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
,
60 static match_table_t f2fs_tokens
= {
61 {Opt_gc_background
, "background_gc=%s"},
62 {Opt_disable_roll_forward
, "disable_roll_forward"},
63 {Opt_discard
, "discard"},
64 {Opt_noheap
, "no_heap"},
65 {Opt_user_xattr
, "user_xattr"},
66 {Opt_nouser_xattr
, "nouser_xattr"},
69 {Opt_active_logs
, "active_logs=%u"},
70 {Opt_disable_ext_identify
, "disable_ext_identify"},
71 {Opt_inline_xattr
, "inline_xattr"},
72 {Opt_inline_data
, "inline_data"},
73 {Opt_inline_dentry
, "inline_dentry"},
74 {Opt_flush_merge
, "flush_merge"},
75 {Opt_nobarrier
, "nobarrier"},
79 /* Sysfs support for f2fs */
81 GC_THREAD
, /* struct f2fs_gc_thread */
82 SM_INFO
, /* struct f2fs_sm_info */
83 NM_INFO
, /* struct f2fs_nm_info */
84 F2FS_SBI
, /* struct f2fs_sb_info */
88 struct attribute attr
;
89 ssize_t (*show
)(struct f2fs_attr
*, struct f2fs_sb_info
*, char *);
90 ssize_t (*store
)(struct f2fs_attr
*, struct f2fs_sb_info
*,
91 const char *, size_t);
96 static unsigned char *__struct_ptr(struct f2fs_sb_info
*sbi
, int struct_type
)
98 if (struct_type
== GC_THREAD
)
99 return (unsigned char *)sbi
->gc_thread
;
100 else if (struct_type
== SM_INFO
)
101 return (unsigned char *)SM_I(sbi
);
102 else if (struct_type
== NM_INFO
)
103 return (unsigned char *)NM_I(sbi
);
104 else if (struct_type
== F2FS_SBI
)
105 return (unsigned char *)sbi
;
109 static ssize_t
f2fs_sbi_show(struct f2fs_attr
*a
,
110 struct f2fs_sb_info
*sbi
, char *buf
)
112 unsigned char *ptr
= NULL
;
115 ptr
= __struct_ptr(sbi
, a
->struct_type
);
119 ui
= (unsigned int *)(ptr
+ a
->offset
);
121 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
124 static ssize_t
f2fs_sbi_store(struct f2fs_attr
*a
,
125 struct f2fs_sb_info
*sbi
,
126 const char *buf
, size_t count
)
133 ptr
= __struct_ptr(sbi
, a
->struct_type
);
137 ui
= (unsigned int *)(ptr
+ a
->offset
);
139 ret
= kstrtoul(skip_spaces(buf
), 0, &t
);
146 static ssize_t
f2fs_attr_show(struct kobject
*kobj
,
147 struct attribute
*attr
, char *buf
)
149 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
151 struct f2fs_attr
*a
= container_of(attr
, struct f2fs_attr
, attr
);
153 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
156 static ssize_t
f2fs_attr_store(struct kobject
*kobj
, struct attribute
*attr
,
157 const char *buf
, size_t len
)
159 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
161 struct f2fs_attr
*a
= container_of(attr
, struct f2fs_attr
, attr
);
163 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
166 static void f2fs_sb_release(struct kobject
*kobj
)
168 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
170 complete(&sbi
->s_kobj_unregister
);
173 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
174 static struct f2fs_attr f2fs_attr_##_name = { \
175 .attr = {.name = __stringify(_name), .mode = _mode }, \
178 .struct_type = _struct_type, \
182 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
183 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
184 f2fs_sbi_show, f2fs_sbi_store, \
185 offsetof(struct struct_name, elname))
187 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_min_sleep_time
, min_sleep_time
);
188 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_max_sleep_time
, max_sleep_time
);
189 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_no_gc_sleep_time
, no_gc_sleep_time
);
190 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_idle
, gc_idle
);
191 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, reclaim_segments
, rec_prefree_segments
);
192 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, max_small_discards
, max_discards
);
193 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, ipu_policy
, ipu_policy
);
194 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, min_ipu_util
, min_ipu_util
);
195 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, min_fsync_blocks
, min_fsync_blocks
);
196 F2FS_RW_ATTR(NM_INFO
, f2fs_nm_info
, ram_thresh
, ram_thresh
);
197 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, max_victim_search
, max_victim_search
);
198 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, dir_level
, dir_level
);
200 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
201 static struct attribute
*f2fs_attrs
[] = {
202 ATTR_LIST(gc_min_sleep_time
),
203 ATTR_LIST(gc_max_sleep_time
),
204 ATTR_LIST(gc_no_gc_sleep_time
),
206 ATTR_LIST(reclaim_segments
),
207 ATTR_LIST(max_small_discards
),
208 ATTR_LIST(ipu_policy
),
209 ATTR_LIST(min_ipu_util
),
210 ATTR_LIST(min_fsync_blocks
),
211 ATTR_LIST(max_victim_search
),
212 ATTR_LIST(dir_level
),
213 ATTR_LIST(ram_thresh
),
217 static const struct sysfs_ops f2fs_attr_ops
= {
218 .show
= f2fs_attr_show
,
219 .store
= f2fs_attr_store
,
222 static struct kobj_type f2fs_ktype
= {
223 .default_attrs
= f2fs_attrs
,
224 .sysfs_ops
= &f2fs_attr_ops
,
225 .release
= f2fs_sb_release
,
228 void f2fs_msg(struct super_block
*sb
, const char *level
, const char *fmt
, ...)
230 struct va_format vaf
;
236 printk("%sF2FS-fs (%s): %pV\n", level
, sb
->s_id
, &vaf
);
240 static void init_once(void *foo
)
242 struct f2fs_inode_info
*fi
= (struct f2fs_inode_info
*) foo
;
244 inode_init_once(&fi
->vfs_inode
);
247 static int parse_options(struct super_block
*sb
, char *options
)
249 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
250 substring_t args
[MAX_OPT_ARGS
];
257 while ((p
= strsep(&options
, ",")) != NULL
) {
262 * Initialize args struct so we know whether arg was
263 * found; some options take optional arguments.
265 args
[0].to
= args
[0].from
= NULL
;
266 token
= match_token(p
, f2fs_tokens
, args
);
269 case Opt_gc_background
:
270 name
= match_strdup(&args
[0]);
274 if (strlen(name
) == 2 && !strncmp(name
, "on", 2))
276 else if (strlen(name
) == 3 && !strncmp(name
, "off", 3))
277 clear_opt(sbi
, BG_GC
);
284 case Opt_disable_roll_forward
:
285 set_opt(sbi
, DISABLE_ROLL_FORWARD
);
288 set_opt(sbi
, DISCARD
);
291 set_opt(sbi
, NOHEAP
);
293 #ifdef CONFIG_F2FS_FS_XATTR
295 set_opt(sbi
, XATTR_USER
);
297 case Opt_nouser_xattr
:
298 clear_opt(sbi
, XATTR_USER
);
300 case Opt_inline_xattr
:
301 set_opt(sbi
, INLINE_XATTR
);
305 f2fs_msg(sb
, KERN_INFO
,
306 "user_xattr options not supported");
308 case Opt_nouser_xattr
:
309 f2fs_msg(sb
, KERN_INFO
,
310 "nouser_xattr options not supported");
312 case Opt_inline_xattr
:
313 f2fs_msg(sb
, KERN_INFO
,
314 "inline_xattr options not supported");
317 #ifdef CONFIG_F2FS_FS_POSIX_ACL
319 set_opt(sbi
, POSIX_ACL
);
322 clear_opt(sbi
, POSIX_ACL
);
326 f2fs_msg(sb
, KERN_INFO
, "acl options not supported");
329 f2fs_msg(sb
, KERN_INFO
, "noacl options not supported");
332 case Opt_active_logs
:
333 if (args
->from
&& match_int(args
, &arg
))
335 if (arg
!= 2 && arg
!= 4 && arg
!= NR_CURSEG_TYPE
)
337 sbi
->active_logs
= arg
;
339 case Opt_disable_ext_identify
:
340 set_opt(sbi
, DISABLE_EXT_IDENTIFY
);
342 case Opt_inline_data
:
343 set_opt(sbi
, INLINE_DATA
);
345 case Opt_inline_dentry
:
346 set_opt(sbi
, INLINE_DENTRY
);
348 case Opt_flush_merge
:
349 set_opt(sbi
, FLUSH_MERGE
);
352 set_opt(sbi
, NOBARRIER
);
355 f2fs_msg(sb
, KERN_ERR
,
356 "Unrecognized mount option \"%s\" or missing value",
364 static struct inode
*f2fs_alloc_inode(struct super_block
*sb
)
366 struct f2fs_inode_info
*fi
;
368 fi
= kmem_cache_alloc(f2fs_inode_cachep
, GFP_F2FS_ZERO
);
372 init_once((void *) fi
);
374 /* Initialize f2fs-specific inode info */
375 fi
->vfs_inode
.i_version
= 1;
376 atomic_set(&fi
->dirty_pages
, 0);
377 fi
->i_current_depth
= 1;
379 rwlock_init(&fi
->ext
.ext_lock
);
380 init_rwsem(&fi
->i_sem
);
381 INIT_RADIX_TREE(&fi
->inmem_root
, GFP_NOFS
);
382 INIT_LIST_HEAD(&fi
->inmem_pages
);
383 mutex_init(&fi
->inmem_lock
);
385 set_inode_flag(fi
, FI_NEW_INODE
);
387 if (test_opt(F2FS_SB(sb
), INLINE_XATTR
))
388 set_inode_flag(fi
, FI_INLINE_XATTR
);
390 /* Will be used by directory only */
391 fi
->i_dir_level
= F2FS_SB(sb
)->dir_level
;
393 return &fi
->vfs_inode
;
396 static int f2fs_drop_inode(struct inode
*inode
)
399 * This is to avoid a deadlock condition like below.
400 * writeback_single_inode(inode)
401 * - f2fs_write_data_page
402 * - f2fs_gc -> iput -> evict
403 * - inode_wait_for_writeback(inode)
405 if (!inode_unhashed(inode
) && inode
->i_state
& I_SYNC
)
407 return generic_drop_inode(inode
);
411 * f2fs_dirty_inode() is called from __mark_inode_dirty()
413 * We should call set_dirty_inode to write the dirty inode through write_inode.
415 static void f2fs_dirty_inode(struct inode
*inode
, int flags
)
417 set_inode_flag(F2FS_I(inode
), FI_DIRTY_INODE
);
420 static void f2fs_i_callback(struct rcu_head
*head
)
422 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
423 kmem_cache_free(f2fs_inode_cachep
, F2FS_I(inode
));
426 static void f2fs_destroy_inode(struct inode
*inode
)
428 call_rcu(&inode
->i_rcu
, f2fs_i_callback
);
431 static void f2fs_put_super(struct super_block
*sb
)
433 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
436 remove_proc_entry("segment_info", sbi
->s_proc
);
437 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
439 kobject_del(&sbi
->s_kobj
);
441 f2fs_destroy_stats(sbi
);
444 /* We don't need to do checkpoint when it's clean */
446 struct cp_control cpc
= {
449 write_checkpoint(sbi
, &cpc
);
453 * normally superblock is clean, so we need to release this.
454 * In addition, EIO will skip do checkpoint, we need this as well.
456 release_dirty_inode(sbi
);
457 release_discard_addrs(sbi
);
459 iput(sbi
->node_inode
);
460 iput(sbi
->meta_inode
);
462 /* destroy f2fs internal modules */
463 destroy_node_manager(sbi
);
464 destroy_segment_manager(sbi
);
467 kobject_put(&sbi
->s_kobj
);
468 wait_for_completion(&sbi
->s_kobj_unregister
);
470 sb
->s_fs_info
= NULL
;
471 brelse(sbi
->raw_super_buf
);
475 int f2fs_sync_fs(struct super_block
*sb
, int sync
)
477 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
479 trace_f2fs_sync_fs(sb
, sync
);
482 struct cp_control cpc
= {
485 mutex_lock(&sbi
->gc_mutex
);
486 write_checkpoint(sbi
, &cpc
);
487 mutex_unlock(&sbi
->gc_mutex
);
489 f2fs_balance_fs(sbi
);
495 static int f2fs_freeze(struct super_block
*sb
)
499 if (f2fs_readonly(sb
))
502 err
= f2fs_sync_fs(sb
, 1);
506 static int f2fs_unfreeze(struct super_block
*sb
)
511 static int f2fs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
513 struct super_block
*sb
= dentry
->d_sb
;
514 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
515 u64 id
= huge_encode_dev(sb
->s_bdev
->bd_dev
);
516 block_t total_count
, user_block_count
, start_count
, ovp_count
;
518 total_count
= le64_to_cpu(sbi
->raw_super
->block_count
);
519 user_block_count
= sbi
->user_block_count
;
520 start_count
= le32_to_cpu(sbi
->raw_super
->segment0_blkaddr
);
521 ovp_count
= SM_I(sbi
)->ovp_segments
<< sbi
->log_blocks_per_seg
;
522 buf
->f_type
= F2FS_SUPER_MAGIC
;
523 buf
->f_bsize
= sbi
->blocksize
;
525 buf
->f_blocks
= total_count
- start_count
;
526 buf
->f_bfree
= buf
->f_blocks
- valid_user_blocks(sbi
) - ovp_count
;
527 buf
->f_bavail
= user_block_count
- valid_user_blocks(sbi
);
529 buf
->f_files
= sbi
->total_node_count
- F2FS_RESERVED_NODE_NUM
;
530 buf
->f_ffree
= buf
->f_files
- valid_inode_count(sbi
);
532 buf
->f_namelen
= F2FS_NAME_LEN
;
533 buf
->f_fsid
.val
[0] = (u32
)id
;
534 buf
->f_fsid
.val
[1] = (u32
)(id
>> 32);
539 static int f2fs_show_options(struct seq_file
*seq
, struct dentry
*root
)
541 struct f2fs_sb_info
*sbi
= F2FS_SB(root
->d_sb
);
543 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, BG_GC
))
544 seq_printf(seq
, ",background_gc=%s", "on");
546 seq_printf(seq
, ",background_gc=%s", "off");
547 if (test_opt(sbi
, DISABLE_ROLL_FORWARD
))
548 seq_puts(seq
, ",disable_roll_forward");
549 if (test_opt(sbi
, DISCARD
))
550 seq_puts(seq
, ",discard");
551 if (test_opt(sbi
, NOHEAP
))
552 seq_puts(seq
, ",no_heap_alloc");
553 #ifdef CONFIG_F2FS_FS_XATTR
554 if (test_opt(sbi
, XATTR_USER
))
555 seq_puts(seq
, ",user_xattr");
557 seq_puts(seq
, ",nouser_xattr");
558 if (test_opt(sbi
, INLINE_XATTR
))
559 seq_puts(seq
, ",inline_xattr");
561 #ifdef CONFIG_F2FS_FS_POSIX_ACL
562 if (test_opt(sbi
, POSIX_ACL
))
563 seq_puts(seq
, ",acl");
565 seq_puts(seq
, ",noacl");
567 if (test_opt(sbi
, DISABLE_EXT_IDENTIFY
))
568 seq_puts(seq
, ",disable_ext_identify");
569 if (test_opt(sbi
, INLINE_DATA
))
570 seq_puts(seq
, ",inline_data");
571 if (test_opt(sbi
, INLINE_DENTRY
))
572 seq_puts(seq
, ",inline_dentry");
573 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, FLUSH_MERGE
))
574 seq_puts(seq
, ",flush_merge");
575 if (test_opt(sbi
, NOBARRIER
))
576 seq_puts(seq
, ",nobarrier");
577 seq_printf(seq
, ",active_logs=%u", sbi
->active_logs
);
582 static int segment_info_seq_show(struct seq_file
*seq
, void *offset
)
584 struct super_block
*sb
= seq
->private;
585 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
586 unsigned int total_segs
=
587 le32_to_cpu(sbi
->raw_super
->segment_count_main
);
590 seq_puts(seq
, "format: segment_type|valid_blocks\n"
591 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
593 for (i
= 0; i
< total_segs
; i
++) {
594 struct seg_entry
*se
= get_seg_entry(sbi
, i
);
597 seq_printf(seq
, "%-5d", i
);
598 seq_printf(seq
, "%d|%-3u", se
->type
,
599 get_valid_blocks(sbi
, i
, 1));
600 if ((i
% 10) == 9 || i
== (total_segs
- 1))
609 static int segment_info_open_fs(struct inode
*inode
, struct file
*file
)
611 return single_open(file
, segment_info_seq_show
, PDE_DATA(inode
));
614 static const struct file_operations f2fs_seq_segment_info_fops
= {
615 .owner
= THIS_MODULE
,
616 .open
= segment_info_open_fs
,
619 .release
= single_release
,
622 static int f2fs_remount(struct super_block
*sb
, int *flags
, char *data
)
624 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
625 struct f2fs_mount_info org_mount_opt
;
626 int err
, active_logs
;
627 bool need_restart_gc
= false;
628 bool need_stop_gc
= false;
633 * Save the old mount options in case we
634 * need to restore them.
636 org_mount_opt
= sbi
->mount_opt
;
637 active_logs
= sbi
->active_logs
;
639 sbi
->mount_opt
.opt
= 0;
640 sbi
->active_logs
= NR_CURSEG_TYPE
;
642 /* parse mount options */
643 err
= parse_options(sb
, data
);
648 * Previous and new state of filesystem is RO,
649 * so skip checking GC and FLUSH_MERGE conditions.
651 if (f2fs_readonly(sb
) && (*flags
& MS_RDONLY
))
655 * We stop the GC thread if FS is mounted as RO
656 * or if background_gc = off is passed in mount
657 * option. Also sync the filesystem.
659 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, BG_GC
)) {
660 if (sbi
->gc_thread
) {
663 need_restart_gc
= true;
665 } else if (test_opt(sbi
, BG_GC
) && !sbi
->gc_thread
) {
666 err
= start_gc_thread(sbi
);
673 * We stop issue flush thread if FS is mounted as RO
674 * or if flush_merge is not passed in mount option.
676 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, FLUSH_MERGE
)) {
677 destroy_flush_cmd_control(sbi
);
678 } else if (test_opt(sbi
, FLUSH_MERGE
) && !SM_I(sbi
)->cmd_control_info
) {
679 err
= create_flush_cmd_control(sbi
);
684 /* Update the POSIXACL Flag */
685 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
686 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
689 if (need_restart_gc
) {
690 if (start_gc_thread(sbi
))
691 f2fs_msg(sbi
->sb
, KERN_WARNING
,
692 "background gc thread has stopped");
693 } else if (need_stop_gc
) {
697 sbi
->mount_opt
= org_mount_opt
;
698 sbi
->active_logs
= active_logs
;
702 static struct super_operations f2fs_sops
= {
703 .alloc_inode
= f2fs_alloc_inode
,
704 .drop_inode
= f2fs_drop_inode
,
705 .destroy_inode
= f2fs_destroy_inode
,
706 .write_inode
= f2fs_write_inode
,
707 .dirty_inode
= f2fs_dirty_inode
,
708 .show_options
= f2fs_show_options
,
709 .evict_inode
= f2fs_evict_inode
,
710 .put_super
= f2fs_put_super
,
711 .sync_fs
= f2fs_sync_fs
,
712 .freeze_fs
= f2fs_freeze
,
713 .unfreeze_fs
= f2fs_unfreeze
,
714 .statfs
= f2fs_statfs
,
715 .remount_fs
= f2fs_remount
,
718 static struct inode
*f2fs_nfs_get_inode(struct super_block
*sb
,
719 u64 ino
, u32 generation
)
721 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
724 if (check_nid_range(sbi
, ino
))
725 return ERR_PTR(-ESTALE
);
728 * f2fs_iget isn't quite right if the inode is currently unallocated!
729 * However f2fs_iget currently does appropriate checks to handle stale
730 * inodes so everything is OK.
732 inode
= f2fs_iget(sb
, ino
);
734 return ERR_CAST(inode
);
735 if (unlikely(generation
&& inode
->i_generation
!= generation
)) {
736 /* we didn't find the right inode.. */
738 return ERR_PTR(-ESTALE
);
743 static struct dentry
*f2fs_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
744 int fh_len
, int fh_type
)
746 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
750 static struct dentry
*f2fs_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
751 int fh_len
, int fh_type
)
753 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
757 static const struct export_operations f2fs_export_ops
= {
758 .fh_to_dentry
= f2fs_fh_to_dentry
,
759 .fh_to_parent
= f2fs_fh_to_parent
,
760 .get_parent
= f2fs_get_parent
,
763 static loff_t
max_file_size(unsigned bits
)
765 loff_t result
= (DEF_ADDRS_PER_INODE
- F2FS_INLINE_XATTR_ADDRS
);
766 loff_t leaf_count
= ADDRS_PER_BLOCK
;
768 /* two direct node blocks */
769 result
+= (leaf_count
* 2);
771 /* two indirect node blocks */
772 leaf_count
*= NIDS_PER_BLOCK
;
773 result
+= (leaf_count
* 2);
775 /* one double indirect node block */
776 leaf_count
*= NIDS_PER_BLOCK
;
777 result
+= leaf_count
;
783 static int sanity_check_raw_super(struct super_block
*sb
,
784 struct f2fs_super_block
*raw_super
)
786 unsigned int blocksize
;
788 if (F2FS_SUPER_MAGIC
!= le32_to_cpu(raw_super
->magic
)) {
789 f2fs_msg(sb
, KERN_INFO
,
790 "Magic Mismatch, valid(0x%x) - read(0x%x)",
791 F2FS_SUPER_MAGIC
, le32_to_cpu(raw_super
->magic
));
795 /* Currently, support only 4KB page cache size */
796 if (F2FS_BLKSIZE
!= PAGE_CACHE_SIZE
) {
797 f2fs_msg(sb
, KERN_INFO
,
798 "Invalid page_cache_size (%lu), supports only 4KB\n",
803 /* Currently, support only 4KB block size */
804 blocksize
= 1 << le32_to_cpu(raw_super
->log_blocksize
);
805 if (blocksize
!= F2FS_BLKSIZE
) {
806 f2fs_msg(sb
, KERN_INFO
,
807 "Invalid blocksize (%u), supports only 4KB\n",
812 /* Currently, support 512/1024/2048/4096 bytes sector size */
813 if (le32_to_cpu(raw_super
->log_sectorsize
) >
814 F2FS_MAX_LOG_SECTOR_SIZE
||
815 le32_to_cpu(raw_super
->log_sectorsize
) <
816 F2FS_MIN_LOG_SECTOR_SIZE
) {
817 f2fs_msg(sb
, KERN_INFO
, "Invalid log sectorsize (%u)",
818 le32_to_cpu(raw_super
->log_sectorsize
));
821 if (le32_to_cpu(raw_super
->log_sectors_per_block
) +
822 le32_to_cpu(raw_super
->log_sectorsize
) !=
823 F2FS_MAX_LOG_SECTOR_SIZE
) {
824 f2fs_msg(sb
, KERN_INFO
,
825 "Invalid log sectors per block(%u) log sectorsize(%u)",
826 le32_to_cpu(raw_super
->log_sectors_per_block
),
827 le32_to_cpu(raw_super
->log_sectorsize
));
833 static int sanity_check_ckpt(struct f2fs_sb_info
*sbi
)
835 unsigned int total
, fsmeta
;
836 struct f2fs_super_block
*raw_super
= F2FS_RAW_SUPER(sbi
);
837 struct f2fs_checkpoint
*ckpt
= F2FS_CKPT(sbi
);
839 total
= le32_to_cpu(raw_super
->segment_count
);
840 fsmeta
= le32_to_cpu(raw_super
->segment_count_ckpt
);
841 fsmeta
+= le32_to_cpu(raw_super
->segment_count_sit
);
842 fsmeta
+= le32_to_cpu(raw_super
->segment_count_nat
);
843 fsmeta
+= le32_to_cpu(ckpt
->rsvd_segment_count
);
844 fsmeta
+= le32_to_cpu(raw_super
->segment_count_ssa
);
846 if (unlikely(fsmeta
>= total
))
849 if (unlikely(f2fs_cp_error(sbi
))) {
850 f2fs_msg(sbi
->sb
, KERN_ERR
, "A bug case: need to run fsck");
856 static void init_sb_info(struct f2fs_sb_info
*sbi
)
858 struct f2fs_super_block
*raw_super
= sbi
->raw_super
;
861 sbi
->log_sectors_per_block
=
862 le32_to_cpu(raw_super
->log_sectors_per_block
);
863 sbi
->log_blocksize
= le32_to_cpu(raw_super
->log_blocksize
);
864 sbi
->blocksize
= 1 << sbi
->log_blocksize
;
865 sbi
->log_blocks_per_seg
= le32_to_cpu(raw_super
->log_blocks_per_seg
);
866 sbi
->blocks_per_seg
= 1 << sbi
->log_blocks_per_seg
;
867 sbi
->segs_per_sec
= le32_to_cpu(raw_super
->segs_per_sec
);
868 sbi
->secs_per_zone
= le32_to_cpu(raw_super
->secs_per_zone
);
869 sbi
->total_sections
= le32_to_cpu(raw_super
->section_count
);
870 sbi
->total_node_count
=
871 (le32_to_cpu(raw_super
->segment_count_nat
) / 2)
872 * sbi
->blocks_per_seg
* NAT_ENTRY_PER_BLOCK
;
873 sbi
->root_ino_num
= le32_to_cpu(raw_super
->root_ino
);
874 sbi
->node_ino_num
= le32_to_cpu(raw_super
->node_ino
);
875 sbi
->meta_ino_num
= le32_to_cpu(raw_super
->meta_ino
);
876 sbi
->cur_victim_sec
= NULL_SECNO
;
877 sbi
->max_victim_search
= DEF_MAX_VICTIM_SEARCH
;
879 for (i
= 0; i
< NR_COUNT_TYPE
; i
++)
880 atomic_set(&sbi
->nr_pages
[i
], 0);
882 sbi
->dir_level
= DEF_DIR_LEVEL
;
883 sbi
->need_fsck
= false;
887 * Read f2fs raw super block.
888 * Because we have two copies of super block, so read the first one at first,
889 * if the first one is invalid, move to read the second one.
891 static int read_raw_super_block(struct super_block
*sb
,
892 struct f2fs_super_block
**raw_super
,
893 struct buffer_head
**raw_super_buf
)
898 *raw_super_buf
= sb_bread(sb
, block
);
899 if (!*raw_super_buf
) {
900 f2fs_msg(sb
, KERN_ERR
, "Unable to read %dth superblock",
910 *raw_super
= (struct f2fs_super_block
*)
911 ((char *)(*raw_super_buf
)->b_data
+ F2FS_SUPER_OFFSET
);
913 /* sanity checking of raw super */
914 if (sanity_check_raw_super(sb
, *raw_super
)) {
915 brelse(*raw_super_buf
);
916 f2fs_msg(sb
, KERN_ERR
,
917 "Can't find valid F2FS filesystem in %dth superblock",
930 static int f2fs_fill_super(struct super_block
*sb
, void *data
, int silent
)
932 struct f2fs_sb_info
*sbi
;
933 struct f2fs_super_block
*raw_super
;
934 struct buffer_head
*raw_super_buf
;
941 /* allocate memory for f2fs-specific super block info */
942 sbi
= kzalloc(sizeof(struct f2fs_sb_info
), GFP_KERNEL
);
946 /* set a block size */
947 if (unlikely(!sb_set_blocksize(sb
, F2FS_BLKSIZE
))) {
948 f2fs_msg(sb
, KERN_ERR
, "unable to set blocksize");
952 err
= read_raw_super_block(sb
, &raw_super
, &raw_super_buf
);
957 /* init some FS parameters */
958 sbi
->active_logs
= NR_CURSEG_TYPE
;
962 #ifdef CONFIG_F2FS_FS_XATTR
963 set_opt(sbi
, XATTR_USER
);
965 #ifdef CONFIG_F2FS_FS_POSIX_ACL
966 set_opt(sbi
, POSIX_ACL
);
968 /* parse mount options */
969 err
= parse_options(sb
, (char *)data
);
973 sb
->s_maxbytes
= max_file_size(le32_to_cpu(raw_super
->log_blocksize
));
974 sb
->s_max_links
= F2FS_LINK_MAX
;
975 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
977 sb
->s_op
= &f2fs_sops
;
978 sb
->s_xattr
= f2fs_xattr_handlers
;
979 sb
->s_export_op
= &f2fs_export_ops
;
980 sb
->s_magic
= F2FS_SUPER_MAGIC
;
982 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
983 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
984 memcpy(sb
->s_uuid
, raw_super
->uuid
, sizeof(raw_super
->uuid
));
986 /* init f2fs-specific super block info */
988 sbi
->raw_super
= raw_super
;
989 sbi
->raw_super_buf
= raw_super_buf
;
990 mutex_init(&sbi
->gc_mutex
);
991 mutex_init(&sbi
->writepages
);
992 mutex_init(&sbi
->cp_mutex
);
993 init_rwsem(&sbi
->node_write
);
994 sbi
->por_doing
= false;
995 spin_lock_init(&sbi
->stat_lock
);
997 init_rwsem(&sbi
->read_io
.io_rwsem
);
998 sbi
->read_io
.sbi
= sbi
;
999 sbi
->read_io
.bio
= NULL
;
1000 for (i
= 0; i
< NR_PAGE_TYPE
; i
++) {
1001 init_rwsem(&sbi
->write_io
[i
].io_rwsem
);
1002 sbi
->write_io
[i
].sbi
= sbi
;
1003 sbi
->write_io
[i
].bio
= NULL
;
1006 init_rwsem(&sbi
->cp_rwsem
);
1007 init_waitqueue_head(&sbi
->cp_wait
);
1010 /* get an inode for meta space */
1011 sbi
->meta_inode
= f2fs_iget(sb
, F2FS_META_INO(sbi
));
1012 if (IS_ERR(sbi
->meta_inode
)) {
1013 f2fs_msg(sb
, KERN_ERR
, "Failed to read F2FS meta data inode");
1014 err
= PTR_ERR(sbi
->meta_inode
);
1018 err
= get_valid_checkpoint(sbi
);
1020 f2fs_msg(sb
, KERN_ERR
, "Failed to get valid F2FS checkpoint");
1021 goto free_meta_inode
;
1024 /* sanity checking of checkpoint */
1026 if (sanity_check_ckpt(sbi
)) {
1027 f2fs_msg(sb
, KERN_ERR
, "Invalid F2FS checkpoint");
1031 sbi
->total_valid_node_count
=
1032 le32_to_cpu(sbi
->ckpt
->valid_node_count
);
1033 sbi
->total_valid_inode_count
=
1034 le32_to_cpu(sbi
->ckpt
->valid_inode_count
);
1035 sbi
->user_block_count
= le64_to_cpu(sbi
->ckpt
->user_block_count
);
1036 sbi
->total_valid_block_count
=
1037 le64_to_cpu(sbi
->ckpt
->valid_block_count
);
1038 sbi
->last_valid_block_count
= sbi
->total_valid_block_count
;
1039 sbi
->alloc_valid_block_count
= 0;
1040 INIT_LIST_HEAD(&sbi
->dir_inode_list
);
1041 spin_lock_init(&sbi
->dir_inode_lock
);
1043 init_ino_entry_info(sbi
);
1045 /* setup f2fs internal modules */
1046 err
= build_segment_manager(sbi
);
1048 f2fs_msg(sb
, KERN_ERR
,
1049 "Failed to initialize F2FS segment manager");
1052 err
= build_node_manager(sbi
);
1054 f2fs_msg(sb
, KERN_ERR
,
1055 "Failed to initialize F2FS node manager");
1059 build_gc_manager(sbi
);
1061 /* get an inode for node space */
1062 sbi
->node_inode
= f2fs_iget(sb
, F2FS_NODE_INO(sbi
));
1063 if (IS_ERR(sbi
->node_inode
)) {
1064 f2fs_msg(sb
, KERN_ERR
, "Failed to read node inode");
1065 err
= PTR_ERR(sbi
->node_inode
);
1069 /* if there are nt orphan nodes free them */
1070 recover_orphan_inodes(sbi
);
1072 /* read root inode and dentry */
1073 root
= f2fs_iget(sb
, F2FS_ROOT_INO(sbi
));
1075 f2fs_msg(sb
, KERN_ERR
, "Failed to read root inode");
1076 err
= PTR_ERR(root
);
1077 goto free_node_inode
;
1079 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
1082 goto free_node_inode
;
1085 sb
->s_root
= d_make_root(root
); /* allocate root dentry */
1088 goto free_root_inode
;
1091 err
= f2fs_build_stats(sbi
);
1093 goto free_root_inode
;
1096 sbi
->s_proc
= proc_mkdir(sb
->s_id
, f2fs_proc_root
);
1099 proc_create_data("segment_info", S_IRUGO
, sbi
->s_proc
,
1100 &f2fs_seq_segment_info_fops
, sb
);
1102 if (test_opt(sbi
, DISCARD
)) {
1103 struct request_queue
*q
= bdev_get_queue(sb
->s_bdev
);
1104 if (!blk_queue_discard(q
))
1105 f2fs_msg(sb
, KERN_WARNING
,
1106 "mounting with \"discard\" option, but "
1107 "the device does not support discard");
1110 sbi
->s_kobj
.kset
= f2fs_kset
;
1111 init_completion(&sbi
->s_kobj_unregister
);
1112 err
= kobject_init_and_add(&sbi
->s_kobj
, &f2fs_ktype
, NULL
,
1118 sbi
->need_fsck
= true;
1120 /* recover fsynced data */
1121 if (!test_opt(sbi
, DISABLE_ROLL_FORWARD
)) {
1122 err
= recover_fsync_data(sbi
);
1124 f2fs_msg(sb
, KERN_ERR
,
1125 "Cannot recover all fsync data errno=%ld", err
);
1131 * If filesystem is not mounted as read-only then
1132 * do start the gc_thread.
1134 if (!f2fs_readonly(sb
)) {
1135 /* After POR, we can run background GC thread.*/
1136 err
= start_gc_thread(sbi
);
1143 kobject_del(&sbi
->s_kobj
);
1146 remove_proc_entry("segment_info", sbi
->s_proc
);
1147 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
1149 f2fs_destroy_stats(sbi
);
1154 iput(sbi
->node_inode
);
1156 destroy_node_manager(sbi
);
1158 destroy_segment_manager(sbi
);
1162 make_bad_inode(sbi
->meta_inode
);
1163 iput(sbi
->meta_inode
);
1165 brelse(raw_super_buf
);
1169 /* give only one another chance */
1172 shrink_dcache_sb(sb
);
1178 static struct dentry
*f2fs_mount(struct file_system_type
*fs_type
, int flags
,
1179 const char *dev_name
, void *data
)
1181 return mount_bdev(fs_type
, flags
, dev_name
, data
, f2fs_fill_super
);
1184 static struct file_system_type f2fs_fs_type
= {
1185 .owner
= THIS_MODULE
,
1187 .mount
= f2fs_mount
,
1188 .kill_sb
= kill_block_super
,
1189 .fs_flags
= FS_REQUIRES_DEV
,
1191 MODULE_ALIAS_FS("f2fs");
1193 static int __init
init_inodecache(void)
1195 f2fs_inode_cachep
= f2fs_kmem_cache_create("f2fs_inode_cache",
1196 sizeof(struct f2fs_inode_info
));
1197 if (!f2fs_inode_cachep
)
1202 static void destroy_inodecache(void)
1205 * Make sure all delayed rcu free inodes are flushed before we
1209 kmem_cache_destroy(f2fs_inode_cachep
);
1212 static int __init
init_f2fs_fs(void)
1216 err
= init_inodecache();
1219 err
= create_node_manager_caches();
1221 goto free_inodecache
;
1222 err
= create_segment_manager_caches();
1224 goto free_node_manager_caches
;
1225 err
= create_gc_caches();
1227 goto free_segment_manager_caches
;
1228 err
= create_checkpoint_caches();
1230 goto free_gc_caches
;
1231 f2fs_kset
= kset_create_and_add("f2fs", NULL
, fs_kobj
);
1234 goto free_checkpoint_caches
;
1236 err
= register_filesystem(&f2fs_fs_type
);
1239 f2fs_create_root_stats();
1240 f2fs_proc_root
= proc_mkdir("fs/f2fs", NULL
);
1244 kset_unregister(f2fs_kset
);
1245 free_checkpoint_caches
:
1246 destroy_checkpoint_caches();
1248 destroy_gc_caches();
1249 free_segment_manager_caches
:
1250 destroy_segment_manager_caches();
1251 free_node_manager_caches
:
1252 destroy_node_manager_caches();
1254 destroy_inodecache();
1259 static void __exit
exit_f2fs_fs(void)
1261 remove_proc_entry("fs/f2fs", NULL
);
1262 f2fs_destroy_root_stats();
1263 unregister_filesystem(&f2fs_fs_type
);
1264 destroy_checkpoint_caches();
1265 destroy_gc_caches();
1266 destroy_segment_manager_caches();
1267 destroy_node_manager_caches();
1268 destroy_inodecache();
1269 kset_unregister(f2fs_kset
);
1272 module_init(init_f2fs_fs
)
1273 module_exit(exit_f2fs_fs
)
1275 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1276 MODULE_DESCRIPTION("Flash Friendly File System");
1277 MODULE_LICENSE("GPL");