xref: /openbmc/linux/fs/f2fs/super.c (revision ca55b2fe)
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 /* f2fs-wide shrinker description */
43 static struct shrinker f2fs_shrinker_info = {
44 	.scan_objects = f2fs_shrink_scan,
45 	.count_objects = f2fs_shrink_count,
46 	.seeks = DEFAULT_SEEKS,
47 };
48 
49 enum {
50 	Opt_gc_background,
51 	Opt_disable_roll_forward,
52 	Opt_norecovery,
53 	Opt_discard,
54 	Opt_noheap,
55 	Opt_user_xattr,
56 	Opt_nouser_xattr,
57 	Opt_acl,
58 	Opt_noacl,
59 	Opt_active_logs,
60 	Opt_disable_ext_identify,
61 	Opt_inline_xattr,
62 	Opt_inline_data,
63 	Opt_inline_dentry,
64 	Opt_flush_merge,
65 	Opt_nobarrier,
66 	Opt_fastboot,
67 	Opt_extent_cache,
68 	Opt_noextent_cache,
69 	Opt_noinline_data,
70 	Opt_err,
71 };
72 
73 static match_table_t f2fs_tokens = {
74 	{Opt_gc_background, "background_gc=%s"},
75 	{Opt_disable_roll_forward, "disable_roll_forward"},
76 	{Opt_norecovery, "norecovery"},
77 	{Opt_discard, "discard"},
78 	{Opt_noheap, "no_heap"},
79 	{Opt_user_xattr, "user_xattr"},
80 	{Opt_nouser_xattr, "nouser_xattr"},
81 	{Opt_acl, "acl"},
82 	{Opt_noacl, "noacl"},
83 	{Opt_active_logs, "active_logs=%u"},
84 	{Opt_disable_ext_identify, "disable_ext_identify"},
85 	{Opt_inline_xattr, "inline_xattr"},
86 	{Opt_inline_data, "inline_data"},
87 	{Opt_inline_dentry, "inline_dentry"},
88 	{Opt_flush_merge, "flush_merge"},
89 	{Opt_nobarrier, "nobarrier"},
90 	{Opt_fastboot, "fastboot"},
91 	{Opt_extent_cache, "extent_cache"},
92 	{Opt_noextent_cache, "noextent_cache"},
93 	{Opt_noinline_data, "noinline_data"},
94 	{Opt_err, NULL},
95 };
96 
97 /* Sysfs support for f2fs */
98 enum {
99 	GC_THREAD,	/* struct f2fs_gc_thread */
100 	SM_INFO,	/* struct f2fs_sm_info */
101 	NM_INFO,	/* struct f2fs_nm_info */
102 	F2FS_SBI,	/* struct f2fs_sb_info */
103 };
104 
105 struct f2fs_attr {
106 	struct attribute attr;
107 	ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
108 	ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
109 			 const char *, size_t);
110 	int struct_type;
111 	int offset;
112 };
113 
114 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
115 {
116 	if (struct_type == GC_THREAD)
117 		return (unsigned char *)sbi->gc_thread;
118 	else if (struct_type == SM_INFO)
119 		return (unsigned char *)SM_I(sbi);
120 	else if (struct_type == NM_INFO)
121 		return (unsigned char *)NM_I(sbi);
122 	else if (struct_type == F2FS_SBI)
123 		return (unsigned char *)sbi;
124 	return NULL;
125 }
126 
127 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
128 			struct f2fs_sb_info *sbi, char *buf)
129 {
130 	unsigned char *ptr = NULL;
131 	unsigned int *ui;
132 
133 	ptr = __struct_ptr(sbi, a->struct_type);
134 	if (!ptr)
135 		return -EINVAL;
136 
137 	ui = (unsigned int *)(ptr + a->offset);
138 
139 	return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
140 }
141 
142 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
143 			struct f2fs_sb_info *sbi,
144 			const char *buf, size_t count)
145 {
146 	unsigned char *ptr;
147 	unsigned long t;
148 	unsigned int *ui;
149 	ssize_t ret;
150 
151 	ptr = __struct_ptr(sbi, a->struct_type);
152 	if (!ptr)
153 		return -EINVAL;
154 
155 	ui = (unsigned int *)(ptr + a->offset);
156 
157 	ret = kstrtoul(skip_spaces(buf), 0, &t);
158 	if (ret < 0)
159 		return ret;
160 	*ui = t;
161 	return count;
162 }
163 
164 static ssize_t f2fs_attr_show(struct kobject *kobj,
165 				struct attribute *attr, char *buf)
166 {
167 	struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
168 								s_kobj);
169 	struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
170 
171 	return a->show ? a->show(a, sbi, buf) : 0;
172 }
173 
174 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
175 						const char *buf, size_t len)
176 {
177 	struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
178 									s_kobj);
179 	struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
180 
181 	return a->store ? a->store(a, sbi, buf, len) : 0;
182 }
183 
184 static void f2fs_sb_release(struct kobject *kobj)
185 {
186 	struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
187 								s_kobj);
188 	complete(&sbi->s_kobj_unregister);
189 }
190 
191 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
192 static struct f2fs_attr f2fs_attr_##_name = {			\
193 	.attr = {.name = __stringify(_name), .mode = _mode },	\
194 	.show	= _show,					\
195 	.store	= _store,					\
196 	.struct_type = _struct_type,				\
197 	.offset = _offset					\
198 }
199 
200 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname)	\
201 	F2FS_ATTR_OFFSET(struct_type, name, 0644,		\
202 		f2fs_sbi_show, f2fs_sbi_store,			\
203 		offsetof(struct struct_name, elname))
204 
205 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
206 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
207 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
208 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
209 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
210 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
211 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, batched_trim_sections, trim_sections);
212 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
213 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
214 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks);
215 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
216 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
217 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
218 
219 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
220 static struct attribute *f2fs_attrs[] = {
221 	ATTR_LIST(gc_min_sleep_time),
222 	ATTR_LIST(gc_max_sleep_time),
223 	ATTR_LIST(gc_no_gc_sleep_time),
224 	ATTR_LIST(gc_idle),
225 	ATTR_LIST(reclaim_segments),
226 	ATTR_LIST(max_small_discards),
227 	ATTR_LIST(batched_trim_sections),
228 	ATTR_LIST(ipu_policy),
229 	ATTR_LIST(min_ipu_util),
230 	ATTR_LIST(min_fsync_blocks),
231 	ATTR_LIST(max_victim_search),
232 	ATTR_LIST(dir_level),
233 	ATTR_LIST(ram_thresh),
234 	NULL,
235 };
236 
237 static const struct sysfs_ops f2fs_attr_ops = {
238 	.show	= f2fs_attr_show,
239 	.store	= f2fs_attr_store,
240 };
241 
242 static struct kobj_type f2fs_ktype = {
243 	.default_attrs	= f2fs_attrs,
244 	.sysfs_ops	= &f2fs_attr_ops,
245 	.release	= f2fs_sb_release,
246 };
247 
248 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
249 {
250 	struct va_format vaf;
251 	va_list args;
252 
253 	va_start(args, fmt);
254 	vaf.fmt = fmt;
255 	vaf.va = &args;
256 	printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
257 	va_end(args);
258 }
259 
260 static void init_once(void *foo)
261 {
262 	struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
263 
264 	inode_init_once(&fi->vfs_inode);
265 }
266 
267 static int parse_options(struct super_block *sb, char *options)
268 {
269 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
270 	struct request_queue *q;
271 	substring_t args[MAX_OPT_ARGS];
272 	char *p, *name;
273 	int arg = 0;
274 
275 	if (!options)
276 		return 0;
277 
278 	while ((p = strsep(&options, ",")) != NULL) {
279 		int token;
280 		if (!*p)
281 			continue;
282 		/*
283 		 * Initialize args struct so we know whether arg was
284 		 * found; some options take optional arguments.
285 		 */
286 		args[0].to = args[0].from = NULL;
287 		token = match_token(p, f2fs_tokens, args);
288 
289 		switch (token) {
290 		case Opt_gc_background:
291 			name = match_strdup(&args[0]);
292 
293 			if (!name)
294 				return -ENOMEM;
295 			if (strlen(name) == 2 && !strncmp(name, "on", 2))
296 				set_opt(sbi, BG_GC);
297 			else if (strlen(name) == 3 && !strncmp(name, "off", 3))
298 				clear_opt(sbi, BG_GC);
299 			else {
300 				kfree(name);
301 				return -EINVAL;
302 			}
303 			kfree(name);
304 			break;
305 		case Opt_disable_roll_forward:
306 			set_opt(sbi, DISABLE_ROLL_FORWARD);
307 			break;
308 		case Opt_norecovery:
309 			/* this option mounts f2fs with ro */
310 			set_opt(sbi, DISABLE_ROLL_FORWARD);
311 			if (!f2fs_readonly(sb))
312 				return -EINVAL;
313 			break;
314 		case Opt_discard:
315 			q = bdev_get_queue(sb->s_bdev);
316 			if (blk_queue_discard(q)) {
317 				set_opt(sbi, DISCARD);
318 			} else {
319 				f2fs_msg(sb, KERN_WARNING,
320 					"mounting with \"discard\" option, but "
321 					"the device does not support discard");
322 			}
323 			break;
324 		case Opt_noheap:
325 			set_opt(sbi, NOHEAP);
326 			break;
327 #ifdef CONFIG_F2FS_FS_XATTR
328 		case Opt_user_xattr:
329 			set_opt(sbi, XATTR_USER);
330 			break;
331 		case Opt_nouser_xattr:
332 			clear_opt(sbi, XATTR_USER);
333 			break;
334 		case Opt_inline_xattr:
335 			set_opt(sbi, INLINE_XATTR);
336 			break;
337 #else
338 		case Opt_user_xattr:
339 			f2fs_msg(sb, KERN_INFO,
340 				"user_xattr options not supported");
341 			break;
342 		case Opt_nouser_xattr:
343 			f2fs_msg(sb, KERN_INFO,
344 				"nouser_xattr options not supported");
345 			break;
346 		case Opt_inline_xattr:
347 			f2fs_msg(sb, KERN_INFO,
348 				"inline_xattr options not supported");
349 			break;
350 #endif
351 #ifdef CONFIG_F2FS_FS_POSIX_ACL
352 		case Opt_acl:
353 			set_opt(sbi, POSIX_ACL);
354 			break;
355 		case Opt_noacl:
356 			clear_opt(sbi, POSIX_ACL);
357 			break;
358 #else
359 		case Opt_acl:
360 			f2fs_msg(sb, KERN_INFO, "acl options not supported");
361 			break;
362 		case Opt_noacl:
363 			f2fs_msg(sb, KERN_INFO, "noacl options not supported");
364 			break;
365 #endif
366 		case Opt_active_logs:
367 			if (args->from && match_int(args, &arg))
368 				return -EINVAL;
369 			if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
370 				return -EINVAL;
371 			sbi->active_logs = arg;
372 			break;
373 		case Opt_disable_ext_identify:
374 			set_opt(sbi, DISABLE_EXT_IDENTIFY);
375 			break;
376 		case Opt_inline_data:
377 			set_opt(sbi, INLINE_DATA);
378 			break;
379 		case Opt_inline_dentry:
380 			set_opt(sbi, INLINE_DENTRY);
381 			break;
382 		case Opt_flush_merge:
383 			set_opt(sbi, FLUSH_MERGE);
384 			break;
385 		case Opt_nobarrier:
386 			set_opt(sbi, NOBARRIER);
387 			break;
388 		case Opt_fastboot:
389 			set_opt(sbi, FASTBOOT);
390 			break;
391 		case Opt_extent_cache:
392 			set_opt(sbi, EXTENT_CACHE);
393 			break;
394 		case Opt_noextent_cache:
395 			clear_opt(sbi, EXTENT_CACHE);
396 			break;
397 		case Opt_noinline_data:
398 			clear_opt(sbi, INLINE_DATA);
399 			break;
400 		default:
401 			f2fs_msg(sb, KERN_ERR,
402 				"Unrecognized mount option \"%s\" or missing value",
403 				p);
404 			return -EINVAL;
405 		}
406 	}
407 	return 0;
408 }
409 
410 static struct inode *f2fs_alloc_inode(struct super_block *sb)
411 {
412 	struct f2fs_inode_info *fi;
413 
414 	fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
415 	if (!fi)
416 		return NULL;
417 
418 	init_once((void *) fi);
419 
420 	/* Initialize f2fs-specific inode info */
421 	fi->vfs_inode.i_version = 1;
422 	atomic_set(&fi->dirty_pages, 0);
423 	fi->i_current_depth = 1;
424 	fi->i_advise = 0;
425 	init_rwsem(&fi->i_sem);
426 	INIT_LIST_HEAD(&fi->inmem_pages);
427 	mutex_init(&fi->inmem_lock);
428 
429 	set_inode_flag(fi, FI_NEW_INODE);
430 
431 	if (test_opt(F2FS_SB(sb), INLINE_XATTR))
432 		set_inode_flag(fi, FI_INLINE_XATTR);
433 
434 	/* Will be used by directory only */
435 	fi->i_dir_level = F2FS_SB(sb)->dir_level;
436 
437 #ifdef CONFIG_F2FS_FS_ENCRYPTION
438 	fi->i_crypt_info = NULL;
439 #endif
440 	return &fi->vfs_inode;
441 }
442 
443 static int f2fs_drop_inode(struct inode *inode)
444 {
445 	/*
446 	 * This is to avoid a deadlock condition like below.
447 	 * writeback_single_inode(inode)
448 	 *  - f2fs_write_data_page
449 	 *    - f2fs_gc -> iput -> evict
450 	 *       - inode_wait_for_writeback(inode)
451 	 */
452 	if (!inode_unhashed(inode) && inode->i_state & I_SYNC) {
453 		if (!inode->i_nlink && !is_bad_inode(inode)) {
454 			/* to avoid evict_inode call simultaneously */
455 			atomic_inc(&inode->i_count);
456 			spin_unlock(&inode->i_lock);
457 
458 			/* some remained atomic pages should discarded */
459 			if (f2fs_is_atomic_file(inode))
460 				commit_inmem_pages(inode, true);
461 
462 			/* should remain fi->extent_tree for writepage */
463 			f2fs_destroy_extent_node(inode);
464 
465 			sb_start_intwrite(inode->i_sb);
466 			i_size_write(inode, 0);
467 
468 			if (F2FS_HAS_BLOCKS(inode))
469 				f2fs_truncate(inode, true);
470 
471 			sb_end_intwrite(inode->i_sb);
472 
473 #ifdef CONFIG_F2FS_FS_ENCRYPTION
474 			if (F2FS_I(inode)->i_crypt_info)
475 				f2fs_free_encryption_info(inode,
476 					F2FS_I(inode)->i_crypt_info);
477 #endif
478 			spin_lock(&inode->i_lock);
479 			atomic_dec(&inode->i_count);
480 		}
481 		return 0;
482 	}
483 	return generic_drop_inode(inode);
484 }
485 
486 /*
487  * f2fs_dirty_inode() is called from __mark_inode_dirty()
488  *
489  * We should call set_dirty_inode to write the dirty inode through write_inode.
490  */
491 static void f2fs_dirty_inode(struct inode *inode, int flags)
492 {
493 	set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
494 }
495 
496 static void f2fs_i_callback(struct rcu_head *head)
497 {
498 	struct inode *inode = container_of(head, struct inode, i_rcu);
499 	kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
500 }
501 
502 static void f2fs_destroy_inode(struct inode *inode)
503 {
504 	call_rcu(&inode->i_rcu, f2fs_i_callback);
505 }
506 
507 static void f2fs_put_super(struct super_block *sb)
508 {
509 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
510 
511 	if (sbi->s_proc) {
512 		remove_proc_entry("segment_info", sbi->s_proc);
513 		remove_proc_entry(sb->s_id, f2fs_proc_root);
514 	}
515 	kobject_del(&sbi->s_kobj);
516 
517 	stop_gc_thread(sbi);
518 
519 	/* prevent remaining shrinker jobs */
520 	mutex_lock(&sbi->umount_mutex);
521 
522 	/*
523 	 * We don't need to do checkpoint when superblock is clean.
524 	 * But, the previous checkpoint was not done by umount, it needs to do
525 	 * clean checkpoint again.
526 	 */
527 	if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
528 			!is_set_ckpt_flags(F2FS_CKPT(sbi), CP_UMOUNT_FLAG)) {
529 		struct cp_control cpc = {
530 			.reason = CP_UMOUNT,
531 		};
532 		write_checkpoint(sbi, &cpc);
533 	}
534 
535 	/* write_checkpoint can update stat informaion */
536 	f2fs_destroy_stats(sbi);
537 
538 	/*
539 	 * normally superblock is clean, so we need to release this.
540 	 * In addition, EIO will skip do checkpoint, we need this as well.
541 	 */
542 	release_dirty_inode(sbi);
543 	release_discard_addrs(sbi);
544 
545 	f2fs_leave_shrinker(sbi);
546 	mutex_unlock(&sbi->umount_mutex);
547 
548 	iput(sbi->node_inode);
549 	iput(sbi->meta_inode);
550 
551 	/* destroy f2fs internal modules */
552 	destroy_node_manager(sbi);
553 	destroy_segment_manager(sbi);
554 
555 	kfree(sbi->ckpt);
556 	kobject_put(&sbi->s_kobj);
557 	wait_for_completion(&sbi->s_kobj_unregister);
558 
559 	sb->s_fs_info = NULL;
560 	brelse(sbi->raw_super_buf);
561 	kfree(sbi);
562 }
563 
564 int f2fs_sync_fs(struct super_block *sb, int sync)
565 {
566 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
567 
568 	trace_f2fs_sync_fs(sb, sync);
569 
570 	if (sync) {
571 		struct cp_control cpc;
572 
573 		cpc.reason = __get_cp_reason(sbi);
574 
575 		mutex_lock(&sbi->gc_mutex);
576 		write_checkpoint(sbi, &cpc);
577 		mutex_unlock(&sbi->gc_mutex);
578 	} else {
579 		f2fs_balance_fs(sbi);
580 	}
581 	f2fs_trace_ios(NULL, 1);
582 
583 	return 0;
584 }
585 
586 static int f2fs_freeze(struct super_block *sb)
587 {
588 	int err;
589 
590 	if (f2fs_readonly(sb))
591 		return 0;
592 
593 	err = f2fs_sync_fs(sb, 1);
594 	return err;
595 }
596 
597 static int f2fs_unfreeze(struct super_block *sb)
598 {
599 	return 0;
600 }
601 
602 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
603 {
604 	struct super_block *sb = dentry->d_sb;
605 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
606 	u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
607 	block_t total_count, user_block_count, start_count, ovp_count;
608 
609 	total_count = le64_to_cpu(sbi->raw_super->block_count);
610 	user_block_count = sbi->user_block_count;
611 	start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
612 	ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
613 	buf->f_type = F2FS_SUPER_MAGIC;
614 	buf->f_bsize = sbi->blocksize;
615 
616 	buf->f_blocks = total_count - start_count;
617 	buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
618 	buf->f_bavail = user_block_count - valid_user_blocks(sbi);
619 
620 	buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
621 	buf->f_ffree = buf->f_files - valid_inode_count(sbi);
622 
623 	buf->f_namelen = F2FS_NAME_LEN;
624 	buf->f_fsid.val[0] = (u32)id;
625 	buf->f_fsid.val[1] = (u32)(id >> 32);
626 
627 	return 0;
628 }
629 
630 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
631 {
632 	struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
633 
634 	if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC))
635 		seq_printf(seq, ",background_gc=%s", "on");
636 	else
637 		seq_printf(seq, ",background_gc=%s", "off");
638 	if (test_opt(sbi, DISABLE_ROLL_FORWARD))
639 		seq_puts(seq, ",disable_roll_forward");
640 	if (test_opt(sbi, DISCARD))
641 		seq_puts(seq, ",discard");
642 	if (test_opt(sbi, NOHEAP))
643 		seq_puts(seq, ",no_heap_alloc");
644 #ifdef CONFIG_F2FS_FS_XATTR
645 	if (test_opt(sbi, XATTR_USER))
646 		seq_puts(seq, ",user_xattr");
647 	else
648 		seq_puts(seq, ",nouser_xattr");
649 	if (test_opt(sbi, INLINE_XATTR))
650 		seq_puts(seq, ",inline_xattr");
651 #endif
652 #ifdef CONFIG_F2FS_FS_POSIX_ACL
653 	if (test_opt(sbi, POSIX_ACL))
654 		seq_puts(seq, ",acl");
655 	else
656 		seq_puts(seq, ",noacl");
657 #endif
658 	if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
659 		seq_puts(seq, ",disable_ext_identify");
660 	if (test_opt(sbi, INLINE_DATA))
661 		seq_puts(seq, ",inline_data");
662 	else
663 		seq_puts(seq, ",noinline_data");
664 	if (test_opt(sbi, INLINE_DENTRY))
665 		seq_puts(seq, ",inline_dentry");
666 	if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
667 		seq_puts(seq, ",flush_merge");
668 	if (test_opt(sbi, NOBARRIER))
669 		seq_puts(seq, ",nobarrier");
670 	if (test_opt(sbi, FASTBOOT))
671 		seq_puts(seq, ",fastboot");
672 	if (test_opt(sbi, EXTENT_CACHE))
673 		seq_puts(seq, ",extent_cache");
674 	else
675 		seq_puts(seq, ",noextent_cache");
676 	seq_printf(seq, ",active_logs=%u", sbi->active_logs);
677 
678 	return 0;
679 }
680 
681 static int segment_info_seq_show(struct seq_file *seq, void *offset)
682 {
683 	struct super_block *sb = seq->private;
684 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
685 	unsigned int total_segs =
686 			le32_to_cpu(sbi->raw_super->segment_count_main);
687 	int i;
688 
689 	seq_puts(seq, "format: segment_type|valid_blocks\n"
690 		"segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
691 
692 	for (i = 0; i < total_segs; i++) {
693 		struct seg_entry *se = get_seg_entry(sbi, i);
694 
695 		if ((i % 10) == 0)
696 			seq_printf(seq, "%-10d", i);
697 		seq_printf(seq, "%d|%-3u", se->type,
698 					get_valid_blocks(sbi, i, 1));
699 		if ((i % 10) == 9 || i == (total_segs - 1))
700 			seq_putc(seq, '\n');
701 		else
702 			seq_putc(seq, ' ');
703 	}
704 
705 	return 0;
706 }
707 
708 static int segment_info_open_fs(struct inode *inode, struct file *file)
709 {
710 	return single_open(file, segment_info_seq_show, PDE_DATA(inode));
711 }
712 
713 static const struct file_operations f2fs_seq_segment_info_fops = {
714 	.owner = THIS_MODULE,
715 	.open = segment_info_open_fs,
716 	.read = seq_read,
717 	.llseek = seq_lseek,
718 	.release = single_release,
719 };
720 
721 static void default_options(struct f2fs_sb_info *sbi)
722 {
723 	/* init some FS parameters */
724 	sbi->active_logs = NR_CURSEG_TYPE;
725 
726 	set_opt(sbi, BG_GC);
727 	set_opt(sbi, INLINE_DATA);
728 	set_opt(sbi, EXTENT_CACHE);
729 
730 #ifdef CONFIG_F2FS_FS_XATTR
731 	set_opt(sbi, XATTR_USER);
732 #endif
733 #ifdef CONFIG_F2FS_FS_POSIX_ACL
734 	set_opt(sbi, POSIX_ACL);
735 #endif
736 }
737 
738 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
739 {
740 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
741 	struct f2fs_mount_info org_mount_opt;
742 	int err, active_logs;
743 	bool need_restart_gc = false;
744 	bool need_stop_gc = false;
745 
746 	sync_filesystem(sb);
747 
748 	/*
749 	 * Save the old mount options in case we
750 	 * need to restore them.
751 	 */
752 	org_mount_opt = sbi->mount_opt;
753 	active_logs = sbi->active_logs;
754 
755 	sbi->mount_opt.opt = 0;
756 	default_options(sbi);
757 
758 	/* parse mount options */
759 	err = parse_options(sb, data);
760 	if (err)
761 		goto restore_opts;
762 
763 	/*
764 	 * Previous and new state of filesystem is RO,
765 	 * so skip checking GC and FLUSH_MERGE conditions.
766 	 */
767 	if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
768 		goto skip;
769 
770 	/*
771 	 * We stop the GC thread if FS is mounted as RO
772 	 * or if background_gc = off is passed in mount
773 	 * option. Also sync the filesystem.
774 	 */
775 	if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
776 		if (sbi->gc_thread) {
777 			stop_gc_thread(sbi);
778 			f2fs_sync_fs(sb, 1);
779 			need_restart_gc = true;
780 		}
781 	} else if (!sbi->gc_thread) {
782 		err = start_gc_thread(sbi);
783 		if (err)
784 			goto restore_opts;
785 		need_stop_gc = true;
786 	}
787 
788 	/*
789 	 * We stop issue flush thread if FS is mounted as RO
790 	 * or if flush_merge is not passed in mount option.
791 	 */
792 	if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
793 		destroy_flush_cmd_control(sbi);
794 	} else if (!SM_I(sbi)->cmd_control_info) {
795 		err = create_flush_cmd_control(sbi);
796 		if (err)
797 			goto restore_gc;
798 	}
799 skip:
800 	/* Update the POSIXACL Flag */
801 	 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
802 		(test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
803 	return 0;
804 restore_gc:
805 	if (need_restart_gc) {
806 		if (start_gc_thread(sbi))
807 			f2fs_msg(sbi->sb, KERN_WARNING,
808 				"background gc thread has stopped");
809 	} else if (need_stop_gc) {
810 		stop_gc_thread(sbi);
811 	}
812 restore_opts:
813 	sbi->mount_opt = org_mount_opt;
814 	sbi->active_logs = active_logs;
815 	return err;
816 }
817 
818 static struct super_operations f2fs_sops = {
819 	.alloc_inode	= f2fs_alloc_inode,
820 	.drop_inode	= f2fs_drop_inode,
821 	.destroy_inode	= f2fs_destroy_inode,
822 	.write_inode	= f2fs_write_inode,
823 	.dirty_inode	= f2fs_dirty_inode,
824 	.show_options	= f2fs_show_options,
825 	.evict_inode	= f2fs_evict_inode,
826 	.put_super	= f2fs_put_super,
827 	.sync_fs	= f2fs_sync_fs,
828 	.freeze_fs	= f2fs_freeze,
829 	.unfreeze_fs	= f2fs_unfreeze,
830 	.statfs		= f2fs_statfs,
831 	.remount_fs	= f2fs_remount,
832 };
833 
834 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
835 		u64 ino, u32 generation)
836 {
837 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
838 	struct inode *inode;
839 
840 	if (check_nid_range(sbi, ino))
841 		return ERR_PTR(-ESTALE);
842 
843 	/*
844 	 * f2fs_iget isn't quite right if the inode is currently unallocated!
845 	 * However f2fs_iget currently does appropriate checks to handle stale
846 	 * inodes so everything is OK.
847 	 */
848 	inode = f2fs_iget(sb, ino);
849 	if (IS_ERR(inode))
850 		return ERR_CAST(inode);
851 	if (unlikely(generation && inode->i_generation != generation)) {
852 		/* we didn't find the right inode.. */
853 		iput(inode);
854 		return ERR_PTR(-ESTALE);
855 	}
856 	return inode;
857 }
858 
859 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
860 		int fh_len, int fh_type)
861 {
862 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
863 				    f2fs_nfs_get_inode);
864 }
865 
866 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
867 		int fh_len, int fh_type)
868 {
869 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
870 				    f2fs_nfs_get_inode);
871 }
872 
873 static const struct export_operations f2fs_export_ops = {
874 	.fh_to_dentry = f2fs_fh_to_dentry,
875 	.fh_to_parent = f2fs_fh_to_parent,
876 	.get_parent = f2fs_get_parent,
877 };
878 
879 static loff_t max_file_size(unsigned bits)
880 {
881 	loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
882 	loff_t leaf_count = ADDRS_PER_BLOCK;
883 
884 	/* two direct node blocks */
885 	result += (leaf_count * 2);
886 
887 	/* two indirect node blocks */
888 	leaf_count *= NIDS_PER_BLOCK;
889 	result += (leaf_count * 2);
890 
891 	/* one double indirect node block */
892 	leaf_count *= NIDS_PER_BLOCK;
893 	result += leaf_count;
894 
895 	result <<= bits;
896 	return result;
897 }
898 
899 static int sanity_check_raw_super(struct super_block *sb,
900 			struct f2fs_super_block *raw_super)
901 {
902 	unsigned int blocksize;
903 
904 	if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
905 		f2fs_msg(sb, KERN_INFO,
906 			"Magic Mismatch, valid(0x%x) - read(0x%x)",
907 			F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
908 		return 1;
909 	}
910 
911 	/* Currently, support only 4KB page cache size */
912 	if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
913 		f2fs_msg(sb, KERN_INFO,
914 			"Invalid page_cache_size (%lu), supports only 4KB\n",
915 			PAGE_CACHE_SIZE);
916 		return 1;
917 	}
918 
919 	/* Currently, support only 4KB block size */
920 	blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
921 	if (blocksize != F2FS_BLKSIZE) {
922 		f2fs_msg(sb, KERN_INFO,
923 			"Invalid blocksize (%u), supports only 4KB\n",
924 			blocksize);
925 		return 1;
926 	}
927 
928 	/* Currently, support 512/1024/2048/4096 bytes sector size */
929 	if (le32_to_cpu(raw_super->log_sectorsize) >
930 				F2FS_MAX_LOG_SECTOR_SIZE ||
931 		le32_to_cpu(raw_super->log_sectorsize) <
932 				F2FS_MIN_LOG_SECTOR_SIZE) {
933 		f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
934 			le32_to_cpu(raw_super->log_sectorsize));
935 		return 1;
936 	}
937 	if (le32_to_cpu(raw_super->log_sectors_per_block) +
938 		le32_to_cpu(raw_super->log_sectorsize) !=
939 			F2FS_MAX_LOG_SECTOR_SIZE) {
940 		f2fs_msg(sb, KERN_INFO,
941 			"Invalid log sectors per block(%u) log sectorsize(%u)",
942 			le32_to_cpu(raw_super->log_sectors_per_block),
943 			le32_to_cpu(raw_super->log_sectorsize));
944 		return 1;
945 	}
946 	return 0;
947 }
948 
949 static int sanity_check_ckpt(struct f2fs_sb_info *sbi)
950 {
951 	unsigned int total, fsmeta;
952 	struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
953 	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
954 
955 	total = le32_to_cpu(raw_super->segment_count);
956 	fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
957 	fsmeta += le32_to_cpu(raw_super->segment_count_sit);
958 	fsmeta += le32_to_cpu(raw_super->segment_count_nat);
959 	fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
960 	fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
961 
962 	if (unlikely(fsmeta >= total))
963 		return 1;
964 
965 	if (unlikely(f2fs_cp_error(sbi))) {
966 		f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
967 		return 1;
968 	}
969 	return 0;
970 }
971 
972 static void init_sb_info(struct f2fs_sb_info *sbi)
973 {
974 	struct f2fs_super_block *raw_super = sbi->raw_super;
975 	int i;
976 
977 	sbi->log_sectors_per_block =
978 		le32_to_cpu(raw_super->log_sectors_per_block);
979 	sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
980 	sbi->blocksize = 1 << sbi->log_blocksize;
981 	sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
982 	sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
983 	sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
984 	sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
985 	sbi->total_sections = le32_to_cpu(raw_super->section_count);
986 	sbi->total_node_count =
987 		(le32_to_cpu(raw_super->segment_count_nat) / 2)
988 			* sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
989 	sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
990 	sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
991 	sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
992 	sbi->cur_victim_sec = NULL_SECNO;
993 	sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
994 
995 	for (i = 0; i < NR_COUNT_TYPE; i++)
996 		atomic_set(&sbi->nr_pages[i], 0);
997 
998 	sbi->dir_level = DEF_DIR_LEVEL;
999 	clear_sbi_flag(sbi, SBI_NEED_FSCK);
1000 
1001 	INIT_LIST_HEAD(&sbi->s_list);
1002 	mutex_init(&sbi->umount_mutex);
1003 }
1004 
1005 /*
1006  * Read f2fs raw super block.
1007  * Because we have two copies of super block, so read the first one at first,
1008  * if the first one is invalid, move to read the second one.
1009  */
1010 static int read_raw_super_block(struct super_block *sb,
1011 			struct f2fs_super_block **raw_super,
1012 			struct buffer_head **raw_super_buf,
1013 			int *recovery)
1014 {
1015 	int block = 0;
1016 	struct buffer_head *buffer;
1017 	struct f2fs_super_block *super;
1018 	int err = 0;
1019 
1020 retry:
1021 	buffer = sb_bread(sb, block);
1022 	if (!buffer) {
1023 		*recovery = 1;
1024 		f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
1025 				block + 1);
1026 		if (block == 0) {
1027 			block++;
1028 			goto retry;
1029 		} else {
1030 			err = -EIO;
1031 			goto out;
1032 		}
1033 	}
1034 
1035 	super = (struct f2fs_super_block *)
1036 		((char *)(buffer)->b_data + F2FS_SUPER_OFFSET);
1037 
1038 	/* sanity checking of raw super */
1039 	if (sanity_check_raw_super(sb, super)) {
1040 		brelse(buffer);
1041 		*recovery = 1;
1042 		f2fs_msg(sb, KERN_ERR,
1043 			"Can't find valid F2FS filesystem in %dth superblock",
1044 								block + 1);
1045 		if (block == 0) {
1046 			block++;
1047 			goto retry;
1048 		} else {
1049 			err = -EINVAL;
1050 			goto out;
1051 		}
1052 	}
1053 
1054 	if (!*raw_super) {
1055 		*raw_super_buf = buffer;
1056 		*raw_super = super;
1057 	} else {
1058 		/* already have a valid superblock */
1059 		brelse(buffer);
1060 	}
1061 
1062 	/* check the validity of the second superblock */
1063 	if (block == 0) {
1064 		block++;
1065 		goto retry;
1066 	}
1067 
1068 out:
1069 	/* No valid superblock */
1070 	if (!*raw_super)
1071 		return err;
1072 
1073 	return 0;
1074 }
1075 
1076 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
1077 {
1078 	struct buffer_head *sbh = sbi->raw_super_buf;
1079 	sector_t block = sbh->b_blocknr;
1080 	int err;
1081 
1082 	/* write back-up superblock first */
1083 	sbh->b_blocknr = block ? 0 : 1;
1084 	mark_buffer_dirty(sbh);
1085 	err = sync_dirty_buffer(sbh);
1086 
1087 	sbh->b_blocknr = block;
1088 
1089 	/* if we are in recovery path, skip writing valid superblock */
1090 	if (recover || err)
1091 		goto out;
1092 
1093 	/* write current valid superblock */
1094 	mark_buffer_dirty(sbh);
1095 	err = sync_dirty_buffer(sbh);
1096 out:
1097 	clear_buffer_write_io_error(sbh);
1098 	set_buffer_uptodate(sbh);
1099 	return err;
1100 }
1101 
1102 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
1103 {
1104 	struct f2fs_sb_info *sbi;
1105 	struct f2fs_super_block *raw_super;
1106 	struct buffer_head *raw_super_buf;
1107 	struct inode *root;
1108 	long err;
1109 	bool retry = true, need_fsck = false;
1110 	char *options = NULL;
1111 	int recovery, i;
1112 
1113 try_onemore:
1114 	err = -EINVAL;
1115 	raw_super = NULL;
1116 	raw_super_buf = NULL;
1117 	recovery = 0;
1118 
1119 	/* allocate memory for f2fs-specific super block info */
1120 	sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
1121 	if (!sbi)
1122 		return -ENOMEM;
1123 
1124 	/* set a block size */
1125 	if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
1126 		f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
1127 		goto free_sbi;
1128 	}
1129 
1130 	err = read_raw_super_block(sb, &raw_super, &raw_super_buf, &recovery);
1131 	if (err)
1132 		goto free_sbi;
1133 
1134 	sb->s_fs_info = sbi;
1135 	default_options(sbi);
1136 	/* parse mount options */
1137 	options = kstrdup((const char *)data, GFP_KERNEL);
1138 	if (data && !options) {
1139 		err = -ENOMEM;
1140 		goto free_sb_buf;
1141 	}
1142 
1143 	err = parse_options(sb, options);
1144 	if (err)
1145 		goto free_options;
1146 
1147 	sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
1148 	sb->s_max_links = F2FS_LINK_MAX;
1149 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1150 
1151 	sb->s_op = &f2fs_sops;
1152 	sb->s_xattr = f2fs_xattr_handlers;
1153 	sb->s_export_op = &f2fs_export_ops;
1154 	sb->s_magic = F2FS_SUPER_MAGIC;
1155 	sb->s_time_gran = 1;
1156 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1157 		(test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1158 	memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1159 
1160 	/* init f2fs-specific super block info */
1161 	sbi->sb = sb;
1162 	sbi->raw_super = raw_super;
1163 	sbi->raw_super_buf = raw_super_buf;
1164 	mutex_init(&sbi->gc_mutex);
1165 	mutex_init(&sbi->writepages);
1166 	mutex_init(&sbi->cp_mutex);
1167 	init_rwsem(&sbi->node_write);
1168 
1169 	/* disallow all the data/node/meta page writes */
1170 	set_sbi_flag(sbi, SBI_POR_DOING);
1171 	spin_lock_init(&sbi->stat_lock);
1172 
1173 	init_rwsem(&sbi->read_io.io_rwsem);
1174 	sbi->read_io.sbi = sbi;
1175 	sbi->read_io.bio = NULL;
1176 	for (i = 0; i < NR_PAGE_TYPE; i++) {
1177 		init_rwsem(&sbi->write_io[i].io_rwsem);
1178 		sbi->write_io[i].sbi = sbi;
1179 		sbi->write_io[i].bio = NULL;
1180 	}
1181 
1182 	init_rwsem(&sbi->cp_rwsem);
1183 	init_waitqueue_head(&sbi->cp_wait);
1184 	init_sb_info(sbi);
1185 
1186 	/* get an inode for meta space */
1187 	sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1188 	if (IS_ERR(sbi->meta_inode)) {
1189 		f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1190 		err = PTR_ERR(sbi->meta_inode);
1191 		goto free_options;
1192 	}
1193 
1194 	err = get_valid_checkpoint(sbi);
1195 	if (err) {
1196 		f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1197 		goto free_meta_inode;
1198 	}
1199 
1200 	/* sanity checking of checkpoint */
1201 	err = -EINVAL;
1202 	if (sanity_check_ckpt(sbi)) {
1203 		f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint");
1204 		goto free_cp;
1205 	}
1206 
1207 	sbi->total_valid_node_count =
1208 				le32_to_cpu(sbi->ckpt->valid_node_count);
1209 	sbi->total_valid_inode_count =
1210 				le32_to_cpu(sbi->ckpt->valid_inode_count);
1211 	sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1212 	sbi->total_valid_block_count =
1213 				le64_to_cpu(sbi->ckpt->valid_block_count);
1214 	sbi->last_valid_block_count = sbi->total_valid_block_count;
1215 	sbi->alloc_valid_block_count = 0;
1216 	INIT_LIST_HEAD(&sbi->dir_inode_list);
1217 	spin_lock_init(&sbi->dir_inode_lock);
1218 
1219 	init_extent_cache_info(sbi);
1220 
1221 	init_ino_entry_info(sbi);
1222 
1223 	/* setup f2fs internal modules */
1224 	err = build_segment_manager(sbi);
1225 	if (err) {
1226 		f2fs_msg(sb, KERN_ERR,
1227 			"Failed to initialize F2FS segment manager");
1228 		goto free_sm;
1229 	}
1230 	err = build_node_manager(sbi);
1231 	if (err) {
1232 		f2fs_msg(sb, KERN_ERR,
1233 			"Failed to initialize F2FS node manager");
1234 		goto free_nm;
1235 	}
1236 
1237 	build_gc_manager(sbi);
1238 
1239 	/* get an inode for node space */
1240 	sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1241 	if (IS_ERR(sbi->node_inode)) {
1242 		f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1243 		err = PTR_ERR(sbi->node_inode);
1244 		goto free_nm;
1245 	}
1246 
1247 	f2fs_join_shrinker(sbi);
1248 
1249 	/* if there are nt orphan nodes free them */
1250 	err = recover_orphan_inodes(sbi);
1251 	if (err)
1252 		goto free_node_inode;
1253 
1254 	/* read root inode and dentry */
1255 	root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1256 	if (IS_ERR(root)) {
1257 		f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1258 		err = PTR_ERR(root);
1259 		goto free_node_inode;
1260 	}
1261 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1262 		iput(root);
1263 		err = -EINVAL;
1264 		goto free_node_inode;
1265 	}
1266 
1267 	sb->s_root = d_make_root(root); /* allocate root dentry */
1268 	if (!sb->s_root) {
1269 		err = -ENOMEM;
1270 		goto free_root_inode;
1271 	}
1272 
1273 	err = f2fs_build_stats(sbi);
1274 	if (err)
1275 		goto free_root_inode;
1276 
1277 	if (f2fs_proc_root)
1278 		sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1279 
1280 	if (sbi->s_proc)
1281 		proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1282 				 &f2fs_seq_segment_info_fops, sb);
1283 
1284 	sbi->s_kobj.kset = f2fs_kset;
1285 	init_completion(&sbi->s_kobj_unregister);
1286 	err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1287 							"%s", sb->s_id);
1288 	if (err)
1289 		goto free_proc;
1290 
1291 	/* recover fsynced data */
1292 	if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1293 		/*
1294 		 * mount should be failed, when device has readonly mode, and
1295 		 * previous checkpoint was not done by clean system shutdown.
1296 		 */
1297 		if (bdev_read_only(sb->s_bdev) &&
1298 				!is_set_ckpt_flags(sbi->ckpt, CP_UMOUNT_FLAG)) {
1299 			err = -EROFS;
1300 			goto free_kobj;
1301 		}
1302 
1303 		if (need_fsck)
1304 			set_sbi_flag(sbi, SBI_NEED_FSCK);
1305 
1306 		err = recover_fsync_data(sbi);
1307 		if (err) {
1308 			need_fsck = true;
1309 			f2fs_msg(sb, KERN_ERR,
1310 				"Cannot recover all fsync data errno=%ld", err);
1311 			goto free_kobj;
1312 		}
1313 	}
1314 	/* recover_fsync_data() cleared this already */
1315 	clear_sbi_flag(sbi, SBI_POR_DOING);
1316 
1317 	/*
1318 	 * If filesystem is not mounted as read-only then
1319 	 * do start the gc_thread.
1320 	 */
1321 	if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
1322 		/* After POR, we can run background GC thread.*/
1323 		err = start_gc_thread(sbi);
1324 		if (err)
1325 			goto free_kobj;
1326 	}
1327 	kfree(options);
1328 
1329 	/* recover broken superblock */
1330 	if (recovery && !f2fs_readonly(sb) && !bdev_read_only(sb->s_bdev)) {
1331 		f2fs_msg(sb, KERN_INFO, "Recover invalid superblock");
1332 		f2fs_commit_super(sbi, true);
1333 	}
1334 
1335 	return 0;
1336 
1337 free_kobj:
1338 	kobject_del(&sbi->s_kobj);
1339 free_proc:
1340 	if (sbi->s_proc) {
1341 		remove_proc_entry("segment_info", sbi->s_proc);
1342 		remove_proc_entry(sb->s_id, f2fs_proc_root);
1343 	}
1344 	f2fs_destroy_stats(sbi);
1345 free_root_inode:
1346 	dput(sb->s_root);
1347 	sb->s_root = NULL;
1348 free_node_inode:
1349 	mutex_lock(&sbi->umount_mutex);
1350 	f2fs_leave_shrinker(sbi);
1351 	iput(sbi->node_inode);
1352 	mutex_unlock(&sbi->umount_mutex);
1353 free_nm:
1354 	destroy_node_manager(sbi);
1355 free_sm:
1356 	destroy_segment_manager(sbi);
1357 free_cp:
1358 	kfree(sbi->ckpt);
1359 free_meta_inode:
1360 	make_bad_inode(sbi->meta_inode);
1361 	iput(sbi->meta_inode);
1362 free_options:
1363 	kfree(options);
1364 free_sb_buf:
1365 	brelse(raw_super_buf);
1366 free_sbi:
1367 	kfree(sbi);
1368 
1369 	/* give only one another chance */
1370 	if (retry) {
1371 		retry = false;
1372 		shrink_dcache_sb(sb);
1373 		goto try_onemore;
1374 	}
1375 	return err;
1376 }
1377 
1378 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1379 			const char *dev_name, void *data)
1380 {
1381 	return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1382 }
1383 
1384 static void kill_f2fs_super(struct super_block *sb)
1385 {
1386 	if (sb->s_root)
1387 		set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
1388 	kill_block_super(sb);
1389 }
1390 
1391 static struct file_system_type f2fs_fs_type = {
1392 	.owner		= THIS_MODULE,
1393 	.name		= "f2fs",
1394 	.mount		= f2fs_mount,
1395 	.kill_sb	= kill_f2fs_super,
1396 	.fs_flags	= FS_REQUIRES_DEV,
1397 };
1398 MODULE_ALIAS_FS("f2fs");
1399 
1400 static int __init init_inodecache(void)
1401 {
1402 	f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
1403 			sizeof(struct f2fs_inode_info));
1404 	if (!f2fs_inode_cachep)
1405 		return -ENOMEM;
1406 	return 0;
1407 }
1408 
1409 static void destroy_inodecache(void)
1410 {
1411 	/*
1412 	 * Make sure all delayed rcu free inodes are flushed before we
1413 	 * destroy cache.
1414 	 */
1415 	rcu_barrier();
1416 	kmem_cache_destroy(f2fs_inode_cachep);
1417 }
1418 
1419 static int __init init_f2fs_fs(void)
1420 {
1421 	int err;
1422 
1423 	f2fs_build_trace_ios();
1424 
1425 	err = init_inodecache();
1426 	if (err)
1427 		goto fail;
1428 	err = create_node_manager_caches();
1429 	if (err)
1430 		goto free_inodecache;
1431 	err = create_segment_manager_caches();
1432 	if (err)
1433 		goto free_node_manager_caches;
1434 	err = create_checkpoint_caches();
1435 	if (err)
1436 		goto free_segment_manager_caches;
1437 	err = create_extent_cache();
1438 	if (err)
1439 		goto free_checkpoint_caches;
1440 	f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1441 	if (!f2fs_kset) {
1442 		err = -ENOMEM;
1443 		goto free_extent_cache;
1444 	}
1445 	err = f2fs_init_crypto();
1446 	if (err)
1447 		goto free_kset;
1448 
1449 	err = register_shrinker(&f2fs_shrinker_info);
1450 	if (err)
1451 		goto free_crypto;
1452 
1453 	err = register_filesystem(&f2fs_fs_type);
1454 	if (err)
1455 		goto free_shrinker;
1456 	f2fs_create_root_stats();
1457 	f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1458 	return 0;
1459 
1460 free_shrinker:
1461 	unregister_shrinker(&f2fs_shrinker_info);
1462 free_crypto:
1463 	f2fs_exit_crypto();
1464 free_kset:
1465 	kset_unregister(f2fs_kset);
1466 free_extent_cache:
1467 	destroy_extent_cache();
1468 free_checkpoint_caches:
1469 	destroy_checkpoint_caches();
1470 free_segment_manager_caches:
1471 	destroy_segment_manager_caches();
1472 free_node_manager_caches:
1473 	destroy_node_manager_caches();
1474 free_inodecache:
1475 	destroy_inodecache();
1476 fail:
1477 	return err;
1478 }
1479 
1480 static void __exit exit_f2fs_fs(void)
1481 {
1482 	remove_proc_entry("fs/f2fs", NULL);
1483 	f2fs_destroy_root_stats();
1484 	unregister_shrinker(&f2fs_shrinker_info);
1485 	unregister_filesystem(&f2fs_fs_type);
1486 	f2fs_exit_crypto();
1487 	destroy_extent_cache();
1488 	destroy_checkpoint_caches();
1489 	destroy_segment_manager_caches();
1490 	destroy_node_manager_caches();
1491 	destroy_inodecache();
1492 	kset_unregister(f2fs_kset);
1493 	f2fs_destroy_trace_ios();
1494 }
1495 
1496 module_init(init_f2fs_fs)
1497 module_exit(exit_f2fs_fs)
1498 
1499 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1500 MODULE_DESCRIPTION("Flash Friendly File System");
1501 MODULE_LICENSE("GPL");
1502