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);
116 * f2fs_dirty_inode() is called from __mark_inode_dirty()
118 * We should call set_dirty_inode to write the dirty inode through write_inode.
120 static void f2fs_dirty_inode(struct inode *inode, int flags)
122 set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
126 static void f2fs_i_callback(struct rcu_head *head)
128 struct inode *inode = container_of(head, struct inode, i_rcu);
129 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
132 static void f2fs_destroy_inode(struct inode *inode)
134 call_rcu(&inode->i_rcu, f2fs_i_callback);
137 static void f2fs_put_super(struct super_block *sb)
139 struct f2fs_sb_info *sbi = F2FS_SB(sb);
141 f2fs_destroy_stats(sbi);
144 write_checkpoint(sbi, true);
146 iput(sbi->node_inode);
147 iput(sbi->meta_inode);
149 /* destroy f2fs internal modules */
150 destroy_node_manager(sbi);
151 destroy_segment_manager(sbi);
155 sb->s_fs_info = NULL;
156 brelse(sbi->raw_super_buf);
160 int f2fs_sync_fs(struct super_block *sb, int sync)
162 struct f2fs_sb_info *sbi = F2FS_SB(sb);
164 trace_f2fs_sync_fs(sb, sync);
166 if (!sbi->s_dirty && !get_pages(sbi, F2FS_DIRTY_NODES))
170 mutex_lock(&sbi->gc_mutex);
171 write_checkpoint(sbi, false);
172 mutex_unlock(&sbi->gc_mutex);
174 f2fs_balance_fs(sbi);
180 static int f2fs_freeze(struct super_block *sb)
184 if (f2fs_readonly(sb))
187 err = f2fs_sync_fs(sb, 1);
191 static int f2fs_unfreeze(struct super_block *sb)
196 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
198 struct super_block *sb = dentry->d_sb;
199 struct f2fs_sb_info *sbi = F2FS_SB(sb);
200 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
201 block_t total_count, user_block_count, start_count, ovp_count;
203 total_count = le64_to_cpu(sbi->raw_super->block_count);
204 user_block_count = sbi->user_block_count;
205 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
206 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
207 buf->f_type = F2FS_SUPER_MAGIC;
208 buf->f_bsize = sbi->blocksize;
210 buf->f_blocks = total_count - start_count;
211 buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
212 buf->f_bavail = user_block_count - valid_user_blocks(sbi);
214 buf->f_files = sbi->total_node_count;
215 buf->f_ffree = sbi->total_node_count - valid_inode_count(sbi);
217 buf->f_namelen = F2FS_NAME_LEN;
218 buf->f_fsid.val[0] = (u32)id;
219 buf->f_fsid.val[1] = (u32)(id >> 32);
224 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
226 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
228 if (test_opt(sbi, BG_GC))
229 seq_puts(seq, ",background_gc_on");
231 seq_puts(seq, ",background_gc_off");
232 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
233 seq_puts(seq, ",disable_roll_forward");
234 if (test_opt(sbi, DISCARD))
235 seq_puts(seq, ",discard");
236 if (test_opt(sbi, NOHEAP))
237 seq_puts(seq, ",no_heap_alloc");
238 #ifdef CONFIG_F2FS_FS_XATTR
239 if (test_opt(sbi, XATTR_USER))
240 seq_puts(seq, ",user_xattr");
242 seq_puts(seq, ",nouser_xattr");
244 #ifdef CONFIG_F2FS_FS_POSIX_ACL
245 if (test_opt(sbi, POSIX_ACL))
246 seq_puts(seq, ",acl");
248 seq_puts(seq, ",noacl");
250 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
251 seq_puts(seq, ",disable_ext_identify");
253 seq_printf(seq, ",active_logs=%u", sbi->active_logs);
258 static struct super_operations f2fs_sops = {
259 .alloc_inode = f2fs_alloc_inode,
260 .drop_inode = f2fs_drop_inode,
261 .destroy_inode = f2fs_destroy_inode,
262 .write_inode = f2fs_write_inode,
263 .dirty_inode = f2fs_dirty_inode,
264 .show_options = f2fs_show_options,
265 .evict_inode = f2fs_evict_inode,
266 .put_super = f2fs_put_super,
267 .sync_fs = f2fs_sync_fs,
268 .freeze_fs = f2fs_freeze,
269 .unfreeze_fs = f2fs_unfreeze,
270 .statfs = f2fs_statfs,
273 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
274 u64 ino, u32 generation)
276 struct f2fs_sb_info *sbi = F2FS_SB(sb);
279 if (ino < F2FS_ROOT_INO(sbi))
280 return ERR_PTR(-ESTALE);
283 * f2fs_iget isn't quite right if the inode is currently unallocated!
284 * However f2fs_iget currently does appropriate checks to handle stale
285 * inodes so everything is OK.
287 inode = f2fs_iget(sb, ino);
289 return ERR_CAST(inode);
290 if (generation && inode->i_generation != generation) {
291 /* we didn't find the right inode.. */
293 return ERR_PTR(-ESTALE);
298 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
299 int fh_len, int fh_type)
301 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
305 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
306 int fh_len, int fh_type)
308 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
312 static const struct export_operations f2fs_export_ops = {
313 .fh_to_dentry = f2fs_fh_to_dentry,
314 .fh_to_parent = f2fs_fh_to_parent,
315 .get_parent = f2fs_get_parent,
318 static int parse_options(struct super_block *sb, char *options)
320 struct f2fs_sb_info *sbi = F2FS_SB(sb);
321 substring_t args[MAX_OPT_ARGS];
328 while ((p = strsep(&options, ",")) != NULL) {
333 * Initialize args struct so we know whether arg was
334 * found; some options take optional arguments.
336 args[0].to = args[0].from = NULL;
337 token = match_token(p, f2fs_tokens, args);
340 case Opt_gc_background_off:
341 clear_opt(sbi, BG_GC);
343 case Opt_disable_roll_forward:
344 set_opt(sbi, DISABLE_ROLL_FORWARD);
347 set_opt(sbi, DISCARD);
350 set_opt(sbi, NOHEAP);
352 #ifdef CONFIG_F2FS_FS_XATTR
353 case Opt_nouser_xattr:
354 clear_opt(sbi, XATTR_USER);
357 case Opt_nouser_xattr:
358 f2fs_msg(sb, KERN_INFO,
359 "nouser_xattr options not supported");
362 #ifdef CONFIG_F2FS_FS_POSIX_ACL
364 clear_opt(sbi, POSIX_ACL);
368 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
371 case Opt_active_logs:
372 if (args->from && match_int(args, &arg))
374 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
376 sbi->active_logs = arg;
378 case Opt_disable_ext_identify:
379 set_opt(sbi, DISABLE_EXT_IDENTIFY);
382 f2fs_msg(sb, KERN_ERR,
383 "Unrecognized mount option \"%s\" or missing value",
391 static loff_t max_file_size(unsigned bits)
393 loff_t result = ADDRS_PER_INODE;
394 loff_t leaf_count = ADDRS_PER_BLOCK;
396 /* two direct node blocks */
397 result += (leaf_count * 2);
399 /* two indirect node blocks */
400 leaf_count *= NIDS_PER_BLOCK;
401 result += (leaf_count * 2);
403 /* one double indirect node block */
404 leaf_count *= NIDS_PER_BLOCK;
405 result += leaf_count;
411 static int sanity_check_raw_super(struct super_block *sb,
412 struct f2fs_super_block *raw_super)
414 unsigned int blocksize;
416 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
417 f2fs_msg(sb, KERN_INFO,
418 "Magic Mismatch, valid(0x%x) - read(0x%x)",
419 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
423 /* Currently, support only 4KB page cache size */
424 if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
425 f2fs_msg(sb, KERN_INFO,
426 "Invalid page_cache_size (%lu), supports only 4KB\n",
431 /* Currently, support only 4KB block size */
432 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
433 if (blocksize != F2FS_BLKSIZE) {
434 f2fs_msg(sb, KERN_INFO,
435 "Invalid blocksize (%u), supports only 4KB\n",
440 if (le32_to_cpu(raw_super->log_sectorsize) !=
441 F2FS_LOG_SECTOR_SIZE) {
442 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize");
445 if (le32_to_cpu(raw_super->log_sectors_per_block) !=
446 F2FS_LOG_SECTORS_PER_BLOCK) {
447 f2fs_msg(sb, KERN_INFO, "Invalid log sectors per block");
453 static int sanity_check_ckpt(struct f2fs_sb_info *sbi)
455 unsigned int total, fsmeta;
456 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
457 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
459 total = le32_to_cpu(raw_super->segment_count);
460 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
461 fsmeta += le32_to_cpu(raw_super->segment_count_sit);
462 fsmeta += le32_to_cpu(raw_super->segment_count_nat);
463 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
464 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
469 if (is_set_ckpt_flags(ckpt, CP_ERROR_FLAG)) {
470 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
476 static void init_sb_info(struct f2fs_sb_info *sbi)
478 struct f2fs_super_block *raw_super = sbi->raw_super;
481 sbi->log_sectors_per_block =
482 le32_to_cpu(raw_super->log_sectors_per_block);
483 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
484 sbi->blocksize = 1 << sbi->log_blocksize;
485 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
486 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
487 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
488 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
489 sbi->total_sections = le32_to_cpu(raw_super->section_count);
490 sbi->total_node_count =
491 (le32_to_cpu(raw_super->segment_count_nat) / 2)
492 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
493 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
494 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
495 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
496 sbi->cur_victim_sec = NULL_SECNO;
498 for (i = 0; i < NR_COUNT_TYPE; i++)
499 atomic_set(&sbi->nr_pages[i], 0);
502 static int validate_superblock(struct super_block *sb,
503 struct f2fs_super_block **raw_super,
504 struct buffer_head **raw_super_buf, sector_t block)
506 const char *super = (block == 0 ? "first" : "second");
508 /* read f2fs raw super block */
509 *raw_super_buf = sb_bread(sb, block);
510 if (!*raw_super_buf) {
511 f2fs_msg(sb, KERN_ERR, "unable to read %s superblock",
516 *raw_super = (struct f2fs_super_block *)
517 ((char *)(*raw_super_buf)->b_data + F2FS_SUPER_OFFSET);
519 /* sanity checking of raw super */
520 if (!sanity_check_raw_super(sb, *raw_super))
523 f2fs_msg(sb, KERN_ERR, "Can't find a valid F2FS filesystem "
524 "in %s superblock", super);
528 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
530 struct f2fs_sb_info *sbi;
531 struct f2fs_super_block *raw_super;
532 struct buffer_head *raw_super_buf;
537 /* allocate memory for f2fs-specific super block info */
538 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
542 /* set a block size */
543 if (!sb_set_blocksize(sb, F2FS_BLKSIZE)) {
544 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
548 err = validate_superblock(sb, &raw_super, &raw_super_buf, 0);
550 brelse(raw_super_buf);
551 /* check secondary superblock when primary failed */
552 err = validate_superblock(sb, &raw_super, &raw_super_buf, 1);
557 /* init some FS parameters */
558 sbi->active_logs = NR_CURSEG_TYPE;
562 #ifdef CONFIG_F2FS_FS_XATTR
563 set_opt(sbi, XATTR_USER);
565 #ifdef CONFIG_F2FS_FS_POSIX_ACL
566 set_opt(sbi, POSIX_ACL);
568 /* parse mount options */
569 err = parse_options(sb, (char *)data);
573 sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
574 sb->s_max_links = F2FS_LINK_MAX;
575 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
577 sb->s_op = &f2fs_sops;
578 sb->s_xattr = f2fs_xattr_handlers;
579 sb->s_export_op = &f2fs_export_ops;
580 sb->s_magic = F2FS_SUPER_MAGIC;
582 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
583 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
584 memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
586 /* init f2fs-specific super block info */
588 sbi->raw_super = raw_super;
589 sbi->raw_super_buf = raw_super_buf;
590 mutex_init(&sbi->gc_mutex);
591 mutex_init(&sbi->writepages);
592 mutex_init(&sbi->cp_mutex);
593 for (i = 0; i < NR_GLOBAL_LOCKS; i++)
594 mutex_init(&sbi->fs_lock[i]);
595 mutex_init(&sbi->node_write);
597 spin_lock_init(&sbi->stat_lock);
598 init_rwsem(&sbi->bio_sem);
601 /* get an inode for meta space */
602 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
603 if (IS_ERR(sbi->meta_inode)) {
604 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
605 err = PTR_ERR(sbi->meta_inode);
609 err = get_valid_checkpoint(sbi);
611 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
612 goto free_meta_inode;
615 /* sanity checking of checkpoint */
617 if (sanity_check_ckpt(sbi)) {
618 f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint");
622 sbi->total_valid_node_count =
623 le32_to_cpu(sbi->ckpt->valid_node_count);
624 sbi->total_valid_inode_count =
625 le32_to_cpu(sbi->ckpt->valid_inode_count);
626 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
627 sbi->total_valid_block_count =
628 le64_to_cpu(sbi->ckpt->valid_block_count);
629 sbi->last_valid_block_count = sbi->total_valid_block_count;
630 sbi->alloc_valid_block_count = 0;
631 INIT_LIST_HEAD(&sbi->dir_inode_list);
632 spin_lock_init(&sbi->dir_inode_lock);
634 init_orphan_info(sbi);
636 /* setup f2fs internal modules */
637 err = build_segment_manager(sbi);
639 f2fs_msg(sb, KERN_ERR,
640 "Failed to initialize F2FS segment manager");
643 err = build_node_manager(sbi);
645 f2fs_msg(sb, KERN_ERR,
646 "Failed to initialize F2FS node manager");
650 build_gc_manager(sbi);
652 /* get an inode for node space */
653 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
654 if (IS_ERR(sbi->node_inode)) {
655 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
656 err = PTR_ERR(sbi->node_inode);
660 /* if there are nt orphan nodes free them */
662 if (recover_orphan_inodes(sbi))
663 goto free_node_inode;
665 /* read root inode and dentry */
666 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
668 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
670 goto free_node_inode;
672 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size)
673 goto free_root_inode;
675 sb->s_root = d_make_root(root); /* allocate root dentry */
678 goto free_root_inode;
681 /* recover fsynced data */
682 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
683 err = recover_fsync_data(sbi);
685 f2fs_msg(sb, KERN_ERR,
686 "Cannot recover all fsync data errno=%ld", err);
689 /* After POR, we can run background GC thread */
690 err = start_gc_thread(sbi);
694 err = f2fs_build_stats(sbi);
698 if (test_opt(sbi, DISCARD)) {
699 struct request_queue *q = bdev_get_queue(sb->s_bdev);
700 if (!blk_queue_discard(q))
701 f2fs_msg(sb, KERN_WARNING,
702 "mounting with \"discard\" option, but "
703 "the device does not support discard");
713 iput(sbi->node_inode);
715 destroy_node_manager(sbi);
717 destroy_segment_manager(sbi);
721 make_bad_inode(sbi->meta_inode);
722 iput(sbi->meta_inode);
724 brelse(raw_super_buf);
730 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
731 const char *dev_name, void *data)
733 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
736 static struct file_system_type f2fs_fs_type = {
737 .owner = THIS_MODULE,
740 .kill_sb = kill_block_super,
741 .fs_flags = FS_REQUIRES_DEV,
743 MODULE_ALIAS_FS("f2fs");
745 static int __init init_inodecache(void)
747 f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
748 sizeof(struct f2fs_inode_info), NULL);
749 if (f2fs_inode_cachep == NULL)
754 static void destroy_inodecache(void)
757 * Make sure all delayed rcu free inodes are flushed before we
761 kmem_cache_destroy(f2fs_inode_cachep);
764 static int __init init_f2fs_fs(void)
768 err = init_inodecache();
771 err = create_node_manager_caches();
774 err = create_gc_caches();
777 err = create_checkpoint_caches();
780 err = register_filesystem(&f2fs_fs_type);
783 f2fs_create_root_stats();
788 static void __exit exit_f2fs_fs(void)
790 f2fs_destroy_root_stats();
791 unregister_filesystem(&f2fs_fs_type);
792 destroy_checkpoint_caches();
794 destroy_node_manager_caches();
795 destroy_inodecache();
798 module_init(init_f2fs_fs)
799 module_exit(exit_f2fs_fs)
801 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
802 MODULE_DESCRIPTION("Flash Friendly File System");
803 MODULE_LICENSE("GPL");