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;
42 /* f2fs-wide shrinker description */
43 static struct shrinker f2fs_shrinker_info = {
44 .scan_objects = f2fs_shrink_scan,
45 .count_objects = f2fs_shrink_count,
46 .seeks = DEFAULT_SEEKS,
51 Opt_disable_roll_forward,
60 Opt_disable_ext_identify,
73 static match_table_t f2fs_tokens = {
74 {Opt_gc_background, "background_gc=%s"},
75 {Opt_disable_roll_forward, "disable_roll_forward"},
76 {Opt_norecovery, "norecovery"},
77 {Opt_discard, "discard"},
78 {Opt_noheap, "no_heap"},
79 {Opt_user_xattr, "user_xattr"},
80 {Opt_nouser_xattr, "nouser_xattr"},
83 {Opt_active_logs, "active_logs=%u"},
84 {Opt_disable_ext_identify, "disable_ext_identify"},
85 {Opt_inline_xattr, "inline_xattr"},
86 {Opt_inline_data, "inline_data"},
87 {Opt_inline_dentry, "inline_dentry"},
88 {Opt_flush_merge, "flush_merge"},
89 {Opt_nobarrier, "nobarrier"},
90 {Opt_fastboot, "fastboot"},
91 {Opt_extent_cache, "extent_cache"},
92 {Opt_noextent_cache, "noextent_cache"},
93 {Opt_noinline_data, "noinline_data"},
97 /* Sysfs support for f2fs */
99 GC_THREAD, /* struct f2fs_gc_thread */
100 SM_INFO, /* struct f2fs_sm_info */
101 NM_INFO, /* struct f2fs_nm_info */
102 F2FS_SBI, /* struct f2fs_sb_info */
106 struct attribute attr;
107 ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
108 ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
109 const char *, size_t);
114 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
116 if (struct_type == GC_THREAD)
117 return (unsigned char *)sbi->gc_thread;
118 else if (struct_type == SM_INFO)
119 return (unsigned char *)SM_I(sbi);
120 else if (struct_type == NM_INFO)
121 return (unsigned char *)NM_I(sbi);
122 else if (struct_type == F2FS_SBI)
123 return (unsigned char *)sbi;
127 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
128 struct f2fs_sb_info *sbi, char *buf)
130 unsigned char *ptr = NULL;
133 ptr = __struct_ptr(sbi, a->struct_type);
137 ui = (unsigned int *)(ptr + a->offset);
139 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
142 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
143 struct f2fs_sb_info *sbi,
144 const char *buf, size_t count)
151 ptr = __struct_ptr(sbi, a->struct_type);
155 ui = (unsigned int *)(ptr + a->offset);
157 ret = kstrtoul(skip_spaces(buf), 0, &t);
164 static ssize_t f2fs_attr_show(struct kobject *kobj,
165 struct attribute *attr, char *buf)
167 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
169 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
171 return a->show ? a->show(a, sbi, buf) : 0;
174 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
175 const char *buf, size_t len)
177 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
179 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
181 return a->store ? a->store(a, sbi, buf, len) : 0;
184 static void f2fs_sb_release(struct kobject *kobj)
186 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
188 complete(&sbi->s_kobj_unregister);
191 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
192 static struct f2fs_attr f2fs_attr_##_name = { \
193 .attr = {.name = __stringify(_name), .mode = _mode }, \
196 .struct_type = _struct_type, \
200 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
201 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
202 f2fs_sbi_show, f2fs_sbi_store, \
203 offsetof(struct struct_name, elname))
205 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
206 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
207 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
208 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
209 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
210 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
211 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, batched_trim_sections, trim_sections);
212 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
213 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
214 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks);
215 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
216 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
217 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
219 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
220 static struct attribute *f2fs_attrs[] = {
221 ATTR_LIST(gc_min_sleep_time),
222 ATTR_LIST(gc_max_sleep_time),
223 ATTR_LIST(gc_no_gc_sleep_time),
225 ATTR_LIST(reclaim_segments),
226 ATTR_LIST(max_small_discards),
227 ATTR_LIST(batched_trim_sections),
228 ATTR_LIST(ipu_policy),
229 ATTR_LIST(min_ipu_util),
230 ATTR_LIST(min_fsync_blocks),
231 ATTR_LIST(max_victim_search),
232 ATTR_LIST(dir_level),
233 ATTR_LIST(ram_thresh),
237 static const struct sysfs_ops f2fs_attr_ops = {
238 .show = f2fs_attr_show,
239 .store = f2fs_attr_store,
242 static struct kobj_type f2fs_ktype = {
243 .default_attrs = f2fs_attrs,
244 .sysfs_ops = &f2fs_attr_ops,
245 .release = f2fs_sb_release,
248 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
250 struct va_format vaf;
256 printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
260 static void init_once(void *foo)
262 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
264 inode_init_once(&fi->vfs_inode);
267 static int parse_options(struct super_block *sb, char *options)
269 struct f2fs_sb_info *sbi = F2FS_SB(sb);
270 struct request_queue *q;
271 substring_t args[MAX_OPT_ARGS];
278 while ((p = strsep(&options, ",")) != NULL) {
283 * Initialize args struct so we know whether arg was
284 * found; some options take optional arguments.
286 args[0].to = args[0].from = NULL;
287 token = match_token(p, f2fs_tokens, args);
290 case Opt_gc_background:
291 name = match_strdup(&args[0]);
295 if (strlen(name) == 2 && !strncmp(name, "on", 2))
297 else if (strlen(name) == 3 && !strncmp(name, "off", 3))
298 clear_opt(sbi, BG_GC);
305 case Opt_disable_roll_forward:
306 set_opt(sbi, DISABLE_ROLL_FORWARD);
309 /* this option mounts f2fs with ro */
310 set_opt(sbi, DISABLE_ROLL_FORWARD);
311 if (!f2fs_readonly(sb))
315 q = bdev_get_queue(sb->s_bdev);
316 if (blk_queue_discard(q)) {
317 set_opt(sbi, DISCARD);
319 f2fs_msg(sb, KERN_WARNING,
320 "mounting with \"discard\" option, but "
321 "the device does not support discard");
325 set_opt(sbi, NOHEAP);
327 #ifdef CONFIG_F2FS_FS_XATTR
329 set_opt(sbi, XATTR_USER);
331 case Opt_nouser_xattr:
332 clear_opt(sbi, XATTR_USER);
334 case Opt_inline_xattr:
335 set_opt(sbi, INLINE_XATTR);
339 f2fs_msg(sb, KERN_INFO,
340 "user_xattr options not supported");
342 case Opt_nouser_xattr:
343 f2fs_msg(sb, KERN_INFO,
344 "nouser_xattr options not supported");
346 case Opt_inline_xattr:
347 f2fs_msg(sb, KERN_INFO,
348 "inline_xattr options not supported");
351 #ifdef CONFIG_F2FS_FS_POSIX_ACL
353 set_opt(sbi, POSIX_ACL);
356 clear_opt(sbi, POSIX_ACL);
360 f2fs_msg(sb, KERN_INFO, "acl options not supported");
363 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
366 case Opt_active_logs:
367 if (args->from && match_int(args, &arg))
369 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
371 sbi->active_logs = arg;
373 case Opt_disable_ext_identify:
374 set_opt(sbi, DISABLE_EXT_IDENTIFY);
376 case Opt_inline_data:
377 set_opt(sbi, INLINE_DATA);
379 case Opt_inline_dentry:
380 set_opt(sbi, INLINE_DENTRY);
382 case Opt_flush_merge:
383 set_opt(sbi, FLUSH_MERGE);
386 set_opt(sbi, NOBARRIER);
389 set_opt(sbi, FASTBOOT);
391 case Opt_extent_cache:
392 set_opt(sbi, EXTENT_CACHE);
394 case Opt_noextent_cache:
395 clear_opt(sbi, EXTENT_CACHE);
397 case Opt_noinline_data:
398 clear_opt(sbi, INLINE_DATA);
401 f2fs_msg(sb, KERN_ERR,
402 "Unrecognized mount option \"%s\" or missing value",
410 static struct inode *f2fs_alloc_inode(struct super_block *sb)
412 struct f2fs_inode_info *fi;
414 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
418 init_once((void *) fi);
420 /* Initialize f2fs-specific inode info */
421 fi->vfs_inode.i_version = 1;
422 atomic_set(&fi->dirty_pages, 0);
423 fi->i_current_depth = 1;
425 init_rwsem(&fi->i_sem);
426 INIT_LIST_HEAD(&fi->inmem_pages);
427 mutex_init(&fi->inmem_lock);
429 set_inode_flag(fi, FI_NEW_INODE);
431 if (test_opt(F2FS_SB(sb), INLINE_XATTR))
432 set_inode_flag(fi, FI_INLINE_XATTR);
434 /* Will be used by directory only */
435 fi->i_dir_level = F2FS_SB(sb)->dir_level;
437 #ifdef CONFIG_F2FS_FS_ENCRYPTION
438 fi->i_crypt_info = NULL;
440 return &fi->vfs_inode;
443 static int f2fs_drop_inode(struct inode *inode)
446 * This is to avoid a deadlock condition like below.
447 * writeback_single_inode(inode)
448 * - f2fs_write_data_page
449 * - f2fs_gc -> iput -> evict
450 * - inode_wait_for_writeback(inode)
452 if (!inode_unhashed(inode) && inode->i_state & I_SYNC) {
453 if (!inode->i_nlink && !is_bad_inode(inode)) {
454 /* to avoid evict_inode call simultaneously */
455 atomic_inc(&inode->i_count);
456 spin_unlock(&inode->i_lock);
458 /* some remained atomic pages should discarded */
459 if (f2fs_is_atomic_file(inode))
460 commit_inmem_pages(inode, true);
462 /* should remain fi->extent_tree for writepage */
463 f2fs_destroy_extent_node(inode);
465 sb_start_intwrite(inode->i_sb);
466 i_size_write(inode, 0);
468 if (F2FS_HAS_BLOCKS(inode))
469 f2fs_truncate(inode, true);
471 sb_end_intwrite(inode->i_sb);
473 #ifdef CONFIG_F2FS_FS_ENCRYPTION
474 if (F2FS_I(inode)->i_crypt_info)
475 f2fs_free_encryption_info(inode,
476 F2FS_I(inode)->i_crypt_info);
478 spin_lock(&inode->i_lock);
479 atomic_dec(&inode->i_count);
483 return generic_drop_inode(inode);
487 * f2fs_dirty_inode() is called from __mark_inode_dirty()
489 * We should call set_dirty_inode to write the dirty inode through write_inode.
491 static void f2fs_dirty_inode(struct inode *inode, int flags)
493 set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
496 static void f2fs_i_callback(struct rcu_head *head)
498 struct inode *inode = container_of(head, struct inode, i_rcu);
499 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
502 static void f2fs_destroy_inode(struct inode *inode)
504 call_rcu(&inode->i_rcu, f2fs_i_callback);
507 static void f2fs_put_super(struct super_block *sb)
509 struct f2fs_sb_info *sbi = F2FS_SB(sb);
512 remove_proc_entry("segment_info", sbi->s_proc);
513 remove_proc_entry(sb->s_id, f2fs_proc_root);
515 kobject_del(&sbi->s_kobj);
519 /* prevent remaining shrinker jobs */
520 mutex_lock(&sbi->umount_mutex);
523 * We don't need to do checkpoint when superblock is clean.
524 * But, the previous checkpoint was not done by umount, it needs to do
525 * clean checkpoint again.
527 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
528 !is_set_ckpt_flags(F2FS_CKPT(sbi), CP_UMOUNT_FLAG)) {
529 struct cp_control cpc = {
532 write_checkpoint(sbi, &cpc);
535 /* write_checkpoint can update stat informaion */
536 f2fs_destroy_stats(sbi);
539 * normally superblock is clean, so we need to release this.
540 * In addition, EIO will skip do checkpoint, we need this as well.
542 release_dirty_inode(sbi);
543 release_discard_addrs(sbi);
545 f2fs_leave_shrinker(sbi);
546 mutex_unlock(&sbi->umount_mutex);
548 iput(sbi->node_inode);
549 iput(sbi->meta_inode);
551 /* destroy f2fs internal modules */
552 destroy_node_manager(sbi);
553 destroy_segment_manager(sbi);
556 kobject_put(&sbi->s_kobj);
557 wait_for_completion(&sbi->s_kobj_unregister);
559 sb->s_fs_info = NULL;
560 brelse(sbi->raw_super_buf);
564 int f2fs_sync_fs(struct super_block *sb, int sync)
566 struct f2fs_sb_info *sbi = F2FS_SB(sb);
568 trace_f2fs_sync_fs(sb, sync);
571 struct cp_control cpc;
573 cpc.reason = __get_cp_reason(sbi);
575 mutex_lock(&sbi->gc_mutex);
576 write_checkpoint(sbi, &cpc);
577 mutex_unlock(&sbi->gc_mutex);
579 f2fs_balance_fs(sbi);
581 f2fs_trace_ios(NULL, 1);
586 static int f2fs_freeze(struct super_block *sb)
590 if (f2fs_readonly(sb))
593 err = f2fs_sync_fs(sb, 1);
597 static int f2fs_unfreeze(struct super_block *sb)
602 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
604 struct super_block *sb = dentry->d_sb;
605 struct f2fs_sb_info *sbi = F2FS_SB(sb);
606 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
607 block_t total_count, user_block_count, start_count, ovp_count;
609 total_count = le64_to_cpu(sbi->raw_super->block_count);
610 user_block_count = sbi->user_block_count;
611 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
612 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
613 buf->f_type = F2FS_SUPER_MAGIC;
614 buf->f_bsize = sbi->blocksize;
616 buf->f_blocks = total_count - start_count;
617 buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
618 buf->f_bavail = user_block_count - valid_user_blocks(sbi);
620 buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
621 buf->f_ffree = buf->f_files - valid_inode_count(sbi);
623 buf->f_namelen = F2FS_NAME_LEN;
624 buf->f_fsid.val[0] = (u32)id;
625 buf->f_fsid.val[1] = (u32)(id >> 32);
630 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
632 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
634 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC))
635 seq_printf(seq, ",background_gc=%s", "on");
637 seq_printf(seq, ",background_gc=%s", "off");
638 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
639 seq_puts(seq, ",disable_roll_forward");
640 if (test_opt(sbi, DISCARD))
641 seq_puts(seq, ",discard");
642 if (test_opt(sbi, NOHEAP))
643 seq_puts(seq, ",no_heap_alloc");
644 #ifdef CONFIG_F2FS_FS_XATTR
645 if (test_opt(sbi, XATTR_USER))
646 seq_puts(seq, ",user_xattr");
648 seq_puts(seq, ",nouser_xattr");
649 if (test_opt(sbi, INLINE_XATTR))
650 seq_puts(seq, ",inline_xattr");
652 #ifdef CONFIG_F2FS_FS_POSIX_ACL
653 if (test_opt(sbi, POSIX_ACL))
654 seq_puts(seq, ",acl");
656 seq_puts(seq, ",noacl");
658 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
659 seq_puts(seq, ",disable_ext_identify");
660 if (test_opt(sbi, INLINE_DATA))
661 seq_puts(seq, ",inline_data");
663 seq_puts(seq, ",noinline_data");
664 if (test_opt(sbi, INLINE_DENTRY))
665 seq_puts(seq, ",inline_dentry");
666 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
667 seq_puts(seq, ",flush_merge");
668 if (test_opt(sbi, NOBARRIER))
669 seq_puts(seq, ",nobarrier");
670 if (test_opt(sbi, FASTBOOT))
671 seq_puts(seq, ",fastboot");
672 if (test_opt(sbi, EXTENT_CACHE))
673 seq_puts(seq, ",extent_cache");
675 seq_puts(seq, ",noextent_cache");
676 seq_printf(seq, ",active_logs=%u", sbi->active_logs);
681 static int segment_info_seq_show(struct seq_file *seq, void *offset)
683 struct super_block *sb = seq->private;
684 struct f2fs_sb_info *sbi = F2FS_SB(sb);
685 unsigned int total_segs =
686 le32_to_cpu(sbi->raw_super->segment_count_main);
689 seq_puts(seq, "format: segment_type|valid_blocks\n"
690 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
692 for (i = 0; i < total_segs; i++) {
693 struct seg_entry *se = get_seg_entry(sbi, i);
696 seq_printf(seq, "%-10d", i);
697 seq_printf(seq, "%d|%-3u", se->type,
698 get_valid_blocks(sbi, i, 1));
699 if ((i % 10) == 9 || i == (total_segs - 1))
708 static int segment_info_open_fs(struct inode *inode, struct file *file)
710 return single_open(file, segment_info_seq_show, PDE_DATA(inode));
713 static const struct file_operations f2fs_seq_segment_info_fops = {
714 .owner = THIS_MODULE,
715 .open = segment_info_open_fs,
718 .release = single_release,
721 static void default_options(struct f2fs_sb_info *sbi)
723 /* init some FS parameters */
724 sbi->active_logs = NR_CURSEG_TYPE;
727 set_opt(sbi, INLINE_DATA);
728 set_opt(sbi, EXTENT_CACHE);
730 #ifdef CONFIG_F2FS_FS_XATTR
731 set_opt(sbi, XATTR_USER);
733 #ifdef CONFIG_F2FS_FS_POSIX_ACL
734 set_opt(sbi, POSIX_ACL);
738 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
740 struct f2fs_sb_info *sbi = F2FS_SB(sb);
741 struct f2fs_mount_info org_mount_opt;
742 int err, active_logs;
743 bool need_restart_gc = false;
744 bool need_stop_gc = false;
749 * Save the old mount options in case we
750 * need to restore them.
752 org_mount_opt = sbi->mount_opt;
753 active_logs = sbi->active_logs;
755 sbi->mount_opt.opt = 0;
756 default_options(sbi);
758 /* parse mount options */
759 err = parse_options(sb, data);
764 * Previous and new state of filesystem is RO,
765 * so skip checking GC and FLUSH_MERGE conditions.
767 if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
771 * We stop the GC thread if FS is mounted as RO
772 * or if background_gc = off is passed in mount
773 * option. Also sync the filesystem.
775 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
776 if (sbi->gc_thread) {
779 need_restart_gc = true;
781 } else if (!sbi->gc_thread) {
782 err = start_gc_thread(sbi);
789 * We stop issue flush thread if FS is mounted as RO
790 * or if flush_merge is not passed in mount option.
792 if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
793 destroy_flush_cmd_control(sbi);
794 } else if (!SM_I(sbi)->cmd_control_info) {
795 err = create_flush_cmd_control(sbi);
800 /* Update the POSIXACL Flag */
801 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
802 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
805 if (need_restart_gc) {
806 if (start_gc_thread(sbi))
807 f2fs_msg(sbi->sb, KERN_WARNING,
808 "background gc thread has stopped");
809 } else if (need_stop_gc) {
813 sbi->mount_opt = org_mount_opt;
814 sbi->active_logs = active_logs;
818 static struct super_operations f2fs_sops = {
819 .alloc_inode = f2fs_alloc_inode,
820 .drop_inode = f2fs_drop_inode,
821 .destroy_inode = f2fs_destroy_inode,
822 .write_inode = f2fs_write_inode,
823 .dirty_inode = f2fs_dirty_inode,
824 .show_options = f2fs_show_options,
825 .evict_inode = f2fs_evict_inode,
826 .put_super = f2fs_put_super,
827 .sync_fs = f2fs_sync_fs,
828 .freeze_fs = f2fs_freeze,
829 .unfreeze_fs = f2fs_unfreeze,
830 .statfs = f2fs_statfs,
831 .remount_fs = f2fs_remount,
834 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
835 u64 ino, u32 generation)
837 struct f2fs_sb_info *sbi = F2FS_SB(sb);
840 if (check_nid_range(sbi, ino))
841 return ERR_PTR(-ESTALE);
844 * f2fs_iget isn't quite right if the inode is currently unallocated!
845 * However f2fs_iget currently does appropriate checks to handle stale
846 * inodes so everything is OK.
848 inode = f2fs_iget(sb, ino);
850 return ERR_CAST(inode);
851 if (unlikely(generation && inode->i_generation != generation)) {
852 /* we didn't find the right inode.. */
854 return ERR_PTR(-ESTALE);
859 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
860 int fh_len, int fh_type)
862 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
866 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
867 int fh_len, int fh_type)
869 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
873 static const struct export_operations f2fs_export_ops = {
874 .fh_to_dentry = f2fs_fh_to_dentry,
875 .fh_to_parent = f2fs_fh_to_parent,
876 .get_parent = f2fs_get_parent,
879 static loff_t max_file_size(unsigned bits)
881 loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
882 loff_t leaf_count = ADDRS_PER_BLOCK;
884 /* two direct node blocks */
885 result += (leaf_count * 2);
887 /* two indirect node blocks */
888 leaf_count *= NIDS_PER_BLOCK;
889 result += (leaf_count * 2);
891 /* one double indirect node block */
892 leaf_count *= NIDS_PER_BLOCK;
893 result += leaf_count;
899 static int sanity_check_raw_super(struct super_block *sb,
900 struct f2fs_super_block *raw_super)
902 unsigned int blocksize;
904 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
905 f2fs_msg(sb, KERN_INFO,
906 "Magic Mismatch, valid(0x%x) - read(0x%x)",
907 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
911 /* Currently, support only 4KB page cache size */
912 if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
913 f2fs_msg(sb, KERN_INFO,
914 "Invalid page_cache_size (%lu), supports only 4KB\n",
919 /* Currently, support only 4KB block size */
920 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
921 if (blocksize != F2FS_BLKSIZE) {
922 f2fs_msg(sb, KERN_INFO,
923 "Invalid blocksize (%u), supports only 4KB\n",
928 /* Currently, support 512/1024/2048/4096 bytes sector size */
929 if (le32_to_cpu(raw_super->log_sectorsize) >
930 F2FS_MAX_LOG_SECTOR_SIZE ||
931 le32_to_cpu(raw_super->log_sectorsize) <
932 F2FS_MIN_LOG_SECTOR_SIZE) {
933 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
934 le32_to_cpu(raw_super->log_sectorsize));
937 if (le32_to_cpu(raw_super->log_sectors_per_block) +
938 le32_to_cpu(raw_super->log_sectorsize) !=
939 F2FS_MAX_LOG_SECTOR_SIZE) {
940 f2fs_msg(sb, KERN_INFO,
941 "Invalid log sectors per block(%u) log sectorsize(%u)",
942 le32_to_cpu(raw_super->log_sectors_per_block),
943 le32_to_cpu(raw_super->log_sectorsize));
949 static int sanity_check_ckpt(struct f2fs_sb_info *sbi)
951 unsigned int total, fsmeta;
952 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
953 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
955 total = le32_to_cpu(raw_super->segment_count);
956 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
957 fsmeta += le32_to_cpu(raw_super->segment_count_sit);
958 fsmeta += le32_to_cpu(raw_super->segment_count_nat);
959 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
960 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
962 if (unlikely(fsmeta >= total))
965 if (unlikely(f2fs_cp_error(sbi))) {
966 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
972 static void init_sb_info(struct f2fs_sb_info *sbi)
974 struct f2fs_super_block *raw_super = sbi->raw_super;
977 sbi->log_sectors_per_block =
978 le32_to_cpu(raw_super->log_sectors_per_block);
979 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
980 sbi->blocksize = 1 << sbi->log_blocksize;
981 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
982 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
983 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
984 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
985 sbi->total_sections = le32_to_cpu(raw_super->section_count);
986 sbi->total_node_count =
987 (le32_to_cpu(raw_super->segment_count_nat) / 2)
988 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
989 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
990 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
991 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
992 sbi->cur_victim_sec = NULL_SECNO;
993 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
995 for (i = 0; i < NR_COUNT_TYPE; i++)
996 atomic_set(&sbi->nr_pages[i], 0);
998 sbi->dir_level = DEF_DIR_LEVEL;
999 clear_sbi_flag(sbi, SBI_NEED_FSCK);
1001 INIT_LIST_HEAD(&sbi->s_list);
1002 mutex_init(&sbi->umount_mutex);
1006 * Read f2fs raw super block.
1007 * Because we have two copies of super block, so read the first one at first,
1008 * if the first one is invalid, move to read the second one.
1010 static int read_raw_super_block(struct super_block *sb,
1011 struct f2fs_super_block **raw_super,
1012 struct buffer_head **raw_super_buf,
1016 struct buffer_head *buffer;
1017 struct f2fs_super_block *super;
1021 buffer = sb_bread(sb, block);
1024 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
1035 super = (struct f2fs_super_block *)
1036 ((char *)(buffer)->b_data + F2FS_SUPER_OFFSET);
1038 /* sanity checking of raw super */
1039 if (sanity_check_raw_super(sb, super)) {
1042 f2fs_msg(sb, KERN_ERR,
1043 "Can't find valid F2FS filesystem in %dth superblock",
1055 *raw_super_buf = buffer;
1058 /* already have a valid superblock */
1062 /* check the validity of the second superblock */
1069 /* No valid superblock */
1076 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
1078 struct buffer_head *sbh = sbi->raw_super_buf;
1079 sector_t block = sbh->b_blocknr;
1082 /* write back-up superblock first */
1083 sbh->b_blocknr = block ? 0 : 1;
1084 mark_buffer_dirty(sbh);
1085 err = sync_dirty_buffer(sbh);
1087 sbh->b_blocknr = block;
1089 /* if we are in recovery path, skip writing valid superblock */
1093 /* write current valid superblock */
1094 mark_buffer_dirty(sbh);
1095 err = sync_dirty_buffer(sbh);
1097 clear_buffer_write_io_error(sbh);
1098 set_buffer_uptodate(sbh);
1102 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
1104 struct f2fs_sb_info *sbi;
1105 struct f2fs_super_block *raw_super;
1106 struct buffer_head *raw_super_buf;
1109 bool retry = true, need_fsck = false;
1110 char *options = NULL;
1116 raw_super_buf = NULL;
1119 /* allocate memory for f2fs-specific super block info */
1120 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
1124 /* set a block size */
1125 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
1126 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
1130 err = read_raw_super_block(sb, &raw_super, &raw_super_buf, &recovery);
1134 sb->s_fs_info = sbi;
1135 default_options(sbi);
1136 /* parse mount options */
1137 options = kstrdup((const char *)data, GFP_KERNEL);
1138 if (data && !options) {
1143 err = parse_options(sb, options);
1147 sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
1148 sb->s_max_links = F2FS_LINK_MAX;
1149 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1151 sb->s_op = &f2fs_sops;
1152 sb->s_xattr = f2fs_xattr_handlers;
1153 sb->s_export_op = &f2fs_export_ops;
1154 sb->s_magic = F2FS_SUPER_MAGIC;
1155 sb->s_time_gran = 1;
1156 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1157 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1158 memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1160 /* init f2fs-specific super block info */
1162 sbi->raw_super = raw_super;
1163 sbi->raw_super_buf = raw_super_buf;
1164 mutex_init(&sbi->gc_mutex);
1165 mutex_init(&sbi->writepages);
1166 mutex_init(&sbi->cp_mutex);
1167 init_rwsem(&sbi->node_write);
1169 /* disallow all the data/node/meta page writes */
1170 set_sbi_flag(sbi, SBI_POR_DOING);
1171 spin_lock_init(&sbi->stat_lock);
1173 init_rwsem(&sbi->read_io.io_rwsem);
1174 sbi->read_io.sbi = sbi;
1175 sbi->read_io.bio = NULL;
1176 for (i = 0; i < NR_PAGE_TYPE; i++) {
1177 init_rwsem(&sbi->write_io[i].io_rwsem);
1178 sbi->write_io[i].sbi = sbi;
1179 sbi->write_io[i].bio = NULL;
1182 init_rwsem(&sbi->cp_rwsem);
1183 init_waitqueue_head(&sbi->cp_wait);
1186 /* get an inode for meta space */
1187 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1188 if (IS_ERR(sbi->meta_inode)) {
1189 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1190 err = PTR_ERR(sbi->meta_inode);
1194 err = get_valid_checkpoint(sbi);
1196 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1197 goto free_meta_inode;
1200 /* sanity checking of checkpoint */
1202 if (sanity_check_ckpt(sbi)) {
1203 f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint");
1207 sbi->total_valid_node_count =
1208 le32_to_cpu(sbi->ckpt->valid_node_count);
1209 sbi->total_valid_inode_count =
1210 le32_to_cpu(sbi->ckpt->valid_inode_count);
1211 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1212 sbi->total_valid_block_count =
1213 le64_to_cpu(sbi->ckpt->valid_block_count);
1214 sbi->last_valid_block_count = sbi->total_valid_block_count;
1215 sbi->alloc_valid_block_count = 0;
1216 INIT_LIST_HEAD(&sbi->dir_inode_list);
1217 spin_lock_init(&sbi->dir_inode_lock);
1219 init_extent_cache_info(sbi);
1221 init_ino_entry_info(sbi);
1223 /* setup f2fs internal modules */
1224 err = build_segment_manager(sbi);
1226 f2fs_msg(sb, KERN_ERR,
1227 "Failed to initialize F2FS segment manager");
1230 err = build_node_manager(sbi);
1232 f2fs_msg(sb, KERN_ERR,
1233 "Failed to initialize F2FS node manager");
1237 build_gc_manager(sbi);
1239 /* get an inode for node space */
1240 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1241 if (IS_ERR(sbi->node_inode)) {
1242 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1243 err = PTR_ERR(sbi->node_inode);
1247 f2fs_join_shrinker(sbi);
1249 /* if there are nt orphan nodes free them */
1250 err = recover_orphan_inodes(sbi);
1252 goto free_node_inode;
1254 /* read root inode and dentry */
1255 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1257 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1258 err = PTR_ERR(root);
1259 goto free_node_inode;
1261 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1264 goto free_node_inode;
1267 sb->s_root = d_make_root(root); /* allocate root dentry */
1270 goto free_root_inode;
1273 err = f2fs_build_stats(sbi);
1275 goto free_root_inode;
1278 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1281 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1282 &f2fs_seq_segment_info_fops, sb);
1284 sbi->s_kobj.kset = f2fs_kset;
1285 init_completion(&sbi->s_kobj_unregister);
1286 err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1291 /* recover fsynced data */
1292 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1294 * mount should be failed, when device has readonly mode, and
1295 * previous checkpoint was not done by clean system shutdown.
1297 if (bdev_read_only(sb->s_bdev) &&
1298 !is_set_ckpt_flags(sbi->ckpt, CP_UMOUNT_FLAG)) {
1304 set_sbi_flag(sbi, SBI_NEED_FSCK);
1306 err = recover_fsync_data(sbi);
1309 f2fs_msg(sb, KERN_ERR,
1310 "Cannot recover all fsync data errno=%ld", err);
1314 /* recover_fsync_data() cleared this already */
1315 clear_sbi_flag(sbi, SBI_POR_DOING);
1318 * If filesystem is not mounted as read-only then
1319 * do start the gc_thread.
1321 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
1322 /* After POR, we can run background GC thread.*/
1323 err = start_gc_thread(sbi);
1329 /* recover broken superblock */
1330 if (recovery && !f2fs_readonly(sb) && !bdev_read_only(sb->s_bdev)) {
1331 f2fs_msg(sb, KERN_INFO, "Recover invalid superblock");
1332 f2fs_commit_super(sbi, true);
1338 kobject_del(&sbi->s_kobj);
1341 remove_proc_entry("segment_info", sbi->s_proc);
1342 remove_proc_entry(sb->s_id, f2fs_proc_root);
1344 f2fs_destroy_stats(sbi);
1349 mutex_lock(&sbi->umount_mutex);
1350 f2fs_leave_shrinker(sbi);
1351 iput(sbi->node_inode);
1352 mutex_unlock(&sbi->umount_mutex);
1354 destroy_node_manager(sbi);
1356 destroy_segment_manager(sbi);
1360 make_bad_inode(sbi->meta_inode);
1361 iput(sbi->meta_inode);
1365 brelse(raw_super_buf);
1369 /* give only one another chance */
1372 shrink_dcache_sb(sb);
1378 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1379 const char *dev_name, void *data)
1381 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1384 static void kill_f2fs_super(struct super_block *sb)
1387 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
1388 kill_block_super(sb);
1391 static struct file_system_type f2fs_fs_type = {
1392 .owner = THIS_MODULE,
1394 .mount = f2fs_mount,
1395 .kill_sb = kill_f2fs_super,
1396 .fs_flags = FS_REQUIRES_DEV,
1398 MODULE_ALIAS_FS("f2fs");
1400 static int __init init_inodecache(void)
1402 f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
1403 sizeof(struct f2fs_inode_info));
1404 if (!f2fs_inode_cachep)
1409 static void destroy_inodecache(void)
1412 * Make sure all delayed rcu free inodes are flushed before we
1416 kmem_cache_destroy(f2fs_inode_cachep);
1419 static int __init init_f2fs_fs(void)
1423 f2fs_build_trace_ios();
1425 err = init_inodecache();
1428 err = create_node_manager_caches();
1430 goto free_inodecache;
1431 err = create_segment_manager_caches();
1433 goto free_node_manager_caches;
1434 err = create_checkpoint_caches();
1436 goto free_segment_manager_caches;
1437 err = create_extent_cache();
1439 goto free_checkpoint_caches;
1440 f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1443 goto free_extent_cache;
1445 err = f2fs_init_crypto();
1449 err = register_shrinker(&f2fs_shrinker_info);
1453 err = register_filesystem(&f2fs_fs_type);
1456 f2fs_create_root_stats();
1457 f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1461 unregister_shrinker(&f2fs_shrinker_info);
1465 kset_unregister(f2fs_kset);
1467 destroy_extent_cache();
1468 free_checkpoint_caches:
1469 destroy_checkpoint_caches();
1470 free_segment_manager_caches:
1471 destroy_segment_manager_caches();
1472 free_node_manager_caches:
1473 destroy_node_manager_caches();
1475 destroy_inodecache();
1480 static void __exit exit_f2fs_fs(void)
1482 remove_proc_entry("fs/f2fs", NULL);
1483 f2fs_destroy_root_stats();
1484 unregister_shrinker(&f2fs_shrinker_info);
1485 unregister_filesystem(&f2fs_fs_type);
1487 destroy_extent_cache();
1488 destroy_checkpoint_caches();
1489 destroy_segment_manager_caches();
1490 destroy_node_manager_caches();
1491 destroy_inodecache();
1492 kset_unregister(f2fs_kset);
1493 f2fs_destroy_trace_ios();
1496 module_init(init_f2fs_fs)
1497 module_exit(exit_f2fs_fs)
1499 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1500 MODULE_DESCRIPTION("Flash Friendly File System");
1501 MODULE_LICENSE("GPL");