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/f2fs_fs.h>
30 static struct kmem_cache
*f2fs_inode_cachep
;
33 Opt_gc_background_off
,
34 Opt_disable_roll_forward
,
40 Opt_disable_ext_identify
,
44 static match_table_t f2fs_tokens
= {
45 {Opt_gc_background_off
, "background_gc_off"},
46 {Opt_disable_roll_forward
, "disable_roll_forward"},
47 {Opt_discard
, "discard"},
48 {Opt_noheap
, "no_heap"},
49 {Opt_nouser_xattr
, "nouser_xattr"},
51 {Opt_active_logs
, "active_logs=%u"},
52 {Opt_disable_ext_identify
, "disable_ext_identify"},
56 void f2fs_msg(struct super_block
*sb
, const char *level
, const char *fmt
, ...)
64 printk("%sF2FS-fs (%s): %pV\n", level
, sb
->s_id
, &vaf
);
68 static void init_once(void *foo
)
70 struct f2fs_inode_info
*fi
= (struct f2fs_inode_info
*) foo
;
72 inode_init_once(&fi
->vfs_inode
);
75 static struct inode
*f2fs_alloc_inode(struct super_block
*sb
)
77 struct f2fs_inode_info
*fi
;
79 fi
= kmem_cache_alloc(f2fs_inode_cachep
, GFP_NOFS
| __GFP_ZERO
);
83 init_once((void *) fi
);
85 /* Initilize f2fs-specific inode info */
86 fi
->vfs_inode
.i_version
= 1;
87 atomic_set(&fi
->dirty_dents
, 0);
88 fi
->i_current_depth
= 1;
90 rwlock_init(&fi
->ext
.ext_lock
);
92 set_inode_flag(fi
, FI_NEW_INODE
);
94 return &fi
->vfs_inode
;
97 static void f2fs_i_callback(struct rcu_head
*head
)
99 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
100 kmem_cache_free(f2fs_inode_cachep
, F2FS_I(inode
));
103 static void f2fs_destroy_inode(struct inode
*inode
)
105 call_rcu(&inode
->i_rcu
, f2fs_i_callback
);
108 static void f2fs_put_super(struct super_block
*sb
)
110 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
112 f2fs_destroy_stats(sbi
);
115 write_checkpoint(sbi
, true);
117 iput(sbi
->node_inode
);
118 iput(sbi
->meta_inode
);
120 /* destroy f2fs internal modules */
121 destroy_node_manager(sbi
);
122 destroy_segment_manager(sbi
);
126 sb
->s_fs_info
= NULL
;
127 brelse(sbi
->raw_super_buf
);
131 int f2fs_sync_fs(struct super_block
*sb
, int sync
)
133 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
135 if (!sbi
->s_dirty
&& !get_pages(sbi
, F2FS_DIRTY_NODES
))
139 write_checkpoint(sbi
, false);
141 f2fs_balance_fs(sbi
);
146 static int f2fs_freeze(struct super_block
*sb
)
150 if (sb
->s_flags
& MS_RDONLY
)
153 err
= f2fs_sync_fs(sb
, 1);
157 static int f2fs_unfreeze(struct super_block
*sb
)
162 static int f2fs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
164 struct super_block
*sb
= dentry
->d_sb
;
165 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
166 u64 id
= huge_encode_dev(sb
->s_bdev
->bd_dev
);
167 block_t total_count
, user_block_count
, start_count
, ovp_count
;
169 total_count
= le64_to_cpu(sbi
->raw_super
->block_count
);
170 user_block_count
= sbi
->user_block_count
;
171 start_count
= le32_to_cpu(sbi
->raw_super
->segment0_blkaddr
);
172 ovp_count
= SM_I(sbi
)->ovp_segments
<< sbi
->log_blocks_per_seg
;
173 buf
->f_type
= F2FS_SUPER_MAGIC
;
174 buf
->f_bsize
= sbi
->blocksize
;
176 buf
->f_blocks
= total_count
- start_count
;
177 buf
->f_bfree
= buf
->f_blocks
- valid_user_blocks(sbi
) - ovp_count
;
178 buf
->f_bavail
= user_block_count
- valid_user_blocks(sbi
);
180 buf
->f_files
= sbi
->total_node_count
;
181 buf
->f_ffree
= sbi
->total_node_count
- valid_inode_count(sbi
);
183 buf
->f_namelen
= F2FS_MAX_NAME_LEN
;
184 buf
->f_fsid
.val
[0] = (u32
)id
;
185 buf
->f_fsid
.val
[1] = (u32
)(id
>> 32);
190 static int f2fs_show_options(struct seq_file
*seq
, struct dentry
*root
)
192 struct f2fs_sb_info
*sbi
= F2FS_SB(root
->d_sb
);
194 if (test_opt(sbi
, BG_GC
))
195 seq_puts(seq
, ",background_gc_on");
197 seq_puts(seq
, ",background_gc_off");
198 if (test_opt(sbi
, DISABLE_ROLL_FORWARD
))
199 seq_puts(seq
, ",disable_roll_forward");
200 if (test_opt(sbi
, DISCARD
))
201 seq_puts(seq
, ",discard");
202 if (test_opt(sbi
, NOHEAP
))
203 seq_puts(seq
, ",no_heap_alloc");
204 #ifdef CONFIG_F2FS_FS_XATTR
205 if (test_opt(sbi
, XATTR_USER
))
206 seq_puts(seq
, ",user_xattr");
208 seq_puts(seq
, ",nouser_xattr");
210 #ifdef CONFIG_F2FS_FS_POSIX_ACL
211 if (test_opt(sbi
, POSIX_ACL
))
212 seq_puts(seq
, ",acl");
214 seq_puts(seq
, ",noacl");
216 if (test_opt(sbi
, DISABLE_EXT_IDENTIFY
))
217 seq_puts(seq
, ",disable_ext_identify");
219 seq_printf(seq
, ",active_logs=%u", sbi
->active_logs
);
224 static struct super_operations f2fs_sops
= {
225 .alloc_inode
= f2fs_alloc_inode
,
226 .destroy_inode
= f2fs_destroy_inode
,
227 .write_inode
= f2fs_write_inode
,
228 .show_options
= f2fs_show_options
,
229 .evict_inode
= f2fs_evict_inode
,
230 .put_super
= f2fs_put_super
,
231 .sync_fs
= f2fs_sync_fs
,
232 .freeze_fs
= f2fs_freeze
,
233 .unfreeze_fs
= f2fs_unfreeze
,
234 .statfs
= f2fs_statfs
,
237 static struct inode
*f2fs_nfs_get_inode(struct super_block
*sb
,
238 u64 ino
, u32 generation
)
240 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
243 if (ino
< F2FS_ROOT_INO(sbi
))
244 return ERR_PTR(-ESTALE
);
247 * f2fs_iget isn't quite right if the inode is currently unallocated!
248 * However f2fs_iget currently does appropriate checks to handle stale
249 * inodes so everything is OK.
251 inode
= f2fs_iget(sb
, ino
);
253 return ERR_CAST(inode
);
254 if (generation
&& inode
->i_generation
!= generation
) {
255 /* we didn't find the right inode.. */
257 return ERR_PTR(-ESTALE
);
262 static struct dentry
*f2fs_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
263 int fh_len
, int fh_type
)
265 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
269 static struct dentry
*f2fs_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
270 int fh_len
, int fh_type
)
272 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
276 static const struct export_operations f2fs_export_ops
= {
277 .fh_to_dentry
= f2fs_fh_to_dentry
,
278 .fh_to_parent
= f2fs_fh_to_parent
,
279 .get_parent
= f2fs_get_parent
,
282 static int parse_options(struct super_block
*sb
, struct f2fs_sb_info
*sbi
,
285 substring_t args
[MAX_OPT_ARGS
];
292 while ((p
= strsep(&options
, ",")) != NULL
) {
297 * Initialize args struct so we know whether arg was
298 * found; some options take optional arguments.
300 args
[0].to
= args
[0].from
= NULL
;
301 token
= match_token(p
, f2fs_tokens
, args
);
304 case Opt_gc_background_off
:
305 clear_opt(sbi
, BG_GC
);
307 case Opt_disable_roll_forward
:
308 set_opt(sbi
, DISABLE_ROLL_FORWARD
);
311 set_opt(sbi
, DISCARD
);
314 set_opt(sbi
, NOHEAP
);
316 #ifdef CONFIG_F2FS_FS_XATTR
317 case Opt_nouser_xattr
:
318 clear_opt(sbi
, XATTR_USER
);
321 case Opt_nouser_xattr
:
322 f2fs_msg(sb
, KERN_INFO
,
323 "nouser_xattr options not supported");
326 #ifdef CONFIG_F2FS_FS_POSIX_ACL
328 clear_opt(sbi
, POSIX_ACL
);
332 f2fs_msg(sb
, KERN_INFO
, "noacl options not supported");
335 case Opt_active_logs
:
336 if (args
->from
&& match_int(args
, &arg
))
338 if (arg
!= 2 && arg
!= 4 && arg
!= NR_CURSEG_TYPE
)
340 sbi
->active_logs
= arg
;
342 case Opt_disable_ext_identify
:
343 set_opt(sbi
, DISABLE_EXT_IDENTIFY
);
346 f2fs_msg(sb
, KERN_ERR
,
347 "Unrecognized mount option \"%s\" or missing value",
355 static loff_t
max_file_size(unsigned bits
)
357 loff_t result
= ADDRS_PER_INODE
;
358 loff_t leaf_count
= ADDRS_PER_BLOCK
;
360 /* two direct node blocks */
361 result
+= (leaf_count
* 2);
363 /* two indirect node blocks */
364 leaf_count
*= NIDS_PER_BLOCK
;
365 result
+= (leaf_count
* 2);
367 /* one double indirect node block */
368 leaf_count
*= NIDS_PER_BLOCK
;
369 result
+= leaf_count
;
375 static int sanity_check_raw_super(struct super_block
*sb
,
376 struct f2fs_super_block
*raw_super
)
378 unsigned int blocksize
;
380 if (F2FS_SUPER_MAGIC
!= le32_to_cpu(raw_super
->magic
)) {
381 f2fs_msg(sb
, KERN_INFO
,
382 "Magic Mismatch, valid(0x%x) - read(0x%x)",
383 F2FS_SUPER_MAGIC
, le32_to_cpu(raw_super
->magic
));
387 /* Currently, support only 4KB page cache size */
388 if (F2FS_BLKSIZE
!= PAGE_CACHE_SIZE
) {
389 f2fs_msg(sb
, KERN_INFO
,
390 "Invalid page_cache_size (%lu), supports only 4KB\n",
395 /* Currently, support only 4KB block size */
396 blocksize
= 1 << le32_to_cpu(raw_super
->log_blocksize
);
397 if (blocksize
!= F2FS_BLKSIZE
) {
398 f2fs_msg(sb
, KERN_INFO
,
399 "Invalid blocksize (%u), supports only 4KB\n",
404 if (le32_to_cpu(raw_super
->log_sectorsize
) !=
405 F2FS_LOG_SECTOR_SIZE
) {
406 f2fs_msg(sb
, KERN_INFO
, "Invalid log sectorsize");
409 if (le32_to_cpu(raw_super
->log_sectors_per_block
) !=
410 F2FS_LOG_SECTORS_PER_BLOCK
) {
411 f2fs_msg(sb
, KERN_INFO
, "Invalid log sectors per block");
417 static int sanity_check_ckpt(struct f2fs_sb_info
*sbi
)
419 unsigned int total
, fsmeta
;
420 struct f2fs_super_block
*raw_super
= F2FS_RAW_SUPER(sbi
);
421 struct f2fs_checkpoint
*ckpt
= F2FS_CKPT(sbi
);
423 total
= le32_to_cpu(raw_super
->segment_count
);
424 fsmeta
= le32_to_cpu(raw_super
->segment_count_ckpt
);
425 fsmeta
+= le32_to_cpu(raw_super
->segment_count_sit
);
426 fsmeta
+= le32_to_cpu(raw_super
->segment_count_nat
);
427 fsmeta
+= le32_to_cpu(ckpt
->rsvd_segment_count
);
428 fsmeta
+= le32_to_cpu(raw_super
->segment_count_ssa
);
433 if (is_set_ckpt_flags(ckpt
, CP_ERROR_FLAG
)) {
434 f2fs_msg(sbi
->sb
, KERN_ERR
, "A bug case: need to run fsck");
440 static void init_sb_info(struct f2fs_sb_info
*sbi
)
442 struct f2fs_super_block
*raw_super
= sbi
->raw_super
;
445 sbi
->log_sectors_per_block
=
446 le32_to_cpu(raw_super
->log_sectors_per_block
);
447 sbi
->log_blocksize
= le32_to_cpu(raw_super
->log_blocksize
);
448 sbi
->blocksize
= 1 << sbi
->log_blocksize
;
449 sbi
->log_blocks_per_seg
= le32_to_cpu(raw_super
->log_blocks_per_seg
);
450 sbi
->blocks_per_seg
= 1 << sbi
->log_blocks_per_seg
;
451 sbi
->segs_per_sec
= le32_to_cpu(raw_super
->segs_per_sec
);
452 sbi
->secs_per_zone
= le32_to_cpu(raw_super
->secs_per_zone
);
453 sbi
->total_sections
= le32_to_cpu(raw_super
->section_count
);
454 sbi
->total_node_count
=
455 (le32_to_cpu(raw_super
->segment_count_nat
) / 2)
456 * sbi
->blocks_per_seg
* NAT_ENTRY_PER_BLOCK
;
457 sbi
->root_ino_num
= le32_to_cpu(raw_super
->root_ino
);
458 sbi
->node_ino_num
= le32_to_cpu(raw_super
->node_ino
);
459 sbi
->meta_ino_num
= le32_to_cpu(raw_super
->meta_ino
);
461 for (i
= 0; i
< NR_COUNT_TYPE
; i
++)
462 atomic_set(&sbi
->nr_pages
[i
], 0);
465 static int validate_superblock(struct super_block
*sb
,
466 struct f2fs_super_block
**raw_super
,
467 struct buffer_head
**raw_super_buf
, sector_t block
)
469 const char *super
= (block
== 0 ? "first" : "second");
471 /* read f2fs raw super block */
472 *raw_super_buf
= sb_bread(sb
, block
);
473 if (!*raw_super_buf
) {
474 f2fs_msg(sb
, KERN_ERR
, "unable to read %s superblock",
479 *raw_super
= (struct f2fs_super_block
*)
480 ((char *)(*raw_super_buf
)->b_data
+ F2FS_SUPER_OFFSET
);
482 /* sanity checking of raw super */
483 if (!sanity_check_raw_super(sb
, *raw_super
))
486 f2fs_msg(sb
, KERN_ERR
, "Can't find a valid F2FS filesystem "
487 "in %s superblock", super
);
491 static int f2fs_fill_super(struct super_block
*sb
, void *data
, int silent
)
493 struct f2fs_sb_info
*sbi
;
494 struct f2fs_super_block
*raw_super
;
495 struct buffer_head
*raw_super_buf
;
500 /* allocate memory for f2fs-specific super block info */
501 sbi
= kzalloc(sizeof(struct f2fs_sb_info
), GFP_KERNEL
);
505 /* set a block size */
506 if (!sb_set_blocksize(sb
, F2FS_BLKSIZE
)) {
507 f2fs_msg(sb
, KERN_ERR
, "unable to set blocksize");
511 if (validate_superblock(sb
, &raw_super
, &raw_super_buf
, 0)) {
512 brelse(raw_super_buf
);
513 if (validate_superblock(sb
, &raw_super
, &raw_super_buf
, 1))
516 /* init some FS parameters */
517 sbi
->active_logs
= NR_CURSEG_TYPE
;
521 #ifdef CONFIG_F2FS_FS_XATTR
522 set_opt(sbi
, XATTR_USER
);
524 #ifdef CONFIG_F2FS_FS_POSIX_ACL
525 set_opt(sbi
, POSIX_ACL
);
527 /* parse mount options */
528 if (parse_options(sb
, sbi
, (char *)data
))
531 sb
->s_maxbytes
= max_file_size(le32_to_cpu(raw_super
->log_blocksize
));
532 sb
->s_max_links
= F2FS_LINK_MAX
;
533 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
535 sb
->s_op
= &f2fs_sops
;
536 sb
->s_xattr
= f2fs_xattr_handlers
;
537 sb
->s_export_op
= &f2fs_export_ops
;
538 sb
->s_magic
= F2FS_SUPER_MAGIC
;
541 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
542 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
543 memcpy(sb
->s_uuid
, raw_super
->uuid
, sizeof(raw_super
->uuid
));
545 /* init f2fs-specific super block info */
547 sbi
->raw_super
= raw_super
;
548 sbi
->raw_super_buf
= raw_super_buf
;
549 mutex_init(&sbi
->gc_mutex
);
550 mutex_init(&sbi
->write_inode
);
551 mutex_init(&sbi
->writepages
);
552 mutex_init(&sbi
->cp_mutex
);
553 for (i
= 0; i
< NR_LOCK_TYPE
; i
++)
554 mutex_init(&sbi
->fs_lock
[i
]);
556 spin_lock_init(&sbi
->stat_lock
);
557 init_rwsem(&sbi
->bio_sem
);
560 /* get an inode for meta space */
561 sbi
->meta_inode
= f2fs_iget(sb
, F2FS_META_INO(sbi
));
562 if (IS_ERR(sbi
->meta_inode
)) {
563 f2fs_msg(sb
, KERN_ERR
, "Failed to read F2FS meta data inode");
564 err
= PTR_ERR(sbi
->meta_inode
);
568 err
= get_valid_checkpoint(sbi
);
570 f2fs_msg(sb
, KERN_ERR
, "Failed to get valid F2FS checkpoint");
571 goto free_meta_inode
;
574 /* sanity checking of checkpoint */
576 if (sanity_check_ckpt(sbi
)) {
577 f2fs_msg(sb
, KERN_ERR
, "Invalid F2FS checkpoint");
581 sbi
->total_valid_node_count
=
582 le32_to_cpu(sbi
->ckpt
->valid_node_count
);
583 sbi
->total_valid_inode_count
=
584 le32_to_cpu(sbi
->ckpt
->valid_inode_count
);
585 sbi
->user_block_count
= le64_to_cpu(sbi
->ckpt
->user_block_count
);
586 sbi
->total_valid_block_count
=
587 le64_to_cpu(sbi
->ckpt
->valid_block_count
);
588 sbi
->last_valid_block_count
= sbi
->total_valid_block_count
;
589 sbi
->alloc_valid_block_count
= 0;
590 INIT_LIST_HEAD(&sbi
->dir_inode_list
);
591 spin_lock_init(&sbi
->dir_inode_lock
);
593 init_orphan_info(sbi
);
595 /* setup f2fs internal modules */
596 err
= build_segment_manager(sbi
);
598 f2fs_msg(sb
, KERN_ERR
,
599 "Failed to initialize F2FS segment manager");
602 err
= build_node_manager(sbi
);
604 f2fs_msg(sb
, KERN_ERR
,
605 "Failed to initialize F2FS node manager");
609 build_gc_manager(sbi
);
611 /* get an inode for node space */
612 sbi
->node_inode
= f2fs_iget(sb
, F2FS_NODE_INO(sbi
));
613 if (IS_ERR(sbi
->node_inode
)) {
614 f2fs_msg(sb
, KERN_ERR
, "Failed to read node inode");
615 err
= PTR_ERR(sbi
->node_inode
);
619 /* if there are nt orphan nodes free them */
621 if (recover_orphan_inodes(sbi
))
622 goto free_node_inode
;
624 /* read root inode and dentry */
625 root
= f2fs_iget(sb
, F2FS_ROOT_INO(sbi
));
627 f2fs_msg(sb
, KERN_ERR
, "Failed to read root inode");
629 goto free_node_inode
;
631 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
)
632 goto free_root_inode
;
634 sb
->s_root
= d_make_root(root
); /* allocate root dentry */
637 goto free_root_inode
;
640 /* recover fsynced data */
641 if (!test_opt(sbi
, DISABLE_ROLL_FORWARD
))
642 recover_fsync_data(sbi
);
644 /* After POR, we can run background GC thread */
645 err
= start_gc_thread(sbi
);
649 err
= f2fs_build_stats(sbi
);
660 iput(sbi
->node_inode
);
662 destroy_node_manager(sbi
);
664 destroy_segment_manager(sbi
);
668 make_bad_inode(sbi
->meta_inode
);
669 iput(sbi
->meta_inode
);
671 brelse(raw_super_buf
);
677 static struct dentry
*f2fs_mount(struct file_system_type
*fs_type
, int flags
,
678 const char *dev_name
, void *data
)
680 return mount_bdev(fs_type
, flags
, dev_name
, data
, f2fs_fill_super
);
683 static struct file_system_type f2fs_fs_type
= {
684 .owner
= THIS_MODULE
,
687 .kill_sb
= kill_block_super
,
688 .fs_flags
= FS_REQUIRES_DEV
,
690 MODULE_ALIAS_FS("f2fs");
692 static int __init
init_inodecache(void)
694 f2fs_inode_cachep
= f2fs_kmem_cache_create("f2fs_inode_cache",
695 sizeof(struct f2fs_inode_info
), NULL
);
696 if (f2fs_inode_cachep
== NULL
)
701 static void destroy_inodecache(void)
704 * Make sure all delayed rcu free inodes are flushed before we
708 kmem_cache_destroy(f2fs_inode_cachep
);
711 static int __init
init_f2fs_fs(void)
715 err
= init_inodecache();
718 err
= create_node_manager_caches();
721 err
= create_gc_caches();
724 err
= create_checkpoint_caches();
727 err
= register_filesystem(&f2fs_fs_type
);
730 f2fs_create_root_stats();
735 static void __exit
exit_f2fs_fs(void)
737 f2fs_destroy_root_stats();
738 unregister_filesystem(&f2fs_fs_type
);
739 destroy_checkpoint_caches();
741 destroy_node_manager_caches();
742 destroy_inodecache();
745 module_init(init_f2fs_fs
)
746 module_exit(exit_f2fs_fs
)
748 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
749 MODULE_DESCRIPTION("Flash Friendly File System");
750 MODULE_LICENSE("GPL");