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
,
72 static match_table_t f2fs_tokens
= {
73 {Opt_gc_background
, "background_gc=%s"},
74 {Opt_disable_roll_forward
, "disable_roll_forward"},
75 {Opt_norecovery
, "norecovery"},
76 {Opt_discard
, "discard"},
77 {Opt_noheap
, "no_heap"},
78 {Opt_user_xattr
, "user_xattr"},
79 {Opt_nouser_xattr
, "nouser_xattr"},
82 {Opt_active_logs
, "active_logs=%u"},
83 {Opt_disable_ext_identify
, "disable_ext_identify"},
84 {Opt_inline_xattr
, "inline_xattr"},
85 {Opt_inline_data
, "inline_data"},
86 {Opt_inline_dentry
, "inline_dentry"},
87 {Opt_flush_merge
, "flush_merge"},
88 {Opt_nobarrier
, "nobarrier"},
89 {Opt_fastboot
, "fastboot"},
90 {Opt_extent_cache
, "extent_cache"},
91 {Opt_noinline_data
, "noinline_data"},
95 /* Sysfs support for f2fs */
97 GC_THREAD
, /* struct f2fs_gc_thread */
98 SM_INFO
, /* struct f2fs_sm_info */
99 NM_INFO
, /* struct f2fs_nm_info */
100 F2FS_SBI
, /* struct f2fs_sb_info */
104 struct attribute attr
;
105 ssize_t (*show
)(struct f2fs_attr
*, struct f2fs_sb_info
*, char *);
106 ssize_t (*store
)(struct f2fs_attr
*, struct f2fs_sb_info
*,
107 const char *, size_t);
112 static unsigned char *__struct_ptr(struct f2fs_sb_info
*sbi
, int struct_type
)
114 if (struct_type
== GC_THREAD
)
115 return (unsigned char *)sbi
->gc_thread
;
116 else if (struct_type
== SM_INFO
)
117 return (unsigned char *)SM_I(sbi
);
118 else if (struct_type
== NM_INFO
)
119 return (unsigned char *)NM_I(sbi
);
120 else if (struct_type
== F2FS_SBI
)
121 return (unsigned char *)sbi
;
125 static ssize_t
f2fs_sbi_show(struct f2fs_attr
*a
,
126 struct f2fs_sb_info
*sbi
, char *buf
)
128 unsigned char *ptr
= NULL
;
131 ptr
= __struct_ptr(sbi
, a
->struct_type
);
135 ui
= (unsigned int *)(ptr
+ a
->offset
);
137 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
140 static ssize_t
f2fs_sbi_store(struct f2fs_attr
*a
,
141 struct f2fs_sb_info
*sbi
,
142 const char *buf
, size_t count
)
149 ptr
= __struct_ptr(sbi
, a
->struct_type
);
153 ui
= (unsigned int *)(ptr
+ a
->offset
);
155 ret
= kstrtoul(skip_spaces(buf
), 0, &t
);
162 static ssize_t
f2fs_attr_show(struct kobject
*kobj
,
163 struct attribute
*attr
, char *buf
)
165 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
167 struct f2fs_attr
*a
= container_of(attr
, struct f2fs_attr
, attr
);
169 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
172 static ssize_t
f2fs_attr_store(struct kobject
*kobj
, struct attribute
*attr
,
173 const char *buf
, size_t len
)
175 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
177 struct f2fs_attr
*a
= container_of(attr
, struct f2fs_attr
, attr
);
179 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
182 static void f2fs_sb_release(struct kobject
*kobj
)
184 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
186 complete(&sbi
->s_kobj_unregister
);
189 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
190 static struct f2fs_attr f2fs_attr_##_name = { \
191 .attr = {.name = __stringify(_name), .mode = _mode }, \
194 .struct_type = _struct_type, \
198 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
199 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
200 f2fs_sbi_show, f2fs_sbi_store, \
201 offsetof(struct struct_name, elname))
203 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_min_sleep_time
, min_sleep_time
);
204 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_max_sleep_time
, max_sleep_time
);
205 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_no_gc_sleep_time
, no_gc_sleep_time
);
206 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_idle
, gc_idle
);
207 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, reclaim_segments
, rec_prefree_segments
);
208 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, max_small_discards
, max_discards
);
209 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, batched_trim_sections
, trim_sections
);
210 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, ipu_policy
, ipu_policy
);
211 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, min_ipu_util
, min_ipu_util
);
212 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, min_fsync_blocks
, min_fsync_blocks
);
213 F2FS_RW_ATTR(NM_INFO
, f2fs_nm_info
, ram_thresh
, ram_thresh
);
214 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, max_victim_search
, max_victim_search
);
215 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, dir_level
, dir_level
);
217 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
218 static struct attribute
*f2fs_attrs
[] = {
219 ATTR_LIST(gc_min_sleep_time
),
220 ATTR_LIST(gc_max_sleep_time
),
221 ATTR_LIST(gc_no_gc_sleep_time
),
223 ATTR_LIST(reclaim_segments
),
224 ATTR_LIST(max_small_discards
),
225 ATTR_LIST(batched_trim_sections
),
226 ATTR_LIST(ipu_policy
),
227 ATTR_LIST(min_ipu_util
),
228 ATTR_LIST(min_fsync_blocks
),
229 ATTR_LIST(max_victim_search
),
230 ATTR_LIST(dir_level
),
231 ATTR_LIST(ram_thresh
),
235 static const struct sysfs_ops f2fs_attr_ops
= {
236 .show
= f2fs_attr_show
,
237 .store
= f2fs_attr_store
,
240 static struct kobj_type f2fs_ktype
= {
241 .default_attrs
= f2fs_attrs
,
242 .sysfs_ops
= &f2fs_attr_ops
,
243 .release
= f2fs_sb_release
,
246 void f2fs_msg(struct super_block
*sb
, const char *level
, const char *fmt
, ...)
248 struct va_format vaf
;
254 printk("%sF2FS-fs (%s): %pV\n", level
, sb
->s_id
, &vaf
);
258 static void init_once(void *foo
)
260 struct f2fs_inode_info
*fi
= (struct f2fs_inode_info
*) foo
;
262 inode_init_once(&fi
->vfs_inode
);
265 static int parse_options(struct super_block
*sb
, char *options
)
267 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
268 struct request_queue
*q
;
269 substring_t args
[MAX_OPT_ARGS
];
276 while ((p
= strsep(&options
, ",")) != NULL
) {
281 * Initialize args struct so we know whether arg was
282 * found; some options take optional arguments.
284 args
[0].to
= args
[0].from
= NULL
;
285 token
= match_token(p
, f2fs_tokens
, args
);
288 case Opt_gc_background
:
289 name
= match_strdup(&args
[0]);
293 if (strlen(name
) == 2 && !strncmp(name
, "on", 2))
295 else if (strlen(name
) == 3 && !strncmp(name
, "off", 3))
296 clear_opt(sbi
, BG_GC
);
303 case Opt_disable_roll_forward
:
304 set_opt(sbi
, DISABLE_ROLL_FORWARD
);
307 /* this option mounts f2fs with ro */
308 set_opt(sbi
, DISABLE_ROLL_FORWARD
);
309 if (!f2fs_readonly(sb
))
313 q
= bdev_get_queue(sb
->s_bdev
);
314 if (blk_queue_discard(q
)) {
315 set_opt(sbi
, DISCARD
);
317 f2fs_msg(sb
, KERN_WARNING
,
318 "mounting with \"discard\" option, but "
319 "the device does not support discard");
323 set_opt(sbi
, NOHEAP
);
325 #ifdef CONFIG_F2FS_FS_XATTR
327 set_opt(sbi
, XATTR_USER
);
329 case Opt_nouser_xattr
:
330 clear_opt(sbi
, XATTR_USER
);
332 case Opt_inline_xattr
:
333 set_opt(sbi
, INLINE_XATTR
);
337 f2fs_msg(sb
, KERN_INFO
,
338 "user_xattr options not supported");
340 case Opt_nouser_xattr
:
341 f2fs_msg(sb
, KERN_INFO
,
342 "nouser_xattr options not supported");
344 case Opt_inline_xattr
:
345 f2fs_msg(sb
, KERN_INFO
,
346 "inline_xattr options not supported");
349 #ifdef CONFIG_F2FS_FS_POSIX_ACL
351 set_opt(sbi
, POSIX_ACL
);
354 clear_opt(sbi
, POSIX_ACL
);
358 f2fs_msg(sb
, KERN_INFO
, "acl options not supported");
361 f2fs_msg(sb
, KERN_INFO
, "noacl options not supported");
364 case Opt_active_logs
:
365 if (args
->from
&& match_int(args
, &arg
))
367 if (arg
!= 2 && arg
!= 4 && arg
!= NR_CURSEG_TYPE
)
369 sbi
->active_logs
= arg
;
371 case Opt_disable_ext_identify
:
372 set_opt(sbi
, DISABLE_EXT_IDENTIFY
);
374 case Opt_inline_data
:
375 set_opt(sbi
, INLINE_DATA
);
377 case Opt_inline_dentry
:
378 set_opt(sbi
, INLINE_DENTRY
);
380 case Opt_flush_merge
:
381 set_opt(sbi
, FLUSH_MERGE
);
384 set_opt(sbi
, NOBARRIER
);
387 set_opt(sbi
, FASTBOOT
);
389 case Opt_extent_cache
:
390 set_opt(sbi
, EXTENT_CACHE
);
392 case Opt_noinline_data
:
393 clear_opt(sbi
, INLINE_DATA
);
396 f2fs_msg(sb
, KERN_ERR
,
397 "Unrecognized mount option \"%s\" or missing value",
405 static struct inode
*f2fs_alloc_inode(struct super_block
*sb
)
407 struct f2fs_inode_info
*fi
;
409 fi
= kmem_cache_alloc(f2fs_inode_cachep
, GFP_F2FS_ZERO
);
413 init_once((void *) fi
);
415 /* Initialize f2fs-specific inode info */
416 fi
->vfs_inode
.i_version
= 1;
417 atomic_set(&fi
->dirty_pages
, 0);
418 fi
->i_current_depth
= 1;
420 rwlock_init(&fi
->ext_lock
);
421 init_rwsem(&fi
->i_sem
);
422 INIT_RADIX_TREE(&fi
->inmem_root
, GFP_NOFS
);
423 INIT_LIST_HEAD(&fi
->inmem_pages
);
424 mutex_init(&fi
->inmem_lock
);
426 set_inode_flag(fi
, FI_NEW_INODE
);
428 if (test_opt(F2FS_SB(sb
), INLINE_XATTR
))
429 set_inode_flag(fi
, FI_INLINE_XATTR
);
431 /* Will be used by directory only */
432 fi
->i_dir_level
= F2FS_SB(sb
)->dir_level
;
434 #ifdef CONFIG_F2FS_FS_ENCRYPTION
435 fi
->i_crypt_info
= NULL
;
437 return &fi
->vfs_inode
;
440 static int f2fs_drop_inode(struct inode
*inode
)
443 * This is to avoid a deadlock condition like below.
444 * writeback_single_inode(inode)
445 * - f2fs_write_data_page
446 * - f2fs_gc -> iput -> evict
447 * - inode_wait_for_writeback(inode)
449 if (!inode_unhashed(inode
) && inode
->i_state
& I_SYNC
) {
450 if (!inode
->i_nlink
&& !is_bad_inode(inode
)) {
451 spin_unlock(&inode
->i_lock
);
453 /* some remained atomic pages should discarded */
454 if (f2fs_is_atomic_file(inode
))
455 commit_inmem_pages(inode
, true);
457 sb_start_intwrite(inode
->i_sb
);
458 i_size_write(inode
, 0);
460 if (F2FS_HAS_BLOCKS(inode
))
461 f2fs_truncate(inode
);
463 sb_end_intwrite(inode
->i_sb
);
465 #ifdef CONFIG_F2FS_FS_ENCRYPTION
466 if (F2FS_I(inode
)->i_crypt_info
)
467 f2fs_free_encryption_info(inode
,
468 F2FS_I(inode
)->i_crypt_info
);
470 spin_lock(&inode
->i_lock
);
474 return generic_drop_inode(inode
);
478 * f2fs_dirty_inode() is called from __mark_inode_dirty()
480 * We should call set_dirty_inode to write the dirty inode through write_inode.
482 static void f2fs_dirty_inode(struct inode
*inode
, int flags
)
484 set_inode_flag(F2FS_I(inode
), FI_DIRTY_INODE
);
487 static void f2fs_i_callback(struct rcu_head
*head
)
489 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
490 kmem_cache_free(f2fs_inode_cachep
, F2FS_I(inode
));
493 static void f2fs_destroy_inode(struct inode
*inode
)
495 call_rcu(&inode
->i_rcu
, f2fs_i_callback
);
498 static void f2fs_put_super(struct super_block
*sb
)
500 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
503 remove_proc_entry("segment_info", sbi
->s_proc
);
504 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
506 kobject_del(&sbi
->s_kobj
);
510 /* prevent remaining shrinker jobs */
511 mutex_lock(&sbi
->umount_mutex
);
514 * We don't need to do checkpoint when superblock is clean.
515 * But, the previous checkpoint was not done by umount, it needs to do
516 * clean checkpoint again.
518 if (is_sbi_flag_set(sbi
, SBI_IS_DIRTY
) ||
519 !is_set_ckpt_flags(F2FS_CKPT(sbi
), CP_UMOUNT_FLAG
)) {
520 struct cp_control cpc
= {
523 write_checkpoint(sbi
, &cpc
);
526 /* write_checkpoint can update stat informaion */
527 f2fs_destroy_stats(sbi
);
530 * normally superblock is clean, so we need to release this.
531 * In addition, EIO will skip do checkpoint, we need this as well.
533 release_dirty_inode(sbi
);
534 release_discard_addrs(sbi
);
536 f2fs_leave_shrinker(sbi
);
537 mutex_unlock(&sbi
->umount_mutex
);
539 iput(sbi
->node_inode
);
540 iput(sbi
->meta_inode
);
542 /* destroy f2fs internal modules */
543 destroy_node_manager(sbi
);
544 destroy_segment_manager(sbi
);
547 kobject_put(&sbi
->s_kobj
);
548 wait_for_completion(&sbi
->s_kobj_unregister
);
550 sb
->s_fs_info
= NULL
;
551 brelse(sbi
->raw_super_buf
);
555 int f2fs_sync_fs(struct super_block
*sb
, int sync
)
557 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
559 trace_f2fs_sync_fs(sb
, sync
);
562 struct cp_control cpc
;
564 cpc
.reason
= __get_cp_reason(sbi
);
566 mutex_lock(&sbi
->gc_mutex
);
567 write_checkpoint(sbi
, &cpc
);
568 mutex_unlock(&sbi
->gc_mutex
);
570 f2fs_balance_fs(sbi
);
572 f2fs_trace_ios(NULL
, 1);
577 static int f2fs_freeze(struct super_block
*sb
)
581 if (f2fs_readonly(sb
))
584 err
= f2fs_sync_fs(sb
, 1);
588 static int f2fs_unfreeze(struct super_block
*sb
)
593 static int f2fs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
595 struct super_block
*sb
= dentry
->d_sb
;
596 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
597 u64 id
= huge_encode_dev(sb
->s_bdev
->bd_dev
);
598 block_t total_count
, user_block_count
, start_count
, ovp_count
;
600 total_count
= le64_to_cpu(sbi
->raw_super
->block_count
);
601 user_block_count
= sbi
->user_block_count
;
602 start_count
= le32_to_cpu(sbi
->raw_super
->segment0_blkaddr
);
603 ovp_count
= SM_I(sbi
)->ovp_segments
<< sbi
->log_blocks_per_seg
;
604 buf
->f_type
= F2FS_SUPER_MAGIC
;
605 buf
->f_bsize
= sbi
->blocksize
;
607 buf
->f_blocks
= total_count
- start_count
;
608 buf
->f_bfree
= buf
->f_blocks
- valid_user_blocks(sbi
) - ovp_count
;
609 buf
->f_bavail
= user_block_count
- valid_user_blocks(sbi
);
611 buf
->f_files
= sbi
->total_node_count
- F2FS_RESERVED_NODE_NUM
;
612 buf
->f_ffree
= buf
->f_files
- valid_inode_count(sbi
);
614 buf
->f_namelen
= F2FS_NAME_LEN
;
615 buf
->f_fsid
.val
[0] = (u32
)id
;
616 buf
->f_fsid
.val
[1] = (u32
)(id
>> 32);
621 static int f2fs_show_options(struct seq_file
*seq
, struct dentry
*root
)
623 struct f2fs_sb_info
*sbi
= F2FS_SB(root
->d_sb
);
625 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, BG_GC
))
626 seq_printf(seq
, ",background_gc=%s", "on");
628 seq_printf(seq
, ",background_gc=%s", "off");
629 if (test_opt(sbi
, DISABLE_ROLL_FORWARD
))
630 seq_puts(seq
, ",disable_roll_forward");
631 if (test_opt(sbi
, DISCARD
))
632 seq_puts(seq
, ",discard");
633 if (test_opt(sbi
, NOHEAP
))
634 seq_puts(seq
, ",no_heap_alloc");
635 #ifdef CONFIG_F2FS_FS_XATTR
636 if (test_opt(sbi
, XATTR_USER
))
637 seq_puts(seq
, ",user_xattr");
639 seq_puts(seq
, ",nouser_xattr");
640 if (test_opt(sbi
, INLINE_XATTR
))
641 seq_puts(seq
, ",inline_xattr");
643 #ifdef CONFIG_F2FS_FS_POSIX_ACL
644 if (test_opt(sbi
, POSIX_ACL
))
645 seq_puts(seq
, ",acl");
647 seq_puts(seq
, ",noacl");
649 if (test_opt(sbi
, DISABLE_EXT_IDENTIFY
))
650 seq_puts(seq
, ",disable_ext_identify");
651 if (test_opt(sbi
, INLINE_DATA
))
652 seq_puts(seq
, ",inline_data");
654 seq_puts(seq
, ",noinline_data");
655 if (test_opt(sbi
, INLINE_DENTRY
))
656 seq_puts(seq
, ",inline_dentry");
657 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, FLUSH_MERGE
))
658 seq_puts(seq
, ",flush_merge");
659 if (test_opt(sbi
, NOBARRIER
))
660 seq_puts(seq
, ",nobarrier");
661 if (test_opt(sbi
, FASTBOOT
))
662 seq_puts(seq
, ",fastboot");
663 if (test_opt(sbi
, EXTENT_CACHE
))
664 seq_puts(seq
, ",extent_cache");
665 seq_printf(seq
, ",active_logs=%u", sbi
->active_logs
);
670 static int segment_info_seq_show(struct seq_file
*seq
, void *offset
)
672 struct super_block
*sb
= seq
->private;
673 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
674 unsigned int total_segs
=
675 le32_to_cpu(sbi
->raw_super
->segment_count_main
);
678 seq_puts(seq
, "format: segment_type|valid_blocks\n"
679 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
681 for (i
= 0; i
< total_segs
; i
++) {
682 struct seg_entry
*se
= get_seg_entry(sbi
, i
);
685 seq_printf(seq
, "%-5d", i
);
686 seq_printf(seq
, "%d|%-3u", se
->type
,
687 get_valid_blocks(sbi
, i
, 1));
688 if ((i
% 10) == 9 || i
== (total_segs
- 1))
697 static int segment_info_open_fs(struct inode
*inode
, struct file
*file
)
699 return single_open(file
, segment_info_seq_show
, PDE_DATA(inode
));
702 static const struct file_operations f2fs_seq_segment_info_fops
= {
703 .owner
= THIS_MODULE
,
704 .open
= segment_info_open_fs
,
707 .release
= single_release
,
710 static void default_options(struct f2fs_sb_info
*sbi
)
712 /* init some FS parameters */
713 sbi
->active_logs
= NR_CURSEG_TYPE
;
716 set_opt(sbi
, INLINE_DATA
);
718 #ifdef CONFIG_F2FS_FS_XATTR
719 set_opt(sbi
, XATTR_USER
);
721 #ifdef CONFIG_F2FS_FS_POSIX_ACL
722 set_opt(sbi
, POSIX_ACL
);
726 static int f2fs_remount(struct super_block
*sb
, int *flags
, char *data
)
728 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
729 struct f2fs_mount_info org_mount_opt
;
730 int err
, active_logs
;
731 bool need_restart_gc
= false;
732 bool need_stop_gc
= false;
737 * Save the old mount options in case we
738 * need to restore them.
740 org_mount_opt
= sbi
->mount_opt
;
741 active_logs
= sbi
->active_logs
;
743 sbi
->mount_opt
.opt
= 0;
744 default_options(sbi
);
746 /* parse mount options */
747 err
= parse_options(sb
, data
);
752 * Previous and new state of filesystem is RO,
753 * so skip checking GC and FLUSH_MERGE conditions.
755 if (f2fs_readonly(sb
) && (*flags
& MS_RDONLY
))
759 * We stop the GC thread if FS is mounted as RO
760 * or if background_gc = off is passed in mount
761 * option. Also sync the filesystem.
763 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, BG_GC
)) {
764 if (sbi
->gc_thread
) {
767 need_restart_gc
= true;
769 } else if (!sbi
->gc_thread
) {
770 err
= start_gc_thread(sbi
);
777 * We stop issue flush thread if FS is mounted as RO
778 * or if flush_merge is not passed in mount option.
780 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, FLUSH_MERGE
)) {
781 destroy_flush_cmd_control(sbi
);
782 } else if (!SM_I(sbi
)->cmd_control_info
) {
783 err
= create_flush_cmd_control(sbi
);
788 /* Update the POSIXACL Flag */
789 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
790 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
793 if (need_restart_gc
) {
794 if (start_gc_thread(sbi
))
795 f2fs_msg(sbi
->sb
, KERN_WARNING
,
796 "background gc thread has stopped");
797 } else if (need_stop_gc
) {
801 sbi
->mount_opt
= org_mount_opt
;
802 sbi
->active_logs
= active_logs
;
806 static struct super_operations f2fs_sops
= {
807 .alloc_inode
= f2fs_alloc_inode
,
808 .drop_inode
= f2fs_drop_inode
,
809 .destroy_inode
= f2fs_destroy_inode
,
810 .write_inode
= f2fs_write_inode
,
811 .dirty_inode
= f2fs_dirty_inode
,
812 .show_options
= f2fs_show_options
,
813 .evict_inode
= f2fs_evict_inode
,
814 .put_super
= f2fs_put_super
,
815 .sync_fs
= f2fs_sync_fs
,
816 .freeze_fs
= f2fs_freeze
,
817 .unfreeze_fs
= f2fs_unfreeze
,
818 .statfs
= f2fs_statfs
,
819 .remount_fs
= f2fs_remount
,
822 static struct inode
*f2fs_nfs_get_inode(struct super_block
*sb
,
823 u64 ino
, u32 generation
)
825 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
828 if (check_nid_range(sbi
, ino
))
829 return ERR_PTR(-ESTALE
);
832 * f2fs_iget isn't quite right if the inode is currently unallocated!
833 * However f2fs_iget currently does appropriate checks to handle stale
834 * inodes so everything is OK.
836 inode
= f2fs_iget(sb
, ino
);
838 return ERR_CAST(inode
);
839 if (unlikely(generation
&& inode
->i_generation
!= generation
)) {
840 /* we didn't find the right inode.. */
842 return ERR_PTR(-ESTALE
);
847 static struct dentry
*f2fs_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
848 int fh_len
, int fh_type
)
850 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
854 static struct dentry
*f2fs_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
855 int fh_len
, int fh_type
)
857 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
861 static const struct export_operations f2fs_export_ops
= {
862 .fh_to_dentry
= f2fs_fh_to_dentry
,
863 .fh_to_parent
= f2fs_fh_to_parent
,
864 .get_parent
= f2fs_get_parent
,
867 static loff_t
max_file_size(unsigned bits
)
869 loff_t result
= (DEF_ADDRS_PER_INODE
- F2FS_INLINE_XATTR_ADDRS
);
870 loff_t leaf_count
= ADDRS_PER_BLOCK
;
872 /* two direct node blocks */
873 result
+= (leaf_count
* 2);
875 /* two indirect node blocks */
876 leaf_count
*= NIDS_PER_BLOCK
;
877 result
+= (leaf_count
* 2);
879 /* one double indirect node block */
880 leaf_count
*= NIDS_PER_BLOCK
;
881 result
+= leaf_count
;
887 static int sanity_check_raw_super(struct super_block
*sb
,
888 struct f2fs_super_block
*raw_super
)
890 unsigned int blocksize
;
892 if (F2FS_SUPER_MAGIC
!= le32_to_cpu(raw_super
->magic
)) {
893 f2fs_msg(sb
, KERN_INFO
,
894 "Magic Mismatch, valid(0x%x) - read(0x%x)",
895 F2FS_SUPER_MAGIC
, le32_to_cpu(raw_super
->magic
));
899 /* Currently, support only 4KB page cache size */
900 if (F2FS_BLKSIZE
!= PAGE_CACHE_SIZE
) {
901 f2fs_msg(sb
, KERN_INFO
,
902 "Invalid page_cache_size (%lu), supports only 4KB\n",
907 /* Currently, support only 4KB block size */
908 blocksize
= 1 << le32_to_cpu(raw_super
->log_blocksize
);
909 if (blocksize
!= F2FS_BLKSIZE
) {
910 f2fs_msg(sb
, KERN_INFO
,
911 "Invalid blocksize (%u), supports only 4KB\n",
916 /* Currently, support 512/1024/2048/4096 bytes sector size */
917 if (le32_to_cpu(raw_super
->log_sectorsize
) >
918 F2FS_MAX_LOG_SECTOR_SIZE
||
919 le32_to_cpu(raw_super
->log_sectorsize
) <
920 F2FS_MIN_LOG_SECTOR_SIZE
) {
921 f2fs_msg(sb
, KERN_INFO
, "Invalid log sectorsize (%u)",
922 le32_to_cpu(raw_super
->log_sectorsize
));
925 if (le32_to_cpu(raw_super
->log_sectors_per_block
) +
926 le32_to_cpu(raw_super
->log_sectorsize
) !=
927 F2FS_MAX_LOG_SECTOR_SIZE
) {
928 f2fs_msg(sb
, KERN_INFO
,
929 "Invalid log sectors per block(%u) log sectorsize(%u)",
930 le32_to_cpu(raw_super
->log_sectors_per_block
),
931 le32_to_cpu(raw_super
->log_sectorsize
));
937 static int sanity_check_ckpt(struct f2fs_sb_info
*sbi
)
939 unsigned int total
, fsmeta
;
940 struct f2fs_super_block
*raw_super
= F2FS_RAW_SUPER(sbi
);
941 struct f2fs_checkpoint
*ckpt
= F2FS_CKPT(sbi
);
943 total
= le32_to_cpu(raw_super
->segment_count
);
944 fsmeta
= le32_to_cpu(raw_super
->segment_count_ckpt
);
945 fsmeta
+= le32_to_cpu(raw_super
->segment_count_sit
);
946 fsmeta
+= le32_to_cpu(raw_super
->segment_count_nat
);
947 fsmeta
+= le32_to_cpu(ckpt
->rsvd_segment_count
);
948 fsmeta
+= le32_to_cpu(raw_super
->segment_count_ssa
);
950 if (unlikely(fsmeta
>= total
))
953 if (unlikely(f2fs_cp_error(sbi
))) {
954 f2fs_msg(sbi
->sb
, KERN_ERR
, "A bug case: need to run fsck");
960 static void init_sb_info(struct f2fs_sb_info
*sbi
)
962 struct f2fs_super_block
*raw_super
= sbi
->raw_super
;
965 sbi
->log_sectors_per_block
=
966 le32_to_cpu(raw_super
->log_sectors_per_block
);
967 sbi
->log_blocksize
= le32_to_cpu(raw_super
->log_blocksize
);
968 sbi
->blocksize
= 1 << sbi
->log_blocksize
;
969 sbi
->log_blocks_per_seg
= le32_to_cpu(raw_super
->log_blocks_per_seg
);
970 sbi
->blocks_per_seg
= 1 << sbi
->log_blocks_per_seg
;
971 sbi
->segs_per_sec
= le32_to_cpu(raw_super
->segs_per_sec
);
972 sbi
->secs_per_zone
= le32_to_cpu(raw_super
->secs_per_zone
);
973 sbi
->total_sections
= le32_to_cpu(raw_super
->section_count
);
974 sbi
->total_node_count
=
975 (le32_to_cpu(raw_super
->segment_count_nat
) / 2)
976 * sbi
->blocks_per_seg
* NAT_ENTRY_PER_BLOCK
;
977 sbi
->root_ino_num
= le32_to_cpu(raw_super
->root_ino
);
978 sbi
->node_ino_num
= le32_to_cpu(raw_super
->node_ino
);
979 sbi
->meta_ino_num
= le32_to_cpu(raw_super
->meta_ino
);
980 sbi
->cur_victim_sec
= NULL_SECNO
;
981 sbi
->max_victim_search
= DEF_MAX_VICTIM_SEARCH
;
983 for (i
= 0; i
< NR_COUNT_TYPE
; i
++)
984 atomic_set(&sbi
->nr_pages
[i
], 0);
986 sbi
->dir_level
= DEF_DIR_LEVEL
;
987 clear_sbi_flag(sbi
, SBI_NEED_FSCK
);
989 INIT_LIST_HEAD(&sbi
->s_list
);
990 mutex_init(&sbi
->umount_mutex
);
994 * Read f2fs raw super block.
995 * Because we have two copies of super block, so read the first one at first,
996 * if the first one is invalid, move to read the second one.
998 static int read_raw_super_block(struct super_block
*sb
,
999 struct f2fs_super_block
**raw_super
,
1000 struct buffer_head
**raw_super_buf
,
1004 struct buffer_head
*buffer
;
1005 struct f2fs_super_block
*super
;
1009 buffer
= sb_bread(sb
, block
);
1012 f2fs_msg(sb
, KERN_ERR
, "Unable to read %dth superblock",
1023 super
= (struct f2fs_super_block
*)
1024 ((char *)(buffer
)->b_data
+ F2FS_SUPER_OFFSET
);
1026 /* sanity checking of raw super */
1027 if (sanity_check_raw_super(sb
, super
)) {
1030 f2fs_msg(sb
, KERN_ERR
,
1031 "Can't find valid F2FS filesystem in %dth superblock",
1043 *raw_super_buf
= buffer
;
1046 /* already have a valid superblock */
1050 /* check the validity of the second superblock */
1057 /* No valid superblock */
1064 int f2fs_commit_super(struct f2fs_sb_info
*sbi
, bool recover
)
1066 struct buffer_head
*sbh
= sbi
->raw_super_buf
;
1067 sector_t block
= sbh
->b_blocknr
;
1070 /* write back-up superblock first */
1071 sbh
->b_blocknr
= block
? 0 : 1;
1072 mark_buffer_dirty(sbh
);
1073 err
= sync_dirty_buffer(sbh
);
1075 sbh
->b_blocknr
= block
;
1077 /* if we are in recovery path, skip writing valid superblock */
1081 /* write current valid superblock */
1082 mark_buffer_dirty(sbh
);
1083 err
= sync_dirty_buffer(sbh
);
1085 clear_buffer_write_io_error(sbh
);
1086 set_buffer_uptodate(sbh
);
1090 static int f2fs_fill_super(struct super_block
*sb
, void *data
, int silent
)
1092 struct f2fs_sb_info
*sbi
;
1093 struct f2fs_super_block
*raw_super
;
1094 struct buffer_head
*raw_super_buf
;
1097 bool retry
= true, need_fsck
= false;
1098 char *options
= NULL
;
1104 raw_super_buf
= NULL
;
1107 /* allocate memory for f2fs-specific super block info */
1108 sbi
= kzalloc(sizeof(struct f2fs_sb_info
), GFP_KERNEL
);
1112 /* set a block size */
1113 if (unlikely(!sb_set_blocksize(sb
, F2FS_BLKSIZE
))) {
1114 f2fs_msg(sb
, KERN_ERR
, "unable to set blocksize");
1118 err
= read_raw_super_block(sb
, &raw_super
, &raw_super_buf
, &recovery
);
1122 sb
->s_fs_info
= sbi
;
1123 default_options(sbi
);
1124 /* parse mount options */
1125 options
= kstrdup((const char *)data
, GFP_KERNEL
);
1126 if (data
&& !options
) {
1131 err
= parse_options(sb
, options
);
1135 sb
->s_maxbytes
= max_file_size(le32_to_cpu(raw_super
->log_blocksize
));
1136 sb
->s_max_links
= F2FS_LINK_MAX
;
1137 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
1139 sb
->s_op
= &f2fs_sops
;
1140 sb
->s_xattr
= f2fs_xattr_handlers
;
1141 sb
->s_export_op
= &f2fs_export_ops
;
1142 sb
->s_magic
= F2FS_SUPER_MAGIC
;
1143 sb
->s_time_gran
= 1;
1144 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
1145 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
1146 memcpy(sb
->s_uuid
, raw_super
->uuid
, sizeof(raw_super
->uuid
));
1148 /* init f2fs-specific super block info */
1150 sbi
->raw_super
= raw_super
;
1151 sbi
->raw_super_buf
= raw_super_buf
;
1152 mutex_init(&sbi
->gc_mutex
);
1153 mutex_init(&sbi
->writepages
);
1154 mutex_init(&sbi
->cp_mutex
);
1155 init_rwsem(&sbi
->node_write
);
1156 clear_sbi_flag(sbi
, SBI_POR_DOING
);
1157 spin_lock_init(&sbi
->stat_lock
);
1159 init_rwsem(&sbi
->read_io
.io_rwsem
);
1160 sbi
->read_io
.sbi
= sbi
;
1161 sbi
->read_io
.bio
= NULL
;
1162 for (i
= 0; i
< NR_PAGE_TYPE
; i
++) {
1163 init_rwsem(&sbi
->write_io
[i
].io_rwsem
);
1164 sbi
->write_io
[i
].sbi
= sbi
;
1165 sbi
->write_io
[i
].bio
= NULL
;
1168 init_rwsem(&sbi
->cp_rwsem
);
1169 init_waitqueue_head(&sbi
->cp_wait
);
1172 /* get an inode for meta space */
1173 sbi
->meta_inode
= f2fs_iget(sb
, F2FS_META_INO(sbi
));
1174 if (IS_ERR(sbi
->meta_inode
)) {
1175 f2fs_msg(sb
, KERN_ERR
, "Failed to read F2FS meta data inode");
1176 err
= PTR_ERR(sbi
->meta_inode
);
1180 err
= get_valid_checkpoint(sbi
);
1182 f2fs_msg(sb
, KERN_ERR
, "Failed to get valid F2FS checkpoint");
1183 goto free_meta_inode
;
1186 /* sanity checking of checkpoint */
1188 if (sanity_check_ckpt(sbi
)) {
1189 f2fs_msg(sb
, KERN_ERR
, "Invalid F2FS checkpoint");
1193 sbi
->total_valid_node_count
=
1194 le32_to_cpu(sbi
->ckpt
->valid_node_count
);
1195 sbi
->total_valid_inode_count
=
1196 le32_to_cpu(sbi
->ckpt
->valid_inode_count
);
1197 sbi
->user_block_count
= le64_to_cpu(sbi
->ckpt
->user_block_count
);
1198 sbi
->total_valid_block_count
=
1199 le64_to_cpu(sbi
->ckpt
->valid_block_count
);
1200 sbi
->last_valid_block_count
= sbi
->total_valid_block_count
;
1201 sbi
->alloc_valid_block_count
= 0;
1202 INIT_LIST_HEAD(&sbi
->dir_inode_list
);
1203 spin_lock_init(&sbi
->dir_inode_lock
);
1205 init_extent_cache_info(sbi
);
1207 init_ino_entry_info(sbi
);
1209 /* setup f2fs internal modules */
1210 err
= build_segment_manager(sbi
);
1212 f2fs_msg(sb
, KERN_ERR
,
1213 "Failed to initialize F2FS segment manager");
1216 err
= build_node_manager(sbi
);
1218 f2fs_msg(sb
, KERN_ERR
,
1219 "Failed to initialize F2FS node manager");
1223 build_gc_manager(sbi
);
1225 /* get an inode for node space */
1226 sbi
->node_inode
= f2fs_iget(sb
, F2FS_NODE_INO(sbi
));
1227 if (IS_ERR(sbi
->node_inode
)) {
1228 f2fs_msg(sb
, KERN_ERR
, "Failed to read node inode");
1229 err
= PTR_ERR(sbi
->node_inode
);
1233 f2fs_join_shrinker(sbi
);
1235 /* if there are nt orphan nodes free them */
1236 recover_orphan_inodes(sbi
);
1238 /* read root inode and dentry */
1239 root
= f2fs_iget(sb
, F2FS_ROOT_INO(sbi
));
1241 f2fs_msg(sb
, KERN_ERR
, "Failed to read root inode");
1242 err
= PTR_ERR(root
);
1243 goto free_node_inode
;
1245 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
1248 goto free_node_inode
;
1251 sb
->s_root
= d_make_root(root
); /* allocate root dentry */
1254 goto free_root_inode
;
1257 err
= f2fs_build_stats(sbi
);
1259 goto free_root_inode
;
1262 sbi
->s_proc
= proc_mkdir(sb
->s_id
, f2fs_proc_root
);
1265 proc_create_data("segment_info", S_IRUGO
, sbi
->s_proc
,
1266 &f2fs_seq_segment_info_fops
, sb
);
1268 sbi
->s_kobj
.kset
= f2fs_kset
;
1269 init_completion(&sbi
->s_kobj_unregister
);
1270 err
= kobject_init_and_add(&sbi
->s_kobj
, &f2fs_ktype
, NULL
,
1275 /* recover fsynced data */
1276 if (!test_opt(sbi
, DISABLE_ROLL_FORWARD
)) {
1278 * mount should be failed, when device has readonly mode, and
1279 * previous checkpoint was not done by clean system shutdown.
1281 if (bdev_read_only(sb
->s_bdev
) &&
1282 !is_set_ckpt_flags(sbi
->ckpt
, CP_UMOUNT_FLAG
)) {
1288 set_sbi_flag(sbi
, SBI_NEED_FSCK
);
1290 err
= recover_fsync_data(sbi
);
1293 f2fs_msg(sb
, KERN_ERR
,
1294 "Cannot recover all fsync data errno=%ld", err
);
1300 * If filesystem is not mounted as read-only then
1301 * do start the gc_thread.
1303 if (test_opt(sbi
, BG_GC
) && !f2fs_readonly(sb
)) {
1304 /* After POR, we can run background GC thread.*/
1305 err
= start_gc_thread(sbi
);
1311 /* recover broken superblock */
1312 if (recovery
&& !f2fs_readonly(sb
) && !bdev_read_only(sb
->s_bdev
)) {
1313 f2fs_msg(sb
, KERN_INFO
, "Recover invalid superblock");
1314 f2fs_commit_super(sbi
, true);
1320 kobject_del(&sbi
->s_kobj
);
1323 remove_proc_entry("segment_info", sbi
->s_proc
);
1324 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
1326 f2fs_destroy_stats(sbi
);
1331 mutex_lock(&sbi
->umount_mutex
);
1332 f2fs_leave_shrinker(sbi
);
1333 iput(sbi
->node_inode
);
1334 mutex_unlock(&sbi
->umount_mutex
);
1336 destroy_node_manager(sbi
);
1338 destroy_segment_manager(sbi
);
1342 make_bad_inode(sbi
->meta_inode
);
1343 iput(sbi
->meta_inode
);
1347 brelse(raw_super_buf
);
1351 /* give only one another chance */
1354 shrink_dcache_sb(sb
);
1360 static struct dentry
*f2fs_mount(struct file_system_type
*fs_type
, int flags
,
1361 const char *dev_name
, void *data
)
1363 return mount_bdev(fs_type
, flags
, dev_name
, data
, f2fs_fill_super
);
1366 static void kill_f2fs_super(struct super_block
*sb
)
1369 set_sbi_flag(F2FS_SB(sb
), SBI_IS_CLOSE
);
1370 kill_block_super(sb
);
1373 static struct file_system_type f2fs_fs_type
= {
1374 .owner
= THIS_MODULE
,
1376 .mount
= f2fs_mount
,
1377 .kill_sb
= kill_f2fs_super
,
1378 .fs_flags
= FS_REQUIRES_DEV
,
1380 MODULE_ALIAS_FS("f2fs");
1382 static int __init
init_inodecache(void)
1384 f2fs_inode_cachep
= f2fs_kmem_cache_create("f2fs_inode_cache",
1385 sizeof(struct f2fs_inode_info
));
1386 if (!f2fs_inode_cachep
)
1391 static void destroy_inodecache(void)
1394 * Make sure all delayed rcu free inodes are flushed before we
1398 kmem_cache_destroy(f2fs_inode_cachep
);
1401 static int __init
init_f2fs_fs(void)
1405 f2fs_build_trace_ios();
1407 err
= init_inodecache();
1410 err
= create_node_manager_caches();
1412 goto free_inodecache
;
1413 err
= create_segment_manager_caches();
1415 goto free_node_manager_caches
;
1416 err
= create_checkpoint_caches();
1418 goto free_segment_manager_caches
;
1419 err
= create_extent_cache();
1421 goto free_checkpoint_caches
;
1422 f2fs_kset
= kset_create_and_add("f2fs", NULL
, fs_kobj
);
1425 goto free_extent_cache
;
1427 err
= f2fs_init_crypto();
1431 err
= register_shrinker(&f2fs_shrinker_info
);
1435 err
= register_filesystem(&f2fs_fs_type
);
1438 f2fs_create_root_stats();
1439 f2fs_proc_root
= proc_mkdir("fs/f2fs", NULL
);
1443 unregister_shrinker(&f2fs_shrinker_info
);
1447 kset_unregister(f2fs_kset
);
1449 destroy_extent_cache();
1450 free_checkpoint_caches
:
1451 destroy_checkpoint_caches();
1452 free_segment_manager_caches
:
1453 destroy_segment_manager_caches();
1454 free_node_manager_caches
:
1455 destroy_node_manager_caches();
1457 destroy_inodecache();
1462 static void __exit
exit_f2fs_fs(void)
1464 remove_proc_entry("fs/f2fs", NULL
);
1465 f2fs_destroy_root_stats();
1466 unregister_shrinker(&f2fs_shrinker_info
);
1467 unregister_filesystem(&f2fs_fs_type
);
1469 destroy_extent_cache();
1470 destroy_checkpoint_caches();
1471 destroy_segment_manager_caches();
1472 destroy_node_manager_caches();
1473 destroy_inodecache();
1474 kset_unregister(f2fs_kset
);
1475 f2fs_destroy_trace_ios();
1478 module_init(init_f2fs_fs
)
1479 module_exit(exit_f2fs_fs
)
1481 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1482 MODULE_DESCRIPTION("Flash Friendly File System");
1483 MODULE_LICENSE("GPL");