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,
64 static match_table_t f2fs_tokens = {
65 {Opt_gc_background, "background_gc=%s"},
66 {Opt_disable_roll_forward, "disable_roll_forward"},
67 {Opt_norecovery, "norecovery"},
68 {Opt_discard, "discard"},
69 {Opt_noheap, "no_heap"},
70 {Opt_user_xattr, "user_xattr"},
71 {Opt_nouser_xattr, "nouser_xattr"},
74 {Opt_active_logs, "active_logs=%u"},
75 {Opt_disable_ext_identify, "disable_ext_identify"},
76 {Opt_inline_xattr, "inline_xattr"},
77 {Opt_inline_data, "inline_data"},
78 {Opt_inline_dentry, "inline_dentry"},
79 {Opt_flush_merge, "flush_merge"},
80 {Opt_nobarrier, "nobarrier"},
81 {Opt_fastboot, "fastboot"},
82 {Opt_extent_cache, "extent_cache"},
86 /* Sysfs support for f2fs */
88 GC_THREAD, /* struct f2fs_gc_thread */
89 SM_INFO, /* struct f2fs_sm_info */
90 NM_INFO, /* struct f2fs_nm_info */
91 F2FS_SBI, /* struct f2fs_sb_info */
95 struct attribute attr;
96 ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
97 ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
98 const char *, size_t);
103 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
105 if (struct_type == GC_THREAD)
106 return (unsigned char *)sbi->gc_thread;
107 else if (struct_type == SM_INFO)
108 return (unsigned char *)SM_I(sbi);
109 else if (struct_type == NM_INFO)
110 return (unsigned char *)NM_I(sbi);
111 else if (struct_type == F2FS_SBI)
112 return (unsigned char *)sbi;
116 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
117 struct f2fs_sb_info *sbi, char *buf)
119 unsigned char *ptr = NULL;
122 ptr = __struct_ptr(sbi, a->struct_type);
126 ui = (unsigned int *)(ptr + a->offset);
128 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
131 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
132 struct f2fs_sb_info *sbi,
133 const char *buf, size_t count)
140 ptr = __struct_ptr(sbi, a->struct_type);
144 ui = (unsigned int *)(ptr + a->offset);
146 ret = kstrtoul(skip_spaces(buf), 0, &t);
153 static ssize_t f2fs_attr_show(struct kobject *kobj,
154 struct attribute *attr, char *buf)
156 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
158 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
160 return a->show ? a->show(a, sbi, buf) : 0;
163 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
164 const char *buf, size_t len)
166 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
168 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
170 return a->store ? a->store(a, sbi, buf, len) : 0;
173 static void f2fs_sb_release(struct kobject *kobj)
175 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
177 complete(&sbi->s_kobj_unregister);
180 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
181 static struct f2fs_attr f2fs_attr_##_name = { \
182 .attr = {.name = __stringify(_name), .mode = _mode }, \
185 .struct_type = _struct_type, \
189 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
190 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
191 f2fs_sbi_show, f2fs_sbi_store, \
192 offsetof(struct struct_name, elname))
194 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
195 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
196 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
197 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
198 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
199 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
200 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, batched_trim_sections, trim_sections);
201 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
202 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
203 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks);
204 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
205 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
206 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
208 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
209 static struct attribute *f2fs_attrs[] = {
210 ATTR_LIST(gc_min_sleep_time),
211 ATTR_LIST(gc_max_sleep_time),
212 ATTR_LIST(gc_no_gc_sleep_time),
214 ATTR_LIST(reclaim_segments),
215 ATTR_LIST(max_small_discards),
216 ATTR_LIST(batched_trim_sections),
217 ATTR_LIST(ipu_policy),
218 ATTR_LIST(min_ipu_util),
219 ATTR_LIST(min_fsync_blocks),
220 ATTR_LIST(max_victim_search),
221 ATTR_LIST(dir_level),
222 ATTR_LIST(ram_thresh),
226 static const struct sysfs_ops f2fs_attr_ops = {
227 .show = f2fs_attr_show,
228 .store = f2fs_attr_store,
231 static struct kobj_type f2fs_ktype = {
232 .default_attrs = f2fs_attrs,
233 .sysfs_ops = &f2fs_attr_ops,
234 .release = f2fs_sb_release,
237 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
239 struct va_format vaf;
245 printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
249 static void init_once(void *foo)
251 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
253 inode_init_once(&fi->vfs_inode);
256 static int parse_options(struct super_block *sb, char *options)
258 struct f2fs_sb_info *sbi = F2FS_SB(sb);
259 substring_t args[MAX_OPT_ARGS];
266 while ((p = strsep(&options, ",")) != NULL) {
271 * Initialize args struct so we know whether arg was
272 * found; some options take optional arguments.
274 args[0].to = args[0].from = NULL;
275 token = match_token(p, f2fs_tokens, args);
278 case Opt_gc_background:
279 name = match_strdup(&args[0]);
283 if (strlen(name) == 2 && !strncmp(name, "on", 2))
285 else if (strlen(name) == 3 && !strncmp(name, "off", 3))
286 clear_opt(sbi, BG_GC);
293 case Opt_disable_roll_forward:
294 set_opt(sbi, DISABLE_ROLL_FORWARD);
297 /* this option mounts f2fs with ro */
298 set_opt(sbi, DISABLE_ROLL_FORWARD);
299 if (!f2fs_readonly(sb))
303 set_opt(sbi, DISCARD);
306 set_opt(sbi, NOHEAP);
308 #ifdef CONFIG_F2FS_FS_XATTR
310 set_opt(sbi, XATTR_USER);
312 case Opt_nouser_xattr:
313 clear_opt(sbi, XATTR_USER);
315 case Opt_inline_xattr:
316 set_opt(sbi, INLINE_XATTR);
320 f2fs_msg(sb, KERN_INFO,
321 "user_xattr options not supported");
323 case Opt_nouser_xattr:
324 f2fs_msg(sb, KERN_INFO,
325 "nouser_xattr options not supported");
327 case Opt_inline_xattr:
328 f2fs_msg(sb, KERN_INFO,
329 "inline_xattr options not supported");
332 #ifdef CONFIG_F2FS_FS_POSIX_ACL
334 set_opt(sbi, POSIX_ACL);
337 clear_opt(sbi, POSIX_ACL);
341 f2fs_msg(sb, KERN_INFO, "acl options not supported");
344 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
347 case Opt_active_logs:
348 if (args->from && match_int(args, &arg))
350 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
352 sbi->active_logs = arg;
354 case Opt_disable_ext_identify:
355 set_opt(sbi, DISABLE_EXT_IDENTIFY);
357 case Opt_inline_data:
358 set_opt(sbi, INLINE_DATA);
360 case Opt_inline_dentry:
361 set_opt(sbi, INLINE_DENTRY);
363 case Opt_flush_merge:
364 set_opt(sbi, FLUSH_MERGE);
367 set_opt(sbi, NOBARRIER);
370 set_opt(sbi, FASTBOOT);
372 case Opt_extent_cache:
373 set_opt(sbi, EXTENT_CACHE);
376 f2fs_msg(sb, KERN_ERR,
377 "Unrecognized mount option \"%s\" or missing value",
385 static struct inode *f2fs_alloc_inode(struct super_block *sb)
387 struct f2fs_inode_info *fi;
389 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
393 init_once((void *) fi);
395 /* Initialize f2fs-specific inode info */
396 fi->vfs_inode.i_version = 1;
397 atomic_set(&fi->dirty_pages, 0);
398 fi->i_current_depth = 1;
400 rwlock_init(&fi->ext_lock);
401 init_rwsem(&fi->i_sem);
402 INIT_RADIX_TREE(&fi->inmem_root, GFP_NOFS);
403 INIT_LIST_HEAD(&fi->inmem_pages);
404 mutex_init(&fi->inmem_lock);
406 set_inode_flag(fi, FI_NEW_INODE);
408 if (test_opt(F2FS_SB(sb), INLINE_XATTR))
409 set_inode_flag(fi, FI_INLINE_XATTR);
411 /* Will be used by directory only */
412 fi->i_dir_level = F2FS_SB(sb)->dir_level;
414 return &fi->vfs_inode;
417 static int f2fs_drop_inode(struct inode *inode)
420 * This is to avoid a deadlock condition like below.
421 * writeback_single_inode(inode)
422 * - f2fs_write_data_page
423 * - f2fs_gc -> iput -> evict
424 * - inode_wait_for_writeback(inode)
426 if (!inode_unhashed(inode) && inode->i_state & I_SYNC)
428 return generic_drop_inode(inode);
432 * f2fs_dirty_inode() is called from __mark_inode_dirty()
434 * We should call set_dirty_inode to write the dirty inode through write_inode.
436 static void f2fs_dirty_inode(struct inode *inode, int flags)
438 set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
441 static void f2fs_i_callback(struct rcu_head *head)
443 struct inode *inode = container_of(head, struct inode, i_rcu);
444 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
447 static void f2fs_destroy_inode(struct inode *inode)
449 call_rcu(&inode->i_rcu, f2fs_i_callback);
452 static void f2fs_put_super(struct super_block *sb)
454 struct f2fs_sb_info *sbi = F2FS_SB(sb);
457 remove_proc_entry("segment_info", sbi->s_proc);
458 remove_proc_entry(sb->s_id, f2fs_proc_root);
460 kobject_del(&sbi->s_kobj);
462 f2fs_destroy_stats(sbi);
466 * We don't need to do checkpoint when superblock is clean.
467 * But, the previous checkpoint was not done by umount, it needs to do
468 * clean checkpoint again.
470 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
471 !is_set_ckpt_flags(F2FS_CKPT(sbi), CP_UMOUNT_FLAG)) {
472 struct cp_control cpc = {
475 write_checkpoint(sbi, &cpc);
479 * normally superblock is clean, so we need to release this.
480 * In addition, EIO will skip do checkpoint, we need this as well.
482 release_dirty_inode(sbi);
483 release_discard_addrs(sbi);
485 iput(sbi->node_inode);
486 iput(sbi->meta_inode);
488 /* destroy f2fs internal modules */
489 destroy_node_manager(sbi);
490 destroy_segment_manager(sbi);
493 kobject_put(&sbi->s_kobj);
494 wait_for_completion(&sbi->s_kobj_unregister);
496 sb->s_fs_info = NULL;
497 brelse(sbi->raw_super_buf);
501 int f2fs_sync_fs(struct super_block *sb, int sync)
503 struct f2fs_sb_info *sbi = F2FS_SB(sb);
505 trace_f2fs_sync_fs(sb, sync);
508 struct cp_control cpc;
510 cpc.reason = __get_cp_reason(sbi);
512 mutex_lock(&sbi->gc_mutex);
513 write_checkpoint(sbi, &cpc);
514 mutex_unlock(&sbi->gc_mutex);
516 f2fs_balance_fs(sbi);
518 f2fs_trace_ios(NULL, NULL, 1);
523 static int f2fs_freeze(struct super_block *sb)
527 if (f2fs_readonly(sb))
530 err = f2fs_sync_fs(sb, 1);
534 static int f2fs_unfreeze(struct super_block *sb)
539 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
541 struct super_block *sb = dentry->d_sb;
542 struct f2fs_sb_info *sbi = F2FS_SB(sb);
543 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
544 block_t total_count, user_block_count, start_count, ovp_count;
546 total_count = le64_to_cpu(sbi->raw_super->block_count);
547 user_block_count = sbi->user_block_count;
548 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
549 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
550 buf->f_type = F2FS_SUPER_MAGIC;
551 buf->f_bsize = sbi->blocksize;
553 buf->f_blocks = total_count - start_count;
554 buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
555 buf->f_bavail = user_block_count - valid_user_blocks(sbi);
557 buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
558 buf->f_ffree = buf->f_files - valid_inode_count(sbi);
560 buf->f_namelen = F2FS_NAME_LEN;
561 buf->f_fsid.val[0] = (u32)id;
562 buf->f_fsid.val[1] = (u32)(id >> 32);
567 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
569 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
571 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC))
572 seq_printf(seq, ",background_gc=%s", "on");
574 seq_printf(seq, ",background_gc=%s", "off");
575 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
576 seq_puts(seq, ",disable_roll_forward");
577 if (test_opt(sbi, DISCARD))
578 seq_puts(seq, ",discard");
579 if (test_opt(sbi, NOHEAP))
580 seq_puts(seq, ",no_heap_alloc");
581 #ifdef CONFIG_F2FS_FS_XATTR
582 if (test_opt(sbi, XATTR_USER))
583 seq_puts(seq, ",user_xattr");
585 seq_puts(seq, ",nouser_xattr");
586 if (test_opt(sbi, INLINE_XATTR))
587 seq_puts(seq, ",inline_xattr");
589 #ifdef CONFIG_F2FS_FS_POSIX_ACL
590 if (test_opt(sbi, POSIX_ACL))
591 seq_puts(seq, ",acl");
593 seq_puts(seq, ",noacl");
595 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
596 seq_puts(seq, ",disable_ext_identify");
597 if (test_opt(sbi, INLINE_DATA))
598 seq_puts(seq, ",inline_data");
599 if (test_opt(sbi, INLINE_DENTRY))
600 seq_puts(seq, ",inline_dentry");
601 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
602 seq_puts(seq, ",flush_merge");
603 if (test_opt(sbi, NOBARRIER))
604 seq_puts(seq, ",nobarrier");
605 if (test_opt(sbi, FASTBOOT))
606 seq_puts(seq, ",fastboot");
607 if (test_opt(sbi, EXTENT_CACHE))
608 seq_puts(seq, ",extent_cache");
609 seq_printf(seq, ",active_logs=%u", sbi->active_logs);
614 static int segment_info_seq_show(struct seq_file *seq, void *offset)
616 struct super_block *sb = seq->private;
617 struct f2fs_sb_info *sbi = F2FS_SB(sb);
618 unsigned int total_segs =
619 le32_to_cpu(sbi->raw_super->segment_count_main);
622 seq_puts(seq, "format: segment_type|valid_blocks\n"
623 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
625 for (i = 0; i < total_segs; i++) {
626 struct seg_entry *se = get_seg_entry(sbi, i);
629 seq_printf(seq, "%-5d", i);
630 seq_printf(seq, "%d|%-3u", se->type,
631 get_valid_blocks(sbi, i, 1));
632 if ((i % 10) == 9 || i == (total_segs - 1))
641 static int segment_info_open_fs(struct inode *inode, struct file *file)
643 return single_open(file, segment_info_seq_show, PDE_DATA(inode));
646 static const struct file_operations f2fs_seq_segment_info_fops = {
647 .owner = THIS_MODULE,
648 .open = segment_info_open_fs,
651 .release = single_release,
654 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
656 struct f2fs_sb_info *sbi = F2FS_SB(sb);
657 struct f2fs_mount_info org_mount_opt;
658 int err, active_logs;
659 bool need_restart_gc = false;
660 bool need_stop_gc = false;
665 * Save the old mount options in case we
666 * need to restore them.
668 org_mount_opt = sbi->mount_opt;
669 active_logs = sbi->active_logs;
671 sbi->mount_opt.opt = 0;
672 sbi->active_logs = NR_CURSEG_TYPE;
674 /* parse mount options */
675 err = parse_options(sb, data);
680 * Previous and new state of filesystem is RO,
681 * so skip checking GC and FLUSH_MERGE conditions.
683 if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
687 * We stop the GC thread if FS is mounted as RO
688 * or if background_gc = off is passed in mount
689 * option. Also sync the filesystem.
691 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
692 if (sbi->gc_thread) {
695 need_restart_gc = true;
697 } else if (!sbi->gc_thread) {
698 err = start_gc_thread(sbi);
705 * We stop issue flush thread if FS is mounted as RO
706 * or if flush_merge is not passed in mount option.
708 if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
709 destroy_flush_cmd_control(sbi);
710 } else if (!SM_I(sbi)->cmd_control_info) {
711 err = create_flush_cmd_control(sbi);
716 /* Update the POSIXACL Flag */
717 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
718 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
721 if (need_restart_gc) {
722 if (start_gc_thread(sbi))
723 f2fs_msg(sbi->sb, KERN_WARNING,
724 "background gc thread has stopped");
725 } else if (need_stop_gc) {
729 sbi->mount_opt = org_mount_opt;
730 sbi->active_logs = active_logs;
734 static struct super_operations f2fs_sops = {
735 .alloc_inode = f2fs_alloc_inode,
736 .drop_inode = f2fs_drop_inode,
737 .destroy_inode = f2fs_destroy_inode,
738 .write_inode = f2fs_write_inode,
739 .dirty_inode = f2fs_dirty_inode,
740 .show_options = f2fs_show_options,
741 .evict_inode = f2fs_evict_inode,
742 .put_super = f2fs_put_super,
743 .sync_fs = f2fs_sync_fs,
744 .freeze_fs = f2fs_freeze,
745 .unfreeze_fs = f2fs_unfreeze,
746 .statfs = f2fs_statfs,
747 .remount_fs = f2fs_remount,
750 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
751 u64 ino, u32 generation)
753 struct f2fs_sb_info *sbi = F2FS_SB(sb);
756 if (check_nid_range(sbi, ino))
757 return ERR_PTR(-ESTALE);
760 * f2fs_iget isn't quite right if the inode is currently unallocated!
761 * However f2fs_iget currently does appropriate checks to handle stale
762 * inodes so everything is OK.
764 inode = f2fs_iget(sb, ino);
766 return ERR_CAST(inode);
767 if (unlikely(generation && inode->i_generation != generation)) {
768 /* we didn't find the right inode.. */
770 return ERR_PTR(-ESTALE);
775 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
776 int fh_len, int fh_type)
778 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
782 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
783 int fh_len, int fh_type)
785 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
789 static const struct export_operations f2fs_export_ops = {
790 .fh_to_dentry = f2fs_fh_to_dentry,
791 .fh_to_parent = f2fs_fh_to_parent,
792 .get_parent = f2fs_get_parent,
795 static loff_t max_file_size(unsigned bits)
797 loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
798 loff_t leaf_count = ADDRS_PER_BLOCK;
800 /* two direct node blocks */
801 result += (leaf_count * 2);
803 /* two indirect node blocks */
804 leaf_count *= NIDS_PER_BLOCK;
805 result += (leaf_count * 2);
807 /* one double indirect node block */
808 leaf_count *= NIDS_PER_BLOCK;
809 result += leaf_count;
815 static int sanity_check_raw_super(struct super_block *sb,
816 struct f2fs_super_block *raw_super)
818 unsigned int blocksize;
820 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
821 f2fs_msg(sb, KERN_INFO,
822 "Magic Mismatch, valid(0x%x) - read(0x%x)",
823 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
827 /* Currently, support only 4KB page cache size */
828 if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
829 f2fs_msg(sb, KERN_INFO,
830 "Invalid page_cache_size (%lu), supports only 4KB\n",
835 /* Currently, support only 4KB block size */
836 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
837 if (blocksize != F2FS_BLKSIZE) {
838 f2fs_msg(sb, KERN_INFO,
839 "Invalid blocksize (%u), supports only 4KB\n",
844 /* Currently, support 512/1024/2048/4096 bytes sector size */
845 if (le32_to_cpu(raw_super->log_sectorsize) >
846 F2FS_MAX_LOG_SECTOR_SIZE ||
847 le32_to_cpu(raw_super->log_sectorsize) <
848 F2FS_MIN_LOG_SECTOR_SIZE) {
849 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
850 le32_to_cpu(raw_super->log_sectorsize));
853 if (le32_to_cpu(raw_super->log_sectors_per_block) +
854 le32_to_cpu(raw_super->log_sectorsize) !=
855 F2FS_MAX_LOG_SECTOR_SIZE) {
856 f2fs_msg(sb, KERN_INFO,
857 "Invalid log sectors per block(%u) log sectorsize(%u)",
858 le32_to_cpu(raw_super->log_sectors_per_block),
859 le32_to_cpu(raw_super->log_sectorsize));
865 static int sanity_check_ckpt(struct f2fs_sb_info *sbi)
867 unsigned int total, fsmeta;
868 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
869 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
871 total = le32_to_cpu(raw_super->segment_count);
872 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
873 fsmeta += le32_to_cpu(raw_super->segment_count_sit);
874 fsmeta += le32_to_cpu(raw_super->segment_count_nat);
875 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
876 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
878 if (unlikely(fsmeta >= total))
881 if (unlikely(f2fs_cp_error(sbi))) {
882 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
888 static void init_sb_info(struct f2fs_sb_info *sbi)
890 struct f2fs_super_block *raw_super = sbi->raw_super;
893 sbi->log_sectors_per_block =
894 le32_to_cpu(raw_super->log_sectors_per_block);
895 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
896 sbi->blocksize = 1 << sbi->log_blocksize;
897 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
898 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
899 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
900 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
901 sbi->total_sections = le32_to_cpu(raw_super->section_count);
902 sbi->total_node_count =
903 (le32_to_cpu(raw_super->segment_count_nat) / 2)
904 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
905 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
906 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
907 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
908 sbi->cur_victim_sec = NULL_SECNO;
909 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
911 for (i = 0; i < NR_COUNT_TYPE; i++)
912 atomic_set(&sbi->nr_pages[i], 0);
914 sbi->dir_level = DEF_DIR_LEVEL;
915 clear_sbi_flag(sbi, SBI_NEED_FSCK);
919 * Read f2fs raw super block.
920 * Because we have two copies of super block, so read the first one at first,
921 * if the first one is invalid, move to read the second one.
923 static int read_raw_super_block(struct super_block *sb,
924 struct f2fs_super_block **raw_super,
925 struct buffer_head **raw_super_buf)
930 *raw_super_buf = sb_bread(sb, block);
931 if (!*raw_super_buf) {
932 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
942 *raw_super = (struct f2fs_super_block *)
943 ((char *)(*raw_super_buf)->b_data + F2FS_SUPER_OFFSET);
945 /* sanity checking of raw super */
946 if (sanity_check_raw_super(sb, *raw_super)) {
947 brelse(*raw_super_buf);
948 f2fs_msg(sb, KERN_ERR,
949 "Can't find valid F2FS filesystem in %dth superblock",
962 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
964 struct f2fs_sb_info *sbi;
965 struct f2fs_super_block *raw_super = NULL;
966 struct buffer_head *raw_super_buf;
970 char *options = NULL;
974 /* allocate memory for f2fs-specific super block info */
975 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
979 /* set a block size */
980 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
981 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
985 err = read_raw_super_block(sb, &raw_super, &raw_super_buf);
990 /* init some FS parameters */
991 sbi->active_logs = NR_CURSEG_TYPE;
995 #ifdef CONFIG_F2FS_FS_XATTR
996 set_opt(sbi, XATTR_USER);
998 #ifdef CONFIG_F2FS_FS_POSIX_ACL
999 set_opt(sbi, POSIX_ACL);
1001 /* parse mount options */
1002 options = kstrdup((const char *)data, GFP_KERNEL);
1003 if (data && !options) {
1008 err = parse_options(sb, options);
1012 sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
1013 sb->s_max_links = F2FS_LINK_MAX;
1014 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1016 sb->s_op = &f2fs_sops;
1017 sb->s_xattr = f2fs_xattr_handlers;
1018 sb->s_export_op = &f2fs_export_ops;
1019 sb->s_magic = F2FS_SUPER_MAGIC;
1020 sb->s_time_gran = 1;
1021 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1022 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1023 memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1025 /* init f2fs-specific super block info */
1027 sbi->raw_super = raw_super;
1028 sbi->raw_super_buf = raw_super_buf;
1029 mutex_init(&sbi->gc_mutex);
1030 mutex_init(&sbi->writepages);
1031 mutex_init(&sbi->cp_mutex);
1032 init_rwsem(&sbi->node_write);
1033 clear_sbi_flag(sbi, SBI_POR_DOING);
1034 spin_lock_init(&sbi->stat_lock);
1036 init_rwsem(&sbi->read_io.io_rwsem);
1037 sbi->read_io.sbi = sbi;
1038 sbi->read_io.bio = NULL;
1039 for (i = 0; i < NR_PAGE_TYPE; i++) {
1040 init_rwsem(&sbi->write_io[i].io_rwsem);
1041 sbi->write_io[i].sbi = sbi;
1042 sbi->write_io[i].bio = NULL;
1045 init_rwsem(&sbi->cp_rwsem);
1046 init_waitqueue_head(&sbi->cp_wait);
1049 /* get an inode for meta space */
1050 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1051 if (IS_ERR(sbi->meta_inode)) {
1052 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1053 err = PTR_ERR(sbi->meta_inode);
1057 err = get_valid_checkpoint(sbi);
1059 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1060 goto free_meta_inode;
1063 /* sanity checking of checkpoint */
1065 if (sanity_check_ckpt(sbi)) {
1066 f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint");
1070 sbi->total_valid_node_count =
1071 le32_to_cpu(sbi->ckpt->valid_node_count);
1072 sbi->total_valid_inode_count =
1073 le32_to_cpu(sbi->ckpt->valid_inode_count);
1074 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1075 sbi->total_valid_block_count =
1076 le64_to_cpu(sbi->ckpt->valid_block_count);
1077 sbi->last_valid_block_count = sbi->total_valid_block_count;
1078 sbi->alloc_valid_block_count = 0;
1079 INIT_LIST_HEAD(&sbi->dir_inode_list);
1080 spin_lock_init(&sbi->dir_inode_lock);
1082 init_extent_cache_info(sbi);
1084 init_ino_entry_info(sbi);
1086 /* setup f2fs internal modules */
1087 err = build_segment_manager(sbi);
1089 f2fs_msg(sb, KERN_ERR,
1090 "Failed to initialize F2FS segment manager");
1093 err = build_node_manager(sbi);
1095 f2fs_msg(sb, KERN_ERR,
1096 "Failed to initialize F2FS node manager");
1100 build_gc_manager(sbi);
1102 /* get an inode for node space */
1103 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1104 if (IS_ERR(sbi->node_inode)) {
1105 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1106 err = PTR_ERR(sbi->node_inode);
1110 /* if there are nt orphan nodes free them */
1111 recover_orphan_inodes(sbi);
1113 /* read root inode and dentry */
1114 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1116 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1117 err = PTR_ERR(root);
1118 goto free_node_inode;
1120 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1123 goto free_node_inode;
1126 sb->s_root = d_make_root(root); /* allocate root dentry */
1129 goto free_root_inode;
1132 err = f2fs_build_stats(sbi);
1134 goto free_root_inode;
1137 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1140 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1141 &f2fs_seq_segment_info_fops, sb);
1143 if (test_opt(sbi, DISCARD)) {
1144 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1145 if (!blk_queue_discard(q))
1146 f2fs_msg(sb, KERN_WARNING,
1147 "mounting with \"discard\" option, but "
1148 "the device does not support discard");
1151 sbi->s_kobj.kset = f2fs_kset;
1152 init_completion(&sbi->s_kobj_unregister);
1153 err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1159 set_sbi_flag(sbi, SBI_NEED_FSCK);
1161 /* recover fsynced data */
1162 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1164 * mount should be failed, when device has readonly mode, and
1165 * previous checkpoint was not done by clean system shutdown.
1167 if (bdev_read_only(sb->s_bdev) &&
1168 !is_set_ckpt_flags(sbi->ckpt, CP_UMOUNT_FLAG)) {
1172 err = recover_fsync_data(sbi);
1174 f2fs_msg(sb, KERN_ERR,
1175 "Cannot recover all fsync data errno=%ld", err);
1181 * If filesystem is not mounted as read-only then
1182 * do start the gc_thread.
1184 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
1185 /* After POR, we can run background GC thread.*/
1186 err = start_gc_thread(sbi);
1194 kobject_del(&sbi->s_kobj);
1197 remove_proc_entry("segment_info", sbi->s_proc);
1198 remove_proc_entry(sb->s_id, f2fs_proc_root);
1200 f2fs_destroy_stats(sbi);
1205 iput(sbi->node_inode);
1207 destroy_node_manager(sbi);
1209 destroy_segment_manager(sbi);
1213 make_bad_inode(sbi->meta_inode);
1214 iput(sbi->meta_inode);
1218 brelse(raw_super_buf);
1222 /* give only one another chance */
1225 shrink_dcache_sb(sb);
1231 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1232 const char *dev_name, void *data)
1234 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1237 static void kill_f2fs_super(struct super_block *sb)
1240 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
1241 kill_block_super(sb);
1244 static struct file_system_type f2fs_fs_type = {
1245 .owner = THIS_MODULE,
1247 .mount = f2fs_mount,
1248 .kill_sb = kill_f2fs_super,
1249 .fs_flags = FS_REQUIRES_DEV,
1251 MODULE_ALIAS_FS("f2fs");
1253 static int __init init_inodecache(void)
1255 f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
1256 sizeof(struct f2fs_inode_info));
1257 if (!f2fs_inode_cachep)
1262 static void destroy_inodecache(void)
1265 * Make sure all delayed rcu free inodes are flushed before we
1269 kmem_cache_destroy(f2fs_inode_cachep);
1272 static int __init init_f2fs_fs(void)
1276 f2fs_build_trace_ios();
1278 err = init_inodecache();
1281 err = create_node_manager_caches();
1283 goto free_inodecache;
1284 err = create_segment_manager_caches();
1286 goto free_node_manager_caches;
1287 err = create_checkpoint_caches();
1289 goto free_segment_manager_caches;
1290 err = create_extent_cache();
1292 goto free_checkpoint_caches;
1293 f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1296 goto free_extent_cache;
1298 err = register_filesystem(&f2fs_fs_type);
1301 f2fs_create_root_stats();
1302 f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1306 kset_unregister(f2fs_kset);
1308 destroy_extent_cache();
1309 free_checkpoint_caches:
1310 destroy_checkpoint_caches();
1311 free_segment_manager_caches:
1312 destroy_segment_manager_caches();
1313 free_node_manager_caches:
1314 destroy_node_manager_caches();
1316 destroy_inodecache();
1321 static void __exit exit_f2fs_fs(void)
1323 remove_proc_entry("fs/f2fs", NULL);
1324 f2fs_destroy_root_stats();
1325 unregister_filesystem(&f2fs_fs_type);
1326 destroy_extent_cache();
1327 destroy_checkpoint_caches();
1328 destroy_segment_manager_caches();
1329 destroy_node_manager_caches();
1330 destroy_inodecache();
1331 kset_unregister(f2fs_kset);
1332 f2fs_destroy_trace_ios();
1335 module_init(init_f2fs_fs)
1336 module_exit(exit_f2fs_fs)
1338 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1339 MODULE_DESCRIPTION("Flash Friendly File System");
1340 MODULE_LICENSE("GPL");