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 if (percpu_counter_init(&fi->dirty_pages, 0, GFP_NOFS)) {
530 kmem_cache_free(f2fs_inode_cachep, fi);
534 /* Initialize f2fs-specific inode info */
535 fi->vfs_inode.i_version = 1;
536 fi->i_current_depth = 1;
538 init_rwsem(&fi->i_sem);
539 INIT_LIST_HEAD(&fi->dirty_list);
540 INIT_LIST_HEAD(&fi->inmem_pages);
541 mutex_init(&fi->inmem_lock);
543 set_inode_flag(fi, FI_NEW_INODE);
545 if (test_opt(F2FS_SB(sb), INLINE_XATTR))
546 set_inode_flag(fi, FI_INLINE_XATTR);
548 /* Will be used by directory only */
549 fi->i_dir_level = F2FS_SB(sb)->dir_level;
550 return &fi->vfs_inode;
553 static int f2fs_drop_inode(struct inode *inode)
556 * This is to avoid a deadlock condition like below.
557 * writeback_single_inode(inode)
558 * - f2fs_write_data_page
559 * - f2fs_gc -> iput -> evict
560 * - inode_wait_for_writeback(inode)
562 if (!inode_unhashed(inode) && inode->i_state & I_SYNC) {
563 if (!inode->i_nlink && !is_bad_inode(inode)) {
564 /* to avoid evict_inode call simultaneously */
565 atomic_inc(&inode->i_count);
566 spin_unlock(&inode->i_lock);
568 /* some remained atomic pages should discarded */
569 if (f2fs_is_atomic_file(inode))
570 drop_inmem_pages(inode);
572 /* should remain fi->extent_tree for writepage */
573 f2fs_destroy_extent_node(inode);
575 sb_start_intwrite(inode->i_sb);
576 i_size_write(inode, 0);
578 if (F2FS_HAS_BLOCKS(inode))
579 f2fs_truncate(inode, true);
581 sb_end_intwrite(inode->i_sb);
583 fscrypt_put_encryption_info(inode, NULL);
584 spin_lock(&inode->i_lock);
585 atomic_dec(&inode->i_count);
589 return generic_drop_inode(inode);
593 * f2fs_dirty_inode() is called from __mark_inode_dirty()
595 * We should call set_dirty_inode to write the dirty inode through write_inode.
597 static void f2fs_dirty_inode(struct inode *inode, int flags)
599 set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
602 static void f2fs_i_callback(struct rcu_head *head)
604 struct inode *inode = container_of(head, struct inode, i_rcu);
605 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
608 static void f2fs_destroy_inode(struct inode *inode)
610 percpu_counter_destroy(&F2FS_I(inode)->dirty_pages);
611 call_rcu(&inode->i_rcu, f2fs_i_callback);
614 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
618 for (i = 0; i < NR_COUNT_TYPE; i++)
619 percpu_counter_destroy(&sbi->nr_pages[i]);
620 percpu_counter_destroy(&sbi->alloc_valid_block_count);
621 percpu_counter_destroy(&sbi->total_valid_inode_count);
624 static void f2fs_put_super(struct super_block *sb)
626 struct f2fs_sb_info *sbi = F2FS_SB(sb);
629 remove_proc_entry("segment_info", sbi->s_proc);
630 remove_proc_entry("segment_bits", sbi->s_proc);
631 remove_proc_entry(sb->s_id, f2fs_proc_root);
633 kobject_del(&sbi->s_kobj);
637 /* prevent remaining shrinker jobs */
638 mutex_lock(&sbi->umount_mutex);
641 * We don't need to do checkpoint when superblock is clean.
642 * But, the previous checkpoint was not done by umount, it needs to do
643 * clean checkpoint again.
645 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
646 !is_set_ckpt_flags(F2FS_CKPT(sbi), CP_UMOUNT_FLAG)) {
647 struct cp_control cpc = {
650 write_checkpoint(sbi, &cpc);
653 /* write_checkpoint can update stat informaion */
654 f2fs_destroy_stats(sbi);
657 * normally superblock is clean, so we need to release this.
658 * In addition, EIO will skip do checkpoint, we need this as well.
660 release_ino_entry(sbi, true);
661 release_discard_addrs(sbi);
663 f2fs_leave_shrinker(sbi);
664 mutex_unlock(&sbi->umount_mutex);
666 /* our cp_error case, we can wait for any writeback page */
667 f2fs_flush_merged_bios(sbi);
669 iput(sbi->node_inode);
670 iput(sbi->meta_inode);
672 /* destroy f2fs internal modules */
673 destroy_node_manager(sbi);
674 destroy_segment_manager(sbi);
677 kobject_put(&sbi->s_kobj);
678 wait_for_completion(&sbi->s_kobj_unregister);
680 sb->s_fs_info = NULL;
681 if (sbi->s_chksum_driver)
682 crypto_free_shash(sbi->s_chksum_driver);
683 kfree(sbi->raw_super);
685 destroy_percpu_info(sbi);
689 int f2fs_sync_fs(struct super_block *sb, int sync)
691 struct f2fs_sb_info *sbi = F2FS_SB(sb);
694 trace_f2fs_sync_fs(sb, sync);
697 struct cp_control cpc;
699 cpc.reason = __get_cp_reason(sbi);
701 mutex_lock(&sbi->gc_mutex);
702 err = write_checkpoint(sbi, &cpc);
703 mutex_unlock(&sbi->gc_mutex);
705 f2fs_trace_ios(NULL, 1);
710 static int f2fs_freeze(struct super_block *sb)
714 if (f2fs_readonly(sb))
717 err = f2fs_sync_fs(sb, 1);
721 static int f2fs_unfreeze(struct super_block *sb)
726 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
728 struct super_block *sb = dentry->d_sb;
729 struct f2fs_sb_info *sbi = F2FS_SB(sb);
730 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
731 block_t total_count, user_block_count, start_count, ovp_count;
733 total_count = le64_to_cpu(sbi->raw_super->block_count);
734 user_block_count = sbi->user_block_count;
735 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
736 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
737 buf->f_type = F2FS_SUPER_MAGIC;
738 buf->f_bsize = sbi->blocksize;
740 buf->f_blocks = total_count - start_count;
741 buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
742 buf->f_bavail = user_block_count - valid_user_blocks(sbi);
744 buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
745 buf->f_ffree = buf->f_files - valid_inode_count(sbi);
747 buf->f_namelen = F2FS_NAME_LEN;
748 buf->f_fsid.val[0] = (u32)id;
749 buf->f_fsid.val[1] = (u32)(id >> 32);
754 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
756 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
758 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
759 if (test_opt(sbi, FORCE_FG_GC))
760 seq_printf(seq, ",background_gc=%s", "sync");
762 seq_printf(seq, ",background_gc=%s", "on");
764 seq_printf(seq, ",background_gc=%s", "off");
766 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
767 seq_puts(seq, ",disable_roll_forward");
768 if (test_opt(sbi, DISCARD))
769 seq_puts(seq, ",discard");
770 if (test_opt(sbi, NOHEAP))
771 seq_puts(seq, ",no_heap_alloc");
772 #ifdef CONFIG_F2FS_FS_XATTR
773 if (test_opt(sbi, XATTR_USER))
774 seq_puts(seq, ",user_xattr");
776 seq_puts(seq, ",nouser_xattr");
777 if (test_opt(sbi, INLINE_XATTR))
778 seq_puts(seq, ",inline_xattr");
780 #ifdef CONFIG_F2FS_FS_POSIX_ACL
781 if (test_opt(sbi, POSIX_ACL))
782 seq_puts(seq, ",acl");
784 seq_puts(seq, ",noacl");
786 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
787 seq_puts(seq, ",disable_ext_identify");
788 if (test_opt(sbi, INLINE_DATA))
789 seq_puts(seq, ",inline_data");
791 seq_puts(seq, ",noinline_data");
792 if (test_opt(sbi, INLINE_DENTRY))
793 seq_puts(seq, ",inline_dentry");
794 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
795 seq_puts(seq, ",flush_merge");
796 if (test_opt(sbi, NOBARRIER))
797 seq_puts(seq, ",nobarrier");
798 if (test_opt(sbi, FASTBOOT))
799 seq_puts(seq, ",fastboot");
800 if (test_opt(sbi, EXTENT_CACHE))
801 seq_puts(seq, ",extent_cache");
803 seq_puts(seq, ",noextent_cache");
804 if (test_opt(sbi, DATA_FLUSH))
805 seq_puts(seq, ",data_flush");
806 seq_printf(seq, ",active_logs=%u", sbi->active_logs);
811 static int segment_info_seq_show(struct seq_file *seq, void *offset)
813 struct super_block *sb = seq->private;
814 struct f2fs_sb_info *sbi = F2FS_SB(sb);
815 unsigned int total_segs =
816 le32_to_cpu(sbi->raw_super->segment_count_main);
819 seq_puts(seq, "format: segment_type|valid_blocks\n"
820 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
822 for (i = 0; i < total_segs; i++) {
823 struct seg_entry *se = get_seg_entry(sbi, i);
826 seq_printf(seq, "%-10d", i);
827 seq_printf(seq, "%d|%-3u", se->type,
828 get_valid_blocks(sbi, i, 1));
829 if ((i % 10) == 9 || i == (total_segs - 1))
838 static int segment_bits_seq_show(struct seq_file *seq, void *offset)
840 struct super_block *sb = seq->private;
841 struct f2fs_sb_info *sbi = F2FS_SB(sb);
842 unsigned int total_segs =
843 le32_to_cpu(sbi->raw_super->segment_count_main);
846 seq_puts(seq, "format: segment_type|valid_blocks|bitmaps\n"
847 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
849 for (i = 0; i < total_segs; i++) {
850 struct seg_entry *se = get_seg_entry(sbi, i);
852 seq_printf(seq, "%-10d", i);
853 seq_printf(seq, "%d|%-3u|", se->type,
854 get_valid_blocks(sbi, i, 1));
855 for (j = 0; j < SIT_VBLOCK_MAP_SIZE; j++)
856 seq_printf(seq, "%x ", se->cur_valid_map[j]);
862 #define F2FS_PROC_FILE_DEF(_name) \
863 static int _name##_open_fs(struct inode *inode, struct file *file) \
865 return single_open(file, _name##_seq_show, PDE_DATA(inode)); \
868 static const struct file_operations f2fs_seq_##_name##_fops = { \
869 .open = _name##_open_fs, \
871 .llseek = seq_lseek, \
872 .release = single_release, \
875 F2FS_PROC_FILE_DEF(segment_info);
876 F2FS_PROC_FILE_DEF(segment_bits);
878 static void default_options(struct f2fs_sb_info *sbi)
880 /* init some FS parameters */
881 sbi->active_logs = NR_CURSEG_TYPE;
884 set_opt(sbi, INLINE_DATA);
885 set_opt(sbi, EXTENT_CACHE);
887 #ifdef CONFIG_F2FS_FS_XATTR
888 set_opt(sbi, XATTR_USER);
890 #ifdef CONFIG_F2FS_FS_POSIX_ACL
891 set_opt(sbi, POSIX_ACL);
895 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
897 struct f2fs_sb_info *sbi = F2FS_SB(sb);
898 struct f2fs_mount_info org_mount_opt;
899 int err, active_logs;
900 bool need_restart_gc = false;
901 bool need_stop_gc = false;
902 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
905 * Save the old mount options in case we
906 * need to restore them.
908 org_mount_opt = sbi->mount_opt;
909 active_logs = sbi->active_logs;
911 /* recover superblocks we couldn't write due to previous RO mount */
912 if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
913 err = f2fs_commit_super(sbi, false);
914 f2fs_msg(sb, KERN_INFO,
915 "Try to recover all the superblocks, ret: %d", err);
917 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
920 sbi->mount_opt.opt = 0;
921 default_options(sbi);
923 /* parse mount options */
924 err = parse_options(sb, data);
929 * Previous and new state of filesystem is RO,
930 * so skip checking GC and FLUSH_MERGE conditions.
932 if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
935 /* disallow enable/disable extent_cache dynamically */
936 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
938 f2fs_msg(sbi->sb, KERN_WARNING,
939 "switch extent_cache option is not allowed");
944 * We stop the GC thread if FS is mounted as RO
945 * or if background_gc = off is passed in mount
946 * option. Also sync the filesystem.
948 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
949 if (sbi->gc_thread) {
951 need_restart_gc = true;
953 } else if (!sbi->gc_thread) {
954 err = start_gc_thread(sbi);
960 if (*flags & MS_RDONLY) {
961 writeback_inodes_sb(sb, WB_REASON_SYNC);
964 set_sbi_flag(sbi, SBI_IS_DIRTY);
965 set_sbi_flag(sbi, SBI_IS_CLOSE);
967 clear_sbi_flag(sbi, SBI_IS_CLOSE);
971 * We stop issue flush thread if FS is mounted as RO
972 * or if flush_merge is not passed in mount option.
974 if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
975 destroy_flush_cmd_control(sbi);
976 } else if (!SM_I(sbi)->cmd_control_info) {
977 err = create_flush_cmd_control(sbi);
982 /* Update the POSIXACL Flag */
983 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
984 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
988 if (need_restart_gc) {
989 if (start_gc_thread(sbi))
990 f2fs_msg(sbi->sb, KERN_WARNING,
991 "background gc thread has stopped");
992 } else if (need_stop_gc) {
996 sbi->mount_opt = org_mount_opt;
997 sbi->active_logs = active_logs;
1001 static struct super_operations f2fs_sops = {
1002 .alloc_inode = f2fs_alloc_inode,
1003 .drop_inode = f2fs_drop_inode,
1004 .destroy_inode = f2fs_destroy_inode,
1005 .write_inode = f2fs_write_inode,
1006 .dirty_inode = f2fs_dirty_inode,
1007 .show_options = f2fs_show_options,
1008 .evict_inode = f2fs_evict_inode,
1009 .put_super = f2fs_put_super,
1010 .sync_fs = f2fs_sync_fs,
1011 .freeze_fs = f2fs_freeze,
1012 .unfreeze_fs = f2fs_unfreeze,
1013 .statfs = f2fs_statfs,
1014 .remount_fs = f2fs_remount,
1017 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1018 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1020 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1021 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1025 static int f2fs_key_prefix(struct inode *inode, u8 **key)
1027 *key = F2FS_I_SB(inode)->key_prefix;
1028 return F2FS_I_SB(inode)->key_prefix_size;
1031 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1034 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1035 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1036 ctx, len, fs_data, XATTR_CREATE);
1039 static unsigned f2fs_max_namelen(struct inode *inode)
1041 return S_ISLNK(inode->i_mode) ?
1042 inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1045 static struct fscrypt_operations f2fs_cryptops = {
1046 .get_context = f2fs_get_context,
1047 .key_prefix = f2fs_key_prefix,
1048 .set_context = f2fs_set_context,
1049 .is_encrypted = f2fs_encrypted_inode,
1050 .empty_dir = f2fs_empty_dir,
1051 .max_namelen = f2fs_max_namelen,
1054 static struct fscrypt_operations f2fs_cryptops = {
1055 .is_encrypted = f2fs_encrypted_inode,
1059 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1060 u64 ino, u32 generation)
1062 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1063 struct inode *inode;
1065 if (check_nid_range(sbi, ino))
1066 return ERR_PTR(-ESTALE);
1069 * f2fs_iget isn't quite right if the inode is currently unallocated!
1070 * However f2fs_iget currently does appropriate checks to handle stale
1071 * inodes so everything is OK.
1073 inode = f2fs_iget(sb, ino);
1075 return ERR_CAST(inode);
1076 if (unlikely(generation && inode->i_generation != generation)) {
1077 /* we didn't find the right inode.. */
1079 return ERR_PTR(-ESTALE);
1084 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1085 int fh_len, int fh_type)
1087 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1088 f2fs_nfs_get_inode);
1091 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1092 int fh_len, int fh_type)
1094 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1095 f2fs_nfs_get_inode);
1098 static const struct export_operations f2fs_export_ops = {
1099 .fh_to_dentry = f2fs_fh_to_dentry,
1100 .fh_to_parent = f2fs_fh_to_parent,
1101 .get_parent = f2fs_get_parent,
1104 static loff_t max_file_blocks(void)
1106 loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
1107 loff_t leaf_count = ADDRS_PER_BLOCK;
1109 /* two direct node blocks */
1110 result += (leaf_count * 2);
1112 /* two indirect node blocks */
1113 leaf_count *= NIDS_PER_BLOCK;
1114 result += (leaf_count * 2);
1116 /* one double indirect node block */
1117 leaf_count *= NIDS_PER_BLOCK;
1118 result += leaf_count;
1123 static int __f2fs_commit_super(struct buffer_head *bh,
1124 struct f2fs_super_block *super)
1128 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
1129 set_buffer_uptodate(bh);
1130 set_buffer_dirty(bh);
1133 /* it's rare case, we can do fua all the time */
1134 return __sync_dirty_buffer(bh, WRITE_FLUSH_FUA);
1137 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
1138 struct buffer_head *bh)
1140 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1141 (bh->b_data + F2FS_SUPER_OFFSET);
1142 struct super_block *sb = sbi->sb;
1143 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
1144 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
1145 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
1146 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
1147 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
1148 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
1149 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
1150 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
1151 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
1152 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
1153 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1154 u32 segment_count = le32_to_cpu(raw_super->segment_count);
1155 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1156 u64 main_end_blkaddr = main_blkaddr +
1157 (segment_count_main << log_blocks_per_seg);
1158 u64 seg_end_blkaddr = segment0_blkaddr +
1159 (segment_count << log_blocks_per_seg);
1161 if (segment0_blkaddr != cp_blkaddr) {
1162 f2fs_msg(sb, KERN_INFO,
1163 "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1164 segment0_blkaddr, cp_blkaddr);
1168 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
1170 f2fs_msg(sb, KERN_INFO,
1171 "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1172 cp_blkaddr, sit_blkaddr,
1173 segment_count_ckpt << log_blocks_per_seg);
1177 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
1179 f2fs_msg(sb, KERN_INFO,
1180 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1181 sit_blkaddr, nat_blkaddr,
1182 segment_count_sit << log_blocks_per_seg);
1186 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
1188 f2fs_msg(sb, KERN_INFO,
1189 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1190 nat_blkaddr, ssa_blkaddr,
1191 segment_count_nat << log_blocks_per_seg);
1195 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
1197 f2fs_msg(sb, KERN_INFO,
1198 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1199 ssa_blkaddr, main_blkaddr,
1200 segment_count_ssa << log_blocks_per_seg);
1204 if (main_end_blkaddr > seg_end_blkaddr) {
1205 f2fs_msg(sb, KERN_INFO,
1206 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1209 (segment_count << log_blocks_per_seg),
1210 segment_count_main << log_blocks_per_seg);
1212 } else if (main_end_blkaddr < seg_end_blkaddr) {
1216 /* fix in-memory information all the time */
1217 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
1218 segment0_blkaddr) >> log_blocks_per_seg);
1220 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
1221 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1224 err = __f2fs_commit_super(bh, NULL);
1225 res = err ? "failed" : "done";
1227 f2fs_msg(sb, KERN_INFO,
1228 "Fix alignment : %s, start(%u) end(%u) block(%u)",
1231 (segment_count << log_blocks_per_seg),
1232 segment_count_main << log_blocks_per_seg);
1239 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
1240 struct buffer_head *bh)
1242 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1243 (bh->b_data + F2FS_SUPER_OFFSET);
1244 struct super_block *sb = sbi->sb;
1245 unsigned int blocksize;
1247 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
1248 f2fs_msg(sb, KERN_INFO,
1249 "Magic Mismatch, valid(0x%x) - read(0x%x)",
1250 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
1254 /* Currently, support only 4KB page cache size */
1255 if (F2FS_BLKSIZE != PAGE_SIZE) {
1256 f2fs_msg(sb, KERN_INFO,
1257 "Invalid page_cache_size (%lu), supports only 4KB\n",
1262 /* Currently, support only 4KB block size */
1263 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
1264 if (blocksize != F2FS_BLKSIZE) {
1265 f2fs_msg(sb, KERN_INFO,
1266 "Invalid blocksize (%u), supports only 4KB\n",
1271 /* check log blocks per segment */
1272 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
1273 f2fs_msg(sb, KERN_INFO,
1274 "Invalid log blocks per segment (%u)\n",
1275 le32_to_cpu(raw_super->log_blocks_per_seg));
1279 /* Currently, support 512/1024/2048/4096 bytes sector size */
1280 if (le32_to_cpu(raw_super->log_sectorsize) >
1281 F2FS_MAX_LOG_SECTOR_SIZE ||
1282 le32_to_cpu(raw_super->log_sectorsize) <
1283 F2FS_MIN_LOG_SECTOR_SIZE) {
1284 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
1285 le32_to_cpu(raw_super->log_sectorsize));
1288 if (le32_to_cpu(raw_super->log_sectors_per_block) +
1289 le32_to_cpu(raw_super->log_sectorsize) !=
1290 F2FS_MAX_LOG_SECTOR_SIZE) {
1291 f2fs_msg(sb, KERN_INFO,
1292 "Invalid log sectors per block(%u) log sectorsize(%u)",
1293 le32_to_cpu(raw_super->log_sectors_per_block),
1294 le32_to_cpu(raw_super->log_sectorsize));
1298 /* check reserved ino info */
1299 if (le32_to_cpu(raw_super->node_ino) != 1 ||
1300 le32_to_cpu(raw_super->meta_ino) != 2 ||
1301 le32_to_cpu(raw_super->root_ino) != 3) {
1302 f2fs_msg(sb, KERN_INFO,
1303 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1304 le32_to_cpu(raw_super->node_ino),
1305 le32_to_cpu(raw_super->meta_ino),
1306 le32_to_cpu(raw_super->root_ino));
1310 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1311 if (sanity_check_area_boundary(sbi, bh))
1317 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1319 unsigned int total, fsmeta;
1320 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1321 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1323 total = le32_to_cpu(raw_super->segment_count);
1324 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
1325 fsmeta += le32_to_cpu(raw_super->segment_count_sit);
1326 fsmeta += le32_to_cpu(raw_super->segment_count_nat);
1327 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
1328 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
1330 if (unlikely(fsmeta >= total))
1333 if (unlikely(f2fs_cp_error(sbi))) {
1334 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
1340 static void init_sb_info(struct f2fs_sb_info *sbi)
1342 struct f2fs_super_block *raw_super = sbi->raw_super;
1344 sbi->log_sectors_per_block =
1345 le32_to_cpu(raw_super->log_sectors_per_block);
1346 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
1347 sbi->blocksize = 1 << sbi->log_blocksize;
1348 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1349 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
1350 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1351 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1352 sbi->total_sections = le32_to_cpu(raw_super->section_count);
1353 sbi->total_node_count =
1354 (le32_to_cpu(raw_super->segment_count_nat) / 2)
1355 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
1356 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
1357 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
1358 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
1359 sbi->cur_victim_sec = NULL_SECNO;
1360 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
1362 sbi->dir_level = DEF_DIR_LEVEL;
1363 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
1364 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
1365 clear_sbi_flag(sbi, SBI_NEED_FSCK);
1367 INIT_LIST_HEAD(&sbi->s_list);
1368 mutex_init(&sbi->umount_mutex);
1370 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1371 memcpy(sbi->key_prefix, F2FS_KEY_DESC_PREFIX,
1372 F2FS_KEY_DESC_PREFIX_SIZE);
1373 sbi->key_prefix_size = F2FS_KEY_DESC_PREFIX_SIZE;
1377 static int init_percpu_info(struct f2fs_sb_info *sbi)
1381 for (i = 0; i < NR_COUNT_TYPE; i++) {
1382 err = percpu_counter_init(&sbi->nr_pages[i], 0, GFP_KERNEL);
1387 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
1391 return percpu_counter_init(&sbi->total_valid_inode_count, 0,
1396 * Read f2fs raw super block.
1397 * Because we have two copies of super block, so read both of them
1398 * to get the first valid one. If any one of them is broken, we pass
1399 * them recovery flag back to the caller.
1401 static int read_raw_super_block(struct f2fs_sb_info *sbi,
1402 struct f2fs_super_block **raw_super,
1403 int *valid_super_block, int *recovery)
1405 struct super_block *sb = sbi->sb;
1407 struct buffer_head *bh;
1408 struct f2fs_super_block *super;
1411 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
1415 for (block = 0; block < 2; block++) {
1416 bh = sb_bread(sb, block);
1418 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
1424 /* sanity checking of raw super */
1425 if (sanity_check_raw_super(sbi, bh)) {
1426 f2fs_msg(sb, KERN_ERR,
1427 "Can't find valid F2FS filesystem in %dth superblock",
1435 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
1437 *valid_super_block = block;
1443 /* Fail to read any one of the superblocks*/
1447 /* No valid superblock */
1456 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
1458 struct buffer_head *bh;
1461 if ((recover && f2fs_readonly(sbi->sb)) ||
1462 bdev_read_only(sbi->sb->s_bdev)) {
1463 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1467 /* write back-up superblock first */
1468 bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1);
1471 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1474 /* if we are in recovery path, skip writing valid superblock */
1478 /* write current valid superblock */
1479 bh = sb_getblk(sbi->sb, sbi->valid_super_block);
1482 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1487 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
1489 struct f2fs_sb_info *sbi;
1490 struct f2fs_super_block *raw_super;
1493 bool retry = true, need_fsck = false;
1494 char *options = NULL;
1495 int recovery, i, valid_super_block;
1496 struct curseg_info *seg_i;
1501 valid_super_block = -1;
1504 /* allocate memory for f2fs-specific super block info */
1505 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
1511 /* Load the checksum driver */
1512 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
1513 if (IS_ERR(sbi->s_chksum_driver)) {
1514 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
1515 err = PTR_ERR(sbi->s_chksum_driver);
1516 sbi->s_chksum_driver = NULL;
1520 /* set a block size */
1521 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
1522 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
1526 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
1531 sb->s_fs_info = sbi;
1532 default_options(sbi);
1533 /* parse mount options */
1534 options = kstrdup((const char *)data, GFP_KERNEL);
1535 if (data && !options) {
1540 err = parse_options(sb, options);
1544 sbi->max_file_blocks = max_file_blocks();
1545 sb->s_maxbytes = sbi->max_file_blocks <<
1546 le32_to_cpu(raw_super->log_blocksize);
1547 sb->s_max_links = F2FS_LINK_MAX;
1548 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1550 sb->s_op = &f2fs_sops;
1551 sb->s_cop = &f2fs_cryptops;
1552 sb->s_xattr = f2fs_xattr_handlers;
1553 sb->s_export_op = &f2fs_export_ops;
1554 sb->s_magic = F2FS_SUPER_MAGIC;
1555 sb->s_time_gran = 1;
1556 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1557 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1558 memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1560 /* init f2fs-specific super block info */
1561 sbi->raw_super = raw_super;
1562 sbi->valid_super_block = valid_super_block;
1563 mutex_init(&sbi->gc_mutex);
1564 mutex_init(&sbi->writepages);
1565 mutex_init(&sbi->cp_mutex);
1566 init_rwsem(&sbi->node_write);
1568 /* disallow all the data/node/meta page writes */
1569 set_sbi_flag(sbi, SBI_POR_DOING);
1570 spin_lock_init(&sbi->stat_lock);
1572 init_rwsem(&sbi->read_io.io_rwsem);
1573 sbi->read_io.sbi = sbi;
1574 sbi->read_io.bio = NULL;
1575 for (i = 0; i < NR_PAGE_TYPE; i++) {
1576 init_rwsem(&sbi->write_io[i].io_rwsem);
1577 sbi->write_io[i].sbi = sbi;
1578 sbi->write_io[i].bio = NULL;
1581 init_rwsem(&sbi->cp_rwsem);
1582 init_waitqueue_head(&sbi->cp_wait);
1585 err = init_percpu_info(sbi);
1589 /* get an inode for meta space */
1590 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1591 if (IS_ERR(sbi->meta_inode)) {
1592 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1593 err = PTR_ERR(sbi->meta_inode);
1597 err = get_valid_checkpoint(sbi);
1599 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1600 goto free_meta_inode;
1603 sbi->total_valid_node_count =
1604 le32_to_cpu(sbi->ckpt->valid_node_count);
1605 percpu_counter_set(&sbi->total_valid_inode_count,
1606 le32_to_cpu(sbi->ckpt->valid_inode_count));
1607 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1608 sbi->total_valid_block_count =
1609 le64_to_cpu(sbi->ckpt->valid_block_count);
1610 sbi->last_valid_block_count = sbi->total_valid_block_count;
1612 for (i = 0; i < NR_INODE_TYPE; i++) {
1613 INIT_LIST_HEAD(&sbi->inode_list[i]);
1614 spin_lock_init(&sbi->inode_lock[i]);
1617 init_extent_cache_info(sbi);
1619 init_ino_entry_info(sbi);
1621 /* setup f2fs internal modules */
1622 err = build_segment_manager(sbi);
1624 f2fs_msg(sb, KERN_ERR,
1625 "Failed to initialize F2FS segment manager");
1628 err = build_node_manager(sbi);
1630 f2fs_msg(sb, KERN_ERR,
1631 "Failed to initialize F2FS node manager");
1635 /* For write statistics */
1636 if (sb->s_bdev->bd_part)
1637 sbi->sectors_written_start =
1638 (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
1640 /* Read accumulated write IO statistics if exists */
1641 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
1642 if (__exist_node_summaries(sbi))
1643 sbi->kbytes_written =
1644 le64_to_cpu(seg_i->journal->info.kbytes_written);
1646 build_gc_manager(sbi);
1648 /* get an inode for node space */
1649 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1650 if (IS_ERR(sbi->node_inode)) {
1651 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1652 err = PTR_ERR(sbi->node_inode);
1656 f2fs_join_shrinker(sbi);
1658 /* if there are nt orphan nodes free them */
1659 err = recover_orphan_inodes(sbi);
1661 goto free_node_inode;
1663 /* read root inode and dentry */
1664 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1666 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1667 err = PTR_ERR(root);
1668 goto free_node_inode;
1670 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1673 goto free_node_inode;
1676 sb->s_root = d_make_root(root); /* allocate root dentry */
1679 goto free_root_inode;
1682 err = f2fs_build_stats(sbi);
1684 goto free_root_inode;
1687 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1690 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1691 &f2fs_seq_segment_info_fops, sb);
1692 proc_create_data("segment_bits", S_IRUGO, sbi->s_proc,
1693 &f2fs_seq_segment_bits_fops, sb);
1696 sbi->s_kobj.kset = f2fs_kset;
1697 init_completion(&sbi->s_kobj_unregister);
1698 err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1703 /* recover fsynced data */
1704 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1706 * mount should be failed, when device has readonly mode, and
1707 * previous checkpoint was not done by clean system shutdown.
1709 if (bdev_read_only(sb->s_bdev) &&
1710 !is_set_ckpt_flags(sbi->ckpt, CP_UMOUNT_FLAG)) {
1716 set_sbi_flag(sbi, SBI_NEED_FSCK);
1718 err = recover_fsync_data(sbi, false);
1721 f2fs_msg(sb, KERN_ERR,
1722 "Cannot recover all fsync data errno=%d", err);
1726 err = recover_fsync_data(sbi, true);
1728 if (!f2fs_readonly(sb) && err > 0) {
1730 f2fs_msg(sb, KERN_ERR,
1731 "Need to recover fsync data");
1736 /* recover_fsync_data() cleared this already */
1737 clear_sbi_flag(sbi, SBI_POR_DOING);
1740 * If filesystem is not mounted as read-only then
1741 * do start the gc_thread.
1743 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
1744 /* After POR, we can run background GC thread.*/
1745 err = start_gc_thread(sbi);
1751 /* recover broken superblock */
1753 err = f2fs_commit_super(sbi, true);
1754 f2fs_msg(sb, KERN_INFO,
1755 "Try to recover %dth superblock, ret: %d",
1756 sbi->valid_super_block ? 1 : 2, err);
1759 f2fs_update_time(sbi, CP_TIME);
1760 f2fs_update_time(sbi, REQ_TIME);
1764 kobject_del(&sbi->s_kobj);
1765 kobject_put(&sbi->s_kobj);
1766 wait_for_completion(&sbi->s_kobj_unregister);
1769 remove_proc_entry("segment_info", sbi->s_proc);
1770 remove_proc_entry("segment_bits", sbi->s_proc);
1771 remove_proc_entry(sb->s_id, f2fs_proc_root);
1773 f2fs_destroy_stats(sbi);
1778 mutex_lock(&sbi->umount_mutex);
1779 f2fs_leave_shrinker(sbi);
1780 iput(sbi->node_inode);
1781 mutex_unlock(&sbi->umount_mutex);
1783 destroy_node_manager(sbi);
1785 destroy_segment_manager(sbi);
1788 make_bad_inode(sbi->meta_inode);
1789 iput(sbi->meta_inode);
1791 destroy_percpu_info(sbi);
1796 if (sbi->s_chksum_driver)
1797 crypto_free_shash(sbi->s_chksum_driver);
1800 /* give only one another chance */
1803 shrink_dcache_sb(sb);
1809 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1810 const char *dev_name, void *data)
1812 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1815 static void kill_f2fs_super(struct super_block *sb)
1818 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
1819 kill_block_super(sb);
1822 static struct file_system_type f2fs_fs_type = {
1823 .owner = THIS_MODULE,
1825 .mount = f2fs_mount,
1826 .kill_sb = kill_f2fs_super,
1827 .fs_flags = FS_REQUIRES_DEV,
1829 MODULE_ALIAS_FS("f2fs");
1831 static int __init init_inodecache(void)
1833 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
1834 sizeof(struct f2fs_inode_info), 0,
1835 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
1836 if (!f2fs_inode_cachep)
1841 static void destroy_inodecache(void)
1844 * Make sure all delayed rcu free inodes are flushed before we
1848 kmem_cache_destroy(f2fs_inode_cachep);
1851 static int __init init_f2fs_fs(void)
1855 f2fs_build_trace_ios();
1857 err = init_inodecache();
1860 err = create_node_manager_caches();
1862 goto free_inodecache;
1863 err = create_segment_manager_caches();
1865 goto free_node_manager_caches;
1866 err = create_checkpoint_caches();
1868 goto free_segment_manager_caches;
1869 err = create_extent_cache();
1871 goto free_checkpoint_caches;
1872 f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1875 goto free_extent_cache;
1877 #ifdef CONFIG_F2FS_FAULT_INJECTION
1878 f2fs_fault_inject.kset = f2fs_kset;
1879 f2fs_build_fault_attr(0);
1880 err = kobject_init_and_add(&f2fs_fault_inject, &f2fs_fault_ktype,
1881 NULL, "fault_injection");
1883 f2fs_fault_inject.kset = NULL;
1887 err = register_shrinker(&f2fs_shrinker_info);
1891 err = register_filesystem(&f2fs_fs_type);
1894 err = f2fs_create_root_stats();
1896 goto free_filesystem;
1897 f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1901 unregister_filesystem(&f2fs_fs_type);
1903 unregister_shrinker(&f2fs_shrinker_info);
1905 #ifdef CONFIG_F2FS_FAULT_INJECTION
1906 if (f2fs_fault_inject.kset)
1907 kobject_put(&f2fs_fault_inject);
1909 kset_unregister(f2fs_kset);
1911 destroy_extent_cache();
1912 free_checkpoint_caches:
1913 destroy_checkpoint_caches();
1914 free_segment_manager_caches:
1915 destroy_segment_manager_caches();
1916 free_node_manager_caches:
1917 destroy_node_manager_caches();
1919 destroy_inodecache();
1924 static void __exit exit_f2fs_fs(void)
1926 remove_proc_entry("fs/f2fs", NULL);
1927 f2fs_destroy_root_stats();
1928 unregister_filesystem(&f2fs_fs_type);
1929 unregister_shrinker(&f2fs_shrinker_info);
1930 #ifdef CONFIG_F2FS_FAULT_INJECTION
1931 kobject_put(&f2fs_fault_inject);
1933 kset_unregister(f2fs_kset);
1934 destroy_extent_cache();
1935 destroy_checkpoint_caches();
1936 destroy_segment_manager_caches();
1937 destroy_node_manager_caches();
1938 destroy_inodecache();
1939 f2fs_destroy_trace_ios();
1942 module_init(init_f2fs_fs)
1943 module_exit(exit_f2fs_fs)
1945 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1946 MODULE_DESCRIPTION("Flash Friendly File System");
1947 MODULE_LICENSE("GPL");