xref: /openbmc/linux/fs/f2fs/super.c (revision 26bf3dc7e25b813ff5c92234f8165941fdc12a63)
1 /*
2  * fs/f2fs/super.c
3  *
4  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5  *             http://www.samsung.com/
6  *
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.
10  */
11 #include <linux/module.h>
12 #include <linux/init.h>
13 #include <linux/fs.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>
27 
28 #include "f2fs.h"
29 #include "node.h"
30 #include "segment.h"
31 #include "xattr.h"
32 #include "gc.h"
33 #include "trace.h"
34 
35 #define CREATE_TRACE_POINTS
36 #include <trace/events/f2fs.h>
37 
38 static struct proc_dir_entry *f2fs_proc_root;
39 static struct kmem_cache *f2fs_inode_cachep;
40 static struct kset *f2fs_kset;
41 
42 enum {
43 	Opt_gc_background,
44 	Opt_disable_roll_forward,
45 	Opt_norecovery,
46 	Opt_discard,
47 	Opt_noheap,
48 	Opt_user_xattr,
49 	Opt_nouser_xattr,
50 	Opt_acl,
51 	Opt_noacl,
52 	Opt_active_logs,
53 	Opt_disable_ext_identify,
54 	Opt_inline_xattr,
55 	Opt_inline_data,
56 	Opt_inline_dentry,
57 	Opt_flush_merge,
58 	Opt_nobarrier,
59 	Opt_fastboot,
60 	Opt_extent_cache,
61 	Opt_noinline_data,
62 	Opt_err,
63 };
64 
65 static match_table_t f2fs_tokens = {
66 	{Opt_gc_background, "background_gc=%s"},
67 	{Opt_disable_roll_forward, "disable_roll_forward"},
68 	{Opt_norecovery, "norecovery"},
69 	{Opt_discard, "discard"},
70 	{Opt_noheap, "no_heap"},
71 	{Opt_user_xattr, "user_xattr"},
72 	{Opt_nouser_xattr, "nouser_xattr"},
73 	{Opt_acl, "acl"},
74 	{Opt_noacl, "noacl"},
75 	{Opt_active_logs, "active_logs=%u"},
76 	{Opt_disable_ext_identify, "disable_ext_identify"},
77 	{Opt_inline_xattr, "inline_xattr"},
78 	{Opt_inline_data, "inline_data"},
79 	{Opt_inline_dentry, "inline_dentry"},
80 	{Opt_flush_merge, "flush_merge"},
81 	{Opt_nobarrier, "nobarrier"},
82 	{Opt_fastboot, "fastboot"},
83 	{Opt_extent_cache, "extent_cache"},
84 	{Opt_noinline_data, "noinline_data"},
85 	{Opt_err, NULL},
86 };
87 
88 /* Sysfs support for f2fs */
89 enum {
90 	GC_THREAD,	/* struct f2fs_gc_thread */
91 	SM_INFO,	/* struct f2fs_sm_info */
92 	NM_INFO,	/* struct f2fs_nm_info */
93 	F2FS_SBI,	/* struct f2fs_sb_info */
94 };
95 
96 struct f2fs_attr {
97 	struct attribute attr;
98 	ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
99 	ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
100 			 const char *, size_t);
101 	int struct_type;
102 	int offset;
103 };
104 
105 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
106 {
107 	if (struct_type == GC_THREAD)
108 		return (unsigned char *)sbi->gc_thread;
109 	else if (struct_type == SM_INFO)
110 		return (unsigned char *)SM_I(sbi);
111 	else if (struct_type == NM_INFO)
112 		return (unsigned char *)NM_I(sbi);
113 	else if (struct_type == F2FS_SBI)
114 		return (unsigned char *)sbi;
115 	return NULL;
116 }
117 
118 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
119 			struct f2fs_sb_info *sbi, char *buf)
120 {
121 	unsigned char *ptr = NULL;
122 	unsigned int *ui;
123 
124 	ptr = __struct_ptr(sbi, a->struct_type);
125 	if (!ptr)
126 		return -EINVAL;
127 
128 	ui = (unsigned int *)(ptr + a->offset);
129 
130 	return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
131 }
132 
133 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
134 			struct f2fs_sb_info *sbi,
135 			const char *buf, size_t count)
136 {
137 	unsigned char *ptr;
138 	unsigned long t;
139 	unsigned int *ui;
140 	ssize_t ret;
141 
142 	ptr = __struct_ptr(sbi, a->struct_type);
143 	if (!ptr)
144 		return -EINVAL;
145 
146 	ui = (unsigned int *)(ptr + a->offset);
147 
148 	ret = kstrtoul(skip_spaces(buf), 0, &t);
149 	if (ret < 0)
150 		return ret;
151 	*ui = t;
152 	return count;
153 }
154 
155 static ssize_t f2fs_attr_show(struct kobject *kobj,
156 				struct attribute *attr, char *buf)
157 {
158 	struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
159 								s_kobj);
160 	struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
161 
162 	return a->show ? a->show(a, sbi, buf) : 0;
163 }
164 
165 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
166 						const char *buf, size_t len)
167 {
168 	struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
169 									s_kobj);
170 	struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
171 
172 	return a->store ? a->store(a, sbi, buf, len) : 0;
173 }
174 
175 static void f2fs_sb_release(struct kobject *kobj)
176 {
177 	struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
178 								s_kobj);
179 	complete(&sbi->s_kobj_unregister);
180 }
181 
182 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
183 static struct f2fs_attr f2fs_attr_##_name = {			\
184 	.attr = {.name = __stringify(_name), .mode = _mode },	\
185 	.show	= _show,					\
186 	.store	= _store,					\
187 	.struct_type = _struct_type,				\
188 	.offset = _offset					\
189 }
190 
191 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname)	\
192 	F2FS_ATTR_OFFSET(struct_type, name, 0644,		\
193 		f2fs_sbi_show, f2fs_sbi_store,			\
194 		offsetof(struct struct_name, elname))
195 
196 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
197 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
198 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
199 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
200 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
201 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
202 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, batched_trim_sections, trim_sections);
203 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
204 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
205 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks);
206 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
207 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
208 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
209 
210 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
211 static struct attribute *f2fs_attrs[] = {
212 	ATTR_LIST(gc_min_sleep_time),
213 	ATTR_LIST(gc_max_sleep_time),
214 	ATTR_LIST(gc_no_gc_sleep_time),
215 	ATTR_LIST(gc_idle),
216 	ATTR_LIST(reclaim_segments),
217 	ATTR_LIST(max_small_discards),
218 	ATTR_LIST(batched_trim_sections),
219 	ATTR_LIST(ipu_policy),
220 	ATTR_LIST(min_ipu_util),
221 	ATTR_LIST(min_fsync_blocks),
222 	ATTR_LIST(max_victim_search),
223 	ATTR_LIST(dir_level),
224 	ATTR_LIST(ram_thresh),
225 	NULL,
226 };
227 
228 static const struct sysfs_ops f2fs_attr_ops = {
229 	.show	= f2fs_attr_show,
230 	.store	= f2fs_attr_store,
231 };
232 
233 static struct kobj_type f2fs_ktype = {
234 	.default_attrs	= f2fs_attrs,
235 	.sysfs_ops	= &f2fs_attr_ops,
236 	.release	= f2fs_sb_release,
237 };
238 
239 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
240 {
241 	struct va_format vaf;
242 	va_list args;
243 
244 	va_start(args, fmt);
245 	vaf.fmt = fmt;
246 	vaf.va = &args;
247 	printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
248 	va_end(args);
249 }
250 
251 static void init_once(void *foo)
252 {
253 	struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
254 
255 	inode_init_once(&fi->vfs_inode);
256 }
257 
258 static int parse_options(struct super_block *sb, char *options)
259 {
260 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
261 	substring_t args[MAX_OPT_ARGS];
262 	char *p, *name;
263 	int arg = 0;
264 
265 	if (!options)
266 		return 0;
267 
268 	while ((p = strsep(&options, ",")) != NULL) {
269 		int token;
270 		if (!*p)
271 			continue;
272 		/*
273 		 * Initialize args struct so we know whether arg was
274 		 * found; some options take optional arguments.
275 		 */
276 		args[0].to = args[0].from = NULL;
277 		token = match_token(p, f2fs_tokens, args);
278 
279 		switch (token) {
280 		case Opt_gc_background:
281 			name = match_strdup(&args[0]);
282 
283 			if (!name)
284 				return -ENOMEM;
285 			if (strlen(name) == 2 && !strncmp(name, "on", 2))
286 				set_opt(sbi, BG_GC);
287 			else if (strlen(name) == 3 && !strncmp(name, "off", 3))
288 				clear_opt(sbi, BG_GC);
289 			else {
290 				kfree(name);
291 				return -EINVAL;
292 			}
293 			kfree(name);
294 			break;
295 		case Opt_disable_roll_forward:
296 			set_opt(sbi, DISABLE_ROLL_FORWARD);
297 			break;
298 		case Opt_norecovery:
299 			/* this option mounts f2fs with ro */
300 			set_opt(sbi, DISABLE_ROLL_FORWARD);
301 			if (!f2fs_readonly(sb))
302 				return -EINVAL;
303 			break;
304 		case Opt_discard:
305 			set_opt(sbi, DISCARD);
306 			break;
307 		case Opt_noheap:
308 			set_opt(sbi, NOHEAP);
309 			break;
310 #ifdef CONFIG_F2FS_FS_XATTR
311 		case Opt_user_xattr:
312 			set_opt(sbi, XATTR_USER);
313 			break;
314 		case Opt_nouser_xattr:
315 			clear_opt(sbi, XATTR_USER);
316 			break;
317 		case Opt_inline_xattr:
318 			set_opt(sbi, INLINE_XATTR);
319 			break;
320 #else
321 		case Opt_user_xattr:
322 			f2fs_msg(sb, KERN_INFO,
323 				"user_xattr options not supported");
324 			break;
325 		case Opt_nouser_xattr:
326 			f2fs_msg(sb, KERN_INFO,
327 				"nouser_xattr options not supported");
328 			break;
329 		case Opt_inline_xattr:
330 			f2fs_msg(sb, KERN_INFO,
331 				"inline_xattr options not supported");
332 			break;
333 #endif
334 #ifdef CONFIG_F2FS_FS_POSIX_ACL
335 		case Opt_acl:
336 			set_opt(sbi, POSIX_ACL);
337 			break;
338 		case Opt_noacl:
339 			clear_opt(sbi, POSIX_ACL);
340 			break;
341 #else
342 		case Opt_acl:
343 			f2fs_msg(sb, KERN_INFO, "acl options not supported");
344 			break;
345 		case Opt_noacl:
346 			f2fs_msg(sb, KERN_INFO, "noacl options not supported");
347 			break;
348 #endif
349 		case Opt_active_logs:
350 			if (args->from && match_int(args, &arg))
351 				return -EINVAL;
352 			if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
353 				return -EINVAL;
354 			sbi->active_logs = arg;
355 			break;
356 		case Opt_disable_ext_identify:
357 			set_opt(sbi, DISABLE_EXT_IDENTIFY);
358 			break;
359 		case Opt_inline_data:
360 			set_opt(sbi, INLINE_DATA);
361 			break;
362 		case Opt_inline_dentry:
363 			set_opt(sbi, INLINE_DENTRY);
364 			break;
365 		case Opt_flush_merge:
366 			set_opt(sbi, FLUSH_MERGE);
367 			break;
368 		case Opt_nobarrier:
369 			set_opt(sbi, NOBARRIER);
370 			break;
371 		case Opt_fastboot:
372 			set_opt(sbi, FASTBOOT);
373 			break;
374 		case Opt_extent_cache:
375 			set_opt(sbi, EXTENT_CACHE);
376 			break;
377 		case Opt_noinline_data:
378 			clear_opt(sbi, INLINE_DATA);
379 			break;
380 		default:
381 			f2fs_msg(sb, KERN_ERR,
382 				"Unrecognized mount option \"%s\" or missing value",
383 				p);
384 			return -EINVAL;
385 		}
386 	}
387 	return 0;
388 }
389 
390 static struct inode *f2fs_alloc_inode(struct super_block *sb)
391 {
392 	struct f2fs_inode_info *fi;
393 
394 	fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
395 	if (!fi)
396 		return NULL;
397 
398 	init_once((void *) fi);
399 
400 	/* Initialize f2fs-specific inode info */
401 	fi->vfs_inode.i_version = 1;
402 	atomic_set(&fi->dirty_pages, 0);
403 	fi->i_current_depth = 1;
404 	fi->i_advise = 0;
405 	rwlock_init(&fi->ext_lock);
406 	init_rwsem(&fi->i_sem);
407 	INIT_RADIX_TREE(&fi->inmem_root, GFP_NOFS);
408 	INIT_LIST_HEAD(&fi->inmem_pages);
409 	mutex_init(&fi->inmem_lock);
410 
411 	set_inode_flag(fi, FI_NEW_INODE);
412 
413 	if (test_opt(F2FS_SB(sb), INLINE_XATTR))
414 		set_inode_flag(fi, FI_INLINE_XATTR);
415 
416 	/* Will be used by directory only */
417 	fi->i_dir_level = F2FS_SB(sb)->dir_level;
418 
419 #ifdef CONFIG_F2FS_FS_ENCRYPTION
420 	fi->i_crypt_info = NULL;
421 #endif
422 	return &fi->vfs_inode;
423 }
424 
425 static int f2fs_drop_inode(struct inode *inode)
426 {
427 	/*
428 	 * This is to avoid a deadlock condition like below.
429 	 * writeback_single_inode(inode)
430 	 *  - f2fs_write_data_page
431 	 *    - f2fs_gc -> iput -> evict
432 	 *       - inode_wait_for_writeback(inode)
433 	 */
434 	if (!inode_unhashed(inode) && inode->i_state & I_SYNC) {
435 		if (!inode->i_nlink && !is_bad_inode(inode)) {
436 			spin_unlock(&inode->i_lock);
437 
438 			/* some remained atomic pages should discarded */
439 			if (f2fs_is_atomic_file(inode))
440 				commit_inmem_pages(inode, true);
441 
442 			sb_start_intwrite(inode->i_sb);
443 			i_size_write(inode, 0);
444 
445 			if (F2FS_HAS_BLOCKS(inode))
446 				f2fs_truncate(inode);
447 
448 			sb_end_intwrite(inode->i_sb);
449 
450 #ifdef CONFIG_F2FS_FS_ENCRYPTION
451 			if (F2FS_I(inode)->i_crypt_info)
452 				f2fs_free_encryption_info(inode,
453 					F2FS_I(inode)->i_crypt_info);
454 #endif
455 			spin_lock(&inode->i_lock);
456 		}
457 		return 0;
458 	}
459 	return generic_drop_inode(inode);
460 }
461 
462 /*
463  * f2fs_dirty_inode() is called from __mark_inode_dirty()
464  *
465  * We should call set_dirty_inode to write the dirty inode through write_inode.
466  */
467 static void f2fs_dirty_inode(struct inode *inode, int flags)
468 {
469 	set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
470 }
471 
472 static void f2fs_i_callback(struct rcu_head *head)
473 {
474 	struct inode *inode = container_of(head, struct inode, i_rcu);
475 	kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
476 }
477 
478 static void f2fs_destroy_inode(struct inode *inode)
479 {
480 	call_rcu(&inode->i_rcu, f2fs_i_callback);
481 }
482 
483 static void f2fs_put_super(struct super_block *sb)
484 {
485 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
486 
487 	if (sbi->s_proc) {
488 		remove_proc_entry("segment_info", sbi->s_proc);
489 		remove_proc_entry(sb->s_id, f2fs_proc_root);
490 	}
491 	kobject_del(&sbi->s_kobj);
492 
493 	f2fs_destroy_stats(sbi);
494 	stop_gc_thread(sbi);
495 
496 	/*
497 	 * We don't need to do checkpoint when superblock is clean.
498 	 * But, the previous checkpoint was not done by umount, it needs to do
499 	 * clean checkpoint again.
500 	 */
501 	if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
502 			!is_set_ckpt_flags(F2FS_CKPT(sbi), CP_UMOUNT_FLAG)) {
503 		struct cp_control cpc = {
504 			.reason = CP_UMOUNT,
505 		};
506 		write_checkpoint(sbi, &cpc);
507 	}
508 
509 	/*
510 	 * normally superblock is clean, so we need to release this.
511 	 * In addition, EIO will skip do checkpoint, we need this as well.
512 	 */
513 	release_dirty_inode(sbi);
514 	release_discard_addrs(sbi);
515 
516 	iput(sbi->node_inode);
517 	iput(sbi->meta_inode);
518 
519 	/* destroy f2fs internal modules */
520 	destroy_node_manager(sbi);
521 	destroy_segment_manager(sbi);
522 
523 	kfree(sbi->ckpt);
524 	kobject_put(&sbi->s_kobj);
525 	wait_for_completion(&sbi->s_kobj_unregister);
526 
527 	sb->s_fs_info = NULL;
528 	brelse(sbi->raw_super_buf);
529 	kfree(sbi);
530 }
531 
532 int f2fs_sync_fs(struct super_block *sb, int sync)
533 {
534 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
535 
536 	trace_f2fs_sync_fs(sb, sync);
537 
538 	if (sync) {
539 		struct cp_control cpc;
540 
541 		cpc.reason = __get_cp_reason(sbi);
542 
543 		mutex_lock(&sbi->gc_mutex);
544 		write_checkpoint(sbi, &cpc);
545 		mutex_unlock(&sbi->gc_mutex);
546 	} else {
547 		f2fs_balance_fs(sbi);
548 	}
549 	f2fs_trace_ios(NULL, 1);
550 
551 	return 0;
552 }
553 
554 static int f2fs_freeze(struct super_block *sb)
555 {
556 	int err;
557 
558 	if (f2fs_readonly(sb))
559 		return 0;
560 
561 	err = f2fs_sync_fs(sb, 1);
562 	return err;
563 }
564 
565 static int f2fs_unfreeze(struct super_block *sb)
566 {
567 	return 0;
568 }
569 
570 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
571 {
572 	struct super_block *sb = dentry->d_sb;
573 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
574 	u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
575 	block_t total_count, user_block_count, start_count, ovp_count;
576 
577 	total_count = le64_to_cpu(sbi->raw_super->block_count);
578 	user_block_count = sbi->user_block_count;
579 	start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
580 	ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
581 	buf->f_type = F2FS_SUPER_MAGIC;
582 	buf->f_bsize = sbi->blocksize;
583 
584 	buf->f_blocks = total_count - start_count;
585 	buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
586 	buf->f_bavail = user_block_count - valid_user_blocks(sbi);
587 
588 	buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
589 	buf->f_ffree = buf->f_files - valid_inode_count(sbi);
590 
591 	buf->f_namelen = F2FS_NAME_LEN;
592 	buf->f_fsid.val[0] = (u32)id;
593 	buf->f_fsid.val[1] = (u32)(id >> 32);
594 
595 	return 0;
596 }
597 
598 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
599 {
600 	struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
601 
602 	if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC))
603 		seq_printf(seq, ",background_gc=%s", "on");
604 	else
605 		seq_printf(seq, ",background_gc=%s", "off");
606 	if (test_opt(sbi, DISABLE_ROLL_FORWARD))
607 		seq_puts(seq, ",disable_roll_forward");
608 	if (test_opt(sbi, DISCARD))
609 		seq_puts(seq, ",discard");
610 	if (test_opt(sbi, NOHEAP))
611 		seq_puts(seq, ",no_heap_alloc");
612 #ifdef CONFIG_F2FS_FS_XATTR
613 	if (test_opt(sbi, XATTR_USER))
614 		seq_puts(seq, ",user_xattr");
615 	else
616 		seq_puts(seq, ",nouser_xattr");
617 	if (test_opt(sbi, INLINE_XATTR))
618 		seq_puts(seq, ",inline_xattr");
619 #endif
620 #ifdef CONFIG_F2FS_FS_POSIX_ACL
621 	if (test_opt(sbi, POSIX_ACL))
622 		seq_puts(seq, ",acl");
623 	else
624 		seq_puts(seq, ",noacl");
625 #endif
626 	if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
627 		seq_puts(seq, ",disable_ext_identify");
628 	if (test_opt(sbi, INLINE_DATA))
629 		seq_puts(seq, ",inline_data");
630 	else
631 		seq_puts(seq, ",noinline_data");
632 	if (test_opt(sbi, INLINE_DENTRY))
633 		seq_puts(seq, ",inline_dentry");
634 	if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
635 		seq_puts(seq, ",flush_merge");
636 	if (test_opt(sbi, NOBARRIER))
637 		seq_puts(seq, ",nobarrier");
638 	if (test_opt(sbi, FASTBOOT))
639 		seq_puts(seq, ",fastboot");
640 	if (test_opt(sbi, EXTENT_CACHE))
641 		seq_puts(seq, ",extent_cache");
642 	seq_printf(seq, ",active_logs=%u", sbi->active_logs);
643 
644 	return 0;
645 }
646 
647 static int segment_info_seq_show(struct seq_file *seq, void *offset)
648 {
649 	struct super_block *sb = seq->private;
650 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
651 	unsigned int total_segs =
652 			le32_to_cpu(sbi->raw_super->segment_count_main);
653 	int i;
654 
655 	seq_puts(seq, "format: segment_type|valid_blocks\n"
656 		"segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
657 
658 	for (i = 0; i < total_segs; i++) {
659 		struct seg_entry *se = get_seg_entry(sbi, i);
660 
661 		if ((i % 10) == 0)
662 			seq_printf(seq, "%-5d", i);
663 		seq_printf(seq, "%d|%-3u", se->type,
664 					get_valid_blocks(sbi, i, 1));
665 		if ((i % 10) == 9 || i == (total_segs - 1))
666 			seq_putc(seq, '\n');
667 		else
668 			seq_putc(seq, ' ');
669 	}
670 
671 	return 0;
672 }
673 
674 static int segment_info_open_fs(struct inode *inode, struct file *file)
675 {
676 	return single_open(file, segment_info_seq_show, PDE_DATA(inode));
677 }
678 
679 static const struct file_operations f2fs_seq_segment_info_fops = {
680 	.owner = THIS_MODULE,
681 	.open = segment_info_open_fs,
682 	.read = seq_read,
683 	.llseek = seq_lseek,
684 	.release = single_release,
685 };
686 
687 static void default_options(struct f2fs_sb_info *sbi)
688 {
689 	/* init some FS parameters */
690 	sbi->active_logs = NR_CURSEG_TYPE;
691 
692 	set_opt(sbi, BG_GC);
693 	set_opt(sbi, INLINE_DATA);
694 
695 #ifdef CONFIG_F2FS_FS_XATTR
696 	set_opt(sbi, XATTR_USER);
697 #endif
698 #ifdef CONFIG_F2FS_FS_POSIX_ACL
699 	set_opt(sbi, POSIX_ACL);
700 #endif
701 }
702 
703 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
704 {
705 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
706 	struct f2fs_mount_info org_mount_opt;
707 	int err, active_logs;
708 	bool need_restart_gc = false;
709 	bool need_stop_gc = false;
710 
711 	sync_filesystem(sb);
712 
713 	/*
714 	 * Save the old mount options in case we
715 	 * need to restore them.
716 	 */
717 	org_mount_opt = sbi->mount_opt;
718 	active_logs = sbi->active_logs;
719 
720 	sbi->mount_opt.opt = 0;
721 	default_options(sbi);
722 
723 	/* parse mount options */
724 	err = parse_options(sb, data);
725 	if (err)
726 		goto restore_opts;
727 
728 	/*
729 	 * Previous and new state of filesystem is RO,
730 	 * so skip checking GC and FLUSH_MERGE conditions.
731 	 */
732 	if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
733 		goto skip;
734 
735 	/*
736 	 * We stop the GC thread if FS is mounted as RO
737 	 * or if background_gc = off is passed in mount
738 	 * option. Also sync the filesystem.
739 	 */
740 	if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
741 		if (sbi->gc_thread) {
742 			stop_gc_thread(sbi);
743 			f2fs_sync_fs(sb, 1);
744 			need_restart_gc = true;
745 		}
746 	} else if (!sbi->gc_thread) {
747 		err = start_gc_thread(sbi);
748 		if (err)
749 			goto restore_opts;
750 		need_stop_gc = true;
751 	}
752 
753 	/*
754 	 * We stop issue flush thread if FS is mounted as RO
755 	 * or if flush_merge is not passed in mount option.
756 	 */
757 	if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
758 		destroy_flush_cmd_control(sbi);
759 	} else if (!SM_I(sbi)->cmd_control_info) {
760 		err = create_flush_cmd_control(sbi);
761 		if (err)
762 			goto restore_gc;
763 	}
764 skip:
765 	/* Update the POSIXACL Flag */
766 	 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
767 		(test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
768 	return 0;
769 restore_gc:
770 	if (need_restart_gc) {
771 		if (start_gc_thread(sbi))
772 			f2fs_msg(sbi->sb, KERN_WARNING,
773 				"background gc thread has stopped");
774 	} else if (need_stop_gc) {
775 		stop_gc_thread(sbi);
776 	}
777 restore_opts:
778 	sbi->mount_opt = org_mount_opt;
779 	sbi->active_logs = active_logs;
780 	return err;
781 }
782 
783 static struct super_operations f2fs_sops = {
784 	.alloc_inode	= f2fs_alloc_inode,
785 	.drop_inode	= f2fs_drop_inode,
786 	.destroy_inode	= f2fs_destroy_inode,
787 	.write_inode	= f2fs_write_inode,
788 	.dirty_inode	= f2fs_dirty_inode,
789 	.show_options	= f2fs_show_options,
790 	.evict_inode	= f2fs_evict_inode,
791 	.put_super	= f2fs_put_super,
792 	.sync_fs	= f2fs_sync_fs,
793 	.freeze_fs	= f2fs_freeze,
794 	.unfreeze_fs	= f2fs_unfreeze,
795 	.statfs		= f2fs_statfs,
796 	.remount_fs	= f2fs_remount,
797 };
798 
799 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
800 		u64 ino, u32 generation)
801 {
802 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
803 	struct inode *inode;
804 
805 	if (check_nid_range(sbi, ino))
806 		return ERR_PTR(-ESTALE);
807 
808 	/*
809 	 * f2fs_iget isn't quite right if the inode is currently unallocated!
810 	 * However f2fs_iget currently does appropriate checks to handle stale
811 	 * inodes so everything is OK.
812 	 */
813 	inode = f2fs_iget(sb, ino);
814 	if (IS_ERR(inode))
815 		return ERR_CAST(inode);
816 	if (unlikely(generation && inode->i_generation != generation)) {
817 		/* we didn't find the right inode.. */
818 		iput(inode);
819 		return ERR_PTR(-ESTALE);
820 	}
821 	return inode;
822 }
823 
824 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
825 		int fh_len, int fh_type)
826 {
827 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
828 				    f2fs_nfs_get_inode);
829 }
830 
831 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
832 		int fh_len, int fh_type)
833 {
834 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
835 				    f2fs_nfs_get_inode);
836 }
837 
838 static const struct export_operations f2fs_export_ops = {
839 	.fh_to_dentry = f2fs_fh_to_dentry,
840 	.fh_to_parent = f2fs_fh_to_parent,
841 	.get_parent = f2fs_get_parent,
842 };
843 
844 static loff_t max_file_size(unsigned bits)
845 {
846 	loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
847 	loff_t leaf_count = ADDRS_PER_BLOCK;
848 
849 	/* two direct node blocks */
850 	result += (leaf_count * 2);
851 
852 	/* two indirect node blocks */
853 	leaf_count *= NIDS_PER_BLOCK;
854 	result += (leaf_count * 2);
855 
856 	/* one double indirect node block */
857 	leaf_count *= NIDS_PER_BLOCK;
858 	result += leaf_count;
859 
860 	result <<= bits;
861 	return result;
862 }
863 
864 static int sanity_check_raw_super(struct super_block *sb,
865 			struct f2fs_super_block *raw_super)
866 {
867 	unsigned int blocksize;
868 
869 	if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
870 		f2fs_msg(sb, KERN_INFO,
871 			"Magic Mismatch, valid(0x%x) - read(0x%x)",
872 			F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
873 		return 1;
874 	}
875 
876 	/* Currently, support only 4KB page cache size */
877 	if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
878 		f2fs_msg(sb, KERN_INFO,
879 			"Invalid page_cache_size (%lu), supports only 4KB\n",
880 			PAGE_CACHE_SIZE);
881 		return 1;
882 	}
883 
884 	/* Currently, support only 4KB block size */
885 	blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
886 	if (blocksize != F2FS_BLKSIZE) {
887 		f2fs_msg(sb, KERN_INFO,
888 			"Invalid blocksize (%u), supports only 4KB\n",
889 			blocksize);
890 		return 1;
891 	}
892 
893 	/* Currently, support 512/1024/2048/4096 bytes sector size */
894 	if (le32_to_cpu(raw_super->log_sectorsize) >
895 				F2FS_MAX_LOG_SECTOR_SIZE ||
896 		le32_to_cpu(raw_super->log_sectorsize) <
897 				F2FS_MIN_LOG_SECTOR_SIZE) {
898 		f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
899 			le32_to_cpu(raw_super->log_sectorsize));
900 		return 1;
901 	}
902 	if (le32_to_cpu(raw_super->log_sectors_per_block) +
903 		le32_to_cpu(raw_super->log_sectorsize) !=
904 			F2FS_MAX_LOG_SECTOR_SIZE) {
905 		f2fs_msg(sb, KERN_INFO,
906 			"Invalid log sectors per block(%u) log sectorsize(%u)",
907 			le32_to_cpu(raw_super->log_sectors_per_block),
908 			le32_to_cpu(raw_super->log_sectorsize));
909 		return 1;
910 	}
911 	return 0;
912 }
913 
914 static int sanity_check_ckpt(struct f2fs_sb_info *sbi)
915 {
916 	unsigned int total, fsmeta;
917 	struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
918 	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
919 
920 	total = le32_to_cpu(raw_super->segment_count);
921 	fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
922 	fsmeta += le32_to_cpu(raw_super->segment_count_sit);
923 	fsmeta += le32_to_cpu(raw_super->segment_count_nat);
924 	fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
925 	fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
926 
927 	if (unlikely(fsmeta >= total))
928 		return 1;
929 
930 	if (unlikely(f2fs_cp_error(sbi))) {
931 		f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
932 		return 1;
933 	}
934 	return 0;
935 }
936 
937 static void init_sb_info(struct f2fs_sb_info *sbi)
938 {
939 	struct f2fs_super_block *raw_super = sbi->raw_super;
940 	int i;
941 
942 	sbi->log_sectors_per_block =
943 		le32_to_cpu(raw_super->log_sectors_per_block);
944 	sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
945 	sbi->blocksize = 1 << sbi->log_blocksize;
946 	sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
947 	sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
948 	sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
949 	sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
950 	sbi->total_sections = le32_to_cpu(raw_super->section_count);
951 	sbi->total_node_count =
952 		(le32_to_cpu(raw_super->segment_count_nat) / 2)
953 			* sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
954 	sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
955 	sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
956 	sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
957 	sbi->cur_victim_sec = NULL_SECNO;
958 	sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
959 
960 	for (i = 0; i < NR_COUNT_TYPE; i++)
961 		atomic_set(&sbi->nr_pages[i], 0);
962 
963 	sbi->dir_level = DEF_DIR_LEVEL;
964 	clear_sbi_flag(sbi, SBI_NEED_FSCK);
965 }
966 
967 /*
968  * Read f2fs raw super block.
969  * Because we have two copies of super block, so read the first one at first,
970  * if the first one is invalid, move to read the second one.
971  */
972 static int read_raw_super_block(struct super_block *sb,
973 			struct f2fs_super_block **raw_super,
974 			struct buffer_head **raw_super_buf,
975 			int *recovery)
976 {
977 	int block = 0;
978 	struct buffer_head *buffer;
979 	struct f2fs_super_block *super;
980 	int err = 0;
981 
982 retry:
983 	buffer = sb_bread(sb, block);
984 	if (!buffer) {
985 		*recovery = 1;
986 		f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
987 				block + 1);
988 		if (block == 0) {
989 			block++;
990 			goto retry;
991 		} else {
992 			err = -EIO;
993 			goto out;
994 		}
995 	}
996 
997 	super = (struct f2fs_super_block *)
998 		((char *)(buffer)->b_data + F2FS_SUPER_OFFSET);
999 
1000 	/* sanity checking of raw super */
1001 	if (sanity_check_raw_super(sb, super)) {
1002 		brelse(buffer);
1003 		*recovery = 1;
1004 		f2fs_msg(sb, KERN_ERR,
1005 			"Can't find valid F2FS filesystem in %dth superblock",
1006 								block + 1);
1007 		if (block == 0) {
1008 			block++;
1009 			goto retry;
1010 		} else {
1011 			err = -EINVAL;
1012 			goto out;
1013 		}
1014 	}
1015 
1016 	if (!*raw_super) {
1017 		*raw_super_buf = buffer;
1018 		*raw_super = super;
1019 	} else {
1020 		/* already have a valid superblock */
1021 		brelse(buffer);
1022 	}
1023 
1024 	/* check the validity of the second superblock */
1025 	if (block == 0) {
1026 		block++;
1027 		goto retry;
1028 	}
1029 
1030 out:
1031 	/* No valid superblock */
1032 	if (!*raw_super)
1033 		return err;
1034 
1035 	return 0;
1036 }
1037 
1038 int f2fs_commit_super(struct f2fs_sb_info *sbi)
1039 {
1040 	struct buffer_head *sbh = sbi->raw_super_buf;
1041 	sector_t block = sbh->b_blocknr;
1042 	int err;
1043 
1044 	/* write back-up superblock first */
1045 	sbh->b_blocknr = block ? 0 : 1;
1046 	mark_buffer_dirty(sbh);
1047 	err = sync_dirty_buffer(sbh);
1048 
1049 	sbh->b_blocknr = block;
1050 	if (err)
1051 		goto out;
1052 
1053 	/* write current valid superblock */
1054 	mark_buffer_dirty(sbh);
1055 	err = sync_dirty_buffer(sbh);
1056 out:
1057 	clear_buffer_write_io_error(sbh);
1058 	set_buffer_uptodate(sbh);
1059 	return err;
1060 }
1061 
1062 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
1063 {
1064 	struct f2fs_sb_info *sbi;
1065 	struct f2fs_super_block *raw_super;
1066 	struct buffer_head *raw_super_buf;
1067 	struct inode *root;
1068 	long err;
1069 	bool retry = true, need_fsck = false;
1070 	char *options = NULL;
1071 	int recovery, i;
1072 
1073 try_onemore:
1074 	err = -EINVAL;
1075 	raw_super = NULL;
1076 	raw_super_buf = NULL;
1077 	recovery = 0;
1078 
1079 	/* allocate memory for f2fs-specific super block info */
1080 	sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
1081 	if (!sbi)
1082 		return -ENOMEM;
1083 
1084 	/* set a block size */
1085 	if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
1086 		f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
1087 		goto free_sbi;
1088 	}
1089 
1090 	err = read_raw_super_block(sb, &raw_super, &raw_super_buf, &recovery);
1091 	if (err)
1092 		goto free_sbi;
1093 
1094 	sb->s_fs_info = sbi;
1095 	default_options(sbi);
1096 	/* parse mount options */
1097 	options = kstrdup((const char *)data, GFP_KERNEL);
1098 	if (data && !options) {
1099 		err = -ENOMEM;
1100 		goto free_sb_buf;
1101 	}
1102 
1103 	err = parse_options(sb, options);
1104 	if (err)
1105 		goto free_options;
1106 
1107 	sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
1108 	sb->s_max_links = F2FS_LINK_MAX;
1109 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1110 
1111 	sb->s_op = &f2fs_sops;
1112 	sb->s_xattr = f2fs_xattr_handlers;
1113 	sb->s_export_op = &f2fs_export_ops;
1114 	sb->s_magic = F2FS_SUPER_MAGIC;
1115 	sb->s_time_gran = 1;
1116 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1117 		(test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1118 	memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1119 
1120 	/* init f2fs-specific super block info */
1121 	sbi->sb = sb;
1122 	sbi->raw_super = raw_super;
1123 	sbi->raw_super_buf = raw_super_buf;
1124 	mutex_init(&sbi->gc_mutex);
1125 	mutex_init(&sbi->writepages);
1126 	mutex_init(&sbi->cp_mutex);
1127 	init_rwsem(&sbi->node_write);
1128 	clear_sbi_flag(sbi, SBI_POR_DOING);
1129 	spin_lock_init(&sbi->stat_lock);
1130 
1131 	init_rwsem(&sbi->read_io.io_rwsem);
1132 	sbi->read_io.sbi = sbi;
1133 	sbi->read_io.bio = NULL;
1134 	for (i = 0; i < NR_PAGE_TYPE; i++) {
1135 		init_rwsem(&sbi->write_io[i].io_rwsem);
1136 		sbi->write_io[i].sbi = sbi;
1137 		sbi->write_io[i].bio = NULL;
1138 	}
1139 
1140 	init_rwsem(&sbi->cp_rwsem);
1141 	init_waitqueue_head(&sbi->cp_wait);
1142 	init_sb_info(sbi);
1143 
1144 	/* get an inode for meta space */
1145 	sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1146 	if (IS_ERR(sbi->meta_inode)) {
1147 		f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1148 		err = PTR_ERR(sbi->meta_inode);
1149 		goto free_options;
1150 	}
1151 
1152 	err = get_valid_checkpoint(sbi);
1153 	if (err) {
1154 		f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1155 		goto free_meta_inode;
1156 	}
1157 
1158 	/* sanity checking of checkpoint */
1159 	err = -EINVAL;
1160 	if (sanity_check_ckpt(sbi)) {
1161 		f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint");
1162 		goto free_cp;
1163 	}
1164 
1165 	sbi->total_valid_node_count =
1166 				le32_to_cpu(sbi->ckpt->valid_node_count);
1167 	sbi->total_valid_inode_count =
1168 				le32_to_cpu(sbi->ckpt->valid_inode_count);
1169 	sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1170 	sbi->total_valid_block_count =
1171 				le64_to_cpu(sbi->ckpt->valid_block_count);
1172 	sbi->last_valid_block_count = sbi->total_valid_block_count;
1173 	sbi->alloc_valid_block_count = 0;
1174 	INIT_LIST_HEAD(&sbi->dir_inode_list);
1175 	spin_lock_init(&sbi->dir_inode_lock);
1176 
1177 	init_extent_cache_info(sbi);
1178 
1179 	init_ino_entry_info(sbi);
1180 
1181 	/* setup f2fs internal modules */
1182 	err = build_segment_manager(sbi);
1183 	if (err) {
1184 		f2fs_msg(sb, KERN_ERR,
1185 			"Failed to initialize F2FS segment manager");
1186 		goto free_sm;
1187 	}
1188 	err = build_node_manager(sbi);
1189 	if (err) {
1190 		f2fs_msg(sb, KERN_ERR,
1191 			"Failed to initialize F2FS node manager");
1192 		goto free_nm;
1193 	}
1194 
1195 	build_gc_manager(sbi);
1196 
1197 	/* get an inode for node space */
1198 	sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1199 	if (IS_ERR(sbi->node_inode)) {
1200 		f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1201 		err = PTR_ERR(sbi->node_inode);
1202 		goto free_nm;
1203 	}
1204 
1205 	/* if there are nt orphan nodes free them */
1206 	recover_orphan_inodes(sbi);
1207 
1208 	/* read root inode and dentry */
1209 	root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1210 	if (IS_ERR(root)) {
1211 		f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1212 		err = PTR_ERR(root);
1213 		goto free_node_inode;
1214 	}
1215 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1216 		iput(root);
1217 		err = -EINVAL;
1218 		goto free_node_inode;
1219 	}
1220 
1221 	sb->s_root = d_make_root(root); /* allocate root dentry */
1222 	if (!sb->s_root) {
1223 		err = -ENOMEM;
1224 		goto free_root_inode;
1225 	}
1226 
1227 	err = f2fs_build_stats(sbi);
1228 	if (err)
1229 		goto free_root_inode;
1230 
1231 	if (f2fs_proc_root)
1232 		sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1233 
1234 	if (sbi->s_proc)
1235 		proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1236 				 &f2fs_seq_segment_info_fops, sb);
1237 
1238 	if (test_opt(sbi, DISCARD)) {
1239 		struct request_queue *q = bdev_get_queue(sb->s_bdev);
1240 		if (!blk_queue_discard(q))
1241 			f2fs_msg(sb, KERN_WARNING,
1242 					"mounting with \"discard\" option, but "
1243 					"the device does not support discard");
1244 	}
1245 
1246 	sbi->s_kobj.kset = f2fs_kset;
1247 	init_completion(&sbi->s_kobj_unregister);
1248 	err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1249 							"%s", sb->s_id);
1250 	if (err)
1251 		goto free_proc;
1252 
1253 	/* recover fsynced data */
1254 	if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1255 		/*
1256 		 * mount should be failed, when device has readonly mode, and
1257 		 * previous checkpoint was not done by clean system shutdown.
1258 		 */
1259 		if (bdev_read_only(sb->s_bdev) &&
1260 				!is_set_ckpt_flags(sbi->ckpt, CP_UMOUNT_FLAG)) {
1261 			err = -EROFS;
1262 			goto free_kobj;
1263 		}
1264 
1265 		if (need_fsck)
1266 			set_sbi_flag(sbi, SBI_NEED_FSCK);
1267 
1268 		err = recover_fsync_data(sbi);
1269 		if (err) {
1270 			need_fsck = true;
1271 			f2fs_msg(sb, KERN_ERR,
1272 				"Cannot recover all fsync data errno=%ld", err);
1273 			goto free_kobj;
1274 		}
1275 	}
1276 
1277 	/*
1278 	 * If filesystem is not mounted as read-only then
1279 	 * do start the gc_thread.
1280 	 */
1281 	if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
1282 		/* After POR, we can run background GC thread.*/
1283 		err = start_gc_thread(sbi);
1284 		if (err)
1285 			goto free_kobj;
1286 	}
1287 	kfree(options);
1288 
1289 	/* recover broken superblock */
1290 	if (recovery && !f2fs_readonly(sb) && !bdev_read_only(sb->s_bdev)) {
1291 		f2fs_msg(sb, KERN_INFO, "Recover invalid superblock");
1292 		f2fs_commit_super(sbi);
1293 	}
1294 
1295 	return 0;
1296 
1297 free_kobj:
1298 	kobject_del(&sbi->s_kobj);
1299 free_proc:
1300 	if (sbi->s_proc) {
1301 		remove_proc_entry("segment_info", sbi->s_proc);
1302 		remove_proc_entry(sb->s_id, f2fs_proc_root);
1303 	}
1304 	f2fs_destroy_stats(sbi);
1305 free_root_inode:
1306 	dput(sb->s_root);
1307 	sb->s_root = NULL;
1308 free_node_inode:
1309 	iput(sbi->node_inode);
1310 free_nm:
1311 	destroy_node_manager(sbi);
1312 free_sm:
1313 	destroy_segment_manager(sbi);
1314 free_cp:
1315 	kfree(sbi->ckpt);
1316 free_meta_inode:
1317 	make_bad_inode(sbi->meta_inode);
1318 	iput(sbi->meta_inode);
1319 free_options:
1320 	kfree(options);
1321 free_sb_buf:
1322 	brelse(raw_super_buf);
1323 free_sbi:
1324 	kfree(sbi);
1325 
1326 	/* give only one another chance */
1327 	if (retry) {
1328 		retry = false;
1329 		shrink_dcache_sb(sb);
1330 		goto try_onemore;
1331 	}
1332 	return err;
1333 }
1334 
1335 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1336 			const char *dev_name, void *data)
1337 {
1338 	return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1339 }
1340 
1341 static void kill_f2fs_super(struct super_block *sb)
1342 {
1343 	if (sb->s_root)
1344 		set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
1345 	kill_block_super(sb);
1346 }
1347 
1348 static struct file_system_type f2fs_fs_type = {
1349 	.owner		= THIS_MODULE,
1350 	.name		= "f2fs",
1351 	.mount		= f2fs_mount,
1352 	.kill_sb	= kill_f2fs_super,
1353 	.fs_flags	= FS_REQUIRES_DEV,
1354 };
1355 MODULE_ALIAS_FS("f2fs");
1356 
1357 static int __init init_inodecache(void)
1358 {
1359 	f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
1360 			sizeof(struct f2fs_inode_info));
1361 	if (!f2fs_inode_cachep)
1362 		return -ENOMEM;
1363 	return 0;
1364 }
1365 
1366 static void destroy_inodecache(void)
1367 {
1368 	/*
1369 	 * Make sure all delayed rcu free inodes are flushed before we
1370 	 * destroy cache.
1371 	 */
1372 	rcu_barrier();
1373 	kmem_cache_destroy(f2fs_inode_cachep);
1374 }
1375 
1376 static int __init init_f2fs_fs(void)
1377 {
1378 	int err;
1379 
1380 	f2fs_build_trace_ios();
1381 
1382 	err = init_inodecache();
1383 	if (err)
1384 		goto fail;
1385 	err = create_node_manager_caches();
1386 	if (err)
1387 		goto free_inodecache;
1388 	err = create_segment_manager_caches();
1389 	if (err)
1390 		goto free_node_manager_caches;
1391 	err = create_checkpoint_caches();
1392 	if (err)
1393 		goto free_segment_manager_caches;
1394 	err = create_extent_cache();
1395 	if (err)
1396 		goto free_checkpoint_caches;
1397 	f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1398 	if (!f2fs_kset) {
1399 		err = -ENOMEM;
1400 		goto free_extent_cache;
1401 	}
1402 	err = f2fs_init_crypto();
1403 	if (err)
1404 		goto free_kset;
1405 	err = register_filesystem(&f2fs_fs_type);
1406 	if (err)
1407 		goto free_crypto;
1408 	f2fs_create_root_stats();
1409 	f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1410 	return 0;
1411 
1412 free_crypto:
1413 	f2fs_exit_crypto();
1414 free_kset:
1415 	kset_unregister(f2fs_kset);
1416 free_extent_cache:
1417 	destroy_extent_cache();
1418 free_checkpoint_caches:
1419 	destroy_checkpoint_caches();
1420 free_segment_manager_caches:
1421 	destroy_segment_manager_caches();
1422 free_node_manager_caches:
1423 	destroy_node_manager_caches();
1424 free_inodecache:
1425 	destroy_inodecache();
1426 fail:
1427 	return err;
1428 }
1429 
1430 static void __exit exit_f2fs_fs(void)
1431 {
1432 	remove_proc_entry("fs/f2fs", NULL);
1433 	f2fs_destroy_root_stats();
1434 	unregister_filesystem(&f2fs_fs_type);
1435 	f2fs_exit_crypto();
1436 	destroy_extent_cache();
1437 	destroy_checkpoint_caches();
1438 	destroy_segment_manager_caches();
1439 	destroy_node_manager_caches();
1440 	destroy_inodecache();
1441 	kset_unregister(f2fs_kset);
1442 	f2fs_destroy_trace_ios();
1443 }
1444 
1445 module_init(init_f2fs_fs)
1446 module_exit(exit_f2fs_fs)
1447 
1448 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1449 MODULE_DESCRIPTION("Flash Friendly File System");
1450 MODULE_LICENSE("GPL");
1451