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",
52 [FAULT_EVICT_INODE] = "evict_inode fail",
55 static void f2fs_build_fault_attr(unsigned int rate)
58 atomic_set(&f2fs_fault.inject_ops, 0);
59 f2fs_fault.inject_rate = rate;
60 f2fs_fault.inject_type = (1 << FAULT_MAX) - 1;
62 memset(&f2fs_fault, 0, sizeof(struct f2fs_fault_info));
67 /* f2fs-wide shrinker description */
68 static struct shrinker f2fs_shrinker_info = {
69 .scan_objects = f2fs_shrink_scan,
70 .count_objects = f2fs_shrink_count,
71 .seeks = DEFAULT_SEEKS,
76 Opt_disable_roll_forward,
85 Opt_disable_ext_identify,
103 static match_table_t f2fs_tokens = {
104 {Opt_gc_background, "background_gc=%s"},
105 {Opt_disable_roll_forward, "disable_roll_forward"},
106 {Opt_norecovery, "norecovery"},
107 {Opt_discard, "discard"},
108 {Opt_noheap, "no_heap"},
109 {Opt_user_xattr, "user_xattr"},
110 {Opt_nouser_xattr, "nouser_xattr"},
112 {Opt_noacl, "noacl"},
113 {Opt_active_logs, "active_logs=%u"},
114 {Opt_disable_ext_identify, "disable_ext_identify"},
115 {Opt_inline_xattr, "inline_xattr"},
116 {Opt_inline_data, "inline_data"},
117 {Opt_inline_dentry, "inline_dentry"},
118 {Opt_flush_merge, "flush_merge"},
119 {Opt_noflush_merge, "noflush_merge"},
120 {Opt_nobarrier, "nobarrier"},
121 {Opt_fastboot, "fastboot"},
122 {Opt_extent_cache, "extent_cache"},
123 {Opt_noextent_cache, "noextent_cache"},
124 {Opt_noinline_data, "noinline_data"},
125 {Opt_data_flush, "data_flush"},
126 {Opt_fault_injection, "fault_injection=%u"},
127 {Opt_lazytime, "lazytime"},
128 {Opt_nolazytime, "nolazytime"},
132 /* Sysfs support for f2fs */
134 GC_THREAD, /* struct f2fs_gc_thread */
135 SM_INFO, /* struct f2fs_sm_info */
136 NM_INFO, /* struct f2fs_nm_info */
137 F2FS_SBI, /* struct f2fs_sb_info */
138 #ifdef CONFIG_F2FS_FAULT_INJECTION
139 FAULT_INFO_RATE, /* struct f2fs_fault_info */
140 FAULT_INFO_TYPE, /* struct f2fs_fault_info */
145 struct attribute attr;
146 ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
147 ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
148 const char *, size_t);
153 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
155 if (struct_type == GC_THREAD)
156 return (unsigned char *)sbi->gc_thread;
157 else if (struct_type == SM_INFO)
158 return (unsigned char *)SM_I(sbi);
159 else if (struct_type == NM_INFO)
160 return (unsigned char *)NM_I(sbi);
161 else if (struct_type == F2FS_SBI)
162 return (unsigned char *)sbi;
163 #ifdef CONFIG_F2FS_FAULT_INJECTION
164 else if (struct_type == FAULT_INFO_RATE ||
165 struct_type == FAULT_INFO_TYPE)
166 return (unsigned char *)&f2fs_fault;
171 static ssize_t lifetime_write_kbytes_show(struct f2fs_attr *a,
172 struct f2fs_sb_info *sbi, char *buf)
174 struct super_block *sb = sbi->sb;
176 if (!sb->s_bdev->bd_part)
177 return snprintf(buf, PAGE_SIZE, "0\n");
179 return snprintf(buf, PAGE_SIZE, "%llu\n",
180 (unsigned long long)(sbi->kbytes_written +
181 BD_PART_WRITTEN(sbi)));
184 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
185 struct f2fs_sb_info *sbi, char *buf)
187 unsigned char *ptr = NULL;
190 ptr = __struct_ptr(sbi, a->struct_type);
194 ui = (unsigned int *)(ptr + a->offset);
196 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
199 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
200 struct f2fs_sb_info *sbi,
201 const char *buf, size_t count)
208 ptr = __struct_ptr(sbi, a->struct_type);
212 ui = (unsigned int *)(ptr + a->offset);
214 ret = kstrtoul(skip_spaces(buf), 0, &t);
217 #ifdef CONFIG_F2FS_FAULT_INJECTION
218 if (a->struct_type == FAULT_INFO_TYPE && t >= (1 << FAULT_MAX))
225 static ssize_t f2fs_attr_show(struct kobject *kobj,
226 struct attribute *attr, char *buf)
228 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
230 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
232 return a->show ? a->show(a, sbi, buf) : 0;
235 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
236 const char *buf, size_t len)
238 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
240 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
242 return a->store ? a->store(a, sbi, buf, len) : 0;
245 static void f2fs_sb_release(struct kobject *kobj)
247 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
249 complete(&sbi->s_kobj_unregister);
252 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
253 static struct f2fs_attr f2fs_attr_##_name = { \
254 .attr = {.name = __stringify(_name), .mode = _mode }, \
257 .struct_type = _struct_type, \
261 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
262 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
263 f2fs_sbi_show, f2fs_sbi_store, \
264 offsetof(struct struct_name, elname))
266 #define F2FS_GENERAL_RO_ATTR(name) \
267 static struct f2fs_attr f2fs_attr_##name = __ATTR(name, 0444, name##_show, NULL)
269 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
270 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
271 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
272 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
273 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
274 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
275 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, batched_trim_sections, trim_sections);
276 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
277 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
278 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks);
279 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
280 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ra_nid_pages, ra_nid_pages);
281 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, dirty_nats_ratio, dirty_nats_ratio);
282 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
283 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
284 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, cp_interval, interval_time[CP_TIME]);
285 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, idle_interval, interval_time[REQ_TIME]);
286 #ifdef CONFIG_F2FS_FAULT_INJECTION
287 F2FS_RW_ATTR(FAULT_INFO_RATE, f2fs_fault_info, inject_rate, inject_rate);
288 F2FS_RW_ATTR(FAULT_INFO_TYPE, f2fs_fault_info, inject_type, inject_type);
290 F2FS_GENERAL_RO_ATTR(lifetime_write_kbytes);
292 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
293 static struct attribute *f2fs_attrs[] = {
294 ATTR_LIST(gc_min_sleep_time),
295 ATTR_LIST(gc_max_sleep_time),
296 ATTR_LIST(gc_no_gc_sleep_time),
298 ATTR_LIST(reclaim_segments),
299 ATTR_LIST(max_small_discards),
300 ATTR_LIST(batched_trim_sections),
301 ATTR_LIST(ipu_policy),
302 ATTR_LIST(min_ipu_util),
303 ATTR_LIST(min_fsync_blocks),
304 ATTR_LIST(max_victim_search),
305 ATTR_LIST(dir_level),
306 ATTR_LIST(ram_thresh),
307 ATTR_LIST(ra_nid_pages),
308 ATTR_LIST(dirty_nats_ratio),
309 ATTR_LIST(cp_interval),
310 ATTR_LIST(idle_interval),
311 ATTR_LIST(lifetime_write_kbytes),
315 static const struct sysfs_ops f2fs_attr_ops = {
316 .show = f2fs_attr_show,
317 .store = f2fs_attr_store,
320 static struct kobj_type f2fs_ktype = {
321 .default_attrs = f2fs_attrs,
322 .sysfs_ops = &f2fs_attr_ops,
323 .release = f2fs_sb_release,
326 #ifdef CONFIG_F2FS_FAULT_INJECTION
327 /* sysfs for f2fs fault injection */
328 static struct kobject f2fs_fault_inject;
330 static struct attribute *f2fs_fault_attrs[] = {
331 ATTR_LIST(inject_rate),
332 ATTR_LIST(inject_type),
336 static struct kobj_type f2fs_fault_ktype = {
337 .default_attrs = f2fs_fault_attrs,
338 .sysfs_ops = &f2fs_attr_ops,
342 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
344 struct va_format vaf;
350 printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
354 static void init_once(void *foo)
356 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
358 inode_init_once(&fi->vfs_inode);
361 static int parse_options(struct super_block *sb, char *options)
363 struct f2fs_sb_info *sbi = F2FS_SB(sb);
364 struct request_queue *q;
365 substring_t args[MAX_OPT_ARGS];
369 #ifdef CONFIG_F2FS_FAULT_INJECTION
370 f2fs_build_fault_attr(0);
376 while ((p = strsep(&options, ",")) != NULL) {
381 * Initialize args struct so we know whether arg was
382 * found; some options take optional arguments.
384 args[0].to = args[0].from = NULL;
385 token = match_token(p, f2fs_tokens, args);
388 case Opt_gc_background:
389 name = match_strdup(&args[0]);
393 if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
395 clear_opt(sbi, FORCE_FG_GC);
396 } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
397 clear_opt(sbi, BG_GC);
398 clear_opt(sbi, FORCE_FG_GC);
399 } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
401 set_opt(sbi, FORCE_FG_GC);
408 case Opt_disable_roll_forward:
409 set_opt(sbi, DISABLE_ROLL_FORWARD);
412 /* this option mounts f2fs with ro */
413 set_opt(sbi, DISABLE_ROLL_FORWARD);
414 if (!f2fs_readonly(sb))
418 q = bdev_get_queue(sb->s_bdev);
419 if (blk_queue_discard(q)) {
420 set_opt(sbi, DISCARD);
422 f2fs_msg(sb, KERN_WARNING,
423 "mounting with \"discard\" option, but "
424 "the device does not support discard");
428 set_opt(sbi, NOHEAP);
430 #ifdef CONFIG_F2FS_FS_XATTR
432 set_opt(sbi, XATTR_USER);
434 case Opt_nouser_xattr:
435 clear_opt(sbi, XATTR_USER);
437 case Opt_inline_xattr:
438 set_opt(sbi, INLINE_XATTR);
442 f2fs_msg(sb, KERN_INFO,
443 "user_xattr options not supported");
445 case Opt_nouser_xattr:
446 f2fs_msg(sb, KERN_INFO,
447 "nouser_xattr options not supported");
449 case Opt_inline_xattr:
450 f2fs_msg(sb, KERN_INFO,
451 "inline_xattr options not supported");
454 #ifdef CONFIG_F2FS_FS_POSIX_ACL
456 set_opt(sbi, POSIX_ACL);
459 clear_opt(sbi, POSIX_ACL);
463 f2fs_msg(sb, KERN_INFO, "acl options not supported");
466 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
469 case Opt_active_logs:
470 if (args->from && match_int(args, &arg))
472 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
474 sbi->active_logs = arg;
476 case Opt_disable_ext_identify:
477 set_opt(sbi, DISABLE_EXT_IDENTIFY);
479 case Opt_inline_data:
480 set_opt(sbi, INLINE_DATA);
482 case Opt_inline_dentry:
483 set_opt(sbi, INLINE_DENTRY);
485 case Opt_flush_merge:
486 set_opt(sbi, FLUSH_MERGE);
488 case Opt_noflush_merge:
489 clear_opt(sbi, FLUSH_MERGE);
492 set_opt(sbi, NOBARRIER);
495 set_opt(sbi, FASTBOOT);
497 case Opt_extent_cache:
498 set_opt(sbi, EXTENT_CACHE);
500 case Opt_noextent_cache:
501 clear_opt(sbi, EXTENT_CACHE);
503 case Opt_noinline_data:
504 clear_opt(sbi, INLINE_DATA);
507 set_opt(sbi, DATA_FLUSH);
509 case Opt_fault_injection:
510 if (args->from && match_int(args, &arg))
512 #ifdef CONFIG_F2FS_FAULT_INJECTION
513 f2fs_build_fault_attr(arg);
515 f2fs_msg(sb, KERN_INFO,
516 "FAULT_INJECTION was not selected");
520 sb->s_flags |= MS_LAZYTIME;
523 sb->s_flags &= ~MS_LAZYTIME;
526 f2fs_msg(sb, KERN_ERR,
527 "Unrecognized mount option \"%s\" or missing value",
535 static struct inode *f2fs_alloc_inode(struct super_block *sb)
537 struct f2fs_inode_info *fi;
539 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
543 init_once((void *) fi);
545 if (percpu_counter_init(&fi->dirty_pages, 0, GFP_NOFS)) {
546 kmem_cache_free(f2fs_inode_cachep, fi);
550 /* Initialize f2fs-specific inode info */
551 fi->vfs_inode.i_version = 1;
552 fi->i_current_depth = 1;
554 init_rwsem(&fi->i_sem);
555 INIT_LIST_HEAD(&fi->dirty_list);
556 INIT_LIST_HEAD(&fi->gdirty_list);
557 INIT_LIST_HEAD(&fi->inmem_pages);
558 mutex_init(&fi->inmem_lock);
560 /* Will be used by directory only */
561 fi->i_dir_level = F2FS_SB(sb)->dir_level;
562 return &fi->vfs_inode;
565 static int f2fs_drop_inode(struct inode *inode)
570 * This is to avoid a deadlock condition like below.
571 * writeback_single_inode(inode)
572 * - f2fs_write_data_page
573 * - f2fs_gc -> iput -> evict
574 * - inode_wait_for_writeback(inode)
576 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
577 if (!inode->i_nlink && !is_bad_inode(inode)) {
578 /* to avoid evict_inode call simultaneously */
579 atomic_inc(&inode->i_count);
580 spin_unlock(&inode->i_lock);
582 /* some remained atomic pages should discarded */
583 if (f2fs_is_atomic_file(inode))
584 drop_inmem_pages(inode);
586 /* should remain fi->extent_tree for writepage */
587 f2fs_destroy_extent_node(inode);
589 sb_start_intwrite(inode->i_sb);
590 f2fs_i_size_write(inode, 0);
592 if (F2FS_HAS_BLOCKS(inode))
593 f2fs_truncate(inode);
595 sb_end_intwrite(inode->i_sb);
597 fscrypt_put_encryption_info(inode, NULL);
598 spin_lock(&inode->i_lock);
599 atomic_dec(&inode->i_count);
604 ret = generic_drop_inode(inode);
605 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
607 inode->i_state |= I_WILL_FREE;
608 spin_unlock(&inode->i_lock);
610 update_inode_page(inode);
612 spin_lock(&inode->i_lock);
614 inode->i_state &= ~I_WILL_FREE;
620 * f2fs_dirty_inode() is called from __mark_inode_dirty()
622 * We should call set_dirty_inode to write the dirty inode through write_inode.
624 static void f2fs_dirty_inode(struct inode *inode, int flags)
626 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
628 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
629 inode->i_ino == F2FS_META_INO(sbi))
632 if (flags == I_DIRTY_TIME)
635 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
636 clear_inode_flag(inode, FI_AUTO_RECOVER);
638 spin_lock(&sbi->inode_lock[DIRTY_META]);
639 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
640 spin_unlock(&sbi->inode_lock[DIRTY_META]);
644 set_inode_flag(inode, FI_DIRTY_INODE);
645 list_add_tail(&F2FS_I(inode)->gdirty_list,
646 &sbi->inode_list[DIRTY_META]);
647 inc_page_count(sbi, F2FS_DIRTY_IMETA);
648 stat_inc_dirty_inode(sbi, DIRTY_META);
649 spin_unlock(&sbi->inode_lock[DIRTY_META]);
652 void f2fs_inode_synced(struct inode *inode)
654 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
656 spin_lock(&sbi->inode_lock[DIRTY_META]);
657 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
658 spin_unlock(&sbi->inode_lock[DIRTY_META]);
661 list_del_init(&F2FS_I(inode)->gdirty_list);
662 clear_inode_flag(inode, FI_DIRTY_INODE);
663 clear_inode_flag(inode, FI_AUTO_RECOVER);
664 dec_page_count(sbi, F2FS_DIRTY_IMETA);
665 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
666 spin_unlock(&sbi->inode_lock[DIRTY_META]);
669 static void f2fs_i_callback(struct rcu_head *head)
671 struct inode *inode = container_of(head, struct inode, i_rcu);
672 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
675 static void f2fs_destroy_inode(struct inode *inode)
677 percpu_counter_destroy(&F2FS_I(inode)->dirty_pages);
678 call_rcu(&inode->i_rcu, f2fs_i_callback);
681 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
685 for (i = 0; i < NR_COUNT_TYPE; i++)
686 percpu_counter_destroy(&sbi->nr_pages[i]);
687 percpu_counter_destroy(&sbi->alloc_valid_block_count);
688 percpu_counter_destroy(&sbi->total_valid_inode_count);
691 static void f2fs_put_super(struct super_block *sb)
693 struct f2fs_sb_info *sbi = F2FS_SB(sb);
696 remove_proc_entry("segment_info", sbi->s_proc);
697 remove_proc_entry("segment_bits", sbi->s_proc);
698 remove_proc_entry(sb->s_id, f2fs_proc_root);
700 kobject_del(&sbi->s_kobj);
704 /* prevent remaining shrinker jobs */
705 mutex_lock(&sbi->umount_mutex);
708 * We don't need to do checkpoint when superblock is clean.
709 * But, the previous checkpoint was not done by umount, it needs to do
710 * clean checkpoint again.
712 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
713 !is_set_ckpt_flags(F2FS_CKPT(sbi), CP_UMOUNT_FLAG)) {
714 struct cp_control cpc = {
717 write_checkpoint(sbi, &cpc);
720 /* write_checkpoint can update stat informaion */
721 f2fs_destroy_stats(sbi);
724 * normally superblock is clean, so we need to release this.
725 * In addition, EIO will skip do checkpoint, we need this as well.
727 release_ino_entry(sbi, true);
728 release_discard_addrs(sbi);
730 f2fs_leave_shrinker(sbi);
731 mutex_unlock(&sbi->umount_mutex);
733 /* our cp_error case, we can wait for any writeback page */
734 f2fs_flush_merged_bios(sbi);
736 iput(sbi->node_inode);
737 iput(sbi->meta_inode);
739 /* destroy f2fs internal modules */
740 destroy_node_manager(sbi);
741 destroy_segment_manager(sbi);
744 kobject_put(&sbi->s_kobj);
745 wait_for_completion(&sbi->s_kobj_unregister);
747 sb->s_fs_info = NULL;
748 if (sbi->s_chksum_driver)
749 crypto_free_shash(sbi->s_chksum_driver);
750 kfree(sbi->raw_super);
752 destroy_percpu_info(sbi);
756 int f2fs_sync_fs(struct super_block *sb, int sync)
758 struct f2fs_sb_info *sbi = F2FS_SB(sb);
761 trace_f2fs_sync_fs(sb, sync);
764 struct cp_control cpc;
766 cpc.reason = __get_cp_reason(sbi);
768 mutex_lock(&sbi->gc_mutex);
769 err = write_checkpoint(sbi, &cpc);
770 mutex_unlock(&sbi->gc_mutex);
772 f2fs_trace_ios(NULL, 1);
777 static int f2fs_freeze(struct super_block *sb)
781 if (f2fs_readonly(sb))
784 err = f2fs_sync_fs(sb, 1);
788 static int f2fs_unfreeze(struct super_block *sb)
793 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
795 struct super_block *sb = dentry->d_sb;
796 struct f2fs_sb_info *sbi = F2FS_SB(sb);
797 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
798 block_t total_count, user_block_count, start_count, ovp_count;
800 total_count = le64_to_cpu(sbi->raw_super->block_count);
801 user_block_count = sbi->user_block_count;
802 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
803 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
804 buf->f_type = F2FS_SUPER_MAGIC;
805 buf->f_bsize = sbi->blocksize;
807 buf->f_blocks = total_count - start_count;
808 buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
809 buf->f_bavail = user_block_count - valid_user_blocks(sbi);
811 buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
812 buf->f_ffree = buf->f_files - valid_inode_count(sbi);
814 buf->f_namelen = F2FS_NAME_LEN;
815 buf->f_fsid.val[0] = (u32)id;
816 buf->f_fsid.val[1] = (u32)(id >> 32);
821 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
823 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
825 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
826 if (test_opt(sbi, FORCE_FG_GC))
827 seq_printf(seq, ",background_gc=%s", "sync");
829 seq_printf(seq, ",background_gc=%s", "on");
831 seq_printf(seq, ",background_gc=%s", "off");
833 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
834 seq_puts(seq, ",disable_roll_forward");
835 if (test_opt(sbi, DISCARD))
836 seq_puts(seq, ",discard");
837 if (test_opt(sbi, NOHEAP))
838 seq_puts(seq, ",no_heap_alloc");
839 #ifdef CONFIG_F2FS_FS_XATTR
840 if (test_opt(sbi, XATTR_USER))
841 seq_puts(seq, ",user_xattr");
843 seq_puts(seq, ",nouser_xattr");
844 if (test_opt(sbi, INLINE_XATTR))
845 seq_puts(seq, ",inline_xattr");
847 #ifdef CONFIG_F2FS_FS_POSIX_ACL
848 if (test_opt(sbi, POSIX_ACL))
849 seq_puts(seq, ",acl");
851 seq_puts(seq, ",noacl");
853 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
854 seq_puts(seq, ",disable_ext_identify");
855 if (test_opt(sbi, INLINE_DATA))
856 seq_puts(seq, ",inline_data");
858 seq_puts(seq, ",noinline_data");
859 if (test_opt(sbi, INLINE_DENTRY))
860 seq_puts(seq, ",inline_dentry");
861 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
862 seq_puts(seq, ",flush_merge");
863 if (test_opt(sbi, NOBARRIER))
864 seq_puts(seq, ",nobarrier");
865 if (test_opt(sbi, FASTBOOT))
866 seq_puts(seq, ",fastboot");
867 if (test_opt(sbi, EXTENT_CACHE))
868 seq_puts(seq, ",extent_cache");
870 seq_puts(seq, ",noextent_cache");
871 if (test_opt(sbi, DATA_FLUSH))
872 seq_puts(seq, ",data_flush");
873 seq_printf(seq, ",active_logs=%u", sbi->active_logs);
878 static int segment_info_seq_show(struct seq_file *seq, void *offset)
880 struct super_block *sb = seq->private;
881 struct f2fs_sb_info *sbi = F2FS_SB(sb);
882 unsigned int total_segs =
883 le32_to_cpu(sbi->raw_super->segment_count_main);
886 seq_puts(seq, "format: segment_type|valid_blocks\n"
887 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
889 for (i = 0; i < total_segs; i++) {
890 struct seg_entry *se = get_seg_entry(sbi, i);
893 seq_printf(seq, "%-10d", i);
894 seq_printf(seq, "%d|%-3u", se->type,
895 get_valid_blocks(sbi, i, 1));
896 if ((i % 10) == 9 || i == (total_segs - 1))
905 static int segment_bits_seq_show(struct seq_file *seq, void *offset)
907 struct super_block *sb = seq->private;
908 struct f2fs_sb_info *sbi = F2FS_SB(sb);
909 unsigned int total_segs =
910 le32_to_cpu(sbi->raw_super->segment_count_main);
913 seq_puts(seq, "format: segment_type|valid_blocks|bitmaps\n"
914 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
916 for (i = 0; i < total_segs; i++) {
917 struct seg_entry *se = get_seg_entry(sbi, i);
919 seq_printf(seq, "%-10d", i);
920 seq_printf(seq, "%d|%-3u|", se->type,
921 get_valid_blocks(sbi, i, 1));
922 for (j = 0; j < SIT_VBLOCK_MAP_SIZE; j++)
923 seq_printf(seq, "%x ", se->cur_valid_map[j]);
929 #define F2FS_PROC_FILE_DEF(_name) \
930 static int _name##_open_fs(struct inode *inode, struct file *file) \
932 return single_open(file, _name##_seq_show, PDE_DATA(inode)); \
935 static const struct file_operations f2fs_seq_##_name##_fops = { \
936 .owner = THIS_MODULE, \
937 .open = _name##_open_fs, \
939 .llseek = seq_lseek, \
940 .release = single_release, \
943 F2FS_PROC_FILE_DEF(segment_info);
944 F2FS_PROC_FILE_DEF(segment_bits);
946 static void default_options(struct f2fs_sb_info *sbi)
948 /* init some FS parameters */
949 sbi->active_logs = NR_CURSEG_TYPE;
952 set_opt(sbi, INLINE_DATA);
953 set_opt(sbi, EXTENT_CACHE);
954 sbi->sb->s_flags |= MS_LAZYTIME;
955 set_opt(sbi, FLUSH_MERGE);
957 #ifdef CONFIG_F2FS_FS_XATTR
958 set_opt(sbi, XATTR_USER);
960 #ifdef CONFIG_F2FS_FS_POSIX_ACL
961 set_opt(sbi, POSIX_ACL);
965 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
967 struct f2fs_sb_info *sbi = F2FS_SB(sb);
968 struct f2fs_mount_info org_mount_opt;
969 int err, active_logs;
970 bool need_restart_gc = false;
971 bool need_stop_gc = false;
972 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
975 * Save the old mount options in case we
976 * need to restore them.
978 org_mount_opt = sbi->mount_opt;
979 active_logs = sbi->active_logs;
981 /* recover superblocks we couldn't write due to previous RO mount */
982 if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
983 err = f2fs_commit_super(sbi, false);
984 f2fs_msg(sb, KERN_INFO,
985 "Try to recover all the superblocks, ret: %d", err);
987 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
990 sbi->mount_opt.opt = 0;
991 default_options(sbi);
993 /* parse mount options */
994 err = parse_options(sb, data);
999 * Previous and new state of filesystem is RO,
1000 * so skip checking GC and FLUSH_MERGE conditions.
1002 if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
1005 /* disallow enable/disable extent_cache dynamically */
1006 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1008 f2fs_msg(sbi->sb, KERN_WARNING,
1009 "switch extent_cache option is not allowed");
1014 * We stop the GC thread if FS is mounted as RO
1015 * or if background_gc = off is passed in mount
1016 * option. Also sync the filesystem.
1018 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
1019 if (sbi->gc_thread) {
1020 stop_gc_thread(sbi);
1021 need_restart_gc = true;
1023 } else if (!sbi->gc_thread) {
1024 err = start_gc_thread(sbi);
1027 need_stop_gc = true;
1030 if (*flags & MS_RDONLY) {
1031 writeback_inodes_sb(sb, WB_REASON_SYNC);
1034 set_sbi_flag(sbi, SBI_IS_DIRTY);
1035 set_sbi_flag(sbi, SBI_IS_CLOSE);
1036 f2fs_sync_fs(sb, 1);
1037 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1041 * We stop issue flush thread if FS is mounted as RO
1042 * or if flush_merge is not passed in mount option.
1044 if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1045 destroy_flush_cmd_control(sbi);
1046 } else if (!SM_I(sbi)->cmd_control_info) {
1047 err = create_flush_cmd_control(sbi);
1052 /* Update the POSIXACL Flag */
1053 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1054 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1058 if (need_restart_gc) {
1059 if (start_gc_thread(sbi))
1060 f2fs_msg(sbi->sb, KERN_WARNING,
1061 "background gc thread has stopped");
1062 } else if (need_stop_gc) {
1063 stop_gc_thread(sbi);
1066 sbi->mount_opt = org_mount_opt;
1067 sbi->active_logs = active_logs;
1071 static struct super_operations f2fs_sops = {
1072 .alloc_inode = f2fs_alloc_inode,
1073 .drop_inode = f2fs_drop_inode,
1074 .destroy_inode = f2fs_destroy_inode,
1075 .write_inode = f2fs_write_inode,
1076 .dirty_inode = f2fs_dirty_inode,
1077 .show_options = f2fs_show_options,
1078 .evict_inode = f2fs_evict_inode,
1079 .put_super = f2fs_put_super,
1080 .sync_fs = f2fs_sync_fs,
1081 .freeze_fs = f2fs_freeze,
1082 .unfreeze_fs = f2fs_unfreeze,
1083 .statfs = f2fs_statfs,
1084 .remount_fs = f2fs_remount,
1087 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1088 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1090 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1091 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1095 static int f2fs_key_prefix(struct inode *inode, u8 **key)
1097 *key = F2FS_I_SB(inode)->key_prefix;
1098 return F2FS_I_SB(inode)->key_prefix_size;
1101 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1104 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1105 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1106 ctx, len, fs_data, XATTR_CREATE);
1109 static unsigned f2fs_max_namelen(struct inode *inode)
1111 return S_ISLNK(inode->i_mode) ?
1112 inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1115 static struct fscrypt_operations f2fs_cryptops = {
1116 .get_context = f2fs_get_context,
1117 .key_prefix = f2fs_key_prefix,
1118 .set_context = f2fs_set_context,
1119 .is_encrypted = f2fs_encrypted_inode,
1120 .empty_dir = f2fs_empty_dir,
1121 .max_namelen = f2fs_max_namelen,
1124 static struct fscrypt_operations f2fs_cryptops = {
1125 .is_encrypted = f2fs_encrypted_inode,
1129 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1130 u64 ino, u32 generation)
1132 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1133 struct inode *inode;
1135 if (check_nid_range(sbi, ino))
1136 return ERR_PTR(-ESTALE);
1139 * f2fs_iget isn't quite right if the inode is currently unallocated!
1140 * However f2fs_iget currently does appropriate checks to handle stale
1141 * inodes so everything is OK.
1143 inode = f2fs_iget(sb, ino);
1145 return ERR_CAST(inode);
1146 if (unlikely(generation && inode->i_generation != generation)) {
1147 /* we didn't find the right inode.. */
1149 return ERR_PTR(-ESTALE);
1154 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1155 int fh_len, int fh_type)
1157 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1158 f2fs_nfs_get_inode);
1161 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1162 int fh_len, int fh_type)
1164 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1165 f2fs_nfs_get_inode);
1168 static const struct export_operations f2fs_export_ops = {
1169 .fh_to_dentry = f2fs_fh_to_dentry,
1170 .fh_to_parent = f2fs_fh_to_parent,
1171 .get_parent = f2fs_get_parent,
1174 static loff_t max_file_blocks(void)
1176 loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
1177 loff_t leaf_count = ADDRS_PER_BLOCK;
1179 /* two direct node blocks */
1180 result += (leaf_count * 2);
1182 /* two indirect node blocks */
1183 leaf_count *= NIDS_PER_BLOCK;
1184 result += (leaf_count * 2);
1186 /* one double indirect node block */
1187 leaf_count *= NIDS_PER_BLOCK;
1188 result += leaf_count;
1193 static int __f2fs_commit_super(struct buffer_head *bh,
1194 struct f2fs_super_block *super)
1198 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
1199 set_buffer_uptodate(bh);
1200 set_buffer_dirty(bh);
1203 /* it's rare case, we can do fua all the time */
1204 return __sync_dirty_buffer(bh, WRITE_FLUSH_FUA);
1207 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
1208 struct buffer_head *bh)
1210 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1211 (bh->b_data + F2FS_SUPER_OFFSET);
1212 struct super_block *sb = sbi->sb;
1213 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
1214 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
1215 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
1216 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
1217 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
1218 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
1219 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
1220 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
1221 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
1222 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
1223 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1224 u32 segment_count = le32_to_cpu(raw_super->segment_count);
1225 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1226 u64 main_end_blkaddr = main_blkaddr +
1227 (segment_count_main << log_blocks_per_seg);
1228 u64 seg_end_blkaddr = segment0_blkaddr +
1229 (segment_count << log_blocks_per_seg);
1231 if (segment0_blkaddr != cp_blkaddr) {
1232 f2fs_msg(sb, KERN_INFO,
1233 "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1234 segment0_blkaddr, cp_blkaddr);
1238 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
1240 f2fs_msg(sb, KERN_INFO,
1241 "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1242 cp_blkaddr, sit_blkaddr,
1243 segment_count_ckpt << log_blocks_per_seg);
1247 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
1249 f2fs_msg(sb, KERN_INFO,
1250 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1251 sit_blkaddr, nat_blkaddr,
1252 segment_count_sit << log_blocks_per_seg);
1256 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
1258 f2fs_msg(sb, KERN_INFO,
1259 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1260 nat_blkaddr, ssa_blkaddr,
1261 segment_count_nat << log_blocks_per_seg);
1265 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
1267 f2fs_msg(sb, KERN_INFO,
1268 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1269 ssa_blkaddr, main_blkaddr,
1270 segment_count_ssa << log_blocks_per_seg);
1274 if (main_end_blkaddr > seg_end_blkaddr) {
1275 f2fs_msg(sb, KERN_INFO,
1276 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1279 (segment_count << log_blocks_per_seg),
1280 segment_count_main << log_blocks_per_seg);
1282 } else if (main_end_blkaddr < seg_end_blkaddr) {
1286 /* fix in-memory information all the time */
1287 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
1288 segment0_blkaddr) >> log_blocks_per_seg);
1290 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
1291 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1294 err = __f2fs_commit_super(bh, NULL);
1295 res = err ? "failed" : "done";
1297 f2fs_msg(sb, KERN_INFO,
1298 "Fix alignment : %s, start(%u) end(%u) block(%u)",
1301 (segment_count << log_blocks_per_seg),
1302 segment_count_main << log_blocks_per_seg);
1309 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
1310 struct buffer_head *bh)
1312 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1313 (bh->b_data + F2FS_SUPER_OFFSET);
1314 struct super_block *sb = sbi->sb;
1315 unsigned int blocksize;
1317 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
1318 f2fs_msg(sb, KERN_INFO,
1319 "Magic Mismatch, valid(0x%x) - read(0x%x)",
1320 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
1324 /* Currently, support only 4KB page cache size */
1325 if (F2FS_BLKSIZE != PAGE_SIZE) {
1326 f2fs_msg(sb, KERN_INFO,
1327 "Invalid page_cache_size (%lu), supports only 4KB\n",
1332 /* Currently, support only 4KB block size */
1333 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
1334 if (blocksize != F2FS_BLKSIZE) {
1335 f2fs_msg(sb, KERN_INFO,
1336 "Invalid blocksize (%u), supports only 4KB\n",
1341 /* check log blocks per segment */
1342 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
1343 f2fs_msg(sb, KERN_INFO,
1344 "Invalid log blocks per segment (%u)\n",
1345 le32_to_cpu(raw_super->log_blocks_per_seg));
1349 /* Currently, support 512/1024/2048/4096 bytes sector size */
1350 if (le32_to_cpu(raw_super->log_sectorsize) >
1351 F2FS_MAX_LOG_SECTOR_SIZE ||
1352 le32_to_cpu(raw_super->log_sectorsize) <
1353 F2FS_MIN_LOG_SECTOR_SIZE) {
1354 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
1355 le32_to_cpu(raw_super->log_sectorsize));
1358 if (le32_to_cpu(raw_super->log_sectors_per_block) +
1359 le32_to_cpu(raw_super->log_sectorsize) !=
1360 F2FS_MAX_LOG_SECTOR_SIZE) {
1361 f2fs_msg(sb, KERN_INFO,
1362 "Invalid log sectors per block(%u) log sectorsize(%u)",
1363 le32_to_cpu(raw_super->log_sectors_per_block),
1364 le32_to_cpu(raw_super->log_sectorsize));
1368 /* check reserved ino info */
1369 if (le32_to_cpu(raw_super->node_ino) != 1 ||
1370 le32_to_cpu(raw_super->meta_ino) != 2 ||
1371 le32_to_cpu(raw_super->root_ino) != 3) {
1372 f2fs_msg(sb, KERN_INFO,
1373 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1374 le32_to_cpu(raw_super->node_ino),
1375 le32_to_cpu(raw_super->meta_ino),
1376 le32_to_cpu(raw_super->root_ino));
1380 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1381 if (sanity_check_area_boundary(sbi, bh))
1387 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1389 unsigned int total, fsmeta;
1390 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1391 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1393 total = le32_to_cpu(raw_super->segment_count);
1394 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
1395 fsmeta += le32_to_cpu(raw_super->segment_count_sit);
1396 fsmeta += le32_to_cpu(raw_super->segment_count_nat);
1397 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
1398 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
1400 if (unlikely(fsmeta >= total))
1403 if (unlikely(f2fs_cp_error(sbi))) {
1404 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
1410 static void init_sb_info(struct f2fs_sb_info *sbi)
1412 struct f2fs_super_block *raw_super = sbi->raw_super;
1414 sbi->log_sectors_per_block =
1415 le32_to_cpu(raw_super->log_sectors_per_block);
1416 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
1417 sbi->blocksize = 1 << sbi->log_blocksize;
1418 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1419 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
1420 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1421 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1422 sbi->total_sections = le32_to_cpu(raw_super->section_count);
1423 sbi->total_node_count =
1424 (le32_to_cpu(raw_super->segment_count_nat) / 2)
1425 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
1426 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
1427 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
1428 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
1429 sbi->cur_victim_sec = NULL_SECNO;
1430 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
1432 sbi->dir_level = DEF_DIR_LEVEL;
1433 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
1434 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
1435 clear_sbi_flag(sbi, SBI_NEED_FSCK);
1437 INIT_LIST_HEAD(&sbi->s_list);
1438 mutex_init(&sbi->umount_mutex);
1439 mutex_init(&sbi->wio_mutex[NODE]);
1440 mutex_init(&sbi->wio_mutex[DATA]);
1442 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1443 memcpy(sbi->key_prefix, F2FS_KEY_DESC_PREFIX,
1444 F2FS_KEY_DESC_PREFIX_SIZE);
1445 sbi->key_prefix_size = F2FS_KEY_DESC_PREFIX_SIZE;
1449 static int init_percpu_info(struct f2fs_sb_info *sbi)
1453 for (i = 0; i < NR_COUNT_TYPE; i++) {
1454 err = percpu_counter_init(&sbi->nr_pages[i], 0, GFP_KERNEL);
1459 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
1463 return percpu_counter_init(&sbi->total_valid_inode_count, 0,
1468 * Read f2fs raw super block.
1469 * Because we have two copies of super block, so read both of them
1470 * to get the first valid one. If any one of them is broken, we pass
1471 * them recovery flag back to the caller.
1473 static int read_raw_super_block(struct f2fs_sb_info *sbi,
1474 struct f2fs_super_block **raw_super,
1475 int *valid_super_block, int *recovery)
1477 struct super_block *sb = sbi->sb;
1479 struct buffer_head *bh;
1480 struct f2fs_super_block *super;
1483 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
1487 for (block = 0; block < 2; block++) {
1488 bh = sb_bread(sb, block);
1490 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
1496 /* sanity checking of raw super */
1497 if (sanity_check_raw_super(sbi, bh)) {
1498 f2fs_msg(sb, KERN_ERR,
1499 "Can't find valid F2FS filesystem in %dth superblock",
1507 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
1509 *valid_super_block = block;
1515 /* Fail to read any one of the superblocks*/
1519 /* No valid superblock */
1528 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
1530 struct buffer_head *bh;
1533 if ((recover && f2fs_readonly(sbi->sb)) ||
1534 bdev_read_only(sbi->sb->s_bdev)) {
1535 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1539 /* write back-up superblock first */
1540 bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1);
1543 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1546 /* if we are in recovery path, skip writing valid superblock */
1550 /* write current valid superblock */
1551 bh = sb_getblk(sbi->sb, sbi->valid_super_block);
1554 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1559 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
1561 struct f2fs_sb_info *sbi;
1562 struct f2fs_super_block *raw_super;
1565 bool retry = true, need_fsck = false;
1566 char *options = NULL;
1567 int recovery, i, valid_super_block;
1568 struct curseg_info *seg_i;
1573 valid_super_block = -1;
1576 /* allocate memory for f2fs-specific super block info */
1577 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
1583 /* Load the checksum driver */
1584 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
1585 if (IS_ERR(sbi->s_chksum_driver)) {
1586 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
1587 err = PTR_ERR(sbi->s_chksum_driver);
1588 sbi->s_chksum_driver = NULL;
1592 /* set a block size */
1593 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
1594 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
1598 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
1603 sb->s_fs_info = sbi;
1604 default_options(sbi);
1605 /* parse mount options */
1606 options = kstrdup((const char *)data, GFP_KERNEL);
1607 if (data && !options) {
1612 err = parse_options(sb, options);
1616 sbi->max_file_blocks = max_file_blocks();
1617 sb->s_maxbytes = sbi->max_file_blocks <<
1618 le32_to_cpu(raw_super->log_blocksize);
1619 sb->s_max_links = F2FS_LINK_MAX;
1620 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1622 sb->s_op = &f2fs_sops;
1623 sb->s_cop = &f2fs_cryptops;
1624 sb->s_xattr = f2fs_xattr_handlers;
1625 sb->s_export_op = &f2fs_export_ops;
1626 sb->s_magic = F2FS_SUPER_MAGIC;
1627 sb->s_time_gran = 1;
1628 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1629 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1630 memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1632 /* init f2fs-specific super block info */
1633 sbi->raw_super = raw_super;
1634 sbi->valid_super_block = valid_super_block;
1635 mutex_init(&sbi->gc_mutex);
1636 mutex_init(&sbi->cp_mutex);
1637 init_rwsem(&sbi->node_write);
1639 /* disallow all the data/node/meta page writes */
1640 set_sbi_flag(sbi, SBI_POR_DOING);
1641 spin_lock_init(&sbi->stat_lock);
1643 init_rwsem(&sbi->read_io.io_rwsem);
1644 sbi->read_io.sbi = sbi;
1645 sbi->read_io.bio = NULL;
1646 for (i = 0; i < NR_PAGE_TYPE; i++) {
1647 init_rwsem(&sbi->write_io[i].io_rwsem);
1648 sbi->write_io[i].sbi = sbi;
1649 sbi->write_io[i].bio = NULL;
1652 init_rwsem(&sbi->cp_rwsem);
1653 init_waitqueue_head(&sbi->cp_wait);
1656 err = init_percpu_info(sbi);
1660 /* get an inode for meta space */
1661 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1662 if (IS_ERR(sbi->meta_inode)) {
1663 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1664 err = PTR_ERR(sbi->meta_inode);
1668 err = get_valid_checkpoint(sbi);
1670 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1671 goto free_meta_inode;
1674 sbi->total_valid_node_count =
1675 le32_to_cpu(sbi->ckpt->valid_node_count);
1676 percpu_counter_set(&sbi->total_valid_inode_count,
1677 le32_to_cpu(sbi->ckpt->valid_inode_count));
1678 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1679 sbi->total_valid_block_count =
1680 le64_to_cpu(sbi->ckpt->valid_block_count);
1681 sbi->last_valid_block_count = sbi->total_valid_block_count;
1683 for (i = 0; i < NR_INODE_TYPE; i++) {
1684 INIT_LIST_HEAD(&sbi->inode_list[i]);
1685 spin_lock_init(&sbi->inode_lock[i]);
1688 init_extent_cache_info(sbi);
1690 init_ino_entry_info(sbi);
1692 /* setup f2fs internal modules */
1693 err = build_segment_manager(sbi);
1695 f2fs_msg(sb, KERN_ERR,
1696 "Failed to initialize F2FS segment manager");
1699 err = build_node_manager(sbi);
1701 f2fs_msg(sb, KERN_ERR,
1702 "Failed to initialize F2FS node manager");
1706 /* For write statistics */
1707 if (sb->s_bdev->bd_part)
1708 sbi->sectors_written_start =
1709 (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
1711 /* Read accumulated write IO statistics if exists */
1712 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
1713 if (__exist_node_summaries(sbi))
1714 sbi->kbytes_written =
1715 le64_to_cpu(seg_i->journal->info.kbytes_written);
1717 build_gc_manager(sbi);
1719 /* get an inode for node space */
1720 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1721 if (IS_ERR(sbi->node_inode)) {
1722 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1723 err = PTR_ERR(sbi->node_inode);
1727 f2fs_join_shrinker(sbi);
1729 /* if there are nt orphan nodes free them */
1730 err = recover_orphan_inodes(sbi);
1732 goto free_node_inode;
1734 /* read root inode and dentry */
1735 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1737 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1738 err = PTR_ERR(root);
1739 goto free_node_inode;
1741 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1744 goto free_node_inode;
1747 sb->s_root = d_make_root(root); /* allocate root dentry */
1750 goto free_root_inode;
1753 err = f2fs_build_stats(sbi);
1755 goto free_root_inode;
1758 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1761 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1762 &f2fs_seq_segment_info_fops, sb);
1763 proc_create_data("segment_bits", S_IRUGO, sbi->s_proc,
1764 &f2fs_seq_segment_bits_fops, sb);
1767 sbi->s_kobj.kset = f2fs_kset;
1768 init_completion(&sbi->s_kobj_unregister);
1769 err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1774 /* recover fsynced data */
1775 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1777 * mount should be failed, when device has readonly mode, and
1778 * previous checkpoint was not done by clean system shutdown.
1780 if (bdev_read_only(sb->s_bdev) &&
1781 !is_set_ckpt_flags(sbi->ckpt, CP_UMOUNT_FLAG)) {
1787 set_sbi_flag(sbi, SBI_NEED_FSCK);
1789 err = recover_fsync_data(sbi, false);
1792 f2fs_msg(sb, KERN_ERR,
1793 "Cannot recover all fsync data errno=%d", err);
1797 err = recover_fsync_data(sbi, true);
1799 if (!f2fs_readonly(sb) && err > 0) {
1801 f2fs_msg(sb, KERN_ERR,
1802 "Need to recover fsync data");
1807 /* recover_fsync_data() cleared this already */
1808 clear_sbi_flag(sbi, SBI_POR_DOING);
1811 * If filesystem is not mounted as read-only then
1812 * do start the gc_thread.
1814 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
1815 /* After POR, we can run background GC thread.*/
1816 err = start_gc_thread(sbi);
1822 /* recover broken superblock */
1824 err = f2fs_commit_super(sbi, true);
1825 f2fs_msg(sb, KERN_INFO,
1826 "Try to recover %dth superblock, ret: %d",
1827 sbi->valid_super_block ? 1 : 2, err);
1830 f2fs_update_time(sbi, CP_TIME);
1831 f2fs_update_time(sbi, REQ_TIME);
1835 f2fs_sync_inode_meta(sbi);
1836 kobject_del(&sbi->s_kobj);
1837 kobject_put(&sbi->s_kobj);
1838 wait_for_completion(&sbi->s_kobj_unregister);
1841 remove_proc_entry("segment_info", sbi->s_proc);
1842 remove_proc_entry("segment_bits", sbi->s_proc);
1843 remove_proc_entry(sb->s_id, f2fs_proc_root);
1845 f2fs_destroy_stats(sbi);
1850 mutex_lock(&sbi->umount_mutex);
1851 f2fs_leave_shrinker(sbi);
1852 iput(sbi->node_inode);
1853 mutex_unlock(&sbi->umount_mutex);
1855 destroy_node_manager(sbi);
1857 destroy_segment_manager(sbi);
1860 make_bad_inode(sbi->meta_inode);
1861 iput(sbi->meta_inode);
1863 destroy_percpu_info(sbi);
1868 if (sbi->s_chksum_driver)
1869 crypto_free_shash(sbi->s_chksum_driver);
1872 /* give only one another chance */
1875 shrink_dcache_sb(sb);
1881 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1882 const char *dev_name, void *data)
1884 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1887 static void kill_f2fs_super(struct super_block *sb)
1890 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
1891 kill_block_super(sb);
1894 static struct file_system_type f2fs_fs_type = {
1895 .owner = THIS_MODULE,
1897 .mount = f2fs_mount,
1898 .kill_sb = kill_f2fs_super,
1899 .fs_flags = FS_REQUIRES_DEV,
1901 MODULE_ALIAS_FS("f2fs");
1903 static int __init init_inodecache(void)
1905 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
1906 sizeof(struct f2fs_inode_info), 0,
1907 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
1908 if (!f2fs_inode_cachep)
1913 static void destroy_inodecache(void)
1916 * Make sure all delayed rcu free inodes are flushed before we
1920 kmem_cache_destroy(f2fs_inode_cachep);
1923 static int __init init_f2fs_fs(void)
1927 f2fs_build_trace_ios();
1929 err = init_inodecache();
1932 err = create_node_manager_caches();
1934 goto free_inodecache;
1935 err = create_segment_manager_caches();
1937 goto free_node_manager_caches;
1938 err = create_checkpoint_caches();
1940 goto free_segment_manager_caches;
1941 err = create_extent_cache();
1943 goto free_checkpoint_caches;
1944 f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1947 goto free_extent_cache;
1949 #ifdef CONFIG_F2FS_FAULT_INJECTION
1950 f2fs_fault_inject.kset = f2fs_kset;
1951 f2fs_build_fault_attr(0);
1952 err = kobject_init_and_add(&f2fs_fault_inject, &f2fs_fault_ktype,
1953 NULL, "fault_injection");
1955 f2fs_fault_inject.kset = NULL;
1959 err = register_shrinker(&f2fs_shrinker_info);
1963 err = register_filesystem(&f2fs_fs_type);
1966 err = f2fs_create_root_stats();
1968 goto free_filesystem;
1969 f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1973 unregister_filesystem(&f2fs_fs_type);
1975 unregister_shrinker(&f2fs_shrinker_info);
1977 #ifdef CONFIG_F2FS_FAULT_INJECTION
1978 if (f2fs_fault_inject.kset)
1979 kobject_put(&f2fs_fault_inject);
1981 kset_unregister(f2fs_kset);
1983 destroy_extent_cache();
1984 free_checkpoint_caches:
1985 destroy_checkpoint_caches();
1986 free_segment_manager_caches:
1987 destroy_segment_manager_caches();
1988 free_node_manager_caches:
1989 destroy_node_manager_caches();
1991 destroy_inodecache();
1996 static void __exit exit_f2fs_fs(void)
1998 remove_proc_entry("fs/f2fs", NULL);
1999 f2fs_destroy_root_stats();
2000 unregister_filesystem(&f2fs_fs_type);
2001 unregister_shrinker(&f2fs_shrinker_info);
2002 #ifdef CONFIG_F2FS_FAULT_INJECTION
2003 kobject_put(&f2fs_fault_inject);
2005 kset_unregister(f2fs_kset);
2006 destroy_extent_cache();
2007 destroy_checkpoint_caches();
2008 destroy_segment_manager_caches();
2009 destroy_node_manager_caches();
2010 destroy_inodecache();
2011 f2fs_destroy_trace_ios();
2014 module_init(init_f2fs_fs)
2015 module_exit(exit_f2fs_fs)
2017 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
2018 MODULE_DESCRIPTION("Flash Friendly File System");
2019 MODULE_LICENSE("GPL");