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,
52 Opt_disable_ext_identify,
62 static match_table_t f2fs_tokens = {
63 {Opt_gc_background, "background_gc=%s"},
64 {Opt_disable_roll_forward, "disable_roll_forward"},
65 {Opt_discard, "discard"},
66 {Opt_noheap, "no_heap"},
67 {Opt_user_xattr, "user_xattr"},
68 {Opt_nouser_xattr, "nouser_xattr"},
71 {Opt_active_logs, "active_logs=%u"},
72 {Opt_disable_ext_identify, "disable_ext_identify"},
73 {Opt_inline_xattr, "inline_xattr"},
74 {Opt_inline_data, "inline_data"},
75 {Opt_inline_dentry, "inline_dentry"},
76 {Opt_flush_merge, "flush_merge"},
77 {Opt_nobarrier, "nobarrier"},
78 {Opt_fastboot, "fastboot"},
82 /* Sysfs support for f2fs */
84 GC_THREAD, /* struct f2fs_gc_thread */
85 SM_INFO, /* struct f2fs_sm_info */
86 NM_INFO, /* struct f2fs_nm_info */
87 F2FS_SBI, /* struct f2fs_sb_info */
91 struct attribute attr;
92 ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
93 ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
94 const char *, size_t);
99 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
101 if (struct_type == GC_THREAD)
102 return (unsigned char *)sbi->gc_thread;
103 else if (struct_type == SM_INFO)
104 return (unsigned char *)SM_I(sbi);
105 else if (struct_type == NM_INFO)
106 return (unsigned char *)NM_I(sbi);
107 else if (struct_type == F2FS_SBI)
108 return (unsigned char *)sbi;
112 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
113 struct f2fs_sb_info *sbi, char *buf)
115 unsigned char *ptr = NULL;
118 ptr = __struct_ptr(sbi, a->struct_type);
122 ui = (unsigned int *)(ptr + a->offset);
124 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
127 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
128 struct f2fs_sb_info *sbi,
129 const char *buf, size_t count)
136 ptr = __struct_ptr(sbi, a->struct_type);
140 ui = (unsigned int *)(ptr + a->offset);
142 ret = kstrtoul(skip_spaces(buf), 0, &t);
149 static ssize_t f2fs_attr_show(struct kobject *kobj,
150 struct attribute *attr, char *buf)
152 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
154 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
156 return a->show ? a->show(a, sbi, buf) : 0;
159 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
160 const char *buf, size_t len)
162 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
164 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
166 return a->store ? a->store(a, sbi, buf, len) : 0;
169 static void f2fs_sb_release(struct kobject *kobj)
171 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
173 complete(&sbi->s_kobj_unregister);
176 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
177 static struct f2fs_attr f2fs_attr_##_name = { \
178 .attr = {.name = __stringify(_name), .mode = _mode }, \
181 .struct_type = _struct_type, \
185 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
186 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
187 f2fs_sbi_show, f2fs_sbi_store, \
188 offsetof(struct struct_name, elname))
190 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
191 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
192 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
193 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
194 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
195 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
196 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
197 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
198 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks);
199 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
200 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
201 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
203 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
204 static struct attribute *f2fs_attrs[] = {
205 ATTR_LIST(gc_min_sleep_time),
206 ATTR_LIST(gc_max_sleep_time),
207 ATTR_LIST(gc_no_gc_sleep_time),
209 ATTR_LIST(reclaim_segments),
210 ATTR_LIST(max_small_discards),
211 ATTR_LIST(ipu_policy),
212 ATTR_LIST(min_ipu_util),
213 ATTR_LIST(min_fsync_blocks),
214 ATTR_LIST(max_victim_search),
215 ATTR_LIST(dir_level),
216 ATTR_LIST(ram_thresh),
220 static const struct sysfs_ops f2fs_attr_ops = {
221 .show = f2fs_attr_show,
222 .store = f2fs_attr_store,
225 static struct kobj_type f2fs_ktype = {
226 .default_attrs = f2fs_attrs,
227 .sysfs_ops = &f2fs_attr_ops,
228 .release = f2fs_sb_release,
231 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
233 struct va_format vaf;
239 printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
243 static void init_once(void *foo)
245 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
247 inode_init_once(&fi->vfs_inode);
250 static int parse_options(struct super_block *sb, char *options)
252 struct f2fs_sb_info *sbi = F2FS_SB(sb);
253 substring_t args[MAX_OPT_ARGS];
260 while ((p = strsep(&options, ",")) != NULL) {
265 * Initialize args struct so we know whether arg was
266 * found; some options take optional arguments.
268 args[0].to = args[0].from = NULL;
269 token = match_token(p, f2fs_tokens, args);
272 case Opt_gc_background:
273 name = match_strdup(&args[0]);
277 if (strlen(name) == 2 && !strncmp(name, "on", 2))
279 else if (strlen(name) == 3 && !strncmp(name, "off", 3))
280 clear_opt(sbi, BG_GC);
287 case Opt_disable_roll_forward:
288 set_opt(sbi, DISABLE_ROLL_FORWARD);
291 set_opt(sbi, DISCARD);
294 set_opt(sbi, NOHEAP);
296 #ifdef CONFIG_F2FS_FS_XATTR
298 set_opt(sbi, XATTR_USER);
300 case Opt_nouser_xattr:
301 clear_opt(sbi, XATTR_USER);
303 case Opt_inline_xattr:
304 set_opt(sbi, INLINE_XATTR);
308 f2fs_msg(sb, KERN_INFO,
309 "user_xattr options not supported");
311 case Opt_nouser_xattr:
312 f2fs_msg(sb, KERN_INFO,
313 "nouser_xattr options not supported");
315 case Opt_inline_xattr:
316 f2fs_msg(sb, KERN_INFO,
317 "inline_xattr options not supported");
320 #ifdef CONFIG_F2FS_FS_POSIX_ACL
322 set_opt(sbi, POSIX_ACL);
325 clear_opt(sbi, POSIX_ACL);
329 f2fs_msg(sb, KERN_INFO, "acl options not supported");
332 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
335 case Opt_active_logs:
336 if (args->from && match_int(args, &arg))
338 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
340 sbi->active_logs = arg;
342 case Opt_disable_ext_identify:
343 set_opt(sbi, DISABLE_EXT_IDENTIFY);
345 case Opt_inline_data:
346 set_opt(sbi, INLINE_DATA);
348 case Opt_inline_dentry:
349 set_opt(sbi, INLINE_DENTRY);
351 case Opt_flush_merge:
352 set_opt(sbi, FLUSH_MERGE);
355 set_opt(sbi, NOBARRIER);
358 set_opt(sbi, FASTBOOT);
361 f2fs_msg(sb, KERN_ERR,
362 "Unrecognized mount option \"%s\" or missing value",
370 static struct inode *f2fs_alloc_inode(struct super_block *sb)
372 struct f2fs_inode_info *fi;
374 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
378 init_once((void *) fi);
380 /* Initialize f2fs-specific inode info */
381 fi->vfs_inode.i_version = 1;
382 atomic_set(&fi->dirty_pages, 0);
383 fi->i_current_depth = 1;
385 rwlock_init(&fi->ext.ext_lock);
386 init_rwsem(&fi->i_sem);
387 INIT_RADIX_TREE(&fi->inmem_root, GFP_NOFS);
388 INIT_LIST_HEAD(&fi->inmem_pages);
389 mutex_init(&fi->inmem_lock);
391 set_inode_flag(fi, FI_NEW_INODE);
393 if (test_opt(F2FS_SB(sb), INLINE_XATTR))
394 set_inode_flag(fi, FI_INLINE_XATTR);
396 /* Will be used by directory only */
397 fi->i_dir_level = F2FS_SB(sb)->dir_level;
399 return &fi->vfs_inode;
402 static int f2fs_drop_inode(struct inode *inode)
405 * This is to avoid a deadlock condition like below.
406 * writeback_single_inode(inode)
407 * - f2fs_write_data_page
408 * - f2fs_gc -> iput -> evict
409 * - inode_wait_for_writeback(inode)
411 if (!inode_unhashed(inode) && inode->i_state & I_SYNC)
413 return generic_drop_inode(inode);
417 * f2fs_dirty_inode() is called from __mark_inode_dirty()
419 * We should call set_dirty_inode to write the dirty inode through write_inode.
421 static void f2fs_dirty_inode(struct inode *inode, int flags)
423 set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
426 static void f2fs_i_callback(struct rcu_head *head)
428 struct inode *inode = container_of(head, struct inode, i_rcu);
429 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
432 static void f2fs_destroy_inode(struct inode *inode)
434 call_rcu(&inode->i_rcu, f2fs_i_callback);
437 static void f2fs_put_super(struct super_block *sb)
439 struct f2fs_sb_info *sbi = F2FS_SB(sb);
442 remove_proc_entry("segment_info", sbi->s_proc);
443 remove_proc_entry(sb->s_id, f2fs_proc_root);
445 kobject_del(&sbi->s_kobj);
447 f2fs_destroy_stats(sbi);
450 /* We don't need to do checkpoint when it's clean */
452 struct cp_control cpc = {
455 write_checkpoint(sbi, &cpc);
459 * normally superblock is clean, so we need to release this.
460 * In addition, EIO will skip do checkpoint, we need this as well.
462 release_dirty_inode(sbi);
463 release_discard_addrs(sbi);
465 iput(sbi->node_inode);
466 iput(sbi->meta_inode);
468 /* destroy f2fs internal modules */
469 destroy_node_manager(sbi);
470 destroy_segment_manager(sbi);
473 kobject_put(&sbi->s_kobj);
474 wait_for_completion(&sbi->s_kobj_unregister);
476 sb->s_fs_info = NULL;
477 brelse(sbi->raw_super_buf);
481 int f2fs_sync_fs(struct super_block *sb, int sync)
483 struct f2fs_sb_info *sbi = F2FS_SB(sb);
485 trace_f2fs_sync_fs(sb, sync);
488 struct cp_control cpc;
490 cpc.reason = test_opt(sbi, FASTBOOT) ? CP_UMOUNT : CP_SYNC;
491 mutex_lock(&sbi->gc_mutex);
492 write_checkpoint(sbi, &cpc);
493 mutex_unlock(&sbi->gc_mutex);
495 f2fs_balance_fs(sbi);
497 f2fs_trace_ios(NULL, NULL, 1);
502 static int f2fs_freeze(struct super_block *sb)
506 if (f2fs_readonly(sb))
509 err = f2fs_sync_fs(sb, 1);
513 static int f2fs_unfreeze(struct super_block *sb)
518 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
520 struct super_block *sb = dentry->d_sb;
521 struct f2fs_sb_info *sbi = F2FS_SB(sb);
522 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
523 block_t total_count, user_block_count, start_count, ovp_count;
525 total_count = le64_to_cpu(sbi->raw_super->block_count);
526 user_block_count = sbi->user_block_count;
527 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
528 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
529 buf->f_type = F2FS_SUPER_MAGIC;
530 buf->f_bsize = sbi->blocksize;
532 buf->f_blocks = total_count - start_count;
533 buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
534 buf->f_bavail = user_block_count - valid_user_blocks(sbi);
536 buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
537 buf->f_ffree = buf->f_files - valid_inode_count(sbi);
539 buf->f_namelen = F2FS_NAME_LEN;
540 buf->f_fsid.val[0] = (u32)id;
541 buf->f_fsid.val[1] = (u32)(id >> 32);
546 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
548 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
550 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC))
551 seq_printf(seq, ",background_gc=%s", "on");
553 seq_printf(seq, ",background_gc=%s", "off");
554 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
555 seq_puts(seq, ",disable_roll_forward");
556 if (test_opt(sbi, DISCARD))
557 seq_puts(seq, ",discard");
558 if (test_opt(sbi, NOHEAP))
559 seq_puts(seq, ",no_heap_alloc");
560 #ifdef CONFIG_F2FS_FS_XATTR
561 if (test_opt(sbi, XATTR_USER))
562 seq_puts(seq, ",user_xattr");
564 seq_puts(seq, ",nouser_xattr");
565 if (test_opt(sbi, INLINE_XATTR))
566 seq_puts(seq, ",inline_xattr");
568 #ifdef CONFIG_F2FS_FS_POSIX_ACL
569 if (test_opt(sbi, POSIX_ACL))
570 seq_puts(seq, ",acl");
572 seq_puts(seq, ",noacl");
574 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
575 seq_puts(seq, ",disable_ext_identify");
576 if (test_opt(sbi, INLINE_DATA))
577 seq_puts(seq, ",inline_data");
578 if (test_opt(sbi, INLINE_DENTRY))
579 seq_puts(seq, ",inline_dentry");
580 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
581 seq_puts(seq, ",flush_merge");
582 if (test_opt(sbi, NOBARRIER))
583 seq_puts(seq, ",nobarrier");
584 if (test_opt(sbi, FASTBOOT))
585 seq_puts(seq, ",fastboot");
586 seq_printf(seq, ",active_logs=%u", sbi->active_logs);
591 static int segment_info_seq_show(struct seq_file *seq, void *offset)
593 struct super_block *sb = seq->private;
594 struct f2fs_sb_info *sbi = F2FS_SB(sb);
595 unsigned int total_segs =
596 le32_to_cpu(sbi->raw_super->segment_count_main);
599 seq_puts(seq, "format: segment_type|valid_blocks\n"
600 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
602 for (i = 0; i < total_segs; i++) {
603 struct seg_entry *se = get_seg_entry(sbi, i);
606 seq_printf(seq, "%-5d", i);
607 seq_printf(seq, "%d|%-3u", se->type,
608 get_valid_blocks(sbi, i, 1));
609 if ((i % 10) == 9 || i == (total_segs - 1))
618 static int segment_info_open_fs(struct inode *inode, struct file *file)
620 return single_open(file, segment_info_seq_show, PDE_DATA(inode));
623 static const struct file_operations f2fs_seq_segment_info_fops = {
624 .owner = THIS_MODULE,
625 .open = segment_info_open_fs,
628 .release = single_release,
631 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
633 struct f2fs_sb_info *sbi = F2FS_SB(sb);
634 struct f2fs_mount_info org_mount_opt;
635 int err, active_logs;
636 bool need_restart_gc = false;
637 bool need_stop_gc = false;
642 * Save the old mount options in case we
643 * need to restore them.
645 org_mount_opt = sbi->mount_opt;
646 active_logs = sbi->active_logs;
648 sbi->mount_opt.opt = 0;
649 sbi->active_logs = NR_CURSEG_TYPE;
651 /* parse mount options */
652 err = parse_options(sb, data);
657 * Previous and new state of filesystem is RO,
658 * so skip checking GC and FLUSH_MERGE conditions.
660 if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
664 * We stop the GC thread if FS is mounted as RO
665 * or if background_gc = off is passed in mount
666 * option. Also sync the filesystem.
668 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
669 if (sbi->gc_thread) {
672 need_restart_gc = true;
674 } else if (!sbi->gc_thread) {
675 err = start_gc_thread(sbi);
682 * We stop issue flush thread if FS is mounted as RO
683 * or if flush_merge is not passed in mount option.
685 if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
686 destroy_flush_cmd_control(sbi);
687 } else if (!SM_I(sbi)->cmd_control_info) {
688 err = create_flush_cmd_control(sbi);
693 /* Update the POSIXACL Flag */
694 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
695 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
698 if (need_restart_gc) {
699 if (start_gc_thread(sbi))
700 f2fs_msg(sbi->sb, KERN_WARNING,
701 "background gc thread has stopped");
702 } else if (need_stop_gc) {
706 sbi->mount_opt = org_mount_opt;
707 sbi->active_logs = active_logs;
711 static struct super_operations f2fs_sops = {
712 .alloc_inode = f2fs_alloc_inode,
713 .drop_inode = f2fs_drop_inode,
714 .destroy_inode = f2fs_destroy_inode,
715 .write_inode = f2fs_write_inode,
716 .dirty_inode = f2fs_dirty_inode,
717 .show_options = f2fs_show_options,
718 .evict_inode = f2fs_evict_inode,
719 .put_super = f2fs_put_super,
720 .sync_fs = f2fs_sync_fs,
721 .freeze_fs = f2fs_freeze,
722 .unfreeze_fs = f2fs_unfreeze,
723 .statfs = f2fs_statfs,
724 .remount_fs = f2fs_remount,
727 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
728 u64 ino, u32 generation)
730 struct f2fs_sb_info *sbi = F2FS_SB(sb);
733 if (check_nid_range(sbi, ino))
734 return ERR_PTR(-ESTALE);
737 * f2fs_iget isn't quite right if the inode is currently unallocated!
738 * However f2fs_iget currently does appropriate checks to handle stale
739 * inodes so everything is OK.
741 inode = f2fs_iget(sb, ino);
743 return ERR_CAST(inode);
744 if (unlikely(generation && inode->i_generation != generation)) {
745 /* we didn't find the right inode.. */
747 return ERR_PTR(-ESTALE);
752 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
753 int fh_len, int fh_type)
755 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
759 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
760 int fh_len, int fh_type)
762 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
766 static const struct export_operations f2fs_export_ops = {
767 .fh_to_dentry = f2fs_fh_to_dentry,
768 .fh_to_parent = f2fs_fh_to_parent,
769 .get_parent = f2fs_get_parent,
772 static loff_t max_file_size(unsigned bits)
774 loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
775 loff_t leaf_count = ADDRS_PER_BLOCK;
777 /* two direct node blocks */
778 result += (leaf_count * 2);
780 /* two indirect node blocks */
781 leaf_count *= NIDS_PER_BLOCK;
782 result += (leaf_count * 2);
784 /* one double indirect node block */
785 leaf_count *= NIDS_PER_BLOCK;
786 result += leaf_count;
792 static int sanity_check_raw_super(struct super_block *sb,
793 struct f2fs_super_block *raw_super)
795 unsigned int blocksize;
797 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
798 f2fs_msg(sb, KERN_INFO,
799 "Magic Mismatch, valid(0x%x) - read(0x%x)",
800 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
804 /* Currently, support only 4KB page cache size */
805 if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
806 f2fs_msg(sb, KERN_INFO,
807 "Invalid page_cache_size (%lu), supports only 4KB\n",
812 /* Currently, support only 4KB block size */
813 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
814 if (blocksize != F2FS_BLKSIZE) {
815 f2fs_msg(sb, KERN_INFO,
816 "Invalid blocksize (%u), supports only 4KB\n",
821 /* Currently, support 512/1024/2048/4096 bytes sector size */
822 if (le32_to_cpu(raw_super->log_sectorsize) >
823 F2FS_MAX_LOG_SECTOR_SIZE ||
824 le32_to_cpu(raw_super->log_sectorsize) <
825 F2FS_MIN_LOG_SECTOR_SIZE) {
826 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
827 le32_to_cpu(raw_super->log_sectorsize));
830 if (le32_to_cpu(raw_super->log_sectors_per_block) +
831 le32_to_cpu(raw_super->log_sectorsize) !=
832 F2FS_MAX_LOG_SECTOR_SIZE) {
833 f2fs_msg(sb, KERN_INFO,
834 "Invalid log sectors per block(%u) log sectorsize(%u)",
835 le32_to_cpu(raw_super->log_sectors_per_block),
836 le32_to_cpu(raw_super->log_sectorsize));
842 static int sanity_check_ckpt(struct f2fs_sb_info *sbi)
844 unsigned int total, fsmeta;
845 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
846 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
848 total = le32_to_cpu(raw_super->segment_count);
849 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
850 fsmeta += le32_to_cpu(raw_super->segment_count_sit);
851 fsmeta += le32_to_cpu(raw_super->segment_count_nat);
852 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
853 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
855 if (unlikely(fsmeta >= total))
858 if (unlikely(f2fs_cp_error(sbi))) {
859 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
865 static void init_sb_info(struct f2fs_sb_info *sbi)
867 struct f2fs_super_block *raw_super = sbi->raw_super;
870 sbi->log_sectors_per_block =
871 le32_to_cpu(raw_super->log_sectors_per_block);
872 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
873 sbi->blocksize = 1 << sbi->log_blocksize;
874 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
875 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
876 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
877 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
878 sbi->total_sections = le32_to_cpu(raw_super->section_count);
879 sbi->total_node_count =
880 (le32_to_cpu(raw_super->segment_count_nat) / 2)
881 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
882 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
883 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
884 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
885 sbi->cur_victim_sec = NULL_SECNO;
886 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
888 for (i = 0; i < NR_COUNT_TYPE; i++)
889 atomic_set(&sbi->nr_pages[i], 0);
891 sbi->dir_level = DEF_DIR_LEVEL;
892 sbi->need_fsck = false;
896 * Read f2fs raw super block.
897 * Because we have two copies of super block, so read the first one at first,
898 * if the first one is invalid, move to read the second one.
900 static int read_raw_super_block(struct super_block *sb,
901 struct f2fs_super_block **raw_super,
902 struct buffer_head **raw_super_buf)
907 *raw_super_buf = sb_bread(sb, block);
908 if (!*raw_super_buf) {
909 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
919 *raw_super = (struct f2fs_super_block *)
920 ((char *)(*raw_super_buf)->b_data + F2FS_SUPER_OFFSET);
922 /* sanity checking of raw super */
923 if (sanity_check_raw_super(sb, *raw_super)) {
924 brelse(*raw_super_buf);
925 f2fs_msg(sb, KERN_ERR,
926 "Can't find valid F2FS filesystem in %dth superblock",
939 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
941 struct f2fs_sb_info *sbi;
942 struct f2fs_super_block *raw_super = NULL;
943 struct buffer_head *raw_super_buf;
950 /* allocate memory for f2fs-specific super block info */
951 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
955 /* set a block size */
956 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
957 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
961 err = read_raw_super_block(sb, &raw_super, &raw_super_buf);
966 /* init some FS parameters */
967 sbi->active_logs = NR_CURSEG_TYPE;
971 #ifdef CONFIG_F2FS_FS_XATTR
972 set_opt(sbi, XATTR_USER);
974 #ifdef CONFIG_F2FS_FS_POSIX_ACL
975 set_opt(sbi, POSIX_ACL);
977 /* parse mount options */
978 err = parse_options(sb, (char *)data);
982 sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
983 sb->s_max_links = F2FS_LINK_MAX;
984 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
986 sb->s_op = &f2fs_sops;
987 sb->s_xattr = f2fs_xattr_handlers;
988 sb->s_export_op = &f2fs_export_ops;
989 sb->s_magic = F2FS_SUPER_MAGIC;
991 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
992 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
993 memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
995 /* init f2fs-specific super block info */
997 sbi->raw_super = raw_super;
998 sbi->raw_super_buf = raw_super_buf;
999 mutex_init(&sbi->gc_mutex);
1000 mutex_init(&sbi->writepages);
1001 mutex_init(&sbi->cp_mutex);
1002 init_rwsem(&sbi->node_write);
1003 sbi->por_doing = false;
1004 spin_lock_init(&sbi->stat_lock);
1006 init_rwsem(&sbi->read_io.io_rwsem);
1007 sbi->read_io.sbi = sbi;
1008 sbi->read_io.bio = NULL;
1009 for (i = 0; i < NR_PAGE_TYPE; i++) {
1010 init_rwsem(&sbi->write_io[i].io_rwsem);
1011 sbi->write_io[i].sbi = sbi;
1012 sbi->write_io[i].bio = NULL;
1015 init_rwsem(&sbi->cp_rwsem);
1016 init_waitqueue_head(&sbi->cp_wait);
1019 /* get an inode for meta space */
1020 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1021 if (IS_ERR(sbi->meta_inode)) {
1022 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1023 err = PTR_ERR(sbi->meta_inode);
1027 err = get_valid_checkpoint(sbi);
1029 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1030 goto free_meta_inode;
1033 /* sanity checking of checkpoint */
1035 if (sanity_check_ckpt(sbi)) {
1036 f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint");
1040 sbi->total_valid_node_count =
1041 le32_to_cpu(sbi->ckpt->valid_node_count);
1042 sbi->total_valid_inode_count =
1043 le32_to_cpu(sbi->ckpt->valid_inode_count);
1044 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1045 sbi->total_valid_block_count =
1046 le64_to_cpu(sbi->ckpt->valid_block_count);
1047 sbi->last_valid_block_count = sbi->total_valid_block_count;
1048 sbi->alloc_valid_block_count = 0;
1049 INIT_LIST_HEAD(&sbi->dir_inode_list);
1050 spin_lock_init(&sbi->dir_inode_lock);
1052 init_ino_entry_info(sbi);
1054 /* setup f2fs internal modules */
1055 err = build_segment_manager(sbi);
1057 f2fs_msg(sb, KERN_ERR,
1058 "Failed to initialize F2FS segment manager");
1061 err = build_node_manager(sbi);
1063 f2fs_msg(sb, KERN_ERR,
1064 "Failed to initialize F2FS node manager");
1068 build_gc_manager(sbi);
1070 /* get an inode for node space */
1071 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1072 if (IS_ERR(sbi->node_inode)) {
1073 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1074 err = PTR_ERR(sbi->node_inode);
1078 /* if there are nt orphan nodes free them */
1079 recover_orphan_inodes(sbi);
1081 /* read root inode and dentry */
1082 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1084 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1085 err = PTR_ERR(root);
1086 goto free_node_inode;
1088 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1091 goto free_node_inode;
1094 sb->s_root = d_make_root(root); /* allocate root dentry */
1097 goto free_root_inode;
1100 err = f2fs_build_stats(sbi);
1102 goto free_root_inode;
1105 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1108 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1109 &f2fs_seq_segment_info_fops, sb);
1111 if (test_opt(sbi, DISCARD)) {
1112 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1113 if (!blk_queue_discard(q))
1114 f2fs_msg(sb, KERN_WARNING,
1115 "mounting with \"discard\" option, but "
1116 "the device does not support discard");
1119 sbi->s_kobj.kset = f2fs_kset;
1120 init_completion(&sbi->s_kobj_unregister);
1121 err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1127 sbi->need_fsck = true;
1129 /* recover fsynced data */
1130 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1131 err = recover_fsync_data(sbi);
1133 f2fs_msg(sb, KERN_ERR,
1134 "Cannot recover all fsync data errno=%ld", err);
1140 * If filesystem is not mounted as read-only then
1141 * do start the gc_thread.
1143 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
1144 /* After POR, we can run background GC thread.*/
1145 err = start_gc_thread(sbi);
1152 kobject_del(&sbi->s_kobj);
1155 remove_proc_entry("segment_info", sbi->s_proc);
1156 remove_proc_entry(sb->s_id, f2fs_proc_root);
1158 f2fs_destroy_stats(sbi);
1163 iput(sbi->node_inode);
1165 destroy_node_manager(sbi);
1167 destroy_segment_manager(sbi);
1171 make_bad_inode(sbi->meta_inode);
1172 iput(sbi->meta_inode);
1174 brelse(raw_super_buf);
1178 /* give only one another chance */
1181 shrink_dcache_sb(sb);
1187 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1188 const char *dev_name, void *data)
1190 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1193 static struct file_system_type f2fs_fs_type = {
1194 .owner = THIS_MODULE,
1196 .mount = f2fs_mount,
1197 .kill_sb = kill_block_super,
1198 .fs_flags = FS_REQUIRES_DEV,
1200 MODULE_ALIAS_FS("f2fs");
1202 static int __init init_inodecache(void)
1204 f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
1205 sizeof(struct f2fs_inode_info));
1206 if (!f2fs_inode_cachep)
1211 static void destroy_inodecache(void)
1214 * Make sure all delayed rcu free inodes are flushed before we
1218 kmem_cache_destroy(f2fs_inode_cachep);
1221 static int __init init_f2fs_fs(void)
1225 err = init_inodecache();
1228 err = create_node_manager_caches();
1230 goto free_inodecache;
1231 err = create_segment_manager_caches();
1233 goto free_node_manager_caches;
1234 err = create_gc_caches();
1236 goto free_segment_manager_caches;
1237 err = create_checkpoint_caches();
1239 goto free_gc_caches;
1240 f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1243 goto free_checkpoint_caches;
1245 err = register_filesystem(&f2fs_fs_type);
1248 f2fs_create_root_stats();
1249 f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1253 kset_unregister(f2fs_kset);
1254 free_checkpoint_caches:
1255 destroy_checkpoint_caches();
1257 destroy_gc_caches();
1258 free_segment_manager_caches:
1259 destroy_segment_manager_caches();
1260 free_node_manager_caches:
1261 destroy_node_manager_caches();
1263 destroy_inodecache();
1268 static void __exit exit_f2fs_fs(void)
1270 remove_proc_entry("fs/f2fs", NULL);
1271 f2fs_destroy_root_stats();
1272 unregister_filesystem(&f2fs_fs_type);
1273 destroy_checkpoint_caches();
1274 destroy_gc_caches();
1275 destroy_segment_manager_caches();
1276 destroy_node_manager_caches();
1277 destroy_inodecache();
1278 kset_unregister(f2fs_kset);
1281 module_init(init_f2fs_fs)
1282 module_exit(exit_f2fs_fs)
1284 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1285 MODULE_DESCRIPTION("Flash Friendly File System");
1286 MODULE_LICENSE("GPL");