4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
11 #include <linux/module.h>
12 #include <linux/init.h>
14 #include <linux/statfs.h>
15 #include <linux/buffer_head.h>
16 #include <linux/backing-dev.h>
17 #include <linux/kthread.h>
18 #include <linux/parser.h>
19 #include <linux/mount.h>
20 #include <linux/seq_file.h>
21 #include <linux/proc_fs.h>
22 #include <linux/random.h>
23 #include <linux/exportfs.h>
24 #include <linux/blkdev.h>
25 #include <linux/quotaops.h>
26 #include <linux/f2fs_fs.h>
27 #include <linux/sysfs.h>
28 #include <linux/quota.h>
37 #define CREATE_TRACE_POINTS
38 #include <trace/events/f2fs.h>
40 static struct kmem_cache
*f2fs_inode_cachep
;
42 #ifdef CONFIG_F2FS_FAULT_INJECTION
44 char *fault_name
[FAULT_MAX
] = {
45 [FAULT_KMALLOC
] = "kmalloc",
46 [FAULT_PAGE_ALLOC
] = "page alloc",
47 [FAULT_PAGE_GET
] = "page get",
48 [FAULT_ALLOC_BIO
] = "alloc bio",
49 [FAULT_ALLOC_NID
] = "alloc nid",
50 [FAULT_ORPHAN
] = "orphan",
51 [FAULT_BLOCK
] = "no more block",
52 [FAULT_DIR_DEPTH
] = "too big dir depth",
53 [FAULT_EVICT_INODE
] = "evict_inode fail",
54 [FAULT_TRUNCATE
] = "truncate fail",
55 [FAULT_IO
] = "IO error",
56 [FAULT_CHECKPOINT
] = "checkpoint error",
59 static void f2fs_build_fault_attr(struct f2fs_sb_info
*sbi
,
62 struct f2fs_fault_info
*ffi
= &sbi
->fault_info
;
65 atomic_set(&ffi
->inject_ops
, 0);
66 ffi
->inject_rate
= rate
;
67 ffi
->inject_type
= (1 << FAULT_MAX
) - 1;
69 memset(ffi
, 0, sizeof(struct f2fs_fault_info
));
74 /* f2fs-wide shrinker description */
75 static struct shrinker f2fs_shrinker_info
= {
76 .scan_objects
= f2fs_shrink_scan
,
77 .count_objects
= f2fs_shrink_count
,
78 .seeks
= DEFAULT_SEEKS
,
83 Opt_disable_roll_forward
,
94 Opt_disable_ext_identify
,
97 Opt_inline_xattr_size
,
131 static match_table_t f2fs_tokens
= {
132 {Opt_gc_background
, "background_gc=%s"},
133 {Opt_disable_roll_forward
, "disable_roll_forward"},
134 {Opt_norecovery
, "norecovery"},
135 {Opt_discard
, "discard"},
136 {Opt_nodiscard
, "nodiscard"},
137 {Opt_noheap
, "no_heap"},
139 {Opt_user_xattr
, "user_xattr"},
140 {Opt_nouser_xattr
, "nouser_xattr"},
142 {Opt_noacl
, "noacl"},
143 {Opt_active_logs
, "active_logs=%u"},
144 {Opt_disable_ext_identify
, "disable_ext_identify"},
145 {Opt_inline_xattr
, "inline_xattr"},
146 {Opt_noinline_xattr
, "noinline_xattr"},
147 {Opt_inline_xattr_size
, "inline_xattr_size=%u"},
148 {Opt_inline_data
, "inline_data"},
149 {Opt_inline_dentry
, "inline_dentry"},
150 {Opt_noinline_dentry
, "noinline_dentry"},
151 {Opt_flush_merge
, "flush_merge"},
152 {Opt_noflush_merge
, "noflush_merge"},
153 {Opt_nobarrier
, "nobarrier"},
154 {Opt_fastboot
, "fastboot"},
155 {Opt_extent_cache
, "extent_cache"},
156 {Opt_noextent_cache
, "noextent_cache"},
157 {Opt_noinline_data
, "noinline_data"},
158 {Opt_data_flush
, "data_flush"},
159 {Opt_mode
, "mode=%s"},
160 {Opt_io_size_bits
, "io_bits=%u"},
161 {Opt_fault_injection
, "fault_injection=%u"},
162 {Opt_lazytime
, "lazytime"},
163 {Opt_nolazytime
, "nolazytime"},
164 {Opt_quota
, "quota"},
165 {Opt_noquota
, "noquota"},
166 {Opt_usrquota
, "usrquota"},
167 {Opt_grpquota
, "grpquota"},
168 {Opt_prjquota
, "prjquota"},
169 {Opt_usrjquota
, "usrjquota=%s"},
170 {Opt_grpjquota
, "grpjquota=%s"},
171 {Opt_prjjquota
, "prjjquota=%s"},
172 {Opt_offusrjquota
, "usrjquota="},
173 {Opt_offgrpjquota
, "grpjquota="},
174 {Opt_offprjjquota
, "prjjquota="},
175 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
176 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
177 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
181 void f2fs_msg(struct super_block
*sb
, const char *level
, const char *fmt
, ...)
183 struct va_format vaf
;
189 printk_ratelimited("%sF2FS-fs (%s): %pV\n", level
, sb
->s_id
, &vaf
);
193 static void init_once(void *foo
)
195 struct f2fs_inode_info
*fi
= (struct f2fs_inode_info
*) foo
;
197 inode_init_once(&fi
->vfs_inode
);
201 static const char * const quotatypes
[] = INITQFNAMES
;
202 #define QTYPE2NAME(t) (quotatypes[t])
203 static int f2fs_set_qf_name(struct super_block
*sb
, int qtype
,
206 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
210 if (sb_any_quota_loaded(sb
) && !sbi
->s_qf_names
[qtype
]) {
211 f2fs_msg(sb
, KERN_ERR
,
212 "Cannot change journaled "
213 "quota options when quota turned on");
216 if (f2fs_sb_has_quota_ino(sb
)) {
217 f2fs_msg(sb
, KERN_INFO
,
218 "QUOTA feature is enabled, so ignore qf_name");
222 qname
= match_strdup(args
);
224 f2fs_msg(sb
, KERN_ERR
,
225 "Not enough memory for storing quotafile name");
228 if (sbi
->s_qf_names
[qtype
]) {
229 if (strcmp(sbi
->s_qf_names
[qtype
], qname
) == 0)
232 f2fs_msg(sb
, KERN_ERR
,
233 "%s quota file already specified",
237 if (strchr(qname
, '/')) {
238 f2fs_msg(sb
, KERN_ERR
,
239 "quotafile must be on filesystem root");
242 sbi
->s_qf_names
[qtype
] = qname
;
250 static int f2fs_clear_qf_name(struct super_block
*sb
, int qtype
)
252 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
254 if (sb_any_quota_loaded(sb
) && sbi
->s_qf_names
[qtype
]) {
255 f2fs_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
256 " when quota turned on");
259 kfree(sbi
->s_qf_names
[qtype
]);
260 sbi
->s_qf_names
[qtype
] = NULL
;
264 static int f2fs_check_quota_options(struct f2fs_sb_info
*sbi
)
267 * We do the test below only for project quotas. 'usrquota' and
268 * 'grpquota' mount options are allowed even without quota feature
269 * to support legacy quotas in quota files.
271 if (test_opt(sbi
, PRJQUOTA
) && !f2fs_sb_has_project_quota(sbi
->sb
)) {
272 f2fs_msg(sbi
->sb
, KERN_ERR
, "Project quota feature not enabled. "
273 "Cannot enable project quota enforcement.");
276 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
] ||
277 sbi
->s_qf_names
[PRJQUOTA
]) {
278 if (test_opt(sbi
, USRQUOTA
) && sbi
->s_qf_names
[USRQUOTA
])
279 clear_opt(sbi
, USRQUOTA
);
281 if (test_opt(sbi
, GRPQUOTA
) && sbi
->s_qf_names
[GRPQUOTA
])
282 clear_opt(sbi
, GRPQUOTA
);
284 if (test_opt(sbi
, PRJQUOTA
) && sbi
->s_qf_names
[PRJQUOTA
])
285 clear_opt(sbi
, PRJQUOTA
);
287 if (test_opt(sbi
, GRPQUOTA
) || test_opt(sbi
, USRQUOTA
) ||
288 test_opt(sbi
, PRJQUOTA
)) {
289 f2fs_msg(sbi
->sb
, KERN_ERR
, "old and new quota "
294 if (!sbi
->s_jquota_fmt
) {
295 f2fs_msg(sbi
->sb
, KERN_ERR
, "journaled quota format "
301 if (f2fs_sb_has_quota_ino(sbi
->sb
) && sbi
->s_jquota_fmt
) {
302 f2fs_msg(sbi
->sb
, KERN_INFO
,
303 "QUOTA feature is enabled, so ignore jquota_fmt");
304 sbi
->s_jquota_fmt
= 0;
306 if (f2fs_sb_has_quota_ino(sbi
->sb
) && sb_rdonly(sbi
->sb
)) {
307 f2fs_msg(sbi
->sb
, KERN_INFO
,
308 "Filesystem with quota feature cannot be mounted RDWR "
309 "without CONFIG_QUOTA");
316 static int parse_options(struct super_block
*sb
, char *options
)
318 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
319 struct request_queue
*q
;
320 substring_t args
[MAX_OPT_ARGS
];
330 while ((p
= strsep(&options
, ",")) != NULL
) {
335 * Initialize args struct so we know whether arg was
336 * found; some options take optional arguments.
338 args
[0].to
= args
[0].from
= NULL
;
339 token
= match_token(p
, f2fs_tokens
, args
);
342 case Opt_gc_background
:
343 name
= match_strdup(&args
[0]);
347 if (strlen(name
) == 2 && !strncmp(name
, "on", 2)) {
349 clear_opt(sbi
, FORCE_FG_GC
);
350 } else if (strlen(name
) == 3 && !strncmp(name
, "off", 3)) {
351 clear_opt(sbi
, BG_GC
);
352 clear_opt(sbi
, FORCE_FG_GC
);
353 } else if (strlen(name
) == 4 && !strncmp(name
, "sync", 4)) {
355 set_opt(sbi
, FORCE_FG_GC
);
362 case Opt_disable_roll_forward
:
363 set_opt(sbi
, DISABLE_ROLL_FORWARD
);
366 /* this option mounts f2fs with ro */
367 set_opt(sbi
, DISABLE_ROLL_FORWARD
);
368 if (!f2fs_readonly(sb
))
372 q
= bdev_get_queue(sb
->s_bdev
);
373 if (blk_queue_discard(q
)) {
374 set_opt(sbi
, DISCARD
);
375 } else if (!f2fs_sb_mounted_blkzoned(sb
)) {
376 f2fs_msg(sb
, KERN_WARNING
,
377 "mounting with \"discard\" option, but "
378 "the device does not support discard");
382 if (f2fs_sb_mounted_blkzoned(sb
)) {
383 f2fs_msg(sb
, KERN_WARNING
,
384 "discard is required for zoned block devices");
387 clear_opt(sbi
, DISCARD
);
390 set_opt(sbi
, NOHEAP
);
393 clear_opt(sbi
, NOHEAP
);
395 #ifdef CONFIG_F2FS_FS_XATTR
397 set_opt(sbi
, XATTR_USER
);
399 case Opt_nouser_xattr
:
400 clear_opt(sbi
, XATTR_USER
);
402 case Opt_inline_xattr
:
403 set_opt(sbi
, INLINE_XATTR
);
405 case Opt_noinline_xattr
:
406 clear_opt(sbi
, INLINE_XATTR
);
408 case Opt_inline_xattr_size
:
409 if (args
->from
&& match_int(args
, &arg
))
411 set_opt(sbi
, INLINE_XATTR_SIZE
);
412 sbi
->inline_xattr_size
= arg
;
416 f2fs_msg(sb
, KERN_INFO
,
417 "user_xattr options not supported");
419 case Opt_nouser_xattr
:
420 f2fs_msg(sb
, KERN_INFO
,
421 "nouser_xattr options not supported");
423 case Opt_inline_xattr
:
424 f2fs_msg(sb
, KERN_INFO
,
425 "inline_xattr options not supported");
427 case Opt_noinline_xattr
:
428 f2fs_msg(sb
, KERN_INFO
,
429 "noinline_xattr options not supported");
432 #ifdef CONFIG_F2FS_FS_POSIX_ACL
434 set_opt(sbi
, POSIX_ACL
);
437 clear_opt(sbi
, POSIX_ACL
);
441 f2fs_msg(sb
, KERN_INFO
, "acl options not supported");
444 f2fs_msg(sb
, KERN_INFO
, "noacl options not supported");
447 case Opt_active_logs
:
448 if (args
->from
&& match_int(args
, &arg
))
450 if (arg
!= 2 && arg
!= 4 && arg
!= NR_CURSEG_TYPE
)
452 sbi
->active_logs
= arg
;
454 case Opt_disable_ext_identify
:
455 set_opt(sbi
, DISABLE_EXT_IDENTIFY
);
457 case Opt_inline_data
:
458 set_opt(sbi
, INLINE_DATA
);
460 case Opt_inline_dentry
:
461 set_opt(sbi
, INLINE_DENTRY
);
463 case Opt_noinline_dentry
:
464 clear_opt(sbi
, INLINE_DENTRY
);
466 case Opt_flush_merge
:
467 set_opt(sbi
, FLUSH_MERGE
);
469 case Opt_noflush_merge
:
470 clear_opt(sbi
, FLUSH_MERGE
);
473 set_opt(sbi
, NOBARRIER
);
476 set_opt(sbi
, FASTBOOT
);
478 case Opt_extent_cache
:
479 set_opt(sbi
, EXTENT_CACHE
);
481 case Opt_noextent_cache
:
482 clear_opt(sbi
, EXTENT_CACHE
);
484 case Opt_noinline_data
:
485 clear_opt(sbi
, INLINE_DATA
);
488 set_opt(sbi
, DATA_FLUSH
);
491 name
= match_strdup(&args
[0]);
495 if (strlen(name
) == 8 &&
496 !strncmp(name
, "adaptive", 8)) {
497 if (f2fs_sb_mounted_blkzoned(sb
)) {
498 f2fs_msg(sb
, KERN_WARNING
,
499 "adaptive mode is not allowed with "
500 "zoned block device feature");
504 set_opt_mode(sbi
, F2FS_MOUNT_ADAPTIVE
);
505 } else if (strlen(name
) == 3 &&
506 !strncmp(name
, "lfs", 3)) {
507 set_opt_mode(sbi
, F2FS_MOUNT_LFS
);
514 case Opt_io_size_bits
:
515 if (args
->from
&& match_int(args
, &arg
))
517 if (arg
> __ilog2_u32(BIO_MAX_PAGES
)) {
518 f2fs_msg(sb
, KERN_WARNING
,
519 "Not support %d, larger than %d",
520 1 << arg
, BIO_MAX_PAGES
);
523 sbi
->write_io_size_bits
= arg
;
525 case Opt_fault_injection
:
526 if (args
->from
&& match_int(args
, &arg
))
528 #ifdef CONFIG_F2FS_FAULT_INJECTION
529 f2fs_build_fault_attr(sbi
, arg
);
530 set_opt(sbi
, FAULT_INJECTION
);
532 f2fs_msg(sb
, KERN_INFO
,
533 "FAULT_INJECTION was not selected");
537 sb
->s_flags
|= MS_LAZYTIME
;
540 sb
->s_flags
&= ~MS_LAZYTIME
;
545 set_opt(sbi
, USRQUOTA
);
548 set_opt(sbi
, GRPQUOTA
);
551 set_opt(sbi
, PRJQUOTA
);
554 ret
= f2fs_set_qf_name(sb
, USRQUOTA
, &args
[0]);
559 ret
= f2fs_set_qf_name(sb
, GRPQUOTA
, &args
[0]);
564 ret
= f2fs_set_qf_name(sb
, PRJQUOTA
, &args
[0]);
568 case Opt_offusrjquota
:
569 ret
= f2fs_clear_qf_name(sb
, USRQUOTA
);
573 case Opt_offgrpjquota
:
574 ret
= f2fs_clear_qf_name(sb
, GRPQUOTA
);
578 case Opt_offprjjquota
:
579 ret
= f2fs_clear_qf_name(sb
, PRJQUOTA
);
583 case Opt_jqfmt_vfsold
:
584 sbi
->s_jquota_fmt
= QFMT_VFS_OLD
;
586 case Opt_jqfmt_vfsv0
:
587 sbi
->s_jquota_fmt
= QFMT_VFS_V0
;
589 case Opt_jqfmt_vfsv1
:
590 sbi
->s_jquota_fmt
= QFMT_VFS_V1
;
593 clear_opt(sbi
, QUOTA
);
594 clear_opt(sbi
, USRQUOTA
);
595 clear_opt(sbi
, GRPQUOTA
);
596 clear_opt(sbi
, PRJQUOTA
);
606 case Opt_offusrjquota
:
607 case Opt_offgrpjquota
:
608 case Opt_offprjjquota
:
609 case Opt_jqfmt_vfsold
:
610 case Opt_jqfmt_vfsv0
:
611 case Opt_jqfmt_vfsv1
:
613 f2fs_msg(sb
, KERN_INFO
,
614 "quota operations not supported");
618 f2fs_msg(sb
, KERN_ERR
,
619 "Unrecognized mount option \"%s\" or missing value",
625 if (f2fs_check_quota_options(sbi
))
629 if (F2FS_IO_SIZE_BITS(sbi
) && !test_opt(sbi
, LFS
)) {
630 f2fs_msg(sb
, KERN_ERR
,
631 "Should set mode=lfs with %uKB-sized IO",
632 F2FS_IO_SIZE_KB(sbi
));
636 if (test_opt(sbi
, INLINE_XATTR_SIZE
)) {
637 if (!test_opt(sbi
, INLINE_XATTR
)) {
638 f2fs_msg(sb
, KERN_ERR
,
639 "inline_xattr_size option should be "
640 "set with inline_xattr option");
643 if (!sbi
->inline_xattr_size
||
644 sbi
->inline_xattr_size
>= DEF_ADDRS_PER_INODE
-
645 F2FS_TOTAL_EXTRA_ATTR_SIZE
-
646 DEF_INLINE_RESERVED_SIZE
-
647 DEF_MIN_INLINE_SIZE
) {
648 f2fs_msg(sb
, KERN_ERR
,
649 "inline xattr size is out of range");
656 static struct inode
*f2fs_alloc_inode(struct super_block
*sb
)
658 struct f2fs_inode_info
*fi
;
660 fi
= kmem_cache_alloc(f2fs_inode_cachep
, GFP_F2FS_ZERO
);
664 init_once((void *) fi
);
666 /* Initialize f2fs-specific inode info */
667 atomic_set(&fi
->dirty_pages
, 0);
668 fi
->i_current_depth
= 1;
670 init_rwsem(&fi
->i_sem
);
671 INIT_LIST_HEAD(&fi
->dirty_list
);
672 INIT_LIST_HEAD(&fi
->gdirty_list
);
673 INIT_LIST_HEAD(&fi
->inmem_ilist
);
674 INIT_LIST_HEAD(&fi
->inmem_pages
);
675 mutex_init(&fi
->inmem_lock
);
676 init_rwsem(&fi
->dio_rwsem
[READ
]);
677 init_rwsem(&fi
->dio_rwsem
[WRITE
]);
678 init_rwsem(&fi
->i_mmap_sem
);
679 init_rwsem(&fi
->i_xattr_sem
);
682 memset(&fi
->i_dquot
, 0, sizeof(fi
->i_dquot
));
683 fi
->i_reserved_quota
= 0;
685 /* Will be used by directory only */
686 fi
->i_dir_level
= F2FS_SB(sb
)->dir_level
;
688 return &fi
->vfs_inode
;
691 static int f2fs_drop_inode(struct inode
*inode
)
695 * This is to avoid a deadlock condition like below.
696 * writeback_single_inode(inode)
697 * - f2fs_write_data_page
698 * - f2fs_gc -> iput -> evict
699 * - inode_wait_for_writeback(inode)
701 if ((!inode_unhashed(inode
) && inode
->i_state
& I_SYNC
)) {
702 if (!inode
->i_nlink
&& !is_bad_inode(inode
)) {
703 /* to avoid evict_inode call simultaneously */
704 atomic_inc(&inode
->i_count
);
705 spin_unlock(&inode
->i_lock
);
707 /* some remained atomic pages should discarded */
708 if (f2fs_is_atomic_file(inode
))
709 drop_inmem_pages(inode
);
711 /* should remain fi->extent_tree for writepage */
712 f2fs_destroy_extent_node(inode
);
714 sb_start_intwrite(inode
->i_sb
);
715 f2fs_i_size_write(inode
, 0);
717 if (F2FS_HAS_BLOCKS(inode
))
718 f2fs_truncate(inode
);
720 sb_end_intwrite(inode
->i_sb
);
722 spin_lock(&inode
->i_lock
);
723 atomic_dec(&inode
->i_count
);
725 trace_f2fs_drop_inode(inode
, 0);
728 ret
= generic_drop_inode(inode
);
729 trace_f2fs_drop_inode(inode
, ret
);
733 int f2fs_inode_dirtied(struct inode
*inode
, bool sync
)
735 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
738 spin_lock(&sbi
->inode_lock
[DIRTY_META
]);
739 if (is_inode_flag_set(inode
, FI_DIRTY_INODE
)) {
742 set_inode_flag(inode
, FI_DIRTY_INODE
);
743 stat_inc_dirty_inode(sbi
, DIRTY_META
);
745 if (sync
&& list_empty(&F2FS_I(inode
)->gdirty_list
)) {
746 list_add_tail(&F2FS_I(inode
)->gdirty_list
,
747 &sbi
->inode_list
[DIRTY_META
]);
748 inc_page_count(sbi
, F2FS_DIRTY_IMETA
);
750 spin_unlock(&sbi
->inode_lock
[DIRTY_META
]);
754 void f2fs_inode_synced(struct inode
*inode
)
756 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
758 spin_lock(&sbi
->inode_lock
[DIRTY_META
]);
759 if (!is_inode_flag_set(inode
, FI_DIRTY_INODE
)) {
760 spin_unlock(&sbi
->inode_lock
[DIRTY_META
]);
763 if (!list_empty(&F2FS_I(inode
)->gdirty_list
)) {
764 list_del_init(&F2FS_I(inode
)->gdirty_list
);
765 dec_page_count(sbi
, F2FS_DIRTY_IMETA
);
767 clear_inode_flag(inode
, FI_DIRTY_INODE
);
768 clear_inode_flag(inode
, FI_AUTO_RECOVER
);
769 stat_dec_dirty_inode(F2FS_I_SB(inode
), DIRTY_META
);
770 spin_unlock(&sbi
->inode_lock
[DIRTY_META
]);
774 * f2fs_dirty_inode() is called from __mark_inode_dirty()
776 * We should call set_dirty_inode to write the dirty inode through write_inode.
778 static void f2fs_dirty_inode(struct inode
*inode
, int flags
)
780 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
782 if (inode
->i_ino
== F2FS_NODE_INO(sbi
) ||
783 inode
->i_ino
== F2FS_META_INO(sbi
))
786 if (flags
== I_DIRTY_TIME
)
789 if (is_inode_flag_set(inode
, FI_AUTO_RECOVER
))
790 clear_inode_flag(inode
, FI_AUTO_RECOVER
);
792 f2fs_inode_dirtied(inode
, false);
795 static void f2fs_i_callback(struct rcu_head
*head
)
797 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
798 kmem_cache_free(f2fs_inode_cachep
, F2FS_I(inode
));
801 static void f2fs_destroy_inode(struct inode
*inode
)
803 call_rcu(&inode
->i_rcu
, f2fs_i_callback
);
806 static void destroy_percpu_info(struct f2fs_sb_info
*sbi
)
808 percpu_counter_destroy(&sbi
->alloc_valid_block_count
);
809 percpu_counter_destroy(&sbi
->total_valid_inode_count
);
812 static void destroy_device_list(struct f2fs_sb_info
*sbi
)
816 for (i
= 0; i
< sbi
->s_ndevs
; i
++) {
817 blkdev_put(FDEV(i
).bdev
, FMODE_EXCL
);
818 #ifdef CONFIG_BLK_DEV_ZONED
819 kfree(FDEV(i
).blkz_type
);
825 static void f2fs_put_super(struct super_block
*sb
)
827 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
831 f2fs_quota_off_umount(sb
);
833 /* prevent remaining shrinker jobs */
834 mutex_lock(&sbi
->umount_mutex
);
837 * We don't need to do checkpoint when superblock is clean.
838 * But, the previous checkpoint was not done by umount, it needs to do
839 * clean checkpoint again.
841 if (is_sbi_flag_set(sbi
, SBI_IS_DIRTY
) ||
842 !is_set_ckpt_flags(sbi
, CP_UMOUNT_FLAG
)) {
843 struct cp_control cpc
= {
846 write_checkpoint(sbi
, &cpc
);
849 /* be sure to wait for any on-going discard commands */
850 dropped
= f2fs_wait_discard_bios(sbi
);
852 if (f2fs_discard_en(sbi
) && !sbi
->discard_blks
&& !dropped
) {
853 struct cp_control cpc
= {
854 .reason
= CP_UMOUNT
| CP_TRIMMED
,
856 write_checkpoint(sbi
, &cpc
);
859 /* write_checkpoint can update stat informaion */
860 f2fs_destroy_stats(sbi
);
863 * normally superblock is clean, so we need to release this.
864 * In addition, EIO will skip do checkpoint, we need this as well.
866 release_ino_entry(sbi
, true);
868 f2fs_leave_shrinker(sbi
);
869 mutex_unlock(&sbi
->umount_mutex
);
871 /* our cp_error case, we can wait for any writeback page */
872 f2fs_flush_merged_writes(sbi
);
874 iput(sbi
->node_inode
);
875 iput(sbi
->meta_inode
);
877 /* destroy f2fs internal modules */
878 destroy_node_manager(sbi
);
879 destroy_segment_manager(sbi
);
883 f2fs_unregister_sysfs(sbi
);
885 sb
->s_fs_info
= NULL
;
886 if (sbi
->s_chksum_driver
)
887 crypto_free_shash(sbi
->s_chksum_driver
);
888 kfree(sbi
->raw_super
);
890 destroy_device_list(sbi
);
891 if (sbi
->write_io_dummy
)
892 mempool_destroy(sbi
->write_io_dummy
);
894 for (i
= 0; i
< MAXQUOTAS
; i
++)
895 kfree(sbi
->s_qf_names
[i
]);
897 destroy_percpu_info(sbi
);
898 for (i
= 0; i
< NR_PAGE_TYPE
; i
++)
899 kfree(sbi
->write_io
[i
]);
903 int f2fs_sync_fs(struct super_block
*sb
, int sync
)
905 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
908 if (unlikely(f2fs_cp_error(sbi
)))
911 trace_f2fs_sync_fs(sb
, sync
);
913 if (unlikely(is_sbi_flag_set(sbi
, SBI_POR_DOING
)))
917 struct cp_control cpc
;
919 cpc
.reason
= __get_cp_reason(sbi
);
921 mutex_lock(&sbi
->gc_mutex
);
922 err
= write_checkpoint(sbi
, &cpc
);
923 mutex_unlock(&sbi
->gc_mutex
);
925 f2fs_trace_ios(NULL
, 1);
930 static int f2fs_freeze(struct super_block
*sb
)
932 if (f2fs_readonly(sb
))
935 /* IO error happened before */
936 if (unlikely(f2fs_cp_error(F2FS_SB(sb
))))
939 /* must be clean, since sync_filesystem() was already called */
940 if (is_sbi_flag_set(F2FS_SB(sb
), SBI_IS_DIRTY
))
945 static int f2fs_unfreeze(struct super_block
*sb
)
951 static int f2fs_statfs_project(struct super_block
*sb
,
952 kprojid_t projid
, struct kstatfs
*buf
)
959 qid
= make_kqid_projid(projid
);
960 dquot
= dqget(sb
, qid
);
962 return PTR_ERR(dquot
);
963 spin_lock(&dq_data_lock
);
965 limit
= (dquot
->dq_dqb
.dqb_bsoftlimit
?
966 dquot
->dq_dqb
.dqb_bsoftlimit
:
967 dquot
->dq_dqb
.dqb_bhardlimit
) >> sb
->s_blocksize_bits
;
968 if (limit
&& buf
->f_blocks
> limit
) {
969 curblock
= dquot
->dq_dqb
.dqb_curspace
>> sb
->s_blocksize_bits
;
970 buf
->f_blocks
= limit
;
971 buf
->f_bfree
= buf
->f_bavail
=
972 (buf
->f_blocks
> curblock
) ?
973 (buf
->f_blocks
- curblock
) : 0;
976 limit
= dquot
->dq_dqb
.dqb_isoftlimit
?
977 dquot
->dq_dqb
.dqb_isoftlimit
:
978 dquot
->dq_dqb
.dqb_ihardlimit
;
979 if (limit
&& buf
->f_files
> limit
) {
980 buf
->f_files
= limit
;
982 (buf
->f_files
> dquot
->dq_dqb
.dqb_curinodes
) ?
983 (buf
->f_files
- dquot
->dq_dqb
.dqb_curinodes
) : 0;
986 spin_unlock(&dq_data_lock
);
992 static int f2fs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
994 struct super_block
*sb
= dentry
->d_sb
;
995 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
996 u64 id
= huge_encode_dev(sb
->s_bdev
->bd_dev
);
997 block_t total_count
, user_block_count
, start_count
, ovp_count
;
998 u64 avail_node_count
;
1000 total_count
= le64_to_cpu(sbi
->raw_super
->block_count
);
1001 user_block_count
= sbi
->user_block_count
;
1002 start_count
= le32_to_cpu(sbi
->raw_super
->segment0_blkaddr
);
1003 ovp_count
= SM_I(sbi
)->ovp_segments
<< sbi
->log_blocks_per_seg
;
1004 buf
->f_type
= F2FS_SUPER_MAGIC
;
1005 buf
->f_bsize
= sbi
->blocksize
;
1007 buf
->f_blocks
= total_count
- start_count
;
1008 buf
->f_bfree
= user_block_count
- valid_user_blocks(sbi
) + ovp_count
;
1009 buf
->f_bavail
= user_block_count
- valid_user_blocks(sbi
) -
1010 sbi
->current_reserved_blocks
;
1012 avail_node_count
= sbi
->total_node_count
- F2FS_RESERVED_NODE_NUM
;
1014 if (avail_node_count
> user_block_count
) {
1015 buf
->f_files
= user_block_count
;
1016 buf
->f_ffree
= buf
->f_bavail
;
1018 buf
->f_files
= avail_node_count
;
1019 buf
->f_ffree
= min(avail_node_count
- valid_node_count(sbi
),
1023 buf
->f_namelen
= F2FS_NAME_LEN
;
1024 buf
->f_fsid
.val
[0] = (u32
)id
;
1025 buf
->f_fsid
.val
[1] = (u32
)(id
>> 32);
1028 if (is_inode_flag_set(dentry
->d_inode
, FI_PROJ_INHERIT
) &&
1029 sb_has_quota_limits_enabled(sb
, PRJQUOTA
)) {
1030 f2fs_statfs_project(sb
, F2FS_I(dentry
->d_inode
)->i_projid
, buf
);
1036 static inline void f2fs_show_quota_options(struct seq_file
*seq
,
1037 struct super_block
*sb
)
1040 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
1042 if (sbi
->s_jquota_fmt
) {
1045 switch (sbi
->s_jquota_fmt
) {
1056 seq_printf(seq
, ",jqfmt=%s", fmtname
);
1059 if (sbi
->s_qf_names
[USRQUOTA
])
1060 seq_show_option(seq
, "usrjquota", sbi
->s_qf_names
[USRQUOTA
]);
1062 if (sbi
->s_qf_names
[GRPQUOTA
])
1063 seq_show_option(seq
, "grpjquota", sbi
->s_qf_names
[GRPQUOTA
]);
1065 if (sbi
->s_qf_names
[PRJQUOTA
])
1066 seq_show_option(seq
, "prjjquota", sbi
->s_qf_names
[PRJQUOTA
]);
1070 static int f2fs_show_options(struct seq_file
*seq
, struct dentry
*root
)
1072 struct f2fs_sb_info
*sbi
= F2FS_SB(root
->d_sb
);
1074 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, BG_GC
)) {
1075 if (test_opt(sbi
, FORCE_FG_GC
))
1076 seq_printf(seq
, ",background_gc=%s", "sync");
1078 seq_printf(seq
, ",background_gc=%s", "on");
1080 seq_printf(seq
, ",background_gc=%s", "off");
1082 if (test_opt(sbi
, DISABLE_ROLL_FORWARD
))
1083 seq_puts(seq
, ",disable_roll_forward");
1084 if (test_opt(sbi
, DISCARD
))
1085 seq_puts(seq
, ",discard");
1086 if (test_opt(sbi
, NOHEAP
))
1087 seq_puts(seq
, ",no_heap");
1089 seq_puts(seq
, ",heap");
1090 #ifdef CONFIG_F2FS_FS_XATTR
1091 if (test_opt(sbi
, XATTR_USER
))
1092 seq_puts(seq
, ",user_xattr");
1094 seq_puts(seq
, ",nouser_xattr");
1095 if (test_opt(sbi
, INLINE_XATTR
))
1096 seq_puts(seq
, ",inline_xattr");
1098 seq_puts(seq
, ",noinline_xattr");
1099 if (test_opt(sbi
, INLINE_XATTR_SIZE
))
1100 seq_printf(seq
, ",inline_xattr_size=%u",
1101 sbi
->inline_xattr_size
);
1103 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1104 if (test_opt(sbi
, POSIX_ACL
))
1105 seq_puts(seq
, ",acl");
1107 seq_puts(seq
, ",noacl");
1109 if (test_opt(sbi
, DISABLE_EXT_IDENTIFY
))
1110 seq_puts(seq
, ",disable_ext_identify");
1111 if (test_opt(sbi
, INLINE_DATA
))
1112 seq_puts(seq
, ",inline_data");
1114 seq_puts(seq
, ",noinline_data");
1115 if (test_opt(sbi
, INLINE_DENTRY
))
1116 seq_puts(seq
, ",inline_dentry");
1118 seq_puts(seq
, ",noinline_dentry");
1119 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, FLUSH_MERGE
))
1120 seq_puts(seq
, ",flush_merge");
1121 if (test_opt(sbi
, NOBARRIER
))
1122 seq_puts(seq
, ",nobarrier");
1123 if (test_opt(sbi
, FASTBOOT
))
1124 seq_puts(seq
, ",fastboot");
1125 if (test_opt(sbi
, EXTENT_CACHE
))
1126 seq_puts(seq
, ",extent_cache");
1128 seq_puts(seq
, ",noextent_cache");
1129 if (test_opt(sbi
, DATA_FLUSH
))
1130 seq_puts(seq
, ",data_flush");
1132 seq_puts(seq
, ",mode=");
1133 if (test_opt(sbi
, ADAPTIVE
))
1134 seq_puts(seq
, "adaptive");
1135 else if (test_opt(sbi
, LFS
))
1136 seq_puts(seq
, "lfs");
1137 seq_printf(seq
, ",active_logs=%u", sbi
->active_logs
);
1138 if (F2FS_IO_SIZE_BITS(sbi
))
1139 seq_printf(seq
, ",io_size=%uKB", F2FS_IO_SIZE_KB(sbi
));
1140 #ifdef CONFIG_F2FS_FAULT_INJECTION
1141 if (test_opt(sbi
, FAULT_INJECTION
))
1142 seq_printf(seq
, ",fault_injection=%u",
1143 sbi
->fault_info
.inject_rate
);
1146 if (test_opt(sbi
, QUOTA
))
1147 seq_puts(seq
, ",quota");
1148 if (test_opt(sbi
, USRQUOTA
))
1149 seq_puts(seq
, ",usrquota");
1150 if (test_opt(sbi
, GRPQUOTA
))
1151 seq_puts(seq
, ",grpquota");
1152 if (test_opt(sbi
, PRJQUOTA
))
1153 seq_puts(seq
, ",prjquota");
1155 f2fs_show_quota_options(seq
, sbi
->sb
);
1160 static void default_options(struct f2fs_sb_info
*sbi
)
1162 /* init some FS parameters */
1163 sbi
->active_logs
= NR_CURSEG_TYPE
;
1164 sbi
->inline_xattr_size
= DEFAULT_INLINE_XATTR_ADDRS
;
1166 set_opt(sbi
, BG_GC
);
1167 set_opt(sbi
, INLINE_XATTR
);
1168 set_opt(sbi
, INLINE_DATA
);
1169 set_opt(sbi
, INLINE_DENTRY
);
1170 set_opt(sbi
, EXTENT_CACHE
);
1171 set_opt(sbi
, NOHEAP
);
1172 sbi
->sb
->s_flags
|= MS_LAZYTIME
;
1173 set_opt(sbi
, FLUSH_MERGE
);
1174 if (f2fs_sb_mounted_blkzoned(sbi
->sb
)) {
1175 set_opt_mode(sbi
, F2FS_MOUNT_LFS
);
1176 set_opt(sbi
, DISCARD
);
1178 set_opt_mode(sbi
, F2FS_MOUNT_ADAPTIVE
);
1181 #ifdef CONFIG_F2FS_FS_XATTR
1182 set_opt(sbi
, XATTR_USER
);
1184 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1185 set_opt(sbi
, POSIX_ACL
);
1188 #ifdef CONFIG_F2FS_FAULT_INJECTION
1189 f2fs_build_fault_attr(sbi
, 0);
1194 static int f2fs_enable_quotas(struct super_block
*sb
);
1196 static int f2fs_remount(struct super_block
*sb
, int *flags
, char *data
)
1198 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
1199 struct f2fs_mount_info org_mount_opt
;
1200 unsigned long old_sb_flags
;
1201 int err
, active_logs
;
1202 bool need_restart_gc
= false;
1203 bool need_stop_gc
= false;
1204 bool no_extent_cache
= !test_opt(sbi
, EXTENT_CACHE
);
1205 #ifdef CONFIG_F2FS_FAULT_INJECTION
1206 struct f2fs_fault_info ffi
= sbi
->fault_info
;
1210 char *s_qf_names
[MAXQUOTAS
];
1215 * Save the old mount options in case we
1216 * need to restore them.
1218 org_mount_opt
= sbi
->mount_opt
;
1219 old_sb_flags
= sb
->s_flags
;
1220 active_logs
= sbi
->active_logs
;
1223 s_jquota_fmt
= sbi
->s_jquota_fmt
;
1224 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1225 if (sbi
->s_qf_names
[i
]) {
1226 s_qf_names
[i
] = kstrdup(sbi
->s_qf_names
[i
],
1228 if (!s_qf_names
[i
]) {
1229 for (j
= 0; j
< i
; j
++)
1230 kfree(s_qf_names
[j
]);
1234 s_qf_names
[i
] = NULL
;
1239 /* recover superblocks we couldn't write due to previous RO mount */
1240 if (!(*flags
& MS_RDONLY
) && is_sbi_flag_set(sbi
, SBI_NEED_SB_WRITE
)) {
1241 err
= f2fs_commit_super(sbi
, false);
1242 f2fs_msg(sb
, KERN_INFO
,
1243 "Try to recover all the superblocks, ret: %d", err
);
1245 clear_sbi_flag(sbi
, SBI_NEED_SB_WRITE
);
1248 default_options(sbi
);
1250 /* parse mount options */
1251 err
= parse_options(sb
, data
);
1256 * Previous and new state of filesystem is RO,
1257 * so skip checking GC and FLUSH_MERGE conditions.
1259 if (f2fs_readonly(sb
) && (*flags
& MS_RDONLY
))
1263 if (!f2fs_readonly(sb
) && (*flags
& MS_RDONLY
)) {
1264 err
= dquot_suspend(sb
, -1);
1268 /* dquot_resume needs RW */
1269 sb
->s_flags
&= ~MS_RDONLY
;
1270 if (sb_any_quota_suspended(sb
)) {
1271 dquot_resume(sb
, -1);
1272 } else if (f2fs_sb_has_quota_ino(sb
)) {
1273 err
= f2fs_enable_quotas(sb
);
1279 /* disallow enable/disable extent_cache dynamically */
1280 if (no_extent_cache
== !!test_opt(sbi
, EXTENT_CACHE
)) {
1282 f2fs_msg(sbi
->sb
, KERN_WARNING
,
1283 "switch extent_cache option is not allowed");
1288 * We stop the GC thread if FS is mounted as RO
1289 * or if background_gc = off is passed in mount
1290 * option. Also sync the filesystem.
1292 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, BG_GC
)) {
1293 if (sbi
->gc_thread
) {
1294 stop_gc_thread(sbi
);
1295 need_restart_gc
= true;
1297 } else if (!sbi
->gc_thread
) {
1298 err
= start_gc_thread(sbi
);
1301 need_stop_gc
= true;
1304 if (*flags
& MS_RDONLY
) {
1305 writeback_inodes_sb(sb
, WB_REASON_SYNC
);
1308 set_sbi_flag(sbi
, SBI_IS_DIRTY
);
1309 set_sbi_flag(sbi
, SBI_IS_CLOSE
);
1310 f2fs_sync_fs(sb
, 1);
1311 clear_sbi_flag(sbi
, SBI_IS_CLOSE
);
1315 * We stop issue flush thread if FS is mounted as RO
1316 * or if flush_merge is not passed in mount option.
1318 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, FLUSH_MERGE
)) {
1319 clear_opt(sbi
, FLUSH_MERGE
);
1320 destroy_flush_cmd_control(sbi
, false);
1322 err
= create_flush_cmd_control(sbi
);
1328 /* Release old quota file names */
1329 for (i
= 0; i
< MAXQUOTAS
; i
++)
1330 kfree(s_qf_names
[i
]);
1332 /* Update the POSIXACL Flag */
1333 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
1334 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
1338 if (need_restart_gc
) {
1339 if (start_gc_thread(sbi
))
1340 f2fs_msg(sbi
->sb
, KERN_WARNING
,
1341 "background gc thread has stopped");
1342 } else if (need_stop_gc
) {
1343 stop_gc_thread(sbi
);
1347 sbi
->s_jquota_fmt
= s_jquota_fmt
;
1348 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1349 kfree(sbi
->s_qf_names
[i
]);
1350 sbi
->s_qf_names
[i
] = s_qf_names
[i
];
1353 sbi
->mount_opt
= org_mount_opt
;
1354 sbi
->active_logs
= active_logs
;
1355 sb
->s_flags
= old_sb_flags
;
1356 #ifdef CONFIG_F2FS_FAULT_INJECTION
1357 sbi
->fault_info
= ffi
;
1363 /* Read data from quotafile */
1364 static ssize_t
f2fs_quota_read(struct super_block
*sb
, int type
, char *data
,
1365 size_t len
, loff_t off
)
1367 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
1368 struct address_space
*mapping
= inode
->i_mapping
;
1369 block_t blkidx
= F2FS_BYTES_TO_BLK(off
);
1370 int offset
= off
& (sb
->s_blocksize
- 1);
1373 loff_t i_size
= i_size_read(inode
);
1380 if (off
+ len
> i_size
)
1383 while (toread
> 0) {
1384 tocopy
= min_t(unsigned long, sb
->s_blocksize
- offset
, toread
);
1386 page
= read_mapping_page(mapping
, blkidx
, NULL
);
1388 if (PTR_ERR(page
) == -ENOMEM
) {
1389 congestion_wait(BLK_RW_ASYNC
, HZ
/50);
1392 return PTR_ERR(page
);
1397 if (unlikely(page
->mapping
!= mapping
)) {
1398 f2fs_put_page(page
, 1);
1401 if (unlikely(!PageUptodate(page
))) {
1402 f2fs_put_page(page
, 1);
1406 kaddr
= kmap_atomic(page
);
1407 memcpy(data
, kaddr
+ offset
, tocopy
);
1408 kunmap_atomic(kaddr
);
1409 f2fs_put_page(page
, 1);
1419 /* Write to quotafile */
1420 static ssize_t
f2fs_quota_write(struct super_block
*sb
, int type
,
1421 const char *data
, size_t len
, loff_t off
)
1423 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
1424 struct address_space
*mapping
= inode
->i_mapping
;
1425 const struct address_space_operations
*a_ops
= mapping
->a_ops
;
1426 int offset
= off
& (sb
->s_blocksize
- 1);
1427 size_t towrite
= len
;
1433 while (towrite
> 0) {
1434 tocopy
= min_t(unsigned long, sb
->s_blocksize
- offset
,
1437 err
= a_ops
->write_begin(NULL
, mapping
, off
, tocopy
, 0,
1439 if (unlikely(err
)) {
1440 if (err
== -ENOMEM
) {
1441 congestion_wait(BLK_RW_ASYNC
, HZ
/50);
1447 kaddr
= kmap_atomic(page
);
1448 memcpy(kaddr
+ offset
, data
, tocopy
);
1449 kunmap_atomic(kaddr
);
1450 flush_dcache_page(page
);
1452 a_ops
->write_end(NULL
, mapping
, off
, tocopy
, tocopy
,
1463 inode
->i_mtime
= inode
->i_ctime
= current_time(inode
);
1464 f2fs_mark_inode_dirty_sync(inode
, false);
1465 return len
- towrite
;
1468 static struct dquot
**f2fs_get_dquots(struct inode
*inode
)
1470 return F2FS_I(inode
)->i_dquot
;
1473 static qsize_t
*f2fs_get_reserved_space(struct inode
*inode
)
1475 return &F2FS_I(inode
)->i_reserved_quota
;
1478 static int f2fs_quota_on_mount(struct f2fs_sb_info
*sbi
, int type
)
1480 return dquot_quota_on_mount(sbi
->sb
, sbi
->s_qf_names
[type
],
1481 sbi
->s_jquota_fmt
, type
);
1484 int f2fs_enable_quota_files(struct f2fs_sb_info
*sbi
, bool rdonly
)
1489 if (f2fs_sb_has_quota_ino(sbi
->sb
) && rdonly
) {
1490 err
= f2fs_enable_quotas(sbi
->sb
);
1492 f2fs_msg(sbi
->sb
, KERN_ERR
,
1493 "Cannot turn on quota_ino: %d", err
);
1499 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1500 if (sbi
->s_qf_names
[i
]) {
1501 err
= f2fs_quota_on_mount(sbi
, i
);
1506 f2fs_msg(sbi
->sb
, KERN_ERR
,
1507 "Cannot turn on quotas: %d on %d", err
, i
);
1513 static int f2fs_quota_enable(struct super_block
*sb
, int type
, int format_id
,
1516 struct inode
*qf_inode
;
1517 unsigned long qf_inum
;
1520 BUG_ON(!f2fs_sb_has_quota_ino(sb
));
1522 qf_inum
= f2fs_qf_ino(sb
, type
);
1526 qf_inode
= f2fs_iget(sb
, qf_inum
);
1527 if (IS_ERR(qf_inode
)) {
1528 f2fs_msg(sb
, KERN_ERR
,
1529 "Bad quota inode %u:%lu", type
, qf_inum
);
1530 return PTR_ERR(qf_inode
);
1533 /* Don't account quota for quota files to avoid recursion */
1534 qf_inode
->i_flags
|= S_NOQUOTA
;
1535 err
= dquot_enable(qf_inode
, type
, format_id
, flags
);
1540 static int f2fs_enable_quotas(struct super_block
*sb
)
1543 unsigned long qf_inum
;
1544 bool quota_mopt
[MAXQUOTAS
] = {
1545 test_opt(F2FS_SB(sb
), USRQUOTA
),
1546 test_opt(F2FS_SB(sb
), GRPQUOTA
),
1547 test_opt(F2FS_SB(sb
), PRJQUOTA
),
1550 sb_dqopt(sb
)->flags
|= DQUOT_QUOTA_SYS_FILE
;
1551 for (type
= 0; type
< MAXQUOTAS
; type
++) {
1552 qf_inum
= f2fs_qf_ino(sb
, type
);
1554 err
= f2fs_quota_enable(sb
, type
, QFMT_VFS_V1
,
1555 DQUOT_USAGE_ENABLED
|
1556 (quota_mopt
[type
] ? DQUOT_LIMITS_ENABLED
: 0));
1558 f2fs_msg(sb
, KERN_ERR
,
1559 "Failed to enable quota tracking "
1560 "(type=%d, err=%d). Please run "
1561 "fsck to fix.", type
, err
);
1562 for (type
--; type
>= 0; type
--)
1563 dquot_quota_off(sb
, type
);
1571 static int f2fs_quota_sync(struct super_block
*sb
, int type
)
1573 struct quota_info
*dqopt
= sb_dqopt(sb
);
1577 ret
= dquot_writeback_dquots(sb
, type
);
1582 * Now when everything is written we can discard the pagecache so
1583 * that userspace sees the changes.
1585 for (cnt
= 0; cnt
< MAXQUOTAS
; cnt
++) {
1586 if (type
!= -1 && cnt
!= type
)
1588 if (!sb_has_quota_active(sb
, cnt
))
1591 ret
= filemap_write_and_wait(dqopt
->files
[cnt
]->i_mapping
);
1595 inode_lock(dqopt
->files
[cnt
]);
1596 truncate_inode_pages(&dqopt
->files
[cnt
]->i_data
, 0);
1597 inode_unlock(dqopt
->files
[cnt
]);
1602 static int f2fs_quota_on(struct super_block
*sb
, int type
, int format_id
,
1605 struct inode
*inode
;
1608 err
= f2fs_quota_sync(sb
, type
);
1612 err
= dquot_quota_on(sb
, type
, format_id
, path
);
1616 inode
= d_inode(path
->dentry
);
1619 F2FS_I(inode
)->i_flags
|= FS_NOATIME_FL
| FS_IMMUTABLE_FL
;
1620 inode_set_flags(inode
, S_NOATIME
| S_IMMUTABLE
,
1621 S_NOATIME
| S_IMMUTABLE
);
1622 inode_unlock(inode
);
1623 f2fs_mark_inode_dirty_sync(inode
, false);
1628 static int f2fs_quota_off(struct super_block
*sb
, int type
)
1630 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
1633 if (!inode
|| !igrab(inode
))
1634 return dquot_quota_off(sb
, type
);
1636 f2fs_quota_sync(sb
, type
);
1638 err
= dquot_quota_off(sb
, type
);
1639 if (err
|| f2fs_sb_has_quota_ino(sb
))
1643 F2FS_I(inode
)->i_flags
&= ~(FS_NOATIME_FL
| FS_IMMUTABLE_FL
);
1644 inode_set_flags(inode
, 0, S_NOATIME
| S_IMMUTABLE
);
1645 inode_unlock(inode
);
1646 f2fs_mark_inode_dirty_sync(inode
, false);
1652 void f2fs_quota_off_umount(struct super_block
*sb
)
1656 for (type
= 0; type
< MAXQUOTAS
; type
++)
1657 f2fs_quota_off(sb
, type
);
1661 int f2fs_get_projid(struct inode
*inode
, kprojid_t
*projid
)
1663 *projid
= F2FS_I(inode
)->i_projid
;
1668 static const struct dquot_operations f2fs_quota_operations
= {
1669 .get_reserved_space
= f2fs_get_reserved_space
,
1670 .write_dquot
= dquot_commit
,
1671 .acquire_dquot
= dquot_acquire
,
1672 .release_dquot
= dquot_release
,
1673 .mark_dirty
= dquot_mark_dquot_dirty
,
1674 .write_info
= dquot_commit_info
,
1675 .alloc_dquot
= dquot_alloc
,
1676 .destroy_dquot
= dquot_destroy
,
1678 .get_projid
= f2fs_get_projid
,
1679 .get_next_id
= dquot_get_next_id
,
1683 static const struct quotactl_ops f2fs_quotactl_ops
= {
1684 .quota_on
= f2fs_quota_on
,
1685 .quota_off
= f2fs_quota_off
,
1686 .quota_sync
= f2fs_quota_sync
,
1687 .get_state
= dquot_get_state
,
1688 .set_info
= dquot_set_dqinfo
,
1689 .get_dqblk
= dquot_get_dqblk
,
1690 .set_dqblk
= dquot_set_dqblk
,
1693 void f2fs_quota_off_umount(struct super_block
*sb
)
1698 static const struct super_operations f2fs_sops
= {
1699 .alloc_inode
= f2fs_alloc_inode
,
1700 .drop_inode
= f2fs_drop_inode
,
1701 .destroy_inode
= f2fs_destroy_inode
,
1702 .write_inode
= f2fs_write_inode
,
1703 .dirty_inode
= f2fs_dirty_inode
,
1704 .show_options
= f2fs_show_options
,
1706 .quota_read
= f2fs_quota_read
,
1707 .quota_write
= f2fs_quota_write
,
1708 .get_dquots
= f2fs_get_dquots
,
1710 .evict_inode
= f2fs_evict_inode
,
1711 .put_super
= f2fs_put_super
,
1712 .sync_fs
= f2fs_sync_fs
,
1713 .freeze_fs
= f2fs_freeze
,
1714 .unfreeze_fs
= f2fs_unfreeze
,
1715 .statfs
= f2fs_statfs
,
1716 .remount_fs
= f2fs_remount
,
1719 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1720 static int f2fs_get_context(struct inode
*inode
, void *ctx
, size_t len
)
1722 return f2fs_getxattr(inode
, F2FS_XATTR_INDEX_ENCRYPTION
,
1723 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT
,
1727 static int f2fs_set_context(struct inode
*inode
, const void *ctx
, size_t len
,
1730 return f2fs_setxattr(inode
, F2FS_XATTR_INDEX_ENCRYPTION
,
1731 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT
,
1732 ctx
, len
, fs_data
, XATTR_CREATE
);
1735 static unsigned f2fs_max_namelen(struct inode
*inode
)
1737 return S_ISLNK(inode
->i_mode
) ?
1738 inode
->i_sb
->s_blocksize
: F2FS_NAME_LEN
;
1741 static const struct fscrypt_operations f2fs_cryptops
= {
1742 .key_prefix
= "f2fs:",
1743 .get_context
= f2fs_get_context
,
1744 .set_context
= f2fs_set_context
,
1745 .is_encrypted
= f2fs_encrypted_inode
,
1746 .empty_dir
= f2fs_empty_dir
,
1747 .max_namelen
= f2fs_max_namelen
,
1750 static const struct fscrypt_operations f2fs_cryptops
= {
1751 .is_encrypted
= f2fs_encrypted_inode
,
1755 static struct inode
*f2fs_nfs_get_inode(struct super_block
*sb
,
1756 u64 ino
, u32 generation
)
1758 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
1759 struct inode
*inode
;
1761 if (check_nid_range(sbi
, ino
))
1762 return ERR_PTR(-ESTALE
);
1765 * f2fs_iget isn't quite right if the inode is currently unallocated!
1766 * However f2fs_iget currently does appropriate checks to handle stale
1767 * inodes so everything is OK.
1769 inode
= f2fs_iget(sb
, ino
);
1771 return ERR_CAST(inode
);
1772 if (unlikely(generation
&& inode
->i_generation
!= generation
)) {
1773 /* we didn't find the right inode.. */
1775 return ERR_PTR(-ESTALE
);
1780 static struct dentry
*f2fs_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1781 int fh_len
, int fh_type
)
1783 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
1784 f2fs_nfs_get_inode
);
1787 static struct dentry
*f2fs_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1788 int fh_len
, int fh_type
)
1790 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1791 f2fs_nfs_get_inode
);
1794 static const struct export_operations f2fs_export_ops
= {
1795 .fh_to_dentry
= f2fs_fh_to_dentry
,
1796 .fh_to_parent
= f2fs_fh_to_parent
,
1797 .get_parent
= f2fs_get_parent
,
1800 static loff_t
max_file_blocks(void)
1803 loff_t leaf_count
= ADDRS_PER_BLOCK
;
1806 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
1807 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
1808 * space in inode.i_addr, it will be more safe to reassign
1812 /* two direct node blocks */
1813 result
+= (leaf_count
* 2);
1815 /* two indirect node blocks */
1816 leaf_count
*= NIDS_PER_BLOCK
;
1817 result
+= (leaf_count
* 2);
1819 /* one double indirect node block */
1820 leaf_count
*= NIDS_PER_BLOCK
;
1821 result
+= leaf_count
;
1826 static int __f2fs_commit_super(struct buffer_head
*bh
,
1827 struct f2fs_super_block
*super
)
1831 memcpy(bh
->b_data
+ F2FS_SUPER_OFFSET
, super
, sizeof(*super
));
1832 set_buffer_uptodate(bh
);
1833 set_buffer_dirty(bh
);
1836 /* it's rare case, we can do fua all the time */
1837 return __sync_dirty_buffer(bh
, WRITE_FLUSH_FUA
);
1840 static inline bool sanity_check_area_boundary(struct f2fs_sb_info
*sbi
,
1841 struct buffer_head
*bh
)
1843 struct f2fs_super_block
*raw_super
= (struct f2fs_super_block
*)
1844 (bh
->b_data
+ F2FS_SUPER_OFFSET
);
1845 struct super_block
*sb
= sbi
->sb
;
1846 u32 segment0_blkaddr
= le32_to_cpu(raw_super
->segment0_blkaddr
);
1847 u32 cp_blkaddr
= le32_to_cpu(raw_super
->cp_blkaddr
);
1848 u32 sit_blkaddr
= le32_to_cpu(raw_super
->sit_blkaddr
);
1849 u32 nat_blkaddr
= le32_to_cpu(raw_super
->nat_blkaddr
);
1850 u32 ssa_blkaddr
= le32_to_cpu(raw_super
->ssa_blkaddr
);
1851 u32 main_blkaddr
= le32_to_cpu(raw_super
->main_blkaddr
);
1852 u32 segment_count_ckpt
= le32_to_cpu(raw_super
->segment_count_ckpt
);
1853 u32 segment_count_sit
= le32_to_cpu(raw_super
->segment_count_sit
);
1854 u32 segment_count_nat
= le32_to_cpu(raw_super
->segment_count_nat
);
1855 u32 segment_count_ssa
= le32_to_cpu(raw_super
->segment_count_ssa
);
1856 u32 segment_count_main
= le32_to_cpu(raw_super
->segment_count_main
);
1857 u32 segment_count
= le32_to_cpu(raw_super
->segment_count
);
1858 u32 log_blocks_per_seg
= le32_to_cpu(raw_super
->log_blocks_per_seg
);
1859 u64 main_end_blkaddr
= main_blkaddr
+
1860 (segment_count_main
<< log_blocks_per_seg
);
1861 u64 seg_end_blkaddr
= segment0_blkaddr
+
1862 (segment_count
<< log_blocks_per_seg
);
1864 if (segment0_blkaddr
!= cp_blkaddr
) {
1865 f2fs_msg(sb
, KERN_INFO
,
1866 "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1867 segment0_blkaddr
, cp_blkaddr
);
1871 if (cp_blkaddr
+ (segment_count_ckpt
<< log_blocks_per_seg
) !=
1873 f2fs_msg(sb
, KERN_INFO
,
1874 "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1875 cp_blkaddr
, sit_blkaddr
,
1876 segment_count_ckpt
<< log_blocks_per_seg
);
1880 if (sit_blkaddr
+ (segment_count_sit
<< log_blocks_per_seg
) !=
1882 f2fs_msg(sb
, KERN_INFO
,
1883 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1884 sit_blkaddr
, nat_blkaddr
,
1885 segment_count_sit
<< log_blocks_per_seg
);
1889 if (nat_blkaddr
+ (segment_count_nat
<< log_blocks_per_seg
) !=
1891 f2fs_msg(sb
, KERN_INFO
,
1892 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1893 nat_blkaddr
, ssa_blkaddr
,
1894 segment_count_nat
<< log_blocks_per_seg
);
1898 if (ssa_blkaddr
+ (segment_count_ssa
<< log_blocks_per_seg
) !=
1900 f2fs_msg(sb
, KERN_INFO
,
1901 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1902 ssa_blkaddr
, main_blkaddr
,
1903 segment_count_ssa
<< log_blocks_per_seg
);
1907 if (main_end_blkaddr
> seg_end_blkaddr
) {
1908 f2fs_msg(sb
, KERN_INFO
,
1909 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1912 (segment_count
<< log_blocks_per_seg
),
1913 segment_count_main
<< log_blocks_per_seg
);
1915 } else if (main_end_blkaddr
< seg_end_blkaddr
) {
1919 /* fix in-memory information all the time */
1920 raw_super
->segment_count
= cpu_to_le32((main_end_blkaddr
-
1921 segment0_blkaddr
) >> log_blocks_per_seg
);
1923 if (f2fs_readonly(sb
) || bdev_read_only(sb
->s_bdev
)) {
1924 set_sbi_flag(sbi
, SBI_NEED_SB_WRITE
);
1927 err
= __f2fs_commit_super(bh
, NULL
);
1928 res
= err
? "failed" : "done";
1930 f2fs_msg(sb
, KERN_INFO
,
1931 "Fix alignment : %s, start(%u) end(%u) block(%u)",
1934 (segment_count
<< log_blocks_per_seg
),
1935 segment_count_main
<< log_blocks_per_seg
);
1942 static int sanity_check_raw_super(struct f2fs_sb_info
*sbi
,
1943 struct buffer_head
*bh
)
1945 struct f2fs_super_block
*raw_super
= (struct f2fs_super_block
*)
1946 (bh
->b_data
+ F2FS_SUPER_OFFSET
);
1947 struct super_block
*sb
= sbi
->sb
;
1948 unsigned int blocksize
;
1950 if (F2FS_SUPER_MAGIC
!= le32_to_cpu(raw_super
->magic
)) {
1951 f2fs_msg(sb
, KERN_INFO
,
1952 "Magic Mismatch, valid(0x%x) - read(0x%x)",
1953 F2FS_SUPER_MAGIC
, le32_to_cpu(raw_super
->magic
));
1957 /* Currently, support only 4KB page cache size */
1958 if (F2FS_BLKSIZE
!= PAGE_SIZE
) {
1959 f2fs_msg(sb
, KERN_INFO
,
1960 "Invalid page_cache_size (%lu), supports only 4KB\n",
1965 /* Currently, support only 4KB block size */
1966 blocksize
= 1 << le32_to_cpu(raw_super
->log_blocksize
);
1967 if (blocksize
!= F2FS_BLKSIZE
) {
1968 f2fs_msg(sb
, KERN_INFO
,
1969 "Invalid blocksize (%u), supports only 4KB\n",
1974 /* check log blocks per segment */
1975 if (le32_to_cpu(raw_super
->log_blocks_per_seg
) != 9) {
1976 f2fs_msg(sb
, KERN_INFO
,
1977 "Invalid log blocks per segment (%u)\n",
1978 le32_to_cpu(raw_super
->log_blocks_per_seg
));
1982 /* Currently, support 512/1024/2048/4096 bytes sector size */
1983 if (le32_to_cpu(raw_super
->log_sectorsize
) >
1984 F2FS_MAX_LOG_SECTOR_SIZE
||
1985 le32_to_cpu(raw_super
->log_sectorsize
) <
1986 F2FS_MIN_LOG_SECTOR_SIZE
) {
1987 f2fs_msg(sb
, KERN_INFO
, "Invalid log sectorsize (%u)",
1988 le32_to_cpu(raw_super
->log_sectorsize
));
1991 if (le32_to_cpu(raw_super
->log_sectors_per_block
) +
1992 le32_to_cpu(raw_super
->log_sectorsize
) !=
1993 F2FS_MAX_LOG_SECTOR_SIZE
) {
1994 f2fs_msg(sb
, KERN_INFO
,
1995 "Invalid log sectors per block(%u) log sectorsize(%u)",
1996 le32_to_cpu(raw_super
->log_sectors_per_block
),
1997 le32_to_cpu(raw_super
->log_sectorsize
));
2001 /* check reserved ino info */
2002 if (le32_to_cpu(raw_super
->node_ino
) != 1 ||
2003 le32_to_cpu(raw_super
->meta_ino
) != 2 ||
2004 le32_to_cpu(raw_super
->root_ino
) != 3) {
2005 f2fs_msg(sb
, KERN_INFO
,
2006 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2007 le32_to_cpu(raw_super
->node_ino
),
2008 le32_to_cpu(raw_super
->meta_ino
),
2009 le32_to_cpu(raw_super
->root_ino
));
2013 if (le32_to_cpu(raw_super
->segment_count
) > F2FS_MAX_SEGMENT
) {
2014 f2fs_msg(sb
, KERN_INFO
,
2015 "Invalid segment count (%u)",
2016 le32_to_cpu(raw_super
->segment_count
));
2020 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2021 if (sanity_check_area_boundary(sbi
, bh
))
2027 int sanity_check_ckpt(struct f2fs_sb_info
*sbi
)
2029 unsigned int total
, fsmeta
;
2030 struct f2fs_super_block
*raw_super
= F2FS_RAW_SUPER(sbi
);
2031 struct f2fs_checkpoint
*ckpt
= F2FS_CKPT(sbi
);
2032 unsigned int ovp_segments
, reserved_segments
;
2033 unsigned int main_segs
, blocks_per_seg
;
2036 total
= le32_to_cpu(raw_super
->segment_count
);
2037 fsmeta
= le32_to_cpu(raw_super
->segment_count_ckpt
);
2038 fsmeta
+= le32_to_cpu(raw_super
->segment_count_sit
);
2039 fsmeta
+= le32_to_cpu(raw_super
->segment_count_nat
);
2040 fsmeta
+= le32_to_cpu(ckpt
->rsvd_segment_count
);
2041 fsmeta
+= le32_to_cpu(raw_super
->segment_count_ssa
);
2043 if (unlikely(fsmeta
>= total
))
2046 ovp_segments
= le32_to_cpu(ckpt
->overprov_segment_count
);
2047 reserved_segments
= le32_to_cpu(ckpt
->rsvd_segment_count
);
2049 if (unlikely(fsmeta
< F2FS_MIN_SEGMENTS
||
2050 ovp_segments
== 0 || reserved_segments
== 0)) {
2051 f2fs_msg(sbi
->sb
, KERN_ERR
,
2052 "Wrong layout: check mkfs.f2fs version");
2056 main_segs
= le32_to_cpu(raw_super
->segment_count_main
);
2057 blocks_per_seg
= sbi
->blocks_per_seg
;
2059 for (i
= 0; i
< NR_CURSEG_NODE_TYPE
; i
++) {
2060 if (le32_to_cpu(ckpt
->cur_node_segno
[i
]) >= main_segs
||
2061 le16_to_cpu(ckpt
->cur_node_blkoff
[i
]) >= blocks_per_seg
)
2064 for (i
= 0; i
< NR_CURSEG_DATA_TYPE
; i
++) {
2065 if (le32_to_cpu(ckpt
->cur_data_segno
[i
]) >= main_segs
||
2066 le16_to_cpu(ckpt
->cur_data_blkoff
[i
]) >= blocks_per_seg
)
2070 if (unlikely(f2fs_cp_error(sbi
))) {
2071 f2fs_msg(sbi
->sb
, KERN_ERR
, "A bug case: need to run fsck");
2077 static void init_sb_info(struct f2fs_sb_info
*sbi
)
2079 struct f2fs_super_block
*raw_super
= sbi
->raw_super
;
2082 sbi
->log_sectors_per_block
=
2083 le32_to_cpu(raw_super
->log_sectors_per_block
);
2084 sbi
->log_blocksize
= le32_to_cpu(raw_super
->log_blocksize
);
2085 sbi
->blocksize
= 1 << sbi
->log_blocksize
;
2086 sbi
->log_blocks_per_seg
= le32_to_cpu(raw_super
->log_blocks_per_seg
);
2087 sbi
->blocks_per_seg
= 1 << sbi
->log_blocks_per_seg
;
2088 sbi
->segs_per_sec
= le32_to_cpu(raw_super
->segs_per_sec
);
2089 sbi
->secs_per_zone
= le32_to_cpu(raw_super
->secs_per_zone
);
2090 sbi
->total_sections
= le32_to_cpu(raw_super
->section_count
);
2091 sbi
->total_node_count
=
2092 (le32_to_cpu(raw_super
->segment_count_nat
) / 2)
2093 * sbi
->blocks_per_seg
* NAT_ENTRY_PER_BLOCK
;
2094 sbi
->root_ino_num
= le32_to_cpu(raw_super
->root_ino
);
2095 sbi
->node_ino_num
= le32_to_cpu(raw_super
->node_ino
);
2096 sbi
->meta_ino_num
= le32_to_cpu(raw_super
->meta_ino
);
2097 sbi
->cur_victim_sec
= NULL_SECNO
;
2098 sbi
->max_victim_search
= DEF_MAX_VICTIM_SEARCH
;
2100 sbi
->dir_level
= DEF_DIR_LEVEL
;
2101 sbi
->interval_time
[CP_TIME
] = DEF_CP_INTERVAL
;
2102 sbi
->interval_time
[REQ_TIME
] = DEF_IDLE_INTERVAL
;
2103 clear_sbi_flag(sbi
, SBI_NEED_FSCK
);
2105 for (i
= 0; i
< NR_COUNT_TYPE
; i
++)
2106 atomic_set(&sbi
->nr_pages
[i
], 0);
2108 atomic_set(&sbi
->wb_sync_req
, 0);
2110 INIT_LIST_HEAD(&sbi
->s_list
);
2111 mutex_init(&sbi
->umount_mutex
);
2112 for (i
= 0; i
< NR_PAGE_TYPE
- 1; i
++)
2113 for (j
= HOT
; j
< NR_TEMP_TYPE
; j
++)
2114 mutex_init(&sbi
->wio_mutex
[i
][j
]);
2115 spin_lock_init(&sbi
->cp_lock
);
2117 sbi
->dirty_device
= 0;
2118 spin_lock_init(&sbi
->dev_lock
);
2121 static int init_percpu_info(struct f2fs_sb_info
*sbi
)
2125 err
= percpu_counter_init(&sbi
->alloc_valid_block_count
, 0, GFP_KERNEL
);
2129 return percpu_counter_init(&sbi
->total_valid_inode_count
, 0,
2133 #ifdef CONFIG_BLK_DEV_ZONED
2134 static int init_blkz_info(struct f2fs_sb_info
*sbi
, int devi
)
2136 struct block_device
*bdev
= FDEV(devi
).bdev
;
2137 sector_t nr_sectors
= bdev
->bd_part
->nr_sects
;
2138 sector_t sector
= 0;
2139 struct blk_zone
*zones
;
2140 unsigned int i
, nr_zones
;
2144 if (!f2fs_sb_mounted_blkzoned(sbi
->sb
))
2147 if (sbi
->blocks_per_blkz
&& sbi
->blocks_per_blkz
!=
2148 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev
)))
2150 sbi
->blocks_per_blkz
= SECTOR_TO_BLOCK(bdev_zone_sectors(bdev
));
2151 if (sbi
->log_blocks_per_blkz
&& sbi
->log_blocks_per_blkz
!=
2152 __ilog2_u32(sbi
->blocks_per_blkz
))
2154 sbi
->log_blocks_per_blkz
= __ilog2_u32(sbi
->blocks_per_blkz
);
2155 FDEV(devi
).nr_blkz
= SECTOR_TO_BLOCK(nr_sectors
) >>
2156 sbi
->log_blocks_per_blkz
;
2157 if (nr_sectors
& (bdev_zone_sectors(bdev
) - 1))
2158 FDEV(devi
).nr_blkz
++;
2160 FDEV(devi
).blkz_type
= kmalloc(FDEV(devi
).nr_blkz
, GFP_KERNEL
);
2161 if (!FDEV(devi
).blkz_type
)
2164 #define F2FS_REPORT_NR_ZONES 4096
2166 zones
= kcalloc(F2FS_REPORT_NR_ZONES
, sizeof(struct blk_zone
),
2171 /* Get block zones type */
2172 while (zones
&& sector
< nr_sectors
) {
2174 nr_zones
= F2FS_REPORT_NR_ZONES
;
2175 err
= blkdev_report_zones(bdev
, sector
,
2185 for (i
= 0; i
< nr_zones
; i
++) {
2186 FDEV(devi
).blkz_type
[n
] = zones
[i
].type
;
2187 sector
+= zones
[i
].len
;
2199 * Read f2fs raw super block.
2200 * Because we have two copies of super block, so read both of them
2201 * to get the first valid one. If any one of them is broken, we pass
2202 * them recovery flag back to the caller.
2204 static int read_raw_super_block(struct f2fs_sb_info
*sbi
,
2205 struct f2fs_super_block
**raw_super
,
2206 int *valid_super_block
, int *recovery
)
2208 struct super_block
*sb
= sbi
->sb
;
2210 struct buffer_head
*bh
;
2211 struct f2fs_super_block
*super
;
2214 super
= kzalloc(sizeof(struct f2fs_super_block
), GFP_KERNEL
);
2218 for (block
= 0; block
< 2; block
++) {
2219 bh
= sb_bread(sb
, block
);
2221 f2fs_msg(sb
, KERN_ERR
, "Unable to read %dth superblock",
2227 /* sanity checking of raw super */
2228 if (sanity_check_raw_super(sbi
, bh
)) {
2229 f2fs_msg(sb
, KERN_ERR
,
2230 "Can't find valid F2FS filesystem in %dth superblock",
2238 memcpy(super
, bh
->b_data
+ F2FS_SUPER_OFFSET
,
2240 *valid_super_block
= block
;
2246 /* Fail to read any one of the superblocks*/
2250 /* No valid superblock */
2259 int f2fs_commit_super(struct f2fs_sb_info
*sbi
, bool recover
)
2261 struct buffer_head
*bh
;
2264 if ((recover
&& f2fs_readonly(sbi
->sb
)) ||
2265 bdev_read_only(sbi
->sb
->s_bdev
)) {
2266 set_sbi_flag(sbi
, SBI_NEED_SB_WRITE
);
2270 /* write back-up superblock first */
2271 bh
= sb_getblk(sbi
->sb
, sbi
->valid_super_block
? 0: 1);
2274 err
= __f2fs_commit_super(bh
, F2FS_RAW_SUPER(sbi
));
2277 /* if we are in recovery path, skip writing valid superblock */
2281 /* write current valid superblock */
2282 bh
= sb_getblk(sbi
->sb
, sbi
->valid_super_block
);
2285 err
= __f2fs_commit_super(bh
, F2FS_RAW_SUPER(sbi
));
2290 static int f2fs_scan_devices(struct f2fs_sb_info
*sbi
)
2292 struct f2fs_super_block
*raw_super
= F2FS_RAW_SUPER(sbi
);
2293 unsigned int max_devices
= MAX_DEVICES
;
2296 /* Initialize single device information */
2297 if (!RDEV(0).path
[0]) {
2298 #ifdef CONFIG_BLK_DEV_ZONED
2299 if (!bdev_is_zoned(sbi
->sb
->s_bdev
))
2308 * Initialize multiple devices information, or single
2309 * zoned block device information.
2311 sbi
->devs
= kcalloc(max_devices
, sizeof(struct f2fs_dev_info
),
2316 for (i
= 0; i
< max_devices
; i
++) {
2318 if (i
> 0 && !RDEV(i
).path
[0])
2321 if (max_devices
== 1) {
2322 /* Single zoned block device mount */
2324 blkdev_get_by_dev(sbi
->sb
->s_bdev
->bd_dev
,
2325 sbi
->sb
->s_mode
, sbi
->sb
->s_type
);
2327 /* Multi-device mount */
2328 memcpy(FDEV(i
).path
, RDEV(i
).path
, MAX_PATH_LEN
);
2329 FDEV(i
).total_segments
=
2330 le32_to_cpu(RDEV(i
).total_segments
);
2332 FDEV(i
).start_blk
= 0;
2333 FDEV(i
).end_blk
= FDEV(i
).start_blk
+
2334 (FDEV(i
).total_segments
<<
2335 sbi
->log_blocks_per_seg
) - 1 +
2336 le32_to_cpu(raw_super
->segment0_blkaddr
);
2338 FDEV(i
).start_blk
= FDEV(i
- 1).end_blk
+ 1;
2339 FDEV(i
).end_blk
= FDEV(i
).start_blk
+
2340 (FDEV(i
).total_segments
<<
2341 sbi
->log_blocks_per_seg
) - 1;
2343 FDEV(i
).bdev
= blkdev_get_by_path(FDEV(i
).path
,
2344 sbi
->sb
->s_mode
, sbi
->sb
->s_type
);
2346 if (IS_ERR(FDEV(i
).bdev
))
2347 return PTR_ERR(FDEV(i
).bdev
);
2349 /* to release errored devices */
2350 sbi
->s_ndevs
= i
+ 1;
2352 #ifdef CONFIG_BLK_DEV_ZONED
2353 if (bdev_zoned_model(FDEV(i
).bdev
) == BLK_ZONED_HM
&&
2354 !f2fs_sb_mounted_blkzoned(sbi
->sb
)) {
2355 f2fs_msg(sbi
->sb
, KERN_ERR
,
2356 "Zoned block device feature not enabled\n");
2359 if (bdev_zoned_model(FDEV(i
).bdev
) != BLK_ZONED_NONE
) {
2360 if (init_blkz_info(sbi
, i
)) {
2361 f2fs_msg(sbi
->sb
, KERN_ERR
,
2362 "Failed to initialize F2FS blkzone information");
2365 if (max_devices
== 1)
2367 f2fs_msg(sbi
->sb
, KERN_INFO
,
2368 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
2370 FDEV(i
).total_segments
,
2371 FDEV(i
).start_blk
, FDEV(i
).end_blk
,
2372 bdev_zoned_model(FDEV(i
).bdev
) == BLK_ZONED_HA
?
2373 "Host-aware" : "Host-managed");
2377 f2fs_msg(sbi
->sb
, KERN_INFO
,
2378 "Mount Device [%2d]: %20s, %8u, %8x - %8x",
2380 FDEV(i
).total_segments
,
2381 FDEV(i
).start_blk
, FDEV(i
).end_blk
);
2383 f2fs_msg(sbi
->sb
, KERN_INFO
,
2384 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi
));
2388 static int f2fs_fill_super(struct super_block
*sb
, void *data
, int silent
)
2390 struct f2fs_sb_info
*sbi
;
2391 struct f2fs_super_block
*raw_super
;
2394 bool retry
= true, need_fsck
= false;
2395 char *options
= NULL
;
2396 int recovery
, i
, valid_super_block
;
2397 struct curseg_info
*seg_i
;
2402 valid_super_block
= -1;
2405 /* allocate memory for f2fs-specific super block info */
2406 sbi
= kzalloc(sizeof(struct f2fs_sb_info
), GFP_KERNEL
);
2412 /* Load the checksum driver */
2413 sbi
->s_chksum_driver
= crypto_alloc_shash("crc32", 0, 0);
2414 if (IS_ERR(sbi
->s_chksum_driver
)) {
2415 f2fs_msg(sb
, KERN_ERR
, "Cannot load crc32 driver.");
2416 err
= PTR_ERR(sbi
->s_chksum_driver
);
2417 sbi
->s_chksum_driver
= NULL
;
2421 /* set a block size */
2422 if (unlikely(!sb_set_blocksize(sb
, F2FS_BLKSIZE
))) {
2423 f2fs_msg(sb
, KERN_ERR
, "unable to set blocksize");
2427 err
= read_raw_super_block(sbi
, &raw_super
, &valid_super_block
,
2432 sb
->s_fs_info
= sbi
;
2433 sbi
->raw_super
= raw_super
;
2435 /* precompute checksum seed for metadata */
2436 if (f2fs_sb_has_inode_chksum(sb
))
2437 sbi
->s_chksum_seed
= f2fs_chksum(sbi
, ~0, raw_super
->uuid
,
2438 sizeof(raw_super
->uuid
));
2441 * The BLKZONED feature indicates that the drive was formatted with
2442 * zone alignment optimization. This is optional for host-aware
2443 * devices, but mandatory for host-managed zoned block devices.
2445 #ifndef CONFIG_BLK_DEV_ZONED
2446 if (f2fs_sb_mounted_blkzoned(sb
)) {
2447 f2fs_msg(sb
, KERN_ERR
,
2448 "Zoned block device support is not enabled\n");
2453 default_options(sbi
);
2454 /* parse mount options */
2455 options
= kstrdup((const char *)data
, GFP_KERNEL
);
2456 if (data
&& !options
) {
2461 err
= parse_options(sb
, options
);
2465 sbi
->max_file_blocks
= max_file_blocks();
2466 sb
->s_maxbytes
= sbi
->max_file_blocks
<<
2467 le32_to_cpu(raw_super
->log_blocksize
);
2468 sb
->s_max_links
= F2FS_LINK_MAX
;
2469 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
2472 sb
->dq_op
= &f2fs_quota_operations
;
2473 if (f2fs_sb_has_quota_ino(sb
))
2474 sb
->s_qcop
= &dquot_quotactl_sysfile_ops
;
2476 sb
->s_qcop
= &f2fs_quotactl_ops
;
2477 sb
->s_quota_types
= QTYPE_MASK_USR
| QTYPE_MASK_GRP
| QTYPE_MASK_PRJ
;
2480 sb
->s_op
= &f2fs_sops
;
2481 sb
->s_cop
= &f2fs_cryptops
;
2482 sb
->s_xattr
= f2fs_xattr_handlers
;
2483 sb
->s_export_op
= &f2fs_export_ops
;
2484 sb
->s_magic
= F2FS_SUPER_MAGIC
;
2485 sb
->s_time_gran
= 1;
2486 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
2487 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
2488 memcpy(sb
->s_uuid
, raw_super
->uuid
, sizeof(raw_super
->uuid
));
2490 /* init f2fs-specific super block info */
2491 sbi
->valid_super_block
= valid_super_block
;
2492 mutex_init(&sbi
->gc_mutex
);
2493 mutex_init(&sbi
->cp_mutex
);
2494 init_rwsem(&sbi
->node_write
);
2495 init_rwsem(&sbi
->node_change
);
2497 /* disallow all the data/node/meta page writes */
2498 set_sbi_flag(sbi
, SBI_POR_DOING
);
2499 spin_lock_init(&sbi
->stat_lock
);
2501 /* init iostat info */
2502 spin_lock_init(&sbi
->iostat_lock
);
2503 sbi
->iostat_enable
= false;
2505 for (i
= 0; i
< NR_PAGE_TYPE
; i
++) {
2506 int n
= (i
== META
) ? 1: NR_TEMP_TYPE
;
2509 sbi
->write_io
[i
] = kmalloc(n
* sizeof(struct f2fs_bio_info
),
2511 if (!sbi
->write_io
[i
]) {
2516 for (j
= HOT
; j
< n
; j
++) {
2517 init_rwsem(&sbi
->write_io
[i
][j
].io_rwsem
);
2518 sbi
->write_io
[i
][j
].sbi
= sbi
;
2519 sbi
->write_io
[i
][j
].bio
= NULL
;
2520 spin_lock_init(&sbi
->write_io
[i
][j
].io_lock
);
2521 INIT_LIST_HEAD(&sbi
->write_io
[i
][j
].io_list
);
2525 init_rwsem(&sbi
->cp_rwsem
);
2526 init_waitqueue_head(&sbi
->cp_wait
);
2529 err
= init_percpu_info(sbi
);
2533 if (F2FS_IO_SIZE(sbi
) > 1) {
2534 sbi
->write_io_dummy
=
2535 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi
) - 1), 0);
2536 if (!sbi
->write_io_dummy
) {
2542 /* get an inode for meta space */
2543 sbi
->meta_inode
= f2fs_iget(sb
, F2FS_META_INO(sbi
));
2544 if (IS_ERR(sbi
->meta_inode
)) {
2545 f2fs_msg(sb
, KERN_ERR
, "Failed to read F2FS meta data inode");
2546 err
= PTR_ERR(sbi
->meta_inode
);
2550 err
= get_valid_checkpoint(sbi
);
2552 f2fs_msg(sb
, KERN_ERR
, "Failed to get valid F2FS checkpoint");
2553 goto free_meta_inode
;
2556 /* Initialize device list */
2557 err
= f2fs_scan_devices(sbi
);
2559 f2fs_msg(sb
, KERN_ERR
, "Failed to find devices");
2563 sbi
->total_valid_node_count
=
2564 le32_to_cpu(sbi
->ckpt
->valid_node_count
);
2565 percpu_counter_set(&sbi
->total_valid_inode_count
,
2566 le32_to_cpu(sbi
->ckpt
->valid_inode_count
));
2567 sbi
->user_block_count
= le64_to_cpu(sbi
->ckpt
->user_block_count
);
2568 sbi
->total_valid_block_count
=
2569 le64_to_cpu(sbi
->ckpt
->valid_block_count
);
2570 sbi
->last_valid_block_count
= sbi
->total_valid_block_count
;
2571 sbi
->reserved_blocks
= 0;
2572 sbi
->current_reserved_blocks
= 0;
2574 for (i
= 0; i
< NR_INODE_TYPE
; i
++) {
2575 INIT_LIST_HEAD(&sbi
->inode_list
[i
]);
2576 spin_lock_init(&sbi
->inode_lock
[i
]);
2579 init_extent_cache_info(sbi
);
2581 init_ino_entry_info(sbi
);
2583 /* setup f2fs internal modules */
2584 err
= build_segment_manager(sbi
);
2586 f2fs_msg(sb
, KERN_ERR
,
2587 "Failed to initialize F2FS segment manager");
2590 err
= build_node_manager(sbi
);
2592 f2fs_msg(sb
, KERN_ERR
,
2593 "Failed to initialize F2FS node manager");
2597 /* For write statistics */
2598 if (sb
->s_bdev
->bd_part
)
2599 sbi
->sectors_written_start
=
2600 (u64
)part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]);
2602 /* Read accumulated write IO statistics if exists */
2603 seg_i
= CURSEG_I(sbi
, CURSEG_HOT_NODE
);
2604 if (__exist_node_summaries(sbi
))
2605 sbi
->kbytes_written
=
2606 le64_to_cpu(seg_i
->journal
->info
.kbytes_written
);
2608 build_gc_manager(sbi
);
2610 /* get an inode for node space */
2611 sbi
->node_inode
= f2fs_iget(sb
, F2FS_NODE_INO(sbi
));
2612 if (IS_ERR(sbi
->node_inode
)) {
2613 f2fs_msg(sb
, KERN_ERR
, "Failed to read node inode");
2614 err
= PTR_ERR(sbi
->node_inode
);
2618 f2fs_join_shrinker(sbi
);
2620 err
= f2fs_build_stats(sbi
);
2624 /* read root inode and dentry */
2625 root
= f2fs_iget(sb
, F2FS_ROOT_INO(sbi
));
2627 f2fs_msg(sb
, KERN_ERR
, "Failed to read root inode");
2628 err
= PTR_ERR(root
);
2629 goto free_node_inode
;
2631 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
2634 goto free_node_inode
;
2637 sb
->s_root
= d_make_root(root
); /* allocate root dentry */
2640 goto free_root_inode
;
2643 err
= f2fs_register_sysfs(sbi
);
2645 goto free_root_inode
;
2649 * Turn on quotas which were not enabled for read-only mounts if
2650 * filesystem has quota feature, so that they are updated correctly.
2652 if (f2fs_sb_has_quota_ino(sb
) && !sb_rdonly(sb
)) {
2653 err
= f2fs_enable_quotas(sb
);
2655 f2fs_msg(sb
, KERN_ERR
,
2656 "Cannot turn on quotas: error %d", err
);
2661 /* if there are nt orphan nodes free them */
2662 err
= recover_orphan_inodes(sbi
);
2666 /* recover fsynced data */
2667 if (!test_opt(sbi
, DISABLE_ROLL_FORWARD
)) {
2669 * mount should be failed, when device has readonly mode, and
2670 * previous checkpoint was not done by clean system shutdown.
2672 if (bdev_read_only(sb
->s_bdev
) &&
2673 !is_set_ckpt_flags(sbi
, CP_UMOUNT_FLAG
)) {
2679 set_sbi_flag(sbi
, SBI_NEED_FSCK
);
2684 err
= recover_fsync_data(sbi
, false);
2687 f2fs_msg(sb
, KERN_ERR
,
2688 "Cannot recover all fsync data errno=%d", err
);
2692 err
= recover_fsync_data(sbi
, true);
2694 if (!f2fs_readonly(sb
) && err
> 0) {
2696 f2fs_msg(sb
, KERN_ERR
,
2697 "Need to recover fsync data");
2702 /* recover_fsync_data() cleared this already */
2703 clear_sbi_flag(sbi
, SBI_POR_DOING
);
2706 * If filesystem is not mounted as read-only then
2707 * do start the gc_thread.
2709 if (test_opt(sbi
, BG_GC
) && !f2fs_readonly(sb
)) {
2710 /* After POR, we can run background GC thread.*/
2711 err
= start_gc_thread(sbi
);
2717 /* recover broken superblock */
2719 err
= f2fs_commit_super(sbi
, true);
2720 f2fs_msg(sb
, KERN_INFO
,
2721 "Try to recover %dth superblock, ret: %d",
2722 sbi
->valid_super_block
? 1 : 2, err
);
2725 f2fs_msg(sbi
->sb
, KERN_NOTICE
, "Mounted with checkpoint version = %llx",
2726 cur_cp_version(F2FS_CKPT(sbi
)));
2727 f2fs_update_time(sbi
, CP_TIME
);
2728 f2fs_update_time(sbi
, REQ_TIME
);
2733 if (f2fs_sb_has_quota_ino(sb
) && !sb_rdonly(sb
))
2734 f2fs_quota_off_umount(sbi
->sb
);
2736 f2fs_sync_inode_meta(sbi
);
2738 * Some dirty meta pages can be produced by recover_orphan_inodes()
2739 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
2740 * followed by write_checkpoint() through f2fs_write_node_pages(), which
2741 * falls into an infinite loop in sync_meta_pages().
2743 truncate_inode_pages_final(META_MAPPING(sbi
));
2747 f2fs_unregister_sysfs(sbi
);
2752 truncate_inode_pages_final(NODE_MAPPING(sbi
));
2753 mutex_lock(&sbi
->umount_mutex
);
2754 release_ino_entry(sbi
, true);
2755 f2fs_leave_shrinker(sbi
);
2756 iput(sbi
->node_inode
);
2757 mutex_unlock(&sbi
->umount_mutex
);
2758 f2fs_destroy_stats(sbi
);
2760 destroy_node_manager(sbi
);
2762 destroy_segment_manager(sbi
);
2764 destroy_device_list(sbi
);
2767 make_bad_inode(sbi
->meta_inode
);
2768 iput(sbi
->meta_inode
);
2770 mempool_destroy(sbi
->write_io_dummy
);
2772 for (i
= 0; i
< NR_PAGE_TYPE
; i
++)
2773 kfree(sbi
->write_io
[i
]);
2774 destroy_percpu_info(sbi
);
2776 for (i
= 0; i
< MAXQUOTAS
; i
++)
2777 kfree(sbi
->s_qf_names
[i
]);
2783 if (sbi
->s_chksum_driver
)
2784 crypto_free_shash(sbi
->s_chksum_driver
);
2787 /* give only one another chance */
2790 shrink_dcache_sb(sb
);
2796 static struct dentry
*f2fs_mount(struct file_system_type
*fs_type
, int flags
,
2797 const char *dev_name
, void *data
)
2799 return mount_bdev(fs_type
, flags
, dev_name
, data
, f2fs_fill_super
);
2802 static void kill_f2fs_super(struct super_block
*sb
)
2805 set_sbi_flag(F2FS_SB(sb
), SBI_IS_CLOSE
);
2806 stop_gc_thread(F2FS_SB(sb
));
2807 stop_discard_thread(F2FS_SB(sb
));
2809 kill_block_super(sb
);
2812 static struct file_system_type f2fs_fs_type
= {
2813 .owner
= THIS_MODULE
,
2815 .mount
= f2fs_mount
,
2816 .kill_sb
= kill_f2fs_super
,
2817 .fs_flags
= FS_REQUIRES_DEV
,
2819 MODULE_ALIAS_FS("f2fs");
2821 static int __init
init_inodecache(void)
2823 f2fs_inode_cachep
= kmem_cache_create("f2fs_inode_cache",
2824 sizeof(struct f2fs_inode_info
), 0,
2825 SLAB_RECLAIM_ACCOUNT
, NULL
);
2826 if (!f2fs_inode_cachep
)
2831 static void destroy_inodecache(void)
2834 * Make sure all delayed rcu free inodes are flushed before we
2838 kmem_cache_destroy(f2fs_inode_cachep
);
2841 static int __init
init_f2fs_fs(void)
2845 f2fs_build_trace_ios();
2847 err
= init_inodecache();
2850 err
= create_node_manager_caches();
2852 goto free_inodecache
;
2853 err
= create_segment_manager_caches();
2855 goto free_node_manager_caches
;
2856 err
= create_checkpoint_caches();
2858 goto free_segment_manager_caches
;
2859 err
= create_extent_cache();
2861 goto free_checkpoint_caches
;
2862 err
= f2fs_init_sysfs();
2864 goto free_extent_cache
;
2865 err
= register_shrinker(&f2fs_shrinker_info
);
2868 err
= register_filesystem(&f2fs_fs_type
);
2871 err
= f2fs_create_root_stats();
2873 goto free_filesystem
;
2877 unregister_filesystem(&f2fs_fs_type
);
2879 unregister_shrinker(&f2fs_shrinker_info
);
2883 destroy_extent_cache();
2884 free_checkpoint_caches
:
2885 destroy_checkpoint_caches();
2886 free_segment_manager_caches
:
2887 destroy_segment_manager_caches();
2888 free_node_manager_caches
:
2889 destroy_node_manager_caches();
2891 destroy_inodecache();
2896 static void __exit
exit_f2fs_fs(void)
2898 f2fs_destroy_root_stats();
2899 unregister_filesystem(&f2fs_fs_type
);
2900 unregister_shrinker(&f2fs_shrinker_info
);
2902 destroy_extent_cache();
2903 destroy_checkpoint_caches();
2904 destroy_segment_manager_caches();
2905 destroy_node_manager_caches();
2906 destroy_inodecache();
2907 f2fs_destroy_trace_ios();
2910 module_init(init_f2fs_fs
)
2911 module_exit(exit_f2fs_fs
)
2913 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
2914 MODULE_DESCRIPTION("Flash Friendly File System");
2915 MODULE_LICENSE("GPL");