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/f2fs_fs.h>
26 #include <linux/sysfs.h>
35 #define CREATE_TRACE_POINTS
36 #include <trace/events/f2fs.h>
38 static struct proc_dir_entry
*f2fs_proc_root
;
39 static struct kmem_cache
*f2fs_inode_cachep
;
40 static struct kset
*f2fs_kset
;
42 /* f2fs-wide shrinker description */
43 static struct shrinker f2fs_shrinker_info
= {
44 .scan_objects
= f2fs_shrink_scan
,
45 .count_objects
= f2fs_shrink_count
,
46 .seeks
= DEFAULT_SEEKS
,
51 Opt_disable_roll_forward
,
60 Opt_disable_ext_identify
,
73 static match_table_t f2fs_tokens
= {
74 {Opt_gc_background
, "background_gc=%s"},
75 {Opt_disable_roll_forward
, "disable_roll_forward"},
76 {Opt_norecovery
, "norecovery"},
77 {Opt_discard
, "discard"},
78 {Opt_noheap
, "no_heap"},
79 {Opt_user_xattr
, "user_xattr"},
80 {Opt_nouser_xattr
, "nouser_xattr"},
83 {Opt_active_logs
, "active_logs=%u"},
84 {Opt_disable_ext_identify
, "disable_ext_identify"},
85 {Opt_inline_xattr
, "inline_xattr"},
86 {Opt_inline_data
, "inline_data"},
87 {Opt_inline_dentry
, "inline_dentry"},
88 {Opt_flush_merge
, "flush_merge"},
89 {Opt_nobarrier
, "nobarrier"},
90 {Opt_fastboot
, "fastboot"},
91 {Opt_extent_cache
, "extent_cache"},
92 {Opt_noextent_cache
, "noextent_cache"},
93 {Opt_noinline_data
, "noinline_data"},
97 /* Sysfs support for f2fs */
99 GC_THREAD
, /* struct f2fs_gc_thread */
100 SM_INFO
, /* struct f2fs_sm_info */
101 NM_INFO
, /* struct f2fs_nm_info */
102 F2FS_SBI
, /* struct f2fs_sb_info */
106 struct attribute attr
;
107 ssize_t (*show
)(struct f2fs_attr
*, struct f2fs_sb_info
*, char *);
108 ssize_t (*store
)(struct f2fs_attr
*, struct f2fs_sb_info
*,
109 const char *, size_t);
114 static unsigned char *__struct_ptr(struct f2fs_sb_info
*sbi
, int struct_type
)
116 if (struct_type
== GC_THREAD
)
117 return (unsigned char *)sbi
->gc_thread
;
118 else if (struct_type
== SM_INFO
)
119 return (unsigned char *)SM_I(sbi
);
120 else if (struct_type
== NM_INFO
)
121 return (unsigned char *)NM_I(sbi
);
122 else if (struct_type
== F2FS_SBI
)
123 return (unsigned char *)sbi
;
127 static ssize_t
f2fs_sbi_show(struct f2fs_attr
*a
,
128 struct f2fs_sb_info
*sbi
, char *buf
)
130 unsigned char *ptr
= NULL
;
133 ptr
= __struct_ptr(sbi
, a
->struct_type
);
137 ui
= (unsigned int *)(ptr
+ a
->offset
);
139 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
142 static ssize_t
f2fs_sbi_store(struct f2fs_attr
*a
,
143 struct f2fs_sb_info
*sbi
,
144 const char *buf
, size_t count
)
151 ptr
= __struct_ptr(sbi
, a
->struct_type
);
155 ui
= (unsigned int *)(ptr
+ a
->offset
);
157 ret
= kstrtoul(skip_spaces(buf
), 0, &t
);
164 static ssize_t
f2fs_attr_show(struct kobject
*kobj
,
165 struct attribute
*attr
, char *buf
)
167 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
169 struct f2fs_attr
*a
= container_of(attr
, struct f2fs_attr
, attr
);
171 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
174 static ssize_t
f2fs_attr_store(struct kobject
*kobj
, struct attribute
*attr
,
175 const char *buf
, size_t len
)
177 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
179 struct f2fs_attr
*a
= container_of(attr
, struct f2fs_attr
, attr
);
181 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
184 static void f2fs_sb_release(struct kobject
*kobj
)
186 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
188 complete(&sbi
->s_kobj_unregister
);
191 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
192 static struct f2fs_attr f2fs_attr_##_name = { \
193 .attr = {.name = __stringify(_name), .mode = _mode }, \
196 .struct_type = _struct_type, \
200 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
201 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
202 f2fs_sbi_show, f2fs_sbi_store, \
203 offsetof(struct struct_name, elname))
205 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_min_sleep_time
, min_sleep_time
);
206 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_max_sleep_time
, max_sleep_time
);
207 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_no_gc_sleep_time
, no_gc_sleep_time
);
208 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_idle
, gc_idle
);
209 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, reclaim_segments
, rec_prefree_segments
);
210 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, max_small_discards
, max_discards
);
211 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, batched_trim_sections
, trim_sections
);
212 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, ipu_policy
, ipu_policy
);
213 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, min_ipu_util
, min_ipu_util
);
214 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, min_fsync_blocks
, min_fsync_blocks
);
215 F2FS_RW_ATTR(NM_INFO
, f2fs_nm_info
, ram_thresh
, ram_thresh
);
216 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, max_victim_search
, max_victim_search
);
217 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, dir_level
, dir_level
);
219 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
220 static struct attribute
*f2fs_attrs
[] = {
221 ATTR_LIST(gc_min_sleep_time
),
222 ATTR_LIST(gc_max_sleep_time
),
223 ATTR_LIST(gc_no_gc_sleep_time
),
225 ATTR_LIST(reclaim_segments
),
226 ATTR_LIST(max_small_discards
),
227 ATTR_LIST(batched_trim_sections
),
228 ATTR_LIST(ipu_policy
),
229 ATTR_LIST(min_ipu_util
),
230 ATTR_LIST(min_fsync_blocks
),
231 ATTR_LIST(max_victim_search
),
232 ATTR_LIST(dir_level
),
233 ATTR_LIST(ram_thresh
),
237 static const struct sysfs_ops f2fs_attr_ops
= {
238 .show
= f2fs_attr_show
,
239 .store
= f2fs_attr_store
,
242 static struct kobj_type f2fs_ktype
= {
243 .default_attrs
= f2fs_attrs
,
244 .sysfs_ops
= &f2fs_attr_ops
,
245 .release
= f2fs_sb_release
,
248 void f2fs_msg(struct super_block
*sb
, const char *level
, const char *fmt
, ...)
250 struct va_format vaf
;
256 printk("%sF2FS-fs (%s): %pV\n", level
, sb
->s_id
, &vaf
);
260 static void init_once(void *foo
)
262 struct f2fs_inode_info
*fi
= (struct f2fs_inode_info
*) foo
;
264 inode_init_once(&fi
->vfs_inode
);
267 static int parse_options(struct super_block
*sb
, char *options
)
269 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
270 struct request_queue
*q
;
271 substring_t args
[MAX_OPT_ARGS
];
278 while ((p
= strsep(&options
, ",")) != NULL
) {
283 * Initialize args struct so we know whether arg was
284 * found; some options take optional arguments.
286 args
[0].to
= args
[0].from
= NULL
;
287 token
= match_token(p
, f2fs_tokens
, args
);
290 case Opt_gc_background
:
291 name
= match_strdup(&args
[0]);
295 if (strlen(name
) == 2 && !strncmp(name
, "on", 2))
297 else if (strlen(name
) == 3 && !strncmp(name
, "off", 3))
298 clear_opt(sbi
, BG_GC
);
305 case Opt_disable_roll_forward
:
306 set_opt(sbi
, DISABLE_ROLL_FORWARD
);
309 /* this option mounts f2fs with ro */
310 set_opt(sbi
, DISABLE_ROLL_FORWARD
);
311 if (!f2fs_readonly(sb
))
315 q
= bdev_get_queue(sb
->s_bdev
);
316 if (blk_queue_discard(q
)) {
317 set_opt(sbi
, DISCARD
);
319 f2fs_msg(sb
, KERN_WARNING
,
320 "mounting with \"discard\" option, but "
321 "the device does not support discard");
325 set_opt(sbi
, NOHEAP
);
327 #ifdef CONFIG_F2FS_FS_XATTR
329 set_opt(sbi
, XATTR_USER
);
331 case Opt_nouser_xattr
:
332 clear_opt(sbi
, XATTR_USER
);
334 case Opt_inline_xattr
:
335 set_opt(sbi
, INLINE_XATTR
);
339 f2fs_msg(sb
, KERN_INFO
,
340 "user_xattr options not supported");
342 case Opt_nouser_xattr
:
343 f2fs_msg(sb
, KERN_INFO
,
344 "nouser_xattr options not supported");
346 case Opt_inline_xattr
:
347 f2fs_msg(sb
, KERN_INFO
,
348 "inline_xattr options not supported");
351 #ifdef CONFIG_F2FS_FS_POSIX_ACL
353 set_opt(sbi
, POSIX_ACL
);
356 clear_opt(sbi
, POSIX_ACL
);
360 f2fs_msg(sb
, KERN_INFO
, "acl options not supported");
363 f2fs_msg(sb
, KERN_INFO
, "noacl options not supported");
366 case Opt_active_logs
:
367 if (args
->from
&& match_int(args
, &arg
))
369 if (arg
!= 2 && arg
!= 4 && arg
!= NR_CURSEG_TYPE
)
371 sbi
->active_logs
= arg
;
373 case Opt_disable_ext_identify
:
374 set_opt(sbi
, DISABLE_EXT_IDENTIFY
);
376 case Opt_inline_data
:
377 set_opt(sbi
, INLINE_DATA
);
379 case Opt_inline_dentry
:
380 set_opt(sbi
, INLINE_DENTRY
);
382 case Opt_flush_merge
:
383 set_opt(sbi
, FLUSH_MERGE
);
386 set_opt(sbi
, NOBARRIER
);
389 set_opt(sbi
, FASTBOOT
);
391 case Opt_extent_cache
:
392 set_opt(sbi
, EXTENT_CACHE
);
394 case Opt_noextent_cache
:
395 clear_opt(sbi
, EXTENT_CACHE
);
397 case Opt_noinline_data
:
398 clear_opt(sbi
, INLINE_DATA
);
401 f2fs_msg(sb
, KERN_ERR
,
402 "Unrecognized mount option \"%s\" or missing value",
410 static struct inode
*f2fs_alloc_inode(struct super_block
*sb
)
412 struct f2fs_inode_info
*fi
;
414 fi
= kmem_cache_alloc(f2fs_inode_cachep
, GFP_F2FS_ZERO
);
418 init_once((void *) fi
);
420 /* Initialize f2fs-specific inode info */
421 fi
->vfs_inode
.i_version
= 1;
422 atomic_set(&fi
->dirty_pages
, 0);
423 fi
->i_current_depth
= 1;
425 rwlock_init(&fi
->ext_lock
);
426 init_rwsem(&fi
->i_sem
);
427 INIT_RADIX_TREE(&fi
->inmem_root
, GFP_NOFS
);
428 INIT_LIST_HEAD(&fi
->inmem_pages
);
429 mutex_init(&fi
->inmem_lock
);
431 set_inode_flag(fi
, FI_NEW_INODE
);
433 if (test_opt(F2FS_SB(sb
), INLINE_XATTR
))
434 set_inode_flag(fi
, FI_INLINE_XATTR
);
436 /* Will be used by directory only */
437 fi
->i_dir_level
= F2FS_SB(sb
)->dir_level
;
439 #ifdef CONFIG_F2FS_FS_ENCRYPTION
440 fi
->i_crypt_info
= NULL
;
442 return &fi
->vfs_inode
;
445 static int f2fs_drop_inode(struct inode
*inode
)
448 * This is to avoid a deadlock condition like below.
449 * writeback_single_inode(inode)
450 * - f2fs_write_data_page
451 * - f2fs_gc -> iput -> evict
452 * - inode_wait_for_writeback(inode)
454 if (!inode_unhashed(inode
) && inode
->i_state
& I_SYNC
) {
455 if (!inode
->i_nlink
&& !is_bad_inode(inode
)) {
456 spin_unlock(&inode
->i_lock
);
458 /* some remained atomic pages should discarded */
459 if (f2fs_is_atomic_file(inode
))
460 commit_inmem_pages(inode
, true);
462 sb_start_intwrite(inode
->i_sb
);
463 i_size_write(inode
, 0);
465 if (F2FS_HAS_BLOCKS(inode
))
466 f2fs_truncate(inode
);
468 sb_end_intwrite(inode
->i_sb
);
470 #ifdef CONFIG_F2FS_FS_ENCRYPTION
471 if (F2FS_I(inode
)->i_crypt_info
)
472 f2fs_free_encryption_info(inode
,
473 F2FS_I(inode
)->i_crypt_info
);
475 spin_lock(&inode
->i_lock
);
479 return generic_drop_inode(inode
);
483 * f2fs_dirty_inode() is called from __mark_inode_dirty()
485 * We should call set_dirty_inode to write the dirty inode through write_inode.
487 static void f2fs_dirty_inode(struct inode
*inode
, int flags
)
489 set_inode_flag(F2FS_I(inode
), FI_DIRTY_INODE
);
492 static void f2fs_i_callback(struct rcu_head
*head
)
494 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
495 kmem_cache_free(f2fs_inode_cachep
, F2FS_I(inode
));
498 static void f2fs_destroy_inode(struct inode
*inode
)
500 call_rcu(&inode
->i_rcu
, f2fs_i_callback
);
503 static void f2fs_put_super(struct super_block
*sb
)
505 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
508 remove_proc_entry("segment_info", sbi
->s_proc
);
509 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
511 kobject_del(&sbi
->s_kobj
);
515 /* prevent remaining shrinker jobs */
516 mutex_lock(&sbi
->umount_mutex
);
519 * We don't need to do checkpoint when superblock is clean.
520 * But, the previous checkpoint was not done by umount, it needs to do
521 * clean checkpoint again.
523 if (is_sbi_flag_set(sbi
, SBI_IS_DIRTY
) ||
524 !is_set_ckpt_flags(F2FS_CKPT(sbi
), CP_UMOUNT_FLAG
)) {
525 struct cp_control cpc
= {
528 write_checkpoint(sbi
, &cpc
);
531 /* write_checkpoint can update stat informaion */
532 f2fs_destroy_stats(sbi
);
535 * normally superblock is clean, so we need to release this.
536 * In addition, EIO will skip do checkpoint, we need this as well.
538 release_dirty_inode(sbi
);
539 release_discard_addrs(sbi
);
541 f2fs_leave_shrinker(sbi
);
542 mutex_unlock(&sbi
->umount_mutex
);
544 iput(sbi
->node_inode
);
545 iput(sbi
->meta_inode
);
547 /* destroy f2fs internal modules */
548 destroy_node_manager(sbi
);
549 destroy_segment_manager(sbi
);
552 kobject_put(&sbi
->s_kobj
);
553 wait_for_completion(&sbi
->s_kobj_unregister
);
555 sb
->s_fs_info
= NULL
;
556 brelse(sbi
->raw_super_buf
);
560 int f2fs_sync_fs(struct super_block
*sb
, int sync
)
562 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
564 trace_f2fs_sync_fs(sb
, sync
);
567 struct cp_control cpc
;
569 cpc
.reason
= __get_cp_reason(sbi
);
571 mutex_lock(&sbi
->gc_mutex
);
572 write_checkpoint(sbi
, &cpc
);
573 mutex_unlock(&sbi
->gc_mutex
);
575 f2fs_balance_fs(sbi
);
577 f2fs_trace_ios(NULL
, 1);
582 static int f2fs_freeze(struct super_block
*sb
)
586 if (f2fs_readonly(sb
))
589 err
= f2fs_sync_fs(sb
, 1);
593 static int f2fs_unfreeze(struct super_block
*sb
)
598 static int f2fs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
600 struct super_block
*sb
= dentry
->d_sb
;
601 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
602 u64 id
= huge_encode_dev(sb
->s_bdev
->bd_dev
);
603 block_t total_count
, user_block_count
, start_count
, ovp_count
;
605 total_count
= le64_to_cpu(sbi
->raw_super
->block_count
);
606 user_block_count
= sbi
->user_block_count
;
607 start_count
= le32_to_cpu(sbi
->raw_super
->segment0_blkaddr
);
608 ovp_count
= SM_I(sbi
)->ovp_segments
<< sbi
->log_blocks_per_seg
;
609 buf
->f_type
= F2FS_SUPER_MAGIC
;
610 buf
->f_bsize
= sbi
->blocksize
;
612 buf
->f_blocks
= total_count
- start_count
;
613 buf
->f_bfree
= buf
->f_blocks
- valid_user_blocks(sbi
) - ovp_count
;
614 buf
->f_bavail
= user_block_count
- valid_user_blocks(sbi
);
616 buf
->f_files
= sbi
->total_node_count
- F2FS_RESERVED_NODE_NUM
;
617 buf
->f_ffree
= buf
->f_files
- valid_inode_count(sbi
);
619 buf
->f_namelen
= F2FS_NAME_LEN
;
620 buf
->f_fsid
.val
[0] = (u32
)id
;
621 buf
->f_fsid
.val
[1] = (u32
)(id
>> 32);
626 static int f2fs_show_options(struct seq_file
*seq
, struct dentry
*root
)
628 struct f2fs_sb_info
*sbi
= F2FS_SB(root
->d_sb
);
630 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, BG_GC
))
631 seq_printf(seq
, ",background_gc=%s", "on");
633 seq_printf(seq
, ",background_gc=%s", "off");
634 if (test_opt(sbi
, DISABLE_ROLL_FORWARD
))
635 seq_puts(seq
, ",disable_roll_forward");
636 if (test_opt(sbi
, DISCARD
))
637 seq_puts(seq
, ",discard");
638 if (test_opt(sbi
, NOHEAP
))
639 seq_puts(seq
, ",no_heap_alloc");
640 #ifdef CONFIG_F2FS_FS_XATTR
641 if (test_opt(sbi
, XATTR_USER
))
642 seq_puts(seq
, ",user_xattr");
644 seq_puts(seq
, ",nouser_xattr");
645 if (test_opt(sbi
, INLINE_XATTR
))
646 seq_puts(seq
, ",inline_xattr");
648 #ifdef CONFIG_F2FS_FS_POSIX_ACL
649 if (test_opt(sbi
, POSIX_ACL
))
650 seq_puts(seq
, ",acl");
652 seq_puts(seq
, ",noacl");
654 if (test_opt(sbi
, DISABLE_EXT_IDENTIFY
))
655 seq_puts(seq
, ",disable_ext_identify");
656 if (test_opt(sbi
, INLINE_DATA
))
657 seq_puts(seq
, ",inline_data");
659 seq_puts(seq
, ",noinline_data");
660 if (test_opt(sbi
, INLINE_DENTRY
))
661 seq_puts(seq
, ",inline_dentry");
662 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, FLUSH_MERGE
))
663 seq_puts(seq
, ",flush_merge");
664 if (test_opt(sbi
, NOBARRIER
))
665 seq_puts(seq
, ",nobarrier");
666 if (test_opt(sbi
, FASTBOOT
))
667 seq_puts(seq
, ",fastboot");
668 if (test_opt(sbi
, EXTENT_CACHE
))
669 seq_puts(seq
, ",extent_cache");
671 seq_puts(seq
, ",noextent_cache");
672 seq_printf(seq
, ",active_logs=%u", sbi
->active_logs
);
677 static int segment_info_seq_show(struct seq_file
*seq
, void *offset
)
679 struct super_block
*sb
= seq
->private;
680 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
681 unsigned int total_segs
=
682 le32_to_cpu(sbi
->raw_super
->segment_count_main
);
685 seq_puts(seq
, "format: segment_type|valid_blocks\n"
686 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
688 for (i
= 0; i
< total_segs
; i
++) {
689 struct seg_entry
*se
= get_seg_entry(sbi
, i
);
692 seq_printf(seq
, "%-5d", i
);
693 seq_printf(seq
, "%d|%-3u", se
->type
,
694 get_valid_blocks(sbi
, i
, 1));
695 if ((i
% 10) == 9 || i
== (total_segs
- 1))
704 static int segment_info_open_fs(struct inode
*inode
, struct file
*file
)
706 return single_open(file
, segment_info_seq_show
, PDE_DATA(inode
));
709 static const struct file_operations f2fs_seq_segment_info_fops
= {
710 .owner
= THIS_MODULE
,
711 .open
= segment_info_open_fs
,
714 .release
= single_release
,
717 static void default_options(struct f2fs_sb_info
*sbi
)
719 /* init some FS parameters */
720 sbi
->active_logs
= NR_CURSEG_TYPE
;
723 set_opt(sbi
, INLINE_DATA
);
725 #ifdef CONFIG_F2FS_FS_XATTR
726 set_opt(sbi
, XATTR_USER
);
728 #ifdef CONFIG_F2FS_FS_POSIX_ACL
729 set_opt(sbi
, POSIX_ACL
);
733 static int f2fs_remount(struct super_block
*sb
, int *flags
, char *data
)
735 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
736 struct f2fs_mount_info org_mount_opt
;
737 int err
, active_logs
;
738 bool need_restart_gc
= false;
739 bool need_stop_gc
= false;
744 * Save the old mount options in case we
745 * need to restore them.
747 org_mount_opt
= sbi
->mount_opt
;
748 active_logs
= sbi
->active_logs
;
750 sbi
->mount_opt
.opt
= 0;
751 default_options(sbi
);
753 /* parse mount options */
754 err
= parse_options(sb
, data
);
759 * Previous and new state of filesystem is RO,
760 * so skip checking GC and FLUSH_MERGE conditions.
762 if (f2fs_readonly(sb
) && (*flags
& MS_RDONLY
))
766 * We stop the GC thread if FS is mounted as RO
767 * or if background_gc = off is passed in mount
768 * option. Also sync the filesystem.
770 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, BG_GC
)) {
771 if (sbi
->gc_thread
) {
774 need_restart_gc
= true;
776 } else if (!sbi
->gc_thread
) {
777 err
= start_gc_thread(sbi
);
784 * We stop issue flush thread if FS is mounted as RO
785 * or if flush_merge is not passed in mount option.
787 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, FLUSH_MERGE
)) {
788 destroy_flush_cmd_control(sbi
);
789 } else if (!SM_I(sbi
)->cmd_control_info
) {
790 err
= create_flush_cmd_control(sbi
);
795 /* Update the POSIXACL Flag */
796 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
797 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
800 if (need_restart_gc
) {
801 if (start_gc_thread(sbi
))
802 f2fs_msg(sbi
->sb
, KERN_WARNING
,
803 "background gc thread has stopped");
804 } else if (need_stop_gc
) {
808 sbi
->mount_opt
= org_mount_opt
;
809 sbi
->active_logs
= active_logs
;
813 static struct super_operations f2fs_sops
= {
814 .alloc_inode
= f2fs_alloc_inode
,
815 .drop_inode
= f2fs_drop_inode
,
816 .destroy_inode
= f2fs_destroy_inode
,
817 .write_inode
= f2fs_write_inode
,
818 .dirty_inode
= f2fs_dirty_inode
,
819 .show_options
= f2fs_show_options
,
820 .evict_inode
= f2fs_evict_inode
,
821 .put_super
= f2fs_put_super
,
822 .sync_fs
= f2fs_sync_fs
,
823 .freeze_fs
= f2fs_freeze
,
824 .unfreeze_fs
= f2fs_unfreeze
,
825 .statfs
= f2fs_statfs
,
826 .remount_fs
= f2fs_remount
,
829 static struct inode
*f2fs_nfs_get_inode(struct super_block
*sb
,
830 u64 ino
, u32 generation
)
832 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
835 if (check_nid_range(sbi
, ino
))
836 return ERR_PTR(-ESTALE
);
839 * f2fs_iget isn't quite right if the inode is currently unallocated!
840 * However f2fs_iget currently does appropriate checks to handle stale
841 * inodes so everything is OK.
843 inode
= f2fs_iget(sb
, ino
);
845 return ERR_CAST(inode
);
846 if (unlikely(generation
&& inode
->i_generation
!= generation
)) {
847 /* we didn't find the right inode.. */
849 return ERR_PTR(-ESTALE
);
854 static struct dentry
*f2fs_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
855 int fh_len
, int fh_type
)
857 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
861 static struct dentry
*f2fs_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
862 int fh_len
, int fh_type
)
864 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
868 static const struct export_operations f2fs_export_ops
= {
869 .fh_to_dentry
= f2fs_fh_to_dentry
,
870 .fh_to_parent
= f2fs_fh_to_parent
,
871 .get_parent
= f2fs_get_parent
,
874 static loff_t
max_file_size(unsigned bits
)
876 loff_t result
= (DEF_ADDRS_PER_INODE
- F2FS_INLINE_XATTR_ADDRS
);
877 loff_t leaf_count
= ADDRS_PER_BLOCK
;
879 /* two direct node blocks */
880 result
+= (leaf_count
* 2);
882 /* two indirect node blocks */
883 leaf_count
*= NIDS_PER_BLOCK
;
884 result
+= (leaf_count
* 2);
886 /* one double indirect node block */
887 leaf_count
*= NIDS_PER_BLOCK
;
888 result
+= leaf_count
;
894 static int sanity_check_raw_super(struct super_block
*sb
,
895 struct f2fs_super_block
*raw_super
)
897 unsigned int blocksize
;
899 if (F2FS_SUPER_MAGIC
!= le32_to_cpu(raw_super
->magic
)) {
900 f2fs_msg(sb
, KERN_INFO
,
901 "Magic Mismatch, valid(0x%x) - read(0x%x)",
902 F2FS_SUPER_MAGIC
, le32_to_cpu(raw_super
->magic
));
906 /* Currently, support only 4KB page cache size */
907 if (F2FS_BLKSIZE
!= PAGE_CACHE_SIZE
) {
908 f2fs_msg(sb
, KERN_INFO
,
909 "Invalid page_cache_size (%lu), supports only 4KB\n",
914 /* Currently, support only 4KB block size */
915 blocksize
= 1 << le32_to_cpu(raw_super
->log_blocksize
);
916 if (blocksize
!= F2FS_BLKSIZE
) {
917 f2fs_msg(sb
, KERN_INFO
,
918 "Invalid blocksize (%u), supports only 4KB\n",
923 /* Currently, support 512/1024/2048/4096 bytes sector size */
924 if (le32_to_cpu(raw_super
->log_sectorsize
) >
925 F2FS_MAX_LOG_SECTOR_SIZE
||
926 le32_to_cpu(raw_super
->log_sectorsize
) <
927 F2FS_MIN_LOG_SECTOR_SIZE
) {
928 f2fs_msg(sb
, KERN_INFO
, "Invalid log sectorsize (%u)",
929 le32_to_cpu(raw_super
->log_sectorsize
));
932 if (le32_to_cpu(raw_super
->log_sectors_per_block
) +
933 le32_to_cpu(raw_super
->log_sectorsize
) !=
934 F2FS_MAX_LOG_SECTOR_SIZE
) {
935 f2fs_msg(sb
, KERN_INFO
,
936 "Invalid log sectors per block(%u) log sectorsize(%u)",
937 le32_to_cpu(raw_super
->log_sectors_per_block
),
938 le32_to_cpu(raw_super
->log_sectorsize
));
944 static int sanity_check_ckpt(struct f2fs_sb_info
*sbi
)
946 unsigned int total
, fsmeta
;
947 struct f2fs_super_block
*raw_super
= F2FS_RAW_SUPER(sbi
);
948 struct f2fs_checkpoint
*ckpt
= F2FS_CKPT(sbi
);
950 total
= le32_to_cpu(raw_super
->segment_count
);
951 fsmeta
= le32_to_cpu(raw_super
->segment_count_ckpt
);
952 fsmeta
+= le32_to_cpu(raw_super
->segment_count_sit
);
953 fsmeta
+= le32_to_cpu(raw_super
->segment_count_nat
);
954 fsmeta
+= le32_to_cpu(ckpt
->rsvd_segment_count
);
955 fsmeta
+= le32_to_cpu(raw_super
->segment_count_ssa
);
957 if (unlikely(fsmeta
>= total
))
960 if (unlikely(f2fs_cp_error(sbi
))) {
961 f2fs_msg(sbi
->sb
, KERN_ERR
, "A bug case: need to run fsck");
967 static void init_sb_info(struct f2fs_sb_info
*sbi
)
969 struct f2fs_super_block
*raw_super
= sbi
->raw_super
;
972 sbi
->log_sectors_per_block
=
973 le32_to_cpu(raw_super
->log_sectors_per_block
);
974 sbi
->log_blocksize
= le32_to_cpu(raw_super
->log_blocksize
);
975 sbi
->blocksize
= 1 << sbi
->log_blocksize
;
976 sbi
->log_blocks_per_seg
= le32_to_cpu(raw_super
->log_blocks_per_seg
);
977 sbi
->blocks_per_seg
= 1 << sbi
->log_blocks_per_seg
;
978 sbi
->segs_per_sec
= le32_to_cpu(raw_super
->segs_per_sec
);
979 sbi
->secs_per_zone
= le32_to_cpu(raw_super
->secs_per_zone
);
980 sbi
->total_sections
= le32_to_cpu(raw_super
->section_count
);
981 sbi
->total_node_count
=
982 (le32_to_cpu(raw_super
->segment_count_nat
) / 2)
983 * sbi
->blocks_per_seg
* NAT_ENTRY_PER_BLOCK
;
984 sbi
->root_ino_num
= le32_to_cpu(raw_super
->root_ino
);
985 sbi
->node_ino_num
= le32_to_cpu(raw_super
->node_ino
);
986 sbi
->meta_ino_num
= le32_to_cpu(raw_super
->meta_ino
);
987 sbi
->cur_victim_sec
= NULL_SECNO
;
988 sbi
->max_victim_search
= DEF_MAX_VICTIM_SEARCH
;
990 for (i
= 0; i
< NR_COUNT_TYPE
; i
++)
991 atomic_set(&sbi
->nr_pages
[i
], 0);
993 sbi
->dir_level
= DEF_DIR_LEVEL
;
994 clear_sbi_flag(sbi
, SBI_NEED_FSCK
);
996 INIT_LIST_HEAD(&sbi
->s_list
);
997 mutex_init(&sbi
->umount_mutex
);
1001 * Read f2fs raw super block.
1002 * Because we have two copies of super block, so read the first one at first,
1003 * if the first one is invalid, move to read the second one.
1005 static int read_raw_super_block(struct super_block
*sb
,
1006 struct f2fs_super_block
**raw_super
,
1007 struct buffer_head
**raw_super_buf
,
1011 struct buffer_head
*buffer
;
1012 struct f2fs_super_block
*super
;
1016 buffer
= sb_bread(sb
, block
);
1019 f2fs_msg(sb
, KERN_ERR
, "Unable to read %dth superblock",
1030 super
= (struct f2fs_super_block
*)
1031 ((char *)(buffer
)->b_data
+ F2FS_SUPER_OFFSET
);
1033 /* sanity checking of raw super */
1034 if (sanity_check_raw_super(sb
, super
)) {
1037 f2fs_msg(sb
, KERN_ERR
,
1038 "Can't find valid F2FS filesystem in %dth superblock",
1050 *raw_super_buf
= buffer
;
1053 /* already have a valid superblock */
1057 /* check the validity of the second superblock */
1064 /* No valid superblock */
1071 int f2fs_commit_super(struct f2fs_sb_info
*sbi
, bool recover
)
1073 struct buffer_head
*sbh
= sbi
->raw_super_buf
;
1074 sector_t block
= sbh
->b_blocknr
;
1077 /* write back-up superblock first */
1078 sbh
->b_blocknr
= block
? 0 : 1;
1079 mark_buffer_dirty(sbh
);
1080 err
= sync_dirty_buffer(sbh
);
1082 sbh
->b_blocknr
= block
;
1084 /* if we are in recovery path, skip writing valid superblock */
1088 /* write current valid superblock */
1089 mark_buffer_dirty(sbh
);
1090 err
= sync_dirty_buffer(sbh
);
1092 clear_buffer_write_io_error(sbh
);
1093 set_buffer_uptodate(sbh
);
1097 static int f2fs_fill_super(struct super_block
*sb
, void *data
, int silent
)
1099 struct f2fs_sb_info
*sbi
;
1100 struct f2fs_super_block
*raw_super
;
1101 struct buffer_head
*raw_super_buf
;
1104 bool retry
= true, need_fsck
= false;
1105 char *options
= NULL
;
1111 raw_super_buf
= NULL
;
1114 /* allocate memory for f2fs-specific super block info */
1115 sbi
= kzalloc(sizeof(struct f2fs_sb_info
), GFP_KERNEL
);
1119 /* set a block size */
1120 if (unlikely(!sb_set_blocksize(sb
, F2FS_BLKSIZE
))) {
1121 f2fs_msg(sb
, KERN_ERR
, "unable to set blocksize");
1125 err
= read_raw_super_block(sb
, &raw_super
, &raw_super_buf
, &recovery
);
1129 sb
->s_fs_info
= sbi
;
1130 default_options(sbi
);
1131 /* parse mount options */
1132 options
= kstrdup((const char *)data
, GFP_KERNEL
);
1133 if (data
&& !options
) {
1138 err
= parse_options(sb
, options
);
1142 sb
->s_maxbytes
= max_file_size(le32_to_cpu(raw_super
->log_blocksize
));
1143 sb
->s_max_links
= F2FS_LINK_MAX
;
1144 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
1146 sb
->s_op
= &f2fs_sops
;
1147 sb
->s_xattr
= f2fs_xattr_handlers
;
1148 sb
->s_export_op
= &f2fs_export_ops
;
1149 sb
->s_magic
= F2FS_SUPER_MAGIC
;
1150 sb
->s_time_gran
= 1;
1151 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
1152 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
1153 memcpy(sb
->s_uuid
, raw_super
->uuid
, sizeof(raw_super
->uuid
));
1155 /* init f2fs-specific super block info */
1157 sbi
->raw_super
= raw_super
;
1158 sbi
->raw_super_buf
= raw_super_buf
;
1159 mutex_init(&sbi
->gc_mutex
);
1160 mutex_init(&sbi
->writepages
);
1161 mutex_init(&sbi
->cp_mutex
);
1162 init_rwsem(&sbi
->node_write
);
1163 clear_sbi_flag(sbi
, SBI_POR_DOING
);
1164 spin_lock_init(&sbi
->stat_lock
);
1166 init_rwsem(&sbi
->read_io
.io_rwsem
);
1167 sbi
->read_io
.sbi
= sbi
;
1168 sbi
->read_io
.bio
= NULL
;
1169 for (i
= 0; i
< NR_PAGE_TYPE
; i
++) {
1170 init_rwsem(&sbi
->write_io
[i
].io_rwsem
);
1171 sbi
->write_io
[i
].sbi
= sbi
;
1172 sbi
->write_io
[i
].bio
= NULL
;
1175 init_rwsem(&sbi
->cp_rwsem
);
1176 init_waitqueue_head(&sbi
->cp_wait
);
1179 /* get an inode for meta space */
1180 sbi
->meta_inode
= f2fs_iget(sb
, F2FS_META_INO(sbi
));
1181 if (IS_ERR(sbi
->meta_inode
)) {
1182 f2fs_msg(sb
, KERN_ERR
, "Failed to read F2FS meta data inode");
1183 err
= PTR_ERR(sbi
->meta_inode
);
1187 err
= get_valid_checkpoint(sbi
);
1189 f2fs_msg(sb
, KERN_ERR
, "Failed to get valid F2FS checkpoint");
1190 goto free_meta_inode
;
1193 /* sanity checking of checkpoint */
1195 if (sanity_check_ckpt(sbi
)) {
1196 f2fs_msg(sb
, KERN_ERR
, "Invalid F2FS checkpoint");
1200 sbi
->total_valid_node_count
=
1201 le32_to_cpu(sbi
->ckpt
->valid_node_count
);
1202 sbi
->total_valid_inode_count
=
1203 le32_to_cpu(sbi
->ckpt
->valid_inode_count
);
1204 sbi
->user_block_count
= le64_to_cpu(sbi
->ckpt
->user_block_count
);
1205 sbi
->total_valid_block_count
=
1206 le64_to_cpu(sbi
->ckpt
->valid_block_count
);
1207 sbi
->last_valid_block_count
= sbi
->total_valid_block_count
;
1208 sbi
->alloc_valid_block_count
= 0;
1209 INIT_LIST_HEAD(&sbi
->dir_inode_list
);
1210 spin_lock_init(&sbi
->dir_inode_lock
);
1212 init_extent_cache_info(sbi
);
1214 init_ino_entry_info(sbi
);
1216 /* setup f2fs internal modules */
1217 err
= build_segment_manager(sbi
);
1219 f2fs_msg(sb
, KERN_ERR
,
1220 "Failed to initialize F2FS segment manager");
1223 err
= build_node_manager(sbi
);
1225 f2fs_msg(sb
, KERN_ERR
,
1226 "Failed to initialize F2FS node manager");
1230 build_gc_manager(sbi
);
1232 /* get an inode for node space */
1233 sbi
->node_inode
= f2fs_iget(sb
, F2FS_NODE_INO(sbi
));
1234 if (IS_ERR(sbi
->node_inode
)) {
1235 f2fs_msg(sb
, KERN_ERR
, "Failed to read node inode");
1236 err
= PTR_ERR(sbi
->node_inode
);
1240 f2fs_join_shrinker(sbi
);
1242 /* if there are nt orphan nodes free them */
1243 recover_orphan_inodes(sbi
);
1245 /* read root inode and dentry */
1246 root
= f2fs_iget(sb
, F2FS_ROOT_INO(sbi
));
1248 f2fs_msg(sb
, KERN_ERR
, "Failed to read root inode");
1249 err
= PTR_ERR(root
);
1250 goto free_node_inode
;
1252 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
1255 goto free_node_inode
;
1258 sb
->s_root
= d_make_root(root
); /* allocate root dentry */
1261 goto free_root_inode
;
1264 err
= f2fs_build_stats(sbi
);
1266 goto free_root_inode
;
1269 sbi
->s_proc
= proc_mkdir(sb
->s_id
, f2fs_proc_root
);
1272 proc_create_data("segment_info", S_IRUGO
, sbi
->s_proc
,
1273 &f2fs_seq_segment_info_fops
, sb
);
1275 sbi
->s_kobj
.kset
= f2fs_kset
;
1276 init_completion(&sbi
->s_kobj_unregister
);
1277 err
= kobject_init_and_add(&sbi
->s_kobj
, &f2fs_ktype
, NULL
,
1282 /* recover fsynced data */
1283 if (!test_opt(sbi
, DISABLE_ROLL_FORWARD
)) {
1285 * mount should be failed, when device has readonly mode, and
1286 * previous checkpoint was not done by clean system shutdown.
1288 if (bdev_read_only(sb
->s_bdev
) &&
1289 !is_set_ckpt_flags(sbi
->ckpt
, CP_UMOUNT_FLAG
)) {
1295 set_sbi_flag(sbi
, SBI_NEED_FSCK
);
1297 err
= recover_fsync_data(sbi
);
1300 f2fs_msg(sb
, KERN_ERR
,
1301 "Cannot recover all fsync data errno=%ld", err
);
1307 * If filesystem is not mounted as read-only then
1308 * do start the gc_thread.
1310 if (test_opt(sbi
, BG_GC
) && !f2fs_readonly(sb
)) {
1311 /* After POR, we can run background GC thread.*/
1312 err
= start_gc_thread(sbi
);
1318 /* recover broken superblock */
1319 if (recovery
&& !f2fs_readonly(sb
) && !bdev_read_only(sb
->s_bdev
)) {
1320 f2fs_msg(sb
, KERN_INFO
, "Recover invalid superblock");
1321 f2fs_commit_super(sbi
, true);
1327 kobject_del(&sbi
->s_kobj
);
1330 remove_proc_entry("segment_info", sbi
->s_proc
);
1331 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
1333 f2fs_destroy_stats(sbi
);
1338 mutex_lock(&sbi
->umount_mutex
);
1339 f2fs_leave_shrinker(sbi
);
1340 iput(sbi
->node_inode
);
1341 mutex_unlock(&sbi
->umount_mutex
);
1343 destroy_node_manager(sbi
);
1345 destroy_segment_manager(sbi
);
1349 make_bad_inode(sbi
->meta_inode
);
1350 iput(sbi
->meta_inode
);
1354 brelse(raw_super_buf
);
1358 /* give only one another chance */
1361 shrink_dcache_sb(sb
);
1367 static struct dentry
*f2fs_mount(struct file_system_type
*fs_type
, int flags
,
1368 const char *dev_name
, void *data
)
1370 return mount_bdev(fs_type
, flags
, dev_name
, data
, f2fs_fill_super
);
1373 static void kill_f2fs_super(struct super_block
*sb
)
1376 set_sbi_flag(F2FS_SB(sb
), SBI_IS_CLOSE
);
1377 kill_block_super(sb
);
1380 static struct file_system_type f2fs_fs_type
= {
1381 .owner
= THIS_MODULE
,
1383 .mount
= f2fs_mount
,
1384 .kill_sb
= kill_f2fs_super
,
1385 .fs_flags
= FS_REQUIRES_DEV
,
1387 MODULE_ALIAS_FS("f2fs");
1389 static int __init
init_inodecache(void)
1391 f2fs_inode_cachep
= f2fs_kmem_cache_create("f2fs_inode_cache",
1392 sizeof(struct f2fs_inode_info
));
1393 if (!f2fs_inode_cachep
)
1398 static void destroy_inodecache(void)
1401 * Make sure all delayed rcu free inodes are flushed before we
1405 kmem_cache_destroy(f2fs_inode_cachep
);
1408 static int __init
init_f2fs_fs(void)
1412 f2fs_build_trace_ios();
1414 err
= init_inodecache();
1417 err
= create_node_manager_caches();
1419 goto free_inodecache
;
1420 err
= create_segment_manager_caches();
1422 goto free_node_manager_caches
;
1423 err
= create_checkpoint_caches();
1425 goto free_segment_manager_caches
;
1426 err
= create_extent_cache();
1428 goto free_checkpoint_caches
;
1429 f2fs_kset
= kset_create_and_add("f2fs", NULL
, fs_kobj
);
1432 goto free_extent_cache
;
1434 err
= f2fs_init_crypto();
1438 err
= register_shrinker(&f2fs_shrinker_info
);
1442 err
= register_filesystem(&f2fs_fs_type
);
1445 f2fs_create_root_stats();
1446 f2fs_proc_root
= proc_mkdir("fs/f2fs", NULL
);
1450 unregister_shrinker(&f2fs_shrinker_info
);
1454 kset_unregister(f2fs_kset
);
1456 destroy_extent_cache();
1457 free_checkpoint_caches
:
1458 destroy_checkpoint_caches();
1459 free_segment_manager_caches
:
1460 destroy_segment_manager_caches();
1461 free_node_manager_caches
:
1462 destroy_node_manager_caches();
1464 destroy_inodecache();
1469 static void __exit
exit_f2fs_fs(void)
1471 remove_proc_entry("fs/f2fs", NULL
);
1472 f2fs_destroy_root_stats();
1473 unregister_shrinker(&f2fs_shrinker_info
);
1474 unregister_filesystem(&f2fs_fs_type
);
1476 destroy_extent_cache();
1477 destroy_checkpoint_caches();
1478 destroy_segment_manager_caches();
1479 destroy_node_manager_caches();
1480 destroy_inodecache();
1481 kset_unregister(f2fs_kset
);
1482 f2fs_destroy_trace_ios();
1485 module_init(init_f2fs_fs
)
1486 module_exit(exit_f2fs_fs
)
1488 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1489 MODULE_DESCRIPTION("Flash Friendly File System");
1490 MODULE_LICENSE("GPL");