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
;
44 Opt_disable_roll_forward
,
53 Opt_disable_ext_identify
,
65 static match_table_t f2fs_tokens
= {
66 {Opt_gc_background
, "background_gc=%s"},
67 {Opt_disable_roll_forward
, "disable_roll_forward"},
68 {Opt_norecovery
, "norecovery"},
69 {Opt_discard
, "discard"},
70 {Opt_noheap
, "no_heap"},
71 {Opt_user_xattr
, "user_xattr"},
72 {Opt_nouser_xattr
, "nouser_xattr"},
75 {Opt_active_logs
, "active_logs=%u"},
76 {Opt_disable_ext_identify
, "disable_ext_identify"},
77 {Opt_inline_xattr
, "inline_xattr"},
78 {Opt_inline_data
, "inline_data"},
79 {Opt_inline_dentry
, "inline_dentry"},
80 {Opt_flush_merge
, "flush_merge"},
81 {Opt_nobarrier
, "nobarrier"},
82 {Opt_fastboot
, "fastboot"},
83 {Opt_extent_cache
, "extent_cache"},
84 {Opt_noinline_data
, "noinline_data"},
88 /* Sysfs support for f2fs */
90 GC_THREAD
, /* struct f2fs_gc_thread */
91 SM_INFO
, /* struct f2fs_sm_info */
92 NM_INFO
, /* struct f2fs_nm_info */
93 F2FS_SBI
, /* struct f2fs_sb_info */
97 struct attribute attr
;
98 ssize_t (*show
)(struct f2fs_attr
*, struct f2fs_sb_info
*, char *);
99 ssize_t (*store
)(struct f2fs_attr
*, struct f2fs_sb_info
*,
100 const char *, size_t);
105 static unsigned char *__struct_ptr(struct f2fs_sb_info
*sbi
, int struct_type
)
107 if (struct_type
== GC_THREAD
)
108 return (unsigned char *)sbi
->gc_thread
;
109 else if (struct_type
== SM_INFO
)
110 return (unsigned char *)SM_I(sbi
);
111 else if (struct_type
== NM_INFO
)
112 return (unsigned char *)NM_I(sbi
);
113 else if (struct_type
== F2FS_SBI
)
114 return (unsigned char *)sbi
;
118 static ssize_t
f2fs_sbi_show(struct f2fs_attr
*a
,
119 struct f2fs_sb_info
*sbi
, char *buf
)
121 unsigned char *ptr
= NULL
;
124 ptr
= __struct_ptr(sbi
, a
->struct_type
);
128 ui
= (unsigned int *)(ptr
+ a
->offset
);
130 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
133 static ssize_t
f2fs_sbi_store(struct f2fs_attr
*a
,
134 struct f2fs_sb_info
*sbi
,
135 const char *buf
, size_t count
)
142 ptr
= __struct_ptr(sbi
, a
->struct_type
);
146 ui
= (unsigned int *)(ptr
+ a
->offset
);
148 ret
= kstrtoul(skip_spaces(buf
), 0, &t
);
155 static ssize_t
f2fs_attr_show(struct kobject
*kobj
,
156 struct attribute
*attr
, char *buf
)
158 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
160 struct f2fs_attr
*a
= container_of(attr
, struct f2fs_attr
, attr
);
162 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
165 static ssize_t
f2fs_attr_store(struct kobject
*kobj
, struct attribute
*attr
,
166 const char *buf
, size_t len
)
168 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
170 struct f2fs_attr
*a
= container_of(attr
, struct f2fs_attr
, attr
);
172 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
175 static void f2fs_sb_release(struct kobject
*kobj
)
177 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
179 complete(&sbi
->s_kobj_unregister
);
182 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
183 static struct f2fs_attr f2fs_attr_##_name = { \
184 .attr = {.name = __stringify(_name), .mode = _mode }, \
187 .struct_type = _struct_type, \
191 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
192 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
193 f2fs_sbi_show, f2fs_sbi_store, \
194 offsetof(struct struct_name, elname))
196 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_min_sleep_time
, min_sleep_time
);
197 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_max_sleep_time
, max_sleep_time
);
198 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_no_gc_sleep_time
, no_gc_sleep_time
);
199 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_idle
, gc_idle
);
200 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, reclaim_segments
, rec_prefree_segments
);
201 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, max_small_discards
, max_discards
);
202 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, batched_trim_sections
, trim_sections
);
203 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, ipu_policy
, ipu_policy
);
204 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, min_ipu_util
, min_ipu_util
);
205 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, min_fsync_blocks
, min_fsync_blocks
);
206 F2FS_RW_ATTR(NM_INFO
, f2fs_nm_info
, ram_thresh
, ram_thresh
);
207 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, max_victim_search
, max_victim_search
);
208 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, dir_level
, dir_level
);
210 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
211 static struct attribute
*f2fs_attrs
[] = {
212 ATTR_LIST(gc_min_sleep_time
),
213 ATTR_LIST(gc_max_sleep_time
),
214 ATTR_LIST(gc_no_gc_sleep_time
),
216 ATTR_LIST(reclaim_segments
),
217 ATTR_LIST(max_small_discards
),
218 ATTR_LIST(batched_trim_sections
),
219 ATTR_LIST(ipu_policy
),
220 ATTR_LIST(min_ipu_util
),
221 ATTR_LIST(min_fsync_blocks
),
222 ATTR_LIST(max_victim_search
),
223 ATTR_LIST(dir_level
),
224 ATTR_LIST(ram_thresh
),
228 static const struct sysfs_ops f2fs_attr_ops
= {
229 .show
= f2fs_attr_show
,
230 .store
= f2fs_attr_store
,
233 static struct kobj_type f2fs_ktype
= {
234 .default_attrs
= f2fs_attrs
,
235 .sysfs_ops
= &f2fs_attr_ops
,
236 .release
= f2fs_sb_release
,
239 void f2fs_msg(struct super_block
*sb
, const char *level
, const char *fmt
, ...)
241 struct va_format vaf
;
247 printk("%sF2FS-fs (%s): %pV\n", level
, sb
->s_id
, &vaf
);
251 static void init_once(void *foo
)
253 struct f2fs_inode_info
*fi
= (struct f2fs_inode_info
*) foo
;
255 inode_init_once(&fi
->vfs_inode
);
258 static int parse_options(struct super_block
*sb
, char *options
)
260 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
261 struct request_queue
*q
;
262 substring_t args
[MAX_OPT_ARGS
];
269 while ((p
= strsep(&options
, ",")) != NULL
) {
274 * Initialize args struct so we know whether arg was
275 * found; some options take optional arguments.
277 args
[0].to
= args
[0].from
= NULL
;
278 token
= match_token(p
, f2fs_tokens
, args
);
281 case Opt_gc_background
:
282 name
= match_strdup(&args
[0]);
286 if (strlen(name
) == 2 && !strncmp(name
, "on", 2))
288 else if (strlen(name
) == 3 && !strncmp(name
, "off", 3))
289 clear_opt(sbi
, BG_GC
);
296 case Opt_disable_roll_forward
:
297 set_opt(sbi
, DISABLE_ROLL_FORWARD
);
300 /* this option mounts f2fs with ro */
301 set_opt(sbi
, DISABLE_ROLL_FORWARD
);
302 if (!f2fs_readonly(sb
))
306 q
= bdev_get_queue(sb
->s_bdev
);
307 if (blk_queue_discard(q
)) {
308 set_opt(sbi
, DISCARD
);
310 f2fs_msg(sb
, KERN_WARNING
,
311 "mounting with \"discard\" option, but "
312 "the device does not support discard");
316 set_opt(sbi
, NOHEAP
);
318 #ifdef CONFIG_F2FS_FS_XATTR
320 set_opt(sbi
, XATTR_USER
);
322 case Opt_nouser_xattr
:
323 clear_opt(sbi
, XATTR_USER
);
325 case Opt_inline_xattr
:
326 set_opt(sbi
, INLINE_XATTR
);
330 f2fs_msg(sb
, KERN_INFO
,
331 "user_xattr options not supported");
333 case Opt_nouser_xattr
:
334 f2fs_msg(sb
, KERN_INFO
,
335 "nouser_xattr options not supported");
337 case Opt_inline_xattr
:
338 f2fs_msg(sb
, KERN_INFO
,
339 "inline_xattr options not supported");
342 #ifdef CONFIG_F2FS_FS_POSIX_ACL
344 set_opt(sbi
, POSIX_ACL
);
347 clear_opt(sbi
, POSIX_ACL
);
351 f2fs_msg(sb
, KERN_INFO
, "acl options not supported");
354 f2fs_msg(sb
, KERN_INFO
, "noacl options not supported");
357 case Opt_active_logs
:
358 if (args
->from
&& match_int(args
, &arg
))
360 if (arg
!= 2 && arg
!= 4 && arg
!= NR_CURSEG_TYPE
)
362 sbi
->active_logs
= arg
;
364 case Opt_disable_ext_identify
:
365 set_opt(sbi
, DISABLE_EXT_IDENTIFY
);
367 case Opt_inline_data
:
368 set_opt(sbi
, INLINE_DATA
);
370 case Opt_inline_dentry
:
371 set_opt(sbi
, INLINE_DENTRY
);
373 case Opt_flush_merge
:
374 set_opt(sbi
, FLUSH_MERGE
);
377 set_opt(sbi
, NOBARRIER
);
380 set_opt(sbi
, FASTBOOT
);
382 case Opt_extent_cache
:
383 set_opt(sbi
, EXTENT_CACHE
);
385 case Opt_noinline_data
:
386 clear_opt(sbi
, INLINE_DATA
);
389 f2fs_msg(sb
, KERN_ERR
,
390 "Unrecognized mount option \"%s\" or missing value",
398 static struct inode
*f2fs_alloc_inode(struct super_block
*sb
)
400 struct f2fs_inode_info
*fi
;
402 fi
= kmem_cache_alloc(f2fs_inode_cachep
, GFP_F2FS_ZERO
);
406 init_once((void *) fi
);
408 /* Initialize f2fs-specific inode info */
409 fi
->vfs_inode
.i_version
= 1;
410 atomic_set(&fi
->dirty_pages
, 0);
411 fi
->i_current_depth
= 1;
413 rwlock_init(&fi
->ext_lock
);
414 init_rwsem(&fi
->i_sem
);
415 INIT_RADIX_TREE(&fi
->inmem_root
, GFP_NOFS
);
416 INIT_LIST_HEAD(&fi
->inmem_pages
);
417 mutex_init(&fi
->inmem_lock
);
419 set_inode_flag(fi
, FI_NEW_INODE
);
421 if (test_opt(F2FS_SB(sb
), INLINE_XATTR
))
422 set_inode_flag(fi
, FI_INLINE_XATTR
);
424 /* Will be used by directory only */
425 fi
->i_dir_level
= F2FS_SB(sb
)->dir_level
;
427 #ifdef CONFIG_F2FS_FS_ENCRYPTION
428 fi
->i_crypt_info
= NULL
;
430 return &fi
->vfs_inode
;
433 static int f2fs_drop_inode(struct inode
*inode
)
436 * This is to avoid a deadlock condition like below.
437 * writeback_single_inode(inode)
438 * - f2fs_write_data_page
439 * - f2fs_gc -> iput -> evict
440 * - inode_wait_for_writeback(inode)
442 if (!inode_unhashed(inode
) && inode
->i_state
& I_SYNC
) {
443 if (!inode
->i_nlink
&& !is_bad_inode(inode
)) {
444 spin_unlock(&inode
->i_lock
);
446 /* some remained atomic pages should discarded */
447 if (f2fs_is_atomic_file(inode
))
448 commit_inmem_pages(inode
, true);
450 sb_start_intwrite(inode
->i_sb
);
451 i_size_write(inode
, 0);
453 if (F2FS_HAS_BLOCKS(inode
))
454 f2fs_truncate(inode
);
456 sb_end_intwrite(inode
->i_sb
);
458 #ifdef CONFIG_F2FS_FS_ENCRYPTION
459 if (F2FS_I(inode
)->i_crypt_info
)
460 f2fs_free_encryption_info(inode
,
461 F2FS_I(inode
)->i_crypt_info
);
463 spin_lock(&inode
->i_lock
);
467 return generic_drop_inode(inode
);
471 * f2fs_dirty_inode() is called from __mark_inode_dirty()
473 * We should call set_dirty_inode to write the dirty inode through write_inode.
475 static void f2fs_dirty_inode(struct inode
*inode
, int flags
)
477 set_inode_flag(F2FS_I(inode
), FI_DIRTY_INODE
);
480 static void f2fs_i_callback(struct rcu_head
*head
)
482 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
483 kmem_cache_free(f2fs_inode_cachep
, F2FS_I(inode
));
486 static void f2fs_destroy_inode(struct inode
*inode
)
488 call_rcu(&inode
->i_rcu
, f2fs_i_callback
);
491 static void f2fs_put_super(struct super_block
*sb
)
493 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
496 remove_proc_entry("segment_info", sbi
->s_proc
);
497 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
499 kobject_del(&sbi
->s_kobj
);
501 f2fs_destroy_stats(sbi
);
505 * We don't need to do checkpoint when superblock is clean.
506 * But, the previous checkpoint was not done by umount, it needs to do
507 * clean checkpoint again.
509 if (is_sbi_flag_set(sbi
, SBI_IS_DIRTY
) ||
510 !is_set_ckpt_flags(F2FS_CKPT(sbi
), CP_UMOUNT_FLAG
)) {
511 struct cp_control cpc
= {
514 write_checkpoint(sbi
, &cpc
);
518 * normally superblock is clean, so we need to release this.
519 * In addition, EIO will skip do checkpoint, we need this as well.
521 release_dirty_inode(sbi
);
522 release_discard_addrs(sbi
);
524 iput(sbi
->node_inode
);
525 iput(sbi
->meta_inode
);
527 /* destroy f2fs internal modules */
528 destroy_node_manager(sbi
);
529 destroy_segment_manager(sbi
);
532 kobject_put(&sbi
->s_kobj
);
533 wait_for_completion(&sbi
->s_kobj_unregister
);
535 sb
->s_fs_info
= NULL
;
536 brelse(sbi
->raw_super_buf
);
540 int f2fs_sync_fs(struct super_block
*sb
, int sync
)
542 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
544 trace_f2fs_sync_fs(sb
, sync
);
547 struct cp_control cpc
;
549 cpc
.reason
= __get_cp_reason(sbi
);
551 mutex_lock(&sbi
->gc_mutex
);
552 write_checkpoint(sbi
, &cpc
);
553 mutex_unlock(&sbi
->gc_mutex
);
555 f2fs_balance_fs(sbi
);
557 f2fs_trace_ios(NULL
, 1);
562 static int f2fs_freeze(struct super_block
*sb
)
566 if (f2fs_readonly(sb
))
569 err
= f2fs_sync_fs(sb
, 1);
573 static int f2fs_unfreeze(struct super_block
*sb
)
578 static int f2fs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
580 struct super_block
*sb
= dentry
->d_sb
;
581 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
582 u64 id
= huge_encode_dev(sb
->s_bdev
->bd_dev
);
583 block_t total_count
, user_block_count
, start_count
, ovp_count
;
585 total_count
= le64_to_cpu(sbi
->raw_super
->block_count
);
586 user_block_count
= sbi
->user_block_count
;
587 start_count
= le32_to_cpu(sbi
->raw_super
->segment0_blkaddr
);
588 ovp_count
= SM_I(sbi
)->ovp_segments
<< sbi
->log_blocks_per_seg
;
589 buf
->f_type
= F2FS_SUPER_MAGIC
;
590 buf
->f_bsize
= sbi
->blocksize
;
592 buf
->f_blocks
= total_count
- start_count
;
593 buf
->f_bfree
= buf
->f_blocks
- valid_user_blocks(sbi
) - ovp_count
;
594 buf
->f_bavail
= user_block_count
- valid_user_blocks(sbi
);
596 buf
->f_files
= sbi
->total_node_count
- F2FS_RESERVED_NODE_NUM
;
597 buf
->f_ffree
= buf
->f_files
- valid_inode_count(sbi
);
599 buf
->f_namelen
= F2FS_NAME_LEN
;
600 buf
->f_fsid
.val
[0] = (u32
)id
;
601 buf
->f_fsid
.val
[1] = (u32
)(id
>> 32);
606 static int f2fs_show_options(struct seq_file
*seq
, struct dentry
*root
)
608 struct f2fs_sb_info
*sbi
= F2FS_SB(root
->d_sb
);
610 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, BG_GC
))
611 seq_printf(seq
, ",background_gc=%s", "on");
613 seq_printf(seq
, ",background_gc=%s", "off");
614 if (test_opt(sbi
, DISABLE_ROLL_FORWARD
))
615 seq_puts(seq
, ",disable_roll_forward");
616 if (test_opt(sbi
, DISCARD
))
617 seq_puts(seq
, ",discard");
618 if (test_opt(sbi
, NOHEAP
))
619 seq_puts(seq
, ",no_heap_alloc");
620 #ifdef CONFIG_F2FS_FS_XATTR
621 if (test_opt(sbi
, XATTR_USER
))
622 seq_puts(seq
, ",user_xattr");
624 seq_puts(seq
, ",nouser_xattr");
625 if (test_opt(sbi
, INLINE_XATTR
))
626 seq_puts(seq
, ",inline_xattr");
628 #ifdef CONFIG_F2FS_FS_POSIX_ACL
629 if (test_opt(sbi
, POSIX_ACL
))
630 seq_puts(seq
, ",acl");
632 seq_puts(seq
, ",noacl");
634 if (test_opt(sbi
, DISABLE_EXT_IDENTIFY
))
635 seq_puts(seq
, ",disable_ext_identify");
636 if (test_opt(sbi
, INLINE_DATA
))
637 seq_puts(seq
, ",inline_data");
639 seq_puts(seq
, ",noinline_data");
640 if (test_opt(sbi
, INLINE_DENTRY
))
641 seq_puts(seq
, ",inline_dentry");
642 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, FLUSH_MERGE
))
643 seq_puts(seq
, ",flush_merge");
644 if (test_opt(sbi
, NOBARRIER
))
645 seq_puts(seq
, ",nobarrier");
646 if (test_opt(sbi
, FASTBOOT
))
647 seq_puts(seq
, ",fastboot");
648 if (test_opt(sbi
, EXTENT_CACHE
))
649 seq_puts(seq
, ",extent_cache");
650 seq_printf(seq
, ",active_logs=%u", sbi
->active_logs
);
655 static int segment_info_seq_show(struct seq_file
*seq
, void *offset
)
657 struct super_block
*sb
= seq
->private;
658 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
659 unsigned int total_segs
=
660 le32_to_cpu(sbi
->raw_super
->segment_count_main
);
663 seq_puts(seq
, "format: segment_type|valid_blocks\n"
664 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
666 for (i
= 0; i
< total_segs
; i
++) {
667 struct seg_entry
*se
= get_seg_entry(sbi
, i
);
670 seq_printf(seq
, "%-5d", i
);
671 seq_printf(seq
, "%d|%-3u", se
->type
,
672 get_valid_blocks(sbi
, i
, 1));
673 if ((i
% 10) == 9 || i
== (total_segs
- 1))
682 static int segment_info_open_fs(struct inode
*inode
, struct file
*file
)
684 return single_open(file
, segment_info_seq_show
, PDE_DATA(inode
));
687 static const struct file_operations f2fs_seq_segment_info_fops
= {
688 .owner
= THIS_MODULE
,
689 .open
= segment_info_open_fs
,
692 .release
= single_release
,
695 static void default_options(struct f2fs_sb_info
*sbi
)
697 /* init some FS parameters */
698 sbi
->active_logs
= NR_CURSEG_TYPE
;
701 set_opt(sbi
, INLINE_DATA
);
703 #ifdef CONFIG_F2FS_FS_XATTR
704 set_opt(sbi
, XATTR_USER
);
706 #ifdef CONFIG_F2FS_FS_POSIX_ACL
707 set_opt(sbi
, POSIX_ACL
);
711 static int f2fs_remount(struct super_block
*sb
, int *flags
, char *data
)
713 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
714 struct f2fs_mount_info org_mount_opt
;
715 int err
, active_logs
;
716 bool need_restart_gc
= false;
717 bool need_stop_gc
= false;
722 * Save the old mount options in case we
723 * need to restore them.
725 org_mount_opt
= sbi
->mount_opt
;
726 active_logs
= sbi
->active_logs
;
728 sbi
->mount_opt
.opt
= 0;
729 default_options(sbi
);
731 /* parse mount options */
732 err
= parse_options(sb
, data
);
737 * Previous and new state of filesystem is RO,
738 * so skip checking GC and FLUSH_MERGE conditions.
740 if (f2fs_readonly(sb
) && (*flags
& MS_RDONLY
))
744 * We stop the GC thread if FS is mounted as RO
745 * or if background_gc = off is passed in mount
746 * option. Also sync the filesystem.
748 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, BG_GC
)) {
749 if (sbi
->gc_thread
) {
752 need_restart_gc
= true;
754 } else if (!sbi
->gc_thread
) {
755 err
= start_gc_thread(sbi
);
762 * We stop issue flush thread if FS is mounted as RO
763 * or if flush_merge is not passed in mount option.
765 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, FLUSH_MERGE
)) {
766 destroy_flush_cmd_control(sbi
);
767 } else if (!SM_I(sbi
)->cmd_control_info
) {
768 err
= create_flush_cmd_control(sbi
);
773 /* Update the POSIXACL Flag */
774 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
775 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
778 if (need_restart_gc
) {
779 if (start_gc_thread(sbi
))
780 f2fs_msg(sbi
->sb
, KERN_WARNING
,
781 "background gc thread has stopped");
782 } else if (need_stop_gc
) {
786 sbi
->mount_opt
= org_mount_opt
;
787 sbi
->active_logs
= active_logs
;
791 static struct super_operations f2fs_sops
= {
792 .alloc_inode
= f2fs_alloc_inode
,
793 .drop_inode
= f2fs_drop_inode
,
794 .destroy_inode
= f2fs_destroy_inode
,
795 .write_inode
= f2fs_write_inode
,
796 .dirty_inode
= f2fs_dirty_inode
,
797 .show_options
= f2fs_show_options
,
798 .evict_inode
= f2fs_evict_inode
,
799 .put_super
= f2fs_put_super
,
800 .sync_fs
= f2fs_sync_fs
,
801 .freeze_fs
= f2fs_freeze
,
802 .unfreeze_fs
= f2fs_unfreeze
,
803 .statfs
= f2fs_statfs
,
804 .remount_fs
= f2fs_remount
,
807 static struct inode
*f2fs_nfs_get_inode(struct super_block
*sb
,
808 u64 ino
, u32 generation
)
810 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
813 if (check_nid_range(sbi
, ino
))
814 return ERR_PTR(-ESTALE
);
817 * f2fs_iget isn't quite right if the inode is currently unallocated!
818 * However f2fs_iget currently does appropriate checks to handle stale
819 * inodes so everything is OK.
821 inode
= f2fs_iget(sb
, ino
);
823 return ERR_CAST(inode
);
824 if (unlikely(generation
&& inode
->i_generation
!= generation
)) {
825 /* we didn't find the right inode.. */
827 return ERR_PTR(-ESTALE
);
832 static struct dentry
*f2fs_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
833 int fh_len
, int fh_type
)
835 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
839 static struct dentry
*f2fs_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
840 int fh_len
, int fh_type
)
842 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
846 static const struct export_operations f2fs_export_ops
= {
847 .fh_to_dentry
= f2fs_fh_to_dentry
,
848 .fh_to_parent
= f2fs_fh_to_parent
,
849 .get_parent
= f2fs_get_parent
,
852 static loff_t
max_file_size(unsigned bits
)
854 loff_t result
= (DEF_ADDRS_PER_INODE
- F2FS_INLINE_XATTR_ADDRS
);
855 loff_t leaf_count
= ADDRS_PER_BLOCK
;
857 /* two direct node blocks */
858 result
+= (leaf_count
* 2);
860 /* two indirect node blocks */
861 leaf_count
*= NIDS_PER_BLOCK
;
862 result
+= (leaf_count
* 2);
864 /* one double indirect node block */
865 leaf_count
*= NIDS_PER_BLOCK
;
866 result
+= leaf_count
;
872 static int sanity_check_raw_super(struct super_block
*sb
,
873 struct f2fs_super_block
*raw_super
)
875 unsigned int blocksize
;
877 if (F2FS_SUPER_MAGIC
!= le32_to_cpu(raw_super
->magic
)) {
878 f2fs_msg(sb
, KERN_INFO
,
879 "Magic Mismatch, valid(0x%x) - read(0x%x)",
880 F2FS_SUPER_MAGIC
, le32_to_cpu(raw_super
->magic
));
884 /* Currently, support only 4KB page cache size */
885 if (F2FS_BLKSIZE
!= PAGE_CACHE_SIZE
) {
886 f2fs_msg(sb
, KERN_INFO
,
887 "Invalid page_cache_size (%lu), supports only 4KB\n",
892 /* Currently, support only 4KB block size */
893 blocksize
= 1 << le32_to_cpu(raw_super
->log_blocksize
);
894 if (blocksize
!= F2FS_BLKSIZE
) {
895 f2fs_msg(sb
, KERN_INFO
,
896 "Invalid blocksize (%u), supports only 4KB\n",
901 /* Currently, support 512/1024/2048/4096 bytes sector size */
902 if (le32_to_cpu(raw_super
->log_sectorsize
) >
903 F2FS_MAX_LOG_SECTOR_SIZE
||
904 le32_to_cpu(raw_super
->log_sectorsize
) <
905 F2FS_MIN_LOG_SECTOR_SIZE
) {
906 f2fs_msg(sb
, KERN_INFO
, "Invalid log sectorsize (%u)",
907 le32_to_cpu(raw_super
->log_sectorsize
));
910 if (le32_to_cpu(raw_super
->log_sectors_per_block
) +
911 le32_to_cpu(raw_super
->log_sectorsize
) !=
912 F2FS_MAX_LOG_SECTOR_SIZE
) {
913 f2fs_msg(sb
, KERN_INFO
,
914 "Invalid log sectors per block(%u) log sectorsize(%u)",
915 le32_to_cpu(raw_super
->log_sectors_per_block
),
916 le32_to_cpu(raw_super
->log_sectorsize
));
922 static int sanity_check_ckpt(struct f2fs_sb_info
*sbi
)
924 unsigned int total
, fsmeta
;
925 struct f2fs_super_block
*raw_super
= F2FS_RAW_SUPER(sbi
);
926 struct f2fs_checkpoint
*ckpt
= F2FS_CKPT(sbi
);
928 total
= le32_to_cpu(raw_super
->segment_count
);
929 fsmeta
= le32_to_cpu(raw_super
->segment_count_ckpt
);
930 fsmeta
+= le32_to_cpu(raw_super
->segment_count_sit
);
931 fsmeta
+= le32_to_cpu(raw_super
->segment_count_nat
);
932 fsmeta
+= le32_to_cpu(ckpt
->rsvd_segment_count
);
933 fsmeta
+= le32_to_cpu(raw_super
->segment_count_ssa
);
935 if (unlikely(fsmeta
>= total
))
938 if (unlikely(f2fs_cp_error(sbi
))) {
939 f2fs_msg(sbi
->sb
, KERN_ERR
, "A bug case: need to run fsck");
945 static void init_sb_info(struct f2fs_sb_info
*sbi
)
947 struct f2fs_super_block
*raw_super
= sbi
->raw_super
;
950 sbi
->log_sectors_per_block
=
951 le32_to_cpu(raw_super
->log_sectors_per_block
);
952 sbi
->log_blocksize
= le32_to_cpu(raw_super
->log_blocksize
);
953 sbi
->blocksize
= 1 << sbi
->log_blocksize
;
954 sbi
->log_blocks_per_seg
= le32_to_cpu(raw_super
->log_blocks_per_seg
);
955 sbi
->blocks_per_seg
= 1 << sbi
->log_blocks_per_seg
;
956 sbi
->segs_per_sec
= le32_to_cpu(raw_super
->segs_per_sec
);
957 sbi
->secs_per_zone
= le32_to_cpu(raw_super
->secs_per_zone
);
958 sbi
->total_sections
= le32_to_cpu(raw_super
->section_count
);
959 sbi
->total_node_count
=
960 (le32_to_cpu(raw_super
->segment_count_nat
) / 2)
961 * sbi
->blocks_per_seg
* NAT_ENTRY_PER_BLOCK
;
962 sbi
->root_ino_num
= le32_to_cpu(raw_super
->root_ino
);
963 sbi
->node_ino_num
= le32_to_cpu(raw_super
->node_ino
);
964 sbi
->meta_ino_num
= le32_to_cpu(raw_super
->meta_ino
);
965 sbi
->cur_victim_sec
= NULL_SECNO
;
966 sbi
->max_victim_search
= DEF_MAX_VICTIM_SEARCH
;
968 for (i
= 0; i
< NR_COUNT_TYPE
; i
++)
969 atomic_set(&sbi
->nr_pages
[i
], 0);
971 sbi
->dir_level
= DEF_DIR_LEVEL
;
972 clear_sbi_flag(sbi
, SBI_NEED_FSCK
);
976 * Read f2fs raw super block.
977 * Because we have two copies of super block, so read the first one at first,
978 * if the first one is invalid, move to read the second one.
980 static int read_raw_super_block(struct super_block
*sb
,
981 struct f2fs_super_block
**raw_super
,
982 struct buffer_head
**raw_super_buf
,
986 struct buffer_head
*buffer
;
987 struct f2fs_super_block
*super
;
991 buffer
= sb_bread(sb
, block
);
994 f2fs_msg(sb
, KERN_ERR
, "Unable to read %dth superblock",
1005 super
= (struct f2fs_super_block
*)
1006 ((char *)(buffer
)->b_data
+ F2FS_SUPER_OFFSET
);
1008 /* sanity checking of raw super */
1009 if (sanity_check_raw_super(sb
, super
)) {
1012 f2fs_msg(sb
, KERN_ERR
,
1013 "Can't find valid F2FS filesystem in %dth superblock",
1025 *raw_super_buf
= buffer
;
1028 /* already have a valid superblock */
1032 /* check the validity of the second superblock */
1039 /* No valid superblock */
1046 int f2fs_commit_super(struct f2fs_sb_info
*sbi
, bool recover
)
1048 struct buffer_head
*sbh
= sbi
->raw_super_buf
;
1049 sector_t block
= sbh
->b_blocknr
;
1052 /* write back-up superblock first */
1053 sbh
->b_blocknr
= block
? 0 : 1;
1054 mark_buffer_dirty(sbh
);
1055 err
= sync_dirty_buffer(sbh
);
1057 sbh
->b_blocknr
= block
;
1059 /* if we are in recovery path, skip writing valid superblock */
1063 /* write current valid superblock */
1064 mark_buffer_dirty(sbh
);
1065 err
= sync_dirty_buffer(sbh
);
1067 clear_buffer_write_io_error(sbh
);
1068 set_buffer_uptodate(sbh
);
1072 static int f2fs_fill_super(struct super_block
*sb
, void *data
, int silent
)
1074 struct f2fs_sb_info
*sbi
;
1075 struct f2fs_super_block
*raw_super
;
1076 struct buffer_head
*raw_super_buf
;
1079 bool retry
= true, need_fsck
= false;
1080 char *options
= NULL
;
1086 raw_super_buf
= NULL
;
1089 /* allocate memory for f2fs-specific super block info */
1090 sbi
= kzalloc(sizeof(struct f2fs_sb_info
), GFP_KERNEL
);
1094 /* set a block size */
1095 if (unlikely(!sb_set_blocksize(sb
, F2FS_BLKSIZE
))) {
1096 f2fs_msg(sb
, KERN_ERR
, "unable to set blocksize");
1100 err
= read_raw_super_block(sb
, &raw_super
, &raw_super_buf
, &recovery
);
1104 sb
->s_fs_info
= sbi
;
1105 default_options(sbi
);
1106 /* parse mount options */
1107 options
= kstrdup((const char *)data
, GFP_KERNEL
);
1108 if (data
&& !options
) {
1113 err
= parse_options(sb
, options
);
1117 sb
->s_maxbytes
= max_file_size(le32_to_cpu(raw_super
->log_blocksize
));
1118 sb
->s_max_links
= F2FS_LINK_MAX
;
1119 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
1121 sb
->s_op
= &f2fs_sops
;
1122 sb
->s_xattr
= f2fs_xattr_handlers
;
1123 sb
->s_export_op
= &f2fs_export_ops
;
1124 sb
->s_magic
= F2FS_SUPER_MAGIC
;
1125 sb
->s_time_gran
= 1;
1126 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
1127 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
1128 memcpy(sb
->s_uuid
, raw_super
->uuid
, sizeof(raw_super
->uuid
));
1130 /* init f2fs-specific super block info */
1132 sbi
->raw_super
= raw_super
;
1133 sbi
->raw_super_buf
= raw_super_buf
;
1134 mutex_init(&sbi
->gc_mutex
);
1135 mutex_init(&sbi
->writepages
);
1136 mutex_init(&sbi
->cp_mutex
);
1137 init_rwsem(&sbi
->node_write
);
1138 clear_sbi_flag(sbi
, SBI_POR_DOING
);
1139 spin_lock_init(&sbi
->stat_lock
);
1141 init_rwsem(&sbi
->read_io
.io_rwsem
);
1142 sbi
->read_io
.sbi
= sbi
;
1143 sbi
->read_io
.bio
= NULL
;
1144 for (i
= 0; i
< NR_PAGE_TYPE
; i
++) {
1145 init_rwsem(&sbi
->write_io
[i
].io_rwsem
);
1146 sbi
->write_io
[i
].sbi
= sbi
;
1147 sbi
->write_io
[i
].bio
= NULL
;
1150 init_rwsem(&sbi
->cp_rwsem
);
1151 init_waitqueue_head(&sbi
->cp_wait
);
1154 /* get an inode for meta space */
1155 sbi
->meta_inode
= f2fs_iget(sb
, F2FS_META_INO(sbi
));
1156 if (IS_ERR(sbi
->meta_inode
)) {
1157 f2fs_msg(sb
, KERN_ERR
, "Failed to read F2FS meta data inode");
1158 err
= PTR_ERR(sbi
->meta_inode
);
1162 err
= get_valid_checkpoint(sbi
);
1164 f2fs_msg(sb
, KERN_ERR
, "Failed to get valid F2FS checkpoint");
1165 goto free_meta_inode
;
1168 /* sanity checking of checkpoint */
1170 if (sanity_check_ckpt(sbi
)) {
1171 f2fs_msg(sb
, KERN_ERR
, "Invalid F2FS checkpoint");
1175 sbi
->total_valid_node_count
=
1176 le32_to_cpu(sbi
->ckpt
->valid_node_count
);
1177 sbi
->total_valid_inode_count
=
1178 le32_to_cpu(sbi
->ckpt
->valid_inode_count
);
1179 sbi
->user_block_count
= le64_to_cpu(sbi
->ckpt
->user_block_count
);
1180 sbi
->total_valid_block_count
=
1181 le64_to_cpu(sbi
->ckpt
->valid_block_count
);
1182 sbi
->last_valid_block_count
= sbi
->total_valid_block_count
;
1183 sbi
->alloc_valid_block_count
= 0;
1184 INIT_LIST_HEAD(&sbi
->dir_inode_list
);
1185 spin_lock_init(&sbi
->dir_inode_lock
);
1187 init_extent_cache_info(sbi
);
1189 init_ino_entry_info(sbi
);
1191 /* setup f2fs internal modules */
1192 err
= build_segment_manager(sbi
);
1194 f2fs_msg(sb
, KERN_ERR
,
1195 "Failed to initialize F2FS segment manager");
1198 err
= build_node_manager(sbi
);
1200 f2fs_msg(sb
, KERN_ERR
,
1201 "Failed to initialize F2FS node manager");
1205 build_gc_manager(sbi
);
1207 /* get an inode for node space */
1208 sbi
->node_inode
= f2fs_iget(sb
, F2FS_NODE_INO(sbi
));
1209 if (IS_ERR(sbi
->node_inode
)) {
1210 f2fs_msg(sb
, KERN_ERR
, "Failed to read node inode");
1211 err
= PTR_ERR(sbi
->node_inode
);
1215 /* if there are nt orphan nodes free them */
1216 recover_orphan_inodes(sbi
);
1218 /* read root inode and dentry */
1219 root
= f2fs_iget(sb
, F2FS_ROOT_INO(sbi
));
1221 f2fs_msg(sb
, KERN_ERR
, "Failed to read root inode");
1222 err
= PTR_ERR(root
);
1223 goto free_node_inode
;
1225 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
1228 goto free_node_inode
;
1231 sb
->s_root
= d_make_root(root
); /* allocate root dentry */
1234 goto free_root_inode
;
1237 err
= f2fs_build_stats(sbi
);
1239 goto free_root_inode
;
1242 sbi
->s_proc
= proc_mkdir(sb
->s_id
, f2fs_proc_root
);
1245 proc_create_data("segment_info", S_IRUGO
, sbi
->s_proc
,
1246 &f2fs_seq_segment_info_fops
, sb
);
1248 sbi
->s_kobj
.kset
= f2fs_kset
;
1249 init_completion(&sbi
->s_kobj_unregister
);
1250 err
= kobject_init_and_add(&sbi
->s_kobj
, &f2fs_ktype
, NULL
,
1255 /* recover fsynced data */
1256 if (!test_opt(sbi
, DISABLE_ROLL_FORWARD
)) {
1258 * mount should be failed, when device has readonly mode, and
1259 * previous checkpoint was not done by clean system shutdown.
1261 if (bdev_read_only(sb
->s_bdev
) &&
1262 !is_set_ckpt_flags(sbi
->ckpt
, CP_UMOUNT_FLAG
)) {
1268 set_sbi_flag(sbi
, SBI_NEED_FSCK
);
1270 err
= recover_fsync_data(sbi
);
1273 f2fs_msg(sb
, KERN_ERR
,
1274 "Cannot recover all fsync data errno=%ld", err
);
1280 * If filesystem is not mounted as read-only then
1281 * do start the gc_thread.
1283 if (test_opt(sbi
, BG_GC
) && !f2fs_readonly(sb
)) {
1284 /* After POR, we can run background GC thread.*/
1285 err
= start_gc_thread(sbi
);
1291 /* recover broken superblock */
1292 if (recovery
&& !f2fs_readonly(sb
) && !bdev_read_only(sb
->s_bdev
)) {
1293 f2fs_msg(sb
, KERN_INFO
, "Recover invalid superblock");
1294 f2fs_commit_super(sbi
, true);
1300 kobject_del(&sbi
->s_kobj
);
1303 remove_proc_entry("segment_info", sbi
->s_proc
);
1304 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
1306 f2fs_destroy_stats(sbi
);
1311 iput(sbi
->node_inode
);
1313 destroy_node_manager(sbi
);
1315 destroy_segment_manager(sbi
);
1319 make_bad_inode(sbi
->meta_inode
);
1320 iput(sbi
->meta_inode
);
1324 brelse(raw_super_buf
);
1328 /* give only one another chance */
1331 shrink_dcache_sb(sb
);
1337 static struct dentry
*f2fs_mount(struct file_system_type
*fs_type
, int flags
,
1338 const char *dev_name
, void *data
)
1340 return mount_bdev(fs_type
, flags
, dev_name
, data
, f2fs_fill_super
);
1343 static void kill_f2fs_super(struct super_block
*sb
)
1346 set_sbi_flag(F2FS_SB(sb
), SBI_IS_CLOSE
);
1347 kill_block_super(sb
);
1350 static struct file_system_type f2fs_fs_type
= {
1351 .owner
= THIS_MODULE
,
1353 .mount
= f2fs_mount
,
1354 .kill_sb
= kill_f2fs_super
,
1355 .fs_flags
= FS_REQUIRES_DEV
,
1357 MODULE_ALIAS_FS("f2fs");
1359 static int __init
init_inodecache(void)
1361 f2fs_inode_cachep
= f2fs_kmem_cache_create("f2fs_inode_cache",
1362 sizeof(struct f2fs_inode_info
));
1363 if (!f2fs_inode_cachep
)
1368 static void destroy_inodecache(void)
1371 * Make sure all delayed rcu free inodes are flushed before we
1375 kmem_cache_destroy(f2fs_inode_cachep
);
1378 static int __init
init_f2fs_fs(void)
1382 f2fs_build_trace_ios();
1384 err
= init_inodecache();
1387 err
= create_node_manager_caches();
1389 goto free_inodecache
;
1390 err
= create_segment_manager_caches();
1392 goto free_node_manager_caches
;
1393 err
= create_checkpoint_caches();
1395 goto free_segment_manager_caches
;
1396 err
= create_extent_cache();
1398 goto free_checkpoint_caches
;
1399 f2fs_kset
= kset_create_and_add("f2fs", NULL
, fs_kobj
);
1402 goto free_extent_cache
;
1404 err
= f2fs_init_crypto();
1407 err
= register_filesystem(&f2fs_fs_type
);
1410 f2fs_create_root_stats();
1411 f2fs_proc_root
= proc_mkdir("fs/f2fs", NULL
);
1417 kset_unregister(f2fs_kset
);
1419 destroy_extent_cache();
1420 free_checkpoint_caches
:
1421 destroy_checkpoint_caches();
1422 free_segment_manager_caches
:
1423 destroy_segment_manager_caches();
1424 free_node_manager_caches
:
1425 destroy_node_manager_caches();
1427 destroy_inodecache();
1432 static void __exit
exit_f2fs_fs(void)
1434 remove_proc_entry("fs/f2fs", NULL
);
1435 f2fs_destroy_root_stats();
1436 unregister_filesystem(&f2fs_fs_type
);
1438 destroy_extent_cache();
1439 destroy_checkpoint_caches();
1440 destroy_segment_manager_caches();
1441 destroy_node_manager_caches();
1442 destroy_inodecache();
1443 kset_unregister(f2fs_kset
);
1444 f2fs_destroy_trace_ios();
1447 module_init(init_f2fs_fs
)
1448 module_exit(exit_f2fs_fs
)
1450 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1451 MODULE_DESCRIPTION("Flash Friendly File System");
1452 MODULE_LICENSE("GPL");