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,
58 static match_table_t f2fs_tokens = {
59 {Opt_gc_background, "background_gc=%s"},
60 {Opt_disable_roll_forward, "disable_roll_forward"},
61 {Opt_discard, "discard"},
62 {Opt_noheap, "no_heap"},
63 {Opt_user_xattr, "user_xattr"},
64 {Opt_nouser_xattr, "nouser_xattr"},
67 {Opt_active_logs, "active_logs=%u"},
68 {Opt_disable_ext_identify, "disable_ext_identify"},
69 {Opt_inline_xattr, "inline_xattr"},
70 {Opt_inline_data, "inline_data"},
71 {Opt_flush_merge, "flush_merge"},
75 /* Sysfs support for f2fs */
77 GC_THREAD, /* struct f2fs_gc_thread */
78 SM_INFO, /* struct f2fs_sm_info */
79 NM_INFO, /* struct f2fs_nm_info */
80 F2FS_SBI, /* struct f2fs_sb_info */
84 struct attribute attr;
85 ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
86 ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
87 const char *, size_t);
92 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
94 if (struct_type == GC_THREAD)
95 return (unsigned char *)sbi->gc_thread;
96 else if (struct_type == SM_INFO)
97 return (unsigned char *)SM_I(sbi);
98 else if (struct_type == NM_INFO)
99 return (unsigned char *)NM_I(sbi);
100 else if (struct_type == F2FS_SBI)
101 return (unsigned char *)sbi;
105 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
106 struct f2fs_sb_info *sbi, char *buf)
108 unsigned char *ptr = NULL;
111 ptr = __struct_ptr(sbi, a->struct_type);
115 ui = (unsigned int *)(ptr + a->offset);
117 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
120 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
121 struct f2fs_sb_info *sbi,
122 const char *buf, size_t count)
129 ptr = __struct_ptr(sbi, a->struct_type);
133 ui = (unsigned int *)(ptr + a->offset);
135 ret = kstrtoul(skip_spaces(buf), 0, &t);
142 static ssize_t f2fs_attr_show(struct kobject *kobj,
143 struct attribute *attr, char *buf)
145 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
147 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
149 return a->show ? a->show(a, sbi, buf) : 0;
152 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
153 const char *buf, size_t len)
155 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
157 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
159 return a->store ? a->store(a, sbi, buf, len) : 0;
162 static void f2fs_sb_release(struct kobject *kobj)
164 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
166 complete(&sbi->s_kobj_unregister);
169 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
170 static struct f2fs_attr f2fs_attr_##_name = { \
171 .attr = {.name = __stringify(_name), .mode = _mode }, \
174 .struct_type = _struct_type, \
178 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
179 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
180 f2fs_sbi_show, f2fs_sbi_store, \
181 offsetof(struct struct_name, elname))
183 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
184 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
185 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
186 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
187 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
188 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
189 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
190 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
191 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
192 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
193 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
195 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
196 static struct attribute *f2fs_attrs[] = {
197 ATTR_LIST(gc_min_sleep_time),
198 ATTR_LIST(gc_max_sleep_time),
199 ATTR_LIST(gc_no_gc_sleep_time),
201 ATTR_LIST(reclaim_segments),
202 ATTR_LIST(max_small_discards),
203 ATTR_LIST(ipu_policy),
204 ATTR_LIST(min_ipu_util),
205 ATTR_LIST(max_victim_search),
206 ATTR_LIST(dir_level),
207 ATTR_LIST(ram_thresh),
211 static const struct sysfs_ops f2fs_attr_ops = {
212 .show = f2fs_attr_show,
213 .store = f2fs_attr_store,
216 static struct kobj_type f2fs_ktype = {
217 .default_attrs = f2fs_attrs,
218 .sysfs_ops = &f2fs_attr_ops,
219 .release = f2fs_sb_release,
222 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
224 struct va_format vaf;
230 printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
234 static void init_once(void *foo)
236 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
238 inode_init_once(&fi->vfs_inode);
241 static int parse_options(struct super_block *sb, char *options)
243 struct f2fs_sb_info *sbi = F2FS_SB(sb);
244 substring_t args[MAX_OPT_ARGS];
251 while ((p = strsep(&options, ",")) != NULL) {
256 * Initialize args struct so we know whether arg was
257 * found; some options take optional arguments.
259 args[0].to = args[0].from = NULL;
260 token = match_token(p, f2fs_tokens, args);
263 case Opt_gc_background:
264 name = match_strdup(&args[0]);
268 if (strlen(name) == 2 && !strncmp(name, "on", 2))
270 else if (strlen(name) == 3 && !strncmp(name, "off", 3))
271 clear_opt(sbi, BG_GC);
278 case Opt_disable_roll_forward:
279 set_opt(sbi, DISABLE_ROLL_FORWARD);
282 set_opt(sbi, DISCARD);
285 set_opt(sbi, NOHEAP);
287 #ifdef CONFIG_F2FS_FS_XATTR
289 set_opt(sbi, XATTR_USER);
291 case Opt_nouser_xattr:
292 clear_opt(sbi, XATTR_USER);
294 case Opt_inline_xattr:
295 set_opt(sbi, INLINE_XATTR);
299 f2fs_msg(sb, KERN_INFO,
300 "user_xattr options not supported");
302 case Opt_nouser_xattr:
303 f2fs_msg(sb, KERN_INFO,
304 "nouser_xattr options not supported");
306 case Opt_inline_xattr:
307 f2fs_msg(sb, KERN_INFO,
308 "inline_xattr options not supported");
311 #ifdef CONFIG_F2FS_FS_POSIX_ACL
313 set_opt(sbi, POSIX_ACL);
316 clear_opt(sbi, POSIX_ACL);
320 f2fs_msg(sb, KERN_INFO, "acl options not supported");
323 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
326 case Opt_active_logs:
327 if (args->from && match_int(args, &arg))
329 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
331 sbi->active_logs = arg;
333 case Opt_disable_ext_identify:
334 set_opt(sbi, DISABLE_EXT_IDENTIFY);
336 case Opt_inline_data:
337 set_opt(sbi, INLINE_DATA);
339 case Opt_flush_merge:
340 set_opt(sbi, FLUSH_MERGE);
343 f2fs_msg(sb, KERN_ERR,
344 "Unrecognized mount option \"%s\" or missing value",
352 static struct inode *f2fs_alloc_inode(struct super_block *sb)
354 struct f2fs_inode_info *fi;
356 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
360 init_once((void *) fi);
362 /* Initialize f2fs-specific inode info */
363 fi->vfs_inode.i_version = 1;
364 atomic_set(&fi->dirty_dents, 0);
365 fi->i_current_depth = 1;
367 rwlock_init(&fi->ext.ext_lock);
368 init_rwsem(&fi->i_sem);
370 set_inode_flag(fi, FI_NEW_INODE);
372 if (test_opt(F2FS_SB(sb), INLINE_XATTR))
373 set_inode_flag(fi, FI_INLINE_XATTR);
375 /* Will be used by directory only */
376 fi->i_dir_level = F2FS_SB(sb)->dir_level;
378 return &fi->vfs_inode;
381 static int f2fs_drop_inode(struct inode *inode)
384 * This is to avoid a deadlock condition like below.
385 * writeback_single_inode(inode)
386 * - f2fs_write_data_page
387 * - f2fs_gc -> iput -> evict
388 * - inode_wait_for_writeback(inode)
390 if (!inode_unhashed(inode) && inode->i_state & I_SYNC)
392 return generic_drop_inode(inode);
396 * f2fs_dirty_inode() is called from __mark_inode_dirty()
398 * We should call set_dirty_inode to write the dirty inode through write_inode.
400 static void f2fs_dirty_inode(struct inode *inode, int flags)
402 set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
405 static void f2fs_i_callback(struct rcu_head *head)
407 struct inode *inode = container_of(head, struct inode, i_rcu);
408 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
411 static void f2fs_destroy_inode(struct inode *inode)
413 call_rcu(&inode->i_rcu, f2fs_i_callback);
416 static void f2fs_put_super(struct super_block *sb)
418 struct f2fs_sb_info *sbi = F2FS_SB(sb);
421 remove_proc_entry("segment_info", sbi->s_proc);
422 remove_proc_entry(sb->s_id, f2fs_proc_root);
424 kobject_del(&sbi->s_kobj);
426 f2fs_destroy_stats(sbi);
429 /* We don't need to do checkpoint when it's clean */
430 if (sbi->s_dirty && get_pages(sbi, F2FS_DIRTY_NODES))
431 write_checkpoint(sbi, true);
433 iput(sbi->node_inode);
434 iput(sbi->meta_inode);
436 /* destroy f2fs internal modules */
437 destroy_node_manager(sbi);
438 destroy_segment_manager(sbi);
441 kobject_put(&sbi->s_kobj);
442 wait_for_completion(&sbi->s_kobj_unregister);
444 sb->s_fs_info = NULL;
445 brelse(sbi->raw_super_buf);
449 int f2fs_sync_fs(struct super_block *sb, int sync)
451 struct f2fs_sb_info *sbi = F2FS_SB(sb);
453 trace_f2fs_sync_fs(sb, sync);
455 if (!sbi->s_dirty && !get_pages(sbi, F2FS_DIRTY_NODES))
459 mutex_lock(&sbi->gc_mutex);
460 write_checkpoint(sbi, false);
461 mutex_unlock(&sbi->gc_mutex);
463 f2fs_balance_fs(sbi);
469 static int f2fs_freeze(struct super_block *sb)
473 if (f2fs_readonly(sb))
476 err = f2fs_sync_fs(sb, 1);
480 static int f2fs_unfreeze(struct super_block *sb)
485 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
487 struct super_block *sb = dentry->d_sb;
488 struct f2fs_sb_info *sbi = F2FS_SB(sb);
489 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
490 block_t total_count, user_block_count, start_count, ovp_count;
492 total_count = le64_to_cpu(sbi->raw_super->block_count);
493 user_block_count = sbi->user_block_count;
494 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
495 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
496 buf->f_type = F2FS_SUPER_MAGIC;
497 buf->f_bsize = sbi->blocksize;
499 buf->f_blocks = total_count - start_count;
500 buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
501 buf->f_bavail = user_block_count - valid_user_blocks(sbi);
503 buf->f_files = sbi->total_node_count;
504 buf->f_ffree = sbi->total_node_count - valid_inode_count(sbi);
506 buf->f_namelen = F2FS_NAME_LEN;
507 buf->f_fsid.val[0] = (u32)id;
508 buf->f_fsid.val[1] = (u32)(id >> 32);
513 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
515 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
517 if (!(root->d_sb->s_flags & MS_RDONLY) && test_opt(sbi, BG_GC))
518 seq_printf(seq, ",background_gc=%s", "on");
520 seq_printf(seq, ",background_gc=%s", "off");
521 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
522 seq_puts(seq, ",disable_roll_forward");
523 if (test_opt(sbi, DISCARD))
524 seq_puts(seq, ",discard");
525 if (test_opt(sbi, NOHEAP))
526 seq_puts(seq, ",no_heap_alloc");
527 #ifdef CONFIG_F2FS_FS_XATTR
528 if (test_opt(sbi, XATTR_USER))
529 seq_puts(seq, ",user_xattr");
531 seq_puts(seq, ",nouser_xattr");
532 if (test_opt(sbi, INLINE_XATTR))
533 seq_puts(seq, ",inline_xattr");
535 #ifdef CONFIG_F2FS_FS_POSIX_ACL
536 if (test_opt(sbi, POSIX_ACL))
537 seq_puts(seq, ",acl");
539 seq_puts(seq, ",noacl");
541 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
542 seq_puts(seq, ",disable_ext_identify");
543 if (test_opt(sbi, INLINE_DATA))
544 seq_puts(seq, ",inline_data");
545 if (test_opt(sbi, FLUSH_MERGE))
546 seq_puts(seq, ",flush_merge");
547 seq_printf(seq, ",active_logs=%u", sbi->active_logs);
552 static int segment_info_seq_show(struct seq_file *seq, void *offset)
554 struct super_block *sb = seq->private;
555 struct f2fs_sb_info *sbi = F2FS_SB(sb);
556 unsigned int total_segs =
557 le32_to_cpu(sbi->raw_super->segment_count_main);
560 seq_puts(seq, "format: segment_type|valid_blocks\n"
561 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
563 for (i = 0; i < total_segs; i++) {
564 struct seg_entry *se = get_seg_entry(sbi, i);
567 seq_printf(seq, "%-5d", i);
568 seq_printf(seq, "%d|%-3u", se->type,
569 get_valid_blocks(sbi, i, 1));
570 if ((i % 10) == 9 || i == (total_segs - 1))
579 static int segment_info_open_fs(struct inode *inode, struct file *file)
581 return single_open(file, segment_info_seq_show, PDE_DATA(inode));
584 static const struct file_operations f2fs_seq_segment_info_fops = {
585 .owner = THIS_MODULE,
586 .open = segment_info_open_fs,
589 .release = single_release,
592 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
594 struct f2fs_sb_info *sbi = F2FS_SB(sb);
595 struct f2fs_mount_info org_mount_opt;
596 int err, active_logs;
599 * Save the old mount options in case we
600 * need to restore them.
602 org_mount_opt = sbi->mount_opt;
603 active_logs = sbi->active_logs;
605 /* parse mount options */
606 err = parse_options(sb, data);
611 * Previous and new state of filesystem is RO,
612 * so no point in checking GC conditions.
614 if ((sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY))
618 * We stop the GC thread if FS is mounted as RO
619 * or if background_gc = off is passed in mount
620 * option. Also sync the filesystem.
622 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
623 if (sbi->gc_thread) {
627 } else if (test_opt(sbi, BG_GC) && !sbi->gc_thread) {
628 err = start_gc_thread(sbi);
633 /* Update the POSIXACL Flag */
634 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
635 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
639 sbi->mount_opt = org_mount_opt;
640 sbi->active_logs = active_logs;
644 static struct super_operations f2fs_sops = {
645 .alloc_inode = f2fs_alloc_inode,
646 .drop_inode = f2fs_drop_inode,
647 .destroy_inode = f2fs_destroy_inode,
648 .write_inode = f2fs_write_inode,
649 .dirty_inode = f2fs_dirty_inode,
650 .show_options = f2fs_show_options,
651 .evict_inode = f2fs_evict_inode,
652 .put_super = f2fs_put_super,
653 .sync_fs = f2fs_sync_fs,
654 .freeze_fs = f2fs_freeze,
655 .unfreeze_fs = f2fs_unfreeze,
656 .statfs = f2fs_statfs,
657 .remount_fs = f2fs_remount,
660 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
661 u64 ino, u32 generation)
663 struct f2fs_sb_info *sbi = F2FS_SB(sb);
666 if (unlikely(ino < F2FS_ROOT_INO(sbi)))
667 return ERR_PTR(-ESTALE);
668 if (unlikely(ino >= NM_I(sbi)->max_nid))
669 return ERR_PTR(-ESTALE);
672 * f2fs_iget isn't quite right if the inode is currently unallocated!
673 * However f2fs_iget currently does appropriate checks to handle stale
674 * inodes so everything is OK.
676 inode = f2fs_iget(sb, ino);
678 return ERR_CAST(inode);
679 if (unlikely(generation && inode->i_generation != generation)) {
680 /* we didn't find the right inode.. */
682 return ERR_PTR(-ESTALE);
687 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
688 int fh_len, int fh_type)
690 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
694 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
695 int fh_len, int fh_type)
697 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
701 static const struct export_operations f2fs_export_ops = {
702 .fh_to_dentry = f2fs_fh_to_dentry,
703 .fh_to_parent = f2fs_fh_to_parent,
704 .get_parent = f2fs_get_parent,
707 static loff_t max_file_size(unsigned bits)
709 loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
710 loff_t leaf_count = ADDRS_PER_BLOCK;
712 /* two direct node blocks */
713 result += (leaf_count * 2);
715 /* two indirect node blocks */
716 leaf_count *= NIDS_PER_BLOCK;
717 result += (leaf_count * 2);
719 /* one double indirect node block */
720 leaf_count *= NIDS_PER_BLOCK;
721 result += leaf_count;
727 static int sanity_check_raw_super(struct super_block *sb,
728 struct f2fs_super_block *raw_super)
730 unsigned int blocksize;
732 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
733 f2fs_msg(sb, KERN_INFO,
734 "Magic Mismatch, valid(0x%x) - read(0x%x)",
735 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
739 /* Currently, support only 4KB page cache size */
740 if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
741 f2fs_msg(sb, KERN_INFO,
742 "Invalid page_cache_size (%lu), supports only 4KB\n",
747 /* Currently, support only 4KB block size */
748 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
749 if (blocksize != F2FS_BLKSIZE) {
750 f2fs_msg(sb, KERN_INFO,
751 "Invalid blocksize (%u), supports only 4KB\n",
756 if (le32_to_cpu(raw_super->log_sectorsize) !=
757 F2FS_LOG_SECTOR_SIZE) {
758 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize");
761 if (le32_to_cpu(raw_super->log_sectors_per_block) !=
762 F2FS_LOG_SECTORS_PER_BLOCK) {
763 f2fs_msg(sb, KERN_INFO, "Invalid log sectors per block");
769 static int sanity_check_ckpt(struct f2fs_sb_info *sbi)
771 unsigned int total, fsmeta;
772 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
773 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
775 total = le32_to_cpu(raw_super->segment_count);
776 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
777 fsmeta += le32_to_cpu(raw_super->segment_count_sit);
778 fsmeta += le32_to_cpu(raw_super->segment_count_nat);
779 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
780 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
782 if (unlikely(fsmeta >= total))
785 if (unlikely(is_set_ckpt_flags(ckpt, CP_ERROR_FLAG))) {
786 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
792 static void init_sb_info(struct f2fs_sb_info *sbi)
794 struct f2fs_super_block *raw_super = sbi->raw_super;
797 sbi->log_sectors_per_block =
798 le32_to_cpu(raw_super->log_sectors_per_block);
799 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
800 sbi->blocksize = 1 << sbi->log_blocksize;
801 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
802 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
803 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
804 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
805 sbi->total_sections = le32_to_cpu(raw_super->section_count);
806 sbi->total_node_count =
807 (le32_to_cpu(raw_super->segment_count_nat) / 2)
808 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
809 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
810 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
811 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
812 sbi->cur_victim_sec = NULL_SECNO;
813 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
815 for (i = 0; i < NR_COUNT_TYPE; i++)
816 atomic_set(&sbi->nr_pages[i], 0);
818 sbi->dir_level = DEF_DIR_LEVEL;
822 * Read f2fs raw super block.
823 * Because we have two copies of super block, so read the first one at first,
824 * if the first one is invalid, move to read the second one.
826 static int read_raw_super_block(struct super_block *sb,
827 struct f2fs_super_block **raw_super,
828 struct buffer_head **raw_super_buf)
833 *raw_super_buf = sb_bread(sb, block);
834 if (!*raw_super_buf) {
835 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
845 *raw_super = (struct f2fs_super_block *)
846 ((char *)(*raw_super_buf)->b_data + F2FS_SUPER_OFFSET);
848 /* sanity checking of raw super */
849 if (sanity_check_raw_super(sb, *raw_super)) {
850 brelse(*raw_super_buf);
851 f2fs_msg(sb, KERN_ERR,
852 "Can't find valid F2FS filesystem in %dth superblock",
865 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
867 struct f2fs_sb_info *sbi;
868 struct f2fs_super_block *raw_super;
869 struct buffer_head *raw_super_buf;
874 /* allocate memory for f2fs-specific super block info */
875 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
879 /* set a block size */
880 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
881 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
885 err = read_raw_super_block(sb, &raw_super, &raw_super_buf);
890 /* init some FS parameters */
891 sbi->active_logs = NR_CURSEG_TYPE;
895 #ifdef CONFIG_F2FS_FS_XATTR
896 set_opt(sbi, XATTR_USER);
898 #ifdef CONFIG_F2FS_FS_POSIX_ACL
899 set_opt(sbi, POSIX_ACL);
901 /* parse mount options */
902 err = parse_options(sb, (char *)data);
906 sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
907 sb->s_max_links = F2FS_LINK_MAX;
908 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
910 sb->s_op = &f2fs_sops;
911 sb->s_xattr = f2fs_xattr_handlers;
912 sb->s_export_op = &f2fs_export_ops;
913 sb->s_magic = F2FS_SUPER_MAGIC;
915 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
916 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
917 memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
919 /* init f2fs-specific super block info */
921 sbi->raw_super = raw_super;
922 sbi->raw_super_buf = raw_super_buf;
923 mutex_init(&sbi->gc_mutex);
924 mutex_init(&sbi->writepages);
925 mutex_init(&sbi->cp_mutex);
926 mutex_init(&sbi->node_write);
927 sbi->por_doing = false;
928 spin_lock_init(&sbi->stat_lock);
930 init_rwsem(&sbi->read_io.io_rwsem);
931 sbi->read_io.sbi = sbi;
932 sbi->read_io.bio = NULL;
933 for (i = 0; i < NR_PAGE_TYPE; i++) {
934 init_rwsem(&sbi->write_io[i].io_rwsem);
935 sbi->write_io[i].sbi = sbi;
936 sbi->write_io[i].bio = NULL;
939 init_rwsem(&sbi->cp_rwsem);
940 init_waitqueue_head(&sbi->cp_wait);
943 /* get an inode for meta space */
944 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
945 if (IS_ERR(sbi->meta_inode)) {
946 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
947 err = PTR_ERR(sbi->meta_inode);
951 err = get_valid_checkpoint(sbi);
953 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
954 goto free_meta_inode;
957 /* sanity checking of checkpoint */
959 if (sanity_check_ckpt(sbi)) {
960 f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint");
964 sbi->total_valid_node_count =
965 le32_to_cpu(sbi->ckpt->valid_node_count);
966 sbi->total_valid_inode_count =
967 le32_to_cpu(sbi->ckpt->valid_inode_count);
968 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
969 sbi->total_valid_block_count =
970 le64_to_cpu(sbi->ckpt->valid_block_count);
971 sbi->last_valid_block_count = sbi->total_valid_block_count;
972 sbi->alloc_valid_block_count = 0;
973 INIT_LIST_HEAD(&sbi->dir_inode_list);
974 spin_lock_init(&sbi->dir_inode_lock);
976 init_orphan_info(sbi);
978 /* setup f2fs internal modules */
979 err = build_segment_manager(sbi);
981 f2fs_msg(sb, KERN_ERR,
982 "Failed to initialize F2FS segment manager");
985 err = build_node_manager(sbi);
987 f2fs_msg(sb, KERN_ERR,
988 "Failed to initialize F2FS node manager");
992 build_gc_manager(sbi);
994 /* get an inode for node space */
995 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
996 if (IS_ERR(sbi->node_inode)) {
997 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
998 err = PTR_ERR(sbi->node_inode);
1002 /* if there are nt orphan nodes free them */
1003 recover_orphan_inodes(sbi);
1005 /* read root inode and dentry */
1006 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1008 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1009 err = PTR_ERR(root);
1010 goto free_node_inode;
1012 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1014 goto free_root_inode;
1017 sb->s_root = d_make_root(root); /* allocate root dentry */
1020 goto free_root_inode;
1023 err = f2fs_build_stats(sbi);
1025 goto free_root_inode;
1028 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1031 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1032 &f2fs_seq_segment_info_fops, sb);
1034 if (test_opt(sbi, DISCARD)) {
1035 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1036 if (!blk_queue_discard(q))
1037 f2fs_msg(sb, KERN_WARNING,
1038 "mounting with \"discard\" option, but "
1039 "the device does not support discard");
1042 sbi->s_kobj.kset = f2fs_kset;
1043 init_completion(&sbi->s_kobj_unregister);
1044 err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1049 /* recover fsynced data */
1050 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1051 err = recover_fsync_data(sbi);
1053 f2fs_msg(sb, KERN_ERR,
1054 "Cannot recover all fsync data errno=%ld", err);
1058 * If filesystem is not mounted as read-only then
1059 * do start the gc_thread.
1061 if (!(sb->s_flags & MS_RDONLY)) {
1062 /* After POR, we can run background GC thread.*/
1063 err = start_gc_thread(sbi);
1070 kobject_del(&sbi->s_kobj);
1073 remove_proc_entry("segment_info", sbi->s_proc);
1074 remove_proc_entry(sb->s_id, f2fs_proc_root);
1076 f2fs_destroy_stats(sbi);
1081 iput(sbi->node_inode);
1083 destroy_node_manager(sbi);
1085 destroy_segment_manager(sbi);
1089 make_bad_inode(sbi->meta_inode);
1090 iput(sbi->meta_inode);
1092 brelse(raw_super_buf);
1098 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1099 const char *dev_name, void *data)
1101 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1104 static struct file_system_type f2fs_fs_type = {
1105 .owner = THIS_MODULE,
1107 .mount = f2fs_mount,
1108 .kill_sb = kill_block_super,
1109 .fs_flags = FS_REQUIRES_DEV,
1111 MODULE_ALIAS_FS("f2fs");
1113 static int __init init_inodecache(void)
1115 f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
1116 sizeof(struct f2fs_inode_info));
1117 if (!f2fs_inode_cachep)
1122 static void destroy_inodecache(void)
1125 * Make sure all delayed rcu free inodes are flushed before we
1129 kmem_cache_destroy(f2fs_inode_cachep);
1132 static int __init init_f2fs_fs(void)
1136 err = init_inodecache();
1139 err = create_node_manager_caches();
1141 goto free_inodecache;
1142 err = create_segment_manager_caches();
1144 goto free_node_manager_caches;
1145 err = create_gc_caches();
1147 goto free_segment_manager_caches;
1148 err = create_checkpoint_caches();
1150 goto free_gc_caches;
1151 f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1154 goto free_checkpoint_caches;
1156 err = register_filesystem(&f2fs_fs_type);
1159 f2fs_create_root_stats();
1160 f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1164 kset_unregister(f2fs_kset);
1165 free_checkpoint_caches:
1166 destroy_checkpoint_caches();
1168 destroy_gc_caches();
1169 free_segment_manager_caches:
1170 destroy_segment_manager_caches();
1171 free_node_manager_caches:
1172 destroy_node_manager_caches();
1174 destroy_inodecache();
1179 static void __exit exit_f2fs_fs(void)
1181 remove_proc_entry("fs/f2fs", NULL);
1182 f2fs_destroy_root_stats();
1183 unregister_filesystem(&f2fs_fs_type);
1184 destroy_checkpoint_caches();
1185 destroy_gc_caches();
1186 destroy_segment_manager_caches();
1187 destroy_node_manager_caches();
1188 destroy_inodecache();
1189 kset_unregister(f2fs_kset);
1192 module_init(init_f2fs_fs)
1193 module_exit(exit_f2fs_fs)
1195 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1196 MODULE_DESCRIPTION("Flash Friendly File System");
1197 MODULE_LICENSE("GPL");