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/random.h>
22 #include <linux/exportfs.h>
23 #include <linux/blkdev.h>
24 #include <linux/f2fs_fs.h>
31 #define CREATE_TRACE_POINTS
32 #include <trace/events/f2fs.h>
34 static struct kmem_cache *f2fs_inode_cachep;
37 Opt_gc_background_off,
38 Opt_disable_roll_forward,
44 Opt_disable_ext_identify,
48 static match_table_t f2fs_tokens = {
49 {Opt_gc_background_off, "background_gc_off"},
50 {Opt_disable_roll_forward, "disable_roll_forward"},
51 {Opt_discard, "discard"},
52 {Opt_noheap, "no_heap"},
53 {Opt_nouser_xattr, "nouser_xattr"},
55 {Opt_active_logs, "active_logs=%u"},
56 {Opt_disable_ext_identify, "disable_ext_identify"},
60 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
68 printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
72 static void init_once(void *foo)
74 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
76 inode_init_once(&fi->vfs_inode);
79 static struct inode *f2fs_alloc_inode(struct super_block *sb)
81 struct f2fs_inode_info *fi;
83 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_NOFS | __GFP_ZERO);
87 init_once((void *) fi);
89 /* Initialize f2fs-specific inode info */
90 fi->vfs_inode.i_version = 1;
91 atomic_set(&fi->dirty_dents, 0);
92 fi->i_current_depth = 1;
94 rwlock_init(&fi->ext.ext_lock);
96 set_inode_flag(fi, FI_NEW_INODE);
98 return &fi->vfs_inode;
101 static int f2fs_drop_inode(struct inode *inode)
104 * This is to avoid a deadlock condition like below.
105 * writeback_single_inode(inode)
106 * - f2fs_write_data_page
107 * - f2fs_gc -> iput -> evict
108 * - inode_wait_for_writeback(inode)
110 if (!inode_unhashed(inode) && inode->i_state & I_SYNC)
112 return generic_drop_inode(inode);
115 static void f2fs_i_callback(struct rcu_head *head)
117 struct inode *inode = container_of(head, struct inode, i_rcu);
118 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
121 static void f2fs_destroy_inode(struct inode *inode)
123 call_rcu(&inode->i_rcu, f2fs_i_callback);
126 static void f2fs_put_super(struct super_block *sb)
128 struct f2fs_sb_info *sbi = F2FS_SB(sb);
130 f2fs_destroy_stats(sbi);
133 write_checkpoint(sbi, true);
135 iput(sbi->node_inode);
136 iput(sbi->meta_inode);
138 /* destroy f2fs internal modules */
139 destroy_node_manager(sbi);
140 destroy_segment_manager(sbi);
144 sb->s_fs_info = NULL;
145 brelse(sbi->raw_super_buf);
149 int f2fs_sync_fs(struct super_block *sb, int sync)
151 struct f2fs_sb_info *sbi = F2FS_SB(sb);
153 trace_f2fs_sync_fs(sb, sync);
155 if (!sbi->s_dirty && !get_pages(sbi, F2FS_DIRTY_NODES))
159 mutex_lock(&sbi->gc_mutex);
160 write_checkpoint(sbi, false);
161 mutex_unlock(&sbi->gc_mutex);
163 f2fs_balance_fs(sbi);
169 static int f2fs_freeze(struct super_block *sb)
173 if (sb->s_flags & MS_RDONLY)
176 err = f2fs_sync_fs(sb, 1);
180 static int f2fs_unfreeze(struct super_block *sb)
185 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
187 struct super_block *sb = dentry->d_sb;
188 struct f2fs_sb_info *sbi = F2FS_SB(sb);
189 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
190 block_t total_count, user_block_count, start_count, ovp_count;
192 total_count = le64_to_cpu(sbi->raw_super->block_count);
193 user_block_count = sbi->user_block_count;
194 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
195 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
196 buf->f_type = F2FS_SUPER_MAGIC;
197 buf->f_bsize = sbi->blocksize;
199 buf->f_blocks = total_count - start_count;
200 buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
201 buf->f_bavail = user_block_count - valid_user_blocks(sbi);
203 buf->f_files = sbi->total_node_count;
204 buf->f_ffree = sbi->total_node_count - valid_inode_count(sbi);
206 buf->f_namelen = F2FS_NAME_LEN;
207 buf->f_fsid.val[0] = (u32)id;
208 buf->f_fsid.val[1] = (u32)(id >> 32);
213 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
215 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
217 if (test_opt(sbi, BG_GC))
218 seq_puts(seq, ",background_gc_on");
220 seq_puts(seq, ",background_gc_off");
221 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
222 seq_puts(seq, ",disable_roll_forward");
223 if (test_opt(sbi, DISCARD))
224 seq_puts(seq, ",discard");
225 if (test_opt(sbi, NOHEAP))
226 seq_puts(seq, ",no_heap_alloc");
227 #ifdef CONFIG_F2FS_FS_XATTR
228 if (test_opt(sbi, XATTR_USER))
229 seq_puts(seq, ",user_xattr");
231 seq_puts(seq, ",nouser_xattr");
233 #ifdef CONFIG_F2FS_FS_POSIX_ACL
234 if (test_opt(sbi, POSIX_ACL))
235 seq_puts(seq, ",acl");
237 seq_puts(seq, ",noacl");
239 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
240 seq_puts(seq, ",disable_ext_identify");
242 seq_printf(seq, ",active_logs=%u", sbi->active_logs);
247 static struct super_operations f2fs_sops = {
248 .alloc_inode = f2fs_alloc_inode,
249 .drop_inode = f2fs_drop_inode,
250 .destroy_inode = f2fs_destroy_inode,
251 .write_inode = f2fs_write_inode,
252 .show_options = f2fs_show_options,
253 .evict_inode = f2fs_evict_inode,
254 .put_super = f2fs_put_super,
255 .sync_fs = f2fs_sync_fs,
256 .freeze_fs = f2fs_freeze,
257 .unfreeze_fs = f2fs_unfreeze,
258 .statfs = f2fs_statfs,
261 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
262 u64 ino, u32 generation)
264 struct f2fs_sb_info *sbi = F2FS_SB(sb);
267 if (ino < F2FS_ROOT_INO(sbi))
268 return ERR_PTR(-ESTALE);
271 * f2fs_iget isn't quite right if the inode is currently unallocated!
272 * However f2fs_iget currently does appropriate checks to handle stale
273 * inodes so everything is OK.
275 inode = f2fs_iget(sb, ino);
277 return ERR_CAST(inode);
278 if (generation && inode->i_generation != generation) {
279 /* we didn't find the right inode.. */
281 return ERR_PTR(-ESTALE);
286 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
287 int fh_len, int fh_type)
289 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
293 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
294 int fh_len, int fh_type)
296 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
300 static const struct export_operations f2fs_export_ops = {
301 .fh_to_dentry = f2fs_fh_to_dentry,
302 .fh_to_parent = f2fs_fh_to_parent,
303 .get_parent = f2fs_get_parent,
306 static int parse_options(struct super_block *sb, struct f2fs_sb_info *sbi,
309 substring_t args[MAX_OPT_ARGS];
316 while ((p = strsep(&options, ",")) != NULL) {
321 * Initialize args struct so we know whether arg was
322 * found; some options take optional arguments.
324 args[0].to = args[0].from = NULL;
325 token = match_token(p, f2fs_tokens, args);
328 case Opt_gc_background_off:
329 clear_opt(sbi, BG_GC);
331 case Opt_disable_roll_forward:
332 set_opt(sbi, DISABLE_ROLL_FORWARD);
335 set_opt(sbi, DISCARD);
338 set_opt(sbi, NOHEAP);
340 #ifdef CONFIG_F2FS_FS_XATTR
341 case Opt_nouser_xattr:
342 clear_opt(sbi, XATTR_USER);
345 case Opt_nouser_xattr:
346 f2fs_msg(sb, KERN_INFO,
347 "nouser_xattr options not supported");
350 #ifdef CONFIG_F2FS_FS_POSIX_ACL
352 clear_opt(sbi, POSIX_ACL);
356 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
359 case Opt_active_logs:
360 if (args->from && match_int(args, &arg))
362 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
364 sbi->active_logs = arg;
366 case Opt_disable_ext_identify:
367 set_opt(sbi, DISABLE_EXT_IDENTIFY);
370 f2fs_msg(sb, KERN_ERR,
371 "Unrecognized mount option \"%s\" or missing value",
379 static loff_t max_file_size(unsigned bits)
381 loff_t result = ADDRS_PER_INODE;
382 loff_t leaf_count = ADDRS_PER_BLOCK;
384 /* two direct node blocks */
385 result += (leaf_count * 2);
387 /* two indirect node blocks */
388 leaf_count *= NIDS_PER_BLOCK;
389 result += (leaf_count * 2);
391 /* one double indirect node block */
392 leaf_count *= NIDS_PER_BLOCK;
393 result += leaf_count;
399 static int sanity_check_raw_super(struct super_block *sb,
400 struct f2fs_super_block *raw_super)
402 unsigned int blocksize;
404 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
405 f2fs_msg(sb, KERN_INFO,
406 "Magic Mismatch, valid(0x%x) - read(0x%x)",
407 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
411 /* Currently, support only 4KB page cache size */
412 if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
413 f2fs_msg(sb, KERN_INFO,
414 "Invalid page_cache_size (%lu), supports only 4KB\n",
419 /* Currently, support only 4KB block size */
420 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
421 if (blocksize != F2FS_BLKSIZE) {
422 f2fs_msg(sb, KERN_INFO,
423 "Invalid blocksize (%u), supports only 4KB\n",
428 if (le32_to_cpu(raw_super->log_sectorsize) !=
429 F2FS_LOG_SECTOR_SIZE) {
430 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize");
433 if (le32_to_cpu(raw_super->log_sectors_per_block) !=
434 F2FS_LOG_SECTORS_PER_BLOCK) {
435 f2fs_msg(sb, KERN_INFO, "Invalid log sectors per block");
441 static int sanity_check_ckpt(struct f2fs_sb_info *sbi)
443 unsigned int total, fsmeta;
444 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
445 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
447 total = le32_to_cpu(raw_super->segment_count);
448 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
449 fsmeta += le32_to_cpu(raw_super->segment_count_sit);
450 fsmeta += le32_to_cpu(raw_super->segment_count_nat);
451 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
452 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
457 if (is_set_ckpt_flags(ckpt, CP_ERROR_FLAG)) {
458 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
464 static void init_sb_info(struct f2fs_sb_info *sbi)
466 struct f2fs_super_block *raw_super = sbi->raw_super;
469 sbi->log_sectors_per_block =
470 le32_to_cpu(raw_super->log_sectors_per_block);
471 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
472 sbi->blocksize = 1 << sbi->log_blocksize;
473 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
474 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
475 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
476 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
477 sbi->total_sections = le32_to_cpu(raw_super->section_count);
478 sbi->total_node_count =
479 (le32_to_cpu(raw_super->segment_count_nat) / 2)
480 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
481 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
482 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
483 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
484 sbi->cur_victim_sec = NULL_SECNO;
486 for (i = 0; i < NR_COUNT_TYPE; i++)
487 atomic_set(&sbi->nr_pages[i], 0);
490 static int validate_superblock(struct super_block *sb,
491 struct f2fs_super_block **raw_super,
492 struct buffer_head **raw_super_buf, sector_t block)
494 const char *super = (block == 0 ? "first" : "second");
496 /* read f2fs raw super block */
497 *raw_super_buf = sb_bread(sb, block);
498 if (!*raw_super_buf) {
499 f2fs_msg(sb, KERN_ERR, "unable to read %s superblock",
504 *raw_super = (struct f2fs_super_block *)
505 ((char *)(*raw_super_buf)->b_data + F2FS_SUPER_OFFSET);
507 /* sanity checking of raw super */
508 if (!sanity_check_raw_super(sb, *raw_super))
511 f2fs_msg(sb, KERN_ERR, "Can't find a valid F2FS filesystem "
512 "in %s superblock", super);
516 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
518 struct f2fs_sb_info *sbi;
519 struct f2fs_super_block *raw_super;
520 struct buffer_head *raw_super_buf;
525 /* allocate memory for f2fs-specific super block info */
526 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
530 /* set a block size */
531 if (!sb_set_blocksize(sb, F2FS_BLKSIZE)) {
532 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
536 err = validate_superblock(sb, &raw_super, &raw_super_buf, 0);
538 brelse(raw_super_buf);
539 /* check secondary superblock when primary failed */
540 err = validate_superblock(sb, &raw_super, &raw_super_buf, 1);
544 /* init some FS parameters */
545 sbi->active_logs = NR_CURSEG_TYPE;
549 #ifdef CONFIG_F2FS_FS_XATTR
550 set_opt(sbi, XATTR_USER);
552 #ifdef CONFIG_F2FS_FS_POSIX_ACL
553 set_opt(sbi, POSIX_ACL);
555 /* parse mount options */
556 err = parse_options(sb, sbi, (char *)data);
560 sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
561 sb->s_max_links = F2FS_LINK_MAX;
562 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
564 sb->s_op = &f2fs_sops;
565 sb->s_xattr = f2fs_xattr_handlers;
566 sb->s_export_op = &f2fs_export_ops;
567 sb->s_magic = F2FS_SUPER_MAGIC;
570 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
571 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
572 memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
574 /* init f2fs-specific super block info */
576 sbi->raw_super = raw_super;
577 sbi->raw_super_buf = raw_super_buf;
578 mutex_init(&sbi->gc_mutex);
579 mutex_init(&sbi->writepages);
580 mutex_init(&sbi->cp_mutex);
581 for (i = 0; i < NR_GLOBAL_LOCKS; i++)
582 mutex_init(&sbi->fs_lock[i]);
583 mutex_init(&sbi->node_write);
585 spin_lock_init(&sbi->stat_lock);
586 init_rwsem(&sbi->bio_sem);
589 /* get an inode for meta space */
590 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
591 if (IS_ERR(sbi->meta_inode)) {
592 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
593 err = PTR_ERR(sbi->meta_inode);
597 err = get_valid_checkpoint(sbi);
599 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
600 goto free_meta_inode;
603 /* sanity checking of checkpoint */
605 if (sanity_check_ckpt(sbi)) {
606 f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint");
610 sbi->total_valid_node_count =
611 le32_to_cpu(sbi->ckpt->valid_node_count);
612 sbi->total_valid_inode_count =
613 le32_to_cpu(sbi->ckpt->valid_inode_count);
614 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
615 sbi->total_valid_block_count =
616 le64_to_cpu(sbi->ckpt->valid_block_count);
617 sbi->last_valid_block_count = sbi->total_valid_block_count;
618 sbi->alloc_valid_block_count = 0;
619 INIT_LIST_HEAD(&sbi->dir_inode_list);
620 spin_lock_init(&sbi->dir_inode_lock);
622 init_orphan_info(sbi);
624 /* setup f2fs internal modules */
625 err = build_segment_manager(sbi);
627 f2fs_msg(sb, KERN_ERR,
628 "Failed to initialize F2FS segment manager");
631 err = build_node_manager(sbi);
633 f2fs_msg(sb, KERN_ERR,
634 "Failed to initialize F2FS node manager");
638 build_gc_manager(sbi);
640 /* get an inode for node space */
641 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
642 if (IS_ERR(sbi->node_inode)) {
643 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
644 err = PTR_ERR(sbi->node_inode);
648 /* if there are nt orphan nodes free them */
650 if (recover_orphan_inodes(sbi))
651 goto free_node_inode;
653 /* read root inode and dentry */
654 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
656 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
658 goto free_node_inode;
660 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size)
661 goto free_root_inode;
663 sb->s_root = d_make_root(root); /* allocate root dentry */
666 goto free_root_inode;
669 /* recover fsynced data */
670 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
671 err = recover_fsync_data(sbi);
673 f2fs_msg(sb, KERN_ERR,
674 "Cannot recover all fsync data errno=%ld", err);
677 /* After POR, we can run background GC thread */
678 err = start_gc_thread(sbi);
682 err = f2fs_build_stats(sbi);
686 if (test_opt(sbi, DISCARD)) {
687 struct request_queue *q = bdev_get_queue(sb->s_bdev);
688 if (!blk_queue_discard(q))
689 f2fs_msg(sb, KERN_WARNING,
690 "mounting with \"discard\" option, but "
691 "the device does not support discard");
701 iput(sbi->node_inode);
703 destroy_node_manager(sbi);
705 destroy_segment_manager(sbi);
709 make_bad_inode(sbi->meta_inode);
710 iput(sbi->meta_inode);
712 brelse(raw_super_buf);
718 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
719 const char *dev_name, void *data)
721 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
724 static struct file_system_type f2fs_fs_type = {
725 .owner = THIS_MODULE,
728 .kill_sb = kill_block_super,
729 .fs_flags = FS_REQUIRES_DEV,
732 static int __init init_inodecache(void)
734 f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
735 sizeof(struct f2fs_inode_info), NULL);
736 if (f2fs_inode_cachep == NULL)
741 static void destroy_inodecache(void)
744 * Make sure all delayed rcu free inodes are flushed before we
748 kmem_cache_destroy(f2fs_inode_cachep);
751 static int __init init_f2fs_fs(void)
755 err = init_inodecache();
758 err = create_node_manager_caches();
761 err = create_gc_caches();
764 err = create_checkpoint_caches();
767 err = register_filesystem(&f2fs_fs_type);
770 f2fs_create_root_stats();
775 static void __exit exit_f2fs_fs(void)
777 f2fs_destroy_root_stats();
778 unregister_filesystem(&f2fs_fs_type);
779 destroy_checkpoint_caches();
781 destroy_node_manager_caches();
782 destroy_inodecache();
785 module_init(init_f2fs_fs)
786 module_exit(exit_f2fs_fs)
788 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
789 MODULE_DESCRIPTION("Flash Friendly File System");
790 MODULE_LICENSE("GPL");