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,
59 static match_table_t f2fs_tokens = {
60 {Opt_gc_background, "background_gc=%s"},
61 {Opt_disable_roll_forward, "disable_roll_forward"},
62 {Opt_discard, "discard"},
63 {Opt_noheap, "no_heap"},
64 {Opt_user_xattr, "user_xattr"},
65 {Opt_nouser_xattr, "nouser_xattr"},
68 {Opt_active_logs, "active_logs=%u"},
69 {Opt_disable_ext_identify, "disable_ext_identify"},
70 {Opt_inline_xattr, "inline_xattr"},
71 {Opt_inline_data, "inline_data"},
72 {Opt_flush_merge, "flush_merge"},
73 {Opt_nobarrier, "nobarrier"},
77 /* Sysfs support for f2fs */
79 GC_THREAD, /* struct f2fs_gc_thread */
80 SM_INFO, /* struct f2fs_sm_info */
81 NM_INFO, /* struct f2fs_nm_info */
82 F2FS_SBI, /* struct f2fs_sb_info */
86 struct attribute attr;
87 ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
88 ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
89 const char *, size_t);
94 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
96 if (struct_type == GC_THREAD)
97 return (unsigned char *)sbi->gc_thread;
98 else if (struct_type == SM_INFO)
99 return (unsigned char *)SM_I(sbi);
100 else if (struct_type == NM_INFO)
101 return (unsigned char *)NM_I(sbi);
102 else if (struct_type == F2FS_SBI)
103 return (unsigned char *)sbi;
107 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
108 struct f2fs_sb_info *sbi, char *buf)
110 unsigned char *ptr = NULL;
113 ptr = __struct_ptr(sbi, a->struct_type);
117 ui = (unsigned int *)(ptr + a->offset);
119 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
122 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
123 struct f2fs_sb_info *sbi,
124 const char *buf, size_t count)
131 ptr = __struct_ptr(sbi, a->struct_type);
135 ui = (unsigned int *)(ptr + a->offset);
137 ret = kstrtoul(skip_spaces(buf), 0, &t);
144 static ssize_t f2fs_attr_show(struct kobject *kobj,
145 struct attribute *attr, char *buf)
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->show ? a->show(a, sbi, buf) : 0;
154 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
155 const char *buf, size_t len)
157 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
159 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
161 return a->store ? a->store(a, sbi, buf, len) : 0;
164 static void f2fs_sb_release(struct kobject *kobj)
166 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
168 complete(&sbi->s_kobj_unregister);
171 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
172 static struct f2fs_attr f2fs_attr_##_name = { \
173 .attr = {.name = __stringify(_name), .mode = _mode }, \
176 .struct_type = _struct_type, \
180 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
181 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
182 f2fs_sbi_show, f2fs_sbi_store, \
183 offsetof(struct struct_name, elname))
185 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
186 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
187 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
188 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
189 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
190 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
191 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
192 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
193 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks);
194 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
195 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
196 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
198 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
199 static struct attribute *f2fs_attrs[] = {
200 ATTR_LIST(gc_min_sleep_time),
201 ATTR_LIST(gc_max_sleep_time),
202 ATTR_LIST(gc_no_gc_sleep_time),
204 ATTR_LIST(reclaim_segments),
205 ATTR_LIST(max_small_discards),
206 ATTR_LIST(ipu_policy),
207 ATTR_LIST(min_ipu_util),
208 ATTR_LIST(min_fsync_blocks),
209 ATTR_LIST(max_victim_search),
210 ATTR_LIST(dir_level),
211 ATTR_LIST(ram_thresh),
215 static const struct sysfs_ops f2fs_attr_ops = {
216 .show = f2fs_attr_show,
217 .store = f2fs_attr_store,
220 static struct kobj_type f2fs_ktype = {
221 .default_attrs = f2fs_attrs,
222 .sysfs_ops = &f2fs_attr_ops,
223 .release = f2fs_sb_release,
226 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
228 struct va_format vaf;
234 printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
238 static void init_once(void *foo)
240 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
242 inode_init_once(&fi->vfs_inode);
245 static int parse_options(struct super_block *sb, char *options)
247 struct f2fs_sb_info *sbi = F2FS_SB(sb);
248 substring_t args[MAX_OPT_ARGS];
255 while ((p = strsep(&options, ",")) != NULL) {
260 * Initialize args struct so we know whether arg was
261 * found; some options take optional arguments.
263 args[0].to = args[0].from = NULL;
264 token = match_token(p, f2fs_tokens, args);
267 case Opt_gc_background:
268 name = match_strdup(&args[0]);
272 if (strlen(name) == 2 && !strncmp(name, "on", 2))
274 else if (strlen(name) == 3 && !strncmp(name, "off", 3))
275 clear_opt(sbi, BG_GC);
282 case Opt_disable_roll_forward:
283 set_opt(sbi, DISABLE_ROLL_FORWARD);
286 set_opt(sbi, DISCARD);
289 set_opt(sbi, NOHEAP);
291 #ifdef CONFIG_F2FS_FS_XATTR
293 set_opt(sbi, XATTR_USER);
295 case Opt_nouser_xattr:
296 clear_opt(sbi, XATTR_USER);
298 case Opt_inline_xattr:
299 set_opt(sbi, INLINE_XATTR);
303 f2fs_msg(sb, KERN_INFO,
304 "user_xattr options not supported");
306 case Opt_nouser_xattr:
307 f2fs_msg(sb, KERN_INFO,
308 "nouser_xattr options not supported");
310 case Opt_inline_xattr:
311 f2fs_msg(sb, KERN_INFO,
312 "inline_xattr options not supported");
315 #ifdef CONFIG_F2FS_FS_POSIX_ACL
317 set_opt(sbi, POSIX_ACL);
320 clear_opt(sbi, POSIX_ACL);
324 f2fs_msg(sb, KERN_INFO, "acl options not supported");
327 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
330 case Opt_active_logs:
331 if (args->from && match_int(args, &arg))
333 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
335 sbi->active_logs = arg;
337 case Opt_disable_ext_identify:
338 set_opt(sbi, DISABLE_EXT_IDENTIFY);
340 case Opt_inline_data:
341 set_opt(sbi, INLINE_DATA);
343 case Opt_flush_merge:
344 set_opt(sbi, FLUSH_MERGE);
347 set_opt(sbi, NOBARRIER);
350 f2fs_msg(sb, KERN_ERR,
351 "Unrecognized mount option \"%s\" or missing value",
359 static struct inode *f2fs_alloc_inode(struct super_block *sb)
361 struct f2fs_inode_info *fi;
363 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
367 init_once((void *) fi);
369 /* Initialize f2fs-specific inode info */
370 fi->vfs_inode.i_version = 1;
371 atomic_set(&fi->dirty_pages, 0);
372 fi->i_current_depth = 1;
374 rwlock_init(&fi->ext.ext_lock);
375 init_rwsem(&fi->i_sem);
377 set_inode_flag(fi, FI_NEW_INODE);
379 if (test_opt(F2FS_SB(sb), INLINE_XATTR))
380 set_inode_flag(fi, FI_INLINE_XATTR);
382 /* Will be used by directory only */
383 fi->i_dir_level = F2FS_SB(sb)->dir_level;
385 return &fi->vfs_inode;
388 static int f2fs_drop_inode(struct inode *inode)
391 * This is to avoid a deadlock condition like below.
392 * writeback_single_inode(inode)
393 * - f2fs_write_data_page
394 * - f2fs_gc -> iput -> evict
395 * - inode_wait_for_writeback(inode)
397 if (!inode_unhashed(inode) && inode->i_state & I_SYNC)
399 return generic_drop_inode(inode);
403 * f2fs_dirty_inode() is called from __mark_inode_dirty()
405 * We should call set_dirty_inode to write the dirty inode through write_inode.
407 static void f2fs_dirty_inode(struct inode *inode, int flags)
409 set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
412 static void f2fs_i_callback(struct rcu_head *head)
414 struct inode *inode = container_of(head, struct inode, i_rcu);
415 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
418 static void f2fs_destroy_inode(struct inode *inode)
420 call_rcu(&inode->i_rcu, f2fs_i_callback);
423 static void f2fs_put_super(struct super_block *sb)
425 struct f2fs_sb_info *sbi = F2FS_SB(sb);
428 remove_proc_entry("segment_info", sbi->s_proc);
429 remove_proc_entry(sb->s_id, f2fs_proc_root);
431 kobject_del(&sbi->s_kobj);
433 f2fs_destroy_stats(sbi);
436 /* We don't need to do checkpoint when it's clean */
438 write_checkpoint(sbi, true);
441 * normally superblock is clean, so we need to release this.
442 * In addition, EIO will skip do checkpoint, we need this as well.
444 release_dirty_inode(sbi);
446 iput(sbi->node_inode);
447 iput(sbi->meta_inode);
449 /* destroy f2fs internal modules */
450 destroy_node_manager(sbi);
451 destroy_segment_manager(sbi);
454 kobject_put(&sbi->s_kobj);
455 wait_for_completion(&sbi->s_kobj_unregister);
457 sb->s_fs_info = NULL;
458 brelse(sbi->raw_super_buf);
462 int f2fs_sync_fs(struct super_block *sb, int sync)
464 struct f2fs_sb_info *sbi = F2FS_SB(sb);
466 trace_f2fs_sync_fs(sb, sync);
469 mutex_lock(&sbi->gc_mutex);
470 write_checkpoint(sbi, false);
471 mutex_unlock(&sbi->gc_mutex);
473 f2fs_balance_fs(sbi);
479 static int f2fs_freeze(struct super_block *sb)
483 if (f2fs_readonly(sb))
486 err = f2fs_sync_fs(sb, 1);
490 static int f2fs_unfreeze(struct super_block *sb)
495 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
497 struct super_block *sb = dentry->d_sb;
498 struct f2fs_sb_info *sbi = F2FS_SB(sb);
499 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
500 block_t total_count, user_block_count, start_count, ovp_count;
502 total_count = le64_to_cpu(sbi->raw_super->block_count);
503 user_block_count = sbi->user_block_count;
504 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
505 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
506 buf->f_type = F2FS_SUPER_MAGIC;
507 buf->f_bsize = sbi->blocksize;
509 buf->f_blocks = total_count - start_count;
510 buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
511 buf->f_bavail = user_block_count - valid_user_blocks(sbi);
513 buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
514 buf->f_ffree = buf->f_files - valid_inode_count(sbi);
516 buf->f_namelen = F2FS_NAME_LEN;
517 buf->f_fsid.val[0] = (u32)id;
518 buf->f_fsid.val[1] = (u32)(id >> 32);
523 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
525 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
527 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC))
528 seq_printf(seq, ",background_gc=%s", "on");
530 seq_printf(seq, ",background_gc=%s", "off");
531 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
532 seq_puts(seq, ",disable_roll_forward");
533 if (test_opt(sbi, DISCARD))
534 seq_puts(seq, ",discard");
535 if (test_opt(sbi, NOHEAP))
536 seq_puts(seq, ",no_heap_alloc");
537 #ifdef CONFIG_F2FS_FS_XATTR
538 if (test_opt(sbi, XATTR_USER))
539 seq_puts(seq, ",user_xattr");
541 seq_puts(seq, ",nouser_xattr");
542 if (test_opt(sbi, INLINE_XATTR))
543 seq_puts(seq, ",inline_xattr");
545 #ifdef CONFIG_F2FS_FS_POSIX_ACL
546 if (test_opt(sbi, POSIX_ACL))
547 seq_puts(seq, ",acl");
549 seq_puts(seq, ",noacl");
551 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
552 seq_puts(seq, ",disable_ext_identify");
553 if (test_opt(sbi, INLINE_DATA))
554 seq_puts(seq, ",inline_data");
555 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
556 seq_puts(seq, ",flush_merge");
557 if (test_opt(sbi, NOBARRIER))
558 seq_puts(seq, ",nobarrier");
559 seq_printf(seq, ",active_logs=%u", sbi->active_logs);
564 static int segment_info_seq_show(struct seq_file *seq, void *offset)
566 struct super_block *sb = seq->private;
567 struct f2fs_sb_info *sbi = F2FS_SB(sb);
568 unsigned int total_segs =
569 le32_to_cpu(sbi->raw_super->segment_count_main);
572 seq_puts(seq, "format: segment_type|valid_blocks\n"
573 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
575 for (i = 0; i < total_segs; i++) {
576 struct seg_entry *se = get_seg_entry(sbi, i);
579 seq_printf(seq, "%-5d", i);
580 seq_printf(seq, "%d|%-3u", se->type,
581 get_valid_blocks(sbi, i, 1));
582 if ((i % 10) == 9 || i == (total_segs - 1))
591 static int segment_info_open_fs(struct inode *inode, struct file *file)
593 return single_open(file, segment_info_seq_show, PDE_DATA(inode));
596 static const struct file_operations f2fs_seq_segment_info_fops = {
597 .owner = THIS_MODULE,
598 .open = segment_info_open_fs,
601 .release = single_release,
604 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
606 struct f2fs_sb_info *sbi = F2FS_SB(sb);
607 struct f2fs_mount_info org_mount_opt;
608 int err, active_logs;
609 bool need_restart_gc = false;
610 bool need_stop_gc = false;
615 * Save the old mount options in case we
616 * need to restore them.
618 org_mount_opt = sbi->mount_opt;
619 active_logs = sbi->active_logs;
621 /* parse mount options */
622 err = parse_options(sb, data);
627 * Previous and new state of filesystem is RO,
628 * so skip checking GC and FLUSH_MERGE conditions.
630 if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
634 * We stop the GC thread if FS is mounted as RO
635 * or if background_gc = off is passed in mount
636 * option. Also sync the filesystem.
638 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
639 if (sbi->gc_thread) {
642 need_restart_gc = true;
644 } else if (test_opt(sbi, BG_GC) && !sbi->gc_thread) {
645 err = start_gc_thread(sbi);
652 * We stop issue flush thread if FS is mounted as RO
653 * or if flush_merge is not passed in mount option.
655 if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
656 destroy_flush_cmd_control(sbi);
657 } else if (test_opt(sbi, FLUSH_MERGE) && !SM_I(sbi)->cmd_control_info) {
658 err = create_flush_cmd_control(sbi);
663 /* Update the POSIXACL Flag */
664 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
665 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
668 if (need_restart_gc) {
669 if (start_gc_thread(sbi))
670 f2fs_msg(sbi->sb, KERN_WARNING,
671 "background gc thread has stopped");
672 } else if (need_stop_gc) {
676 sbi->mount_opt = org_mount_opt;
677 sbi->active_logs = active_logs;
681 static struct super_operations f2fs_sops = {
682 .alloc_inode = f2fs_alloc_inode,
683 .drop_inode = f2fs_drop_inode,
684 .destroy_inode = f2fs_destroy_inode,
685 .write_inode = f2fs_write_inode,
686 .dirty_inode = f2fs_dirty_inode,
687 .show_options = f2fs_show_options,
688 .evict_inode = f2fs_evict_inode,
689 .put_super = f2fs_put_super,
690 .sync_fs = f2fs_sync_fs,
691 .freeze_fs = f2fs_freeze,
692 .unfreeze_fs = f2fs_unfreeze,
693 .statfs = f2fs_statfs,
694 .remount_fs = f2fs_remount,
697 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
698 u64 ino, u32 generation)
700 struct f2fs_sb_info *sbi = F2FS_SB(sb);
703 if (check_nid_range(sbi, ino))
704 return ERR_PTR(-ESTALE);
707 * f2fs_iget isn't quite right if the inode is currently unallocated!
708 * However f2fs_iget currently does appropriate checks to handle stale
709 * inodes so everything is OK.
711 inode = f2fs_iget(sb, ino);
713 return ERR_CAST(inode);
714 if (unlikely(generation && inode->i_generation != generation)) {
715 /* we didn't find the right inode.. */
717 return ERR_PTR(-ESTALE);
722 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
723 int fh_len, int fh_type)
725 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
729 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
730 int fh_len, int fh_type)
732 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
736 static const struct export_operations f2fs_export_ops = {
737 .fh_to_dentry = f2fs_fh_to_dentry,
738 .fh_to_parent = f2fs_fh_to_parent,
739 .get_parent = f2fs_get_parent,
742 static loff_t max_file_size(unsigned bits)
744 loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
745 loff_t leaf_count = ADDRS_PER_BLOCK;
747 /* two direct node blocks */
748 result += (leaf_count * 2);
750 /* two indirect node blocks */
751 leaf_count *= NIDS_PER_BLOCK;
752 result += (leaf_count * 2);
754 /* one double indirect node block */
755 leaf_count *= NIDS_PER_BLOCK;
756 result += leaf_count;
762 static int sanity_check_raw_super(struct super_block *sb,
763 struct f2fs_super_block *raw_super)
765 unsigned int blocksize;
767 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
768 f2fs_msg(sb, KERN_INFO,
769 "Magic Mismatch, valid(0x%x) - read(0x%x)",
770 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
774 /* Currently, support only 4KB page cache size */
775 if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
776 f2fs_msg(sb, KERN_INFO,
777 "Invalid page_cache_size (%lu), supports only 4KB\n",
782 /* Currently, support only 4KB block size */
783 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
784 if (blocksize != F2FS_BLKSIZE) {
785 f2fs_msg(sb, KERN_INFO,
786 "Invalid blocksize (%u), supports only 4KB\n",
791 if (le32_to_cpu(raw_super->log_sectorsize) !=
792 F2FS_LOG_SECTOR_SIZE) {
793 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize");
796 if (le32_to_cpu(raw_super->log_sectors_per_block) !=
797 F2FS_LOG_SECTORS_PER_BLOCK) {
798 f2fs_msg(sb, KERN_INFO, "Invalid log sectors per block");
804 static int sanity_check_ckpt(struct f2fs_sb_info *sbi)
806 unsigned int total, fsmeta;
807 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
808 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
810 total = le32_to_cpu(raw_super->segment_count);
811 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
812 fsmeta += le32_to_cpu(raw_super->segment_count_sit);
813 fsmeta += le32_to_cpu(raw_super->segment_count_nat);
814 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
815 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
817 if (unlikely(fsmeta >= total))
820 if (unlikely(f2fs_cp_error(sbi))) {
821 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
827 static void init_sb_info(struct f2fs_sb_info *sbi)
829 struct f2fs_super_block *raw_super = sbi->raw_super;
832 sbi->log_sectors_per_block =
833 le32_to_cpu(raw_super->log_sectors_per_block);
834 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
835 sbi->blocksize = 1 << sbi->log_blocksize;
836 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
837 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
838 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
839 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
840 sbi->total_sections = le32_to_cpu(raw_super->section_count);
841 sbi->total_node_count =
842 (le32_to_cpu(raw_super->segment_count_nat) / 2)
843 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
844 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
845 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
846 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
847 sbi->cur_victim_sec = NULL_SECNO;
848 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
850 for (i = 0; i < NR_COUNT_TYPE; i++)
851 atomic_set(&sbi->nr_pages[i], 0);
853 sbi->dir_level = DEF_DIR_LEVEL;
854 sbi->need_fsck = false;
858 * Read f2fs raw super block.
859 * Because we have two copies of super block, so read the first one at first,
860 * if the first one is invalid, move to read the second one.
862 static int read_raw_super_block(struct super_block *sb,
863 struct f2fs_super_block **raw_super,
864 struct buffer_head **raw_super_buf)
869 *raw_super_buf = sb_bread(sb, block);
870 if (!*raw_super_buf) {
871 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
881 *raw_super = (struct f2fs_super_block *)
882 ((char *)(*raw_super_buf)->b_data + F2FS_SUPER_OFFSET);
884 /* sanity checking of raw super */
885 if (sanity_check_raw_super(sb, *raw_super)) {
886 brelse(*raw_super_buf);
887 f2fs_msg(sb, KERN_ERR,
888 "Can't find valid F2FS filesystem in %dth superblock",
901 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
903 struct f2fs_sb_info *sbi;
904 struct f2fs_super_block *raw_super;
905 struct buffer_head *raw_super_buf;
912 /* allocate memory for f2fs-specific super block info */
913 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
917 /* set a block size */
918 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
919 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
923 err = read_raw_super_block(sb, &raw_super, &raw_super_buf);
928 /* init some FS parameters */
929 sbi->active_logs = NR_CURSEG_TYPE;
933 #ifdef CONFIG_F2FS_FS_XATTR
934 set_opt(sbi, XATTR_USER);
936 #ifdef CONFIG_F2FS_FS_POSIX_ACL
937 set_opt(sbi, POSIX_ACL);
939 /* parse mount options */
940 err = parse_options(sb, (char *)data);
944 sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
945 sb->s_max_links = F2FS_LINK_MAX;
946 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
948 sb->s_op = &f2fs_sops;
949 sb->s_xattr = f2fs_xattr_handlers;
950 sb->s_export_op = &f2fs_export_ops;
951 sb->s_magic = F2FS_SUPER_MAGIC;
953 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
954 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
955 memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
957 /* init f2fs-specific super block info */
959 sbi->raw_super = raw_super;
960 sbi->raw_super_buf = raw_super_buf;
961 mutex_init(&sbi->gc_mutex);
962 mutex_init(&sbi->writepages);
963 mutex_init(&sbi->cp_mutex);
964 init_rwsem(&sbi->node_write);
965 sbi->por_doing = false;
966 spin_lock_init(&sbi->stat_lock);
968 init_rwsem(&sbi->read_io.io_rwsem);
969 sbi->read_io.sbi = sbi;
970 sbi->read_io.bio = NULL;
971 for (i = 0; i < NR_PAGE_TYPE; i++) {
972 init_rwsem(&sbi->write_io[i].io_rwsem);
973 sbi->write_io[i].sbi = sbi;
974 sbi->write_io[i].bio = NULL;
977 init_rwsem(&sbi->cp_rwsem);
978 init_waitqueue_head(&sbi->cp_wait);
981 /* get an inode for meta space */
982 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
983 if (IS_ERR(sbi->meta_inode)) {
984 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
985 err = PTR_ERR(sbi->meta_inode);
989 err = get_valid_checkpoint(sbi);
991 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
992 goto free_meta_inode;
995 /* sanity checking of checkpoint */
997 if (sanity_check_ckpt(sbi)) {
998 f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint");
1002 sbi->total_valid_node_count =
1003 le32_to_cpu(sbi->ckpt->valid_node_count);
1004 sbi->total_valid_inode_count =
1005 le32_to_cpu(sbi->ckpt->valid_inode_count);
1006 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1007 sbi->total_valid_block_count =
1008 le64_to_cpu(sbi->ckpt->valid_block_count);
1009 sbi->last_valid_block_count = sbi->total_valid_block_count;
1010 sbi->alloc_valid_block_count = 0;
1011 INIT_LIST_HEAD(&sbi->dir_inode_list);
1012 spin_lock_init(&sbi->dir_inode_lock);
1014 init_ino_entry_info(sbi);
1016 /* setup f2fs internal modules */
1017 err = build_segment_manager(sbi);
1019 f2fs_msg(sb, KERN_ERR,
1020 "Failed to initialize F2FS segment manager");
1023 err = build_node_manager(sbi);
1025 f2fs_msg(sb, KERN_ERR,
1026 "Failed to initialize F2FS node manager");
1030 build_gc_manager(sbi);
1032 /* get an inode for node space */
1033 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1034 if (IS_ERR(sbi->node_inode)) {
1035 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1036 err = PTR_ERR(sbi->node_inode);
1040 /* if there are nt orphan nodes free them */
1041 recover_orphan_inodes(sbi);
1043 /* read root inode and dentry */
1044 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1046 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1047 err = PTR_ERR(root);
1048 goto free_node_inode;
1050 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1053 goto free_node_inode;
1056 sb->s_root = d_make_root(root); /* allocate root dentry */
1059 goto free_root_inode;
1062 err = f2fs_build_stats(sbi);
1064 goto free_root_inode;
1067 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1070 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1071 &f2fs_seq_segment_info_fops, sb);
1073 if (test_opt(sbi, DISCARD)) {
1074 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1075 if (!blk_queue_discard(q))
1076 f2fs_msg(sb, KERN_WARNING,
1077 "mounting with \"discard\" option, but "
1078 "the device does not support discard");
1081 sbi->s_kobj.kset = f2fs_kset;
1082 init_completion(&sbi->s_kobj_unregister);
1083 err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1089 sbi->need_fsck = true;
1091 /* recover fsynced data */
1092 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1093 err = recover_fsync_data(sbi);
1095 f2fs_msg(sb, KERN_ERR,
1096 "Cannot recover all fsync data errno=%ld", err);
1102 * If filesystem is not mounted as read-only then
1103 * do start the gc_thread.
1105 if (!f2fs_readonly(sb)) {
1106 /* After POR, we can run background GC thread.*/
1107 err = start_gc_thread(sbi);
1114 kobject_del(&sbi->s_kobj);
1117 remove_proc_entry("segment_info", sbi->s_proc);
1118 remove_proc_entry(sb->s_id, f2fs_proc_root);
1120 f2fs_destroy_stats(sbi);
1125 iput(sbi->node_inode);
1127 destroy_node_manager(sbi);
1129 destroy_segment_manager(sbi);
1133 make_bad_inode(sbi->meta_inode);
1134 iput(sbi->meta_inode);
1136 brelse(raw_super_buf);
1140 /* give only one another chance */
1143 shrink_dcache_sb(sb);
1149 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1150 const char *dev_name, void *data)
1152 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1155 static struct file_system_type f2fs_fs_type = {
1156 .owner = THIS_MODULE,
1158 .mount = f2fs_mount,
1159 .kill_sb = kill_block_super,
1160 .fs_flags = FS_REQUIRES_DEV,
1162 MODULE_ALIAS_FS("f2fs");
1164 static int __init init_inodecache(void)
1166 f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
1167 sizeof(struct f2fs_inode_info));
1168 if (!f2fs_inode_cachep)
1173 static void destroy_inodecache(void)
1176 * Make sure all delayed rcu free inodes are flushed before we
1180 kmem_cache_destroy(f2fs_inode_cachep);
1183 static int __init init_f2fs_fs(void)
1187 err = init_inodecache();
1190 err = create_node_manager_caches();
1192 goto free_inodecache;
1193 err = create_segment_manager_caches();
1195 goto free_node_manager_caches;
1196 err = create_gc_caches();
1198 goto free_segment_manager_caches;
1199 err = create_checkpoint_caches();
1201 goto free_gc_caches;
1202 f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1205 goto free_checkpoint_caches;
1207 err = register_filesystem(&f2fs_fs_type);
1210 f2fs_create_root_stats();
1211 f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1215 kset_unregister(f2fs_kset);
1216 free_checkpoint_caches:
1217 destroy_checkpoint_caches();
1219 destroy_gc_caches();
1220 free_segment_manager_caches:
1221 destroy_segment_manager_caches();
1222 free_node_manager_caches:
1223 destroy_node_manager_caches();
1225 destroy_inodecache();
1230 static void __exit exit_f2fs_fs(void)
1232 remove_proc_entry("fs/f2fs", NULL);
1233 f2fs_destroy_root_stats();
1234 unregister_filesystem(&f2fs_fs_type);
1235 destroy_checkpoint_caches();
1236 destroy_gc_caches();
1237 destroy_segment_manager_caches();
1238 destroy_node_manager_caches();
1239 destroy_inodecache();
1240 kset_unregister(f2fs_kset);
1243 module_init(init_f2fs_fs)
1244 module_exit(exit_f2fs_fs)
1246 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1247 MODULE_DESCRIPTION("Flash Friendly File System");
1248 MODULE_LICENSE("GPL");