4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
11 #include <linux/module.h>
12 #include <linux/init.h>
14 #include <linux/statfs.h>
15 #include <linux/buffer_head.h>
16 #include <linux/backing-dev.h>
17 #include <linux/kthread.h>
18 #include <linux/parser.h>
19 #include <linux/mount.h>
20 #include <linux/seq_file.h>
21 #include <linux/proc_fs.h>
22 #include <linux/random.h>
23 #include <linux/exportfs.h>
24 #include <linux/blkdev.h>
25 #include <linux/f2fs_fs.h>
26 #include <linux/sysfs.h>
35 #define CREATE_TRACE_POINTS
36 #include <trace/events/f2fs.h>
38 static struct proc_dir_entry
*f2fs_proc_root
;
39 static struct kmem_cache
*f2fs_inode_cachep
;
40 static struct kset
*f2fs_kset
;
42 /* f2fs-wide shrinker description */
43 static struct shrinker f2fs_shrinker_info
= {
44 .scan_objects
= f2fs_shrink_scan
,
45 .count_objects
= f2fs_shrink_count
,
46 .seeks
= DEFAULT_SEEKS
,
51 Opt_disable_roll_forward
,
60 Opt_disable_ext_identify
,
73 static match_table_t f2fs_tokens
= {
74 {Opt_gc_background
, "background_gc=%s"},
75 {Opt_disable_roll_forward
, "disable_roll_forward"},
76 {Opt_norecovery
, "norecovery"},
77 {Opt_discard
, "discard"},
78 {Opt_noheap
, "no_heap"},
79 {Opt_user_xattr
, "user_xattr"},
80 {Opt_nouser_xattr
, "nouser_xattr"},
83 {Opt_active_logs
, "active_logs=%u"},
84 {Opt_disable_ext_identify
, "disable_ext_identify"},
85 {Opt_inline_xattr
, "inline_xattr"},
86 {Opt_inline_data
, "inline_data"},
87 {Opt_inline_dentry
, "inline_dentry"},
88 {Opt_flush_merge
, "flush_merge"},
89 {Opt_nobarrier
, "nobarrier"},
90 {Opt_fastboot
, "fastboot"},
91 {Opt_extent_cache
, "extent_cache"},
92 {Opt_noextent_cache
, "noextent_cache"},
93 {Opt_noinline_data
, "noinline_data"},
97 /* Sysfs support for f2fs */
99 GC_THREAD
, /* struct f2fs_gc_thread */
100 SM_INFO
, /* struct f2fs_sm_info */
101 NM_INFO
, /* struct f2fs_nm_info */
102 F2FS_SBI
, /* struct f2fs_sb_info */
106 struct attribute attr
;
107 ssize_t (*show
)(struct f2fs_attr
*, struct f2fs_sb_info
*, char *);
108 ssize_t (*store
)(struct f2fs_attr
*, struct f2fs_sb_info
*,
109 const char *, size_t);
114 static unsigned char *__struct_ptr(struct f2fs_sb_info
*sbi
, int struct_type
)
116 if (struct_type
== GC_THREAD
)
117 return (unsigned char *)sbi
->gc_thread
;
118 else if (struct_type
== SM_INFO
)
119 return (unsigned char *)SM_I(sbi
);
120 else if (struct_type
== NM_INFO
)
121 return (unsigned char *)NM_I(sbi
);
122 else if (struct_type
== F2FS_SBI
)
123 return (unsigned char *)sbi
;
127 static ssize_t
f2fs_sbi_show(struct f2fs_attr
*a
,
128 struct f2fs_sb_info
*sbi
, char *buf
)
130 unsigned char *ptr
= NULL
;
133 ptr
= __struct_ptr(sbi
, a
->struct_type
);
137 ui
= (unsigned int *)(ptr
+ a
->offset
);
139 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
142 static ssize_t
f2fs_sbi_store(struct f2fs_attr
*a
,
143 struct f2fs_sb_info
*sbi
,
144 const char *buf
, size_t count
)
151 ptr
= __struct_ptr(sbi
, a
->struct_type
);
155 ui
= (unsigned int *)(ptr
+ a
->offset
);
157 ret
= kstrtoul(skip_spaces(buf
), 0, &t
);
164 static ssize_t
f2fs_attr_show(struct kobject
*kobj
,
165 struct attribute
*attr
, char *buf
)
167 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
169 struct f2fs_attr
*a
= container_of(attr
, struct f2fs_attr
, attr
);
171 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
174 static ssize_t
f2fs_attr_store(struct kobject
*kobj
, struct attribute
*attr
,
175 const char *buf
, size_t len
)
177 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
179 struct f2fs_attr
*a
= container_of(attr
, struct f2fs_attr
, attr
);
181 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
184 static void f2fs_sb_release(struct kobject
*kobj
)
186 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
188 complete(&sbi
->s_kobj_unregister
);
191 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
192 static struct f2fs_attr f2fs_attr_##_name = { \
193 .attr = {.name = __stringify(_name), .mode = _mode }, \
196 .struct_type = _struct_type, \
200 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
201 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
202 f2fs_sbi_show, f2fs_sbi_store, \
203 offsetof(struct struct_name, elname))
205 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_min_sleep_time
, min_sleep_time
);
206 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_max_sleep_time
, max_sleep_time
);
207 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_no_gc_sleep_time
, no_gc_sleep_time
);
208 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_idle
, gc_idle
);
209 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, reclaim_segments
, rec_prefree_segments
);
210 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, max_small_discards
, max_discards
);
211 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, batched_trim_sections
, trim_sections
);
212 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, ipu_policy
, ipu_policy
);
213 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, min_ipu_util
, min_ipu_util
);
214 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, min_fsync_blocks
, min_fsync_blocks
);
215 F2FS_RW_ATTR(NM_INFO
, f2fs_nm_info
, ram_thresh
, ram_thresh
);
216 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, max_victim_search
, max_victim_search
);
217 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, dir_level
, dir_level
);
219 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
220 static struct attribute
*f2fs_attrs
[] = {
221 ATTR_LIST(gc_min_sleep_time
),
222 ATTR_LIST(gc_max_sleep_time
),
223 ATTR_LIST(gc_no_gc_sleep_time
),
225 ATTR_LIST(reclaim_segments
),
226 ATTR_LIST(max_small_discards
),
227 ATTR_LIST(batched_trim_sections
),
228 ATTR_LIST(ipu_policy
),
229 ATTR_LIST(min_ipu_util
),
230 ATTR_LIST(min_fsync_blocks
),
231 ATTR_LIST(max_victim_search
),
232 ATTR_LIST(dir_level
),
233 ATTR_LIST(ram_thresh
),
237 static const struct sysfs_ops f2fs_attr_ops
= {
238 .show
= f2fs_attr_show
,
239 .store
= f2fs_attr_store
,
242 static struct kobj_type f2fs_ktype
= {
243 .default_attrs
= f2fs_attrs
,
244 .sysfs_ops
= &f2fs_attr_ops
,
245 .release
= f2fs_sb_release
,
248 void f2fs_msg(struct super_block
*sb
, const char *level
, const char *fmt
, ...)
250 struct va_format vaf
;
256 printk("%sF2FS-fs (%s): %pV\n", level
, sb
->s_id
, &vaf
);
260 static void init_once(void *foo
)
262 struct f2fs_inode_info
*fi
= (struct f2fs_inode_info
*) foo
;
264 inode_init_once(&fi
->vfs_inode
);
267 static int parse_options(struct super_block
*sb
, char *options
)
269 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
270 struct request_queue
*q
;
271 substring_t args
[MAX_OPT_ARGS
];
278 while ((p
= strsep(&options
, ",")) != NULL
) {
283 * Initialize args struct so we know whether arg was
284 * found; some options take optional arguments.
286 args
[0].to
= args
[0].from
= NULL
;
287 token
= match_token(p
, f2fs_tokens
, args
);
290 case Opt_gc_background
:
291 name
= match_strdup(&args
[0]);
295 if (strlen(name
) == 2 && !strncmp(name
, "on", 2))
297 else if (strlen(name
) == 3 && !strncmp(name
, "off", 3))
298 clear_opt(sbi
, BG_GC
);
305 case Opt_disable_roll_forward
:
306 set_opt(sbi
, DISABLE_ROLL_FORWARD
);
309 /* this option mounts f2fs with ro */
310 set_opt(sbi
, DISABLE_ROLL_FORWARD
);
311 if (!f2fs_readonly(sb
))
315 q
= bdev_get_queue(sb
->s_bdev
);
316 if (blk_queue_discard(q
)) {
317 set_opt(sbi
, DISCARD
);
319 f2fs_msg(sb
, KERN_WARNING
,
320 "mounting with \"discard\" option, but "
321 "the device does not support discard");
325 set_opt(sbi
, NOHEAP
);
327 #ifdef CONFIG_F2FS_FS_XATTR
329 set_opt(sbi
, XATTR_USER
);
331 case Opt_nouser_xattr
:
332 clear_opt(sbi
, XATTR_USER
);
334 case Opt_inline_xattr
:
335 set_opt(sbi
, INLINE_XATTR
);
339 f2fs_msg(sb
, KERN_INFO
,
340 "user_xattr options not supported");
342 case Opt_nouser_xattr
:
343 f2fs_msg(sb
, KERN_INFO
,
344 "nouser_xattr options not supported");
346 case Opt_inline_xattr
:
347 f2fs_msg(sb
, KERN_INFO
,
348 "inline_xattr options not supported");
351 #ifdef CONFIG_F2FS_FS_POSIX_ACL
353 set_opt(sbi
, POSIX_ACL
);
356 clear_opt(sbi
, POSIX_ACL
);
360 f2fs_msg(sb
, KERN_INFO
, "acl options not supported");
363 f2fs_msg(sb
, KERN_INFO
, "noacl options not supported");
366 case Opt_active_logs
:
367 if (args
->from
&& match_int(args
, &arg
))
369 if (arg
!= 2 && arg
!= 4 && arg
!= NR_CURSEG_TYPE
)
371 sbi
->active_logs
= arg
;
373 case Opt_disable_ext_identify
:
374 set_opt(sbi
, DISABLE_EXT_IDENTIFY
);
376 case Opt_inline_data
:
377 set_opt(sbi
, INLINE_DATA
);
379 case Opt_inline_dentry
:
380 set_opt(sbi
, INLINE_DENTRY
);
382 case Opt_flush_merge
:
383 set_opt(sbi
, FLUSH_MERGE
);
386 set_opt(sbi
, NOBARRIER
);
389 set_opt(sbi
, FASTBOOT
);
391 case Opt_extent_cache
:
392 set_opt(sbi
, EXTENT_CACHE
);
394 case Opt_noextent_cache
:
395 clear_opt(sbi
, EXTENT_CACHE
);
397 case Opt_noinline_data
:
398 clear_opt(sbi
, INLINE_DATA
);
401 f2fs_msg(sb
, KERN_ERR
,
402 "Unrecognized mount option \"%s\" or missing value",
410 static struct inode
*f2fs_alloc_inode(struct super_block
*sb
)
412 struct f2fs_inode_info
*fi
;
414 fi
= kmem_cache_alloc(f2fs_inode_cachep
, GFP_F2FS_ZERO
);
418 init_once((void *) fi
);
420 /* Initialize f2fs-specific inode info */
421 fi
->vfs_inode
.i_version
= 1;
422 atomic_set(&fi
->dirty_pages
, 0);
423 fi
->i_current_depth
= 1;
425 init_rwsem(&fi
->i_sem
);
426 INIT_RADIX_TREE(&fi
->inmem_root
, GFP_NOFS
);
427 INIT_LIST_HEAD(&fi
->inmem_pages
);
428 mutex_init(&fi
->inmem_lock
);
430 set_inode_flag(fi
, FI_NEW_INODE
);
432 if (test_opt(F2FS_SB(sb
), INLINE_XATTR
))
433 set_inode_flag(fi
, FI_INLINE_XATTR
);
435 /* Will be used by directory only */
436 fi
->i_dir_level
= F2FS_SB(sb
)->dir_level
;
438 #ifdef CONFIG_F2FS_FS_ENCRYPTION
439 fi
->i_crypt_info
= NULL
;
441 return &fi
->vfs_inode
;
444 static int f2fs_drop_inode(struct inode
*inode
)
447 * This is to avoid a deadlock condition like below.
448 * writeback_single_inode(inode)
449 * - f2fs_write_data_page
450 * - f2fs_gc -> iput -> evict
451 * - inode_wait_for_writeback(inode)
453 if (!inode_unhashed(inode
) && inode
->i_state
& I_SYNC
) {
454 if (!inode
->i_nlink
&& !is_bad_inode(inode
)) {
455 /* to avoid evict_inode call simultaneously */
456 atomic_inc(&inode
->i_count
);
457 spin_unlock(&inode
->i_lock
);
459 /* some remained atomic pages should discarded */
460 if (f2fs_is_atomic_file(inode
))
461 commit_inmem_pages(inode
, true);
463 /* should remain fi->extent_tree for writepage */
464 f2fs_destroy_extent_node(inode
);
466 sb_start_intwrite(inode
->i_sb
);
467 i_size_write(inode
, 0);
469 if (F2FS_HAS_BLOCKS(inode
))
470 f2fs_truncate(inode
);
472 sb_end_intwrite(inode
->i_sb
);
474 #ifdef CONFIG_F2FS_FS_ENCRYPTION
475 if (F2FS_I(inode
)->i_crypt_info
)
476 f2fs_free_encryption_info(inode
,
477 F2FS_I(inode
)->i_crypt_info
);
479 spin_lock(&inode
->i_lock
);
480 atomic_dec(&inode
->i_count
);
484 return generic_drop_inode(inode
);
488 * f2fs_dirty_inode() is called from __mark_inode_dirty()
490 * We should call set_dirty_inode to write the dirty inode through write_inode.
492 static void f2fs_dirty_inode(struct inode
*inode
, int flags
)
494 set_inode_flag(F2FS_I(inode
), FI_DIRTY_INODE
);
497 static void f2fs_i_callback(struct rcu_head
*head
)
499 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
500 kmem_cache_free(f2fs_inode_cachep
, F2FS_I(inode
));
503 static void f2fs_destroy_inode(struct inode
*inode
)
505 call_rcu(&inode
->i_rcu
, f2fs_i_callback
);
508 static void f2fs_put_super(struct super_block
*sb
)
510 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
513 remove_proc_entry("segment_info", sbi
->s_proc
);
514 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
516 kobject_del(&sbi
->s_kobj
);
520 /* prevent remaining shrinker jobs */
521 mutex_lock(&sbi
->umount_mutex
);
524 * We don't need to do checkpoint when superblock is clean.
525 * But, the previous checkpoint was not done by umount, it needs to do
526 * clean checkpoint again.
528 if (is_sbi_flag_set(sbi
, SBI_IS_DIRTY
) ||
529 !is_set_ckpt_flags(F2FS_CKPT(sbi
), CP_UMOUNT_FLAG
)) {
530 struct cp_control cpc
= {
533 write_checkpoint(sbi
, &cpc
);
536 /* write_checkpoint can update stat informaion */
537 f2fs_destroy_stats(sbi
);
540 * normally superblock is clean, so we need to release this.
541 * In addition, EIO will skip do checkpoint, we need this as well.
543 release_dirty_inode(sbi
);
544 release_discard_addrs(sbi
);
546 f2fs_leave_shrinker(sbi
);
547 mutex_unlock(&sbi
->umount_mutex
);
549 iput(sbi
->node_inode
);
550 iput(sbi
->meta_inode
);
552 /* destroy f2fs internal modules */
553 destroy_node_manager(sbi
);
554 destroy_segment_manager(sbi
);
557 kobject_put(&sbi
->s_kobj
);
558 wait_for_completion(&sbi
->s_kobj_unregister
);
560 sb
->s_fs_info
= NULL
;
561 brelse(sbi
->raw_super_buf
);
565 int f2fs_sync_fs(struct super_block
*sb
, int sync
)
567 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
569 trace_f2fs_sync_fs(sb
, sync
);
572 struct cp_control cpc
;
574 cpc
.reason
= __get_cp_reason(sbi
);
576 mutex_lock(&sbi
->gc_mutex
);
577 write_checkpoint(sbi
, &cpc
);
578 mutex_unlock(&sbi
->gc_mutex
);
580 f2fs_balance_fs(sbi
);
582 f2fs_trace_ios(NULL
, 1);
587 static int f2fs_freeze(struct super_block
*sb
)
591 if (f2fs_readonly(sb
))
594 err
= f2fs_sync_fs(sb
, 1);
598 static int f2fs_unfreeze(struct super_block
*sb
)
603 static int f2fs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
605 struct super_block
*sb
= dentry
->d_sb
;
606 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
607 u64 id
= huge_encode_dev(sb
->s_bdev
->bd_dev
);
608 block_t total_count
, user_block_count
, start_count
, ovp_count
;
610 total_count
= le64_to_cpu(sbi
->raw_super
->block_count
);
611 user_block_count
= sbi
->user_block_count
;
612 start_count
= le32_to_cpu(sbi
->raw_super
->segment0_blkaddr
);
613 ovp_count
= SM_I(sbi
)->ovp_segments
<< sbi
->log_blocks_per_seg
;
614 buf
->f_type
= F2FS_SUPER_MAGIC
;
615 buf
->f_bsize
= sbi
->blocksize
;
617 buf
->f_blocks
= total_count
- start_count
;
618 buf
->f_bfree
= buf
->f_blocks
- valid_user_blocks(sbi
) - ovp_count
;
619 buf
->f_bavail
= user_block_count
- valid_user_blocks(sbi
);
621 buf
->f_files
= sbi
->total_node_count
- F2FS_RESERVED_NODE_NUM
;
622 buf
->f_ffree
= buf
->f_files
- valid_inode_count(sbi
);
624 buf
->f_namelen
= F2FS_NAME_LEN
;
625 buf
->f_fsid
.val
[0] = (u32
)id
;
626 buf
->f_fsid
.val
[1] = (u32
)(id
>> 32);
631 static int f2fs_show_options(struct seq_file
*seq
, struct dentry
*root
)
633 struct f2fs_sb_info
*sbi
= F2FS_SB(root
->d_sb
);
635 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, BG_GC
))
636 seq_printf(seq
, ",background_gc=%s", "on");
638 seq_printf(seq
, ",background_gc=%s", "off");
639 if (test_opt(sbi
, DISABLE_ROLL_FORWARD
))
640 seq_puts(seq
, ",disable_roll_forward");
641 if (test_opt(sbi
, DISCARD
))
642 seq_puts(seq
, ",discard");
643 if (test_opt(sbi
, NOHEAP
))
644 seq_puts(seq
, ",no_heap_alloc");
645 #ifdef CONFIG_F2FS_FS_XATTR
646 if (test_opt(sbi
, XATTR_USER
))
647 seq_puts(seq
, ",user_xattr");
649 seq_puts(seq
, ",nouser_xattr");
650 if (test_opt(sbi
, INLINE_XATTR
))
651 seq_puts(seq
, ",inline_xattr");
653 #ifdef CONFIG_F2FS_FS_POSIX_ACL
654 if (test_opt(sbi
, POSIX_ACL
))
655 seq_puts(seq
, ",acl");
657 seq_puts(seq
, ",noacl");
659 if (test_opt(sbi
, DISABLE_EXT_IDENTIFY
))
660 seq_puts(seq
, ",disable_ext_identify");
661 if (test_opt(sbi
, INLINE_DATA
))
662 seq_puts(seq
, ",inline_data");
664 seq_puts(seq
, ",noinline_data");
665 if (test_opt(sbi
, INLINE_DENTRY
))
666 seq_puts(seq
, ",inline_dentry");
667 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, FLUSH_MERGE
))
668 seq_puts(seq
, ",flush_merge");
669 if (test_opt(sbi
, NOBARRIER
))
670 seq_puts(seq
, ",nobarrier");
671 if (test_opt(sbi
, FASTBOOT
))
672 seq_puts(seq
, ",fastboot");
673 if (test_opt(sbi
, EXTENT_CACHE
))
674 seq_puts(seq
, ",extent_cache");
676 seq_puts(seq
, ",noextent_cache");
677 seq_printf(seq
, ",active_logs=%u", sbi
->active_logs
);
682 static int segment_info_seq_show(struct seq_file
*seq
, void *offset
)
684 struct super_block
*sb
= seq
->private;
685 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
686 unsigned int total_segs
=
687 le32_to_cpu(sbi
->raw_super
->segment_count_main
);
690 seq_puts(seq
, "format: segment_type|valid_blocks\n"
691 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
693 for (i
= 0; i
< total_segs
; i
++) {
694 struct seg_entry
*se
= get_seg_entry(sbi
, i
);
697 seq_printf(seq
, "%-5d", i
);
698 seq_printf(seq
, "%d|%-3u", se
->type
,
699 get_valid_blocks(sbi
, i
, 1));
700 if ((i
% 10) == 9 || i
== (total_segs
- 1))
709 static int segment_info_open_fs(struct inode
*inode
, struct file
*file
)
711 return single_open(file
, segment_info_seq_show
, PDE_DATA(inode
));
714 static const struct file_operations f2fs_seq_segment_info_fops
= {
715 .owner
= THIS_MODULE
,
716 .open
= segment_info_open_fs
,
719 .release
= single_release
,
722 static void default_options(struct f2fs_sb_info
*sbi
)
724 /* init some FS parameters */
725 sbi
->active_logs
= NR_CURSEG_TYPE
;
728 set_opt(sbi
, INLINE_DATA
);
729 set_opt(sbi
, EXTENT_CACHE
);
731 #ifdef CONFIG_F2FS_FS_XATTR
732 set_opt(sbi
, XATTR_USER
);
734 #ifdef CONFIG_F2FS_FS_POSIX_ACL
735 set_opt(sbi
, POSIX_ACL
);
739 static int f2fs_remount(struct super_block
*sb
, int *flags
, char *data
)
741 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
742 struct f2fs_mount_info org_mount_opt
;
743 int err
, active_logs
;
744 bool need_restart_gc
= false;
745 bool need_stop_gc
= false;
750 * Save the old mount options in case we
751 * need to restore them.
753 org_mount_opt
= sbi
->mount_opt
;
754 active_logs
= sbi
->active_logs
;
756 sbi
->mount_opt
.opt
= 0;
757 default_options(sbi
);
759 /* parse mount options */
760 err
= parse_options(sb
, data
);
765 * Previous and new state of filesystem is RO,
766 * so skip checking GC and FLUSH_MERGE conditions.
768 if (f2fs_readonly(sb
) && (*flags
& MS_RDONLY
))
772 * We stop the GC thread if FS is mounted as RO
773 * or if background_gc = off is passed in mount
774 * option. Also sync the filesystem.
776 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, BG_GC
)) {
777 if (sbi
->gc_thread
) {
780 need_restart_gc
= true;
782 } else if (!sbi
->gc_thread
) {
783 err
= start_gc_thread(sbi
);
790 * We stop issue flush thread if FS is mounted as RO
791 * or if flush_merge is not passed in mount option.
793 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, FLUSH_MERGE
)) {
794 destroy_flush_cmd_control(sbi
);
795 } else if (!SM_I(sbi
)->cmd_control_info
) {
796 err
= create_flush_cmd_control(sbi
);
801 /* Update the POSIXACL Flag */
802 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
803 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
806 if (need_restart_gc
) {
807 if (start_gc_thread(sbi
))
808 f2fs_msg(sbi
->sb
, KERN_WARNING
,
809 "background gc thread has stopped");
810 } else if (need_stop_gc
) {
814 sbi
->mount_opt
= org_mount_opt
;
815 sbi
->active_logs
= active_logs
;
819 static struct super_operations f2fs_sops
= {
820 .alloc_inode
= f2fs_alloc_inode
,
821 .drop_inode
= f2fs_drop_inode
,
822 .destroy_inode
= f2fs_destroy_inode
,
823 .write_inode
= f2fs_write_inode
,
824 .dirty_inode
= f2fs_dirty_inode
,
825 .show_options
= f2fs_show_options
,
826 .evict_inode
= f2fs_evict_inode
,
827 .put_super
= f2fs_put_super
,
828 .sync_fs
= f2fs_sync_fs
,
829 .freeze_fs
= f2fs_freeze
,
830 .unfreeze_fs
= f2fs_unfreeze
,
831 .statfs
= f2fs_statfs
,
832 .remount_fs
= f2fs_remount
,
835 static struct inode
*f2fs_nfs_get_inode(struct super_block
*sb
,
836 u64 ino
, u32 generation
)
838 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
841 if (check_nid_range(sbi
, ino
))
842 return ERR_PTR(-ESTALE
);
845 * f2fs_iget isn't quite right if the inode is currently unallocated!
846 * However f2fs_iget currently does appropriate checks to handle stale
847 * inodes so everything is OK.
849 inode
= f2fs_iget(sb
, ino
);
851 return ERR_CAST(inode
);
852 if (unlikely(generation
&& inode
->i_generation
!= generation
)) {
853 /* we didn't find the right inode.. */
855 return ERR_PTR(-ESTALE
);
860 static struct dentry
*f2fs_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
861 int fh_len
, int fh_type
)
863 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
867 static struct dentry
*f2fs_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
868 int fh_len
, int fh_type
)
870 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
874 static const struct export_operations f2fs_export_ops
= {
875 .fh_to_dentry
= f2fs_fh_to_dentry
,
876 .fh_to_parent
= f2fs_fh_to_parent
,
877 .get_parent
= f2fs_get_parent
,
880 static loff_t
max_file_size(unsigned bits
)
882 loff_t result
= (DEF_ADDRS_PER_INODE
- F2FS_INLINE_XATTR_ADDRS
);
883 loff_t leaf_count
= ADDRS_PER_BLOCK
;
885 /* two direct node blocks */
886 result
+= (leaf_count
* 2);
888 /* two indirect node blocks */
889 leaf_count
*= NIDS_PER_BLOCK
;
890 result
+= (leaf_count
* 2);
892 /* one double indirect node block */
893 leaf_count
*= NIDS_PER_BLOCK
;
894 result
+= leaf_count
;
900 static int sanity_check_raw_super(struct super_block
*sb
,
901 struct f2fs_super_block
*raw_super
)
903 unsigned int blocksize
;
905 if (F2FS_SUPER_MAGIC
!= le32_to_cpu(raw_super
->magic
)) {
906 f2fs_msg(sb
, KERN_INFO
,
907 "Magic Mismatch, valid(0x%x) - read(0x%x)",
908 F2FS_SUPER_MAGIC
, le32_to_cpu(raw_super
->magic
));
912 /* Currently, support only 4KB page cache size */
913 if (F2FS_BLKSIZE
!= PAGE_CACHE_SIZE
) {
914 f2fs_msg(sb
, KERN_INFO
,
915 "Invalid page_cache_size (%lu), supports only 4KB\n",
920 /* Currently, support only 4KB block size */
921 blocksize
= 1 << le32_to_cpu(raw_super
->log_blocksize
);
922 if (blocksize
!= F2FS_BLKSIZE
) {
923 f2fs_msg(sb
, KERN_INFO
,
924 "Invalid blocksize (%u), supports only 4KB\n",
929 /* Currently, support 512/1024/2048/4096 bytes sector size */
930 if (le32_to_cpu(raw_super
->log_sectorsize
) >
931 F2FS_MAX_LOG_SECTOR_SIZE
||
932 le32_to_cpu(raw_super
->log_sectorsize
) <
933 F2FS_MIN_LOG_SECTOR_SIZE
) {
934 f2fs_msg(sb
, KERN_INFO
, "Invalid log sectorsize (%u)",
935 le32_to_cpu(raw_super
->log_sectorsize
));
938 if (le32_to_cpu(raw_super
->log_sectors_per_block
) +
939 le32_to_cpu(raw_super
->log_sectorsize
) !=
940 F2FS_MAX_LOG_SECTOR_SIZE
) {
941 f2fs_msg(sb
, KERN_INFO
,
942 "Invalid log sectors per block(%u) log sectorsize(%u)",
943 le32_to_cpu(raw_super
->log_sectors_per_block
),
944 le32_to_cpu(raw_super
->log_sectorsize
));
950 static int sanity_check_ckpt(struct f2fs_sb_info
*sbi
)
952 unsigned int total
, fsmeta
;
953 struct f2fs_super_block
*raw_super
= F2FS_RAW_SUPER(sbi
);
954 struct f2fs_checkpoint
*ckpt
= F2FS_CKPT(sbi
);
956 total
= le32_to_cpu(raw_super
->segment_count
);
957 fsmeta
= le32_to_cpu(raw_super
->segment_count_ckpt
);
958 fsmeta
+= le32_to_cpu(raw_super
->segment_count_sit
);
959 fsmeta
+= le32_to_cpu(raw_super
->segment_count_nat
);
960 fsmeta
+= le32_to_cpu(ckpt
->rsvd_segment_count
);
961 fsmeta
+= le32_to_cpu(raw_super
->segment_count_ssa
);
963 if (unlikely(fsmeta
>= total
))
966 if (unlikely(f2fs_cp_error(sbi
))) {
967 f2fs_msg(sbi
->sb
, KERN_ERR
, "A bug case: need to run fsck");
973 static void init_sb_info(struct f2fs_sb_info
*sbi
)
975 struct f2fs_super_block
*raw_super
= sbi
->raw_super
;
978 sbi
->log_sectors_per_block
=
979 le32_to_cpu(raw_super
->log_sectors_per_block
);
980 sbi
->log_blocksize
= le32_to_cpu(raw_super
->log_blocksize
);
981 sbi
->blocksize
= 1 << sbi
->log_blocksize
;
982 sbi
->log_blocks_per_seg
= le32_to_cpu(raw_super
->log_blocks_per_seg
);
983 sbi
->blocks_per_seg
= 1 << sbi
->log_blocks_per_seg
;
984 sbi
->segs_per_sec
= le32_to_cpu(raw_super
->segs_per_sec
);
985 sbi
->secs_per_zone
= le32_to_cpu(raw_super
->secs_per_zone
);
986 sbi
->total_sections
= le32_to_cpu(raw_super
->section_count
);
987 sbi
->total_node_count
=
988 (le32_to_cpu(raw_super
->segment_count_nat
) / 2)
989 * sbi
->blocks_per_seg
* NAT_ENTRY_PER_BLOCK
;
990 sbi
->root_ino_num
= le32_to_cpu(raw_super
->root_ino
);
991 sbi
->node_ino_num
= le32_to_cpu(raw_super
->node_ino
);
992 sbi
->meta_ino_num
= le32_to_cpu(raw_super
->meta_ino
);
993 sbi
->cur_victim_sec
= NULL_SECNO
;
994 sbi
->max_victim_search
= DEF_MAX_VICTIM_SEARCH
;
996 for (i
= 0; i
< NR_COUNT_TYPE
; i
++)
997 atomic_set(&sbi
->nr_pages
[i
], 0);
999 sbi
->dir_level
= DEF_DIR_LEVEL
;
1000 clear_sbi_flag(sbi
, SBI_NEED_FSCK
);
1002 INIT_LIST_HEAD(&sbi
->s_list
);
1003 mutex_init(&sbi
->umount_mutex
);
1007 * Read f2fs raw super block.
1008 * Because we have two copies of super block, so read the first one at first,
1009 * if the first one is invalid, move to read the second one.
1011 static int read_raw_super_block(struct super_block
*sb
,
1012 struct f2fs_super_block
**raw_super
,
1013 struct buffer_head
**raw_super_buf
,
1017 struct buffer_head
*buffer
;
1018 struct f2fs_super_block
*super
;
1022 buffer
= sb_bread(sb
, block
);
1025 f2fs_msg(sb
, KERN_ERR
, "Unable to read %dth superblock",
1036 super
= (struct f2fs_super_block
*)
1037 ((char *)(buffer
)->b_data
+ F2FS_SUPER_OFFSET
);
1039 /* sanity checking of raw super */
1040 if (sanity_check_raw_super(sb
, super
)) {
1043 f2fs_msg(sb
, KERN_ERR
,
1044 "Can't find valid F2FS filesystem in %dth superblock",
1056 *raw_super_buf
= buffer
;
1059 /* already have a valid superblock */
1063 /* check the validity of the second superblock */
1070 /* No valid superblock */
1077 int f2fs_commit_super(struct f2fs_sb_info
*sbi
, bool recover
)
1079 struct buffer_head
*sbh
= sbi
->raw_super_buf
;
1080 sector_t block
= sbh
->b_blocknr
;
1083 /* write back-up superblock first */
1084 sbh
->b_blocknr
= block
? 0 : 1;
1085 mark_buffer_dirty(sbh
);
1086 err
= sync_dirty_buffer(sbh
);
1088 sbh
->b_blocknr
= block
;
1090 /* if we are in recovery path, skip writing valid superblock */
1094 /* write current valid superblock */
1095 mark_buffer_dirty(sbh
);
1096 err
= sync_dirty_buffer(sbh
);
1098 clear_buffer_write_io_error(sbh
);
1099 set_buffer_uptodate(sbh
);
1103 static int f2fs_fill_super(struct super_block
*sb
, void *data
, int silent
)
1105 struct f2fs_sb_info
*sbi
;
1106 struct f2fs_super_block
*raw_super
;
1107 struct buffer_head
*raw_super_buf
;
1110 bool retry
= true, need_fsck
= false;
1111 char *options
= NULL
;
1117 raw_super_buf
= NULL
;
1120 /* allocate memory for f2fs-specific super block info */
1121 sbi
= kzalloc(sizeof(struct f2fs_sb_info
), GFP_KERNEL
);
1125 /* set a block size */
1126 if (unlikely(!sb_set_blocksize(sb
, F2FS_BLKSIZE
))) {
1127 f2fs_msg(sb
, KERN_ERR
, "unable to set blocksize");
1131 err
= read_raw_super_block(sb
, &raw_super
, &raw_super_buf
, &recovery
);
1135 sb
->s_fs_info
= sbi
;
1136 default_options(sbi
);
1137 /* parse mount options */
1138 options
= kstrdup((const char *)data
, GFP_KERNEL
);
1139 if (data
&& !options
) {
1144 err
= parse_options(sb
, options
);
1148 sb
->s_maxbytes
= max_file_size(le32_to_cpu(raw_super
->log_blocksize
));
1149 sb
->s_max_links
= F2FS_LINK_MAX
;
1150 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
1152 sb
->s_op
= &f2fs_sops
;
1153 sb
->s_xattr
= f2fs_xattr_handlers
;
1154 sb
->s_export_op
= &f2fs_export_ops
;
1155 sb
->s_magic
= F2FS_SUPER_MAGIC
;
1156 sb
->s_time_gran
= 1;
1157 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
1158 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
1159 memcpy(sb
->s_uuid
, raw_super
->uuid
, sizeof(raw_super
->uuid
));
1161 /* init f2fs-specific super block info */
1163 sbi
->raw_super
= raw_super
;
1164 sbi
->raw_super_buf
= raw_super_buf
;
1165 mutex_init(&sbi
->gc_mutex
);
1166 mutex_init(&sbi
->writepages
);
1167 mutex_init(&sbi
->cp_mutex
);
1168 init_rwsem(&sbi
->node_write
);
1169 clear_sbi_flag(sbi
, SBI_POR_DOING
);
1170 spin_lock_init(&sbi
->stat_lock
);
1172 init_rwsem(&sbi
->read_io
.io_rwsem
);
1173 sbi
->read_io
.sbi
= sbi
;
1174 sbi
->read_io
.bio
= NULL
;
1175 for (i
= 0; i
< NR_PAGE_TYPE
; i
++) {
1176 init_rwsem(&sbi
->write_io
[i
].io_rwsem
);
1177 sbi
->write_io
[i
].sbi
= sbi
;
1178 sbi
->write_io
[i
].bio
= NULL
;
1181 init_rwsem(&sbi
->cp_rwsem
);
1182 init_waitqueue_head(&sbi
->cp_wait
);
1185 /* get an inode for meta space */
1186 sbi
->meta_inode
= f2fs_iget(sb
, F2FS_META_INO(sbi
));
1187 if (IS_ERR(sbi
->meta_inode
)) {
1188 f2fs_msg(sb
, KERN_ERR
, "Failed to read F2FS meta data inode");
1189 err
= PTR_ERR(sbi
->meta_inode
);
1193 err
= get_valid_checkpoint(sbi
);
1195 f2fs_msg(sb
, KERN_ERR
, "Failed to get valid F2FS checkpoint");
1196 goto free_meta_inode
;
1199 /* sanity checking of checkpoint */
1201 if (sanity_check_ckpt(sbi
)) {
1202 f2fs_msg(sb
, KERN_ERR
, "Invalid F2FS checkpoint");
1206 sbi
->total_valid_node_count
=
1207 le32_to_cpu(sbi
->ckpt
->valid_node_count
);
1208 sbi
->total_valid_inode_count
=
1209 le32_to_cpu(sbi
->ckpt
->valid_inode_count
);
1210 sbi
->user_block_count
= le64_to_cpu(sbi
->ckpt
->user_block_count
);
1211 sbi
->total_valid_block_count
=
1212 le64_to_cpu(sbi
->ckpt
->valid_block_count
);
1213 sbi
->last_valid_block_count
= sbi
->total_valid_block_count
;
1214 sbi
->alloc_valid_block_count
= 0;
1215 INIT_LIST_HEAD(&sbi
->dir_inode_list
);
1216 spin_lock_init(&sbi
->dir_inode_lock
);
1218 init_extent_cache_info(sbi
);
1220 init_ino_entry_info(sbi
);
1222 /* setup f2fs internal modules */
1223 err
= build_segment_manager(sbi
);
1225 f2fs_msg(sb
, KERN_ERR
,
1226 "Failed to initialize F2FS segment manager");
1229 err
= build_node_manager(sbi
);
1231 f2fs_msg(sb
, KERN_ERR
,
1232 "Failed to initialize F2FS node manager");
1236 build_gc_manager(sbi
);
1238 /* get an inode for node space */
1239 sbi
->node_inode
= f2fs_iget(sb
, F2FS_NODE_INO(sbi
));
1240 if (IS_ERR(sbi
->node_inode
)) {
1241 f2fs_msg(sb
, KERN_ERR
, "Failed to read node inode");
1242 err
= PTR_ERR(sbi
->node_inode
);
1246 f2fs_join_shrinker(sbi
);
1248 /* if there are nt orphan nodes free them */
1249 recover_orphan_inodes(sbi
);
1251 /* read root inode and dentry */
1252 root
= f2fs_iget(sb
, F2FS_ROOT_INO(sbi
));
1254 f2fs_msg(sb
, KERN_ERR
, "Failed to read root inode");
1255 err
= PTR_ERR(root
);
1256 goto free_node_inode
;
1258 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
1261 goto free_node_inode
;
1264 sb
->s_root
= d_make_root(root
); /* allocate root dentry */
1267 goto free_root_inode
;
1270 err
= f2fs_build_stats(sbi
);
1272 goto free_root_inode
;
1275 sbi
->s_proc
= proc_mkdir(sb
->s_id
, f2fs_proc_root
);
1278 proc_create_data("segment_info", S_IRUGO
, sbi
->s_proc
,
1279 &f2fs_seq_segment_info_fops
, sb
);
1281 sbi
->s_kobj
.kset
= f2fs_kset
;
1282 init_completion(&sbi
->s_kobj_unregister
);
1283 err
= kobject_init_and_add(&sbi
->s_kobj
, &f2fs_ktype
, NULL
,
1288 /* recover fsynced data */
1289 if (!test_opt(sbi
, DISABLE_ROLL_FORWARD
)) {
1291 * mount should be failed, when device has readonly mode, and
1292 * previous checkpoint was not done by clean system shutdown.
1294 if (bdev_read_only(sb
->s_bdev
) &&
1295 !is_set_ckpt_flags(sbi
->ckpt
, CP_UMOUNT_FLAG
)) {
1301 set_sbi_flag(sbi
, SBI_NEED_FSCK
);
1303 err
= recover_fsync_data(sbi
);
1306 f2fs_msg(sb
, KERN_ERR
,
1307 "Cannot recover all fsync data errno=%ld", err
);
1313 * If filesystem is not mounted as read-only then
1314 * do start the gc_thread.
1316 if (test_opt(sbi
, BG_GC
) && !f2fs_readonly(sb
)) {
1317 /* After POR, we can run background GC thread.*/
1318 err
= start_gc_thread(sbi
);
1324 /* recover broken superblock */
1325 if (recovery
&& !f2fs_readonly(sb
) && !bdev_read_only(sb
->s_bdev
)) {
1326 f2fs_msg(sb
, KERN_INFO
, "Recover invalid superblock");
1327 f2fs_commit_super(sbi
, true);
1333 kobject_del(&sbi
->s_kobj
);
1336 remove_proc_entry("segment_info", sbi
->s_proc
);
1337 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
1339 f2fs_destroy_stats(sbi
);
1344 mutex_lock(&sbi
->umount_mutex
);
1345 f2fs_leave_shrinker(sbi
);
1346 iput(sbi
->node_inode
);
1347 mutex_unlock(&sbi
->umount_mutex
);
1349 destroy_node_manager(sbi
);
1351 destroy_segment_manager(sbi
);
1355 make_bad_inode(sbi
->meta_inode
);
1356 iput(sbi
->meta_inode
);
1360 brelse(raw_super_buf
);
1364 /* give only one another chance */
1367 shrink_dcache_sb(sb
);
1373 static struct dentry
*f2fs_mount(struct file_system_type
*fs_type
, int flags
,
1374 const char *dev_name
, void *data
)
1376 return mount_bdev(fs_type
, flags
, dev_name
, data
, f2fs_fill_super
);
1379 static void kill_f2fs_super(struct super_block
*sb
)
1382 set_sbi_flag(F2FS_SB(sb
), SBI_IS_CLOSE
);
1383 kill_block_super(sb
);
1386 static struct file_system_type f2fs_fs_type
= {
1387 .owner
= THIS_MODULE
,
1389 .mount
= f2fs_mount
,
1390 .kill_sb
= kill_f2fs_super
,
1391 .fs_flags
= FS_REQUIRES_DEV
,
1393 MODULE_ALIAS_FS("f2fs");
1395 static int __init
init_inodecache(void)
1397 f2fs_inode_cachep
= f2fs_kmem_cache_create("f2fs_inode_cache",
1398 sizeof(struct f2fs_inode_info
));
1399 if (!f2fs_inode_cachep
)
1404 static void destroy_inodecache(void)
1407 * Make sure all delayed rcu free inodes are flushed before we
1411 kmem_cache_destroy(f2fs_inode_cachep
);
1414 static int __init
init_f2fs_fs(void)
1418 f2fs_build_trace_ios();
1420 err
= init_inodecache();
1423 err
= create_node_manager_caches();
1425 goto free_inodecache
;
1426 err
= create_segment_manager_caches();
1428 goto free_node_manager_caches
;
1429 err
= create_checkpoint_caches();
1431 goto free_segment_manager_caches
;
1432 err
= create_extent_cache();
1434 goto free_checkpoint_caches
;
1435 f2fs_kset
= kset_create_and_add("f2fs", NULL
, fs_kobj
);
1438 goto free_extent_cache
;
1440 err
= f2fs_init_crypto();
1444 err
= register_shrinker(&f2fs_shrinker_info
);
1448 err
= register_filesystem(&f2fs_fs_type
);
1451 f2fs_create_root_stats();
1452 f2fs_proc_root
= proc_mkdir("fs/f2fs", NULL
);
1456 unregister_shrinker(&f2fs_shrinker_info
);
1460 kset_unregister(f2fs_kset
);
1462 destroy_extent_cache();
1463 free_checkpoint_caches
:
1464 destroy_checkpoint_caches();
1465 free_segment_manager_caches
:
1466 destroy_segment_manager_caches();
1467 free_node_manager_caches
:
1468 destroy_node_manager_caches();
1470 destroy_inodecache();
1475 static void __exit
exit_f2fs_fs(void)
1477 remove_proc_entry("fs/f2fs", NULL
);
1478 f2fs_destroy_root_stats();
1479 unregister_shrinker(&f2fs_shrinker_info
);
1480 unregister_filesystem(&f2fs_fs_type
);
1482 destroy_extent_cache();
1483 destroy_checkpoint_caches();
1484 destroy_segment_manager_caches();
1485 destroy_node_manager_caches();
1486 destroy_inodecache();
1487 kset_unregister(f2fs_kset
);
1488 f2fs_destroy_trace_ios();
1491 module_init(init_f2fs_fs
)
1492 module_exit(exit_f2fs_fs
)
1494 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1495 MODULE_DESCRIPTION("Flash Friendly File System");
1496 MODULE_LICENSE("GPL");