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
,
61 static match_table_t f2fs_tokens
= {
62 {Opt_gc_background
, "background_gc=%s"},
63 {Opt_disable_roll_forward
, "disable_roll_forward"},
64 {Opt_discard
, "discard"},
65 {Opt_noheap
, "no_heap"},
66 {Opt_user_xattr
, "user_xattr"},
67 {Opt_nouser_xattr
, "nouser_xattr"},
70 {Opt_active_logs
, "active_logs=%u"},
71 {Opt_disable_ext_identify
, "disable_ext_identify"},
72 {Opt_inline_xattr
, "inline_xattr"},
73 {Opt_inline_data
, "inline_data"},
74 {Opt_inline_dentry
, "inline_dentry"},
75 {Opt_flush_merge
, "flush_merge"},
76 {Opt_nobarrier
, "nobarrier"},
77 {Opt_fastboot
, "fastboot"},
81 /* Sysfs support for f2fs */
83 GC_THREAD
, /* struct f2fs_gc_thread */
84 SM_INFO
, /* struct f2fs_sm_info */
85 NM_INFO
, /* struct f2fs_nm_info */
86 F2FS_SBI
, /* struct f2fs_sb_info */
90 struct attribute attr
;
91 ssize_t (*show
)(struct f2fs_attr
*, struct f2fs_sb_info
*, char *);
92 ssize_t (*store
)(struct f2fs_attr
*, struct f2fs_sb_info
*,
93 const char *, size_t);
98 static unsigned char *__struct_ptr(struct f2fs_sb_info
*sbi
, int struct_type
)
100 if (struct_type
== GC_THREAD
)
101 return (unsigned char *)sbi
->gc_thread
;
102 else if (struct_type
== SM_INFO
)
103 return (unsigned char *)SM_I(sbi
);
104 else if (struct_type
== NM_INFO
)
105 return (unsigned char *)NM_I(sbi
);
106 else if (struct_type
== F2FS_SBI
)
107 return (unsigned char *)sbi
;
111 static ssize_t
f2fs_sbi_show(struct f2fs_attr
*a
,
112 struct f2fs_sb_info
*sbi
, char *buf
)
114 unsigned char *ptr
= NULL
;
117 ptr
= __struct_ptr(sbi
, a
->struct_type
);
121 ui
= (unsigned int *)(ptr
+ a
->offset
);
123 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
126 static ssize_t
f2fs_sbi_store(struct f2fs_attr
*a
,
127 struct f2fs_sb_info
*sbi
,
128 const char *buf
, size_t count
)
135 ptr
= __struct_ptr(sbi
, a
->struct_type
);
139 ui
= (unsigned int *)(ptr
+ a
->offset
);
141 ret
= kstrtoul(skip_spaces(buf
), 0, &t
);
148 static ssize_t
f2fs_attr_show(struct kobject
*kobj
,
149 struct attribute
*attr
, char *buf
)
151 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
153 struct f2fs_attr
*a
= container_of(attr
, struct f2fs_attr
, attr
);
155 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
158 static ssize_t
f2fs_attr_store(struct kobject
*kobj
, struct attribute
*attr
,
159 const char *buf
, size_t len
)
161 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
163 struct f2fs_attr
*a
= container_of(attr
, struct f2fs_attr
, attr
);
165 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
168 static void f2fs_sb_release(struct kobject
*kobj
)
170 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
172 complete(&sbi
->s_kobj_unregister
);
175 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
176 static struct f2fs_attr f2fs_attr_##_name = { \
177 .attr = {.name = __stringify(_name), .mode = _mode }, \
180 .struct_type = _struct_type, \
184 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
185 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
186 f2fs_sbi_show, f2fs_sbi_store, \
187 offsetof(struct struct_name, elname))
189 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_min_sleep_time
, min_sleep_time
);
190 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_max_sleep_time
, max_sleep_time
);
191 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_no_gc_sleep_time
, no_gc_sleep_time
);
192 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_idle
, gc_idle
);
193 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, reclaim_segments
, rec_prefree_segments
);
194 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, max_small_discards
, max_discards
);
195 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, ipu_policy
, ipu_policy
);
196 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, min_ipu_util
, min_ipu_util
);
197 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, min_fsync_blocks
, min_fsync_blocks
);
198 F2FS_RW_ATTR(NM_INFO
, f2fs_nm_info
, ram_thresh
, ram_thresh
);
199 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, max_victim_search
, max_victim_search
);
200 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, dir_level
, dir_level
);
202 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
203 static struct attribute
*f2fs_attrs
[] = {
204 ATTR_LIST(gc_min_sleep_time
),
205 ATTR_LIST(gc_max_sleep_time
),
206 ATTR_LIST(gc_no_gc_sleep_time
),
208 ATTR_LIST(reclaim_segments
),
209 ATTR_LIST(max_small_discards
),
210 ATTR_LIST(ipu_policy
),
211 ATTR_LIST(min_ipu_util
),
212 ATTR_LIST(min_fsync_blocks
),
213 ATTR_LIST(max_victim_search
),
214 ATTR_LIST(dir_level
),
215 ATTR_LIST(ram_thresh
),
219 static const struct sysfs_ops f2fs_attr_ops
= {
220 .show
= f2fs_attr_show
,
221 .store
= f2fs_attr_store
,
224 static struct kobj_type f2fs_ktype
= {
225 .default_attrs
= f2fs_attrs
,
226 .sysfs_ops
= &f2fs_attr_ops
,
227 .release
= f2fs_sb_release
,
230 void f2fs_msg(struct super_block
*sb
, const char *level
, const char *fmt
, ...)
232 struct va_format vaf
;
238 printk("%sF2FS-fs (%s): %pV\n", level
, sb
->s_id
, &vaf
);
242 static void init_once(void *foo
)
244 struct f2fs_inode_info
*fi
= (struct f2fs_inode_info
*) foo
;
246 inode_init_once(&fi
->vfs_inode
);
249 static int parse_options(struct super_block
*sb
, char *options
)
251 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
252 substring_t args
[MAX_OPT_ARGS
];
259 while ((p
= strsep(&options
, ",")) != NULL
) {
264 * Initialize args struct so we know whether arg was
265 * found; some options take optional arguments.
267 args
[0].to
= args
[0].from
= NULL
;
268 token
= match_token(p
, f2fs_tokens
, args
);
271 case Opt_gc_background
:
272 name
= match_strdup(&args
[0]);
276 if (strlen(name
) == 2 && !strncmp(name
, "on", 2))
278 else if (strlen(name
) == 3 && !strncmp(name
, "off", 3))
279 clear_opt(sbi
, BG_GC
);
286 case Opt_disable_roll_forward
:
287 set_opt(sbi
, DISABLE_ROLL_FORWARD
);
290 set_opt(sbi
, DISCARD
);
293 set_opt(sbi
, NOHEAP
);
295 #ifdef CONFIG_F2FS_FS_XATTR
297 set_opt(sbi
, XATTR_USER
);
299 case Opt_nouser_xattr
:
300 clear_opt(sbi
, XATTR_USER
);
302 case Opt_inline_xattr
:
303 set_opt(sbi
, INLINE_XATTR
);
307 f2fs_msg(sb
, KERN_INFO
,
308 "user_xattr options not supported");
310 case Opt_nouser_xattr
:
311 f2fs_msg(sb
, KERN_INFO
,
312 "nouser_xattr options not supported");
314 case Opt_inline_xattr
:
315 f2fs_msg(sb
, KERN_INFO
,
316 "inline_xattr options not supported");
319 #ifdef CONFIG_F2FS_FS_POSIX_ACL
321 set_opt(sbi
, POSIX_ACL
);
324 clear_opt(sbi
, POSIX_ACL
);
328 f2fs_msg(sb
, KERN_INFO
, "acl options not supported");
331 f2fs_msg(sb
, KERN_INFO
, "noacl options not supported");
334 case Opt_active_logs
:
335 if (args
->from
&& match_int(args
, &arg
))
337 if (arg
!= 2 && arg
!= 4 && arg
!= NR_CURSEG_TYPE
)
339 sbi
->active_logs
= arg
;
341 case Opt_disable_ext_identify
:
342 set_opt(sbi
, DISABLE_EXT_IDENTIFY
);
344 case Opt_inline_data
:
345 set_opt(sbi
, INLINE_DATA
);
347 case Opt_inline_dentry
:
348 set_opt(sbi
, INLINE_DENTRY
);
350 case Opt_flush_merge
:
351 set_opt(sbi
, FLUSH_MERGE
);
354 set_opt(sbi
, NOBARRIER
);
357 set_opt(sbi
, FASTBOOT
);
360 f2fs_msg(sb
, KERN_ERR
,
361 "Unrecognized mount option \"%s\" or missing value",
369 static struct inode
*f2fs_alloc_inode(struct super_block
*sb
)
371 struct f2fs_inode_info
*fi
;
373 fi
= kmem_cache_alloc(f2fs_inode_cachep
, GFP_F2FS_ZERO
);
377 init_once((void *) fi
);
379 /* Initialize f2fs-specific inode info */
380 fi
->vfs_inode
.i_version
= 1;
381 atomic_set(&fi
->dirty_pages
, 0);
382 fi
->i_current_depth
= 1;
384 rwlock_init(&fi
->ext
.ext_lock
);
385 init_rwsem(&fi
->i_sem
);
386 INIT_RADIX_TREE(&fi
->inmem_root
, GFP_NOFS
);
387 INIT_LIST_HEAD(&fi
->inmem_pages
);
388 mutex_init(&fi
->inmem_lock
);
390 set_inode_flag(fi
, FI_NEW_INODE
);
392 if (test_opt(F2FS_SB(sb
), INLINE_XATTR
))
393 set_inode_flag(fi
, FI_INLINE_XATTR
);
395 /* Will be used by directory only */
396 fi
->i_dir_level
= F2FS_SB(sb
)->dir_level
;
398 return &fi
->vfs_inode
;
401 static int f2fs_drop_inode(struct inode
*inode
)
404 * This is to avoid a deadlock condition like below.
405 * writeback_single_inode(inode)
406 * - f2fs_write_data_page
407 * - f2fs_gc -> iput -> evict
408 * - inode_wait_for_writeback(inode)
410 if (!inode_unhashed(inode
) && inode
->i_state
& I_SYNC
)
412 return generic_drop_inode(inode
);
416 * f2fs_dirty_inode() is called from __mark_inode_dirty()
418 * We should call set_dirty_inode to write the dirty inode through write_inode.
420 static void f2fs_dirty_inode(struct inode
*inode
, int flags
)
422 set_inode_flag(F2FS_I(inode
), FI_DIRTY_INODE
);
425 static void f2fs_i_callback(struct rcu_head
*head
)
427 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
428 kmem_cache_free(f2fs_inode_cachep
, F2FS_I(inode
));
431 static void f2fs_destroy_inode(struct inode
*inode
)
433 call_rcu(&inode
->i_rcu
, f2fs_i_callback
);
436 static void f2fs_put_super(struct super_block
*sb
)
438 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
441 remove_proc_entry("segment_info", sbi
->s_proc
);
442 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
444 kobject_del(&sbi
->s_kobj
);
446 f2fs_destroy_stats(sbi
);
449 /* We don't need to do checkpoint when it's clean */
451 struct cp_control cpc
= {
454 write_checkpoint(sbi
, &cpc
);
458 * normally superblock is clean, so we need to release this.
459 * In addition, EIO will skip do checkpoint, we need this as well.
461 release_dirty_inode(sbi
);
462 release_discard_addrs(sbi
);
464 iput(sbi
->node_inode
);
465 iput(sbi
->meta_inode
);
467 /* destroy f2fs internal modules */
468 destroy_node_manager(sbi
);
469 destroy_segment_manager(sbi
);
472 kobject_put(&sbi
->s_kobj
);
473 wait_for_completion(&sbi
->s_kobj_unregister
);
475 sb
->s_fs_info
= NULL
;
476 brelse(sbi
->raw_super_buf
);
480 int f2fs_sync_fs(struct super_block
*sb
, int sync
)
482 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
484 trace_f2fs_sync_fs(sb
, sync
);
487 struct cp_control cpc
;
489 cpc
.reason
= test_opt(sbi
, FASTBOOT
) ? CP_UMOUNT
: CP_SYNC
;
490 mutex_lock(&sbi
->gc_mutex
);
491 write_checkpoint(sbi
, &cpc
);
492 mutex_unlock(&sbi
->gc_mutex
);
494 f2fs_balance_fs(sbi
);
500 static int f2fs_freeze(struct super_block
*sb
)
504 if (f2fs_readonly(sb
))
507 err
= f2fs_sync_fs(sb
, 1);
511 static int f2fs_unfreeze(struct super_block
*sb
)
516 static int f2fs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
518 struct super_block
*sb
= dentry
->d_sb
;
519 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
520 u64 id
= huge_encode_dev(sb
->s_bdev
->bd_dev
);
521 block_t total_count
, user_block_count
, start_count
, ovp_count
;
523 total_count
= le64_to_cpu(sbi
->raw_super
->block_count
);
524 user_block_count
= sbi
->user_block_count
;
525 start_count
= le32_to_cpu(sbi
->raw_super
->segment0_blkaddr
);
526 ovp_count
= SM_I(sbi
)->ovp_segments
<< sbi
->log_blocks_per_seg
;
527 buf
->f_type
= F2FS_SUPER_MAGIC
;
528 buf
->f_bsize
= sbi
->blocksize
;
530 buf
->f_blocks
= total_count
- start_count
;
531 buf
->f_bfree
= buf
->f_blocks
- valid_user_blocks(sbi
) - ovp_count
;
532 buf
->f_bavail
= user_block_count
- valid_user_blocks(sbi
);
534 buf
->f_files
= sbi
->total_node_count
- F2FS_RESERVED_NODE_NUM
;
535 buf
->f_ffree
= buf
->f_files
- valid_inode_count(sbi
);
537 buf
->f_namelen
= F2FS_NAME_LEN
;
538 buf
->f_fsid
.val
[0] = (u32
)id
;
539 buf
->f_fsid
.val
[1] = (u32
)(id
>> 32);
544 static int f2fs_show_options(struct seq_file
*seq
, struct dentry
*root
)
546 struct f2fs_sb_info
*sbi
= F2FS_SB(root
->d_sb
);
548 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, BG_GC
))
549 seq_printf(seq
, ",background_gc=%s", "on");
551 seq_printf(seq
, ",background_gc=%s", "off");
552 if (test_opt(sbi
, DISABLE_ROLL_FORWARD
))
553 seq_puts(seq
, ",disable_roll_forward");
554 if (test_opt(sbi
, DISCARD
))
555 seq_puts(seq
, ",discard");
556 if (test_opt(sbi
, NOHEAP
))
557 seq_puts(seq
, ",no_heap_alloc");
558 #ifdef CONFIG_F2FS_FS_XATTR
559 if (test_opt(sbi
, XATTR_USER
))
560 seq_puts(seq
, ",user_xattr");
562 seq_puts(seq
, ",nouser_xattr");
563 if (test_opt(sbi
, INLINE_XATTR
))
564 seq_puts(seq
, ",inline_xattr");
566 #ifdef CONFIG_F2FS_FS_POSIX_ACL
567 if (test_opt(sbi
, POSIX_ACL
))
568 seq_puts(seq
, ",acl");
570 seq_puts(seq
, ",noacl");
572 if (test_opt(sbi
, DISABLE_EXT_IDENTIFY
))
573 seq_puts(seq
, ",disable_ext_identify");
574 if (test_opt(sbi
, INLINE_DATA
))
575 seq_puts(seq
, ",inline_data");
576 if (test_opt(sbi
, INLINE_DENTRY
))
577 seq_puts(seq
, ",inline_dentry");
578 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, FLUSH_MERGE
))
579 seq_puts(seq
, ",flush_merge");
580 if (test_opt(sbi
, NOBARRIER
))
581 seq_puts(seq
, ",nobarrier");
582 if (test_opt(sbi
, FASTBOOT
))
583 seq_puts(seq
, ",fastboot");
584 seq_printf(seq
, ",active_logs=%u", sbi
->active_logs
);
589 static int segment_info_seq_show(struct seq_file
*seq
, void *offset
)
591 struct super_block
*sb
= seq
->private;
592 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
593 unsigned int total_segs
=
594 le32_to_cpu(sbi
->raw_super
->segment_count_main
);
597 seq_puts(seq
, "format: segment_type|valid_blocks\n"
598 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
600 for (i
= 0; i
< total_segs
; i
++) {
601 struct seg_entry
*se
= get_seg_entry(sbi
, i
);
604 seq_printf(seq
, "%-5d", i
);
605 seq_printf(seq
, "%d|%-3u", se
->type
,
606 get_valid_blocks(sbi
, i
, 1));
607 if ((i
% 10) == 9 || i
== (total_segs
- 1))
616 static int segment_info_open_fs(struct inode
*inode
, struct file
*file
)
618 return single_open(file
, segment_info_seq_show
, PDE_DATA(inode
));
621 static const struct file_operations f2fs_seq_segment_info_fops
= {
622 .owner
= THIS_MODULE
,
623 .open
= segment_info_open_fs
,
626 .release
= single_release
,
629 static int f2fs_remount(struct super_block
*sb
, int *flags
, char *data
)
631 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
632 struct f2fs_mount_info org_mount_opt
;
633 int err
, active_logs
;
634 bool need_restart_gc
= false;
635 bool need_stop_gc
= false;
640 * Save the old mount options in case we
641 * need to restore them.
643 org_mount_opt
= sbi
->mount_opt
;
644 active_logs
= sbi
->active_logs
;
646 sbi
->mount_opt
.opt
= 0;
647 sbi
->active_logs
= NR_CURSEG_TYPE
;
649 /* parse mount options */
650 err
= parse_options(sb
, data
);
655 * Previous and new state of filesystem is RO,
656 * so skip checking GC and FLUSH_MERGE conditions.
658 if (f2fs_readonly(sb
) && (*flags
& MS_RDONLY
))
662 * We stop the GC thread if FS is mounted as RO
663 * or if background_gc = off is passed in mount
664 * option. Also sync the filesystem.
666 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, BG_GC
)) {
667 if (sbi
->gc_thread
) {
670 need_restart_gc
= true;
672 } else if (!sbi
->gc_thread
) {
673 err
= start_gc_thread(sbi
);
680 * We stop issue flush thread if FS is mounted as RO
681 * or if flush_merge is not passed in mount option.
683 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, FLUSH_MERGE
)) {
684 destroy_flush_cmd_control(sbi
);
685 } else if (!SM_I(sbi
)->cmd_control_info
) {
686 err
= create_flush_cmd_control(sbi
);
691 /* Update the POSIXACL Flag */
692 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
693 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
696 if (need_restart_gc
) {
697 if (start_gc_thread(sbi
))
698 f2fs_msg(sbi
->sb
, KERN_WARNING
,
699 "background gc thread has stopped");
700 } else if (need_stop_gc
) {
704 sbi
->mount_opt
= org_mount_opt
;
705 sbi
->active_logs
= active_logs
;
709 static struct super_operations f2fs_sops
= {
710 .alloc_inode
= f2fs_alloc_inode
,
711 .drop_inode
= f2fs_drop_inode
,
712 .destroy_inode
= f2fs_destroy_inode
,
713 .write_inode
= f2fs_write_inode
,
714 .dirty_inode
= f2fs_dirty_inode
,
715 .show_options
= f2fs_show_options
,
716 .evict_inode
= f2fs_evict_inode
,
717 .put_super
= f2fs_put_super
,
718 .sync_fs
= f2fs_sync_fs
,
719 .freeze_fs
= f2fs_freeze
,
720 .unfreeze_fs
= f2fs_unfreeze
,
721 .statfs
= f2fs_statfs
,
722 .remount_fs
= f2fs_remount
,
725 static struct inode
*f2fs_nfs_get_inode(struct super_block
*sb
,
726 u64 ino
, u32 generation
)
728 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
731 if (check_nid_range(sbi
, ino
))
732 return ERR_PTR(-ESTALE
);
735 * f2fs_iget isn't quite right if the inode is currently unallocated!
736 * However f2fs_iget currently does appropriate checks to handle stale
737 * inodes so everything is OK.
739 inode
= f2fs_iget(sb
, ino
);
741 return ERR_CAST(inode
);
742 if (unlikely(generation
&& inode
->i_generation
!= generation
)) {
743 /* we didn't find the right inode.. */
745 return ERR_PTR(-ESTALE
);
750 static struct dentry
*f2fs_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
751 int fh_len
, int fh_type
)
753 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
757 static struct dentry
*f2fs_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
758 int fh_len
, int fh_type
)
760 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
764 static const struct export_operations f2fs_export_ops
= {
765 .fh_to_dentry
= f2fs_fh_to_dentry
,
766 .fh_to_parent
= f2fs_fh_to_parent
,
767 .get_parent
= f2fs_get_parent
,
770 static loff_t
max_file_size(unsigned bits
)
772 loff_t result
= (DEF_ADDRS_PER_INODE
- F2FS_INLINE_XATTR_ADDRS
);
773 loff_t leaf_count
= ADDRS_PER_BLOCK
;
775 /* two direct node blocks */
776 result
+= (leaf_count
* 2);
778 /* two indirect node blocks */
779 leaf_count
*= NIDS_PER_BLOCK
;
780 result
+= (leaf_count
* 2);
782 /* one double indirect node block */
783 leaf_count
*= NIDS_PER_BLOCK
;
784 result
+= leaf_count
;
790 static int sanity_check_raw_super(struct super_block
*sb
,
791 struct f2fs_super_block
*raw_super
)
793 unsigned int blocksize
;
795 if (F2FS_SUPER_MAGIC
!= le32_to_cpu(raw_super
->magic
)) {
796 f2fs_msg(sb
, KERN_INFO
,
797 "Magic Mismatch, valid(0x%x) - read(0x%x)",
798 F2FS_SUPER_MAGIC
, le32_to_cpu(raw_super
->magic
));
802 /* Currently, support only 4KB page cache size */
803 if (F2FS_BLKSIZE
!= PAGE_CACHE_SIZE
) {
804 f2fs_msg(sb
, KERN_INFO
,
805 "Invalid page_cache_size (%lu), supports only 4KB\n",
810 /* Currently, support only 4KB block size */
811 blocksize
= 1 << le32_to_cpu(raw_super
->log_blocksize
);
812 if (blocksize
!= F2FS_BLKSIZE
) {
813 f2fs_msg(sb
, KERN_INFO
,
814 "Invalid blocksize (%u), supports only 4KB\n",
819 /* Currently, support 512/1024/2048/4096 bytes sector size */
820 if (le32_to_cpu(raw_super
->log_sectorsize
) >
821 F2FS_MAX_LOG_SECTOR_SIZE
||
822 le32_to_cpu(raw_super
->log_sectorsize
) <
823 F2FS_MIN_LOG_SECTOR_SIZE
) {
824 f2fs_msg(sb
, KERN_INFO
, "Invalid log sectorsize (%u)",
825 le32_to_cpu(raw_super
->log_sectorsize
));
828 if (le32_to_cpu(raw_super
->log_sectors_per_block
) +
829 le32_to_cpu(raw_super
->log_sectorsize
) !=
830 F2FS_MAX_LOG_SECTOR_SIZE
) {
831 f2fs_msg(sb
, KERN_INFO
,
832 "Invalid log sectors per block(%u) log sectorsize(%u)",
833 le32_to_cpu(raw_super
->log_sectors_per_block
),
834 le32_to_cpu(raw_super
->log_sectorsize
));
840 static int sanity_check_ckpt(struct f2fs_sb_info
*sbi
)
842 unsigned int total
, fsmeta
;
843 struct f2fs_super_block
*raw_super
= F2FS_RAW_SUPER(sbi
);
844 struct f2fs_checkpoint
*ckpt
= F2FS_CKPT(sbi
);
846 total
= le32_to_cpu(raw_super
->segment_count
);
847 fsmeta
= le32_to_cpu(raw_super
->segment_count_ckpt
);
848 fsmeta
+= le32_to_cpu(raw_super
->segment_count_sit
);
849 fsmeta
+= le32_to_cpu(raw_super
->segment_count_nat
);
850 fsmeta
+= le32_to_cpu(ckpt
->rsvd_segment_count
);
851 fsmeta
+= le32_to_cpu(raw_super
->segment_count_ssa
);
853 if (unlikely(fsmeta
>= total
))
856 if (unlikely(f2fs_cp_error(sbi
))) {
857 f2fs_msg(sbi
->sb
, KERN_ERR
, "A bug case: need to run fsck");
863 static void init_sb_info(struct f2fs_sb_info
*sbi
)
865 struct f2fs_super_block
*raw_super
= sbi
->raw_super
;
868 sbi
->log_sectors_per_block
=
869 le32_to_cpu(raw_super
->log_sectors_per_block
);
870 sbi
->log_blocksize
= le32_to_cpu(raw_super
->log_blocksize
);
871 sbi
->blocksize
= 1 << sbi
->log_blocksize
;
872 sbi
->log_blocks_per_seg
= le32_to_cpu(raw_super
->log_blocks_per_seg
);
873 sbi
->blocks_per_seg
= 1 << sbi
->log_blocks_per_seg
;
874 sbi
->segs_per_sec
= le32_to_cpu(raw_super
->segs_per_sec
);
875 sbi
->secs_per_zone
= le32_to_cpu(raw_super
->secs_per_zone
);
876 sbi
->total_sections
= le32_to_cpu(raw_super
->section_count
);
877 sbi
->total_node_count
=
878 (le32_to_cpu(raw_super
->segment_count_nat
) / 2)
879 * sbi
->blocks_per_seg
* NAT_ENTRY_PER_BLOCK
;
880 sbi
->root_ino_num
= le32_to_cpu(raw_super
->root_ino
);
881 sbi
->node_ino_num
= le32_to_cpu(raw_super
->node_ino
);
882 sbi
->meta_ino_num
= le32_to_cpu(raw_super
->meta_ino
);
883 sbi
->cur_victim_sec
= NULL_SECNO
;
884 sbi
->max_victim_search
= DEF_MAX_VICTIM_SEARCH
;
886 for (i
= 0; i
< NR_COUNT_TYPE
; i
++)
887 atomic_set(&sbi
->nr_pages
[i
], 0);
889 sbi
->dir_level
= DEF_DIR_LEVEL
;
890 sbi
->need_fsck
= false;
894 * Read f2fs raw super block.
895 * Because we have two copies of super block, so read the first one at first,
896 * if the first one is invalid, move to read the second one.
898 static int read_raw_super_block(struct super_block
*sb
,
899 struct f2fs_super_block
**raw_super
,
900 struct buffer_head
**raw_super_buf
)
905 *raw_super_buf
= sb_bread(sb
, block
);
906 if (!*raw_super_buf
) {
907 f2fs_msg(sb
, KERN_ERR
, "Unable to read %dth superblock",
917 *raw_super
= (struct f2fs_super_block
*)
918 ((char *)(*raw_super_buf
)->b_data
+ F2FS_SUPER_OFFSET
);
920 /* sanity checking of raw super */
921 if (sanity_check_raw_super(sb
, *raw_super
)) {
922 brelse(*raw_super_buf
);
923 f2fs_msg(sb
, KERN_ERR
,
924 "Can't find valid F2FS filesystem in %dth superblock",
937 static int f2fs_fill_super(struct super_block
*sb
, void *data
, int silent
)
939 struct f2fs_sb_info
*sbi
;
940 struct f2fs_super_block
*raw_super
= NULL
;
941 struct buffer_head
*raw_super_buf
;
948 /* allocate memory for f2fs-specific super block info */
949 sbi
= kzalloc(sizeof(struct f2fs_sb_info
), GFP_KERNEL
);
953 /* set a block size */
954 if (unlikely(!sb_set_blocksize(sb
, F2FS_BLKSIZE
))) {
955 f2fs_msg(sb
, KERN_ERR
, "unable to set blocksize");
959 err
= read_raw_super_block(sb
, &raw_super
, &raw_super_buf
);
964 /* init some FS parameters */
965 sbi
->active_logs
= NR_CURSEG_TYPE
;
969 #ifdef CONFIG_F2FS_FS_XATTR
970 set_opt(sbi
, XATTR_USER
);
972 #ifdef CONFIG_F2FS_FS_POSIX_ACL
973 set_opt(sbi
, POSIX_ACL
);
975 /* parse mount options */
976 err
= parse_options(sb
, (char *)data
);
980 sb
->s_maxbytes
= max_file_size(le32_to_cpu(raw_super
->log_blocksize
));
981 sb
->s_max_links
= F2FS_LINK_MAX
;
982 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
984 sb
->s_op
= &f2fs_sops
;
985 sb
->s_xattr
= f2fs_xattr_handlers
;
986 sb
->s_export_op
= &f2fs_export_ops
;
987 sb
->s_magic
= F2FS_SUPER_MAGIC
;
989 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
990 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
991 memcpy(sb
->s_uuid
, raw_super
->uuid
, sizeof(raw_super
->uuid
));
993 /* init f2fs-specific super block info */
995 sbi
->raw_super
= raw_super
;
996 sbi
->raw_super_buf
= raw_super_buf
;
997 mutex_init(&sbi
->gc_mutex
);
998 mutex_init(&sbi
->writepages
);
999 mutex_init(&sbi
->cp_mutex
);
1000 init_rwsem(&sbi
->node_write
);
1001 sbi
->por_doing
= false;
1002 spin_lock_init(&sbi
->stat_lock
);
1004 init_rwsem(&sbi
->read_io
.io_rwsem
);
1005 sbi
->read_io
.sbi
= sbi
;
1006 sbi
->read_io
.bio
= NULL
;
1007 for (i
= 0; i
< NR_PAGE_TYPE
; i
++) {
1008 init_rwsem(&sbi
->write_io
[i
].io_rwsem
);
1009 sbi
->write_io
[i
].sbi
= sbi
;
1010 sbi
->write_io
[i
].bio
= NULL
;
1013 init_rwsem(&sbi
->cp_rwsem
);
1014 init_waitqueue_head(&sbi
->cp_wait
);
1017 /* get an inode for meta space */
1018 sbi
->meta_inode
= f2fs_iget(sb
, F2FS_META_INO(sbi
));
1019 if (IS_ERR(sbi
->meta_inode
)) {
1020 f2fs_msg(sb
, KERN_ERR
, "Failed to read F2FS meta data inode");
1021 err
= PTR_ERR(sbi
->meta_inode
);
1025 err
= get_valid_checkpoint(sbi
);
1027 f2fs_msg(sb
, KERN_ERR
, "Failed to get valid F2FS checkpoint");
1028 goto free_meta_inode
;
1031 /* sanity checking of checkpoint */
1033 if (sanity_check_ckpt(sbi
)) {
1034 f2fs_msg(sb
, KERN_ERR
, "Invalid F2FS checkpoint");
1038 sbi
->total_valid_node_count
=
1039 le32_to_cpu(sbi
->ckpt
->valid_node_count
);
1040 sbi
->total_valid_inode_count
=
1041 le32_to_cpu(sbi
->ckpt
->valid_inode_count
);
1042 sbi
->user_block_count
= le64_to_cpu(sbi
->ckpt
->user_block_count
);
1043 sbi
->total_valid_block_count
=
1044 le64_to_cpu(sbi
->ckpt
->valid_block_count
);
1045 sbi
->last_valid_block_count
= sbi
->total_valid_block_count
;
1046 sbi
->alloc_valid_block_count
= 0;
1047 INIT_LIST_HEAD(&sbi
->dir_inode_list
);
1048 spin_lock_init(&sbi
->dir_inode_lock
);
1050 init_ino_entry_info(sbi
);
1052 /* setup f2fs internal modules */
1053 err
= build_segment_manager(sbi
);
1055 f2fs_msg(sb
, KERN_ERR
,
1056 "Failed to initialize F2FS segment manager");
1059 err
= build_node_manager(sbi
);
1061 f2fs_msg(sb
, KERN_ERR
,
1062 "Failed to initialize F2FS node manager");
1066 build_gc_manager(sbi
);
1068 /* get an inode for node space */
1069 sbi
->node_inode
= f2fs_iget(sb
, F2FS_NODE_INO(sbi
));
1070 if (IS_ERR(sbi
->node_inode
)) {
1071 f2fs_msg(sb
, KERN_ERR
, "Failed to read node inode");
1072 err
= PTR_ERR(sbi
->node_inode
);
1076 /* if there are nt orphan nodes free them */
1077 recover_orphan_inodes(sbi
);
1079 /* read root inode and dentry */
1080 root
= f2fs_iget(sb
, F2FS_ROOT_INO(sbi
));
1082 f2fs_msg(sb
, KERN_ERR
, "Failed to read root inode");
1083 err
= PTR_ERR(root
);
1084 goto free_node_inode
;
1086 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
1089 goto free_node_inode
;
1092 sb
->s_root
= d_make_root(root
); /* allocate root dentry */
1095 goto free_root_inode
;
1098 err
= f2fs_build_stats(sbi
);
1100 goto free_root_inode
;
1103 sbi
->s_proc
= proc_mkdir(sb
->s_id
, f2fs_proc_root
);
1106 proc_create_data("segment_info", S_IRUGO
, sbi
->s_proc
,
1107 &f2fs_seq_segment_info_fops
, sb
);
1109 if (test_opt(sbi
, DISCARD
)) {
1110 struct request_queue
*q
= bdev_get_queue(sb
->s_bdev
);
1111 if (!blk_queue_discard(q
))
1112 f2fs_msg(sb
, KERN_WARNING
,
1113 "mounting with \"discard\" option, but "
1114 "the device does not support discard");
1117 sbi
->s_kobj
.kset
= f2fs_kset
;
1118 init_completion(&sbi
->s_kobj_unregister
);
1119 err
= kobject_init_and_add(&sbi
->s_kobj
, &f2fs_ktype
, NULL
,
1125 sbi
->need_fsck
= true;
1127 /* recover fsynced data */
1128 if (!test_opt(sbi
, DISABLE_ROLL_FORWARD
)) {
1129 err
= recover_fsync_data(sbi
);
1131 f2fs_msg(sb
, KERN_ERR
,
1132 "Cannot recover all fsync data errno=%ld", err
);
1138 * If filesystem is not mounted as read-only then
1139 * do start the gc_thread.
1141 if (test_opt(sbi
, BG_GC
) && !f2fs_readonly(sb
)) {
1142 /* After POR, we can run background GC thread.*/
1143 err
= start_gc_thread(sbi
);
1150 kobject_del(&sbi
->s_kobj
);
1153 remove_proc_entry("segment_info", sbi
->s_proc
);
1154 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
1156 f2fs_destroy_stats(sbi
);
1161 iput(sbi
->node_inode
);
1163 destroy_node_manager(sbi
);
1165 destroy_segment_manager(sbi
);
1169 make_bad_inode(sbi
->meta_inode
);
1170 iput(sbi
->meta_inode
);
1172 brelse(raw_super_buf
);
1176 /* give only one another chance */
1179 shrink_dcache_sb(sb
);
1185 static struct dentry
*f2fs_mount(struct file_system_type
*fs_type
, int flags
,
1186 const char *dev_name
, void *data
)
1188 return mount_bdev(fs_type
, flags
, dev_name
, data
, f2fs_fill_super
);
1191 static struct file_system_type f2fs_fs_type
= {
1192 .owner
= THIS_MODULE
,
1194 .mount
= f2fs_mount
,
1195 .kill_sb
= kill_block_super
,
1196 .fs_flags
= FS_REQUIRES_DEV
,
1198 MODULE_ALIAS_FS("f2fs");
1200 static int __init
init_inodecache(void)
1202 f2fs_inode_cachep
= f2fs_kmem_cache_create("f2fs_inode_cache",
1203 sizeof(struct f2fs_inode_info
));
1204 if (!f2fs_inode_cachep
)
1209 static void destroy_inodecache(void)
1212 * Make sure all delayed rcu free inodes are flushed before we
1216 kmem_cache_destroy(f2fs_inode_cachep
);
1219 static int __init
init_f2fs_fs(void)
1223 err
= init_inodecache();
1226 err
= create_node_manager_caches();
1228 goto free_inodecache
;
1229 err
= create_segment_manager_caches();
1231 goto free_node_manager_caches
;
1232 err
= create_gc_caches();
1234 goto free_segment_manager_caches
;
1235 err
= create_checkpoint_caches();
1237 goto free_gc_caches
;
1238 f2fs_kset
= kset_create_and_add("f2fs", NULL
, fs_kobj
);
1241 goto free_checkpoint_caches
;
1243 err
= register_filesystem(&f2fs_fs_type
);
1246 f2fs_create_root_stats();
1247 f2fs_proc_root
= proc_mkdir("fs/f2fs", NULL
);
1251 kset_unregister(f2fs_kset
);
1252 free_checkpoint_caches
:
1253 destroy_checkpoint_caches();
1255 destroy_gc_caches();
1256 free_segment_manager_caches
:
1257 destroy_segment_manager_caches();
1258 free_node_manager_caches
:
1259 destroy_node_manager_caches();
1261 destroy_inodecache();
1266 static void __exit
exit_f2fs_fs(void)
1268 remove_proc_entry("fs/f2fs", NULL
);
1269 f2fs_destroy_root_stats();
1270 unregister_filesystem(&f2fs_fs_type
);
1271 destroy_checkpoint_caches();
1272 destroy_gc_caches();
1273 destroy_segment_manager_caches();
1274 destroy_node_manager_caches();
1275 destroy_inodecache();
1276 kset_unregister(f2fs_kset
);
1279 module_init(init_f2fs_fs
)
1280 module_exit(exit_f2fs_fs
)
1282 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1283 MODULE_DESCRIPTION("Flash Friendly File System");
1284 MODULE_LICENSE("GPL");