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/proc_fs.h>
16 #include <linux/buffer_head.h>
17 #include <linux/backing-dev.h>
18 #include <linux/kthread.h>
19 #include <linux/parser.h>
20 #include <linux/mount.h>
21 #include <linux/seq_file.h>
22 #include <linux/random.h>
23 #include <linux/exportfs.h>
24 #include <linux/blkdev.h>
25 #include <linux/f2fs_fs.h>
32 static struct kmem_cache
*f2fs_inode_cachep
;
35 Opt_gc_background_off
,
36 Opt_disable_roll_forward
,
42 Opt_disable_ext_identify
,
46 static match_table_t f2fs_tokens
= {
47 {Opt_gc_background_off
, "background_gc_off"},
48 {Opt_disable_roll_forward
, "disable_roll_forward"},
49 {Opt_discard
, "discard"},
50 {Opt_noheap
, "no_heap"},
51 {Opt_nouser_xattr
, "nouser_xattr"},
53 {Opt_active_logs
, "active_logs=%u"},
54 {Opt_disable_ext_identify
, "disable_ext_identify"},
58 void f2fs_msg(struct super_block
*sb
, const char *level
, const char *fmt
, ...)
66 printk("%sF2FS-fs (%s): %pV\n", level
, sb
->s_id
, &vaf
);
70 static void init_once(void *foo
)
72 struct f2fs_inode_info
*fi
= (struct f2fs_inode_info
*) foo
;
74 inode_init_once(&fi
->vfs_inode
);
77 static struct inode
*f2fs_alloc_inode(struct super_block
*sb
)
79 struct f2fs_inode_info
*fi
;
81 fi
= kmem_cache_alloc(f2fs_inode_cachep
, GFP_NOFS
| __GFP_ZERO
);
85 init_once((void *) fi
);
87 /* Initialize f2fs-specific inode info */
88 fi
->vfs_inode
.i_version
= 1;
89 atomic_set(&fi
->dirty_dents
, 0);
90 fi
->i_current_depth
= 1;
92 rwlock_init(&fi
->ext
.ext_lock
);
94 set_inode_flag(fi
, FI_NEW_INODE
);
96 return &fi
->vfs_inode
;
99 static void f2fs_i_callback(struct rcu_head
*head
)
101 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
102 kmem_cache_free(f2fs_inode_cachep
, F2FS_I(inode
));
105 static void f2fs_destroy_inode(struct inode
*inode
)
107 call_rcu(&inode
->i_rcu
, f2fs_i_callback
);
110 static void f2fs_put_super(struct super_block
*sb
)
112 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
114 f2fs_destroy_stats(sbi
);
117 write_checkpoint(sbi
, true);
119 iput(sbi
->node_inode
);
120 iput(sbi
->meta_inode
);
122 /* destroy f2fs internal modules */
123 destroy_node_manager(sbi
);
124 destroy_segment_manager(sbi
);
128 sb
->s_fs_info
= NULL
;
129 brelse(sbi
->raw_super_buf
);
133 int f2fs_sync_fs(struct super_block
*sb
, int sync
)
135 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
137 if (!sbi
->s_dirty
&& !get_pages(sbi
, F2FS_DIRTY_NODES
))
141 mutex_lock(&sbi
->gc_mutex
);
142 write_checkpoint(sbi
, false);
143 mutex_unlock(&sbi
->gc_mutex
);
145 f2fs_balance_fs(sbi
);
151 static int f2fs_freeze(struct super_block
*sb
)
155 if (sb
->s_flags
& MS_RDONLY
)
158 err
= f2fs_sync_fs(sb
, 1);
162 static int f2fs_unfreeze(struct super_block
*sb
)
167 static int f2fs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
169 struct super_block
*sb
= dentry
->d_sb
;
170 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
171 u64 id
= huge_encode_dev(sb
->s_bdev
->bd_dev
);
172 block_t total_count
, user_block_count
, start_count
, ovp_count
;
174 total_count
= le64_to_cpu(sbi
->raw_super
->block_count
);
175 user_block_count
= sbi
->user_block_count
;
176 start_count
= le32_to_cpu(sbi
->raw_super
->segment0_blkaddr
);
177 ovp_count
= SM_I(sbi
)->ovp_segments
<< sbi
->log_blocks_per_seg
;
178 buf
->f_type
= F2FS_SUPER_MAGIC
;
179 buf
->f_bsize
= sbi
->blocksize
;
181 buf
->f_blocks
= total_count
- start_count
;
182 buf
->f_bfree
= buf
->f_blocks
- valid_user_blocks(sbi
) - ovp_count
;
183 buf
->f_bavail
= user_block_count
- valid_user_blocks(sbi
);
185 buf
->f_files
= sbi
->total_node_count
;
186 buf
->f_ffree
= sbi
->total_node_count
- valid_inode_count(sbi
);
188 buf
->f_namelen
= F2FS_NAME_LEN
;
189 buf
->f_fsid
.val
[0] = (u32
)id
;
190 buf
->f_fsid
.val
[1] = (u32
)(id
>> 32);
195 static int f2fs_show_options(struct seq_file
*seq
, struct dentry
*root
)
197 struct f2fs_sb_info
*sbi
= F2FS_SB(root
->d_sb
);
199 if (test_opt(sbi
, BG_GC
))
200 seq_puts(seq
, ",background_gc_on");
202 seq_puts(seq
, ",background_gc_off");
203 if (test_opt(sbi
, DISABLE_ROLL_FORWARD
))
204 seq_puts(seq
, ",disable_roll_forward");
205 if (test_opt(sbi
, DISCARD
))
206 seq_puts(seq
, ",discard");
207 if (test_opt(sbi
, NOHEAP
))
208 seq_puts(seq
, ",no_heap_alloc");
209 #ifdef CONFIG_F2FS_FS_XATTR
210 if (test_opt(sbi
, XATTR_USER
))
211 seq_puts(seq
, ",user_xattr");
213 seq_puts(seq
, ",nouser_xattr");
215 #ifdef CONFIG_F2FS_FS_POSIX_ACL
216 if (test_opt(sbi
, POSIX_ACL
))
217 seq_puts(seq
, ",acl");
219 seq_puts(seq
, ",noacl");
221 if (test_opt(sbi
, DISABLE_EXT_IDENTIFY
))
222 seq_puts(seq
, ",disable_ext_identify");
224 seq_printf(seq
, ",active_logs=%u", sbi
->active_logs
);
229 static struct super_operations f2fs_sops
= {
230 .alloc_inode
= f2fs_alloc_inode
,
231 .destroy_inode
= f2fs_destroy_inode
,
232 .write_inode
= f2fs_write_inode
,
233 .show_options
= f2fs_show_options
,
234 .evict_inode
= f2fs_evict_inode
,
235 .put_super
= f2fs_put_super
,
236 .sync_fs
= f2fs_sync_fs
,
237 .freeze_fs
= f2fs_freeze
,
238 .unfreeze_fs
= f2fs_unfreeze
,
239 .statfs
= f2fs_statfs
,
242 static struct inode
*f2fs_nfs_get_inode(struct super_block
*sb
,
243 u64 ino
, u32 generation
)
245 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
248 if (ino
< F2FS_ROOT_INO(sbi
))
249 return ERR_PTR(-ESTALE
);
252 * f2fs_iget isn't quite right if the inode is currently unallocated!
253 * However f2fs_iget currently does appropriate checks to handle stale
254 * inodes so everything is OK.
256 inode
= f2fs_iget(sb
, ino
);
258 return ERR_CAST(inode
);
259 if (generation
&& inode
->i_generation
!= generation
) {
260 /* we didn't find the right inode.. */
262 return ERR_PTR(-ESTALE
);
267 static struct dentry
*f2fs_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
268 int fh_len
, int fh_type
)
270 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
274 static struct dentry
*f2fs_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
275 int fh_len
, int fh_type
)
277 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
281 static const struct export_operations f2fs_export_ops
= {
282 .fh_to_dentry
= f2fs_fh_to_dentry
,
283 .fh_to_parent
= f2fs_fh_to_parent
,
284 .get_parent
= f2fs_get_parent
,
287 static int parse_options(struct super_block
*sb
, struct f2fs_sb_info
*sbi
,
290 substring_t args
[MAX_OPT_ARGS
];
297 while ((p
= strsep(&options
, ",")) != NULL
) {
302 * Initialize args struct so we know whether arg was
303 * found; some options take optional arguments.
305 args
[0].to
= args
[0].from
= NULL
;
306 token
= match_token(p
, f2fs_tokens
, args
);
309 case Opt_gc_background_off
:
310 clear_opt(sbi
, BG_GC
);
312 case Opt_disable_roll_forward
:
313 set_opt(sbi
, DISABLE_ROLL_FORWARD
);
316 set_opt(sbi
, DISCARD
);
319 set_opt(sbi
, NOHEAP
);
321 #ifdef CONFIG_F2FS_FS_XATTR
322 case Opt_nouser_xattr
:
323 clear_opt(sbi
, XATTR_USER
);
326 case Opt_nouser_xattr
:
327 f2fs_msg(sb
, KERN_INFO
,
328 "nouser_xattr options not supported");
331 #ifdef CONFIG_F2FS_FS_POSIX_ACL
333 clear_opt(sbi
, POSIX_ACL
);
337 f2fs_msg(sb
, KERN_INFO
, "noacl options not supported");
340 case Opt_active_logs
:
341 if (args
->from
&& match_int(args
, &arg
))
343 if (arg
!= 2 && arg
!= 4 && arg
!= NR_CURSEG_TYPE
)
345 sbi
->active_logs
= arg
;
347 case Opt_disable_ext_identify
:
348 set_opt(sbi
, DISABLE_EXT_IDENTIFY
);
351 f2fs_msg(sb
, KERN_ERR
,
352 "Unrecognized mount option \"%s\" or missing value",
360 static loff_t
max_file_size(unsigned bits
)
362 loff_t result
= ADDRS_PER_INODE
;
363 loff_t leaf_count
= ADDRS_PER_BLOCK
;
365 /* two direct node blocks */
366 result
+= (leaf_count
* 2);
368 /* two indirect node blocks */
369 leaf_count
*= NIDS_PER_BLOCK
;
370 result
+= (leaf_count
* 2);
372 /* one double indirect node block */
373 leaf_count
*= NIDS_PER_BLOCK
;
374 result
+= leaf_count
;
380 static int sanity_check_raw_super(struct super_block
*sb
,
381 struct f2fs_super_block
*raw_super
)
383 unsigned int blocksize
;
385 if (F2FS_SUPER_MAGIC
!= le32_to_cpu(raw_super
->magic
)) {
386 f2fs_msg(sb
, KERN_INFO
,
387 "Magic Mismatch, valid(0x%x) - read(0x%x)",
388 F2FS_SUPER_MAGIC
, le32_to_cpu(raw_super
->magic
));
392 /* Currently, support only 4KB page cache size */
393 if (F2FS_BLKSIZE
!= PAGE_CACHE_SIZE
) {
394 f2fs_msg(sb
, KERN_INFO
,
395 "Invalid page_cache_size (%lu), supports only 4KB\n",
400 /* Currently, support only 4KB block size */
401 blocksize
= 1 << le32_to_cpu(raw_super
->log_blocksize
);
402 if (blocksize
!= F2FS_BLKSIZE
) {
403 f2fs_msg(sb
, KERN_INFO
,
404 "Invalid blocksize (%u), supports only 4KB\n",
409 if (le32_to_cpu(raw_super
->log_sectorsize
) !=
410 F2FS_LOG_SECTOR_SIZE
) {
411 f2fs_msg(sb
, KERN_INFO
, "Invalid log sectorsize");
414 if (le32_to_cpu(raw_super
->log_sectors_per_block
) !=
415 F2FS_LOG_SECTORS_PER_BLOCK
) {
416 f2fs_msg(sb
, KERN_INFO
, "Invalid log sectors per block");
422 static int sanity_check_ckpt(struct f2fs_sb_info
*sbi
)
424 unsigned int total
, fsmeta
;
425 struct f2fs_super_block
*raw_super
= F2FS_RAW_SUPER(sbi
);
426 struct f2fs_checkpoint
*ckpt
= F2FS_CKPT(sbi
);
428 total
= le32_to_cpu(raw_super
->segment_count
);
429 fsmeta
= le32_to_cpu(raw_super
->segment_count_ckpt
);
430 fsmeta
+= le32_to_cpu(raw_super
->segment_count_sit
);
431 fsmeta
+= le32_to_cpu(raw_super
->segment_count_nat
);
432 fsmeta
+= le32_to_cpu(ckpt
->rsvd_segment_count
);
433 fsmeta
+= le32_to_cpu(raw_super
->segment_count_ssa
);
438 if (is_set_ckpt_flags(ckpt
, CP_ERROR_FLAG
)) {
439 f2fs_msg(sbi
->sb
, KERN_ERR
, "A bug case: need to run fsck");
445 static void init_sb_info(struct f2fs_sb_info
*sbi
)
447 struct f2fs_super_block
*raw_super
= sbi
->raw_super
;
450 sbi
->log_sectors_per_block
=
451 le32_to_cpu(raw_super
->log_sectors_per_block
);
452 sbi
->log_blocksize
= le32_to_cpu(raw_super
->log_blocksize
);
453 sbi
->blocksize
= 1 << sbi
->log_blocksize
;
454 sbi
->log_blocks_per_seg
= le32_to_cpu(raw_super
->log_blocks_per_seg
);
455 sbi
->blocks_per_seg
= 1 << sbi
->log_blocks_per_seg
;
456 sbi
->segs_per_sec
= le32_to_cpu(raw_super
->segs_per_sec
);
457 sbi
->secs_per_zone
= le32_to_cpu(raw_super
->secs_per_zone
);
458 sbi
->total_sections
= le32_to_cpu(raw_super
->section_count
);
459 sbi
->total_node_count
=
460 (le32_to_cpu(raw_super
->segment_count_nat
) / 2)
461 * sbi
->blocks_per_seg
* NAT_ENTRY_PER_BLOCK
;
462 sbi
->root_ino_num
= le32_to_cpu(raw_super
->root_ino
);
463 sbi
->node_ino_num
= le32_to_cpu(raw_super
->node_ino
);
464 sbi
->meta_ino_num
= le32_to_cpu(raw_super
->meta_ino
);
465 sbi
->cur_victim_sec
= NULL_SECNO
;
467 for (i
= 0; i
< NR_COUNT_TYPE
; i
++)
468 atomic_set(&sbi
->nr_pages
[i
], 0);
471 static int validate_superblock(struct super_block
*sb
,
472 struct f2fs_super_block
**raw_super
,
473 struct buffer_head
**raw_super_buf
, sector_t block
)
475 const char *super
= (block
== 0 ? "first" : "second");
477 /* read f2fs raw super block */
478 *raw_super_buf
= sb_bread(sb
, block
);
479 if (!*raw_super_buf
) {
480 f2fs_msg(sb
, KERN_ERR
, "unable to read %s superblock",
485 *raw_super
= (struct f2fs_super_block
*)
486 ((char *)(*raw_super_buf
)->b_data
+ F2FS_SUPER_OFFSET
);
488 /* sanity checking of raw super */
489 if (!sanity_check_raw_super(sb
, *raw_super
))
492 f2fs_msg(sb
, KERN_ERR
, "Can't find a valid F2FS filesystem "
493 "in %s superblock", super
);
497 static int f2fs_fill_super(struct super_block
*sb
, void *data
, int silent
)
499 struct f2fs_sb_info
*sbi
;
500 struct f2fs_super_block
*raw_super
;
501 struct buffer_head
*raw_super_buf
;
506 /* allocate memory for f2fs-specific super block info */
507 sbi
= kzalloc(sizeof(struct f2fs_sb_info
), GFP_KERNEL
);
511 /* set a block size */
512 if (!sb_set_blocksize(sb
, F2FS_BLKSIZE
)) {
513 f2fs_msg(sb
, KERN_ERR
, "unable to set blocksize");
517 err
= validate_superblock(sb
, &raw_super
, &raw_super_buf
, 0);
519 brelse(raw_super_buf
);
520 /* check secondary superblock when primary failed */
521 err
= validate_superblock(sb
, &raw_super
, &raw_super_buf
, 1);
525 /* init some FS parameters */
526 sbi
->active_logs
= NR_CURSEG_TYPE
;
530 #ifdef CONFIG_F2FS_FS_XATTR
531 set_opt(sbi
, XATTR_USER
);
533 #ifdef CONFIG_F2FS_FS_POSIX_ACL
534 set_opt(sbi
, POSIX_ACL
);
536 /* parse mount options */
537 if (parse_options(sb
, sbi
, (char *)data
))
540 sb
->s_maxbytes
= max_file_size(le32_to_cpu(raw_super
->log_blocksize
));
541 sb
->s_max_links
= F2FS_LINK_MAX
;
542 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
544 sb
->s_op
= &f2fs_sops
;
545 sb
->s_xattr
= f2fs_xattr_handlers
;
546 sb
->s_export_op
= &f2fs_export_ops
;
547 sb
->s_magic
= F2FS_SUPER_MAGIC
;
550 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
551 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
552 memcpy(sb
->s_uuid
, raw_super
->uuid
, sizeof(raw_super
->uuid
));
554 /* init f2fs-specific super block info */
556 sbi
->raw_super
= raw_super
;
557 sbi
->raw_super_buf
= raw_super_buf
;
558 mutex_init(&sbi
->gc_mutex
);
559 mutex_init(&sbi
->writepages
);
560 mutex_init(&sbi
->cp_mutex
);
561 for (i
= 0; i
< NR_GLOBAL_LOCKS
; i
++)
562 mutex_init(&sbi
->fs_lock
[i
]);
563 mutex_init(&sbi
->node_write
);
565 spin_lock_init(&sbi
->stat_lock
);
566 init_rwsem(&sbi
->bio_sem
);
569 /* get an inode for meta space */
570 sbi
->meta_inode
= f2fs_iget(sb
, F2FS_META_INO(sbi
));
571 if (IS_ERR(sbi
->meta_inode
)) {
572 f2fs_msg(sb
, KERN_ERR
, "Failed to read F2FS meta data inode");
573 err
= PTR_ERR(sbi
->meta_inode
);
577 err
= get_valid_checkpoint(sbi
);
579 f2fs_msg(sb
, KERN_ERR
, "Failed to get valid F2FS checkpoint");
580 goto free_meta_inode
;
583 /* sanity checking of checkpoint */
585 if (sanity_check_ckpt(sbi
)) {
586 f2fs_msg(sb
, KERN_ERR
, "Invalid F2FS checkpoint");
590 sbi
->total_valid_node_count
=
591 le32_to_cpu(sbi
->ckpt
->valid_node_count
);
592 sbi
->total_valid_inode_count
=
593 le32_to_cpu(sbi
->ckpt
->valid_inode_count
);
594 sbi
->user_block_count
= le64_to_cpu(sbi
->ckpt
->user_block_count
);
595 sbi
->total_valid_block_count
=
596 le64_to_cpu(sbi
->ckpt
->valid_block_count
);
597 sbi
->last_valid_block_count
= sbi
->total_valid_block_count
;
598 sbi
->alloc_valid_block_count
= 0;
599 INIT_LIST_HEAD(&sbi
->dir_inode_list
);
600 spin_lock_init(&sbi
->dir_inode_lock
);
602 init_orphan_info(sbi
);
604 /* setup f2fs internal modules */
605 err
= build_segment_manager(sbi
);
607 f2fs_msg(sb
, KERN_ERR
,
608 "Failed to initialize F2FS segment manager");
611 err
= build_node_manager(sbi
);
613 f2fs_msg(sb
, KERN_ERR
,
614 "Failed to initialize F2FS node manager");
618 build_gc_manager(sbi
);
620 /* get an inode for node space */
621 sbi
->node_inode
= f2fs_iget(sb
, F2FS_NODE_INO(sbi
));
622 if (IS_ERR(sbi
->node_inode
)) {
623 f2fs_msg(sb
, KERN_ERR
, "Failed to read node inode");
624 err
= PTR_ERR(sbi
->node_inode
);
628 /* if there are nt orphan nodes free them */
630 if (recover_orphan_inodes(sbi
))
631 goto free_node_inode
;
633 /* read root inode and dentry */
634 root
= f2fs_iget(sb
, F2FS_ROOT_INO(sbi
));
636 f2fs_msg(sb
, KERN_ERR
, "Failed to read root inode");
638 goto free_node_inode
;
640 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
)
641 goto free_root_inode
;
643 sb
->s_root
= d_make_root(root
); /* allocate root dentry */
646 goto free_root_inode
;
649 /* recover fsynced data */
650 if (!test_opt(sbi
, DISABLE_ROLL_FORWARD
)) {
651 err
= recover_fsync_data(sbi
);
653 f2fs_msg(sb
, KERN_ERR
, "Failed to recover fsync data");
654 goto free_root_inode
;
658 /* After POR, we can run background GC thread */
659 err
= start_gc_thread(sbi
);
663 err
= f2fs_build_stats(sbi
);
667 if (test_opt(sbi
, DISCARD
)) {
668 struct request_queue
*q
= bdev_get_queue(sb
->s_bdev
);
669 if (!blk_queue_discard(q
))
670 f2fs_msg(sb
, KERN_WARNING
,
671 "mounting with \"discard\" option, but "
672 "the device does not support discard");
682 iput(sbi
->node_inode
);
684 destroy_node_manager(sbi
);
686 destroy_segment_manager(sbi
);
690 make_bad_inode(sbi
->meta_inode
);
691 iput(sbi
->meta_inode
);
693 brelse(raw_super_buf
);
699 static struct dentry
*f2fs_mount(struct file_system_type
*fs_type
, int flags
,
700 const char *dev_name
, void *data
)
702 return mount_bdev(fs_type
, flags
, dev_name
, data
, f2fs_fill_super
);
705 static struct file_system_type f2fs_fs_type
= {
706 .owner
= THIS_MODULE
,
709 .kill_sb
= kill_block_super
,
710 .fs_flags
= FS_REQUIRES_DEV
,
713 static int __init
init_inodecache(void)
715 f2fs_inode_cachep
= f2fs_kmem_cache_create("f2fs_inode_cache",
716 sizeof(struct f2fs_inode_info
), NULL
);
717 if (f2fs_inode_cachep
== NULL
)
722 static void destroy_inodecache(void)
725 * Make sure all delayed rcu free inodes are flushed before we
729 kmem_cache_destroy(f2fs_inode_cachep
);
732 static int __init
init_f2fs_fs(void)
736 err
= init_inodecache();
739 err
= create_node_manager_caches();
742 err
= create_gc_caches();
745 err
= create_checkpoint_caches();
748 err
= register_filesystem(&f2fs_fs_type
);
751 f2fs_create_root_stats();
756 static void __exit
exit_f2fs_fs(void)
758 f2fs_destroy_root_stats();
759 unregister_filesystem(&f2fs_fs_type
);
760 destroy_checkpoint_caches();
762 destroy_node_manager_caches();
763 destroy_inodecache();
766 module_init(init_f2fs_fs
)
767 module_exit(exit_f2fs_fs
)
769 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
770 MODULE_DESCRIPTION("Flash Friendly File System");
771 MODULE_LICENSE("GPL");