xref: /openbmc/linux/fs/f2fs/super.c (revision 6dfcd296)
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 #ifdef CONFIG_F2FS_FAULT_INJECTION
43 
44 char *fault_name[FAULT_MAX] = {
45 	[FAULT_KMALLOC]		= "kmalloc",
46 	[FAULT_PAGE_ALLOC]	= "page alloc",
47 	[FAULT_ALLOC_NID]	= "alloc nid",
48 	[FAULT_ORPHAN]		= "orphan",
49 	[FAULT_BLOCK]		= "no more block",
50 	[FAULT_DIR_DEPTH]	= "too big dir depth",
51 	[FAULT_EVICT_INODE]	= "evict_inode fail",
52 	[FAULT_IO]		= "IO error",
53 	[FAULT_CHECKPOINT]	= "checkpoint error",
54 };
55 
56 static void f2fs_build_fault_attr(struct f2fs_sb_info *sbi,
57 						unsigned int rate)
58 {
59 	struct f2fs_fault_info *ffi = &sbi->fault_info;
60 
61 	if (rate) {
62 		atomic_set(&ffi->inject_ops, 0);
63 		ffi->inject_rate = rate;
64 		ffi->inject_type = (1 << FAULT_MAX) - 1;
65 	} else {
66 		memset(ffi, 0, sizeof(struct f2fs_fault_info));
67 	}
68 }
69 #endif
70 
71 /* f2fs-wide shrinker description */
72 static struct shrinker f2fs_shrinker_info = {
73 	.scan_objects = f2fs_shrink_scan,
74 	.count_objects = f2fs_shrink_count,
75 	.seeks = DEFAULT_SEEKS,
76 };
77 
78 enum {
79 	Opt_gc_background,
80 	Opt_disable_roll_forward,
81 	Opt_norecovery,
82 	Opt_discard,
83 	Opt_nodiscard,
84 	Opt_noheap,
85 	Opt_user_xattr,
86 	Opt_nouser_xattr,
87 	Opt_acl,
88 	Opt_noacl,
89 	Opt_active_logs,
90 	Opt_disable_ext_identify,
91 	Opt_inline_xattr,
92 	Opt_inline_data,
93 	Opt_inline_dentry,
94 	Opt_noinline_dentry,
95 	Opt_flush_merge,
96 	Opt_noflush_merge,
97 	Opt_nobarrier,
98 	Opt_fastboot,
99 	Opt_extent_cache,
100 	Opt_noextent_cache,
101 	Opt_noinline_data,
102 	Opt_data_flush,
103 	Opt_mode,
104 	Opt_fault_injection,
105 	Opt_lazytime,
106 	Opt_nolazytime,
107 	Opt_err,
108 };
109 
110 static match_table_t f2fs_tokens = {
111 	{Opt_gc_background, "background_gc=%s"},
112 	{Opt_disable_roll_forward, "disable_roll_forward"},
113 	{Opt_norecovery, "norecovery"},
114 	{Opt_discard, "discard"},
115 	{Opt_nodiscard, "nodiscard"},
116 	{Opt_noheap, "no_heap"},
117 	{Opt_user_xattr, "user_xattr"},
118 	{Opt_nouser_xattr, "nouser_xattr"},
119 	{Opt_acl, "acl"},
120 	{Opt_noacl, "noacl"},
121 	{Opt_active_logs, "active_logs=%u"},
122 	{Opt_disable_ext_identify, "disable_ext_identify"},
123 	{Opt_inline_xattr, "inline_xattr"},
124 	{Opt_inline_data, "inline_data"},
125 	{Opt_inline_dentry, "inline_dentry"},
126 	{Opt_noinline_dentry, "noinline_dentry"},
127 	{Opt_flush_merge, "flush_merge"},
128 	{Opt_noflush_merge, "noflush_merge"},
129 	{Opt_nobarrier, "nobarrier"},
130 	{Opt_fastboot, "fastboot"},
131 	{Opt_extent_cache, "extent_cache"},
132 	{Opt_noextent_cache, "noextent_cache"},
133 	{Opt_noinline_data, "noinline_data"},
134 	{Opt_data_flush, "data_flush"},
135 	{Opt_mode, "mode=%s"},
136 	{Opt_fault_injection, "fault_injection=%u"},
137 	{Opt_lazytime, "lazytime"},
138 	{Opt_nolazytime, "nolazytime"},
139 	{Opt_err, NULL},
140 };
141 
142 /* Sysfs support for f2fs */
143 enum {
144 	GC_THREAD,	/* struct f2fs_gc_thread */
145 	SM_INFO,	/* struct f2fs_sm_info */
146 	NM_INFO,	/* struct f2fs_nm_info */
147 	F2FS_SBI,	/* struct f2fs_sb_info */
148 #ifdef CONFIG_F2FS_FAULT_INJECTION
149 	FAULT_INFO_RATE,	/* struct f2fs_fault_info */
150 	FAULT_INFO_TYPE,	/* struct f2fs_fault_info */
151 #endif
152 };
153 
154 struct f2fs_attr {
155 	struct attribute attr;
156 	ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
157 	ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
158 			 const char *, size_t);
159 	int struct_type;
160 	int offset;
161 };
162 
163 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
164 {
165 	if (struct_type == GC_THREAD)
166 		return (unsigned char *)sbi->gc_thread;
167 	else if (struct_type == SM_INFO)
168 		return (unsigned char *)SM_I(sbi);
169 	else if (struct_type == NM_INFO)
170 		return (unsigned char *)NM_I(sbi);
171 	else if (struct_type == F2FS_SBI)
172 		return (unsigned char *)sbi;
173 #ifdef CONFIG_F2FS_FAULT_INJECTION
174 	else if (struct_type == FAULT_INFO_RATE ||
175 					struct_type == FAULT_INFO_TYPE)
176 		return (unsigned char *)&sbi->fault_info;
177 #endif
178 	return NULL;
179 }
180 
181 static ssize_t lifetime_write_kbytes_show(struct f2fs_attr *a,
182 		struct f2fs_sb_info *sbi, char *buf)
183 {
184 	struct super_block *sb = sbi->sb;
185 
186 	if (!sb->s_bdev->bd_part)
187 		return snprintf(buf, PAGE_SIZE, "0\n");
188 
189 	return snprintf(buf, PAGE_SIZE, "%llu\n",
190 		(unsigned long long)(sbi->kbytes_written +
191 			BD_PART_WRITTEN(sbi)));
192 }
193 
194 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
195 			struct f2fs_sb_info *sbi, char *buf)
196 {
197 	unsigned char *ptr = NULL;
198 	unsigned int *ui;
199 
200 	ptr = __struct_ptr(sbi, a->struct_type);
201 	if (!ptr)
202 		return -EINVAL;
203 
204 	ui = (unsigned int *)(ptr + a->offset);
205 
206 	return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
207 }
208 
209 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
210 			struct f2fs_sb_info *sbi,
211 			const char *buf, size_t count)
212 {
213 	unsigned char *ptr;
214 	unsigned long t;
215 	unsigned int *ui;
216 	ssize_t ret;
217 
218 	ptr = __struct_ptr(sbi, a->struct_type);
219 	if (!ptr)
220 		return -EINVAL;
221 
222 	ui = (unsigned int *)(ptr + a->offset);
223 
224 	ret = kstrtoul(skip_spaces(buf), 0, &t);
225 	if (ret < 0)
226 		return ret;
227 #ifdef CONFIG_F2FS_FAULT_INJECTION
228 	if (a->struct_type == FAULT_INFO_TYPE && t >= (1 << FAULT_MAX))
229 		return -EINVAL;
230 #endif
231 	*ui = t;
232 	return count;
233 }
234 
235 static ssize_t f2fs_attr_show(struct kobject *kobj,
236 				struct attribute *attr, char *buf)
237 {
238 	struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
239 								s_kobj);
240 	struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
241 
242 	return a->show ? a->show(a, sbi, buf) : 0;
243 }
244 
245 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
246 						const char *buf, size_t len)
247 {
248 	struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
249 									s_kobj);
250 	struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
251 
252 	return a->store ? a->store(a, sbi, buf, len) : 0;
253 }
254 
255 static void f2fs_sb_release(struct kobject *kobj)
256 {
257 	struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
258 								s_kobj);
259 	complete(&sbi->s_kobj_unregister);
260 }
261 
262 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
263 static struct f2fs_attr f2fs_attr_##_name = {			\
264 	.attr = {.name = __stringify(_name), .mode = _mode },	\
265 	.show	= _show,					\
266 	.store	= _store,					\
267 	.struct_type = _struct_type,				\
268 	.offset = _offset					\
269 }
270 
271 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname)	\
272 	F2FS_ATTR_OFFSET(struct_type, name, 0644,		\
273 		f2fs_sbi_show, f2fs_sbi_store,			\
274 		offsetof(struct struct_name, elname))
275 
276 #define F2FS_GENERAL_RO_ATTR(name) \
277 static struct f2fs_attr f2fs_attr_##name = __ATTR(name, 0444, name##_show, NULL)
278 
279 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
280 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
281 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
282 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
283 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
284 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
285 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, batched_trim_sections, trim_sections);
286 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
287 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
288 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks);
289 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
290 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ra_nid_pages, ra_nid_pages);
291 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, dirty_nats_ratio, dirty_nats_ratio);
292 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
293 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
294 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, cp_interval, interval_time[CP_TIME]);
295 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, idle_interval, interval_time[REQ_TIME]);
296 #ifdef CONFIG_F2FS_FAULT_INJECTION
297 F2FS_RW_ATTR(FAULT_INFO_RATE, f2fs_fault_info, inject_rate, inject_rate);
298 F2FS_RW_ATTR(FAULT_INFO_TYPE, f2fs_fault_info, inject_type, inject_type);
299 #endif
300 F2FS_GENERAL_RO_ATTR(lifetime_write_kbytes);
301 
302 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
303 static struct attribute *f2fs_attrs[] = {
304 	ATTR_LIST(gc_min_sleep_time),
305 	ATTR_LIST(gc_max_sleep_time),
306 	ATTR_LIST(gc_no_gc_sleep_time),
307 	ATTR_LIST(gc_idle),
308 	ATTR_LIST(reclaim_segments),
309 	ATTR_LIST(max_small_discards),
310 	ATTR_LIST(batched_trim_sections),
311 	ATTR_LIST(ipu_policy),
312 	ATTR_LIST(min_ipu_util),
313 	ATTR_LIST(min_fsync_blocks),
314 	ATTR_LIST(max_victim_search),
315 	ATTR_LIST(dir_level),
316 	ATTR_LIST(ram_thresh),
317 	ATTR_LIST(ra_nid_pages),
318 	ATTR_LIST(dirty_nats_ratio),
319 	ATTR_LIST(cp_interval),
320 	ATTR_LIST(idle_interval),
321 #ifdef CONFIG_F2FS_FAULT_INJECTION
322 	ATTR_LIST(inject_rate),
323 	ATTR_LIST(inject_type),
324 #endif
325 	ATTR_LIST(lifetime_write_kbytes),
326 	NULL,
327 };
328 
329 static const struct sysfs_ops f2fs_attr_ops = {
330 	.show	= f2fs_attr_show,
331 	.store	= f2fs_attr_store,
332 };
333 
334 static struct kobj_type f2fs_ktype = {
335 	.default_attrs	= f2fs_attrs,
336 	.sysfs_ops	= &f2fs_attr_ops,
337 	.release	= f2fs_sb_release,
338 };
339 
340 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
341 {
342 	struct va_format vaf;
343 	va_list args;
344 
345 	va_start(args, fmt);
346 	vaf.fmt = fmt;
347 	vaf.va = &args;
348 	printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
349 	va_end(args);
350 }
351 
352 static void init_once(void *foo)
353 {
354 	struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
355 
356 	inode_init_once(&fi->vfs_inode);
357 }
358 
359 static int parse_options(struct super_block *sb, char *options)
360 {
361 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
362 	struct request_queue *q;
363 	substring_t args[MAX_OPT_ARGS];
364 	char *p, *name;
365 	int arg = 0;
366 
367 	if (!options)
368 		return 0;
369 
370 	while ((p = strsep(&options, ",")) != NULL) {
371 		int token;
372 		if (!*p)
373 			continue;
374 		/*
375 		 * Initialize args struct so we know whether arg was
376 		 * found; some options take optional arguments.
377 		 */
378 		args[0].to = args[0].from = NULL;
379 		token = match_token(p, f2fs_tokens, args);
380 
381 		switch (token) {
382 		case Opt_gc_background:
383 			name = match_strdup(&args[0]);
384 
385 			if (!name)
386 				return -ENOMEM;
387 			if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
388 				set_opt(sbi, BG_GC);
389 				clear_opt(sbi, FORCE_FG_GC);
390 			} else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
391 				clear_opt(sbi, BG_GC);
392 				clear_opt(sbi, FORCE_FG_GC);
393 			} else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
394 				set_opt(sbi, BG_GC);
395 				set_opt(sbi, FORCE_FG_GC);
396 			} else {
397 				kfree(name);
398 				return -EINVAL;
399 			}
400 			kfree(name);
401 			break;
402 		case Opt_disable_roll_forward:
403 			set_opt(sbi, DISABLE_ROLL_FORWARD);
404 			break;
405 		case Opt_norecovery:
406 			/* this option mounts f2fs with ro */
407 			set_opt(sbi, DISABLE_ROLL_FORWARD);
408 			if (!f2fs_readonly(sb))
409 				return -EINVAL;
410 			break;
411 		case Opt_discard:
412 			q = bdev_get_queue(sb->s_bdev);
413 			if (blk_queue_discard(q)) {
414 				set_opt(sbi, DISCARD);
415 			} else {
416 				f2fs_msg(sb, KERN_WARNING,
417 					"mounting with \"discard\" option, but "
418 					"the device does not support discard");
419 			}
420 			break;
421 		case Opt_nodiscard:
422 			clear_opt(sbi, DISCARD);
423 		case Opt_noheap:
424 			set_opt(sbi, NOHEAP);
425 			break;
426 #ifdef CONFIG_F2FS_FS_XATTR
427 		case Opt_user_xattr:
428 			set_opt(sbi, XATTR_USER);
429 			break;
430 		case Opt_nouser_xattr:
431 			clear_opt(sbi, XATTR_USER);
432 			break;
433 		case Opt_inline_xattr:
434 			set_opt(sbi, INLINE_XATTR);
435 			break;
436 #else
437 		case Opt_user_xattr:
438 			f2fs_msg(sb, KERN_INFO,
439 				"user_xattr options not supported");
440 			break;
441 		case Opt_nouser_xattr:
442 			f2fs_msg(sb, KERN_INFO,
443 				"nouser_xattr options not supported");
444 			break;
445 		case Opt_inline_xattr:
446 			f2fs_msg(sb, KERN_INFO,
447 				"inline_xattr options not supported");
448 			break;
449 #endif
450 #ifdef CONFIG_F2FS_FS_POSIX_ACL
451 		case Opt_acl:
452 			set_opt(sbi, POSIX_ACL);
453 			break;
454 		case Opt_noacl:
455 			clear_opt(sbi, POSIX_ACL);
456 			break;
457 #else
458 		case Opt_acl:
459 			f2fs_msg(sb, KERN_INFO, "acl options not supported");
460 			break;
461 		case Opt_noacl:
462 			f2fs_msg(sb, KERN_INFO, "noacl options not supported");
463 			break;
464 #endif
465 		case Opt_active_logs:
466 			if (args->from && match_int(args, &arg))
467 				return -EINVAL;
468 			if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
469 				return -EINVAL;
470 			sbi->active_logs = arg;
471 			break;
472 		case Opt_disable_ext_identify:
473 			set_opt(sbi, DISABLE_EXT_IDENTIFY);
474 			break;
475 		case Opt_inline_data:
476 			set_opt(sbi, INLINE_DATA);
477 			break;
478 		case Opt_inline_dentry:
479 			set_opt(sbi, INLINE_DENTRY);
480 			break;
481 		case Opt_noinline_dentry:
482 			clear_opt(sbi, INLINE_DENTRY);
483 			break;
484 		case Opt_flush_merge:
485 			set_opt(sbi, FLUSH_MERGE);
486 			break;
487 		case Opt_noflush_merge:
488 			clear_opt(sbi, FLUSH_MERGE);
489 			break;
490 		case Opt_nobarrier:
491 			set_opt(sbi, NOBARRIER);
492 			break;
493 		case Opt_fastboot:
494 			set_opt(sbi, FASTBOOT);
495 			break;
496 		case Opt_extent_cache:
497 			set_opt(sbi, EXTENT_CACHE);
498 			break;
499 		case Opt_noextent_cache:
500 			clear_opt(sbi, EXTENT_CACHE);
501 			break;
502 		case Opt_noinline_data:
503 			clear_opt(sbi, INLINE_DATA);
504 			break;
505 		case Opt_data_flush:
506 			set_opt(sbi, DATA_FLUSH);
507 			break;
508 		case Opt_mode:
509 			name = match_strdup(&args[0]);
510 
511 			if (!name)
512 				return -ENOMEM;
513 			if (strlen(name) == 8 &&
514 					!strncmp(name, "adaptive", 8)) {
515 				set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
516 			} else if (strlen(name) == 3 &&
517 					!strncmp(name, "lfs", 3)) {
518 				set_opt_mode(sbi, F2FS_MOUNT_LFS);
519 			} else {
520 				kfree(name);
521 				return -EINVAL;
522 			}
523 			kfree(name);
524 			break;
525 		case Opt_fault_injection:
526 			if (args->from && match_int(args, &arg))
527 				return -EINVAL;
528 #ifdef CONFIG_F2FS_FAULT_INJECTION
529 			f2fs_build_fault_attr(sbi, arg);
530 #else
531 			f2fs_msg(sb, KERN_INFO,
532 				"FAULT_INJECTION was not selected");
533 #endif
534 			break;
535 		case Opt_lazytime:
536 			sb->s_flags |= MS_LAZYTIME;
537 			break;
538 		case Opt_nolazytime:
539 			sb->s_flags &= ~MS_LAZYTIME;
540 			break;
541 		default:
542 			f2fs_msg(sb, KERN_ERR,
543 				"Unrecognized mount option \"%s\" or missing value",
544 				p);
545 			return -EINVAL;
546 		}
547 	}
548 	return 0;
549 }
550 
551 static struct inode *f2fs_alloc_inode(struct super_block *sb)
552 {
553 	struct f2fs_inode_info *fi;
554 
555 	fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
556 	if (!fi)
557 		return NULL;
558 
559 	init_once((void *) fi);
560 
561 	if (percpu_counter_init(&fi->dirty_pages, 0, GFP_NOFS)) {
562 		kmem_cache_free(f2fs_inode_cachep, fi);
563 		return NULL;
564 	}
565 
566 	/* Initialize f2fs-specific inode info */
567 	fi->vfs_inode.i_version = 1;
568 	fi->i_current_depth = 1;
569 	fi->i_advise = 0;
570 	init_rwsem(&fi->i_sem);
571 	INIT_LIST_HEAD(&fi->dirty_list);
572 	INIT_LIST_HEAD(&fi->gdirty_list);
573 	INIT_LIST_HEAD(&fi->inmem_pages);
574 	mutex_init(&fi->inmem_lock);
575 	init_rwsem(&fi->dio_rwsem[READ]);
576 	init_rwsem(&fi->dio_rwsem[WRITE]);
577 
578 	/* Will be used by directory only */
579 	fi->i_dir_level = F2FS_SB(sb)->dir_level;
580 	return &fi->vfs_inode;
581 }
582 
583 static int f2fs_drop_inode(struct inode *inode)
584 {
585 	/*
586 	 * This is to avoid a deadlock condition like below.
587 	 * writeback_single_inode(inode)
588 	 *  - f2fs_write_data_page
589 	 *    - f2fs_gc -> iput -> evict
590 	 *       - inode_wait_for_writeback(inode)
591 	 */
592 	if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
593 		if (!inode->i_nlink && !is_bad_inode(inode)) {
594 			/* to avoid evict_inode call simultaneously */
595 			atomic_inc(&inode->i_count);
596 			spin_unlock(&inode->i_lock);
597 
598 			/* some remained atomic pages should discarded */
599 			if (f2fs_is_atomic_file(inode))
600 				drop_inmem_pages(inode);
601 
602 			/* should remain fi->extent_tree for writepage */
603 			f2fs_destroy_extent_node(inode);
604 
605 			sb_start_intwrite(inode->i_sb);
606 			f2fs_i_size_write(inode, 0);
607 
608 			if (F2FS_HAS_BLOCKS(inode))
609 				f2fs_truncate(inode);
610 
611 			sb_end_intwrite(inode->i_sb);
612 
613 			fscrypt_put_encryption_info(inode, NULL);
614 			spin_lock(&inode->i_lock);
615 			atomic_dec(&inode->i_count);
616 		}
617 		return 0;
618 	}
619 
620 	return generic_drop_inode(inode);
621 }
622 
623 int f2fs_inode_dirtied(struct inode *inode)
624 {
625 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
626 
627 	spin_lock(&sbi->inode_lock[DIRTY_META]);
628 	if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
629 		spin_unlock(&sbi->inode_lock[DIRTY_META]);
630 		return 1;
631 	}
632 
633 	set_inode_flag(inode, FI_DIRTY_INODE);
634 	list_add_tail(&F2FS_I(inode)->gdirty_list,
635 				&sbi->inode_list[DIRTY_META]);
636 	inc_page_count(sbi, F2FS_DIRTY_IMETA);
637 	stat_inc_dirty_inode(sbi, DIRTY_META);
638 	spin_unlock(&sbi->inode_lock[DIRTY_META]);
639 
640 	return 0;
641 }
642 
643 void f2fs_inode_synced(struct inode *inode)
644 {
645 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
646 
647 	spin_lock(&sbi->inode_lock[DIRTY_META]);
648 	if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
649 		spin_unlock(&sbi->inode_lock[DIRTY_META]);
650 		return;
651 	}
652 	list_del_init(&F2FS_I(inode)->gdirty_list);
653 	clear_inode_flag(inode, FI_DIRTY_INODE);
654 	clear_inode_flag(inode, FI_AUTO_RECOVER);
655 	dec_page_count(sbi, F2FS_DIRTY_IMETA);
656 	stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
657 	spin_unlock(&sbi->inode_lock[DIRTY_META]);
658 }
659 
660 /*
661  * f2fs_dirty_inode() is called from __mark_inode_dirty()
662  *
663  * We should call set_dirty_inode to write the dirty inode through write_inode.
664  */
665 static void f2fs_dirty_inode(struct inode *inode, int flags)
666 {
667 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
668 
669 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
670 			inode->i_ino == F2FS_META_INO(sbi))
671 		return;
672 
673 	if (flags == I_DIRTY_TIME)
674 		return;
675 
676 	if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
677 		clear_inode_flag(inode, FI_AUTO_RECOVER);
678 
679 	f2fs_inode_dirtied(inode);
680 }
681 
682 static void f2fs_i_callback(struct rcu_head *head)
683 {
684 	struct inode *inode = container_of(head, struct inode, i_rcu);
685 	kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
686 }
687 
688 static void f2fs_destroy_inode(struct inode *inode)
689 {
690 	percpu_counter_destroy(&F2FS_I(inode)->dirty_pages);
691 	call_rcu(&inode->i_rcu, f2fs_i_callback);
692 }
693 
694 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
695 {
696 	int i;
697 
698 	for (i = 0; i < NR_COUNT_TYPE; i++)
699 		percpu_counter_destroy(&sbi->nr_pages[i]);
700 	percpu_counter_destroy(&sbi->alloc_valid_block_count);
701 	percpu_counter_destroy(&sbi->total_valid_inode_count);
702 }
703 
704 static void f2fs_put_super(struct super_block *sb)
705 {
706 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
707 
708 	if (sbi->s_proc) {
709 		remove_proc_entry("segment_info", sbi->s_proc);
710 		remove_proc_entry("segment_bits", sbi->s_proc);
711 		remove_proc_entry(sb->s_id, f2fs_proc_root);
712 	}
713 	kobject_del(&sbi->s_kobj);
714 
715 	stop_gc_thread(sbi);
716 
717 	/* prevent remaining shrinker jobs */
718 	mutex_lock(&sbi->umount_mutex);
719 
720 	/*
721 	 * We don't need to do checkpoint when superblock is clean.
722 	 * But, the previous checkpoint was not done by umount, it needs to do
723 	 * clean checkpoint again.
724 	 */
725 	if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
726 			!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
727 		struct cp_control cpc = {
728 			.reason = CP_UMOUNT,
729 		};
730 		write_checkpoint(sbi, &cpc);
731 	}
732 
733 	/* write_checkpoint can update stat informaion */
734 	f2fs_destroy_stats(sbi);
735 
736 	/*
737 	 * normally superblock is clean, so we need to release this.
738 	 * In addition, EIO will skip do checkpoint, we need this as well.
739 	 */
740 	release_ino_entry(sbi, true);
741 	release_discard_addrs(sbi);
742 
743 	f2fs_leave_shrinker(sbi);
744 	mutex_unlock(&sbi->umount_mutex);
745 
746 	/* our cp_error case, we can wait for any writeback page */
747 	f2fs_flush_merged_bios(sbi);
748 
749 	iput(sbi->node_inode);
750 	iput(sbi->meta_inode);
751 
752 	/* destroy f2fs internal modules */
753 	destroy_node_manager(sbi);
754 	destroy_segment_manager(sbi);
755 
756 	kfree(sbi->ckpt);
757 	kobject_put(&sbi->s_kobj);
758 	wait_for_completion(&sbi->s_kobj_unregister);
759 
760 	sb->s_fs_info = NULL;
761 	if (sbi->s_chksum_driver)
762 		crypto_free_shash(sbi->s_chksum_driver);
763 	kfree(sbi->raw_super);
764 
765 	destroy_percpu_info(sbi);
766 	kfree(sbi);
767 }
768 
769 int f2fs_sync_fs(struct super_block *sb, int sync)
770 {
771 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
772 	int err = 0;
773 
774 	trace_f2fs_sync_fs(sb, sync);
775 
776 	if (sync) {
777 		struct cp_control cpc;
778 
779 		cpc.reason = __get_cp_reason(sbi);
780 
781 		mutex_lock(&sbi->gc_mutex);
782 		err = write_checkpoint(sbi, &cpc);
783 		mutex_unlock(&sbi->gc_mutex);
784 	}
785 	f2fs_trace_ios(NULL, 1);
786 
787 	return err;
788 }
789 
790 static int f2fs_freeze(struct super_block *sb)
791 {
792 	int err;
793 
794 	if (f2fs_readonly(sb))
795 		return 0;
796 
797 	err = f2fs_sync_fs(sb, 1);
798 	return err;
799 }
800 
801 static int f2fs_unfreeze(struct super_block *sb)
802 {
803 	return 0;
804 }
805 
806 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
807 {
808 	struct super_block *sb = dentry->d_sb;
809 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
810 	u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
811 	block_t total_count, user_block_count, start_count, ovp_count;
812 
813 	total_count = le64_to_cpu(sbi->raw_super->block_count);
814 	user_block_count = sbi->user_block_count;
815 	start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
816 	ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
817 	buf->f_type = F2FS_SUPER_MAGIC;
818 	buf->f_bsize = sbi->blocksize;
819 
820 	buf->f_blocks = total_count - start_count;
821 	buf->f_bfree = user_block_count - valid_user_blocks(sbi) + ovp_count;
822 	buf->f_bavail = user_block_count - valid_user_blocks(sbi);
823 
824 	buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
825 	buf->f_ffree = buf->f_files - valid_inode_count(sbi);
826 
827 	buf->f_namelen = F2FS_NAME_LEN;
828 	buf->f_fsid.val[0] = (u32)id;
829 	buf->f_fsid.val[1] = (u32)(id >> 32);
830 
831 	return 0;
832 }
833 
834 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
835 {
836 	struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
837 
838 	if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
839 		if (test_opt(sbi, FORCE_FG_GC))
840 			seq_printf(seq, ",background_gc=%s", "sync");
841 		else
842 			seq_printf(seq, ",background_gc=%s", "on");
843 	} else {
844 		seq_printf(seq, ",background_gc=%s", "off");
845 	}
846 	if (test_opt(sbi, DISABLE_ROLL_FORWARD))
847 		seq_puts(seq, ",disable_roll_forward");
848 	if (test_opt(sbi, DISCARD))
849 		seq_puts(seq, ",discard");
850 	if (test_opt(sbi, NOHEAP))
851 		seq_puts(seq, ",no_heap_alloc");
852 #ifdef CONFIG_F2FS_FS_XATTR
853 	if (test_opt(sbi, XATTR_USER))
854 		seq_puts(seq, ",user_xattr");
855 	else
856 		seq_puts(seq, ",nouser_xattr");
857 	if (test_opt(sbi, INLINE_XATTR))
858 		seq_puts(seq, ",inline_xattr");
859 #endif
860 #ifdef CONFIG_F2FS_FS_POSIX_ACL
861 	if (test_opt(sbi, POSIX_ACL))
862 		seq_puts(seq, ",acl");
863 	else
864 		seq_puts(seq, ",noacl");
865 #endif
866 	if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
867 		seq_puts(seq, ",disable_ext_identify");
868 	if (test_opt(sbi, INLINE_DATA))
869 		seq_puts(seq, ",inline_data");
870 	else
871 		seq_puts(seq, ",noinline_data");
872 	if (test_opt(sbi, INLINE_DENTRY))
873 		seq_puts(seq, ",inline_dentry");
874 	else
875 		seq_puts(seq, ",noinline_dentry");
876 	if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
877 		seq_puts(seq, ",flush_merge");
878 	if (test_opt(sbi, NOBARRIER))
879 		seq_puts(seq, ",nobarrier");
880 	if (test_opt(sbi, FASTBOOT))
881 		seq_puts(seq, ",fastboot");
882 	if (test_opt(sbi, EXTENT_CACHE))
883 		seq_puts(seq, ",extent_cache");
884 	else
885 		seq_puts(seq, ",noextent_cache");
886 	if (test_opt(sbi, DATA_FLUSH))
887 		seq_puts(seq, ",data_flush");
888 
889 	seq_puts(seq, ",mode=");
890 	if (test_opt(sbi, ADAPTIVE))
891 		seq_puts(seq, "adaptive");
892 	else if (test_opt(sbi, LFS))
893 		seq_puts(seq, "lfs");
894 	seq_printf(seq, ",active_logs=%u", sbi->active_logs);
895 
896 	return 0;
897 }
898 
899 static int segment_info_seq_show(struct seq_file *seq, void *offset)
900 {
901 	struct super_block *sb = seq->private;
902 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
903 	unsigned int total_segs =
904 			le32_to_cpu(sbi->raw_super->segment_count_main);
905 	int i;
906 
907 	seq_puts(seq, "format: segment_type|valid_blocks\n"
908 		"segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
909 
910 	for (i = 0; i < total_segs; i++) {
911 		struct seg_entry *se = get_seg_entry(sbi, i);
912 
913 		if ((i % 10) == 0)
914 			seq_printf(seq, "%-10d", i);
915 		seq_printf(seq, "%d|%-3u", se->type,
916 					get_valid_blocks(sbi, i, 1));
917 		if ((i % 10) == 9 || i == (total_segs - 1))
918 			seq_putc(seq, '\n');
919 		else
920 			seq_putc(seq, ' ');
921 	}
922 
923 	return 0;
924 }
925 
926 static int segment_bits_seq_show(struct seq_file *seq, void *offset)
927 {
928 	struct super_block *sb = seq->private;
929 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
930 	unsigned int total_segs =
931 			le32_to_cpu(sbi->raw_super->segment_count_main);
932 	int i, j;
933 
934 	seq_puts(seq, "format: segment_type|valid_blocks|bitmaps\n"
935 		"segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
936 
937 	for (i = 0; i < total_segs; i++) {
938 		struct seg_entry *se = get_seg_entry(sbi, i);
939 
940 		seq_printf(seq, "%-10d", i);
941 		seq_printf(seq, "%d|%-3u|", se->type,
942 					get_valid_blocks(sbi, i, 1));
943 		for (j = 0; j < SIT_VBLOCK_MAP_SIZE; j++)
944 			seq_printf(seq, " %.2x", se->cur_valid_map[j]);
945 		seq_putc(seq, '\n');
946 	}
947 	return 0;
948 }
949 
950 #define F2FS_PROC_FILE_DEF(_name)					\
951 static int _name##_open_fs(struct inode *inode, struct file *file)	\
952 {									\
953 	return single_open(file, _name##_seq_show, PDE_DATA(inode));	\
954 }									\
955 									\
956 static const struct file_operations f2fs_seq_##_name##_fops = {		\
957 	.open = _name##_open_fs,					\
958 	.read = seq_read,						\
959 	.llseek = seq_lseek,						\
960 	.release = single_release,					\
961 };
962 
963 F2FS_PROC_FILE_DEF(segment_info);
964 F2FS_PROC_FILE_DEF(segment_bits);
965 
966 static void default_options(struct f2fs_sb_info *sbi)
967 {
968 	/* init some FS parameters */
969 	sbi->active_logs = NR_CURSEG_TYPE;
970 
971 	set_opt(sbi, BG_GC);
972 	set_opt(sbi, INLINE_DATA);
973 	set_opt(sbi, INLINE_DENTRY);
974 	set_opt(sbi, EXTENT_CACHE);
975 	sbi->sb->s_flags |= MS_LAZYTIME;
976 	set_opt(sbi, FLUSH_MERGE);
977 	if (f2fs_sb_mounted_hmsmr(sbi->sb)) {
978 		set_opt_mode(sbi, F2FS_MOUNT_LFS);
979 		set_opt(sbi, DISCARD);
980 	} else {
981 		set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
982 	}
983 
984 #ifdef CONFIG_F2FS_FS_XATTR
985 	set_opt(sbi, XATTR_USER);
986 #endif
987 #ifdef CONFIG_F2FS_FS_POSIX_ACL
988 	set_opt(sbi, POSIX_ACL);
989 #endif
990 
991 #ifdef CONFIG_F2FS_FAULT_INJECTION
992 	f2fs_build_fault_attr(sbi, 0);
993 #endif
994 }
995 
996 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
997 {
998 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
999 	struct f2fs_mount_info org_mount_opt;
1000 	int err, active_logs;
1001 	bool need_restart_gc = false;
1002 	bool need_stop_gc = false;
1003 	bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1004 #ifdef CONFIG_F2FS_FAULT_INJECTION
1005 	struct f2fs_fault_info ffi = sbi->fault_info;
1006 #endif
1007 
1008 	/*
1009 	 * Save the old mount options in case we
1010 	 * need to restore them.
1011 	 */
1012 	org_mount_opt = sbi->mount_opt;
1013 	active_logs = sbi->active_logs;
1014 
1015 	/* recover superblocks we couldn't write due to previous RO mount */
1016 	if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1017 		err = f2fs_commit_super(sbi, false);
1018 		f2fs_msg(sb, KERN_INFO,
1019 			"Try to recover all the superblocks, ret: %d", err);
1020 		if (!err)
1021 			clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1022 	}
1023 
1024 	sbi->mount_opt.opt = 0;
1025 	default_options(sbi);
1026 
1027 	/* parse mount options */
1028 	err = parse_options(sb, data);
1029 	if (err)
1030 		goto restore_opts;
1031 
1032 	/*
1033 	 * Previous and new state of filesystem is RO,
1034 	 * so skip checking GC and FLUSH_MERGE conditions.
1035 	 */
1036 	if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
1037 		goto skip;
1038 
1039 	/* disallow enable/disable extent_cache dynamically */
1040 	if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1041 		err = -EINVAL;
1042 		f2fs_msg(sbi->sb, KERN_WARNING,
1043 				"switch extent_cache option is not allowed");
1044 		goto restore_opts;
1045 	}
1046 
1047 	/*
1048 	 * We stop the GC thread if FS is mounted as RO
1049 	 * or if background_gc = off is passed in mount
1050 	 * option. Also sync the filesystem.
1051 	 */
1052 	if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
1053 		if (sbi->gc_thread) {
1054 			stop_gc_thread(sbi);
1055 			need_restart_gc = true;
1056 		}
1057 	} else if (!sbi->gc_thread) {
1058 		err = start_gc_thread(sbi);
1059 		if (err)
1060 			goto restore_opts;
1061 		need_stop_gc = true;
1062 	}
1063 
1064 	if (*flags & MS_RDONLY) {
1065 		writeback_inodes_sb(sb, WB_REASON_SYNC);
1066 		sync_inodes_sb(sb);
1067 
1068 		set_sbi_flag(sbi, SBI_IS_DIRTY);
1069 		set_sbi_flag(sbi, SBI_IS_CLOSE);
1070 		f2fs_sync_fs(sb, 1);
1071 		clear_sbi_flag(sbi, SBI_IS_CLOSE);
1072 	}
1073 
1074 	/*
1075 	 * We stop issue flush thread if FS is mounted as RO
1076 	 * or if flush_merge is not passed in mount option.
1077 	 */
1078 	if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1079 		destroy_flush_cmd_control(sbi);
1080 	} else if (!SM_I(sbi)->cmd_control_info) {
1081 		err = create_flush_cmd_control(sbi);
1082 		if (err)
1083 			goto restore_gc;
1084 	}
1085 skip:
1086 	/* Update the POSIXACL Flag */
1087 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1088 		(test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1089 
1090 	return 0;
1091 restore_gc:
1092 	if (need_restart_gc) {
1093 		if (start_gc_thread(sbi))
1094 			f2fs_msg(sbi->sb, KERN_WARNING,
1095 				"background gc thread has stopped");
1096 	} else if (need_stop_gc) {
1097 		stop_gc_thread(sbi);
1098 	}
1099 restore_opts:
1100 	sbi->mount_opt = org_mount_opt;
1101 	sbi->active_logs = active_logs;
1102 #ifdef CONFIG_F2FS_FAULT_INJECTION
1103 	sbi->fault_info = ffi;
1104 #endif
1105 	return err;
1106 }
1107 
1108 static struct super_operations f2fs_sops = {
1109 	.alloc_inode	= f2fs_alloc_inode,
1110 	.drop_inode	= f2fs_drop_inode,
1111 	.destroy_inode	= f2fs_destroy_inode,
1112 	.write_inode	= f2fs_write_inode,
1113 	.dirty_inode	= f2fs_dirty_inode,
1114 	.show_options	= f2fs_show_options,
1115 	.evict_inode	= f2fs_evict_inode,
1116 	.put_super	= f2fs_put_super,
1117 	.sync_fs	= f2fs_sync_fs,
1118 	.freeze_fs	= f2fs_freeze,
1119 	.unfreeze_fs	= f2fs_unfreeze,
1120 	.statfs		= f2fs_statfs,
1121 	.remount_fs	= f2fs_remount,
1122 };
1123 
1124 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1125 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1126 {
1127 	return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1128 				F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1129 				ctx, len, NULL);
1130 }
1131 
1132 static int f2fs_key_prefix(struct inode *inode, u8 **key)
1133 {
1134 	*key = F2FS_I_SB(inode)->key_prefix;
1135 	return F2FS_I_SB(inode)->key_prefix_size;
1136 }
1137 
1138 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1139 							void *fs_data)
1140 {
1141 	return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1142 				F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1143 				ctx, len, fs_data, XATTR_CREATE);
1144 }
1145 
1146 static unsigned f2fs_max_namelen(struct inode *inode)
1147 {
1148 	return S_ISLNK(inode->i_mode) ?
1149 			inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1150 }
1151 
1152 static struct fscrypt_operations f2fs_cryptops = {
1153 	.get_context	= f2fs_get_context,
1154 	.key_prefix	= f2fs_key_prefix,
1155 	.set_context	= f2fs_set_context,
1156 	.is_encrypted	= f2fs_encrypted_inode,
1157 	.empty_dir	= f2fs_empty_dir,
1158 	.max_namelen	= f2fs_max_namelen,
1159 };
1160 #else
1161 static struct fscrypt_operations f2fs_cryptops = {
1162 	.is_encrypted	= f2fs_encrypted_inode,
1163 };
1164 #endif
1165 
1166 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1167 		u64 ino, u32 generation)
1168 {
1169 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1170 	struct inode *inode;
1171 
1172 	if (check_nid_range(sbi, ino))
1173 		return ERR_PTR(-ESTALE);
1174 
1175 	/*
1176 	 * f2fs_iget isn't quite right if the inode is currently unallocated!
1177 	 * However f2fs_iget currently does appropriate checks to handle stale
1178 	 * inodes so everything is OK.
1179 	 */
1180 	inode = f2fs_iget(sb, ino);
1181 	if (IS_ERR(inode))
1182 		return ERR_CAST(inode);
1183 	if (unlikely(generation && inode->i_generation != generation)) {
1184 		/* we didn't find the right inode.. */
1185 		iput(inode);
1186 		return ERR_PTR(-ESTALE);
1187 	}
1188 	return inode;
1189 }
1190 
1191 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1192 		int fh_len, int fh_type)
1193 {
1194 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1195 				    f2fs_nfs_get_inode);
1196 }
1197 
1198 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1199 		int fh_len, int fh_type)
1200 {
1201 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1202 				    f2fs_nfs_get_inode);
1203 }
1204 
1205 static const struct export_operations f2fs_export_ops = {
1206 	.fh_to_dentry = f2fs_fh_to_dentry,
1207 	.fh_to_parent = f2fs_fh_to_parent,
1208 	.get_parent = f2fs_get_parent,
1209 };
1210 
1211 static loff_t max_file_blocks(void)
1212 {
1213 	loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
1214 	loff_t leaf_count = ADDRS_PER_BLOCK;
1215 
1216 	/* two direct node blocks */
1217 	result += (leaf_count * 2);
1218 
1219 	/* two indirect node blocks */
1220 	leaf_count *= NIDS_PER_BLOCK;
1221 	result += (leaf_count * 2);
1222 
1223 	/* one double indirect node block */
1224 	leaf_count *= NIDS_PER_BLOCK;
1225 	result += leaf_count;
1226 
1227 	return result;
1228 }
1229 
1230 static int __f2fs_commit_super(struct buffer_head *bh,
1231 			struct f2fs_super_block *super)
1232 {
1233 	lock_buffer(bh);
1234 	if (super)
1235 		memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
1236 	set_buffer_uptodate(bh);
1237 	set_buffer_dirty(bh);
1238 	unlock_buffer(bh);
1239 
1240 	/* it's rare case, we can do fua all the time */
1241 	return __sync_dirty_buffer(bh, WRITE_FLUSH_FUA);
1242 }
1243 
1244 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
1245 					struct buffer_head *bh)
1246 {
1247 	struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1248 					(bh->b_data + F2FS_SUPER_OFFSET);
1249 	struct super_block *sb = sbi->sb;
1250 	u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
1251 	u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
1252 	u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
1253 	u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
1254 	u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
1255 	u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
1256 	u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
1257 	u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
1258 	u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
1259 	u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
1260 	u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1261 	u32 segment_count = le32_to_cpu(raw_super->segment_count);
1262 	u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1263 	u64 main_end_blkaddr = main_blkaddr +
1264 				(segment_count_main << log_blocks_per_seg);
1265 	u64 seg_end_blkaddr = segment0_blkaddr +
1266 				(segment_count << log_blocks_per_seg);
1267 
1268 	if (segment0_blkaddr != cp_blkaddr) {
1269 		f2fs_msg(sb, KERN_INFO,
1270 			"Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1271 			segment0_blkaddr, cp_blkaddr);
1272 		return true;
1273 	}
1274 
1275 	if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
1276 							sit_blkaddr) {
1277 		f2fs_msg(sb, KERN_INFO,
1278 			"Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1279 			cp_blkaddr, sit_blkaddr,
1280 			segment_count_ckpt << log_blocks_per_seg);
1281 		return true;
1282 	}
1283 
1284 	if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
1285 							nat_blkaddr) {
1286 		f2fs_msg(sb, KERN_INFO,
1287 			"Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1288 			sit_blkaddr, nat_blkaddr,
1289 			segment_count_sit << log_blocks_per_seg);
1290 		return true;
1291 	}
1292 
1293 	if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
1294 							ssa_blkaddr) {
1295 		f2fs_msg(sb, KERN_INFO,
1296 			"Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1297 			nat_blkaddr, ssa_blkaddr,
1298 			segment_count_nat << log_blocks_per_seg);
1299 		return true;
1300 	}
1301 
1302 	if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
1303 							main_blkaddr) {
1304 		f2fs_msg(sb, KERN_INFO,
1305 			"Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1306 			ssa_blkaddr, main_blkaddr,
1307 			segment_count_ssa << log_blocks_per_seg);
1308 		return true;
1309 	}
1310 
1311 	if (main_end_blkaddr > seg_end_blkaddr) {
1312 		f2fs_msg(sb, KERN_INFO,
1313 			"Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1314 			main_blkaddr,
1315 			segment0_blkaddr +
1316 				(segment_count << log_blocks_per_seg),
1317 			segment_count_main << log_blocks_per_seg);
1318 		return true;
1319 	} else if (main_end_blkaddr < seg_end_blkaddr) {
1320 		int err = 0;
1321 		char *res;
1322 
1323 		/* fix in-memory information all the time */
1324 		raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
1325 				segment0_blkaddr) >> log_blocks_per_seg);
1326 
1327 		if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
1328 			set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1329 			res = "internally";
1330 		} else {
1331 			err = __f2fs_commit_super(bh, NULL);
1332 			res = err ? "failed" : "done";
1333 		}
1334 		f2fs_msg(sb, KERN_INFO,
1335 			"Fix alignment : %s, start(%u) end(%u) block(%u)",
1336 			res, main_blkaddr,
1337 			segment0_blkaddr +
1338 				(segment_count << log_blocks_per_seg),
1339 			segment_count_main << log_blocks_per_seg);
1340 		if (err)
1341 			return true;
1342 	}
1343 	return false;
1344 }
1345 
1346 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
1347 				struct buffer_head *bh)
1348 {
1349 	struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1350 					(bh->b_data + F2FS_SUPER_OFFSET);
1351 	struct super_block *sb = sbi->sb;
1352 	unsigned int blocksize;
1353 
1354 	if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
1355 		f2fs_msg(sb, KERN_INFO,
1356 			"Magic Mismatch, valid(0x%x) - read(0x%x)",
1357 			F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
1358 		return 1;
1359 	}
1360 
1361 	/* Currently, support only 4KB page cache size */
1362 	if (F2FS_BLKSIZE != PAGE_SIZE) {
1363 		f2fs_msg(sb, KERN_INFO,
1364 			"Invalid page_cache_size (%lu), supports only 4KB\n",
1365 			PAGE_SIZE);
1366 		return 1;
1367 	}
1368 
1369 	/* Currently, support only 4KB block size */
1370 	blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
1371 	if (blocksize != F2FS_BLKSIZE) {
1372 		f2fs_msg(sb, KERN_INFO,
1373 			"Invalid blocksize (%u), supports only 4KB\n",
1374 			blocksize);
1375 		return 1;
1376 	}
1377 
1378 	/* check log blocks per segment */
1379 	if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
1380 		f2fs_msg(sb, KERN_INFO,
1381 			"Invalid log blocks per segment (%u)\n",
1382 			le32_to_cpu(raw_super->log_blocks_per_seg));
1383 		return 1;
1384 	}
1385 
1386 	/* Currently, support 512/1024/2048/4096 bytes sector size */
1387 	if (le32_to_cpu(raw_super->log_sectorsize) >
1388 				F2FS_MAX_LOG_SECTOR_SIZE ||
1389 		le32_to_cpu(raw_super->log_sectorsize) <
1390 				F2FS_MIN_LOG_SECTOR_SIZE) {
1391 		f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
1392 			le32_to_cpu(raw_super->log_sectorsize));
1393 		return 1;
1394 	}
1395 	if (le32_to_cpu(raw_super->log_sectors_per_block) +
1396 		le32_to_cpu(raw_super->log_sectorsize) !=
1397 			F2FS_MAX_LOG_SECTOR_SIZE) {
1398 		f2fs_msg(sb, KERN_INFO,
1399 			"Invalid log sectors per block(%u) log sectorsize(%u)",
1400 			le32_to_cpu(raw_super->log_sectors_per_block),
1401 			le32_to_cpu(raw_super->log_sectorsize));
1402 		return 1;
1403 	}
1404 
1405 	/* check reserved ino info */
1406 	if (le32_to_cpu(raw_super->node_ino) != 1 ||
1407 		le32_to_cpu(raw_super->meta_ino) != 2 ||
1408 		le32_to_cpu(raw_super->root_ino) != 3) {
1409 		f2fs_msg(sb, KERN_INFO,
1410 			"Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1411 			le32_to_cpu(raw_super->node_ino),
1412 			le32_to_cpu(raw_super->meta_ino),
1413 			le32_to_cpu(raw_super->root_ino));
1414 		return 1;
1415 	}
1416 
1417 	/* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1418 	if (sanity_check_area_boundary(sbi, bh))
1419 		return 1;
1420 
1421 	return 0;
1422 }
1423 
1424 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1425 {
1426 	unsigned int total, fsmeta;
1427 	struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1428 	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1429 
1430 	total = le32_to_cpu(raw_super->segment_count);
1431 	fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
1432 	fsmeta += le32_to_cpu(raw_super->segment_count_sit);
1433 	fsmeta += le32_to_cpu(raw_super->segment_count_nat);
1434 	fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
1435 	fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
1436 
1437 	if (unlikely(fsmeta >= total))
1438 		return 1;
1439 
1440 	if (unlikely(f2fs_cp_error(sbi))) {
1441 		f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
1442 		return 1;
1443 	}
1444 	return 0;
1445 }
1446 
1447 static void init_sb_info(struct f2fs_sb_info *sbi)
1448 {
1449 	struct f2fs_super_block *raw_super = sbi->raw_super;
1450 
1451 	sbi->log_sectors_per_block =
1452 		le32_to_cpu(raw_super->log_sectors_per_block);
1453 	sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
1454 	sbi->blocksize = 1 << sbi->log_blocksize;
1455 	sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1456 	sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
1457 	sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1458 	sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1459 	sbi->total_sections = le32_to_cpu(raw_super->section_count);
1460 	sbi->total_node_count =
1461 		(le32_to_cpu(raw_super->segment_count_nat) / 2)
1462 			* sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
1463 	sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
1464 	sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
1465 	sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
1466 	sbi->cur_victim_sec = NULL_SECNO;
1467 	sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
1468 
1469 	sbi->dir_level = DEF_DIR_LEVEL;
1470 	sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
1471 	sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
1472 	clear_sbi_flag(sbi, SBI_NEED_FSCK);
1473 
1474 	INIT_LIST_HEAD(&sbi->s_list);
1475 	mutex_init(&sbi->umount_mutex);
1476 	mutex_init(&sbi->wio_mutex[NODE]);
1477 	mutex_init(&sbi->wio_mutex[DATA]);
1478 	spin_lock_init(&sbi->cp_lock);
1479 
1480 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1481 	memcpy(sbi->key_prefix, F2FS_KEY_DESC_PREFIX,
1482 				F2FS_KEY_DESC_PREFIX_SIZE);
1483 	sbi->key_prefix_size = F2FS_KEY_DESC_PREFIX_SIZE;
1484 #endif
1485 }
1486 
1487 static int init_percpu_info(struct f2fs_sb_info *sbi)
1488 {
1489 	int i, err;
1490 
1491 	for (i = 0; i < NR_COUNT_TYPE; i++) {
1492 		err = percpu_counter_init(&sbi->nr_pages[i], 0, GFP_KERNEL);
1493 		if (err)
1494 			return err;
1495 	}
1496 
1497 	err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
1498 	if (err)
1499 		return err;
1500 
1501 	return percpu_counter_init(&sbi->total_valid_inode_count, 0,
1502 								GFP_KERNEL);
1503 }
1504 
1505 /*
1506  * Read f2fs raw super block.
1507  * Because we have two copies of super block, so read both of them
1508  * to get the first valid one. If any one of them is broken, we pass
1509  * them recovery flag back to the caller.
1510  */
1511 static int read_raw_super_block(struct f2fs_sb_info *sbi,
1512 			struct f2fs_super_block **raw_super,
1513 			int *valid_super_block, int *recovery)
1514 {
1515 	struct super_block *sb = sbi->sb;
1516 	int block;
1517 	struct buffer_head *bh;
1518 	struct f2fs_super_block *super;
1519 	int err = 0;
1520 
1521 	super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
1522 	if (!super)
1523 		return -ENOMEM;
1524 
1525 	for (block = 0; block < 2; block++) {
1526 		bh = sb_bread(sb, block);
1527 		if (!bh) {
1528 			f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
1529 				block + 1);
1530 			err = -EIO;
1531 			continue;
1532 		}
1533 
1534 		/* sanity checking of raw super */
1535 		if (sanity_check_raw_super(sbi, bh)) {
1536 			f2fs_msg(sb, KERN_ERR,
1537 				"Can't find valid F2FS filesystem in %dth superblock",
1538 				block + 1);
1539 			err = -EINVAL;
1540 			brelse(bh);
1541 			continue;
1542 		}
1543 
1544 		if (!*raw_super) {
1545 			memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
1546 							sizeof(*super));
1547 			*valid_super_block = block;
1548 			*raw_super = super;
1549 		}
1550 		brelse(bh);
1551 	}
1552 
1553 	/* Fail to read any one of the superblocks*/
1554 	if (err < 0)
1555 		*recovery = 1;
1556 
1557 	/* No valid superblock */
1558 	if (!*raw_super)
1559 		kfree(super);
1560 	else
1561 		err = 0;
1562 
1563 	return err;
1564 }
1565 
1566 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
1567 {
1568 	struct buffer_head *bh;
1569 	int err;
1570 
1571 	if ((recover && f2fs_readonly(sbi->sb)) ||
1572 				bdev_read_only(sbi->sb->s_bdev)) {
1573 		set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1574 		return -EROFS;
1575 	}
1576 
1577 	/* write back-up superblock first */
1578 	bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1);
1579 	if (!bh)
1580 		return -EIO;
1581 	err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1582 	brelse(bh);
1583 
1584 	/* if we are in recovery path, skip writing valid superblock */
1585 	if (recover || err)
1586 		return err;
1587 
1588 	/* write current valid superblock */
1589 	bh = sb_getblk(sbi->sb, sbi->valid_super_block);
1590 	if (!bh)
1591 		return -EIO;
1592 	err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1593 	brelse(bh);
1594 	return err;
1595 }
1596 
1597 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
1598 {
1599 	struct f2fs_sb_info *sbi;
1600 	struct f2fs_super_block *raw_super;
1601 	struct inode *root;
1602 	int err;
1603 	bool retry = true, need_fsck = false;
1604 	char *options = NULL;
1605 	int recovery, i, valid_super_block;
1606 	struct curseg_info *seg_i;
1607 
1608 try_onemore:
1609 	err = -EINVAL;
1610 	raw_super = NULL;
1611 	valid_super_block = -1;
1612 	recovery = 0;
1613 
1614 	/* allocate memory for f2fs-specific super block info */
1615 	sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
1616 	if (!sbi)
1617 		return -ENOMEM;
1618 
1619 	sbi->sb = sb;
1620 
1621 	/* Load the checksum driver */
1622 	sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
1623 	if (IS_ERR(sbi->s_chksum_driver)) {
1624 		f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
1625 		err = PTR_ERR(sbi->s_chksum_driver);
1626 		sbi->s_chksum_driver = NULL;
1627 		goto free_sbi;
1628 	}
1629 
1630 	/* set a block size */
1631 	if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
1632 		f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
1633 		goto free_sbi;
1634 	}
1635 
1636 	err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
1637 								&recovery);
1638 	if (err)
1639 		goto free_sbi;
1640 
1641 	sb->s_fs_info = sbi;
1642 	sbi->raw_super = raw_super;
1643 
1644 	default_options(sbi);
1645 	/* parse mount options */
1646 	options = kstrdup((const char *)data, GFP_KERNEL);
1647 	if (data && !options) {
1648 		err = -ENOMEM;
1649 		goto free_sb_buf;
1650 	}
1651 
1652 	err = parse_options(sb, options);
1653 	if (err)
1654 		goto free_options;
1655 
1656 	sbi->max_file_blocks = max_file_blocks();
1657 	sb->s_maxbytes = sbi->max_file_blocks <<
1658 				le32_to_cpu(raw_super->log_blocksize);
1659 	sb->s_max_links = F2FS_LINK_MAX;
1660 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1661 
1662 	sb->s_op = &f2fs_sops;
1663 	sb->s_cop = &f2fs_cryptops;
1664 	sb->s_xattr = f2fs_xattr_handlers;
1665 	sb->s_export_op = &f2fs_export_ops;
1666 	sb->s_magic = F2FS_SUPER_MAGIC;
1667 	sb->s_time_gran = 1;
1668 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1669 		(test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1670 	memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1671 
1672 	/* init f2fs-specific super block info */
1673 	sbi->valid_super_block = valid_super_block;
1674 	mutex_init(&sbi->gc_mutex);
1675 	mutex_init(&sbi->cp_mutex);
1676 	init_rwsem(&sbi->node_write);
1677 
1678 	/* disallow all the data/node/meta page writes */
1679 	set_sbi_flag(sbi, SBI_POR_DOING);
1680 	spin_lock_init(&sbi->stat_lock);
1681 
1682 	init_rwsem(&sbi->read_io.io_rwsem);
1683 	sbi->read_io.sbi = sbi;
1684 	sbi->read_io.bio = NULL;
1685 	for (i = 0; i < NR_PAGE_TYPE; i++) {
1686 		init_rwsem(&sbi->write_io[i].io_rwsem);
1687 		sbi->write_io[i].sbi = sbi;
1688 		sbi->write_io[i].bio = NULL;
1689 	}
1690 
1691 	init_rwsem(&sbi->cp_rwsem);
1692 	init_waitqueue_head(&sbi->cp_wait);
1693 	init_sb_info(sbi);
1694 
1695 	err = init_percpu_info(sbi);
1696 	if (err)
1697 		goto free_options;
1698 
1699 	/* get an inode for meta space */
1700 	sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1701 	if (IS_ERR(sbi->meta_inode)) {
1702 		f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1703 		err = PTR_ERR(sbi->meta_inode);
1704 		goto free_options;
1705 	}
1706 
1707 	err = get_valid_checkpoint(sbi);
1708 	if (err) {
1709 		f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1710 		goto free_meta_inode;
1711 	}
1712 
1713 	sbi->total_valid_node_count =
1714 				le32_to_cpu(sbi->ckpt->valid_node_count);
1715 	percpu_counter_set(&sbi->total_valid_inode_count,
1716 				le32_to_cpu(sbi->ckpt->valid_inode_count));
1717 	sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1718 	sbi->total_valid_block_count =
1719 				le64_to_cpu(sbi->ckpt->valid_block_count);
1720 	sbi->last_valid_block_count = sbi->total_valid_block_count;
1721 
1722 	for (i = 0; i < NR_INODE_TYPE; i++) {
1723 		INIT_LIST_HEAD(&sbi->inode_list[i]);
1724 		spin_lock_init(&sbi->inode_lock[i]);
1725 	}
1726 
1727 	init_extent_cache_info(sbi);
1728 
1729 	init_ino_entry_info(sbi);
1730 
1731 	/* setup f2fs internal modules */
1732 	err = build_segment_manager(sbi);
1733 	if (err) {
1734 		f2fs_msg(sb, KERN_ERR,
1735 			"Failed to initialize F2FS segment manager");
1736 		goto free_sm;
1737 	}
1738 	err = build_node_manager(sbi);
1739 	if (err) {
1740 		f2fs_msg(sb, KERN_ERR,
1741 			"Failed to initialize F2FS node manager");
1742 		goto free_nm;
1743 	}
1744 
1745 	/* For write statistics */
1746 	if (sb->s_bdev->bd_part)
1747 		sbi->sectors_written_start =
1748 			(u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
1749 
1750 	/* Read accumulated write IO statistics if exists */
1751 	seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
1752 	if (__exist_node_summaries(sbi))
1753 		sbi->kbytes_written =
1754 			le64_to_cpu(seg_i->journal->info.kbytes_written);
1755 
1756 	build_gc_manager(sbi);
1757 
1758 	/* get an inode for node space */
1759 	sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1760 	if (IS_ERR(sbi->node_inode)) {
1761 		f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1762 		err = PTR_ERR(sbi->node_inode);
1763 		goto free_nm;
1764 	}
1765 
1766 	f2fs_join_shrinker(sbi);
1767 
1768 	/* if there are nt orphan nodes free them */
1769 	err = recover_orphan_inodes(sbi);
1770 	if (err)
1771 		goto free_node_inode;
1772 
1773 	/* read root inode and dentry */
1774 	root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1775 	if (IS_ERR(root)) {
1776 		f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1777 		err = PTR_ERR(root);
1778 		goto free_node_inode;
1779 	}
1780 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1781 		iput(root);
1782 		err = -EINVAL;
1783 		goto free_node_inode;
1784 	}
1785 
1786 	sb->s_root = d_make_root(root); /* allocate root dentry */
1787 	if (!sb->s_root) {
1788 		err = -ENOMEM;
1789 		goto free_root_inode;
1790 	}
1791 
1792 	err = f2fs_build_stats(sbi);
1793 	if (err)
1794 		goto free_root_inode;
1795 
1796 	if (f2fs_proc_root)
1797 		sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1798 
1799 	if (sbi->s_proc) {
1800 		proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1801 				 &f2fs_seq_segment_info_fops, sb);
1802 		proc_create_data("segment_bits", S_IRUGO, sbi->s_proc,
1803 				 &f2fs_seq_segment_bits_fops, sb);
1804 	}
1805 
1806 	sbi->s_kobj.kset = f2fs_kset;
1807 	init_completion(&sbi->s_kobj_unregister);
1808 	err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1809 							"%s", sb->s_id);
1810 	if (err)
1811 		goto free_proc;
1812 
1813 	/* recover fsynced data */
1814 	if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1815 		/*
1816 		 * mount should be failed, when device has readonly mode, and
1817 		 * previous checkpoint was not done by clean system shutdown.
1818 		 */
1819 		if (bdev_read_only(sb->s_bdev) &&
1820 				!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
1821 			err = -EROFS;
1822 			goto free_kobj;
1823 		}
1824 
1825 		if (need_fsck)
1826 			set_sbi_flag(sbi, SBI_NEED_FSCK);
1827 
1828 		if (!retry)
1829 			goto skip_recovery;
1830 
1831 		err = recover_fsync_data(sbi, false);
1832 		if (err < 0) {
1833 			need_fsck = true;
1834 			f2fs_msg(sb, KERN_ERR,
1835 				"Cannot recover all fsync data errno=%d", err);
1836 			goto free_kobj;
1837 		}
1838 	} else {
1839 		err = recover_fsync_data(sbi, true);
1840 
1841 		if (!f2fs_readonly(sb) && err > 0) {
1842 			err = -EINVAL;
1843 			f2fs_msg(sb, KERN_ERR,
1844 				"Need to recover fsync data");
1845 			goto free_kobj;
1846 		}
1847 	}
1848 skip_recovery:
1849 	/* recover_fsync_data() cleared this already */
1850 	clear_sbi_flag(sbi, SBI_POR_DOING);
1851 
1852 	/*
1853 	 * If filesystem is not mounted as read-only then
1854 	 * do start the gc_thread.
1855 	 */
1856 	if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
1857 		/* After POR, we can run background GC thread.*/
1858 		err = start_gc_thread(sbi);
1859 		if (err)
1860 			goto free_kobj;
1861 	}
1862 	kfree(options);
1863 
1864 	/* recover broken superblock */
1865 	if (recovery) {
1866 		err = f2fs_commit_super(sbi, true);
1867 		f2fs_msg(sb, KERN_INFO,
1868 			"Try to recover %dth superblock, ret: %d",
1869 			sbi->valid_super_block ? 1 : 2, err);
1870 	}
1871 
1872 	f2fs_update_time(sbi, CP_TIME);
1873 	f2fs_update_time(sbi, REQ_TIME);
1874 	return 0;
1875 
1876 free_kobj:
1877 	f2fs_sync_inode_meta(sbi);
1878 	kobject_del(&sbi->s_kobj);
1879 	kobject_put(&sbi->s_kobj);
1880 	wait_for_completion(&sbi->s_kobj_unregister);
1881 free_proc:
1882 	if (sbi->s_proc) {
1883 		remove_proc_entry("segment_info", sbi->s_proc);
1884 		remove_proc_entry("segment_bits", sbi->s_proc);
1885 		remove_proc_entry(sb->s_id, f2fs_proc_root);
1886 	}
1887 	f2fs_destroy_stats(sbi);
1888 free_root_inode:
1889 	dput(sb->s_root);
1890 	sb->s_root = NULL;
1891 free_node_inode:
1892 	truncate_inode_pages_final(NODE_MAPPING(sbi));
1893 	mutex_lock(&sbi->umount_mutex);
1894 	release_ino_entry(sbi, true);
1895 	f2fs_leave_shrinker(sbi);
1896 	iput(sbi->node_inode);
1897 	mutex_unlock(&sbi->umount_mutex);
1898 free_nm:
1899 	destroy_node_manager(sbi);
1900 free_sm:
1901 	destroy_segment_manager(sbi);
1902 	kfree(sbi->ckpt);
1903 free_meta_inode:
1904 	make_bad_inode(sbi->meta_inode);
1905 	iput(sbi->meta_inode);
1906 free_options:
1907 	destroy_percpu_info(sbi);
1908 	kfree(options);
1909 free_sb_buf:
1910 	kfree(raw_super);
1911 free_sbi:
1912 	if (sbi->s_chksum_driver)
1913 		crypto_free_shash(sbi->s_chksum_driver);
1914 	kfree(sbi);
1915 
1916 	/* give only one another chance */
1917 	if (retry) {
1918 		retry = false;
1919 		shrink_dcache_sb(sb);
1920 		goto try_onemore;
1921 	}
1922 	return err;
1923 }
1924 
1925 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1926 			const char *dev_name, void *data)
1927 {
1928 	return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1929 }
1930 
1931 static void kill_f2fs_super(struct super_block *sb)
1932 {
1933 	if (sb->s_root)
1934 		set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
1935 	kill_block_super(sb);
1936 }
1937 
1938 static struct file_system_type f2fs_fs_type = {
1939 	.owner		= THIS_MODULE,
1940 	.name		= "f2fs",
1941 	.mount		= f2fs_mount,
1942 	.kill_sb	= kill_f2fs_super,
1943 	.fs_flags	= FS_REQUIRES_DEV,
1944 };
1945 MODULE_ALIAS_FS("f2fs");
1946 
1947 static int __init init_inodecache(void)
1948 {
1949 	f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
1950 			sizeof(struct f2fs_inode_info), 0,
1951 			SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
1952 	if (!f2fs_inode_cachep)
1953 		return -ENOMEM;
1954 	return 0;
1955 }
1956 
1957 static void destroy_inodecache(void)
1958 {
1959 	/*
1960 	 * Make sure all delayed rcu free inodes are flushed before we
1961 	 * destroy cache.
1962 	 */
1963 	rcu_barrier();
1964 	kmem_cache_destroy(f2fs_inode_cachep);
1965 }
1966 
1967 static int __init init_f2fs_fs(void)
1968 {
1969 	int err;
1970 
1971 	f2fs_build_trace_ios();
1972 
1973 	err = init_inodecache();
1974 	if (err)
1975 		goto fail;
1976 	err = create_node_manager_caches();
1977 	if (err)
1978 		goto free_inodecache;
1979 	err = create_segment_manager_caches();
1980 	if (err)
1981 		goto free_node_manager_caches;
1982 	err = create_checkpoint_caches();
1983 	if (err)
1984 		goto free_segment_manager_caches;
1985 	err = create_extent_cache();
1986 	if (err)
1987 		goto free_checkpoint_caches;
1988 	f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1989 	if (!f2fs_kset) {
1990 		err = -ENOMEM;
1991 		goto free_extent_cache;
1992 	}
1993 	err = register_shrinker(&f2fs_shrinker_info);
1994 	if (err)
1995 		goto free_kset;
1996 
1997 	err = register_filesystem(&f2fs_fs_type);
1998 	if (err)
1999 		goto free_shrinker;
2000 	err = f2fs_create_root_stats();
2001 	if (err)
2002 		goto free_filesystem;
2003 	f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
2004 	return 0;
2005 
2006 free_filesystem:
2007 	unregister_filesystem(&f2fs_fs_type);
2008 free_shrinker:
2009 	unregister_shrinker(&f2fs_shrinker_info);
2010 free_kset:
2011 	kset_unregister(f2fs_kset);
2012 free_extent_cache:
2013 	destroy_extent_cache();
2014 free_checkpoint_caches:
2015 	destroy_checkpoint_caches();
2016 free_segment_manager_caches:
2017 	destroy_segment_manager_caches();
2018 free_node_manager_caches:
2019 	destroy_node_manager_caches();
2020 free_inodecache:
2021 	destroy_inodecache();
2022 fail:
2023 	return err;
2024 }
2025 
2026 static void __exit exit_f2fs_fs(void)
2027 {
2028 	remove_proc_entry("fs/f2fs", NULL);
2029 	f2fs_destroy_root_stats();
2030 	unregister_filesystem(&f2fs_fs_type);
2031 	unregister_shrinker(&f2fs_shrinker_info);
2032 	kset_unregister(f2fs_kset);
2033 	destroy_extent_cache();
2034 	destroy_checkpoint_caches();
2035 	destroy_segment_manager_caches();
2036 	destroy_node_manager_caches();
2037 	destroy_inodecache();
2038 	f2fs_destroy_trace_ios();
2039 }
2040 
2041 module_init(init_f2fs_fs)
2042 module_exit(exit_f2fs_fs)
2043 
2044 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
2045 MODULE_DESCRIPTION("Flash Friendly File System");
2046 MODULE_LICENSE("GPL");
2047