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>
34 #define CREATE_TRACE_POINTS
35 #include <trace/events/f2fs.h>
37 static struct proc_dir_entry
*f2fs_proc_root
;
38 static struct kmem_cache
*f2fs_inode_cachep
;
39 static struct kset
*f2fs_kset
;
43 Opt_disable_roll_forward
,
51 Opt_disable_ext_identify
,
59 static match_table_t f2fs_tokens
= {
60 {Opt_gc_background
, "background_gc=%s"},
61 {Opt_disable_roll_forward
, "disable_roll_forward"},
62 {Opt_discard
, "discard"},
63 {Opt_noheap
, "no_heap"},
64 {Opt_user_xattr
, "user_xattr"},
65 {Opt_nouser_xattr
, "nouser_xattr"},
68 {Opt_active_logs
, "active_logs=%u"},
69 {Opt_disable_ext_identify
, "disable_ext_identify"},
70 {Opt_inline_xattr
, "inline_xattr"},
71 {Opt_inline_data
, "inline_data"},
72 {Opt_flush_merge
, "flush_merge"},
73 {Opt_nobarrier
, "nobarrier"},
77 /* Sysfs support for f2fs */
79 GC_THREAD
, /* struct f2fs_gc_thread */
80 SM_INFO
, /* struct f2fs_sm_info */
81 NM_INFO
, /* struct f2fs_nm_info */
82 F2FS_SBI
, /* struct f2fs_sb_info */
86 struct attribute attr
;
87 ssize_t (*show
)(struct f2fs_attr
*, struct f2fs_sb_info
*, char *);
88 ssize_t (*store
)(struct f2fs_attr
*, struct f2fs_sb_info
*,
89 const char *, size_t);
94 static unsigned char *__struct_ptr(struct f2fs_sb_info
*sbi
, int struct_type
)
96 if (struct_type
== GC_THREAD
)
97 return (unsigned char *)sbi
->gc_thread
;
98 else if (struct_type
== SM_INFO
)
99 return (unsigned char *)SM_I(sbi
);
100 else if (struct_type
== NM_INFO
)
101 return (unsigned char *)NM_I(sbi
);
102 else if (struct_type
== F2FS_SBI
)
103 return (unsigned char *)sbi
;
107 static ssize_t
f2fs_sbi_show(struct f2fs_attr
*a
,
108 struct f2fs_sb_info
*sbi
, char *buf
)
110 unsigned char *ptr
= NULL
;
113 ptr
= __struct_ptr(sbi
, a
->struct_type
);
117 ui
= (unsigned int *)(ptr
+ a
->offset
);
119 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
122 static ssize_t
f2fs_sbi_store(struct f2fs_attr
*a
,
123 struct f2fs_sb_info
*sbi
,
124 const char *buf
, size_t count
)
131 ptr
= __struct_ptr(sbi
, a
->struct_type
);
135 ui
= (unsigned int *)(ptr
+ a
->offset
);
137 ret
= kstrtoul(skip_spaces(buf
), 0, &t
);
144 static ssize_t
f2fs_attr_show(struct kobject
*kobj
,
145 struct attribute
*attr
, char *buf
)
147 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
149 struct f2fs_attr
*a
= container_of(attr
, struct f2fs_attr
, attr
);
151 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
154 static ssize_t
f2fs_attr_store(struct kobject
*kobj
, struct attribute
*attr
,
155 const char *buf
, size_t len
)
157 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
159 struct f2fs_attr
*a
= container_of(attr
, struct f2fs_attr
, attr
);
161 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
164 static void f2fs_sb_release(struct kobject
*kobj
)
166 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
168 complete(&sbi
->s_kobj_unregister
);
171 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
172 static struct f2fs_attr f2fs_attr_##_name = { \
173 .attr = {.name = __stringify(_name), .mode = _mode }, \
176 .struct_type = _struct_type, \
180 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
181 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
182 f2fs_sbi_show, f2fs_sbi_store, \
183 offsetof(struct struct_name, elname))
185 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_min_sleep_time
, min_sleep_time
);
186 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_max_sleep_time
, max_sleep_time
);
187 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_no_gc_sleep_time
, no_gc_sleep_time
);
188 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_idle
, gc_idle
);
189 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, reclaim_segments
, rec_prefree_segments
);
190 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, max_small_discards
, max_discards
);
191 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, ipu_policy
, ipu_policy
);
192 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, min_ipu_util
, min_ipu_util
);
193 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, min_fsync_blocks
, min_fsync_blocks
);
194 F2FS_RW_ATTR(NM_INFO
, f2fs_nm_info
, ram_thresh
, ram_thresh
);
195 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, max_victim_search
, max_victim_search
);
196 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, dir_level
, dir_level
);
198 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
199 static struct attribute
*f2fs_attrs
[] = {
200 ATTR_LIST(gc_min_sleep_time
),
201 ATTR_LIST(gc_max_sleep_time
),
202 ATTR_LIST(gc_no_gc_sleep_time
),
204 ATTR_LIST(reclaim_segments
),
205 ATTR_LIST(max_small_discards
),
206 ATTR_LIST(ipu_policy
),
207 ATTR_LIST(min_ipu_util
),
208 ATTR_LIST(min_fsync_blocks
),
209 ATTR_LIST(max_victim_search
),
210 ATTR_LIST(dir_level
),
211 ATTR_LIST(ram_thresh
),
215 static const struct sysfs_ops f2fs_attr_ops
= {
216 .show
= f2fs_attr_show
,
217 .store
= f2fs_attr_store
,
220 static struct kobj_type f2fs_ktype
= {
221 .default_attrs
= f2fs_attrs
,
222 .sysfs_ops
= &f2fs_attr_ops
,
223 .release
= f2fs_sb_release
,
226 void f2fs_msg(struct super_block
*sb
, const char *level
, const char *fmt
, ...)
228 struct va_format vaf
;
234 printk("%sF2FS-fs (%s): %pV\n", level
, sb
->s_id
, &vaf
);
238 static void init_once(void *foo
)
240 struct f2fs_inode_info
*fi
= (struct f2fs_inode_info
*) foo
;
242 inode_init_once(&fi
->vfs_inode
);
245 static int parse_options(struct super_block
*sb
, char *options
)
247 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
248 substring_t args
[MAX_OPT_ARGS
];
255 while ((p
= strsep(&options
, ",")) != NULL
) {
260 * Initialize args struct so we know whether arg was
261 * found; some options take optional arguments.
263 args
[0].to
= args
[0].from
= NULL
;
264 token
= match_token(p
, f2fs_tokens
, args
);
267 case Opt_gc_background
:
268 name
= match_strdup(&args
[0]);
272 if (strlen(name
) == 2 && !strncmp(name
, "on", 2))
274 else if (strlen(name
) == 3 && !strncmp(name
, "off", 3))
275 clear_opt(sbi
, BG_GC
);
282 case Opt_disable_roll_forward
:
283 set_opt(sbi
, DISABLE_ROLL_FORWARD
);
286 set_opt(sbi
, DISCARD
);
289 set_opt(sbi
, NOHEAP
);
291 #ifdef CONFIG_F2FS_FS_XATTR
293 set_opt(sbi
, XATTR_USER
);
295 case Opt_nouser_xattr
:
296 clear_opt(sbi
, XATTR_USER
);
298 case Opt_inline_xattr
:
299 set_opt(sbi
, INLINE_XATTR
);
303 f2fs_msg(sb
, KERN_INFO
,
304 "user_xattr options not supported");
306 case Opt_nouser_xattr
:
307 f2fs_msg(sb
, KERN_INFO
,
308 "nouser_xattr options not supported");
310 case Opt_inline_xattr
:
311 f2fs_msg(sb
, KERN_INFO
,
312 "inline_xattr options not supported");
315 #ifdef CONFIG_F2FS_FS_POSIX_ACL
317 set_opt(sbi
, POSIX_ACL
);
320 clear_opt(sbi
, POSIX_ACL
);
324 f2fs_msg(sb
, KERN_INFO
, "acl options not supported");
327 f2fs_msg(sb
, KERN_INFO
, "noacl options not supported");
330 case Opt_active_logs
:
331 if (args
->from
&& match_int(args
, &arg
))
333 if (arg
!= 2 && arg
!= 4 && arg
!= NR_CURSEG_TYPE
)
335 sbi
->active_logs
= arg
;
337 case Opt_disable_ext_identify
:
338 set_opt(sbi
, DISABLE_EXT_IDENTIFY
);
340 case Opt_inline_data
:
341 set_opt(sbi
, INLINE_DATA
);
343 case Opt_flush_merge
:
344 set_opt(sbi
, FLUSH_MERGE
);
347 set_opt(sbi
, NOBARRIER
);
350 f2fs_msg(sb
, KERN_ERR
,
351 "Unrecognized mount option \"%s\" or missing value",
359 static struct inode
*f2fs_alloc_inode(struct super_block
*sb
)
361 struct f2fs_inode_info
*fi
;
363 fi
= kmem_cache_alloc(f2fs_inode_cachep
, GFP_F2FS_ZERO
);
367 init_once((void *) fi
);
369 /* Initialize f2fs-specific inode info */
370 fi
->vfs_inode
.i_version
= 1;
371 atomic_set(&fi
->dirty_pages
, 0);
372 fi
->i_current_depth
= 1;
374 rwlock_init(&fi
->ext
.ext_lock
);
375 init_rwsem(&fi
->i_sem
);
377 set_inode_flag(fi
, FI_NEW_INODE
);
379 if (test_opt(F2FS_SB(sb
), INLINE_XATTR
))
380 set_inode_flag(fi
, FI_INLINE_XATTR
);
382 /* Will be used by directory only */
383 fi
->i_dir_level
= F2FS_SB(sb
)->dir_level
;
385 return &fi
->vfs_inode
;
388 static int f2fs_drop_inode(struct inode
*inode
)
391 * This is to avoid a deadlock condition like below.
392 * writeback_single_inode(inode)
393 * - f2fs_write_data_page
394 * - f2fs_gc -> iput -> evict
395 * - inode_wait_for_writeback(inode)
397 if (!inode_unhashed(inode
) && inode
->i_state
& I_SYNC
)
399 return generic_drop_inode(inode
);
403 * f2fs_dirty_inode() is called from __mark_inode_dirty()
405 * We should call set_dirty_inode to write the dirty inode through write_inode.
407 static void f2fs_dirty_inode(struct inode
*inode
, int flags
)
409 set_inode_flag(F2FS_I(inode
), FI_DIRTY_INODE
);
412 static void f2fs_i_callback(struct rcu_head
*head
)
414 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
415 kmem_cache_free(f2fs_inode_cachep
, F2FS_I(inode
));
418 static void f2fs_destroy_inode(struct inode
*inode
)
420 call_rcu(&inode
->i_rcu
, f2fs_i_callback
);
423 static void f2fs_put_super(struct super_block
*sb
)
425 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
428 remove_proc_entry("segment_info", sbi
->s_proc
);
429 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
431 kobject_del(&sbi
->s_kobj
);
433 f2fs_destroy_stats(sbi
);
436 /* We don't need to do checkpoint when it's clean */
438 struct cp_control cpc
= {
441 write_checkpoint(sbi
, &cpc
);
445 * normally superblock is clean, so we need to release this.
446 * In addition, EIO will skip do checkpoint, we need this as well.
448 release_dirty_inode(sbi
);
450 iput(sbi
->node_inode
);
451 iput(sbi
->meta_inode
);
453 /* destroy f2fs internal modules */
454 destroy_node_manager(sbi
);
455 destroy_segment_manager(sbi
);
458 kobject_put(&sbi
->s_kobj
);
459 wait_for_completion(&sbi
->s_kobj_unregister
);
461 sb
->s_fs_info
= NULL
;
462 brelse(sbi
->raw_super_buf
);
466 int f2fs_sync_fs(struct super_block
*sb
, int sync
)
468 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
470 trace_f2fs_sync_fs(sb
, sync
);
473 struct cp_control cpc
= {
476 mutex_lock(&sbi
->gc_mutex
);
477 write_checkpoint(sbi
, &cpc
);
478 mutex_unlock(&sbi
->gc_mutex
);
480 f2fs_balance_fs(sbi
);
486 static int f2fs_freeze(struct super_block
*sb
)
490 if (f2fs_readonly(sb
))
493 err
= f2fs_sync_fs(sb
, 1);
497 static int f2fs_unfreeze(struct super_block
*sb
)
502 static int f2fs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
504 struct super_block
*sb
= dentry
->d_sb
;
505 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
506 u64 id
= huge_encode_dev(sb
->s_bdev
->bd_dev
);
507 block_t total_count
, user_block_count
, start_count
, ovp_count
;
509 total_count
= le64_to_cpu(sbi
->raw_super
->block_count
);
510 user_block_count
= sbi
->user_block_count
;
511 start_count
= le32_to_cpu(sbi
->raw_super
->segment0_blkaddr
);
512 ovp_count
= SM_I(sbi
)->ovp_segments
<< sbi
->log_blocks_per_seg
;
513 buf
->f_type
= F2FS_SUPER_MAGIC
;
514 buf
->f_bsize
= sbi
->blocksize
;
516 buf
->f_blocks
= total_count
- start_count
;
517 buf
->f_bfree
= buf
->f_blocks
- valid_user_blocks(sbi
) - ovp_count
;
518 buf
->f_bavail
= user_block_count
- valid_user_blocks(sbi
);
520 buf
->f_files
= sbi
->total_node_count
- F2FS_RESERVED_NODE_NUM
;
521 buf
->f_ffree
= buf
->f_files
- valid_inode_count(sbi
);
523 buf
->f_namelen
= F2FS_NAME_LEN
;
524 buf
->f_fsid
.val
[0] = (u32
)id
;
525 buf
->f_fsid
.val
[1] = (u32
)(id
>> 32);
530 static int f2fs_show_options(struct seq_file
*seq
, struct dentry
*root
)
532 struct f2fs_sb_info
*sbi
= F2FS_SB(root
->d_sb
);
534 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, BG_GC
))
535 seq_printf(seq
, ",background_gc=%s", "on");
537 seq_printf(seq
, ",background_gc=%s", "off");
538 if (test_opt(sbi
, DISABLE_ROLL_FORWARD
))
539 seq_puts(seq
, ",disable_roll_forward");
540 if (test_opt(sbi
, DISCARD
))
541 seq_puts(seq
, ",discard");
542 if (test_opt(sbi
, NOHEAP
))
543 seq_puts(seq
, ",no_heap_alloc");
544 #ifdef CONFIG_F2FS_FS_XATTR
545 if (test_opt(sbi
, XATTR_USER
))
546 seq_puts(seq
, ",user_xattr");
548 seq_puts(seq
, ",nouser_xattr");
549 if (test_opt(sbi
, INLINE_XATTR
))
550 seq_puts(seq
, ",inline_xattr");
552 #ifdef CONFIG_F2FS_FS_POSIX_ACL
553 if (test_opt(sbi
, POSIX_ACL
))
554 seq_puts(seq
, ",acl");
556 seq_puts(seq
, ",noacl");
558 if (test_opt(sbi
, DISABLE_EXT_IDENTIFY
))
559 seq_puts(seq
, ",disable_ext_identify");
560 if (test_opt(sbi
, INLINE_DATA
))
561 seq_puts(seq
, ",inline_data");
562 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, FLUSH_MERGE
))
563 seq_puts(seq
, ",flush_merge");
564 if (test_opt(sbi
, NOBARRIER
))
565 seq_puts(seq
, ",nobarrier");
566 seq_printf(seq
, ",active_logs=%u", sbi
->active_logs
);
571 static int segment_info_seq_show(struct seq_file
*seq
, void *offset
)
573 struct super_block
*sb
= seq
->private;
574 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
575 unsigned int total_segs
=
576 le32_to_cpu(sbi
->raw_super
->segment_count_main
);
579 seq_puts(seq
, "format: segment_type|valid_blocks\n"
580 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
582 for (i
= 0; i
< total_segs
; i
++) {
583 struct seg_entry
*se
= get_seg_entry(sbi
, i
);
586 seq_printf(seq
, "%-5d", i
);
587 seq_printf(seq
, "%d|%-3u", se
->type
,
588 get_valid_blocks(sbi
, i
, 1));
589 if ((i
% 10) == 9 || i
== (total_segs
- 1))
598 static int segment_info_open_fs(struct inode
*inode
, struct file
*file
)
600 return single_open(file
, segment_info_seq_show
, PDE_DATA(inode
));
603 static const struct file_operations f2fs_seq_segment_info_fops
= {
604 .owner
= THIS_MODULE
,
605 .open
= segment_info_open_fs
,
608 .release
= single_release
,
611 static int f2fs_remount(struct super_block
*sb
, int *flags
, char *data
)
613 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
614 struct f2fs_mount_info org_mount_opt
;
615 int err
, active_logs
;
616 bool need_restart_gc
= false;
617 bool need_stop_gc
= false;
622 * Save the old mount options in case we
623 * need to restore them.
625 org_mount_opt
= sbi
->mount_opt
;
626 active_logs
= sbi
->active_logs
;
628 sbi
->mount_opt
.opt
= 0;
629 sbi
->active_logs
= NR_CURSEG_TYPE
;
631 /* parse mount options */
632 err
= parse_options(sb
, data
);
637 * Previous and new state of filesystem is RO,
638 * so skip checking GC and FLUSH_MERGE conditions.
640 if (f2fs_readonly(sb
) && (*flags
& MS_RDONLY
))
644 * We stop the GC thread if FS is mounted as RO
645 * or if background_gc = off is passed in mount
646 * option. Also sync the filesystem.
648 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, BG_GC
)) {
649 if (sbi
->gc_thread
) {
652 need_restart_gc
= true;
654 } else if (test_opt(sbi
, BG_GC
) && !sbi
->gc_thread
) {
655 err
= start_gc_thread(sbi
);
662 * We stop issue flush thread if FS is mounted as RO
663 * or if flush_merge is not passed in mount option.
665 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, FLUSH_MERGE
)) {
666 destroy_flush_cmd_control(sbi
);
667 } else if (test_opt(sbi
, FLUSH_MERGE
) && !SM_I(sbi
)->cmd_control_info
) {
668 err
= create_flush_cmd_control(sbi
);
673 /* Update the POSIXACL Flag */
674 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
675 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
678 if (need_restart_gc
) {
679 if (start_gc_thread(sbi
))
680 f2fs_msg(sbi
->sb
, KERN_WARNING
,
681 "background gc thread has stopped");
682 } else if (need_stop_gc
) {
686 sbi
->mount_opt
= org_mount_opt
;
687 sbi
->active_logs
= active_logs
;
691 static struct super_operations f2fs_sops
= {
692 .alloc_inode
= f2fs_alloc_inode
,
693 .drop_inode
= f2fs_drop_inode
,
694 .destroy_inode
= f2fs_destroy_inode
,
695 .write_inode
= f2fs_write_inode
,
696 .dirty_inode
= f2fs_dirty_inode
,
697 .show_options
= f2fs_show_options
,
698 .evict_inode
= f2fs_evict_inode
,
699 .put_super
= f2fs_put_super
,
700 .sync_fs
= f2fs_sync_fs
,
701 .freeze_fs
= f2fs_freeze
,
702 .unfreeze_fs
= f2fs_unfreeze
,
703 .statfs
= f2fs_statfs
,
704 .remount_fs
= f2fs_remount
,
707 static struct inode
*f2fs_nfs_get_inode(struct super_block
*sb
,
708 u64 ino
, u32 generation
)
710 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
713 if (check_nid_range(sbi
, ino
))
714 return ERR_PTR(-ESTALE
);
717 * f2fs_iget isn't quite right if the inode is currently unallocated!
718 * However f2fs_iget currently does appropriate checks to handle stale
719 * inodes so everything is OK.
721 inode
= f2fs_iget(sb
, ino
);
723 return ERR_CAST(inode
);
724 if (unlikely(generation
&& inode
->i_generation
!= generation
)) {
725 /* we didn't find the right inode.. */
727 return ERR_PTR(-ESTALE
);
732 static struct dentry
*f2fs_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
733 int fh_len
, int fh_type
)
735 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
739 static struct dentry
*f2fs_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
740 int fh_len
, int fh_type
)
742 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
746 static const struct export_operations f2fs_export_ops
= {
747 .fh_to_dentry
= f2fs_fh_to_dentry
,
748 .fh_to_parent
= f2fs_fh_to_parent
,
749 .get_parent
= f2fs_get_parent
,
752 static loff_t
max_file_size(unsigned bits
)
754 loff_t result
= (DEF_ADDRS_PER_INODE
- F2FS_INLINE_XATTR_ADDRS
);
755 loff_t leaf_count
= ADDRS_PER_BLOCK
;
757 /* two direct node blocks */
758 result
+= (leaf_count
* 2);
760 /* two indirect node blocks */
761 leaf_count
*= NIDS_PER_BLOCK
;
762 result
+= (leaf_count
* 2);
764 /* one double indirect node block */
765 leaf_count
*= NIDS_PER_BLOCK
;
766 result
+= leaf_count
;
772 static int sanity_check_raw_super(struct super_block
*sb
,
773 struct f2fs_super_block
*raw_super
)
775 unsigned int blocksize
;
777 if (F2FS_SUPER_MAGIC
!= le32_to_cpu(raw_super
->magic
)) {
778 f2fs_msg(sb
, KERN_INFO
,
779 "Magic Mismatch, valid(0x%x) - read(0x%x)",
780 F2FS_SUPER_MAGIC
, le32_to_cpu(raw_super
->magic
));
784 /* Currently, support only 4KB page cache size */
785 if (F2FS_BLKSIZE
!= PAGE_CACHE_SIZE
) {
786 f2fs_msg(sb
, KERN_INFO
,
787 "Invalid page_cache_size (%lu), supports only 4KB\n",
792 /* Currently, support only 4KB block size */
793 blocksize
= 1 << le32_to_cpu(raw_super
->log_blocksize
);
794 if (blocksize
!= F2FS_BLKSIZE
) {
795 f2fs_msg(sb
, KERN_INFO
,
796 "Invalid blocksize (%u), supports only 4KB\n",
801 /* Currently, support 512/1024/2048/4096 bytes sector size */
802 if (le32_to_cpu(raw_super
->log_sectorsize
) >
803 F2FS_MAX_LOG_SECTOR_SIZE
||
804 le32_to_cpu(raw_super
->log_sectorsize
) <
805 F2FS_MIN_LOG_SECTOR_SIZE
) {
806 f2fs_msg(sb
, KERN_INFO
, "Invalid log sectorsize (%u)",
807 le32_to_cpu(raw_super
->log_sectorsize
));
810 if (le32_to_cpu(raw_super
->log_sectors_per_block
) +
811 le32_to_cpu(raw_super
->log_sectorsize
) !=
812 F2FS_MAX_LOG_SECTOR_SIZE
) {
813 f2fs_msg(sb
, KERN_INFO
,
814 "Invalid log sectors per block(%u) log sectorsize(%u)",
815 le32_to_cpu(raw_super
->log_sectors_per_block
),
816 le32_to_cpu(raw_super
->log_sectorsize
));
822 static int sanity_check_ckpt(struct f2fs_sb_info
*sbi
)
824 unsigned int total
, fsmeta
;
825 struct f2fs_super_block
*raw_super
= F2FS_RAW_SUPER(sbi
);
826 struct f2fs_checkpoint
*ckpt
= F2FS_CKPT(sbi
);
828 total
= le32_to_cpu(raw_super
->segment_count
);
829 fsmeta
= le32_to_cpu(raw_super
->segment_count_ckpt
);
830 fsmeta
+= le32_to_cpu(raw_super
->segment_count_sit
);
831 fsmeta
+= le32_to_cpu(raw_super
->segment_count_nat
);
832 fsmeta
+= le32_to_cpu(ckpt
->rsvd_segment_count
);
833 fsmeta
+= le32_to_cpu(raw_super
->segment_count_ssa
);
835 if (unlikely(fsmeta
>= total
))
838 if (unlikely(f2fs_cp_error(sbi
))) {
839 f2fs_msg(sbi
->sb
, KERN_ERR
, "A bug case: need to run fsck");
845 static void init_sb_info(struct f2fs_sb_info
*sbi
)
847 struct f2fs_super_block
*raw_super
= sbi
->raw_super
;
850 sbi
->log_sectors_per_block
=
851 le32_to_cpu(raw_super
->log_sectors_per_block
);
852 sbi
->log_blocksize
= le32_to_cpu(raw_super
->log_blocksize
);
853 sbi
->blocksize
= 1 << sbi
->log_blocksize
;
854 sbi
->log_blocks_per_seg
= le32_to_cpu(raw_super
->log_blocks_per_seg
);
855 sbi
->blocks_per_seg
= 1 << sbi
->log_blocks_per_seg
;
856 sbi
->segs_per_sec
= le32_to_cpu(raw_super
->segs_per_sec
);
857 sbi
->secs_per_zone
= le32_to_cpu(raw_super
->secs_per_zone
);
858 sbi
->total_sections
= le32_to_cpu(raw_super
->section_count
);
859 sbi
->total_node_count
=
860 (le32_to_cpu(raw_super
->segment_count_nat
) / 2)
861 * sbi
->blocks_per_seg
* NAT_ENTRY_PER_BLOCK
;
862 sbi
->root_ino_num
= le32_to_cpu(raw_super
->root_ino
);
863 sbi
->node_ino_num
= le32_to_cpu(raw_super
->node_ino
);
864 sbi
->meta_ino_num
= le32_to_cpu(raw_super
->meta_ino
);
865 sbi
->cur_victim_sec
= NULL_SECNO
;
866 sbi
->max_victim_search
= DEF_MAX_VICTIM_SEARCH
;
868 for (i
= 0; i
< NR_COUNT_TYPE
; i
++)
869 atomic_set(&sbi
->nr_pages
[i
], 0);
871 sbi
->dir_level
= DEF_DIR_LEVEL
;
872 sbi
->need_fsck
= false;
876 * Read f2fs raw super block.
877 * Because we have two copies of super block, so read the first one at first,
878 * if the first one is invalid, move to read the second one.
880 static int read_raw_super_block(struct super_block
*sb
,
881 struct f2fs_super_block
**raw_super
,
882 struct buffer_head
**raw_super_buf
)
887 *raw_super_buf
= sb_bread(sb
, block
);
888 if (!*raw_super_buf
) {
889 f2fs_msg(sb
, KERN_ERR
, "Unable to read %dth superblock",
899 *raw_super
= (struct f2fs_super_block
*)
900 ((char *)(*raw_super_buf
)->b_data
+ F2FS_SUPER_OFFSET
);
902 /* sanity checking of raw super */
903 if (sanity_check_raw_super(sb
, *raw_super
)) {
904 brelse(*raw_super_buf
);
905 f2fs_msg(sb
, KERN_ERR
,
906 "Can't find valid F2FS filesystem in %dth superblock",
919 static int f2fs_fill_super(struct super_block
*sb
, void *data
, int silent
)
921 struct f2fs_sb_info
*sbi
;
922 struct f2fs_super_block
*raw_super
;
923 struct buffer_head
*raw_super_buf
;
930 /* allocate memory for f2fs-specific super block info */
931 sbi
= kzalloc(sizeof(struct f2fs_sb_info
), GFP_KERNEL
);
935 /* set a block size */
936 if (unlikely(!sb_set_blocksize(sb
, F2FS_BLKSIZE
))) {
937 f2fs_msg(sb
, KERN_ERR
, "unable to set blocksize");
941 err
= read_raw_super_block(sb
, &raw_super
, &raw_super_buf
);
946 /* init some FS parameters */
947 sbi
->active_logs
= NR_CURSEG_TYPE
;
951 #ifdef CONFIG_F2FS_FS_XATTR
952 set_opt(sbi
, XATTR_USER
);
954 #ifdef CONFIG_F2FS_FS_POSIX_ACL
955 set_opt(sbi
, POSIX_ACL
);
957 /* parse mount options */
958 err
= parse_options(sb
, (char *)data
);
962 sb
->s_maxbytes
= max_file_size(le32_to_cpu(raw_super
->log_blocksize
));
963 sb
->s_max_links
= F2FS_LINK_MAX
;
964 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
966 sb
->s_op
= &f2fs_sops
;
967 sb
->s_xattr
= f2fs_xattr_handlers
;
968 sb
->s_export_op
= &f2fs_export_ops
;
969 sb
->s_magic
= F2FS_SUPER_MAGIC
;
971 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
972 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
973 memcpy(sb
->s_uuid
, raw_super
->uuid
, sizeof(raw_super
->uuid
));
975 /* init f2fs-specific super block info */
977 sbi
->raw_super
= raw_super
;
978 sbi
->raw_super_buf
= raw_super_buf
;
979 mutex_init(&sbi
->gc_mutex
);
980 mutex_init(&sbi
->writepages
);
981 mutex_init(&sbi
->cp_mutex
);
982 init_rwsem(&sbi
->node_write
);
983 sbi
->por_doing
= false;
984 spin_lock_init(&sbi
->stat_lock
);
986 init_rwsem(&sbi
->read_io
.io_rwsem
);
987 sbi
->read_io
.sbi
= sbi
;
988 sbi
->read_io
.bio
= NULL
;
989 for (i
= 0; i
< NR_PAGE_TYPE
; i
++) {
990 init_rwsem(&sbi
->write_io
[i
].io_rwsem
);
991 sbi
->write_io
[i
].sbi
= sbi
;
992 sbi
->write_io
[i
].bio
= NULL
;
995 init_rwsem(&sbi
->cp_rwsem
);
996 init_waitqueue_head(&sbi
->cp_wait
);
999 /* get an inode for meta space */
1000 sbi
->meta_inode
= f2fs_iget(sb
, F2FS_META_INO(sbi
));
1001 if (IS_ERR(sbi
->meta_inode
)) {
1002 f2fs_msg(sb
, KERN_ERR
, "Failed to read F2FS meta data inode");
1003 err
= PTR_ERR(sbi
->meta_inode
);
1007 err
= get_valid_checkpoint(sbi
);
1009 f2fs_msg(sb
, KERN_ERR
, "Failed to get valid F2FS checkpoint");
1010 goto free_meta_inode
;
1013 /* sanity checking of checkpoint */
1015 if (sanity_check_ckpt(sbi
)) {
1016 f2fs_msg(sb
, KERN_ERR
, "Invalid F2FS checkpoint");
1020 sbi
->total_valid_node_count
=
1021 le32_to_cpu(sbi
->ckpt
->valid_node_count
);
1022 sbi
->total_valid_inode_count
=
1023 le32_to_cpu(sbi
->ckpt
->valid_inode_count
);
1024 sbi
->user_block_count
= le64_to_cpu(sbi
->ckpt
->user_block_count
);
1025 sbi
->total_valid_block_count
=
1026 le64_to_cpu(sbi
->ckpt
->valid_block_count
);
1027 sbi
->last_valid_block_count
= sbi
->total_valid_block_count
;
1028 sbi
->alloc_valid_block_count
= 0;
1029 INIT_LIST_HEAD(&sbi
->dir_inode_list
);
1030 spin_lock_init(&sbi
->dir_inode_lock
);
1032 init_ino_entry_info(sbi
);
1034 /* setup f2fs internal modules */
1035 err
= build_segment_manager(sbi
);
1037 f2fs_msg(sb
, KERN_ERR
,
1038 "Failed to initialize F2FS segment manager");
1041 err
= build_node_manager(sbi
);
1043 f2fs_msg(sb
, KERN_ERR
,
1044 "Failed to initialize F2FS node manager");
1048 build_gc_manager(sbi
);
1050 /* get an inode for node space */
1051 sbi
->node_inode
= f2fs_iget(sb
, F2FS_NODE_INO(sbi
));
1052 if (IS_ERR(sbi
->node_inode
)) {
1053 f2fs_msg(sb
, KERN_ERR
, "Failed to read node inode");
1054 err
= PTR_ERR(sbi
->node_inode
);
1058 /* if there are nt orphan nodes free them */
1059 recover_orphan_inodes(sbi
);
1061 /* read root inode and dentry */
1062 root
= f2fs_iget(sb
, F2FS_ROOT_INO(sbi
));
1064 f2fs_msg(sb
, KERN_ERR
, "Failed to read root inode");
1065 err
= PTR_ERR(root
);
1066 goto free_node_inode
;
1068 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
1071 goto free_node_inode
;
1074 sb
->s_root
= d_make_root(root
); /* allocate root dentry */
1077 goto free_root_inode
;
1080 err
= f2fs_build_stats(sbi
);
1082 goto free_root_inode
;
1085 sbi
->s_proc
= proc_mkdir(sb
->s_id
, f2fs_proc_root
);
1088 proc_create_data("segment_info", S_IRUGO
, sbi
->s_proc
,
1089 &f2fs_seq_segment_info_fops
, sb
);
1091 if (test_opt(sbi
, DISCARD
)) {
1092 struct request_queue
*q
= bdev_get_queue(sb
->s_bdev
);
1093 if (!blk_queue_discard(q
))
1094 f2fs_msg(sb
, KERN_WARNING
,
1095 "mounting with \"discard\" option, but "
1096 "the device does not support discard");
1099 sbi
->s_kobj
.kset
= f2fs_kset
;
1100 init_completion(&sbi
->s_kobj_unregister
);
1101 err
= kobject_init_and_add(&sbi
->s_kobj
, &f2fs_ktype
, NULL
,
1107 sbi
->need_fsck
= true;
1109 /* recover fsynced data */
1110 if (!test_opt(sbi
, DISABLE_ROLL_FORWARD
)) {
1111 err
= recover_fsync_data(sbi
);
1113 f2fs_msg(sb
, KERN_ERR
,
1114 "Cannot recover all fsync data errno=%ld", err
);
1120 * If filesystem is not mounted as read-only then
1121 * do start the gc_thread.
1123 if (!f2fs_readonly(sb
)) {
1124 /* After POR, we can run background GC thread.*/
1125 err
= start_gc_thread(sbi
);
1132 kobject_del(&sbi
->s_kobj
);
1135 remove_proc_entry("segment_info", sbi
->s_proc
);
1136 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
1138 f2fs_destroy_stats(sbi
);
1143 iput(sbi
->node_inode
);
1145 destroy_node_manager(sbi
);
1147 destroy_segment_manager(sbi
);
1151 make_bad_inode(sbi
->meta_inode
);
1152 iput(sbi
->meta_inode
);
1154 brelse(raw_super_buf
);
1158 /* give only one another chance */
1161 shrink_dcache_sb(sb
);
1167 static struct dentry
*f2fs_mount(struct file_system_type
*fs_type
, int flags
,
1168 const char *dev_name
, void *data
)
1170 return mount_bdev(fs_type
, flags
, dev_name
, data
, f2fs_fill_super
);
1173 static struct file_system_type f2fs_fs_type
= {
1174 .owner
= THIS_MODULE
,
1176 .mount
= f2fs_mount
,
1177 .kill_sb
= kill_block_super
,
1178 .fs_flags
= FS_REQUIRES_DEV
,
1180 MODULE_ALIAS_FS("f2fs");
1182 static int __init
init_inodecache(void)
1184 f2fs_inode_cachep
= f2fs_kmem_cache_create("f2fs_inode_cache",
1185 sizeof(struct f2fs_inode_info
));
1186 if (!f2fs_inode_cachep
)
1191 static void destroy_inodecache(void)
1194 * Make sure all delayed rcu free inodes are flushed before we
1198 kmem_cache_destroy(f2fs_inode_cachep
);
1201 static int __init
init_f2fs_fs(void)
1205 err
= init_inodecache();
1208 err
= create_node_manager_caches();
1210 goto free_inodecache
;
1211 err
= create_segment_manager_caches();
1213 goto free_node_manager_caches
;
1214 err
= create_gc_caches();
1216 goto free_segment_manager_caches
;
1217 err
= create_checkpoint_caches();
1219 goto free_gc_caches
;
1220 f2fs_kset
= kset_create_and_add("f2fs", NULL
, fs_kobj
);
1223 goto free_checkpoint_caches
;
1225 err
= register_filesystem(&f2fs_fs_type
);
1228 f2fs_create_root_stats();
1229 f2fs_proc_root
= proc_mkdir("fs/f2fs", NULL
);
1233 kset_unregister(f2fs_kset
);
1234 free_checkpoint_caches
:
1235 destroy_checkpoint_caches();
1237 destroy_gc_caches();
1238 free_segment_manager_caches
:
1239 destroy_segment_manager_caches();
1240 free_node_manager_caches
:
1241 destroy_node_manager_caches();
1243 destroy_inodecache();
1248 static void __exit
exit_f2fs_fs(void)
1250 remove_proc_entry("fs/f2fs", NULL
);
1251 f2fs_destroy_root_stats();
1252 unregister_filesystem(&f2fs_fs_type
);
1253 destroy_checkpoint_caches();
1254 destroy_gc_caches();
1255 destroy_segment_manager_caches();
1256 destroy_node_manager_caches();
1257 destroy_inodecache();
1258 kset_unregister(f2fs_kset
);
1261 module_init(init_f2fs_fs
)
1262 module_exit(exit_f2fs_fs
)
1264 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1265 MODULE_DESCRIPTION("Flash Friendly File System");
1266 MODULE_LICENSE("GPL");