery flag before the fs is unlocked. */
5122 ext4_set_feature_journal_needs_recovery(sb
);
5125 ext4_commit_super(sb
, 1);
5130 * Structure to save mount options for ext4_remount's benefit
5132 struct ext4_mount_options
{
5133 unsigned long s_mount_opt
;
5134 unsigned long s_mount_opt2
;
5137 unsigned long s_commit_interval
;
5138 u32 s_min_batch_time
, s_max_batch_time
;
5141 char *s_qf_names
[EXT4_MAXQUOTAS
];
5145 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
5147 struct ext4_super_block
*es
;
5148 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
5149 unsigned long old_sb_flags
;
5150 struct ext4_mount_options old_opts
;
5151 int enable_quota
= 0;
5153 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
5157 char *to_free
[EXT4_MAXQUOTAS
];
5159 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
5161 if (data
&& !orig_data
)
5164 /* Store the original options */
5165 old_sb_flags
= sb
->s_flags
;
5166 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
5167 old_opts
.s_mount_opt2
= sbi
->s_mount_opt2
;
5168 old_opts
.s_resuid
= sbi
->s_resuid
;
5169 old_opts
.s_resgid
= sbi
->s_resgid
;
5170 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
5171 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
5172 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
5174 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
5175 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
5176 if (sbi
->s_qf_names
[i
]) {
5177 char *qf_name
= get_qf_name(sb
, sbi
, i
);
5179 old_opts
.s_qf_names
[i
] = kstrdup(qf_name
, GFP_KERNEL
);
5180 if (!old_opts
.s_qf_names
[i
]) {
5181 for (j
= 0; j
< i
; j
++)
5182 kfree(old_opts
.s_qf_names
[j
]);
5187 old_opts
.s_qf_names
[i
] = NULL
;
5189 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
5190 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
5192 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
, 1)) {
5197 if ((old_opts
.s_mount_opt
& EXT4_MOUNT_JOURNAL_CHECKSUM
) ^
5198 test_opt(sb
, JOURNAL_CHECKSUM
)) {
5199 ext4_msg(sb
, KERN_ERR
, "changing journal_checksum "
5200 "during remount not supported; ignoring");
5201 sbi
->s_mount_opt
^= EXT4_MOUNT_JOURNAL_CHECKSUM
;
5204 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
5205 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
5206 ext4_msg(sb
, KERN_ERR
, "can't mount with "
5207 "both data=journal and delalloc");
5211 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
5212 ext4_msg(sb
, KERN_ERR
, "can't mount with "
5213 "both data=journal and dioread_nolock");
5217 if (test_opt(sb
, DAX
)) {
5218 ext4_msg(sb
, KERN_ERR
, "can't mount with "
5219 "both data=journal and dax");
5223 } else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
) {
5224 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
5225 ext4_msg(sb
, KERN_ERR
, "can't mount with "
5226 "journal_async_commit in data=ordered mode");
5232 if ((sbi
->s_mount_opt
^ old_opts
.s_mount_opt
) & EXT4_MOUNT_NO_MBCACHE
) {
5233 ext4_msg(sb
, KERN_ERR
, "can't enable nombcache during remount");
5238 if ((sbi
->s_mount_opt
^ old_opts
.s_mount_opt
) & EXT4_MOUNT_DAX
) {
5239 ext4_msg(sb
, KERN_WARNING
, "warning: refusing change of "
5240 "dax flag with busy inodes while remounting");
5241 sbi
->s_mount_opt
^= EXT4_MOUNT_DAX
;
5244 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
)
5245 ext4_abort(sb
, "Abort forced by user");
5247 sb
->s_flags
= (sb
->s_flags
& ~SB_POSIXACL
) |
5248 (test_opt(sb
, POSIX_ACL
) ? SB_POSIXACL
: 0);
5252 if (sbi
->s_journal
) {
5253 ext4_init_journal_params(sb
, sbi
->s_journal
);
5254 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
5257 if (*flags
& SB_LAZYTIME
)
5258 sb
->s_flags
|= SB_LAZYTIME
;
5260 if ((bool)(*flags
& SB_RDONLY
) != sb_rdonly(sb
)) {
5261 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
) {
5266 if (*flags
& SB_RDONLY
) {
5267 err
= sync_filesystem(sb
);
5270 err
= dquot_suspend(sb
, -1);
5275 * First of all, the unconditional stuff we have to do
5276 * to disable replay of the journal when we next remount
5278 sb
->s_flags
|= SB_RDONLY
;
5281 * OK, test if we are remounting a valid rw partition
5282 * readonly, and if so set the rdonly flag and then
5283 * mark the partition as valid again.
5285 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
5286 (sbi
->s_mount_state
& EXT4_VALID_FS
))
5287 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
5290 ext4_mark_recovery_complete(sb
, es
);
5292 kthread_stop(sbi
->s_mmp_tsk
);
5294 /* Make sure we can mount this feature set readwrite */
5295 if (ext4_has_feature_readonly(sb
) ||
5296 !ext4_feature_set_ok(sb
, 0)) {
5301 * Make sure the group descriptor checksums
5302 * are sane. If they aren't, refuse to remount r/w.
5304 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
5305 struct ext4_group_desc
*gdp
=
5306 ext4_get_group_desc(sb
, g
, NULL
);
5308 if (!ext4_group_desc_csum_verify(sb
, g
, gdp
)) {
5309 ext4_msg(sb
, KERN_ERR
,
5310 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5311 g
, le16_to_cpu(ext4_group_desc_csum(sb
, g
, gdp
)),
5312 le16_to_cpu(gdp
->bg_checksum
));
5319 * If we have an unprocessed orphan list hanging
5320 * around from a previously readonly bdev mount,
5321 * require a full umount/remount for now.
5323 if (es
->s_last_orphan
) {
5324 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
5325 "remount RDWR because of unprocessed "
5326 "orphan inode list. Please "
5327 "umount/remount instead");
5333 * Mounting a RDONLY partition read-write, so reread
5334 * and store the current valid flag. (It may have
5335 * been changed by e2fsck since we originally mounted
5339 ext4_clear_journal_err(sb
, es
);
5340 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
5342 err
= ext4_setup_super(sb
, es
, 0);
5346 sb
->s_flags
&= ~SB_RDONLY
;
5347 if (ext4_has_feature_mmp(sb
))
5348 if (ext4_multi_mount_protect(sb
,
5349 le64_to_cpu(es
->s_mmp_block
))) {
5358 * Reinitialize lazy itable initialization thread based on
5361 if (sb_rdonly(sb
) || !test_opt(sb
, INIT_INODE_TABLE
))
5362 ext4_unregister_li_request(sb
);
5364 ext4_group_t first_not_zeroed
;
5365 first_not_zeroed
= ext4_has_uninit_itable(sb
);
5366 ext4_register_li_request(sb
, first_not_zeroed
);
5369 ext4_setup_system_zone(sb
);
5370 if (sbi
->s_journal
== NULL
&& !(old_sb_flags
& SB_RDONLY
)) {
5371 err
= ext4_commit_super(sb
, 1);
5377 /* Release old quota file names */
5378 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
5379 kfree(old_opts
.s_qf_names
[i
]);
5381 if (sb_any_quota_suspended(sb
))
5382 dquot_resume(sb
, -1);
5383 else if (ext4_has_feature_quota(sb
)) {
5384 err
= ext4_enable_quotas(sb
);
5391 *flags
= (*flags
& ~SB_LAZYTIME
) | (sb
->s_flags
& SB_LAZYTIME
);
5392 ext4_msg(sb
, KERN_INFO
, "re-mounted. Opts: %s", orig_data
);
5397 sb
->s_flags
= old_sb_flags
;
5398 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
5399 sbi
->s_mount_opt2
= old_opts
.s_mount_opt2
;
5400 sbi
->s_resuid
= old_opts
.s_resuid
;
5401 sbi
->s_resgid
= old_opts
.s_resgid
;
5402 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
5403 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
5404 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
5406 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
5407 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
5408 to_free
[i
] = get_qf_name(sb
, sbi
, i
);
5409 rcu_assign_pointer(sbi
->s_qf_names
[i
], old_opts
.s_qf_names
[i
]);
5412 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
5420 static int ext4_statfs_project(struct super_block
*sb
,
5421 kprojid_t projid
, struct kstatfs
*buf
)
5424 struct dquot
*dquot
;
5428 qid
= make_kqid_projid(projid
);
5429 dquot
= dqget(sb
, qid
);
5431 return PTR_ERR(dquot
);
5432 spin_lock(&dquot
->dq_dqb_lock
);
5434 limit
= (dquot
->dq_dqb
.dqb_bsoftlimit
?
5435 dquot
->dq_dqb
.dqb_bsoftlimit
:
5436 dquot
->dq_dqb
.dqb_bhardlimit
) >> sb
->s_blocksize_bits
;
5437 if (limit
&& buf
->f_blocks
> limit
) {
5438 curblock
= (dquot
->dq_dqb
.dqb_curspace
+
5439 dquot
->dq_dqb
.dqb_rsvspace
) >> sb
->s_blocksize_bits
;
5440 buf
->f_blocks
= limit
;
5441 buf
->f_bfree
= buf
->f_bavail
=
5442 (buf
->f_blocks
> curblock
) ?
5443 (buf
->f_blocks
- curblock
) : 0;
5446 limit
= dquot
->dq_dqb
.dqb_isoftlimit
?
5447 dquot
->dq_dqb
.dqb_isoftlimit
:
5448 dquot
->dq_dqb
.dqb_ihardlimit
;
5449 if (limit
&& buf
->f_files
> limit
) {
5450 buf
->f_files
= limit
;
5452 (buf
->f_files
> dquot
->dq_dqb
.dqb_curinodes
) ?
5453 (buf
->f_files
- dquot
->dq_dqb
.dqb_curinodes
) : 0;
5456 spin_unlock(&dquot
->dq_dqb_lock
);
5462 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
5464 struct super_block
*sb
= dentry
->d_sb
;
5465 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
5466 struct ext4_super_block
*es
= sbi
->s_es
;
5467 ext4_fsblk_t overhead
= 0, resv_blocks
;
5470 resv_blocks
= EXT4_C2B(sbi
, atomic64_read(&sbi
->s_resv_clusters
));
5472 if (!test_opt(sb
, MINIX_DF
))
5473 overhead
= sbi
->s_overhead
;
5475 buf
->f_type
= EXT4_SUPER_MAGIC
;
5476 buf
->f_bsize
= sb
->s_blocksize
;
5477 buf
->f_blocks
= ext4_blocks_count(es
) - EXT4_C2B(sbi
, overhead
);
5478 bfree
= percpu_counter_sum_positive(&sbi
->s_freeclusters_counter
) -
5479 percpu_counter_sum_positive(&sbi
->s_dirtyclusters_counter
);
5480 /* prevent underflow in case that few free space is available */
5481 buf
->f_bfree
= EXT4_C2B(sbi
, max_t(s64
, bfree
, 0));
5482 buf
->f_bavail
= buf
->f_bfree
-
5483 (ext4_r_blocks_count(es
) + resv_blocks
);
5484 if (buf
->f_bfree
< (ext4_r_blocks_count(es
) + resv_blocks
))
5486 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
5487 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
5488 buf
->f_namelen
= EXT4_NAME_LEN
;
5489 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
5490 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
5491 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
5492 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
5495 if (ext4_test_inode_flag(dentry
->d_inode
, EXT4_INODE_PROJINHERIT
) &&
5496 sb_has_quota_limits_enabled(sb
, PRJQUOTA
))
5497 ext4_statfs_project(sb
, EXT4_I(dentry
->d_inode
)->i_projid
, buf
);
5506 * Helper functions so that transaction is started before we acquire dqio_sem
5507 * to keep correct lock ordering of transaction > dqio_sem
5509 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
5511 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_id
.type
];
5514 static int ext4_write_dquot(struct dquot
*dquot
)
5518 struct inode
*inode
;
5520 inode
= dquot_to_inode(dquot
);
5521 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
,
5522 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
5524 return PTR_ERR(handle
);
5525 ret
= dquot_commit(dquot
);
5526 err
= ext4_journal_stop(handle
);
5532 static int ext4_acquire_dquot(struct dquot
*dquot
)
5537 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
5538 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
5540 return PTR_ERR(handle
);
5541 ret
= dquot_acquire(dquot
);
5542 err
= ext4_journal_stop(handle
);
5548 static int ext4_release_dquot(struct dquot
*dquot
)
5553 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
5554 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
5555 if (IS_ERR(handle
)) {
5556 /* Release dquot anyway to avoid endless cycle in dqput() */
5557 dquot_release(dquot
);
5558 return PTR_ERR(handle
);
5560 ret
= dquot_release(dquot
);
5561 err
= ext4_journal_stop(handle
);
5567 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
5569 struct super_block
*sb
= dquot
->dq_sb
;
5570 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
5572 /* Are we journaling quotas? */
5573 if (ext4_has_feature_quota(sb
) ||
5574 sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
5575 dquot_mark_dquot_dirty(dquot
);
5576 return ext4_write_dquot(dquot
);
5578 return dquot_mark_dquot_dirty(dquot
);
5582 static int ext4_write_info(struct super_block
*sb
, int type
)
5587 /* Data block + inode block */
5588 handle
= ext4_journal_start(d_inode(sb
->s_root
), EXT4_HT_QUOTA
, 2);
5590 return PTR_ERR(handle
);
5591 ret
= dquot_commit_info(sb
, type
);
5592 err
= ext4_journal_stop(handle
);
5599 * Turn on quotas during mount time - we need to find
5600 * the quota file and such...
5602 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
5604 return dquot_quota_on_mount(sb
, get_qf_name(sb
, EXT4_SB(sb
), type
),
5605 EXT4_SB(sb
)->s_jquota_fmt
, type
);
5608 static void lockdep_set_quota_inode(struct inode
*inode
, int subclass
)
5610 struct ext4_inode_info
*ei
= EXT4_I(inode
);
5612 /* The first argument of lockdep_set_subclass has to be
5613 * *exactly* the same as the argument to init_rwsem() --- in
5614 * this case, in init_once() --- or lockdep gets unhappy
5615 * because the name of the lock is set using the
5616 * stringification of the argument to init_rwsem().
5618 (void) ei
; /* shut up clang warning if !CONFIG_LOCKDEP */
5619 lockdep_set_subclass(&ei
->i_data_sem
, subclass
);
5623 * Standard function to be called on quota_on
5625 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
5626 const struct path
*path
)
5630 if (!test_opt(sb
, QUOTA
))
5633 /* Quotafile not on the same filesystem? */
5634 if (path
->dentry
->d_sb
!= sb
)
5636 /* Journaling quota? */
5637 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
5638 /* Quotafile not in fs root? */
5639 if (path
->dentry
->d_parent
!= sb
->s_root
)
5640 ext4_msg(sb
, KERN_WARNING
,
5641 "Quota file not on filesystem root. "
5642 "Journaled quota will not work");
5643 sb_dqopt(sb
)->flags
|= DQUOT_NOLIST_DIRTY
;
5646 * Clear the flag just in case mount options changed since
5649 sb_dqopt(sb
)->flags
&= ~DQUOT_NOLIST_DIRTY
;
5653 * When we journal data on quota file, we have to flush journal to see
5654 * all updates to the file when we bypass pagecache...
5656 if (EXT4_SB(sb
)->s_journal
&&
5657 ext4_should_journal_data(d_inode(path
->dentry
))) {
5659 * We don't need to lock updates but journal_flush() could
5660 * otherwise be livelocked...
5662 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
5663 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
5664 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
5669 lockdep_set_quota_inode(path
->dentry
->d_inode
, I_DATA_SEM_QUOTA
);
5670 err
= dquot_quota_on(sb
, type
, format_id
, path
);
5672 lockdep_set_quota_inode(path
->dentry
->d_inode
,
5675 struct inode
*inode
= d_inode(path
->dentry
);
5679 * Set inode flags to prevent userspace from messing with quota
5680 * files. If this fails, we return success anyway since quotas
5681 * are already enabled and this is not a hard failure.
5684 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
, 1);
5687 EXT4_I(inode
)->i_flags
|= EXT4_NOATIME_FL
| EXT4_IMMUTABLE_FL
;
5688 inode_set_flags(inode
, S_NOATIME
| S_IMMUTABLE
,
5689 S_NOATIME
| S_IMMUTABLE
);
5690 ext4_mark_inode_dirty(handle
, inode
);
5691 ext4_journal_stop(handle
);
5693 inode_unlock(inode
);
5698 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
5702 struct inode
*qf_inode
;
5703 unsigned long qf_inums
[EXT4_MAXQUOTAS
] = {
5704 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
5705 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
),
5706 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_prj_quota_inum
)
5709 BUG_ON(!ext4_has_feature_quota(sb
));
5711 if (!qf_inums
[type
])
5714 qf_inode
= ext4_iget(sb
, qf_inums
[type
], EXT4_IGET_SPECIAL
);
5715 if (IS_ERR(qf_inode
)) {
5716 ext4_error(sb
, "Bad quota inode # %lu", qf_inums
[type
]);
5717 return PTR_ERR(qf_inode
);
5720 /* Don't account quota for quota files to avoid recursion */
5721 qf_inode
->i_flags
|= S_NOQUOTA
;
5722 lockdep_set_quota_inode(qf_inode
, I_DATA_SEM_QUOTA
);
5723 err
= dquot_enable(qf_inode
, type
, format_id
, flags
);
5725 lockdep_set_quota_inode(qf_inode
, I_DATA_SEM_NORMAL
);
5731 /* Enable usage tracking for all quota types. */
5732 static int ext4_enable_quotas(struct super_block
*sb
)
5735 unsigned long qf_inums
[EXT4_MAXQUOTAS
] = {
5736 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
5737 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
),
5738 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_prj_quota_inum
)
5740 bool quota_mopt
[EXT4_MAXQUOTAS
] = {
5741 test_opt(sb
, USRQUOTA
),
5742 test_opt(sb
, GRPQUOTA
),
5743 test_opt(sb
, PRJQUOTA
),
5746 sb_dqopt(sb
)->flags
|= DQUOT_QUOTA_SYS_FILE
| DQUOT_NOLIST_DIRTY
;
5747 for (type
= 0; type
< EXT4_MAXQUOTAS
; type
++) {
5748 if (qf_inums
[type
]) {
5749 err
= ext4_quota_enable(sb
, type
, QFMT_VFS_V1
,
5750 DQUOT_USAGE_ENABLED
|
5751 (quota_mopt
[type
] ? DQUOT_LIMITS_ENABLED
: 0));
5754 "Failed to enable quota tracking "
5755 "(type=%d, err=%d). Please run "
5756 "e2fsck to fix.", type
, err
);
5757 for (type
--; type
>= 0; type
--)
5758 dquot_quota_off(sb
, type
);
5767 static int ext4_quota_off(struct super_block
*sb
, int type
)
5769 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5773 /* Force all delayed allocation blocks to be allocated.
5774 * Caller already holds s_umount sem */
5775 if (test_opt(sb
, DELALLOC
))
5776 sync_filesystem(sb
);
5778 if (!inode
|| !igrab(inode
))
5781 err
= dquot_quota_off(sb
, type
);
5782 if (err
|| ext4_has_feature_quota(sb
))
5787 * Update modification times of quota files when userspace can
5788 * start looking at them. If we fail, we return success anyway since
5789 * this is not a hard failure and quotas are already disabled.
5791 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
, 1);
5794 EXT4_I(inode
)->i_flags
&= ~(EXT4_NOATIME_FL
| EXT4_IMMUTABLE_FL
);
5795 inode_set_flags(inode
, 0, S_NOATIME
| S_IMMUTABLE
);
5796 inode
->i_mtime
= inode
->i_ctime
= current_time(inode
);
5797 ext4_mark_inode_dirty(handle
, inode
);
5798 ext4_journal_stop(handle
);
5800 inode_unlock(inode
);
5802 lockdep_set_quota_inode(inode
, I_DATA_SEM_NORMAL
);
5806 return dquot_quota_off(sb
, type
);
5809 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5810 * acquiring the locks... As quota files are never truncated and quota code
5811 * itself serializes the operations (and no one else should touch the files)
5812 * we don't have to be afraid of races */
5813 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
5814 size_t len
, loff_t off
)
5816 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5817 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5818 int offset
= off
& (sb
->s_blocksize
- 1);
5821 struct buffer_head
*bh
;
5822 loff_t i_size
= i_size_read(inode
);
5826 if (off
+len
> i_size
)
5829 while (toread
> 0) {
5830 tocopy
= sb
->s_blocksize
- offset
< toread
?
5831 sb
->s_blocksize
- offset
: toread
;
5832 bh
= ext4_bread(NULL
, inode
, blk
, 0);
5835 if (!bh
) /* A hole? */
5836 memset(data
, 0, tocopy
);
5838 memcpy(data
, bh
->b_data
+offset
, tocopy
);
5848 /* Write to quotafile (we know the transaction is already started and has
5849 * enough credits) */
5850 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
5851 const char *data
, size_t len
, loff_t off
)
5853 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5854 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5855 int err
, offset
= off
& (sb
->s_blocksize
- 1);
5857 struct buffer_head
*bh
;
5858 handle_t
*handle
= journal_current_handle();
5860 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
5861 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5862 " cancelled because transaction is not started",
5863 (unsigned long long)off
, (unsigned long long)len
);
5867 * Since we account only one data block in transaction credits,
5868 * then it is impossible to cross a block boundary.
5870 if (sb
->s_blocksize
- offset
< len
) {
5871 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5872 " cancelled because not block aligned",
5873 (unsigned long long)off
, (unsigned long long)len
);
5878 bh
= ext4_bread(handle
, inode
, blk
,
5879 EXT4_GET_BLOCKS_CREATE
|
5880 EXT4_GET_BLOCKS_METADATA_NOFAIL
);
5881 } while (IS_ERR(bh
) && (PTR_ERR(bh
) == -ENOSPC
) &&
5882 ext4_should_retry_alloc(inode
->i_sb
, &retries
));
5887 BUFFER_TRACE(bh
, "get write access");
5888 err
= ext4_journal_get_write_access(handle
, bh
);
5894 memcpy(bh
->b_data
+offset
, data
, len
);
5895 flush_dcache_page(bh
->b_page
);
5897 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
5900 if (inode
->i_size
< off
+ len
) {
5901 i_size_write(inode
, off
+ len
);
5902 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
5903 ext4_mark_inode_dirty(handle
, inode
);
5908 static int ext4_get_next_id(struct super_block
*sb
, struct kqid
*qid
)
5910 const struct quota_format_ops
*ops
;
5912 if (!sb_has_quota_loaded(sb
, qid
->type
))
5914 ops
= sb_dqopt(sb
)->ops
[qid
->type
];
5915 if (!ops
|| !ops
->get_next_id
)
5917 return dquot_get_next_id(sb
, qid
);
5921 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
5922 const char *dev_name
, void *data
)
5924 return mount_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
);
5927 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5928 static inline void register_as_ext2(void)
5930 int err
= register_filesystem(&ext2_fs_type
);
5933 "EXT4-fs: Unable to register as ext2 (%d)\n", err
);
5936 static inline void unregister_as_ext2(void)
5938 unregister_filesystem(&ext2_fs_type
);
5941 static inline int ext2_feature_set_ok(struct super_block
*sb
)
5943 if (ext4_has_unknown_ext2_incompat_features(sb
))
5947 if (ext4_has_unknown_ext2_ro_compat_features(sb
))
5952 static inline void register_as_ext2(void) { }
5953 static inline void unregister_as_ext2(void) { }
5954 static inline int ext2_feature_set_ok(struct super_block
*sb
) { return 0; }
5957 static inline void register_as_ext3(void)
5959 int err
= register_filesystem(&ext3_fs_type
);
5962 "EXT4-fs: Unable to register as ext3 (%d)\n", err
);
5965 static inline void unregister_as_ext3(void)
5967 unregister_filesystem(&ext3_fs_type
);
5970 static inline int ext3_feature_set_ok(struct super_block
*sb
)
5972 if (ext4_has_unknown_ext3_incompat_features(sb
))
5974 if (!ext4_has_feature_journal(sb
))
5978 if (ext4_has_unknown_ext3_ro_compat_features(sb
))
5983 static struct file_system_type ext4_fs_type
= {
5984 .owner
= THIS_MODULE
,
5986 .mount
= ext4_mount
,
5987 .kill_sb
= kill_block_super
,
5988 .fs_flags
= FS_REQUIRES_DEV
,
5990 MODULE_ALIAS_FS("ext4");
5992 /* Shared across all ext4 file systems */
5993 wait_queue_head_t ext4__ioend_wq
[EXT4_WQ_HASH_SZ
];
5995 static int __init
ext4_init_fs(void)
5999 ratelimit_state_init(&ext4_mount_msg_ratelimit
, 30 * HZ
, 64);
6000 ext4_li_info
= NULL
;
6001 mutex_init(&ext4_li_mtx
);
6003 /* Build-time check for flags consistency */
6004 ext4_check_flag_values();
6006 for (i
= 0; i
< EXT4_WQ_HASH_SZ
; i
++)
6007 init_waitqueue_head(&ext4__ioend_wq
[i
]);
6009 err
= ext4_init_es();
6013 err
= ext4_init_pending();
6017 err
= ext4_init_pageio();
6021 err
= ext4_init_system_zone();
6025 err
= ext4_init_sysfs();
6029 err
= ext4_init_mballoc();
6032 err
= init_inodecache();
6037 err
= register_filesystem(&ext4_fs_type
);
6043 unregister_as_ext2();
6044 unregister_as_ext3();
6045 destroy_inodecache();
6047 ext4_exit_mballoc();
6051 ext4_exit_system_zone();
6055 ext4_exit_pending();
6062 static void __exit
ext4_exit_fs(void)
6064 ext4_destroy_lazyinit_thread();
6065 unregister_as_ext2();
6066 unregister_as_ext3();
6067 unregister_filesystem(&ext4_fs_type
);
6068 destroy_inodecache();
6069 ext4_exit_mballoc();
6071 ext4_exit_system_zone();
6074 ext4_exit_pending();
6077 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6078 MODULE_DESCRIPTION("Fourth Extended Filesystem");
6079 MODULE_LICENSE("GPL");
6080 MODULE_SOFTDEP("pre: crc32c");
6081 module_init(ext4_init_fs
)
6082 module_exit(ext4_exit_fs
)