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>
35 #define CREATE_TRACE_POINTS
36 #include <trace/events/f2fs.h>
38 static struct proc_dir_entry *f2fs_proc_root;
39 static struct kmem_cache *f2fs_inode_cachep;
40 static struct kset *f2fs_kset;
44 Opt_disable_roll_forward,
53 Opt_disable_ext_identify,
63 static match_table_t f2fs_tokens = {
64 {Opt_gc_background, "background_gc=%s"},
65 {Opt_disable_roll_forward, "disable_roll_forward"},
66 {Opt_norecovery, "norecovery"},
67 {Opt_discard, "discard"},
68 {Opt_noheap, "no_heap"},
69 {Opt_user_xattr, "user_xattr"},
70 {Opt_nouser_xattr, "nouser_xattr"},
73 {Opt_active_logs, "active_logs=%u"},
74 {Opt_disable_ext_identify, "disable_ext_identify"},
75 {Opt_inline_xattr, "inline_xattr"},
76 {Opt_inline_data, "inline_data"},
77 {Opt_inline_dentry, "inline_dentry"},
78 {Opt_flush_merge, "flush_merge"},
79 {Opt_nobarrier, "nobarrier"},
80 {Opt_fastboot, "fastboot"},
84 /* Sysfs support for f2fs */
86 GC_THREAD, /* struct f2fs_gc_thread */
87 SM_INFO, /* struct f2fs_sm_info */
88 NM_INFO, /* struct f2fs_nm_info */
89 F2FS_SBI, /* struct f2fs_sb_info */
93 struct attribute attr;
94 ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
95 ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
96 const char *, size_t);
101 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
103 if (struct_type == GC_THREAD)
104 return (unsigned char *)sbi->gc_thread;
105 else if (struct_type == SM_INFO)
106 return (unsigned char *)SM_I(sbi);
107 else if (struct_type == NM_INFO)
108 return (unsigned char *)NM_I(sbi);
109 else if (struct_type == F2FS_SBI)
110 return (unsigned char *)sbi;
114 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
115 struct f2fs_sb_info *sbi, char *buf)
117 unsigned char *ptr = NULL;
120 ptr = __struct_ptr(sbi, a->struct_type);
124 ui = (unsigned int *)(ptr + a->offset);
126 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
129 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
130 struct f2fs_sb_info *sbi,
131 const char *buf, size_t count)
138 ptr = __struct_ptr(sbi, a->struct_type);
142 ui = (unsigned int *)(ptr + a->offset);
144 ret = kstrtoul(skip_spaces(buf), 0, &t);
151 static ssize_t f2fs_attr_show(struct kobject *kobj,
152 struct attribute *attr, char *buf)
154 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
156 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
158 return a->show ? a->show(a, sbi, buf) : 0;
161 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
162 const char *buf, size_t len)
164 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
166 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
168 return a->store ? a->store(a, sbi, buf, len) : 0;
171 static void f2fs_sb_release(struct kobject *kobj)
173 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
175 complete(&sbi->s_kobj_unregister);
178 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
179 static struct f2fs_attr f2fs_attr_##_name = { \
180 .attr = {.name = __stringify(_name), .mode = _mode }, \
183 .struct_type = _struct_type, \
187 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
188 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
189 f2fs_sbi_show, f2fs_sbi_store, \
190 offsetof(struct struct_name, elname))
192 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
193 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
194 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
195 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
196 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
197 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
198 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
199 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
200 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks);
201 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
202 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
203 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
205 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
206 static struct attribute *f2fs_attrs[] = {
207 ATTR_LIST(gc_min_sleep_time),
208 ATTR_LIST(gc_max_sleep_time),
209 ATTR_LIST(gc_no_gc_sleep_time),
211 ATTR_LIST(reclaim_segments),
212 ATTR_LIST(max_small_discards),
213 ATTR_LIST(ipu_policy),
214 ATTR_LIST(min_ipu_util),
215 ATTR_LIST(min_fsync_blocks),
216 ATTR_LIST(max_victim_search),
217 ATTR_LIST(dir_level),
218 ATTR_LIST(ram_thresh),
222 static const struct sysfs_ops f2fs_attr_ops = {
223 .show = f2fs_attr_show,
224 .store = f2fs_attr_store,
227 static struct kobj_type f2fs_ktype = {
228 .default_attrs = f2fs_attrs,
229 .sysfs_ops = &f2fs_attr_ops,
230 .release = f2fs_sb_release,
233 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
235 struct va_format vaf;
241 printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
245 static void init_once(void *foo)
247 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
249 inode_init_once(&fi->vfs_inode);
252 static int parse_options(struct super_block *sb, char *options)
254 struct f2fs_sb_info *sbi = F2FS_SB(sb);
255 substring_t args[MAX_OPT_ARGS];
262 while ((p = strsep(&options, ",")) != NULL) {
267 * Initialize args struct so we know whether arg was
268 * found; some options take optional arguments.
270 args[0].to = args[0].from = NULL;
271 token = match_token(p, f2fs_tokens, args);
274 case Opt_gc_background:
275 name = match_strdup(&args[0]);
279 if (strlen(name) == 2 && !strncmp(name, "on", 2))
281 else if (strlen(name) == 3 && !strncmp(name, "off", 3))
282 clear_opt(sbi, BG_GC);
289 case Opt_disable_roll_forward:
290 set_opt(sbi, DISABLE_ROLL_FORWARD);
293 /* this option mounts f2fs with ro */
294 set_opt(sbi, DISABLE_ROLL_FORWARD);
295 if (!f2fs_readonly(sb))
299 set_opt(sbi, DISCARD);
302 set_opt(sbi, NOHEAP);
304 #ifdef CONFIG_F2FS_FS_XATTR
306 set_opt(sbi, XATTR_USER);
308 case Opt_nouser_xattr:
309 clear_opt(sbi, XATTR_USER);
311 case Opt_inline_xattr:
312 set_opt(sbi, INLINE_XATTR);
316 f2fs_msg(sb, KERN_INFO,
317 "user_xattr options not supported");
319 case Opt_nouser_xattr:
320 f2fs_msg(sb, KERN_INFO,
321 "nouser_xattr options not supported");
323 case Opt_inline_xattr:
324 f2fs_msg(sb, KERN_INFO,
325 "inline_xattr options not supported");
328 #ifdef CONFIG_F2FS_FS_POSIX_ACL
330 set_opt(sbi, POSIX_ACL);
333 clear_opt(sbi, POSIX_ACL);
337 f2fs_msg(sb, KERN_INFO, "acl options not supported");
340 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
343 case Opt_active_logs:
344 if (args->from && match_int(args, &arg))
346 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
348 sbi->active_logs = arg;
350 case Opt_disable_ext_identify:
351 set_opt(sbi, DISABLE_EXT_IDENTIFY);
353 case Opt_inline_data:
354 set_opt(sbi, INLINE_DATA);
356 case Opt_inline_dentry:
357 set_opt(sbi, INLINE_DENTRY);
359 case Opt_flush_merge:
360 set_opt(sbi, FLUSH_MERGE);
363 set_opt(sbi, NOBARRIER);
366 set_opt(sbi, FASTBOOT);
369 f2fs_msg(sb, KERN_ERR,
370 "Unrecognized mount option \"%s\" or missing value",
378 static struct inode *f2fs_alloc_inode(struct super_block *sb)
380 struct f2fs_inode_info *fi;
382 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
386 init_once((void *) fi);
388 /* Initialize f2fs-specific inode info */
389 fi->vfs_inode.i_version = 1;
390 atomic_set(&fi->dirty_pages, 0);
391 fi->i_current_depth = 1;
393 rwlock_init(&fi->ext.ext_lock);
394 init_rwsem(&fi->i_sem);
395 INIT_RADIX_TREE(&fi->inmem_root, GFP_NOFS);
396 INIT_LIST_HEAD(&fi->inmem_pages);
397 mutex_init(&fi->inmem_lock);
399 set_inode_flag(fi, FI_NEW_INODE);
401 if (test_opt(F2FS_SB(sb), INLINE_XATTR))
402 set_inode_flag(fi, FI_INLINE_XATTR);
404 /* Will be used by directory only */
405 fi->i_dir_level = F2FS_SB(sb)->dir_level;
407 return &fi->vfs_inode;
410 static int f2fs_drop_inode(struct inode *inode)
413 * This is to avoid a deadlock condition like below.
414 * writeback_single_inode(inode)
415 * - f2fs_write_data_page
416 * - f2fs_gc -> iput -> evict
417 * - inode_wait_for_writeback(inode)
419 if (!inode_unhashed(inode) && inode->i_state & I_SYNC)
421 return generic_drop_inode(inode);
425 * f2fs_dirty_inode() is called from __mark_inode_dirty()
427 * We should call set_dirty_inode to write the dirty inode through write_inode.
429 static void f2fs_dirty_inode(struct inode *inode, int flags)
431 set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
434 static void f2fs_i_callback(struct rcu_head *head)
436 struct inode *inode = container_of(head, struct inode, i_rcu);
437 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
440 static void f2fs_destroy_inode(struct inode *inode)
442 call_rcu(&inode->i_rcu, f2fs_i_callback);
445 static void f2fs_put_super(struct super_block *sb)
447 struct f2fs_sb_info *sbi = F2FS_SB(sb);
450 remove_proc_entry("segment_info", sbi->s_proc);
451 remove_proc_entry(sb->s_id, f2fs_proc_root);
453 kobject_del(&sbi->s_kobj);
455 f2fs_destroy_stats(sbi);
459 * We don't need to do checkpoint when superblock is clean.
460 * But, the previous checkpoint was not done by umount, it needs to do
461 * clean checkpoint again.
463 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
464 !is_set_ckpt_flags(F2FS_CKPT(sbi), CP_UMOUNT_FLAG)) {
465 struct cp_control cpc = {
468 write_checkpoint(sbi, &cpc);
472 * normally superblock is clean, so we need to release this.
473 * In addition, EIO will skip do checkpoint, we need this as well.
475 release_dirty_inode(sbi);
476 release_discard_addrs(sbi);
478 iput(sbi->node_inode);
479 iput(sbi->meta_inode);
481 /* destroy f2fs internal modules */
482 destroy_node_manager(sbi);
483 destroy_segment_manager(sbi);
486 kobject_put(&sbi->s_kobj);
487 wait_for_completion(&sbi->s_kobj_unregister);
489 sb->s_fs_info = NULL;
490 brelse(sbi->raw_super_buf);
494 int f2fs_sync_fs(struct super_block *sb, int sync)
496 struct f2fs_sb_info *sbi = F2FS_SB(sb);
498 trace_f2fs_sync_fs(sb, sync);
501 struct cp_control cpc;
503 cpc.reason = __get_cp_reason(sbi);
505 mutex_lock(&sbi->gc_mutex);
506 write_checkpoint(sbi, &cpc);
507 mutex_unlock(&sbi->gc_mutex);
509 f2fs_balance_fs(sbi);
511 f2fs_trace_ios(NULL, NULL, 1);
516 static int f2fs_freeze(struct super_block *sb)
520 if (f2fs_readonly(sb))
523 err = f2fs_sync_fs(sb, 1);
527 static int f2fs_unfreeze(struct super_block *sb)
532 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
534 struct super_block *sb = dentry->d_sb;
535 struct f2fs_sb_info *sbi = F2FS_SB(sb);
536 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
537 block_t total_count, user_block_count, start_count, ovp_count;
539 total_count = le64_to_cpu(sbi->raw_super->block_count);
540 user_block_count = sbi->user_block_count;
541 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
542 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
543 buf->f_type = F2FS_SUPER_MAGIC;
544 buf->f_bsize = sbi->blocksize;
546 buf->f_blocks = total_count - start_count;
547 buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
548 buf->f_bavail = user_block_count - valid_user_blocks(sbi);
550 buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
551 buf->f_ffree = buf->f_files - valid_inode_count(sbi);
553 buf->f_namelen = F2FS_NAME_LEN;
554 buf->f_fsid.val[0] = (u32)id;
555 buf->f_fsid.val[1] = (u32)(id >> 32);
560 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
562 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
564 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC))
565 seq_printf(seq, ",background_gc=%s", "on");
567 seq_printf(seq, ",background_gc=%s", "off");
568 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
569 seq_puts(seq, ",disable_roll_forward");
570 if (test_opt(sbi, DISCARD))
571 seq_puts(seq, ",discard");
572 if (test_opt(sbi, NOHEAP))
573 seq_puts(seq, ",no_heap_alloc");
574 #ifdef CONFIG_F2FS_FS_XATTR
575 if (test_opt(sbi, XATTR_USER))
576 seq_puts(seq, ",user_xattr");
578 seq_puts(seq, ",nouser_xattr");
579 if (test_opt(sbi, INLINE_XATTR))
580 seq_puts(seq, ",inline_xattr");
582 #ifdef CONFIG_F2FS_FS_POSIX_ACL
583 if (test_opt(sbi, POSIX_ACL))
584 seq_puts(seq, ",acl");
586 seq_puts(seq, ",noacl");
588 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
589 seq_puts(seq, ",disable_ext_identify");
590 if (test_opt(sbi, INLINE_DATA))
591 seq_puts(seq, ",inline_data");
592 if (test_opt(sbi, INLINE_DENTRY))
593 seq_puts(seq, ",inline_dentry");
594 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
595 seq_puts(seq, ",flush_merge");
596 if (test_opt(sbi, NOBARRIER))
597 seq_puts(seq, ",nobarrier");
598 if (test_opt(sbi, FASTBOOT))
599 seq_puts(seq, ",fastboot");
600 seq_printf(seq, ",active_logs=%u", sbi->active_logs);
605 static int segment_info_seq_show(struct seq_file *seq, void *offset)
607 struct super_block *sb = seq->private;
608 struct f2fs_sb_info *sbi = F2FS_SB(sb);
609 unsigned int total_segs =
610 le32_to_cpu(sbi->raw_super->segment_count_main);
613 seq_puts(seq, "format: segment_type|valid_blocks\n"
614 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
616 for (i = 0; i < total_segs; i++) {
617 struct seg_entry *se = get_seg_entry(sbi, i);
620 seq_printf(seq, "%-5d", i);
621 seq_printf(seq, "%d|%-3u", se->type,
622 get_valid_blocks(sbi, i, 1));
623 if ((i % 10) == 9 || i == (total_segs - 1))
632 static int segment_info_open_fs(struct inode *inode, struct file *file)
634 return single_open(file, segment_info_seq_show, PDE_DATA(inode));
637 static const struct file_operations f2fs_seq_segment_info_fops = {
638 .owner = THIS_MODULE,
639 .open = segment_info_open_fs,
642 .release = single_release,
645 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
647 struct f2fs_sb_info *sbi = F2FS_SB(sb);
648 struct f2fs_mount_info org_mount_opt;
649 int err, active_logs;
650 bool need_restart_gc = false;
651 bool need_stop_gc = false;
656 * Save the old mount options in case we
657 * need to restore them.
659 org_mount_opt = sbi->mount_opt;
660 active_logs = sbi->active_logs;
662 sbi->mount_opt.opt = 0;
663 sbi->active_logs = NR_CURSEG_TYPE;
665 /* parse mount options */
666 err = parse_options(sb, data);
671 * Previous and new state of filesystem is RO,
672 * so skip checking GC and FLUSH_MERGE conditions.
674 if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
678 * We stop the GC thread if FS is mounted as RO
679 * or if background_gc = off is passed in mount
680 * option. Also sync the filesystem.
682 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
683 if (sbi->gc_thread) {
686 need_restart_gc = true;
688 } else if (!sbi->gc_thread) {
689 err = start_gc_thread(sbi);
696 * We stop issue flush thread if FS is mounted as RO
697 * or if flush_merge is not passed in mount option.
699 if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
700 destroy_flush_cmd_control(sbi);
701 } else if (!SM_I(sbi)->cmd_control_info) {
702 err = create_flush_cmd_control(sbi);
707 /* Update the POSIXACL Flag */
708 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
709 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
712 if (need_restart_gc) {
713 if (start_gc_thread(sbi))
714 f2fs_msg(sbi->sb, KERN_WARNING,
715 "background gc thread has stopped");
716 } else if (need_stop_gc) {
720 sbi->mount_opt = org_mount_opt;
721 sbi->active_logs = active_logs;
725 static struct super_operations f2fs_sops = {
726 .alloc_inode = f2fs_alloc_inode,
727 .drop_inode = f2fs_drop_inode,
728 .destroy_inode = f2fs_destroy_inode,
729 .write_inode = f2fs_write_inode,
730 .dirty_inode = f2fs_dirty_inode,
731 .show_options = f2fs_show_options,
732 .evict_inode = f2fs_evict_inode,
733 .put_super = f2fs_put_super,
734 .sync_fs = f2fs_sync_fs,
735 .freeze_fs = f2fs_freeze,
736 .unfreeze_fs = f2fs_unfreeze,
737 .statfs = f2fs_statfs,
738 .remount_fs = f2fs_remount,
741 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
742 u64 ino, u32 generation)
744 struct f2fs_sb_info *sbi = F2FS_SB(sb);
747 if (check_nid_range(sbi, ino))
748 return ERR_PTR(-ESTALE);
751 * f2fs_iget isn't quite right if the inode is currently unallocated!
752 * However f2fs_iget currently does appropriate checks to handle stale
753 * inodes so everything is OK.
755 inode = f2fs_iget(sb, ino);
757 return ERR_CAST(inode);
758 if (unlikely(generation && inode->i_generation != generation)) {
759 /* we didn't find the right inode.. */
761 return ERR_PTR(-ESTALE);
766 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
767 int fh_len, int fh_type)
769 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
773 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
774 int fh_len, int fh_type)
776 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
780 static const struct export_operations f2fs_export_ops = {
781 .fh_to_dentry = f2fs_fh_to_dentry,
782 .fh_to_parent = f2fs_fh_to_parent,
783 .get_parent = f2fs_get_parent,
786 static loff_t max_file_size(unsigned bits)
788 loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
789 loff_t leaf_count = ADDRS_PER_BLOCK;
791 /* two direct node blocks */
792 result += (leaf_count * 2);
794 /* two indirect node blocks */
795 leaf_count *= NIDS_PER_BLOCK;
796 result += (leaf_count * 2);
798 /* one double indirect node block */
799 leaf_count *= NIDS_PER_BLOCK;
800 result += leaf_count;
806 static int sanity_check_raw_super(struct super_block *sb,
807 struct f2fs_super_block *raw_super)
809 unsigned int blocksize;
811 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
812 f2fs_msg(sb, KERN_INFO,
813 "Magic Mismatch, valid(0x%x) - read(0x%x)",
814 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
818 /* Currently, support only 4KB page cache size */
819 if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
820 f2fs_msg(sb, KERN_INFO,
821 "Invalid page_cache_size (%lu), supports only 4KB\n",
826 /* Currently, support only 4KB block size */
827 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
828 if (blocksize != F2FS_BLKSIZE) {
829 f2fs_msg(sb, KERN_INFO,
830 "Invalid blocksize (%u), supports only 4KB\n",
835 /* Currently, support 512/1024/2048/4096 bytes sector size */
836 if (le32_to_cpu(raw_super->log_sectorsize) >
837 F2FS_MAX_LOG_SECTOR_SIZE ||
838 le32_to_cpu(raw_super->log_sectorsize) <
839 F2FS_MIN_LOG_SECTOR_SIZE) {
840 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
841 le32_to_cpu(raw_super->log_sectorsize));
844 if (le32_to_cpu(raw_super->log_sectors_per_block) +
845 le32_to_cpu(raw_super->log_sectorsize) !=
846 F2FS_MAX_LOG_SECTOR_SIZE) {
847 f2fs_msg(sb, KERN_INFO,
848 "Invalid log sectors per block(%u) log sectorsize(%u)",
849 le32_to_cpu(raw_super->log_sectors_per_block),
850 le32_to_cpu(raw_super->log_sectorsize));
856 static int sanity_check_ckpt(struct f2fs_sb_info *sbi)
858 unsigned int total, fsmeta;
859 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
860 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
862 total = le32_to_cpu(raw_super->segment_count);
863 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
864 fsmeta += le32_to_cpu(raw_super->segment_count_sit);
865 fsmeta += le32_to_cpu(raw_super->segment_count_nat);
866 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
867 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
869 if (unlikely(fsmeta >= total))
872 if (unlikely(f2fs_cp_error(sbi))) {
873 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
879 static void init_sb_info(struct f2fs_sb_info *sbi)
881 struct f2fs_super_block *raw_super = sbi->raw_super;
884 sbi->log_sectors_per_block =
885 le32_to_cpu(raw_super->log_sectors_per_block);
886 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
887 sbi->blocksize = 1 << sbi->log_blocksize;
888 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
889 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
890 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
891 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
892 sbi->total_sections = le32_to_cpu(raw_super->section_count);
893 sbi->total_node_count =
894 (le32_to_cpu(raw_super->segment_count_nat) / 2)
895 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
896 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
897 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
898 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
899 sbi->cur_victim_sec = NULL_SECNO;
900 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
902 for (i = 0; i < NR_COUNT_TYPE; i++)
903 atomic_set(&sbi->nr_pages[i], 0);
905 sbi->dir_level = DEF_DIR_LEVEL;
906 clear_sbi_flag(sbi, SBI_NEED_FSCK);
910 * Read f2fs raw super block.
911 * Because we have two copies of super block, so read the first one at first,
912 * if the first one is invalid, move to read the second one.
914 static int read_raw_super_block(struct super_block *sb,
915 struct f2fs_super_block **raw_super,
916 struct buffer_head **raw_super_buf)
921 *raw_super_buf = sb_bread(sb, block);
922 if (!*raw_super_buf) {
923 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
933 *raw_super = (struct f2fs_super_block *)
934 ((char *)(*raw_super_buf)->b_data + F2FS_SUPER_OFFSET);
936 /* sanity checking of raw super */
937 if (sanity_check_raw_super(sb, *raw_super)) {
938 brelse(*raw_super_buf);
939 f2fs_msg(sb, KERN_ERR,
940 "Can't find valid F2FS filesystem in %dth superblock",
953 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
955 struct f2fs_sb_info *sbi;
956 struct f2fs_super_block *raw_super = NULL;
957 struct buffer_head *raw_super_buf;
961 char *options = NULL;
965 /* allocate memory for f2fs-specific super block info */
966 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
970 /* set a block size */
971 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
972 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
976 err = read_raw_super_block(sb, &raw_super, &raw_super_buf);
981 /* init some FS parameters */
982 sbi->active_logs = NR_CURSEG_TYPE;
986 #ifdef CONFIG_F2FS_FS_XATTR
987 set_opt(sbi, XATTR_USER);
989 #ifdef CONFIG_F2FS_FS_POSIX_ACL
990 set_opt(sbi, POSIX_ACL);
992 /* parse mount options */
993 options = kstrdup((const char *)data, GFP_KERNEL);
994 if (data && !options) {
999 err = parse_options(sb, options);
1003 sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
1004 sb->s_max_links = F2FS_LINK_MAX;
1005 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1007 sb->s_op = &f2fs_sops;
1008 sb->s_xattr = f2fs_xattr_handlers;
1009 sb->s_export_op = &f2fs_export_ops;
1010 sb->s_magic = F2FS_SUPER_MAGIC;
1011 sb->s_time_gran = 1;
1012 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1013 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1014 memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1016 /* init f2fs-specific super block info */
1018 sbi->raw_super = raw_super;
1019 sbi->raw_super_buf = raw_super_buf;
1020 mutex_init(&sbi->gc_mutex);
1021 mutex_init(&sbi->writepages);
1022 mutex_init(&sbi->cp_mutex);
1023 init_rwsem(&sbi->node_write);
1024 clear_sbi_flag(sbi, SBI_POR_DOING);
1025 spin_lock_init(&sbi->stat_lock);
1027 init_rwsem(&sbi->read_io.io_rwsem);
1028 sbi->read_io.sbi = sbi;
1029 sbi->read_io.bio = NULL;
1030 for (i = 0; i < NR_PAGE_TYPE; i++) {
1031 init_rwsem(&sbi->write_io[i].io_rwsem);
1032 sbi->write_io[i].sbi = sbi;
1033 sbi->write_io[i].bio = NULL;
1036 init_rwsem(&sbi->cp_rwsem);
1037 init_waitqueue_head(&sbi->cp_wait);
1040 /* get an inode for meta space */
1041 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1042 if (IS_ERR(sbi->meta_inode)) {
1043 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1044 err = PTR_ERR(sbi->meta_inode);
1048 err = get_valid_checkpoint(sbi);
1050 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1051 goto free_meta_inode;
1054 /* sanity checking of checkpoint */
1056 if (sanity_check_ckpt(sbi)) {
1057 f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint");
1061 sbi->total_valid_node_count =
1062 le32_to_cpu(sbi->ckpt->valid_node_count);
1063 sbi->total_valid_inode_count =
1064 le32_to_cpu(sbi->ckpt->valid_inode_count);
1065 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1066 sbi->total_valid_block_count =
1067 le64_to_cpu(sbi->ckpt->valid_block_count);
1068 sbi->last_valid_block_count = sbi->total_valid_block_count;
1069 sbi->alloc_valid_block_count = 0;
1070 INIT_LIST_HEAD(&sbi->dir_inode_list);
1071 spin_lock_init(&sbi->dir_inode_lock);
1073 init_ino_entry_info(sbi);
1075 /* setup f2fs internal modules */
1076 err = build_segment_manager(sbi);
1078 f2fs_msg(sb, KERN_ERR,
1079 "Failed to initialize F2FS segment manager");
1082 err = build_node_manager(sbi);
1084 f2fs_msg(sb, KERN_ERR,
1085 "Failed to initialize F2FS node manager");
1089 build_gc_manager(sbi);
1091 /* get an inode for node space */
1092 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1093 if (IS_ERR(sbi->node_inode)) {
1094 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1095 err = PTR_ERR(sbi->node_inode);
1099 /* if there are nt orphan nodes free them */
1100 recover_orphan_inodes(sbi);
1102 /* read root inode and dentry */
1103 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1105 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1106 err = PTR_ERR(root);
1107 goto free_node_inode;
1109 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1112 goto free_node_inode;
1115 sb->s_root = d_make_root(root); /* allocate root dentry */
1118 goto free_root_inode;
1121 err = f2fs_build_stats(sbi);
1123 goto free_root_inode;
1126 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1129 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1130 &f2fs_seq_segment_info_fops, sb);
1132 if (test_opt(sbi, DISCARD)) {
1133 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1134 if (!blk_queue_discard(q))
1135 f2fs_msg(sb, KERN_WARNING,
1136 "mounting with \"discard\" option, but "
1137 "the device does not support discard");
1140 sbi->s_kobj.kset = f2fs_kset;
1141 init_completion(&sbi->s_kobj_unregister);
1142 err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1148 set_sbi_flag(sbi, SBI_NEED_FSCK);
1150 /* recover fsynced data */
1151 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1152 err = recover_fsync_data(sbi);
1154 f2fs_msg(sb, KERN_ERR,
1155 "Cannot recover all fsync data errno=%ld", err);
1161 * If filesystem is not mounted as read-only then
1162 * do start the gc_thread.
1164 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
1165 /* After POR, we can run background GC thread.*/
1166 err = start_gc_thread(sbi);
1174 kobject_del(&sbi->s_kobj);
1177 remove_proc_entry("segment_info", sbi->s_proc);
1178 remove_proc_entry(sb->s_id, f2fs_proc_root);
1180 f2fs_destroy_stats(sbi);
1185 iput(sbi->node_inode);
1187 destroy_node_manager(sbi);
1189 destroy_segment_manager(sbi);
1193 make_bad_inode(sbi->meta_inode);
1194 iput(sbi->meta_inode);
1198 brelse(raw_super_buf);
1202 /* give only one another chance */
1205 shrink_dcache_sb(sb);
1211 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1212 const char *dev_name, void *data)
1214 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1217 static void kill_f2fs_super(struct super_block *sb)
1220 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
1221 kill_block_super(sb);
1224 static struct file_system_type f2fs_fs_type = {
1225 .owner = THIS_MODULE,
1227 .mount = f2fs_mount,
1228 .kill_sb = kill_f2fs_super,
1229 .fs_flags = FS_REQUIRES_DEV,
1231 MODULE_ALIAS_FS("f2fs");
1233 static int __init init_inodecache(void)
1235 f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
1236 sizeof(struct f2fs_inode_info));
1237 if (!f2fs_inode_cachep)
1242 static void destroy_inodecache(void)
1245 * Make sure all delayed rcu free inodes are flushed before we
1249 kmem_cache_destroy(f2fs_inode_cachep);
1252 static int __init init_f2fs_fs(void)
1256 f2fs_build_trace_ios();
1258 err = init_inodecache();
1261 err = create_node_manager_caches();
1263 goto free_inodecache;
1264 err = create_segment_manager_caches();
1266 goto free_node_manager_caches;
1267 err = create_checkpoint_caches();
1269 goto free_segment_manager_caches;
1270 f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1273 goto free_checkpoint_caches;
1275 err = register_filesystem(&f2fs_fs_type);
1278 f2fs_create_root_stats();
1279 f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1283 kset_unregister(f2fs_kset);
1284 free_checkpoint_caches:
1285 destroy_checkpoint_caches();
1286 free_segment_manager_caches:
1287 destroy_segment_manager_caches();
1288 free_node_manager_caches:
1289 destroy_node_manager_caches();
1291 destroy_inodecache();
1296 static void __exit exit_f2fs_fs(void)
1298 remove_proc_entry("fs/f2fs", NULL);
1299 f2fs_destroy_root_stats();
1300 unregister_filesystem(&f2fs_fs_type);
1301 destroy_checkpoint_caches();
1302 destroy_segment_manager_caches();
1303 destroy_node_manager_caches();
1304 destroy_inodecache();
1305 kset_unregister(f2fs_kset);
1306 f2fs_destroy_trace_ios();
1309 module_init(init_f2fs_fs)
1310 module_exit(exit_f2fs_fs)
1312 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1313 MODULE_DESCRIPTION("Flash Friendly File System");
1314 MODULE_LICENSE("GPL");