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>
34 #define CREATE_TRACE_POINTS
35 #include <trace/events/f2fs.h>
37 static struct proc_dir_entry *f2fs_proc_root;
38 static struct kmem_cache *f2fs_inode_cachep;
39 static struct kset *f2fs_kset;
43 Opt_disable_roll_forward,
51 Opt_disable_ext_identify,
57 static match_table_t f2fs_tokens = {
58 {Opt_gc_background, "background_gc=%s"},
59 {Opt_disable_roll_forward, "disable_roll_forward"},
60 {Opt_discard, "discard"},
61 {Opt_noheap, "no_heap"},
62 {Opt_user_xattr, "user_xattr"},
63 {Opt_nouser_xattr, "nouser_xattr"},
66 {Opt_active_logs, "active_logs=%u"},
67 {Opt_disable_ext_identify, "disable_ext_identify"},
68 {Opt_inline_xattr, "inline_xattr"},
69 {Opt_inline_data, "inline_data"},
73 /* Sysfs support for f2fs */
75 GC_THREAD, /* struct f2fs_gc_thread */
76 SM_INFO, /* struct f2fs_sm_info */
80 struct attribute attr;
81 ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
82 ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
83 const char *, size_t);
88 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
90 if (struct_type == GC_THREAD)
91 return (unsigned char *)sbi->gc_thread;
92 else if (struct_type == SM_INFO)
93 return (unsigned char *)SM_I(sbi);
97 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
98 struct f2fs_sb_info *sbi, char *buf)
100 unsigned char *ptr = NULL;
103 ptr = __struct_ptr(sbi, a->struct_type);
107 ui = (unsigned int *)(ptr + a->offset);
109 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
112 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
113 struct f2fs_sb_info *sbi,
114 const char *buf, size_t count)
121 ptr = __struct_ptr(sbi, a->struct_type);
125 ui = (unsigned int *)(ptr + a->offset);
127 ret = kstrtoul(skip_spaces(buf), 0, &t);
134 static ssize_t f2fs_attr_show(struct kobject *kobj,
135 struct attribute *attr, char *buf)
137 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
139 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
141 return a->show ? a->show(a, sbi, buf) : 0;
144 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
145 const char *buf, size_t len)
147 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
149 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
151 return a->store ? a->store(a, sbi, buf, len) : 0;
154 static void f2fs_sb_release(struct kobject *kobj)
156 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
158 complete(&sbi->s_kobj_unregister);
161 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
162 static struct f2fs_attr f2fs_attr_##_name = { \
163 .attr = {.name = __stringify(_name), .mode = _mode }, \
166 .struct_type = _struct_type, \
170 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
171 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
172 f2fs_sbi_show, f2fs_sbi_store, \
173 offsetof(struct struct_name, elname))
175 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
176 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
177 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
178 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
179 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
180 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
182 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
183 static struct attribute *f2fs_attrs[] = {
184 ATTR_LIST(gc_min_sleep_time),
185 ATTR_LIST(gc_max_sleep_time),
186 ATTR_LIST(gc_no_gc_sleep_time),
188 ATTR_LIST(reclaim_segments),
189 ATTR_LIST(max_small_discards),
193 static const struct sysfs_ops f2fs_attr_ops = {
194 .show = f2fs_attr_show,
195 .store = f2fs_attr_store,
198 static struct kobj_type f2fs_ktype = {
199 .default_attrs = f2fs_attrs,
200 .sysfs_ops = &f2fs_attr_ops,
201 .release = f2fs_sb_release,
204 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
206 struct va_format vaf;
212 printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
216 static void init_once(void *foo)
218 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
220 inode_init_once(&fi->vfs_inode);
223 static int parse_options(struct super_block *sb, char *options)
225 struct f2fs_sb_info *sbi = F2FS_SB(sb);
226 substring_t args[MAX_OPT_ARGS];
233 while ((p = strsep(&options, ",")) != NULL) {
238 * Initialize args struct so we know whether arg was
239 * found; some options take optional arguments.
241 args[0].to = args[0].from = NULL;
242 token = match_token(p, f2fs_tokens, args);
245 case Opt_gc_background:
246 name = match_strdup(&args[0]);
250 if (!strncmp(name, "on", 2))
252 else if (!strncmp(name, "off", 3))
253 clear_opt(sbi, BG_GC);
260 case Opt_disable_roll_forward:
261 set_opt(sbi, DISABLE_ROLL_FORWARD);
264 set_opt(sbi, DISCARD);
267 set_opt(sbi, NOHEAP);
269 #ifdef CONFIG_F2FS_FS_XATTR
271 set_opt(sbi, XATTR_USER);
273 case Opt_nouser_xattr:
274 clear_opt(sbi, XATTR_USER);
276 case Opt_inline_xattr:
277 set_opt(sbi, INLINE_XATTR);
281 f2fs_msg(sb, KERN_INFO,
282 "user_xattr options not supported");
284 case Opt_nouser_xattr:
285 f2fs_msg(sb, KERN_INFO,
286 "nouser_xattr options not supported");
288 case Opt_inline_xattr:
289 f2fs_msg(sb, KERN_INFO,
290 "inline_xattr options not supported");
293 #ifdef CONFIG_F2FS_FS_POSIX_ACL
295 set_opt(sbi, POSIX_ACL);
298 clear_opt(sbi, POSIX_ACL);
302 f2fs_msg(sb, KERN_INFO, "acl options not supported");
305 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
308 case Opt_active_logs:
309 if (args->from && match_int(args, &arg))
311 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
313 sbi->active_logs = arg;
315 case Opt_disable_ext_identify:
316 set_opt(sbi, DISABLE_EXT_IDENTIFY);
318 case Opt_inline_data:
319 set_opt(sbi, INLINE_DATA);
322 f2fs_msg(sb, KERN_ERR,
323 "Unrecognized mount option \"%s\" or missing value",
331 static struct inode *f2fs_alloc_inode(struct super_block *sb)
333 struct f2fs_inode_info *fi;
335 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_NOFS | __GFP_ZERO);
339 init_once((void *) fi);
341 /* Initialize f2fs-specific inode info */
342 fi->vfs_inode.i_version = 1;
343 atomic_set(&fi->dirty_dents, 0);
344 fi->i_current_depth = 1;
346 rwlock_init(&fi->ext.ext_lock);
348 set_inode_flag(fi, FI_NEW_INODE);
350 if (test_opt(F2FS_SB(sb), INLINE_XATTR))
351 set_inode_flag(fi, FI_INLINE_XATTR);
353 return &fi->vfs_inode;
356 static int f2fs_drop_inode(struct inode *inode)
359 * This is to avoid a deadlock condition like below.
360 * writeback_single_inode(inode)
361 * - f2fs_write_data_page
362 * - f2fs_gc -> iput -> evict
363 * - inode_wait_for_writeback(inode)
365 if (!inode_unhashed(inode) && inode->i_state & I_SYNC)
367 return generic_drop_inode(inode);
371 * f2fs_dirty_inode() is called from __mark_inode_dirty()
373 * We should call set_dirty_inode to write the dirty inode through write_inode.
375 static void f2fs_dirty_inode(struct inode *inode, int flags)
377 set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
380 static void f2fs_i_callback(struct rcu_head *head)
382 struct inode *inode = container_of(head, struct inode, i_rcu);
383 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
386 static void f2fs_destroy_inode(struct inode *inode)
388 call_rcu(&inode->i_rcu, f2fs_i_callback);
391 static void f2fs_put_super(struct super_block *sb)
393 struct f2fs_sb_info *sbi = F2FS_SB(sb);
396 remove_proc_entry("segment_info", sbi->s_proc);
397 remove_proc_entry(sb->s_id, f2fs_proc_root);
399 kobject_del(&sbi->s_kobj);
401 f2fs_destroy_stats(sbi);
404 /* We don't need to do checkpoint when it's clean */
405 if (sbi->s_dirty && get_pages(sbi, F2FS_DIRTY_NODES))
406 write_checkpoint(sbi, true);
408 iput(sbi->node_inode);
409 iput(sbi->meta_inode);
411 /* destroy f2fs internal modules */
412 destroy_node_manager(sbi);
413 destroy_segment_manager(sbi);
416 kobject_put(&sbi->s_kobj);
417 wait_for_completion(&sbi->s_kobj_unregister);
419 sb->s_fs_info = NULL;
420 brelse(sbi->raw_super_buf);
424 int f2fs_sync_fs(struct super_block *sb, int sync)
426 struct f2fs_sb_info *sbi = F2FS_SB(sb);
428 trace_f2fs_sync_fs(sb, sync);
430 if (!sbi->s_dirty && !get_pages(sbi, F2FS_DIRTY_NODES))
434 mutex_lock(&sbi->gc_mutex);
435 write_checkpoint(sbi, false);
436 mutex_unlock(&sbi->gc_mutex);
438 f2fs_balance_fs(sbi);
444 static int f2fs_freeze(struct super_block *sb)
448 if (f2fs_readonly(sb))
451 err = f2fs_sync_fs(sb, 1);
455 static int f2fs_unfreeze(struct super_block *sb)
460 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
462 struct super_block *sb = dentry->d_sb;
463 struct f2fs_sb_info *sbi = F2FS_SB(sb);
464 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
465 block_t total_count, user_block_count, start_count, ovp_count;
467 total_count = le64_to_cpu(sbi->raw_super->block_count);
468 user_block_count = sbi->user_block_count;
469 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
470 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
471 buf->f_type = F2FS_SUPER_MAGIC;
472 buf->f_bsize = sbi->blocksize;
474 buf->f_blocks = total_count - start_count;
475 buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
476 buf->f_bavail = user_block_count - valid_user_blocks(sbi);
478 buf->f_files = sbi->total_node_count;
479 buf->f_ffree = sbi->total_node_count - valid_inode_count(sbi);
481 buf->f_namelen = F2FS_NAME_LEN;
482 buf->f_fsid.val[0] = (u32)id;
483 buf->f_fsid.val[1] = (u32)(id >> 32);
488 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
490 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
492 if (!(root->d_sb->s_flags & MS_RDONLY) && test_opt(sbi, BG_GC))
493 seq_printf(seq, ",background_gc=%s", "on");
495 seq_printf(seq, ",background_gc=%s", "off");
496 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
497 seq_puts(seq, ",disable_roll_forward");
498 if (test_opt(sbi, DISCARD))
499 seq_puts(seq, ",discard");
500 if (test_opt(sbi, NOHEAP))
501 seq_puts(seq, ",no_heap_alloc");
502 #ifdef CONFIG_F2FS_FS_XATTR
503 if (test_opt(sbi, XATTR_USER))
504 seq_puts(seq, ",user_xattr");
506 seq_puts(seq, ",nouser_xattr");
507 if (test_opt(sbi, INLINE_XATTR))
508 seq_puts(seq, ",inline_xattr");
510 #ifdef CONFIG_F2FS_FS_POSIX_ACL
511 if (test_opt(sbi, POSIX_ACL))
512 seq_puts(seq, ",acl");
514 seq_puts(seq, ",noacl");
516 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
517 seq_puts(seq, ",disable_ext_identify");
518 if (test_opt(sbi, INLINE_DATA))
519 seq_puts(seq, ",inline_data");
520 seq_printf(seq, ",active_logs=%u", sbi->active_logs);
525 static int segment_info_seq_show(struct seq_file *seq, void *offset)
527 struct super_block *sb = seq->private;
528 struct f2fs_sb_info *sbi = F2FS_SB(sb);
529 unsigned int total_segs = le32_to_cpu(sbi->raw_super->segment_count_main);
532 for (i = 0; i < total_segs; i++) {
533 seq_printf(seq, "%u", get_valid_blocks(sbi, i, 1));
534 if (i != 0 && (i % 10) == 0)
542 static int segment_info_open_fs(struct inode *inode, struct file *file)
544 return single_open(file, segment_info_seq_show, PDE_DATA(inode));
547 static const struct file_operations f2fs_seq_segment_info_fops = {
548 .owner = THIS_MODULE,
549 .open = segment_info_open_fs,
552 .release = single_release,
555 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
557 struct f2fs_sb_info *sbi = F2FS_SB(sb);
558 struct f2fs_mount_info org_mount_opt;
559 int err, active_logs;
562 * Save the old mount options in case we
563 * need to restore them.
565 org_mount_opt = sbi->mount_opt;
566 active_logs = sbi->active_logs;
568 /* parse mount options */
569 err = parse_options(sb, data);
574 * Previous and new state of filesystem is RO,
575 * so no point in checking GC conditions.
577 if ((sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY))
581 * We stop the GC thread if FS is mounted as RO
582 * or if background_gc = off is passed in mount
583 * option. Also sync the filesystem.
585 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
586 if (sbi->gc_thread) {
590 } else if (test_opt(sbi, BG_GC) && !sbi->gc_thread) {
591 err = start_gc_thread(sbi);
596 /* Update the POSIXACL Flag */
597 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
598 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
602 sbi->mount_opt = org_mount_opt;
603 sbi->active_logs = active_logs;
607 static struct super_operations f2fs_sops = {
608 .alloc_inode = f2fs_alloc_inode,
609 .drop_inode = f2fs_drop_inode,
610 .destroy_inode = f2fs_destroy_inode,
611 .write_inode = f2fs_write_inode,
612 .dirty_inode = f2fs_dirty_inode,
613 .show_options = f2fs_show_options,
614 .evict_inode = f2fs_evict_inode,
615 .put_super = f2fs_put_super,
616 .sync_fs = f2fs_sync_fs,
617 .freeze_fs = f2fs_freeze,
618 .unfreeze_fs = f2fs_unfreeze,
619 .statfs = f2fs_statfs,
620 .remount_fs = f2fs_remount,
623 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
624 u64 ino, u32 generation)
626 struct f2fs_sb_info *sbi = F2FS_SB(sb);
629 if (ino < F2FS_ROOT_INO(sbi))
630 return ERR_PTR(-ESTALE);
633 * f2fs_iget isn't quite right if the inode is currently unallocated!
634 * However f2fs_iget currently does appropriate checks to handle stale
635 * inodes so everything is OK.
637 inode = f2fs_iget(sb, ino);
639 return ERR_CAST(inode);
640 if (generation && inode->i_generation != generation) {
641 /* we didn't find the right inode.. */
643 return ERR_PTR(-ESTALE);
648 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
649 int fh_len, int fh_type)
651 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
655 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
656 int fh_len, int fh_type)
658 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
662 static const struct export_operations f2fs_export_ops = {
663 .fh_to_dentry = f2fs_fh_to_dentry,
664 .fh_to_parent = f2fs_fh_to_parent,
665 .get_parent = f2fs_get_parent,
668 static loff_t max_file_size(unsigned bits)
670 loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
671 loff_t leaf_count = ADDRS_PER_BLOCK;
673 /* two direct node blocks */
674 result += (leaf_count * 2);
676 /* two indirect node blocks */
677 leaf_count *= NIDS_PER_BLOCK;
678 result += (leaf_count * 2);
680 /* one double indirect node block */
681 leaf_count *= NIDS_PER_BLOCK;
682 result += leaf_count;
688 static int sanity_check_raw_super(struct super_block *sb,
689 struct f2fs_super_block *raw_super)
691 unsigned int blocksize;
693 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
694 f2fs_msg(sb, KERN_INFO,
695 "Magic Mismatch, valid(0x%x) - read(0x%x)",
696 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
700 /* Currently, support only 4KB page cache size */
701 if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
702 f2fs_msg(sb, KERN_INFO,
703 "Invalid page_cache_size (%lu), supports only 4KB\n",
708 /* Currently, support only 4KB block size */
709 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
710 if (blocksize != F2FS_BLKSIZE) {
711 f2fs_msg(sb, KERN_INFO,
712 "Invalid blocksize (%u), supports only 4KB\n",
717 if (le32_to_cpu(raw_super->log_sectorsize) !=
718 F2FS_LOG_SECTOR_SIZE) {
719 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize");
722 if (le32_to_cpu(raw_super->log_sectors_per_block) !=
723 F2FS_LOG_SECTORS_PER_BLOCK) {
724 f2fs_msg(sb, KERN_INFO, "Invalid log sectors per block");
730 static int sanity_check_ckpt(struct f2fs_sb_info *sbi)
732 unsigned int total, fsmeta;
733 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
734 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
736 total = le32_to_cpu(raw_super->segment_count);
737 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
738 fsmeta += le32_to_cpu(raw_super->segment_count_sit);
739 fsmeta += le32_to_cpu(raw_super->segment_count_nat);
740 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
741 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
746 if (is_set_ckpt_flags(ckpt, CP_ERROR_FLAG)) {
747 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
753 static void init_sb_info(struct f2fs_sb_info *sbi)
755 struct f2fs_super_block *raw_super = sbi->raw_super;
758 sbi->log_sectors_per_block =
759 le32_to_cpu(raw_super->log_sectors_per_block);
760 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
761 sbi->blocksize = 1 << sbi->log_blocksize;
762 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
763 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
764 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
765 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
766 sbi->total_sections = le32_to_cpu(raw_super->section_count);
767 sbi->total_node_count =
768 (le32_to_cpu(raw_super->segment_count_nat) / 2)
769 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
770 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
771 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
772 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
773 sbi->cur_victim_sec = NULL_SECNO;
775 for (i = 0; i < NR_COUNT_TYPE; i++)
776 atomic_set(&sbi->nr_pages[i], 0);
780 * Read f2fs raw super block.
781 * Because we have two copies of super block, so read the first one at first,
782 * if the first one is invalid, move to read the second one.
784 static int read_raw_super_block(struct super_block *sb,
785 struct f2fs_super_block **raw_super,
786 struct buffer_head **raw_super_buf)
791 *raw_super_buf = sb_bread(sb, block);
792 if (!*raw_super_buf) {
793 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
803 *raw_super = (struct f2fs_super_block *)
804 ((char *)(*raw_super_buf)->b_data + F2FS_SUPER_OFFSET);
806 /* sanity checking of raw super */
807 if (sanity_check_raw_super(sb, *raw_super)) {
808 brelse(*raw_super_buf);
809 f2fs_msg(sb, KERN_ERR, "Can't find a valid F2FS filesystem "
810 "in %dth superblock", block + 1);
822 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
824 struct f2fs_sb_info *sbi;
825 struct f2fs_super_block *raw_super;
826 struct buffer_head *raw_super_buf;
831 /* allocate memory for f2fs-specific super block info */
832 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
836 /* set a block size */
837 if (!sb_set_blocksize(sb, F2FS_BLKSIZE)) {
838 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
842 err = read_raw_super_block(sb, &raw_super, &raw_super_buf);
847 /* init some FS parameters */
848 sbi->active_logs = NR_CURSEG_TYPE;
852 #ifdef CONFIG_F2FS_FS_XATTR
853 set_opt(sbi, XATTR_USER);
855 #ifdef CONFIG_F2FS_FS_POSIX_ACL
856 set_opt(sbi, POSIX_ACL);
858 /* parse mount options */
859 err = parse_options(sb, (char *)data);
863 sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
864 sb->s_max_links = F2FS_LINK_MAX;
865 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
867 sb->s_op = &f2fs_sops;
868 sb->s_xattr = f2fs_xattr_handlers;
869 sb->s_export_op = &f2fs_export_ops;
870 sb->s_magic = F2FS_SUPER_MAGIC;
872 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
873 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
874 memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
876 /* init f2fs-specific super block info */
878 sbi->raw_super = raw_super;
879 sbi->raw_super_buf = raw_super_buf;
880 mutex_init(&sbi->gc_mutex);
881 mutex_init(&sbi->writepages);
882 mutex_init(&sbi->cp_mutex);
883 mutex_init(&sbi->node_write);
884 sbi->por_doing = false;
885 spin_lock_init(&sbi->stat_lock);
887 mutex_init(&sbi->read_io.io_mutex);
888 for (i = 0; i < NR_PAGE_TYPE; i++)
889 mutex_init(&sbi->write_io[i].io_mutex);
891 init_rwsem(&sbi->cp_rwsem);
892 init_waitqueue_head(&sbi->cp_wait);
895 /* get an inode for meta space */
896 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
897 if (IS_ERR(sbi->meta_inode)) {
898 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
899 err = PTR_ERR(sbi->meta_inode);
903 err = get_valid_checkpoint(sbi);
905 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
906 goto free_meta_inode;
909 /* sanity checking of checkpoint */
911 if (sanity_check_ckpt(sbi)) {
912 f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint");
916 sbi->total_valid_node_count =
917 le32_to_cpu(sbi->ckpt->valid_node_count);
918 sbi->total_valid_inode_count =
919 le32_to_cpu(sbi->ckpt->valid_inode_count);
920 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
921 sbi->total_valid_block_count =
922 le64_to_cpu(sbi->ckpt->valid_block_count);
923 sbi->last_valid_block_count = sbi->total_valid_block_count;
924 sbi->alloc_valid_block_count = 0;
925 INIT_LIST_HEAD(&sbi->dir_inode_list);
926 spin_lock_init(&sbi->dir_inode_lock);
928 init_orphan_info(sbi);
930 /* setup f2fs internal modules */
931 err = build_segment_manager(sbi);
933 f2fs_msg(sb, KERN_ERR,
934 "Failed to initialize F2FS segment manager");
937 err = build_node_manager(sbi);
939 f2fs_msg(sb, KERN_ERR,
940 "Failed to initialize F2FS node manager");
944 build_gc_manager(sbi);
946 /* get an inode for node space */
947 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
948 if (IS_ERR(sbi->node_inode)) {
949 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
950 err = PTR_ERR(sbi->node_inode);
954 /* if there are nt orphan nodes free them */
955 recover_orphan_inodes(sbi);
957 /* read root inode and dentry */
958 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
960 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
962 goto free_node_inode;
964 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
966 goto free_root_inode;
969 sb->s_root = d_make_root(root); /* allocate root dentry */
972 goto free_root_inode;
975 /* recover fsynced data */
976 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
977 err = recover_fsync_data(sbi);
979 f2fs_msg(sb, KERN_ERR,
980 "Cannot recover all fsync data errno=%ld", err);
984 * If filesystem is not mounted as read-only then
985 * do start the gc_thread.
987 if (!(sb->s_flags & MS_RDONLY)) {
988 /* After POR, we can run background GC thread.*/
989 err = start_gc_thread(sbi);
994 err = f2fs_build_stats(sbi);
999 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1002 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1003 &f2fs_seq_segment_info_fops, sb);
1005 if (test_opt(sbi, DISCARD)) {
1006 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1007 if (!blk_queue_discard(q))
1008 f2fs_msg(sb, KERN_WARNING,
1009 "mounting with \"discard\" option, but "
1010 "the device does not support discard");
1013 sbi->s_kobj.kset = f2fs_kset;
1014 init_completion(&sbi->s_kobj_unregister);
1015 err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1023 remove_proc_entry("segment_info", sbi->s_proc);
1024 remove_proc_entry(sb->s_id, f2fs_proc_root);
1026 f2fs_destroy_stats(sbi);
1028 stop_gc_thread(sbi);
1033 iput(sbi->node_inode);
1035 destroy_node_manager(sbi);
1037 destroy_segment_manager(sbi);
1041 make_bad_inode(sbi->meta_inode);
1042 iput(sbi->meta_inode);
1044 brelse(raw_super_buf);
1050 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1051 const char *dev_name, void *data)
1053 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1056 static struct file_system_type f2fs_fs_type = {
1057 .owner = THIS_MODULE,
1059 .mount = f2fs_mount,
1060 .kill_sb = kill_block_super,
1061 .fs_flags = FS_REQUIRES_DEV,
1063 MODULE_ALIAS_FS("f2fs");
1065 static int __init init_inodecache(void)
1067 f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
1068 sizeof(struct f2fs_inode_info), NULL);
1069 if (f2fs_inode_cachep == NULL)
1074 static void destroy_inodecache(void)
1077 * Make sure all delayed rcu free inodes are flushed before we
1081 kmem_cache_destroy(f2fs_inode_cachep);
1084 static int __init init_f2fs_fs(void)
1088 err = init_inodecache();
1091 err = create_node_manager_caches();
1093 goto free_inodecache;
1094 err = create_segment_manager_caches();
1096 goto free_node_manager_caches;
1097 err = create_gc_caches();
1099 goto free_segment_manager_caches;
1100 err = create_checkpoint_caches();
1102 goto free_gc_caches;
1103 f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1106 goto free_checkpoint_caches;
1108 err = register_filesystem(&f2fs_fs_type);
1111 f2fs_create_root_stats();
1112 f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1116 kset_unregister(f2fs_kset);
1117 free_checkpoint_caches:
1118 destroy_checkpoint_caches();
1120 destroy_gc_caches();
1121 free_segment_manager_caches:
1122 destroy_segment_manager_caches();
1123 free_node_manager_caches:
1124 destroy_node_manager_caches();
1126 destroy_inodecache();
1131 static void __exit exit_f2fs_fs(void)
1133 remove_proc_entry("fs/f2fs", NULL);
1134 f2fs_destroy_root_stats();
1135 unregister_filesystem(&f2fs_fs_type);
1136 destroy_checkpoint_caches();
1137 destroy_gc_caches();
1138 destroy_node_manager_caches();
1139 destroy_inodecache();
1140 kset_unregister(f2fs_kset);
1143 module_init(init_f2fs_fs)
1144 module_exit(exit_f2fs_fs)
1146 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1147 MODULE_DESCRIPTION("Flash Friendly File System");
1148 MODULE_LICENSE("GPL");