4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
11 #include <linux/module.h>
12 #include <linux/init.h>
14 #include <linux/statfs.h>
15 #include <linux/buffer_head.h>
16 #include <linux/backing-dev.h>
17 #include <linux/kthread.h>
18 #include <linux/parser.h>
19 #include <linux/mount.h>
20 #include <linux/seq_file.h>
21 #include <linux/proc_fs.h>
22 #include <linux/random.h>
23 #include <linux/exportfs.h>
24 #include <linux/blkdev.h>
25 #include <linux/f2fs_fs.h>
26 #include <linux/sysfs.h>
35 #define CREATE_TRACE_POINTS
36 #include <trace/events/f2fs.h>
38 static struct proc_dir_entry *f2fs_proc_root;
39 static struct kmem_cache *f2fs_inode_cachep;
40 static struct kset *f2fs_kset;
42 #ifdef CONFIG_F2FS_FAULT_INJECTION
43 struct f2fs_fault_info f2fs_fault;
45 char *fault_name[FAULT_MAX] = {
46 [FAULT_KMALLOC] = "kmalloc",
47 [FAULT_PAGE_ALLOC] = "page alloc",
48 [FAULT_ALLOC_NID] = "alloc nid",
49 [FAULT_ORPHAN] = "orphan",
50 [FAULT_BLOCK] = "no more block",
51 [FAULT_DIR_DEPTH] = "too big dir depth",
54 static void f2fs_build_fault_attr(unsigned int rate)
57 atomic_set(&f2fs_fault.inject_ops, 0);
58 f2fs_fault.inject_rate = rate;
59 f2fs_fault.inject_type = (1 << FAULT_MAX) - 1;
61 memset(&f2fs_fault, 0, sizeof(struct f2fs_fault_info));
66 /* f2fs-wide shrinker description */
67 static struct shrinker f2fs_shrinker_info = {
68 .scan_objects = f2fs_shrink_scan,
69 .count_objects = f2fs_shrink_count,
70 .seeks = DEFAULT_SEEKS,
75 Opt_disable_roll_forward,
84 Opt_disable_ext_identify,
102 static match_table_t f2fs_tokens = {
103 {Opt_gc_background, "background_gc=%s"},
104 {Opt_disable_roll_forward, "disable_roll_forward"},
105 {Opt_norecovery, "norecovery"},
106 {Opt_discard, "discard"},
107 {Opt_noheap, "no_heap"},
108 {Opt_user_xattr, "user_xattr"},
109 {Opt_nouser_xattr, "nouser_xattr"},
111 {Opt_noacl, "noacl"},
112 {Opt_active_logs, "active_logs=%u"},
113 {Opt_disable_ext_identify, "disable_ext_identify"},
114 {Opt_inline_xattr, "inline_xattr"},
115 {Opt_inline_data, "inline_data"},
116 {Opt_inline_dentry, "inline_dentry"},
117 {Opt_flush_merge, "flush_merge"},
118 {Opt_noflush_merge, "noflush_merge"},
119 {Opt_nobarrier, "nobarrier"},
120 {Opt_fastboot, "fastboot"},
121 {Opt_extent_cache, "extent_cache"},
122 {Opt_noextent_cache, "noextent_cache"},
123 {Opt_noinline_data, "noinline_data"},
124 {Opt_data_flush, "data_flush"},
125 {Opt_fault_injection, "fault_injection=%u"},
126 {Opt_lazytime, "lazytime"},
127 {Opt_nolazytime, "nolazytime"},
131 /* Sysfs support for f2fs */
133 GC_THREAD, /* struct f2fs_gc_thread */
134 SM_INFO, /* struct f2fs_sm_info */
135 NM_INFO, /* struct f2fs_nm_info */
136 F2FS_SBI, /* struct f2fs_sb_info */
137 #ifdef CONFIG_F2FS_FAULT_INJECTION
138 FAULT_INFO_RATE, /* struct f2fs_fault_info */
139 FAULT_INFO_TYPE, /* struct f2fs_fault_info */
144 struct attribute attr;
145 ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
146 ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
147 const char *, size_t);
152 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
154 if (struct_type == GC_THREAD)
155 return (unsigned char *)sbi->gc_thread;
156 else if (struct_type == SM_INFO)
157 return (unsigned char *)SM_I(sbi);
158 else if (struct_type == NM_INFO)
159 return (unsigned char *)NM_I(sbi);
160 else if (struct_type == F2FS_SBI)
161 return (unsigned char *)sbi;
162 #ifdef CONFIG_F2FS_FAULT_INJECTION
163 else if (struct_type == FAULT_INFO_RATE ||
164 struct_type == FAULT_INFO_TYPE)
165 return (unsigned char *)&f2fs_fault;
170 static ssize_t lifetime_write_kbytes_show(struct f2fs_attr *a,
171 struct f2fs_sb_info *sbi, char *buf)
173 struct super_block *sb = sbi->sb;
175 if (!sb->s_bdev->bd_part)
176 return snprintf(buf, PAGE_SIZE, "0\n");
178 return snprintf(buf, PAGE_SIZE, "%llu\n",
179 (unsigned long long)(sbi->kbytes_written +
180 BD_PART_WRITTEN(sbi)));
183 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
184 struct f2fs_sb_info *sbi, char *buf)
186 unsigned char *ptr = NULL;
189 ptr = __struct_ptr(sbi, a->struct_type);
193 ui = (unsigned int *)(ptr + a->offset);
195 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
198 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
199 struct f2fs_sb_info *sbi,
200 const char *buf, size_t count)
207 ptr = __struct_ptr(sbi, a->struct_type);
211 ui = (unsigned int *)(ptr + a->offset);
213 ret = kstrtoul(skip_spaces(buf), 0, &t);
216 #ifdef CONFIG_F2FS_FAULT_INJECTION
217 if (a->struct_type == FAULT_INFO_TYPE && t >= (1 << FAULT_MAX))
224 static ssize_t f2fs_attr_show(struct kobject *kobj,
225 struct attribute *attr, char *buf)
227 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
229 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
231 return a->show ? a->show(a, sbi, buf) : 0;
234 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
235 const char *buf, size_t len)
237 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
239 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
241 return a->store ? a->store(a, sbi, buf, len) : 0;
244 static void f2fs_sb_release(struct kobject *kobj)
246 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
248 complete(&sbi->s_kobj_unregister);
251 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
252 static struct f2fs_attr f2fs_attr_##_name = { \
253 .attr = {.name = __stringify(_name), .mode = _mode }, \
256 .struct_type = _struct_type, \
260 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
261 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
262 f2fs_sbi_show, f2fs_sbi_store, \
263 offsetof(struct struct_name, elname))
265 #define F2FS_GENERAL_RO_ATTR(name) \
266 static struct f2fs_attr f2fs_attr_##name = __ATTR(name, 0444, name##_show, NULL)
268 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
269 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
270 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
271 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
272 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
273 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
274 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, batched_trim_sections, trim_sections);
275 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
276 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
277 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks);
278 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
279 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ra_nid_pages, ra_nid_pages);
280 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, dirty_nats_ratio, dirty_nats_ratio);
281 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
282 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
283 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, cp_interval, interval_time[CP_TIME]);
284 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, idle_interval, interval_time[REQ_TIME]);
285 #ifdef CONFIG_F2FS_FAULT_INJECTION
286 F2FS_RW_ATTR(FAULT_INFO_RATE, f2fs_fault_info, inject_rate, inject_rate);
287 F2FS_RW_ATTR(FAULT_INFO_TYPE, f2fs_fault_info, inject_type, inject_type);
289 F2FS_GENERAL_RO_ATTR(lifetime_write_kbytes);
291 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
292 static struct attribute *f2fs_attrs[] = {
293 ATTR_LIST(gc_min_sleep_time),
294 ATTR_LIST(gc_max_sleep_time),
295 ATTR_LIST(gc_no_gc_sleep_time),
297 ATTR_LIST(reclaim_segments),
298 ATTR_LIST(max_small_discards),
299 ATTR_LIST(batched_trim_sections),
300 ATTR_LIST(ipu_policy),
301 ATTR_LIST(min_ipu_util),
302 ATTR_LIST(min_fsync_blocks),
303 ATTR_LIST(max_victim_search),
304 ATTR_LIST(dir_level),
305 ATTR_LIST(ram_thresh),
306 ATTR_LIST(ra_nid_pages),
307 ATTR_LIST(dirty_nats_ratio),
308 ATTR_LIST(cp_interval),
309 ATTR_LIST(idle_interval),
310 ATTR_LIST(lifetime_write_kbytes),
314 static const struct sysfs_ops f2fs_attr_ops = {
315 .show = f2fs_attr_show,
316 .store = f2fs_attr_store,
319 static struct kobj_type f2fs_ktype = {
320 .default_attrs = f2fs_attrs,
321 .sysfs_ops = &f2fs_attr_ops,
322 .release = f2fs_sb_release,
325 #ifdef CONFIG_F2FS_FAULT_INJECTION
326 /* sysfs for f2fs fault injection */
327 static struct kobject f2fs_fault_inject;
329 static struct attribute *f2fs_fault_attrs[] = {
330 ATTR_LIST(inject_rate),
331 ATTR_LIST(inject_type),
335 static struct kobj_type f2fs_fault_ktype = {
336 .default_attrs = f2fs_fault_attrs,
337 .sysfs_ops = &f2fs_attr_ops,
341 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
343 struct va_format vaf;
349 printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
353 static void init_once(void *foo)
355 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
357 inode_init_once(&fi->vfs_inode);
360 static int parse_options(struct super_block *sb, char *options)
362 struct f2fs_sb_info *sbi = F2FS_SB(sb);
363 struct request_queue *q;
364 substring_t args[MAX_OPT_ARGS];
368 #ifdef CONFIG_F2FS_FAULT_INJECTION
369 f2fs_build_fault_attr(0);
375 while ((p = strsep(&options, ",")) != NULL) {
380 * Initialize args struct so we know whether arg was
381 * found; some options take optional arguments.
383 args[0].to = args[0].from = NULL;
384 token = match_token(p, f2fs_tokens, args);
387 case Opt_gc_background:
388 name = match_strdup(&args[0]);
392 if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
394 clear_opt(sbi, FORCE_FG_GC);
395 } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
396 clear_opt(sbi, BG_GC);
397 clear_opt(sbi, FORCE_FG_GC);
398 } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
400 set_opt(sbi, FORCE_FG_GC);
407 case Opt_disable_roll_forward:
408 set_opt(sbi, DISABLE_ROLL_FORWARD);
411 /* this option mounts f2fs with ro */
412 set_opt(sbi, DISABLE_ROLL_FORWARD);
413 if (!f2fs_readonly(sb))
417 q = bdev_get_queue(sb->s_bdev);
418 if (blk_queue_discard(q)) {
419 set_opt(sbi, DISCARD);
421 f2fs_msg(sb, KERN_WARNING,
422 "mounting with \"discard\" option, but "
423 "the device does not support discard");
427 set_opt(sbi, NOHEAP);
429 #ifdef CONFIG_F2FS_FS_XATTR
431 set_opt(sbi, XATTR_USER);
433 case Opt_nouser_xattr:
434 clear_opt(sbi, XATTR_USER);
436 case Opt_inline_xattr:
437 set_opt(sbi, INLINE_XATTR);
441 f2fs_msg(sb, KERN_INFO,
442 "user_xattr options not supported");
444 case Opt_nouser_xattr:
445 f2fs_msg(sb, KERN_INFO,
446 "nouser_xattr options not supported");
448 case Opt_inline_xattr:
449 f2fs_msg(sb, KERN_INFO,
450 "inline_xattr options not supported");
453 #ifdef CONFIG_F2FS_FS_POSIX_ACL
455 set_opt(sbi, POSIX_ACL);
458 clear_opt(sbi, POSIX_ACL);
462 f2fs_msg(sb, KERN_INFO, "acl options not supported");
465 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
468 case Opt_active_logs:
469 if (args->from && match_int(args, &arg))
471 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
473 sbi->active_logs = arg;
475 case Opt_disable_ext_identify:
476 set_opt(sbi, DISABLE_EXT_IDENTIFY);
478 case Opt_inline_data:
479 set_opt(sbi, INLINE_DATA);
481 case Opt_inline_dentry:
482 set_opt(sbi, INLINE_DENTRY);
484 case Opt_flush_merge:
485 set_opt(sbi, FLUSH_MERGE);
487 case Opt_noflush_merge:
488 clear_opt(sbi, FLUSH_MERGE);
491 set_opt(sbi, NOBARRIER);
494 set_opt(sbi, FASTBOOT);
496 case Opt_extent_cache:
497 set_opt(sbi, EXTENT_CACHE);
499 case Opt_noextent_cache:
500 clear_opt(sbi, EXTENT_CACHE);
502 case Opt_noinline_data:
503 clear_opt(sbi, INLINE_DATA);
506 set_opt(sbi, DATA_FLUSH);
508 case Opt_fault_injection:
509 if (args->from && match_int(args, &arg))
511 #ifdef CONFIG_F2FS_FAULT_INJECTION
512 f2fs_build_fault_attr(arg);
514 f2fs_msg(sb, KERN_INFO,
515 "FAULT_INJECTION was not selected");
519 sb->s_flags |= MS_LAZYTIME;
522 sb->s_flags &= ~MS_LAZYTIME;
525 f2fs_msg(sb, KERN_ERR,
526 "Unrecognized mount option \"%s\" or missing value",
534 static struct inode *f2fs_alloc_inode(struct super_block *sb)
536 struct f2fs_inode_info *fi;
538 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
542 init_once((void *) fi);
544 if (percpu_counter_init(&fi->dirty_pages, 0, GFP_NOFS)) {
545 kmem_cache_free(f2fs_inode_cachep, fi);
549 /* Initialize f2fs-specific inode info */
550 fi->vfs_inode.i_version = 1;
551 fi->i_current_depth = 1;
553 init_rwsem(&fi->i_sem);
554 INIT_LIST_HEAD(&fi->dirty_list);
555 INIT_LIST_HEAD(&fi->gdirty_list);
556 INIT_LIST_HEAD(&fi->inmem_pages);
557 mutex_init(&fi->inmem_lock);
559 /* Will be used by directory only */
560 fi->i_dir_level = F2FS_SB(sb)->dir_level;
561 return &fi->vfs_inode;
564 static int f2fs_drop_inode(struct inode *inode)
569 * This is to avoid a deadlock condition like below.
570 * writeback_single_inode(inode)
571 * - f2fs_write_data_page
572 * - f2fs_gc -> iput -> evict
573 * - inode_wait_for_writeback(inode)
575 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
576 if (!inode->i_nlink && !is_bad_inode(inode)) {
577 /* to avoid evict_inode call simultaneously */
578 atomic_inc(&inode->i_count);
579 spin_unlock(&inode->i_lock);
581 /* some remained atomic pages should discarded */
582 if (f2fs_is_atomic_file(inode))
583 drop_inmem_pages(inode);
585 /* should remain fi->extent_tree for writepage */
586 f2fs_destroy_extent_node(inode);
588 sb_start_intwrite(inode->i_sb);
589 f2fs_i_size_write(inode, 0);
591 if (F2FS_HAS_BLOCKS(inode))
592 f2fs_truncate(inode, true);
594 sb_end_intwrite(inode->i_sb);
596 fscrypt_put_encryption_info(inode, NULL);
597 spin_lock(&inode->i_lock);
598 atomic_dec(&inode->i_count);
603 ret = generic_drop_inode(inode);
604 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
606 inode->i_state |= I_WILL_FREE;
607 spin_unlock(&inode->i_lock);
609 update_inode_page(inode);
611 spin_lock(&inode->i_lock);
613 inode->i_state &= ~I_WILL_FREE;
619 * f2fs_dirty_inode() is called from __mark_inode_dirty()
621 * We should call set_dirty_inode to write the dirty inode through write_inode.
623 static void f2fs_dirty_inode(struct inode *inode, int flags)
625 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
627 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
628 inode->i_ino == F2FS_META_INO(sbi))
631 if (flags == I_DIRTY_TIME)
634 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
635 clear_inode_flag(inode, FI_AUTO_RECOVER);
637 spin_lock(&sbi->inode_lock[DIRTY_META]);
638 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
639 spin_unlock(&sbi->inode_lock[DIRTY_META]);
643 set_inode_flag(inode, FI_DIRTY_INODE);
644 list_add_tail(&F2FS_I(inode)->gdirty_list,
645 &sbi->inode_list[DIRTY_META]);
646 inc_page_count(sbi, F2FS_DIRTY_IMETA);
647 spin_unlock(&sbi->inode_lock[DIRTY_META]);
648 stat_inc_dirty_inode(sbi, DIRTY_META);
651 void f2fs_inode_synced(struct inode *inode)
653 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
655 spin_lock(&sbi->inode_lock[DIRTY_META]);
656 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
657 spin_unlock(&sbi->inode_lock[DIRTY_META]);
660 list_del_init(&F2FS_I(inode)->gdirty_list);
661 clear_inode_flag(inode, FI_DIRTY_INODE);
662 clear_inode_flag(inode, FI_AUTO_RECOVER);
663 dec_page_count(sbi, F2FS_DIRTY_IMETA);
664 spin_unlock(&sbi->inode_lock[DIRTY_META]);
665 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
668 static void f2fs_i_callback(struct rcu_head *head)
670 struct inode *inode = container_of(head, struct inode, i_rcu);
671 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
674 static void f2fs_destroy_inode(struct inode *inode)
676 percpu_counter_destroy(&F2FS_I(inode)->dirty_pages);
677 call_rcu(&inode->i_rcu, f2fs_i_callback);
680 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
684 for (i = 0; i < NR_COUNT_TYPE; i++)
685 percpu_counter_destroy(&sbi->nr_pages[i]);
686 percpu_counter_destroy(&sbi->alloc_valid_block_count);
687 percpu_counter_destroy(&sbi->total_valid_inode_count);
690 static void f2fs_put_super(struct super_block *sb)
692 struct f2fs_sb_info *sbi = F2FS_SB(sb);
695 remove_proc_entry("segment_info", sbi->s_proc);
696 remove_proc_entry("segment_bits", sbi->s_proc);
697 remove_proc_entry(sb->s_id, f2fs_proc_root);
699 kobject_del(&sbi->s_kobj);
703 /* prevent remaining shrinker jobs */
704 mutex_lock(&sbi->umount_mutex);
707 * We don't need to do checkpoint when superblock is clean.
708 * But, the previous checkpoint was not done by umount, it needs to do
709 * clean checkpoint again.
711 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
712 !is_set_ckpt_flags(F2FS_CKPT(sbi), CP_UMOUNT_FLAG)) {
713 struct cp_control cpc = {
716 write_checkpoint(sbi, &cpc);
719 /* write_checkpoint can update stat informaion */
720 f2fs_destroy_stats(sbi);
723 * normally superblock is clean, so we need to release this.
724 * In addition, EIO will skip do checkpoint, we need this as well.
726 release_ino_entry(sbi, true);
727 release_discard_addrs(sbi);
729 f2fs_leave_shrinker(sbi);
730 mutex_unlock(&sbi->umount_mutex);
732 /* our cp_error case, we can wait for any writeback page */
733 f2fs_flush_merged_bios(sbi);
735 iput(sbi->node_inode);
736 iput(sbi->meta_inode);
738 /* destroy f2fs internal modules */
739 destroy_node_manager(sbi);
740 destroy_segment_manager(sbi);
743 kobject_put(&sbi->s_kobj);
744 wait_for_completion(&sbi->s_kobj_unregister);
746 sb->s_fs_info = NULL;
747 if (sbi->s_chksum_driver)
748 crypto_free_shash(sbi->s_chksum_driver);
749 kfree(sbi->raw_super);
751 destroy_percpu_info(sbi);
755 int f2fs_sync_fs(struct super_block *sb, int sync)
757 struct f2fs_sb_info *sbi = F2FS_SB(sb);
760 trace_f2fs_sync_fs(sb, sync);
763 struct cp_control cpc;
765 cpc.reason = __get_cp_reason(sbi);
767 mutex_lock(&sbi->gc_mutex);
768 err = write_checkpoint(sbi, &cpc);
769 mutex_unlock(&sbi->gc_mutex);
771 f2fs_trace_ios(NULL, 1);
776 static int f2fs_freeze(struct super_block *sb)
780 if (f2fs_readonly(sb))
783 err = f2fs_sync_fs(sb, 1);
787 static int f2fs_unfreeze(struct super_block *sb)
792 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
794 struct super_block *sb = dentry->d_sb;
795 struct f2fs_sb_info *sbi = F2FS_SB(sb);
796 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
797 block_t total_count, user_block_count, start_count, ovp_count;
799 total_count = le64_to_cpu(sbi->raw_super->block_count);
800 user_block_count = sbi->user_block_count;
801 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
802 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
803 buf->f_type = F2FS_SUPER_MAGIC;
804 buf->f_bsize = sbi->blocksize;
806 buf->f_blocks = total_count - start_count;
807 buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
808 buf->f_bavail = user_block_count - valid_user_blocks(sbi);
810 buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
811 buf->f_ffree = buf->f_files - valid_inode_count(sbi);
813 buf->f_namelen = F2FS_NAME_LEN;
814 buf->f_fsid.val[0] = (u32)id;
815 buf->f_fsid.val[1] = (u32)(id >> 32);
820 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
822 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
824 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
825 if (test_opt(sbi, FORCE_FG_GC))
826 seq_printf(seq, ",background_gc=%s", "sync");
828 seq_printf(seq, ",background_gc=%s", "on");
830 seq_printf(seq, ",background_gc=%s", "off");
832 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
833 seq_puts(seq, ",disable_roll_forward");
834 if (test_opt(sbi, DISCARD))
835 seq_puts(seq, ",discard");
836 if (test_opt(sbi, NOHEAP))
837 seq_puts(seq, ",no_heap_alloc");
838 #ifdef CONFIG_F2FS_FS_XATTR
839 if (test_opt(sbi, XATTR_USER))
840 seq_puts(seq, ",user_xattr");
842 seq_puts(seq, ",nouser_xattr");
843 if (test_opt(sbi, INLINE_XATTR))
844 seq_puts(seq, ",inline_xattr");
846 #ifdef CONFIG_F2FS_FS_POSIX_ACL
847 if (test_opt(sbi, POSIX_ACL))
848 seq_puts(seq, ",acl");
850 seq_puts(seq, ",noacl");
852 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
853 seq_puts(seq, ",disable_ext_identify");
854 if (test_opt(sbi, INLINE_DATA))
855 seq_puts(seq, ",inline_data");
857 seq_puts(seq, ",noinline_data");
858 if (test_opt(sbi, INLINE_DENTRY))
859 seq_puts(seq, ",inline_dentry");
860 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
861 seq_puts(seq, ",flush_merge");
862 if (test_opt(sbi, NOBARRIER))
863 seq_puts(seq, ",nobarrier");
864 if (test_opt(sbi, FASTBOOT))
865 seq_puts(seq, ",fastboot");
866 if (test_opt(sbi, EXTENT_CACHE))
867 seq_puts(seq, ",extent_cache");
869 seq_puts(seq, ",noextent_cache");
870 if (test_opt(sbi, DATA_FLUSH))
871 seq_puts(seq, ",data_flush");
872 seq_printf(seq, ",active_logs=%u", sbi->active_logs);
877 static int segment_info_seq_show(struct seq_file *seq, void *offset)
879 struct super_block *sb = seq->private;
880 struct f2fs_sb_info *sbi = F2FS_SB(sb);
881 unsigned int total_segs =
882 le32_to_cpu(sbi->raw_super->segment_count_main);
885 seq_puts(seq, "format: segment_type|valid_blocks\n"
886 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
888 for (i = 0; i < total_segs; i++) {
889 struct seg_entry *se = get_seg_entry(sbi, i);
892 seq_printf(seq, "%-10d", i);
893 seq_printf(seq, "%d|%-3u", se->type,
894 get_valid_blocks(sbi, i, 1));
895 if ((i % 10) == 9 || i == (total_segs - 1))
904 static int segment_bits_seq_show(struct seq_file *seq, void *offset)
906 struct super_block *sb = seq->private;
907 struct f2fs_sb_info *sbi = F2FS_SB(sb);
908 unsigned int total_segs =
909 le32_to_cpu(sbi->raw_super->segment_count_main);
912 seq_puts(seq, "format: segment_type|valid_blocks|bitmaps\n"
913 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
915 for (i = 0; i < total_segs; i++) {
916 struct seg_entry *se = get_seg_entry(sbi, i);
918 seq_printf(seq, "%-10d", i);
919 seq_printf(seq, "%d|%-3u|", se->type,
920 get_valid_blocks(sbi, i, 1));
921 for (j = 0; j < SIT_VBLOCK_MAP_SIZE; j++)
922 seq_printf(seq, "%x ", se->cur_valid_map[j]);
928 #define F2FS_PROC_FILE_DEF(_name) \
929 static int _name##_open_fs(struct inode *inode, struct file *file) \
931 return single_open(file, _name##_seq_show, PDE_DATA(inode)); \
934 static const struct file_operations f2fs_seq_##_name##_fops = { \
935 .owner = THIS_MODULE, \
936 .open = _name##_open_fs, \
938 .llseek = seq_lseek, \
939 .release = single_release, \
942 F2FS_PROC_FILE_DEF(segment_info);
943 F2FS_PROC_FILE_DEF(segment_bits);
945 static void default_options(struct f2fs_sb_info *sbi)
947 /* init some FS parameters */
948 sbi->active_logs = NR_CURSEG_TYPE;
951 set_opt(sbi, INLINE_DATA);
952 set_opt(sbi, EXTENT_CACHE);
953 sbi->sb->s_flags |= MS_LAZYTIME;
954 set_opt(sbi, FLUSH_MERGE);
956 #ifdef CONFIG_F2FS_FS_XATTR
957 set_opt(sbi, XATTR_USER);
959 #ifdef CONFIG_F2FS_FS_POSIX_ACL
960 set_opt(sbi, POSIX_ACL);
964 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
966 struct f2fs_sb_info *sbi = F2FS_SB(sb);
967 struct f2fs_mount_info org_mount_opt;
968 int err, active_logs;
969 bool need_restart_gc = false;
970 bool need_stop_gc = false;
971 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
974 * Save the old mount options in case we
975 * need to restore them.
977 org_mount_opt = sbi->mount_opt;
978 active_logs = sbi->active_logs;
980 /* recover superblocks we couldn't write due to previous RO mount */
981 if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
982 err = f2fs_commit_super(sbi, false);
983 f2fs_msg(sb, KERN_INFO,
984 "Try to recover all the superblocks, ret: %d", err);
986 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
989 sbi->mount_opt.opt = 0;
990 default_options(sbi);
992 /* parse mount options */
993 err = parse_options(sb, data);
998 * Previous and new state of filesystem is RO,
999 * so skip checking GC and FLUSH_MERGE conditions.
1001 if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
1004 /* disallow enable/disable extent_cache dynamically */
1005 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1007 f2fs_msg(sbi->sb, KERN_WARNING,
1008 "switch extent_cache option is not allowed");
1013 * We stop the GC thread if FS is mounted as RO
1014 * or if background_gc = off is passed in mount
1015 * option. Also sync the filesystem.
1017 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
1018 if (sbi->gc_thread) {
1019 stop_gc_thread(sbi);
1020 need_restart_gc = true;
1022 } else if (!sbi->gc_thread) {
1023 err = start_gc_thread(sbi);
1026 need_stop_gc = true;
1029 if (*flags & MS_RDONLY) {
1030 writeback_inodes_sb(sb, WB_REASON_SYNC);
1033 set_sbi_flag(sbi, SBI_IS_DIRTY);
1034 set_sbi_flag(sbi, SBI_IS_CLOSE);
1035 f2fs_sync_fs(sb, 1);
1036 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1040 * We stop issue flush thread if FS is mounted as RO
1041 * or if flush_merge is not passed in mount option.
1043 if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1044 destroy_flush_cmd_control(sbi);
1045 } else if (!SM_I(sbi)->cmd_control_info) {
1046 err = create_flush_cmd_control(sbi);
1051 /* Update the POSIXACL Flag */
1052 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1053 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1057 if (need_restart_gc) {
1058 if (start_gc_thread(sbi))
1059 f2fs_msg(sbi->sb, KERN_WARNING,
1060 "background gc thread has stopped");
1061 } else if (need_stop_gc) {
1062 stop_gc_thread(sbi);
1065 sbi->mount_opt = org_mount_opt;
1066 sbi->active_logs = active_logs;
1070 static struct super_operations f2fs_sops = {
1071 .alloc_inode = f2fs_alloc_inode,
1072 .drop_inode = f2fs_drop_inode,
1073 .destroy_inode = f2fs_destroy_inode,
1074 .write_inode = f2fs_write_inode,
1075 .dirty_inode = f2fs_dirty_inode,
1076 .show_options = f2fs_show_options,
1077 .evict_inode = f2fs_evict_inode,
1078 .put_super = f2fs_put_super,
1079 .sync_fs = f2fs_sync_fs,
1080 .freeze_fs = f2fs_freeze,
1081 .unfreeze_fs = f2fs_unfreeze,
1082 .statfs = f2fs_statfs,
1083 .remount_fs = f2fs_remount,
1086 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1087 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1089 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1090 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1094 static int f2fs_key_prefix(struct inode *inode, u8 **key)
1096 *key = F2FS_I_SB(inode)->key_prefix;
1097 return F2FS_I_SB(inode)->key_prefix_size;
1100 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1103 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1104 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1105 ctx, len, fs_data, XATTR_CREATE);
1108 static unsigned f2fs_max_namelen(struct inode *inode)
1110 return S_ISLNK(inode->i_mode) ?
1111 inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1114 static struct fscrypt_operations f2fs_cryptops = {
1115 .get_context = f2fs_get_context,
1116 .key_prefix = f2fs_key_prefix,
1117 .set_context = f2fs_set_context,
1118 .is_encrypted = f2fs_encrypted_inode,
1119 .empty_dir = f2fs_empty_dir,
1120 .max_namelen = f2fs_max_namelen,
1123 static struct fscrypt_operations f2fs_cryptops = {
1124 .is_encrypted = f2fs_encrypted_inode,
1128 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1129 u64 ino, u32 generation)
1131 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1132 struct inode *inode;
1134 if (check_nid_range(sbi, ino))
1135 return ERR_PTR(-ESTALE);
1138 * f2fs_iget isn't quite right if the inode is currently unallocated!
1139 * However f2fs_iget currently does appropriate checks to handle stale
1140 * inodes so everything is OK.
1142 inode = f2fs_iget(sb, ino);
1144 return ERR_CAST(inode);
1145 if (unlikely(generation && inode->i_generation != generation)) {
1146 /* we didn't find the right inode.. */
1148 return ERR_PTR(-ESTALE);
1153 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1154 int fh_len, int fh_type)
1156 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1157 f2fs_nfs_get_inode);
1160 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1161 int fh_len, int fh_type)
1163 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1164 f2fs_nfs_get_inode);
1167 static const struct export_operations f2fs_export_ops = {
1168 .fh_to_dentry = f2fs_fh_to_dentry,
1169 .fh_to_parent = f2fs_fh_to_parent,
1170 .get_parent = f2fs_get_parent,
1173 static loff_t max_file_blocks(void)
1175 loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
1176 loff_t leaf_count = ADDRS_PER_BLOCK;
1178 /* two direct node blocks */
1179 result += (leaf_count * 2);
1181 /* two indirect node blocks */
1182 leaf_count *= NIDS_PER_BLOCK;
1183 result += (leaf_count * 2);
1185 /* one double indirect node block */
1186 leaf_count *= NIDS_PER_BLOCK;
1187 result += leaf_count;
1192 static int __f2fs_commit_super(struct buffer_head *bh,
1193 struct f2fs_super_block *super)
1197 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
1198 set_buffer_uptodate(bh);
1199 set_buffer_dirty(bh);
1202 /* it's rare case, we can do fua all the time */
1203 return __sync_dirty_buffer(bh, WRITE_FLUSH_FUA);
1206 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
1207 struct buffer_head *bh)
1209 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1210 (bh->b_data + F2FS_SUPER_OFFSET);
1211 struct super_block *sb = sbi->sb;
1212 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
1213 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
1214 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
1215 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
1216 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
1217 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
1218 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
1219 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
1220 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
1221 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
1222 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1223 u32 segment_count = le32_to_cpu(raw_super->segment_count);
1224 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1225 u64 main_end_blkaddr = main_blkaddr +
1226 (segment_count_main << log_blocks_per_seg);
1227 u64 seg_end_blkaddr = segment0_blkaddr +
1228 (segment_count << log_blocks_per_seg);
1230 if (segment0_blkaddr != cp_blkaddr) {
1231 f2fs_msg(sb, KERN_INFO,
1232 "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1233 segment0_blkaddr, cp_blkaddr);
1237 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
1239 f2fs_msg(sb, KERN_INFO,
1240 "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1241 cp_blkaddr, sit_blkaddr,
1242 segment_count_ckpt << log_blocks_per_seg);
1246 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
1248 f2fs_msg(sb, KERN_INFO,
1249 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1250 sit_blkaddr, nat_blkaddr,
1251 segment_count_sit << log_blocks_per_seg);
1255 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
1257 f2fs_msg(sb, KERN_INFO,
1258 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1259 nat_blkaddr, ssa_blkaddr,
1260 segment_count_nat << log_blocks_per_seg);
1264 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
1266 f2fs_msg(sb, KERN_INFO,
1267 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1268 ssa_blkaddr, main_blkaddr,
1269 segment_count_ssa << log_blocks_per_seg);
1273 if (main_end_blkaddr > seg_end_blkaddr) {
1274 f2fs_msg(sb, KERN_INFO,
1275 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1278 (segment_count << log_blocks_per_seg),
1279 segment_count_main << log_blocks_per_seg);
1281 } else if (main_end_blkaddr < seg_end_blkaddr) {
1285 /* fix in-memory information all the time */
1286 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
1287 segment0_blkaddr) >> log_blocks_per_seg);
1289 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
1290 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1293 err = __f2fs_commit_super(bh, NULL);
1294 res = err ? "failed" : "done";
1296 f2fs_msg(sb, KERN_INFO,
1297 "Fix alignment : %s, start(%u) end(%u) block(%u)",
1300 (segment_count << log_blocks_per_seg),
1301 segment_count_main << log_blocks_per_seg);
1308 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
1309 struct buffer_head *bh)
1311 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1312 (bh->b_data + F2FS_SUPER_OFFSET);
1313 struct super_block *sb = sbi->sb;
1314 unsigned int blocksize;
1316 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
1317 f2fs_msg(sb, KERN_INFO,
1318 "Magic Mismatch, valid(0x%x) - read(0x%x)",
1319 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
1323 /* Currently, support only 4KB page cache size */
1324 if (F2FS_BLKSIZE != PAGE_SIZE) {
1325 f2fs_msg(sb, KERN_INFO,
1326 "Invalid page_cache_size (%lu), supports only 4KB\n",
1331 /* Currently, support only 4KB block size */
1332 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
1333 if (blocksize != F2FS_BLKSIZE) {
1334 f2fs_msg(sb, KERN_INFO,
1335 "Invalid blocksize (%u), supports only 4KB\n",
1340 /* check log blocks per segment */
1341 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
1342 f2fs_msg(sb, KERN_INFO,
1343 "Invalid log blocks per segment (%u)\n",
1344 le32_to_cpu(raw_super->log_blocks_per_seg));
1348 /* Currently, support 512/1024/2048/4096 bytes sector size */
1349 if (le32_to_cpu(raw_super->log_sectorsize) >
1350 F2FS_MAX_LOG_SECTOR_SIZE ||
1351 le32_to_cpu(raw_super->log_sectorsize) <
1352 F2FS_MIN_LOG_SECTOR_SIZE) {
1353 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
1354 le32_to_cpu(raw_super->log_sectorsize));
1357 if (le32_to_cpu(raw_super->log_sectors_per_block) +
1358 le32_to_cpu(raw_super->log_sectorsize) !=
1359 F2FS_MAX_LOG_SECTOR_SIZE) {
1360 f2fs_msg(sb, KERN_INFO,
1361 "Invalid log sectors per block(%u) log sectorsize(%u)",
1362 le32_to_cpu(raw_super->log_sectors_per_block),
1363 le32_to_cpu(raw_super->log_sectorsize));
1367 /* check reserved ino info */
1368 if (le32_to_cpu(raw_super->node_ino) != 1 ||
1369 le32_to_cpu(raw_super->meta_ino) != 2 ||
1370 le32_to_cpu(raw_super->root_ino) != 3) {
1371 f2fs_msg(sb, KERN_INFO,
1372 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1373 le32_to_cpu(raw_super->node_ino),
1374 le32_to_cpu(raw_super->meta_ino),
1375 le32_to_cpu(raw_super->root_ino));
1379 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1380 if (sanity_check_area_boundary(sbi, bh))
1386 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1388 unsigned int total, fsmeta;
1389 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1390 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1392 total = le32_to_cpu(raw_super->segment_count);
1393 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
1394 fsmeta += le32_to_cpu(raw_super->segment_count_sit);
1395 fsmeta += le32_to_cpu(raw_super->segment_count_nat);
1396 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
1397 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
1399 if (unlikely(fsmeta >= total))
1402 if (unlikely(f2fs_cp_error(sbi))) {
1403 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
1409 static void init_sb_info(struct f2fs_sb_info *sbi)
1411 struct f2fs_super_block *raw_super = sbi->raw_super;
1413 sbi->log_sectors_per_block =
1414 le32_to_cpu(raw_super->log_sectors_per_block);
1415 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
1416 sbi->blocksize = 1 << sbi->log_blocksize;
1417 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1418 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
1419 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1420 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1421 sbi->total_sections = le32_to_cpu(raw_super->section_count);
1422 sbi->total_node_count =
1423 (le32_to_cpu(raw_super->segment_count_nat) / 2)
1424 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
1425 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
1426 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
1427 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
1428 sbi->cur_victim_sec = NULL_SECNO;
1429 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
1431 sbi->dir_level = DEF_DIR_LEVEL;
1432 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
1433 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
1434 clear_sbi_flag(sbi, SBI_NEED_FSCK);
1436 INIT_LIST_HEAD(&sbi->s_list);
1437 mutex_init(&sbi->umount_mutex);
1439 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1440 memcpy(sbi->key_prefix, F2FS_KEY_DESC_PREFIX,
1441 F2FS_KEY_DESC_PREFIX_SIZE);
1442 sbi->key_prefix_size = F2FS_KEY_DESC_PREFIX_SIZE;
1446 static int init_percpu_info(struct f2fs_sb_info *sbi)
1450 for (i = 0; i < NR_COUNT_TYPE; i++) {
1451 err = percpu_counter_init(&sbi->nr_pages[i], 0, GFP_KERNEL);
1456 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
1460 return percpu_counter_init(&sbi->total_valid_inode_count, 0,
1465 * Read f2fs raw super block.
1466 * Because we have two copies of super block, so read both of them
1467 * to get the first valid one. If any one of them is broken, we pass
1468 * them recovery flag back to the caller.
1470 static int read_raw_super_block(struct f2fs_sb_info *sbi,
1471 struct f2fs_super_block **raw_super,
1472 int *valid_super_block, int *recovery)
1474 struct super_block *sb = sbi->sb;
1476 struct buffer_head *bh;
1477 struct f2fs_super_block *super;
1480 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
1484 for (block = 0; block < 2; block++) {
1485 bh = sb_bread(sb, block);
1487 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
1493 /* sanity checking of raw super */
1494 if (sanity_check_raw_super(sbi, bh)) {
1495 f2fs_msg(sb, KERN_ERR,
1496 "Can't find valid F2FS filesystem in %dth superblock",
1504 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
1506 *valid_super_block = block;
1512 /* Fail to read any one of the superblocks*/
1516 /* No valid superblock */
1525 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
1527 struct buffer_head *bh;
1530 if ((recover && f2fs_readonly(sbi->sb)) ||
1531 bdev_read_only(sbi->sb->s_bdev)) {
1532 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1536 /* write back-up superblock first */
1537 bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1);
1540 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1543 /* if we are in recovery path, skip writing valid superblock */
1547 /* write current valid superblock */
1548 bh = sb_getblk(sbi->sb, sbi->valid_super_block);
1551 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1556 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
1558 struct f2fs_sb_info *sbi;
1559 struct f2fs_super_block *raw_super;
1562 bool retry = true, need_fsck = false;
1563 char *options = NULL;
1564 int recovery, i, valid_super_block;
1565 struct curseg_info *seg_i;
1570 valid_super_block = -1;
1573 /* allocate memory for f2fs-specific super block info */
1574 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
1580 /* Load the checksum driver */
1581 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
1582 if (IS_ERR(sbi->s_chksum_driver)) {
1583 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
1584 err = PTR_ERR(sbi->s_chksum_driver);
1585 sbi->s_chksum_driver = NULL;
1589 /* set a block size */
1590 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
1591 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
1595 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
1600 sb->s_fs_info = sbi;
1601 default_options(sbi);
1602 /* parse mount options */
1603 options = kstrdup((const char *)data, GFP_KERNEL);
1604 if (data && !options) {
1609 err = parse_options(sb, options);
1613 sbi->max_file_blocks = max_file_blocks();
1614 sb->s_maxbytes = sbi->max_file_blocks <<
1615 le32_to_cpu(raw_super->log_blocksize);
1616 sb->s_max_links = F2FS_LINK_MAX;
1617 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1619 sb->s_op = &f2fs_sops;
1620 sb->s_cop = &f2fs_cryptops;
1621 sb->s_xattr = f2fs_xattr_handlers;
1622 sb->s_export_op = &f2fs_export_ops;
1623 sb->s_magic = F2FS_SUPER_MAGIC;
1624 sb->s_time_gran = 1;
1625 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1626 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1627 memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1629 /* init f2fs-specific super block info */
1630 sbi->raw_super = raw_super;
1631 sbi->valid_super_block = valid_super_block;
1632 mutex_init(&sbi->gc_mutex);
1633 mutex_init(&sbi->writepages);
1634 mutex_init(&sbi->cp_mutex);
1635 init_rwsem(&sbi->node_write);
1637 /* disallow all the data/node/meta page writes */
1638 set_sbi_flag(sbi, SBI_POR_DOING);
1639 spin_lock_init(&sbi->stat_lock);
1641 init_rwsem(&sbi->read_io.io_rwsem);
1642 sbi->read_io.sbi = sbi;
1643 sbi->read_io.bio = NULL;
1644 for (i = 0; i < NR_PAGE_TYPE; i++) {
1645 init_rwsem(&sbi->write_io[i].io_rwsem);
1646 sbi->write_io[i].sbi = sbi;
1647 sbi->write_io[i].bio = NULL;
1650 init_rwsem(&sbi->cp_rwsem);
1651 init_waitqueue_head(&sbi->cp_wait);
1654 err = init_percpu_info(sbi);
1658 /* get an inode for meta space */
1659 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1660 if (IS_ERR(sbi->meta_inode)) {
1661 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1662 err = PTR_ERR(sbi->meta_inode);
1666 err = get_valid_checkpoint(sbi);
1668 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1669 goto free_meta_inode;
1672 sbi->total_valid_node_count =
1673 le32_to_cpu(sbi->ckpt->valid_node_count);
1674 percpu_counter_set(&sbi->total_valid_inode_count,
1675 le32_to_cpu(sbi->ckpt->valid_inode_count));
1676 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1677 sbi->total_valid_block_count =
1678 le64_to_cpu(sbi->ckpt->valid_block_count);
1679 sbi->last_valid_block_count = sbi->total_valid_block_count;
1681 for (i = 0; i < NR_INODE_TYPE; i++) {
1682 INIT_LIST_HEAD(&sbi->inode_list[i]);
1683 spin_lock_init(&sbi->inode_lock[i]);
1686 init_extent_cache_info(sbi);
1688 init_ino_entry_info(sbi);
1690 /* setup f2fs internal modules */
1691 err = build_segment_manager(sbi);
1693 f2fs_msg(sb, KERN_ERR,
1694 "Failed to initialize F2FS segment manager");
1697 err = build_node_manager(sbi);
1699 f2fs_msg(sb, KERN_ERR,
1700 "Failed to initialize F2FS node manager");
1704 /* For write statistics */
1705 if (sb->s_bdev->bd_part)
1706 sbi->sectors_written_start =
1707 (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
1709 /* Read accumulated write IO statistics if exists */
1710 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
1711 if (__exist_node_summaries(sbi))
1712 sbi->kbytes_written =
1713 le64_to_cpu(seg_i->journal->info.kbytes_written);
1715 build_gc_manager(sbi);
1717 /* get an inode for node space */
1718 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1719 if (IS_ERR(sbi->node_inode)) {
1720 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1721 err = PTR_ERR(sbi->node_inode);
1725 f2fs_join_shrinker(sbi);
1727 /* if there are nt orphan nodes free them */
1728 err = recover_orphan_inodes(sbi);
1730 goto free_node_inode;
1732 /* read root inode and dentry */
1733 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1735 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1736 err = PTR_ERR(root);
1737 goto free_node_inode;
1739 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1742 goto free_node_inode;
1745 sb->s_root = d_make_root(root); /* allocate root dentry */
1748 goto free_root_inode;
1751 err = f2fs_build_stats(sbi);
1753 goto free_root_inode;
1756 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1759 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1760 &f2fs_seq_segment_info_fops, sb);
1761 proc_create_data("segment_bits", S_IRUGO, sbi->s_proc,
1762 &f2fs_seq_segment_bits_fops, sb);
1765 sbi->s_kobj.kset = f2fs_kset;
1766 init_completion(&sbi->s_kobj_unregister);
1767 err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1772 /* recover fsynced data */
1773 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1775 * mount should be failed, when device has readonly mode, and
1776 * previous checkpoint was not done by clean system shutdown.
1778 if (bdev_read_only(sb->s_bdev) &&
1779 !is_set_ckpt_flags(sbi->ckpt, CP_UMOUNT_FLAG)) {
1785 set_sbi_flag(sbi, SBI_NEED_FSCK);
1787 err = recover_fsync_data(sbi, false);
1790 f2fs_msg(sb, KERN_ERR,
1791 "Cannot recover all fsync data errno=%d", err);
1795 err = recover_fsync_data(sbi, true);
1797 if (!f2fs_readonly(sb) && err > 0) {
1799 f2fs_msg(sb, KERN_ERR,
1800 "Need to recover fsync data");
1805 /* recover_fsync_data() cleared this already */
1806 clear_sbi_flag(sbi, SBI_POR_DOING);
1809 * If filesystem is not mounted as read-only then
1810 * do start the gc_thread.
1812 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
1813 /* After POR, we can run background GC thread.*/
1814 err = start_gc_thread(sbi);
1820 /* recover broken superblock */
1822 err = f2fs_commit_super(sbi, true);
1823 f2fs_msg(sb, KERN_INFO,
1824 "Try to recover %dth superblock, ret: %d",
1825 sbi->valid_super_block ? 1 : 2, err);
1828 f2fs_update_time(sbi, CP_TIME);
1829 f2fs_update_time(sbi, REQ_TIME);
1833 f2fs_sync_inode_meta(sbi);
1834 kobject_del(&sbi->s_kobj);
1835 kobject_put(&sbi->s_kobj);
1836 wait_for_completion(&sbi->s_kobj_unregister);
1839 remove_proc_entry("segment_info", sbi->s_proc);
1840 remove_proc_entry("segment_bits", sbi->s_proc);
1841 remove_proc_entry(sb->s_id, f2fs_proc_root);
1843 f2fs_destroy_stats(sbi);
1848 mutex_lock(&sbi->umount_mutex);
1849 f2fs_leave_shrinker(sbi);
1850 iput(sbi->node_inode);
1851 mutex_unlock(&sbi->umount_mutex);
1853 destroy_node_manager(sbi);
1855 destroy_segment_manager(sbi);
1858 make_bad_inode(sbi->meta_inode);
1859 iput(sbi->meta_inode);
1861 destroy_percpu_info(sbi);
1866 if (sbi->s_chksum_driver)
1867 crypto_free_shash(sbi->s_chksum_driver);
1870 /* give only one another chance */
1873 shrink_dcache_sb(sb);
1879 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1880 const char *dev_name, void *data)
1882 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1885 static void kill_f2fs_super(struct super_block *sb)
1888 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
1889 kill_block_super(sb);
1892 static struct file_system_type f2fs_fs_type = {
1893 .owner = THIS_MODULE,
1895 .mount = f2fs_mount,
1896 .kill_sb = kill_f2fs_super,
1897 .fs_flags = FS_REQUIRES_DEV,
1899 MODULE_ALIAS_FS("f2fs");
1901 static int __init init_inodecache(void)
1903 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
1904 sizeof(struct f2fs_inode_info), 0,
1905 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
1906 if (!f2fs_inode_cachep)
1911 static void destroy_inodecache(void)
1914 * Make sure all delayed rcu free inodes are flushed before we
1918 kmem_cache_destroy(f2fs_inode_cachep);
1921 static int __init init_f2fs_fs(void)
1925 f2fs_build_trace_ios();
1927 err = init_inodecache();
1930 err = create_node_manager_caches();
1932 goto free_inodecache;
1933 err = create_segment_manager_caches();
1935 goto free_node_manager_caches;
1936 err = create_checkpoint_caches();
1938 goto free_segment_manager_caches;
1939 err = create_extent_cache();
1941 goto free_checkpoint_caches;
1942 f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1945 goto free_extent_cache;
1947 #ifdef CONFIG_F2FS_FAULT_INJECTION
1948 f2fs_fault_inject.kset = f2fs_kset;
1949 f2fs_build_fault_attr(0);
1950 err = kobject_init_and_add(&f2fs_fault_inject, &f2fs_fault_ktype,
1951 NULL, "fault_injection");
1953 f2fs_fault_inject.kset = NULL;
1957 err = register_shrinker(&f2fs_shrinker_info);
1961 err = register_filesystem(&f2fs_fs_type);
1964 err = f2fs_create_root_stats();
1966 goto free_filesystem;
1967 f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1971 unregister_filesystem(&f2fs_fs_type);
1973 unregister_shrinker(&f2fs_shrinker_info);
1975 #ifdef CONFIG_F2FS_FAULT_INJECTION
1976 if (f2fs_fault_inject.kset)
1977 kobject_put(&f2fs_fault_inject);
1979 kset_unregister(f2fs_kset);
1981 destroy_extent_cache();
1982 free_checkpoint_caches:
1983 destroy_checkpoint_caches();
1984 free_segment_manager_caches:
1985 destroy_segment_manager_caches();
1986 free_node_manager_caches:
1987 destroy_node_manager_caches();
1989 destroy_inodecache();
1994 static void __exit exit_f2fs_fs(void)
1996 remove_proc_entry("fs/f2fs", NULL);
1997 f2fs_destroy_root_stats();
1998 unregister_filesystem(&f2fs_fs_type);
1999 unregister_shrinker(&f2fs_shrinker_info);
2000 #ifdef CONFIG_F2FS_FAULT_INJECTION
2001 kobject_put(&f2fs_fault_inject);
2003 kset_unregister(f2fs_kset);
2004 destroy_extent_cache();
2005 destroy_checkpoint_caches();
2006 destroy_segment_manager_caches();
2007 destroy_node_manager_caches();
2008 destroy_inodecache();
2009 f2fs_destroy_trace_ios();
2012 module_init(init_f2fs_fs)
2013 module_exit(exit_f2fs_fs)
2015 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
2016 MODULE_DESCRIPTION("Flash Friendly File System");
2017 MODULE_LICENSE("GPL");