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,
99 static match_table_t f2fs_tokens = {
100 {Opt_gc_background, "background_gc=%s"},
101 {Opt_disable_roll_forward, "disable_roll_forward"},
102 {Opt_norecovery, "norecovery"},
103 {Opt_discard, "discard"},
104 {Opt_noheap, "no_heap"},
105 {Opt_user_xattr, "user_xattr"},
106 {Opt_nouser_xattr, "nouser_xattr"},
108 {Opt_noacl, "noacl"},
109 {Opt_active_logs, "active_logs=%u"},
110 {Opt_disable_ext_identify, "disable_ext_identify"},
111 {Opt_inline_xattr, "inline_xattr"},
112 {Opt_inline_data, "inline_data"},
113 {Opt_inline_dentry, "inline_dentry"},
114 {Opt_flush_merge, "flush_merge"},
115 {Opt_nobarrier, "nobarrier"},
116 {Opt_fastboot, "fastboot"},
117 {Opt_extent_cache, "extent_cache"},
118 {Opt_noextent_cache, "noextent_cache"},
119 {Opt_noinline_data, "noinline_data"},
120 {Opt_data_flush, "data_flush"},
121 {Opt_fault_injection, "fault_injection=%u"},
125 /* Sysfs support for f2fs */
127 GC_THREAD, /* struct f2fs_gc_thread */
128 SM_INFO, /* struct f2fs_sm_info */
129 NM_INFO, /* struct f2fs_nm_info */
130 F2FS_SBI, /* struct f2fs_sb_info */
131 #ifdef CONFIG_F2FS_FAULT_INJECTION
132 FAULT_INFO_RATE, /* struct f2fs_fault_info */
133 FAULT_INFO_TYPE, /* struct f2fs_fault_info */
138 struct attribute attr;
139 ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
140 ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
141 const char *, size_t);
146 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
148 if (struct_type == GC_THREAD)
149 return (unsigned char *)sbi->gc_thread;
150 else if (struct_type == SM_INFO)
151 return (unsigned char *)SM_I(sbi);
152 else if (struct_type == NM_INFO)
153 return (unsigned char *)NM_I(sbi);
154 else if (struct_type == F2FS_SBI)
155 return (unsigned char *)sbi;
156 #ifdef CONFIG_F2FS_FAULT_INJECTION
157 else if (struct_type == FAULT_INFO_RATE ||
158 struct_type == FAULT_INFO_TYPE)
159 return (unsigned char *)&f2fs_fault;
164 static ssize_t lifetime_write_kbytes_show(struct f2fs_attr *a,
165 struct f2fs_sb_info *sbi, char *buf)
167 struct super_block *sb = sbi->sb;
169 if (!sb->s_bdev->bd_part)
170 return snprintf(buf, PAGE_SIZE, "0\n");
172 return snprintf(buf, PAGE_SIZE, "%llu\n",
173 (unsigned long long)(sbi->kbytes_written +
174 BD_PART_WRITTEN(sbi)));
177 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
178 struct f2fs_sb_info *sbi, char *buf)
180 unsigned char *ptr = NULL;
183 ptr = __struct_ptr(sbi, a->struct_type);
187 ui = (unsigned int *)(ptr + a->offset);
189 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
192 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
193 struct f2fs_sb_info *sbi,
194 const char *buf, size_t count)
201 ptr = __struct_ptr(sbi, a->struct_type);
205 ui = (unsigned int *)(ptr + a->offset);
207 ret = kstrtoul(skip_spaces(buf), 0, &t);
210 #ifdef CONFIG_F2FS_FAULT_INJECTION
211 if (a->struct_type == FAULT_INFO_TYPE && t >= (1 << FAULT_MAX))
218 static ssize_t f2fs_attr_show(struct kobject *kobj,
219 struct attribute *attr, char *buf)
221 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
223 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
225 return a->show ? a->show(a, sbi, buf) : 0;
228 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
229 const char *buf, size_t len)
231 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
233 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
235 return a->store ? a->store(a, sbi, buf, len) : 0;
238 static void f2fs_sb_release(struct kobject *kobj)
240 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
242 complete(&sbi->s_kobj_unregister);
245 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
246 static struct f2fs_attr f2fs_attr_##_name = { \
247 .attr = {.name = __stringify(_name), .mode = _mode }, \
250 .struct_type = _struct_type, \
254 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
255 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
256 f2fs_sbi_show, f2fs_sbi_store, \
257 offsetof(struct struct_name, elname))
259 #define F2FS_GENERAL_RO_ATTR(name) \
260 static struct f2fs_attr f2fs_attr_##name = __ATTR(name, 0444, name##_show, NULL)
262 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
263 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
264 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
265 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
266 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
267 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
268 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, batched_trim_sections, trim_sections);
269 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
270 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
271 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks);
272 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
273 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ra_nid_pages, ra_nid_pages);
274 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, dirty_nats_ratio, dirty_nats_ratio);
275 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
276 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
277 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, cp_interval, interval_time[CP_TIME]);
278 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, idle_interval, interval_time[REQ_TIME]);
279 #ifdef CONFIG_F2FS_FAULT_INJECTION
280 F2FS_RW_ATTR(FAULT_INFO_RATE, f2fs_fault_info, inject_rate, inject_rate);
281 F2FS_RW_ATTR(FAULT_INFO_TYPE, f2fs_fault_info, inject_type, inject_type);
283 F2FS_GENERAL_RO_ATTR(lifetime_write_kbytes);
285 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
286 static struct attribute *f2fs_attrs[] = {
287 ATTR_LIST(gc_min_sleep_time),
288 ATTR_LIST(gc_max_sleep_time),
289 ATTR_LIST(gc_no_gc_sleep_time),
291 ATTR_LIST(reclaim_segments),
292 ATTR_LIST(max_small_discards),
293 ATTR_LIST(batched_trim_sections),
294 ATTR_LIST(ipu_policy),
295 ATTR_LIST(min_ipu_util),
296 ATTR_LIST(min_fsync_blocks),
297 ATTR_LIST(max_victim_search),
298 ATTR_LIST(dir_level),
299 ATTR_LIST(ram_thresh),
300 ATTR_LIST(ra_nid_pages),
301 ATTR_LIST(dirty_nats_ratio),
302 ATTR_LIST(cp_interval),
303 ATTR_LIST(idle_interval),
304 ATTR_LIST(lifetime_write_kbytes),
308 static const struct sysfs_ops f2fs_attr_ops = {
309 .show = f2fs_attr_show,
310 .store = f2fs_attr_store,
313 static struct kobj_type f2fs_ktype = {
314 .default_attrs = f2fs_attrs,
315 .sysfs_ops = &f2fs_attr_ops,
316 .release = f2fs_sb_release,
319 #ifdef CONFIG_F2FS_FAULT_INJECTION
320 /* sysfs for f2fs fault injection */
321 static struct kobject f2fs_fault_inject;
323 static struct attribute *f2fs_fault_attrs[] = {
324 ATTR_LIST(inject_rate),
325 ATTR_LIST(inject_type),
329 static struct kobj_type f2fs_fault_ktype = {
330 .default_attrs = f2fs_fault_attrs,
331 .sysfs_ops = &f2fs_attr_ops,
335 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
337 struct va_format vaf;
343 printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
347 static void init_once(void *foo)
349 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
351 inode_init_once(&fi->vfs_inode);
354 static int parse_options(struct super_block *sb, char *options)
356 struct f2fs_sb_info *sbi = F2FS_SB(sb);
357 struct request_queue *q;
358 substring_t args[MAX_OPT_ARGS];
362 #ifdef CONFIG_F2FS_FAULT_INJECTION
363 f2fs_build_fault_attr(0);
369 while ((p = strsep(&options, ",")) != NULL) {
374 * Initialize args struct so we know whether arg was
375 * found; some options take optional arguments.
377 args[0].to = args[0].from = NULL;
378 token = match_token(p, f2fs_tokens, args);
381 case Opt_gc_background:
382 name = match_strdup(&args[0]);
386 if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
388 clear_opt(sbi, FORCE_FG_GC);
389 } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
390 clear_opt(sbi, BG_GC);
391 clear_opt(sbi, FORCE_FG_GC);
392 } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
394 set_opt(sbi, FORCE_FG_GC);
401 case Opt_disable_roll_forward:
402 set_opt(sbi, DISABLE_ROLL_FORWARD);
405 /* this option mounts f2fs with ro */
406 set_opt(sbi, DISABLE_ROLL_FORWARD);
407 if (!f2fs_readonly(sb))
411 q = bdev_get_queue(sb->s_bdev);
412 if (blk_queue_discard(q)) {
413 set_opt(sbi, DISCARD);
415 f2fs_msg(sb, KERN_WARNING,
416 "mounting with \"discard\" option, but "
417 "the device does not support discard");
421 set_opt(sbi, NOHEAP);
423 #ifdef CONFIG_F2FS_FS_XATTR
425 set_opt(sbi, XATTR_USER);
427 case Opt_nouser_xattr:
428 clear_opt(sbi, XATTR_USER);
430 case Opt_inline_xattr:
431 set_opt(sbi, INLINE_XATTR);
435 f2fs_msg(sb, KERN_INFO,
436 "user_xattr options not supported");
438 case Opt_nouser_xattr:
439 f2fs_msg(sb, KERN_INFO,
440 "nouser_xattr options not supported");
442 case Opt_inline_xattr:
443 f2fs_msg(sb, KERN_INFO,
444 "inline_xattr options not supported");
447 #ifdef CONFIG_F2FS_FS_POSIX_ACL
449 set_opt(sbi, POSIX_ACL);
452 clear_opt(sbi, POSIX_ACL);
456 f2fs_msg(sb, KERN_INFO, "acl options not supported");
459 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
462 case Opt_active_logs:
463 if (args->from && match_int(args, &arg))
465 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
467 sbi->active_logs = arg;
469 case Opt_disable_ext_identify:
470 set_opt(sbi, DISABLE_EXT_IDENTIFY);
472 case Opt_inline_data:
473 set_opt(sbi, INLINE_DATA);
475 case Opt_inline_dentry:
476 set_opt(sbi, INLINE_DENTRY);
478 case Opt_flush_merge:
479 set_opt(sbi, FLUSH_MERGE);
482 set_opt(sbi, NOBARRIER);
485 set_opt(sbi, FASTBOOT);
487 case Opt_extent_cache:
488 set_opt(sbi, EXTENT_CACHE);
490 case Opt_noextent_cache:
491 clear_opt(sbi, EXTENT_CACHE);
493 case Opt_noinline_data:
494 clear_opt(sbi, INLINE_DATA);
497 set_opt(sbi, DATA_FLUSH);
499 case Opt_fault_injection:
500 if (args->from && match_int(args, &arg))
502 #ifdef CONFIG_F2FS_FAULT_INJECTION
503 f2fs_build_fault_attr(arg);
505 f2fs_msg(sb, KERN_INFO,
506 "FAULT_INJECTION was not selected");
510 f2fs_msg(sb, KERN_ERR,
511 "Unrecognized mount option \"%s\" or missing value",
519 static struct inode *f2fs_alloc_inode(struct super_block *sb)
521 struct f2fs_inode_info *fi;
523 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
527 init_once((void *) fi);
529 /* Initialize f2fs-specific inode info */
530 fi->vfs_inode.i_version = 1;
531 atomic_set(&fi->dirty_pages, 0);
532 fi->i_current_depth = 1;
534 init_rwsem(&fi->i_sem);
535 INIT_LIST_HEAD(&fi->dirty_list);
536 INIT_LIST_HEAD(&fi->inmem_pages);
537 mutex_init(&fi->inmem_lock);
539 set_inode_flag(fi, FI_NEW_INODE);
541 if (test_opt(F2FS_SB(sb), INLINE_XATTR))
542 set_inode_flag(fi, FI_INLINE_XATTR);
544 /* Will be used by directory only */
545 fi->i_dir_level = F2FS_SB(sb)->dir_level;
546 return &fi->vfs_inode;
549 static int f2fs_drop_inode(struct inode *inode)
552 * This is to avoid a deadlock condition like below.
553 * writeback_single_inode(inode)
554 * - f2fs_write_data_page
555 * - f2fs_gc -> iput -> evict
556 * - inode_wait_for_writeback(inode)
558 if (!inode_unhashed(inode) && inode->i_state & I_SYNC) {
559 if (!inode->i_nlink && !is_bad_inode(inode)) {
560 /* to avoid evict_inode call simultaneously */
561 atomic_inc(&inode->i_count);
562 spin_unlock(&inode->i_lock);
564 /* some remained atomic pages should discarded */
565 if (f2fs_is_atomic_file(inode))
566 drop_inmem_pages(inode);
568 /* should remain fi->extent_tree for writepage */
569 f2fs_destroy_extent_node(inode);
571 sb_start_intwrite(inode->i_sb);
572 i_size_write(inode, 0);
574 if (F2FS_HAS_BLOCKS(inode))
575 f2fs_truncate(inode, true);
577 sb_end_intwrite(inode->i_sb);
579 fscrypt_put_encryption_info(inode, NULL);
580 spin_lock(&inode->i_lock);
581 atomic_dec(&inode->i_count);
585 return generic_drop_inode(inode);
589 * f2fs_dirty_inode() is called from __mark_inode_dirty()
591 * We should call set_dirty_inode to write the dirty inode through write_inode.
593 static void f2fs_dirty_inode(struct inode *inode, int flags)
595 set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
598 static void f2fs_i_callback(struct rcu_head *head)
600 struct inode *inode = container_of(head, struct inode, i_rcu);
601 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
604 static void f2fs_destroy_inode(struct inode *inode)
606 call_rcu(&inode->i_rcu, f2fs_i_callback);
609 static void f2fs_put_super(struct super_block *sb)
611 struct f2fs_sb_info *sbi = F2FS_SB(sb);
614 remove_proc_entry("segment_info", sbi->s_proc);
615 remove_proc_entry("segment_bits", sbi->s_proc);
616 remove_proc_entry(sb->s_id, f2fs_proc_root);
618 kobject_del(&sbi->s_kobj);
622 /* prevent remaining shrinker jobs */
623 mutex_lock(&sbi->umount_mutex);
626 * We don't need to do checkpoint when superblock is clean.
627 * But, the previous checkpoint was not done by umount, it needs to do
628 * clean checkpoint again.
630 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
631 !is_set_ckpt_flags(F2FS_CKPT(sbi), CP_UMOUNT_FLAG)) {
632 struct cp_control cpc = {
635 write_checkpoint(sbi, &cpc);
638 /* write_checkpoint can update stat informaion */
639 f2fs_destroy_stats(sbi);
642 * normally superblock is clean, so we need to release this.
643 * In addition, EIO will skip do checkpoint, we need this as well.
645 release_ino_entry(sbi, true);
646 release_discard_addrs(sbi);
648 f2fs_leave_shrinker(sbi);
649 mutex_unlock(&sbi->umount_mutex);
651 /* our cp_error case, we can wait for any writeback page */
652 if (get_pages(sbi, F2FS_WRITEBACK))
653 f2fs_flush_merged_bios(sbi);
655 iput(sbi->node_inode);
656 iput(sbi->meta_inode);
658 /* destroy f2fs internal modules */
659 destroy_node_manager(sbi);
660 destroy_segment_manager(sbi);
663 kobject_put(&sbi->s_kobj);
664 wait_for_completion(&sbi->s_kobj_unregister);
666 sb->s_fs_info = NULL;
667 if (sbi->s_chksum_driver)
668 crypto_free_shash(sbi->s_chksum_driver);
669 kfree(sbi->raw_super);
673 int f2fs_sync_fs(struct super_block *sb, int sync)
675 struct f2fs_sb_info *sbi = F2FS_SB(sb);
678 trace_f2fs_sync_fs(sb, sync);
681 struct cp_control cpc;
683 cpc.reason = __get_cp_reason(sbi);
685 mutex_lock(&sbi->gc_mutex);
686 err = write_checkpoint(sbi, &cpc);
687 mutex_unlock(&sbi->gc_mutex);
689 f2fs_trace_ios(NULL, 1);
694 static int f2fs_freeze(struct super_block *sb)
698 if (f2fs_readonly(sb))
701 err = f2fs_sync_fs(sb, 1);
705 static int f2fs_unfreeze(struct super_block *sb)
710 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
712 struct super_block *sb = dentry->d_sb;
713 struct f2fs_sb_info *sbi = F2FS_SB(sb);
714 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
715 block_t total_count, user_block_count, start_count, ovp_count;
717 total_count = le64_to_cpu(sbi->raw_super->block_count);
718 user_block_count = sbi->user_block_count;
719 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
720 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
721 buf->f_type = F2FS_SUPER_MAGIC;
722 buf->f_bsize = sbi->blocksize;
724 buf->f_blocks = total_count - start_count;
725 buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
726 buf->f_bavail = user_block_count - valid_user_blocks(sbi);
728 buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
729 buf->f_ffree = buf->f_files - valid_inode_count(sbi);
731 buf->f_namelen = F2FS_NAME_LEN;
732 buf->f_fsid.val[0] = (u32)id;
733 buf->f_fsid.val[1] = (u32)(id >> 32);
738 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
740 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
742 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
743 if (test_opt(sbi, FORCE_FG_GC))
744 seq_printf(seq, ",background_gc=%s", "sync");
746 seq_printf(seq, ",background_gc=%s", "on");
748 seq_printf(seq, ",background_gc=%s", "off");
750 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
751 seq_puts(seq, ",disable_roll_forward");
752 if (test_opt(sbi, DISCARD))
753 seq_puts(seq, ",discard");
754 if (test_opt(sbi, NOHEAP))
755 seq_puts(seq, ",no_heap_alloc");
756 #ifdef CONFIG_F2FS_FS_XATTR
757 if (test_opt(sbi, XATTR_USER))
758 seq_puts(seq, ",user_xattr");
760 seq_puts(seq, ",nouser_xattr");
761 if (test_opt(sbi, INLINE_XATTR))
762 seq_puts(seq, ",inline_xattr");
764 #ifdef CONFIG_F2FS_FS_POSIX_ACL
765 if (test_opt(sbi, POSIX_ACL))
766 seq_puts(seq, ",acl");
768 seq_puts(seq, ",noacl");
770 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
771 seq_puts(seq, ",disable_ext_identify");
772 if (test_opt(sbi, INLINE_DATA))
773 seq_puts(seq, ",inline_data");
775 seq_puts(seq, ",noinline_data");
776 if (test_opt(sbi, INLINE_DENTRY))
777 seq_puts(seq, ",inline_dentry");
778 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
779 seq_puts(seq, ",flush_merge");
780 if (test_opt(sbi, NOBARRIER))
781 seq_puts(seq, ",nobarrier");
782 if (test_opt(sbi, FASTBOOT))
783 seq_puts(seq, ",fastboot");
784 if (test_opt(sbi, EXTENT_CACHE))
785 seq_puts(seq, ",extent_cache");
787 seq_puts(seq, ",noextent_cache");
788 if (test_opt(sbi, DATA_FLUSH))
789 seq_puts(seq, ",data_flush");
790 seq_printf(seq, ",active_logs=%u", sbi->active_logs);
795 static int segment_info_seq_show(struct seq_file *seq, void *offset)
797 struct super_block *sb = seq->private;
798 struct f2fs_sb_info *sbi = F2FS_SB(sb);
799 unsigned int total_segs =
800 le32_to_cpu(sbi->raw_super->segment_count_main);
803 seq_puts(seq, "format: segment_type|valid_blocks\n"
804 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
806 for (i = 0; i < total_segs; i++) {
807 struct seg_entry *se = get_seg_entry(sbi, i);
810 seq_printf(seq, "%-10d", i);
811 seq_printf(seq, "%d|%-3u", se->type,
812 get_valid_blocks(sbi, i, 1));
813 if ((i % 10) == 9 || i == (total_segs - 1))
822 static int segment_bits_seq_show(struct seq_file *seq, void *offset)
824 struct super_block *sb = seq->private;
825 struct f2fs_sb_info *sbi = F2FS_SB(sb);
826 unsigned int total_segs =
827 le32_to_cpu(sbi->raw_super->segment_count_main);
830 seq_puts(seq, "format: segment_type|valid_blocks|bitmaps\n"
831 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
833 for (i = 0; i < total_segs; i++) {
834 struct seg_entry *se = get_seg_entry(sbi, i);
836 seq_printf(seq, "%-10d", i);
837 seq_printf(seq, "%d|%-3u|", se->type,
838 get_valid_blocks(sbi, i, 1));
839 for (j = 0; j < SIT_VBLOCK_MAP_SIZE; j++)
840 seq_printf(seq, "%x ", se->cur_valid_map[j]);
846 #define F2FS_PROC_FILE_DEF(_name) \
847 static int _name##_open_fs(struct inode *inode, struct file *file) \
849 return single_open(file, _name##_seq_show, PDE_DATA(inode)); \
852 static const struct file_operations f2fs_seq_##_name##_fops = { \
853 .owner = THIS_MODULE, \
854 .open = _name##_open_fs, \
856 .llseek = seq_lseek, \
857 .release = single_release, \
860 F2FS_PROC_FILE_DEF(segment_info);
861 F2FS_PROC_FILE_DEF(segment_bits);
863 static void default_options(struct f2fs_sb_info *sbi)
865 /* init some FS parameters */
866 sbi->active_logs = NR_CURSEG_TYPE;
869 set_opt(sbi, INLINE_DATA);
870 set_opt(sbi, EXTENT_CACHE);
872 #ifdef CONFIG_F2FS_FS_XATTR
873 set_opt(sbi, XATTR_USER);
875 #ifdef CONFIG_F2FS_FS_POSIX_ACL
876 set_opt(sbi, POSIX_ACL);
880 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
882 struct f2fs_sb_info *sbi = F2FS_SB(sb);
883 struct f2fs_mount_info org_mount_opt;
884 int err, active_logs;
885 bool need_restart_gc = false;
886 bool need_stop_gc = false;
887 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
890 * Save the old mount options in case we
891 * need to restore them.
893 org_mount_opt = sbi->mount_opt;
894 active_logs = sbi->active_logs;
896 /* recover superblocks we couldn't write due to previous RO mount */
897 if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
898 err = f2fs_commit_super(sbi, false);
899 f2fs_msg(sb, KERN_INFO,
900 "Try to recover all the superblocks, ret: %d", err);
902 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
905 sbi->mount_opt.opt = 0;
906 default_options(sbi);
908 /* parse mount options */
909 err = parse_options(sb, data);
914 * Previous and new state of filesystem is RO,
915 * so skip checking GC and FLUSH_MERGE conditions.
917 if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
920 /* disallow enable/disable extent_cache dynamically */
921 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
923 f2fs_msg(sbi->sb, KERN_WARNING,
924 "switch extent_cache option is not allowed");
929 * We stop the GC thread if FS is mounted as RO
930 * or if background_gc = off is passed in mount
931 * option. Also sync the filesystem.
933 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
934 if (sbi->gc_thread) {
936 need_restart_gc = true;
938 } else if (!sbi->gc_thread) {
939 err = start_gc_thread(sbi);
945 if (*flags & MS_RDONLY) {
946 writeback_inodes_sb(sb, WB_REASON_SYNC);
949 set_sbi_flag(sbi, SBI_IS_DIRTY);
950 set_sbi_flag(sbi, SBI_IS_CLOSE);
952 clear_sbi_flag(sbi, SBI_IS_CLOSE);
956 * We stop issue flush thread if FS is mounted as RO
957 * or if flush_merge is not passed in mount option.
959 if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
960 destroy_flush_cmd_control(sbi);
961 } else if (!SM_I(sbi)->cmd_control_info) {
962 err = create_flush_cmd_control(sbi);
967 /* Update the POSIXACL Flag */
968 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
969 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
973 if (need_restart_gc) {
974 if (start_gc_thread(sbi))
975 f2fs_msg(sbi->sb, KERN_WARNING,
976 "background gc thread has stopped");
977 } else if (need_stop_gc) {
981 sbi->mount_opt = org_mount_opt;
982 sbi->active_logs = active_logs;
986 static struct super_operations f2fs_sops = {
987 .alloc_inode = f2fs_alloc_inode,
988 .drop_inode = f2fs_drop_inode,
989 .destroy_inode = f2fs_destroy_inode,
990 .write_inode = f2fs_write_inode,
991 .dirty_inode = f2fs_dirty_inode,
992 .show_options = f2fs_show_options,
993 .evict_inode = f2fs_evict_inode,
994 .put_super = f2fs_put_super,
995 .sync_fs = f2fs_sync_fs,
996 .freeze_fs = f2fs_freeze,
997 .unfreeze_fs = f2fs_unfreeze,
998 .statfs = f2fs_statfs,
999 .remount_fs = f2fs_remount,
1002 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1003 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1005 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1006 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1010 static int f2fs_key_prefix(struct inode *inode, u8 **key)
1012 *key = F2FS_I_SB(inode)->key_prefix;
1013 return F2FS_I_SB(inode)->key_prefix_size;
1016 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1019 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1020 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1021 ctx, len, fs_data, XATTR_CREATE);
1024 static unsigned f2fs_max_namelen(struct inode *inode)
1026 return S_ISLNK(inode->i_mode) ?
1027 inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1030 static struct fscrypt_operations f2fs_cryptops = {
1031 .get_context = f2fs_get_context,
1032 .key_prefix = f2fs_key_prefix,
1033 .set_context = f2fs_set_context,
1034 .is_encrypted = f2fs_encrypted_inode,
1035 .empty_dir = f2fs_empty_dir,
1036 .max_namelen = f2fs_max_namelen,
1039 static struct fscrypt_operations f2fs_cryptops = {
1040 .is_encrypted = f2fs_encrypted_inode,
1044 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1045 u64 ino, u32 generation)
1047 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1048 struct inode *inode;
1050 if (check_nid_range(sbi, ino))
1051 return ERR_PTR(-ESTALE);
1054 * f2fs_iget isn't quite right if the inode is currently unallocated!
1055 * However f2fs_iget currently does appropriate checks to handle stale
1056 * inodes so everything is OK.
1058 inode = f2fs_iget(sb, ino);
1060 return ERR_CAST(inode);
1061 if (unlikely(generation && inode->i_generation != generation)) {
1062 /* we didn't find the right inode.. */
1064 return ERR_PTR(-ESTALE);
1069 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1070 int fh_len, int fh_type)
1072 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1073 f2fs_nfs_get_inode);
1076 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1077 int fh_len, int fh_type)
1079 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1080 f2fs_nfs_get_inode);
1083 static const struct export_operations f2fs_export_ops = {
1084 .fh_to_dentry = f2fs_fh_to_dentry,
1085 .fh_to_parent = f2fs_fh_to_parent,
1086 .get_parent = f2fs_get_parent,
1089 static loff_t max_file_blocks(void)
1091 loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
1092 loff_t leaf_count = ADDRS_PER_BLOCK;
1094 /* two direct node blocks */
1095 result += (leaf_count * 2);
1097 /* two indirect node blocks */
1098 leaf_count *= NIDS_PER_BLOCK;
1099 result += (leaf_count * 2);
1101 /* one double indirect node block */
1102 leaf_count *= NIDS_PER_BLOCK;
1103 result += leaf_count;
1108 static int __f2fs_commit_super(struct buffer_head *bh,
1109 struct f2fs_super_block *super)
1113 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
1114 set_buffer_uptodate(bh);
1115 set_buffer_dirty(bh);
1118 /* it's rare case, we can do fua all the time */
1119 return __sync_dirty_buffer(bh, WRITE_FLUSH_FUA);
1122 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
1123 struct buffer_head *bh)
1125 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1126 (bh->b_data + F2FS_SUPER_OFFSET);
1127 struct super_block *sb = sbi->sb;
1128 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
1129 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
1130 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
1131 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
1132 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
1133 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
1134 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
1135 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
1136 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
1137 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
1138 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1139 u32 segment_count = le32_to_cpu(raw_super->segment_count);
1140 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1141 u64 main_end_blkaddr = main_blkaddr +
1142 (segment_count_main << log_blocks_per_seg);
1143 u64 seg_end_blkaddr = segment0_blkaddr +
1144 (segment_count << log_blocks_per_seg);
1146 if (segment0_blkaddr != cp_blkaddr) {
1147 f2fs_msg(sb, KERN_INFO,
1148 "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1149 segment0_blkaddr, cp_blkaddr);
1153 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
1155 f2fs_msg(sb, KERN_INFO,
1156 "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1157 cp_blkaddr, sit_blkaddr,
1158 segment_count_ckpt << log_blocks_per_seg);
1162 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
1164 f2fs_msg(sb, KERN_INFO,
1165 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1166 sit_blkaddr, nat_blkaddr,
1167 segment_count_sit << log_blocks_per_seg);
1171 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
1173 f2fs_msg(sb, KERN_INFO,
1174 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1175 nat_blkaddr, ssa_blkaddr,
1176 segment_count_nat << log_blocks_per_seg);
1180 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
1182 f2fs_msg(sb, KERN_INFO,
1183 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1184 ssa_blkaddr, main_blkaddr,
1185 segment_count_ssa << log_blocks_per_seg);
1189 if (main_end_blkaddr > seg_end_blkaddr) {
1190 f2fs_msg(sb, KERN_INFO,
1191 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1194 (segment_count << log_blocks_per_seg),
1195 segment_count_main << log_blocks_per_seg);
1197 } else if (main_end_blkaddr < seg_end_blkaddr) {
1201 /* fix in-memory information all the time */
1202 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
1203 segment0_blkaddr) >> log_blocks_per_seg);
1205 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
1206 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1209 err = __f2fs_commit_super(bh, NULL);
1210 res = err ? "failed" : "done";
1212 f2fs_msg(sb, KERN_INFO,
1213 "Fix alignment : %s, start(%u) end(%u) block(%u)",
1216 (segment_count << log_blocks_per_seg),
1217 segment_count_main << log_blocks_per_seg);
1224 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
1225 struct buffer_head *bh)
1227 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1228 (bh->b_data + F2FS_SUPER_OFFSET);
1229 struct super_block *sb = sbi->sb;
1230 unsigned int blocksize;
1232 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
1233 f2fs_msg(sb, KERN_INFO,
1234 "Magic Mismatch, valid(0x%x) - read(0x%x)",
1235 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
1239 /* Currently, support only 4KB page cache size */
1240 if (F2FS_BLKSIZE != PAGE_SIZE) {
1241 f2fs_msg(sb, KERN_INFO,
1242 "Invalid page_cache_size (%lu), supports only 4KB\n",
1247 /* Currently, support only 4KB block size */
1248 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
1249 if (blocksize != F2FS_BLKSIZE) {
1250 f2fs_msg(sb, KERN_INFO,
1251 "Invalid blocksize (%u), supports only 4KB\n",
1256 /* check log blocks per segment */
1257 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
1258 f2fs_msg(sb, KERN_INFO,
1259 "Invalid log blocks per segment (%u)\n",
1260 le32_to_cpu(raw_super->log_blocks_per_seg));
1264 /* Currently, support 512/1024/2048/4096 bytes sector size */
1265 if (le32_to_cpu(raw_super->log_sectorsize) >
1266 F2FS_MAX_LOG_SECTOR_SIZE ||
1267 le32_to_cpu(raw_super->log_sectorsize) <
1268 F2FS_MIN_LOG_SECTOR_SIZE) {
1269 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
1270 le32_to_cpu(raw_super->log_sectorsize));
1273 if (le32_to_cpu(raw_super->log_sectors_per_block) +
1274 le32_to_cpu(raw_super->log_sectorsize) !=
1275 F2FS_MAX_LOG_SECTOR_SIZE) {
1276 f2fs_msg(sb, KERN_INFO,
1277 "Invalid log sectors per block(%u) log sectorsize(%u)",
1278 le32_to_cpu(raw_super->log_sectors_per_block),
1279 le32_to_cpu(raw_super->log_sectorsize));
1283 /* check reserved ino info */
1284 if (le32_to_cpu(raw_super->node_ino) != 1 ||
1285 le32_to_cpu(raw_super->meta_ino) != 2 ||
1286 le32_to_cpu(raw_super->root_ino) != 3) {
1287 f2fs_msg(sb, KERN_INFO,
1288 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1289 le32_to_cpu(raw_super->node_ino),
1290 le32_to_cpu(raw_super->meta_ino),
1291 le32_to_cpu(raw_super->root_ino));
1295 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1296 if (sanity_check_area_boundary(sbi, bh))
1302 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1304 unsigned int total, fsmeta;
1305 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1306 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1308 total = le32_to_cpu(raw_super->segment_count);
1309 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
1310 fsmeta += le32_to_cpu(raw_super->segment_count_sit);
1311 fsmeta += le32_to_cpu(raw_super->segment_count_nat);
1312 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
1313 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
1315 if (unlikely(fsmeta >= total))
1318 if (unlikely(f2fs_cp_error(sbi))) {
1319 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
1325 static void init_sb_info(struct f2fs_sb_info *sbi)
1327 struct f2fs_super_block *raw_super = sbi->raw_super;
1330 sbi->log_sectors_per_block =
1331 le32_to_cpu(raw_super->log_sectors_per_block);
1332 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
1333 sbi->blocksize = 1 << sbi->log_blocksize;
1334 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1335 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
1336 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1337 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1338 sbi->total_sections = le32_to_cpu(raw_super->section_count);
1339 sbi->total_node_count =
1340 (le32_to_cpu(raw_super->segment_count_nat) / 2)
1341 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
1342 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
1343 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
1344 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
1345 sbi->cur_victim_sec = NULL_SECNO;
1346 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
1348 for (i = 0; i < NR_COUNT_TYPE; i++)
1349 atomic_set(&sbi->nr_pages[i], 0);
1351 sbi->dir_level = DEF_DIR_LEVEL;
1352 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
1353 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
1354 clear_sbi_flag(sbi, SBI_NEED_FSCK);
1356 INIT_LIST_HEAD(&sbi->s_list);
1357 mutex_init(&sbi->umount_mutex);
1359 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1360 memcpy(sbi->key_prefix, F2FS_KEY_DESC_PREFIX,
1361 F2FS_KEY_DESC_PREFIX_SIZE);
1362 sbi->key_prefix_size = F2FS_KEY_DESC_PREFIX_SIZE;
1367 * Read f2fs raw super block.
1368 * Because we have two copies of super block, so read both of them
1369 * to get the first valid one. If any one of them is broken, we pass
1370 * them recovery flag back to the caller.
1372 static int read_raw_super_block(struct f2fs_sb_info *sbi,
1373 struct f2fs_super_block **raw_super,
1374 int *valid_super_block, int *recovery)
1376 struct super_block *sb = sbi->sb;
1378 struct buffer_head *bh;
1379 struct f2fs_super_block *super;
1382 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
1386 for (block = 0; block < 2; block++) {
1387 bh = sb_bread(sb, block);
1389 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
1395 /* sanity checking of raw super */
1396 if (sanity_check_raw_super(sbi, bh)) {
1397 f2fs_msg(sb, KERN_ERR,
1398 "Can't find valid F2FS filesystem in %dth superblock",
1406 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
1408 *valid_super_block = block;
1414 /* Fail to read any one of the superblocks*/
1418 /* No valid superblock */
1427 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
1429 struct buffer_head *bh;
1432 if ((recover && f2fs_readonly(sbi->sb)) ||
1433 bdev_read_only(sbi->sb->s_bdev)) {
1434 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1438 /* write back-up superblock first */
1439 bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1);
1442 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1445 /* if we are in recovery path, skip writing valid superblock */
1449 /* write current valid superblock */
1450 bh = sb_getblk(sbi->sb, sbi->valid_super_block);
1453 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1458 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
1460 struct f2fs_sb_info *sbi;
1461 struct f2fs_super_block *raw_super;
1464 bool retry = true, need_fsck = false;
1465 char *options = NULL;
1466 int recovery, i, valid_super_block;
1467 struct curseg_info *seg_i;
1472 valid_super_block = -1;
1475 /* allocate memory for f2fs-specific super block info */
1476 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
1482 /* Load the checksum driver */
1483 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
1484 if (IS_ERR(sbi->s_chksum_driver)) {
1485 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
1486 err = PTR_ERR(sbi->s_chksum_driver);
1487 sbi->s_chksum_driver = NULL;
1491 /* set a block size */
1492 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
1493 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
1497 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
1502 sb->s_fs_info = sbi;
1503 default_options(sbi);
1504 /* parse mount options */
1505 options = kstrdup((const char *)data, GFP_KERNEL);
1506 if (data && !options) {
1511 err = parse_options(sb, options);
1515 sbi->max_file_blocks = max_file_blocks();
1516 sb->s_maxbytes = sbi->max_file_blocks <<
1517 le32_to_cpu(raw_super->log_blocksize);
1518 sb->s_max_links = F2FS_LINK_MAX;
1519 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1521 sb->s_op = &f2fs_sops;
1522 sb->s_cop = &f2fs_cryptops;
1523 sb->s_xattr = f2fs_xattr_handlers;
1524 sb->s_export_op = &f2fs_export_ops;
1525 sb->s_magic = F2FS_SUPER_MAGIC;
1526 sb->s_time_gran = 1;
1527 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1528 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1529 memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1531 /* init f2fs-specific super block info */
1532 sbi->raw_super = raw_super;
1533 sbi->valid_super_block = valid_super_block;
1534 mutex_init(&sbi->gc_mutex);
1535 mutex_init(&sbi->writepages);
1536 mutex_init(&sbi->cp_mutex);
1537 init_rwsem(&sbi->node_write);
1539 /* disallow all the data/node/meta page writes */
1540 set_sbi_flag(sbi, SBI_POR_DOING);
1541 spin_lock_init(&sbi->stat_lock);
1543 init_rwsem(&sbi->read_io.io_rwsem);
1544 sbi->read_io.sbi = sbi;
1545 sbi->read_io.bio = NULL;
1546 for (i = 0; i < NR_PAGE_TYPE; i++) {
1547 init_rwsem(&sbi->write_io[i].io_rwsem);
1548 sbi->write_io[i].sbi = sbi;
1549 sbi->write_io[i].bio = NULL;
1552 init_rwsem(&sbi->cp_rwsem);
1553 init_waitqueue_head(&sbi->cp_wait);
1556 /* get an inode for meta space */
1557 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1558 if (IS_ERR(sbi->meta_inode)) {
1559 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1560 err = PTR_ERR(sbi->meta_inode);
1564 err = get_valid_checkpoint(sbi);
1566 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1567 goto free_meta_inode;
1570 sbi->total_valid_node_count =
1571 le32_to_cpu(sbi->ckpt->valid_node_count);
1572 sbi->total_valid_inode_count =
1573 le32_to_cpu(sbi->ckpt->valid_inode_count);
1574 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1575 sbi->total_valid_block_count =
1576 le64_to_cpu(sbi->ckpt->valid_block_count);
1577 sbi->last_valid_block_count = sbi->total_valid_block_count;
1578 sbi->alloc_valid_block_count = 0;
1579 for (i = 0; i < NR_INODE_TYPE; i++) {
1580 INIT_LIST_HEAD(&sbi->inode_list[i]);
1581 spin_lock_init(&sbi->inode_lock[i]);
1584 init_extent_cache_info(sbi);
1586 init_ino_entry_info(sbi);
1588 /* setup f2fs internal modules */
1589 err = build_segment_manager(sbi);
1591 f2fs_msg(sb, KERN_ERR,
1592 "Failed to initialize F2FS segment manager");
1595 err = build_node_manager(sbi);
1597 f2fs_msg(sb, KERN_ERR,
1598 "Failed to initialize F2FS node manager");
1602 /* For write statistics */
1603 if (sb->s_bdev->bd_part)
1604 sbi->sectors_written_start =
1605 (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
1607 /* Read accumulated write IO statistics if exists */
1608 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
1609 if (__exist_node_summaries(sbi))
1610 sbi->kbytes_written =
1611 le64_to_cpu(seg_i->journal->info.kbytes_written);
1613 build_gc_manager(sbi);
1615 /* get an inode for node space */
1616 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1617 if (IS_ERR(sbi->node_inode)) {
1618 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1619 err = PTR_ERR(sbi->node_inode);
1623 f2fs_join_shrinker(sbi);
1625 /* if there are nt orphan nodes free them */
1626 err = recover_orphan_inodes(sbi);
1628 goto free_node_inode;
1630 /* read root inode and dentry */
1631 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1633 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1634 err = PTR_ERR(root);
1635 goto free_node_inode;
1637 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1640 goto free_node_inode;
1643 sb->s_root = d_make_root(root); /* allocate root dentry */
1646 goto free_root_inode;
1649 err = f2fs_build_stats(sbi);
1651 goto free_root_inode;
1654 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1657 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1658 &f2fs_seq_segment_info_fops, sb);
1659 proc_create_data("segment_bits", S_IRUGO, sbi->s_proc,
1660 &f2fs_seq_segment_bits_fops, sb);
1663 sbi->s_kobj.kset = f2fs_kset;
1664 init_completion(&sbi->s_kobj_unregister);
1665 err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1670 /* recover fsynced data */
1671 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1673 * mount should be failed, when device has readonly mode, and
1674 * previous checkpoint was not done by clean system shutdown.
1676 if (bdev_read_only(sb->s_bdev) &&
1677 !is_set_ckpt_flags(sbi->ckpt, CP_UMOUNT_FLAG)) {
1683 set_sbi_flag(sbi, SBI_NEED_FSCK);
1685 err = recover_fsync_data(sbi, false);
1688 f2fs_msg(sb, KERN_ERR,
1689 "Cannot recover all fsync data errno=%d", err);
1693 err = recover_fsync_data(sbi, true);
1695 if (!f2fs_readonly(sb) && err > 0) {
1697 f2fs_msg(sb, KERN_ERR,
1698 "Need to recover fsync data");
1703 /* recover_fsync_data() cleared this already */
1704 clear_sbi_flag(sbi, SBI_POR_DOING);
1707 * If filesystem is not mounted as read-only then
1708 * do start the gc_thread.
1710 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
1711 /* After POR, we can run background GC thread.*/
1712 err = start_gc_thread(sbi);
1718 /* recover broken superblock */
1720 err = f2fs_commit_super(sbi, true);
1721 f2fs_msg(sb, KERN_INFO,
1722 "Try to recover %dth superblock, ret: %d",
1723 sbi->valid_super_block ? 1 : 2, err);
1726 f2fs_update_time(sbi, CP_TIME);
1727 f2fs_update_time(sbi, REQ_TIME);
1731 kobject_del(&sbi->s_kobj);
1732 kobject_put(&sbi->s_kobj);
1733 wait_for_completion(&sbi->s_kobj_unregister);
1736 remove_proc_entry("segment_info", sbi->s_proc);
1737 remove_proc_entry("segment_bits", sbi->s_proc);
1738 remove_proc_entry(sb->s_id, f2fs_proc_root);
1740 f2fs_destroy_stats(sbi);
1745 mutex_lock(&sbi->umount_mutex);
1746 f2fs_leave_shrinker(sbi);
1747 iput(sbi->node_inode);
1748 mutex_unlock(&sbi->umount_mutex);
1750 destroy_node_manager(sbi);
1752 destroy_segment_manager(sbi);
1755 make_bad_inode(sbi->meta_inode);
1756 iput(sbi->meta_inode);
1762 if (sbi->s_chksum_driver)
1763 crypto_free_shash(sbi->s_chksum_driver);
1766 /* give only one another chance */
1769 shrink_dcache_sb(sb);
1775 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1776 const char *dev_name, void *data)
1778 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1781 static void kill_f2fs_super(struct super_block *sb)
1784 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
1785 kill_block_super(sb);
1788 static struct file_system_type f2fs_fs_type = {
1789 .owner = THIS_MODULE,
1791 .mount = f2fs_mount,
1792 .kill_sb = kill_f2fs_super,
1793 .fs_flags = FS_REQUIRES_DEV,
1795 MODULE_ALIAS_FS("f2fs");
1797 static int __init init_inodecache(void)
1799 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
1800 sizeof(struct f2fs_inode_info), 0,
1801 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
1802 if (!f2fs_inode_cachep)
1807 static void destroy_inodecache(void)
1810 * Make sure all delayed rcu free inodes are flushed before we
1814 kmem_cache_destroy(f2fs_inode_cachep);
1817 static int __init init_f2fs_fs(void)
1821 f2fs_build_trace_ios();
1823 err = init_inodecache();
1826 err = create_node_manager_caches();
1828 goto free_inodecache;
1829 err = create_segment_manager_caches();
1831 goto free_node_manager_caches;
1832 err = create_checkpoint_caches();
1834 goto free_segment_manager_caches;
1835 err = create_extent_cache();
1837 goto free_checkpoint_caches;
1838 f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1841 goto free_extent_cache;
1843 #ifdef CONFIG_F2FS_FAULT_INJECTION
1844 f2fs_fault_inject.kset = f2fs_kset;
1845 f2fs_build_fault_attr(0);
1846 err = kobject_init_and_add(&f2fs_fault_inject, &f2fs_fault_ktype,
1847 NULL, "fault_injection");
1849 f2fs_fault_inject.kset = NULL;
1853 err = register_shrinker(&f2fs_shrinker_info);
1857 err = register_filesystem(&f2fs_fs_type);
1860 err = f2fs_create_root_stats();
1862 goto free_filesystem;
1863 f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1867 unregister_filesystem(&f2fs_fs_type);
1869 unregister_shrinker(&f2fs_shrinker_info);
1871 #ifdef CONFIG_F2FS_FAULT_INJECTION
1872 if (f2fs_fault_inject.kset)
1873 kobject_put(&f2fs_fault_inject);
1875 kset_unregister(f2fs_kset);
1877 destroy_extent_cache();
1878 free_checkpoint_caches:
1879 destroy_checkpoint_caches();
1880 free_segment_manager_caches:
1881 destroy_segment_manager_caches();
1882 free_node_manager_caches:
1883 destroy_node_manager_caches();
1885 destroy_inodecache();
1890 static void __exit exit_f2fs_fs(void)
1892 remove_proc_entry("fs/f2fs", NULL);
1893 f2fs_destroy_root_stats();
1894 unregister_shrinker(&f2fs_shrinker_info);
1895 unregister_filesystem(&f2fs_fs_type);
1896 destroy_extent_cache();
1897 destroy_checkpoint_caches();
1898 destroy_segment_manager_caches();
1899 destroy_node_manager_caches();
1900 destroy_inodecache();
1901 #ifdef CONFIG_F2FS_FAULT_INJECTION
1902 kobject_put(&f2fs_fault_inject);
1904 kset_unregister(f2fs_kset);
1905 f2fs_destroy_trace_ios();
1908 module_init(init_f2fs_fs)
1909 module_exit(exit_f2fs_fs)
1911 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1912 MODULE_DESCRIPTION("Flash Friendly File System");
1913 MODULE_LICENSE("GPL");