xref: /openbmc/linux/fs/f2fs/super.c (revision 7c45729a4d6d1c90879e6c5c2df325c2f6db7191)
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, bool sync)
624 {
625 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
626 	int ret = 0;
627 
628 	spin_lock(&sbi->inode_lock[DIRTY_META]);
629 	if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
630 		ret = 1;
631 	} else {
632 		set_inode_flag(inode, FI_DIRTY_INODE);
633 		stat_inc_dirty_inode(sbi, DIRTY_META);
634 	}
635 	if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
636 		list_add_tail(&F2FS_I(inode)->gdirty_list,
637 				&sbi->inode_list[DIRTY_META]);
638 		inc_page_count(sbi, F2FS_DIRTY_IMETA);
639 	}
640 	spin_unlock(&sbi->inode_lock[DIRTY_META]);
641 	return ret;
642 }
643 
644 void f2fs_inode_synced(struct inode *inode)
645 {
646 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
647 
648 	spin_lock(&sbi->inode_lock[DIRTY_META]);
649 	if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
650 		spin_unlock(&sbi->inode_lock[DIRTY_META]);
651 		return;
652 	}
653 	if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
654 		list_del_init(&F2FS_I(inode)->gdirty_list);
655 		dec_page_count(sbi, F2FS_DIRTY_IMETA);
656 	}
657 	clear_inode_flag(inode, FI_DIRTY_INODE);
658 	clear_inode_flag(inode, FI_AUTO_RECOVER);
659 	stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
660 	spin_unlock(&sbi->inode_lock[DIRTY_META]);
661 }
662 
663 /*
664  * f2fs_dirty_inode() is called from __mark_inode_dirty()
665  *
666  * We should call set_dirty_inode to write the dirty inode through write_inode.
667  */
668 static void f2fs_dirty_inode(struct inode *inode, int flags)
669 {
670 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
671 
672 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
673 			inode->i_ino == F2FS_META_INO(sbi))
674 		return;
675 
676 	if (flags == I_DIRTY_TIME)
677 		return;
678 
679 	if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
680 		clear_inode_flag(inode, FI_AUTO_RECOVER);
681 
682 	f2fs_inode_dirtied(inode, false);
683 }
684 
685 static void f2fs_i_callback(struct rcu_head *head)
686 {
687 	struct inode *inode = container_of(head, struct inode, i_rcu);
688 	kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
689 }
690 
691 static void f2fs_destroy_inode(struct inode *inode)
692 {
693 	percpu_counter_destroy(&F2FS_I(inode)->dirty_pages);
694 	call_rcu(&inode->i_rcu, f2fs_i_callback);
695 }
696 
697 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
698 {
699 	int i;
700 
701 	for (i = 0; i < NR_COUNT_TYPE; i++)
702 		percpu_counter_destroy(&sbi->nr_pages[i]);
703 	percpu_counter_destroy(&sbi->alloc_valid_block_count);
704 	percpu_counter_destroy(&sbi->total_valid_inode_count);
705 }
706 
707 static void f2fs_put_super(struct super_block *sb)
708 {
709 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
710 
711 	if (sbi->s_proc) {
712 		remove_proc_entry("segment_info", sbi->s_proc);
713 		remove_proc_entry("segment_bits", sbi->s_proc);
714 		remove_proc_entry(sb->s_id, f2fs_proc_root);
715 	}
716 	kobject_del(&sbi->s_kobj);
717 
718 	stop_gc_thread(sbi);
719 
720 	/* prevent remaining shrinker jobs */
721 	mutex_lock(&sbi->umount_mutex);
722 
723 	/*
724 	 * We don't need to do checkpoint when superblock is clean.
725 	 * But, the previous checkpoint was not done by umount, it needs to do
726 	 * clean checkpoint again.
727 	 */
728 	if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
729 			!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
730 		struct cp_control cpc = {
731 			.reason = CP_UMOUNT,
732 		};
733 		write_checkpoint(sbi, &cpc);
734 	}
735 
736 	/* write_checkpoint can update stat informaion */
737 	f2fs_destroy_stats(sbi);
738 
739 	/*
740 	 * normally superblock is clean, so we need to release this.
741 	 * In addition, EIO will skip do checkpoint, we need this as well.
742 	 */
743 	release_ino_entry(sbi, true);
744 
745 	f2fs_leave_shrinker(sbi);
746 	mutex_unlock(&sbi->umount_mutex);
747 
748 	/* our cp_error case, we can wait for any writeback page */
749 	f2fs_flush_merged_bios(sbi);
750 
751 	iput(sbi->node_inode);
752 	iput(sbi->meta_inode);
753 
754 	/* destroy f2fs internal modules */
755 	destroy_node_manager(sbi);
756 	destroy_segment_manager(sbi);
757 
758 	kfree(sbi->ckpt);
759 	kobject_put(&sbi->s_kobj);
760 	wait_for_completion(&sbi->s_kobj_unregister);
761 
762 	sb->s_fs_info = NULL;
763 	if (sbi->s_chksum_driver)
764 		crypto_free_shash(sbi->s_chksum_driver);
765 	kfree(sbi->raw_super);
766 
767 	destroy_percpu_info(sbi);
768 	kfree(sbi);
769 }
770 
771 int f2fs_sync_fs(struct super_block *sb, int sync)
772 {
773 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
774 	int err = 0;
775 
776 	trace_f2fs_sync_fs(sb, sync);
777 
778 	if (sync) {
779 		struct cp_control cpc;
780 
781 		cpc.reason = __get_cp_reason(sbi);
782 
783 		mutex_lock(&sbi->gc_mutex);
784 		err = write_checkpoint(sbi, &cpc);
785 		mutex_unlock(&sbi->gc_mutex);
786 	}
787 	f2fs_trace_ios(NULL, 1);
788 
789 	return err;
790 }
791 
792 static int f2fs_freeze(struct super_block *sb)
793 {
794 	int err;
795 
796 	if (f2fs_readonly(sb))
797 		return 0;
798 
799 	err = f2fs_sync_fs(sb, 1);
800 	return err;
801 }
802 
803 static int f2fs_unfreeze(struct super_block *sb)
804 {
805 	return 0;
806 }
807 
808 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
809 {
810 	struct super_block *sb = dentry->d_sb;
811 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
812 	u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
813 	block_t total_count, user_block_count, start_count, ovp_count;
814 
815 	total_count = le64_to_cpu(sbi->raw_super->block_count);
816 	user_block_count = sbi->user_block_count;
817 	start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
818 	ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
819 	buf->f_type = F2FS_SUPER_MAGIC;
820 	buf->f_bsize = sbi->blocksize;
821 
822 	buf->f_blocks = total_count - start_count;
823 	buf->f_bfree = user_block_count - valid_user_blocks(sbi) + ovp_count;
824 	buf->f_bavail = user_block_count - valid_user_blocks(sbi);
825 
826 	buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
827 	buf->f_ffree = buf->f_files - valid_inode_count(sbi);
828 
829 	buf->f_namelen = F2FS_NAME_LEN;
830 	buf->f_fsid.val[0] = (u32)id;
831 	buf->f_fsid.val[1] = (u32)(id >> 32);
832 
833 	return 0;
834 }
835 
836 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
837 {
838 	struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
839 
840 	if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
841 		if (test_opt(sbi, FORCE_FG_GC))
842 			seq_printf(seq, ",background_gc=%s", "sync");
843 		else
844 			seq_printf(seq, ",background_gc=%s", "on");
845 	} else {
846 		seq_printf(seq, ",background_gc=%s", "off");
847 	}
848 	if (test_opt(sbi, DISABLE_ROLL_FORWARD))
849 		seq_puts(seq, ",disable_roll_forward");
850 	if (test_opt(sbi, DISCARD))
851 		seq_puts(seq, ",discard");
852 	if (test_opt(sbi, NOHEAP))
853 		seq_puts(seq, ",no_heap_alloc");
854 #ifdef CONFIG_F2FS_FS_XATTR
855 	if (test_opt(sbi, XATTR_USER))
856 		seq_puts(seq, ",user_xattr");
857 	else
858 		seq_puts(seq, ",nouser_xattr");
859 	if (test_opt(sbi, INLINE_XATTR))
860 		seq_puts(seq, ",inline_xattr");
861 #endif
862 #ifdef CONFIG_F2FS_FS_POSIX_ACL
863 	if (test_opt(sbi, POSIX_ACL))
864 		seq_puts(seq, ",acl");
865 	else
866 		seq_puts(seq, ",noacl");
867 #endif
868 	if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
869 		seq_puts(seq, ",disable_ext_identify");
870 	if (test_opt(sbi, INLINE_DATA))
871 		seq_puts(seq, ",inline_data");
872 	else
873 		seq_puts(seq, ",noinline_data");
874 	if (test_opt(sbi, INLINE_DENTRY))
875 		seq_puts(seq, ",inline_dentry");
876 	else
877 		seq_puts(seq, ",noinline_dentry");
878 	if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
879 		seq_puts(seq, ",flush_merge");
880 	if (test_opt(sbi, NOBARRIER))
881 		seq_puts(seq, ",nobarrier");
882 	if (test_opt(sbi, FASTBOOT))
883 		seq_puts(seq, ",fastboot");
884 	if (test_opt(sbi, EXTENT_CACHE))
885 		seq_puts(seq, ",extent_cache");
886 	else
887 		seq_puts(seq, ",noextent_cache");
888 	if (test_opt(sbi, DATA_FLUSH))
889 		seq_puts(seq, ",data_flush");
890 
891 	seq_puts(seq, ",mode=");
892 	if (test_opt(sbi, ADAPTIVE))
893 		seq_puts(seq, "adaptive");
894 	else if (test_opt(sbi, LFS))
895 		seq_puts(seq, "lfs");
896 	seq_printf(seq, ",active_logs=%u", sbi->active_logs);
897 
898 	return 0;
899 }
900 
901 static int segment_info_seq_show(struct seq_file *seq, void *offset)
902 {
903 	struct super_block *sb = seq->private;
904 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
905 	unsigned int total_segs =
906 			le32_to_cpu(sbi->raw_super->segment_count_main);
907 	int i;
908 
909 	seq_puts(seq, "format: segment_type|valid_blocks\n"
910 		"segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
911 
912 	for (i = 0; i < total_segs; i++) {
913 		struct seg_entry *se = get_seg_entry(sbi, i);
914 
915 		if ((i % 10) == 0)
916 			seq_printf(seq, "%-10d", i);
917 		seq_printf(seq, "%d|%-3u", se->type,
918 					get_valid_blocks(sbi, i, 1));
919 		if ((i % 10) == 9 || i == (total_segs - 1))
920 			seq_putc(seq, '\n');
921 		else
922 			seq_putc(seq, ' ');
923 	}
924 
925 	return 0;
926 }
927 
928 static int segment_bits_seq_show(struct seq_file *seq, void *offset)
929 {
930 	struct super_block *sb = seq->private;
931 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
932 	unsigned int total_segs =
933 			le32_to_cpu(sbi->raw_super->segment_count_main);
934 	int i, j;
935 
936 	seq_puts(seq, "format: segment_type|valid_blocks|bitmaps\n"
937 		"segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
938 
939 	for (i = 0; i < total_segs; i++) {
940 		struct seg_entry *se = get_seg_entry(sbi, i);
941 
942 		seq_printf(seq, "%-10d", i);
943 		seq_printf(seq, "%d|%-3u|", se->type,
944 					get_valid_blocks(sbi, i, 1));
945 		for (j = 0; j < SIT_VBLOCK_MAP_SIZE; j++)
946 			seq_printf(seq, " %.2x", se->cur_valid_map[j]);
947 		seq_putc(seq, '\n');
948 	}
949 	return 0;
950 }
951 
952 #define F2FS_PROC_FILE_DEF(_name)					\
953 static int _name##_open_fs(struct inode *inode, struct file *file)	\
954 {									\
955 	return single_open(file, _name##_seq_show, PDE_DATA(inode));	\
956 }									\
957 									\
958 static const struct file_operations f2fs_seq_##_name##_fops = {		\
959 	.open = _name##_open_fs,					\
960 	.read = seq_read,						\
961 	.llseek = seq_lseek,						\
962 	.release = single_release,					\
963 };
964 
965 F2FS_PROC_FILE_DEF(segment_info);
966 F2FS_PROC_FILE_DEF(segment_bits);
967 
968 static void default_options(struct f2fs_sb_info *sbi)
969 {
970 	/* init some FS parameters */
971 	sbi->active_logs = NR_CURSEG_TYPE;
972 
973 	set_opt(sbi, BG_GC);
974 	set_opt(sbi, INLINE_DATA);
975 	set_opt(sbi, INLINE_DENTRY);
976 	set_opt(sbi, EXTENT_CACHE);
977 	sbi->sb->s_flags |= MS_LAZYTIME;
978 	set_opt(sbi, FLUSH_MERGE);
979 	if (f2fs_sb_mounted_hmsmr(sbi->sb)) {
980 		set_opt_mode(sbi, F2FS_MOUNT_LFS);
981 		set_opt(sbi, DISCARD);
982 	} else {
983 		set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
984 	}
985 
986 #ifdef CONFIG_F2FS_FS_XATTR
987 	set_opt(sbi, XATTR_USER);
988 #endif
989 #ifdef CONFIG_F2FS_FS_POSIX_ACL
990 	set_opt(sbi, POSIX_ACL);
991 #endif
992 
993 #ifdef CONFIG_F2FS_FAULT_INJECTION
994 	f2fs_build_fault_attr(sbi, 0);
995 #endif
996 }
997 
998 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
999 {
1000 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1001 	struct f2fs_mount_info org_mount_opt;
1002 	int err, active_logs;
1003 	bool need_restart_gc = false;
1004 	bool need_stop_gc = false;
1005 	bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1006 #ifdef CONFIG_F2FS_FAULT_INJECTION
1007 	struct f2fs_fault_info ffi = sbi->fault_info;
1008 #endif
1009 
1010 	/*
1011 	 * Save the old mount options in case we
1012 	 * need to restore them.
1013 	 */
1014 	org_mount_opt = sbi->mount_opt;
1015 	active_logs = sbi->active_logs;
1016 
1017 	/* recover superblocks we couldn't write due to previous RO mount */
1018 	if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1019 		err = f2fs_commit_super(sbi, false);
1020 		f2fs_msg(sb, KERN_INFO,
1021 			"Try to recover all the superblocks, ret: %d", err);
1022 		if (!err)
1023 			clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1024 	}
1025 
1026 	sbi->mount_opt.opt = 0;
1027 	default_options(sbi);
1028 
1029 	/* parse mount options */
1030 	err = parse_options(sb, data);
1031 	if (err)
1032 		goto restore_opts;
1033 
1034 	/*
1035 	 * Previous and new state of filesystem is RO,
1036 	 * so skip checking GC and FLUSH_MERGE conditions.
1037 	 */
1038 	if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
1039 		goto skip;
1040 
1041 	/* disallow enable/disable extent_cache dynamically */
1042 	if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1043 		err = -EINVAL;
1044 		f2fs_msg(sbi->sb, KERN_WARNING,
1045 				"switch extent_cache option is not allowed");
1046 		goto restore_opts;
1047 	}
1048 
1049 	/*
1050 	 * We stop the GC thread if FS is mounted as RO
1051 	 * or if background_gc = off is passed in mount
1052 	 * option. Also sync the filesystem.
1053 	 */
1054 	if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
1055 		if (sbi->gc_thread) {
1056 			stop_gc_thread(sbi);
1057 			need_restart_gc = true;
1058 		}
1059 	} else if (!sbi->gc_thread) {
1060 		err = start_gc_thread(sbi);
1061 		if (err)
1062 			goto restore_opts;
1063 		need_stop_gc = true;
1064 	}
1065 
1066 	if (*flags & MS_RDONLY) {
1067 		writeback_inodes_sb(sb, WB_REASON_SYNC);
1068 		sync_inodes_sb(sb);
1069 
1070 		set_sbi_flag(sbi, SBI_IS_DIRTY);
1071 		set_sbi_flag(sbi, SBI_IS_CLOSE);
1072 		f2fs_sync_fs(sb, 1);
1073 		clear_sbi_flag(sbi, SBI_IS_CLOSE);
1074 	}
1075 
1076 	/*
1077 	 * We stop issue flush thread if FS is mounted as RO
1078 	 * or if flush_merge is not passed in mount option.
1079 	 */
1080 	if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1081 		destroy_flush_cmd_control(sbi);
1082 	} else if (!SM_I(sbi)->cmd_control_info) {
1083 		err = create_flush_cmd_control(sbi);
1084 		if (err)
1085 			goto restore_gc;
1086 	}
1087 skip:
1088 	/* Update the POSIXACL Flag */
1089 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1090 		(test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1091 
1092 	return 0;
1093 restore_gc:
1094 	if (need_restart_gc) {
1095 		if (start_gc_thread(sbi))
1096 			f2fs_msg(sbi->sb, KERN_WARNING,
1097 				"background gc thread has stopped");
1098 	} else if (need_stop_gc) {
1099 		stop_gc_thread(sbi);
1100 	}
1101 restore_opts:
1102 	sbi->mount_opt = org_mount_opt;
1103 	sbi->active_logs = active_logs;
1104 #ifdef CONFIG_F2FS_FAULT_INJECTION
1105 	sbi->fault_info = ffi;
1106 #endif
1107 	return err;
1108 }
1109 
1110 static struct super_operations f2fs_sops = {
1111 	.alloc_inode	= f2fs_alloc_inode,
1112 	.drop_inode	= f2fs_drop_inode,
1113 	.destroy_inode	= f2fs_destroy_inode,
1114 	.write_inode	= f2fs_write_inode,
1115 	.dirty_inode	= f2fs_dirty_inode,
1116 	.show_options	= f2fs_show_options,
1117 	.evict_inode	= f2fs_evict_inode,
1118 	.put_super	= f2fs_put_super,
1119 	.sync_fs	= f2fs_sync_fs,
1120 	.freeze_fs	= f2fs_freeze,
1121 	.unfreeze_fs	= f2fs_unfreeze,
1122 	.statfs		= f2fs_statfs,
1123 	.remount_fs	= f2fs_remount,
1124 };
1125 
1126 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1127 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1128 {
1129 	return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1130 				F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1131 				ctx, len, NULL);
1132 }
1133 
1134 static int f2fs_key_prefix(struct inode *inode, u8 **key)
1135 {
1136 	*key = F2FS_I_SB(inode)->key_prefix;
1137 	return F2FS_I_SB(inode)->key_prefix_size;
1138 }
1139 
1140 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1141 							void *fs_data)
1142 {
1143 	return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1144 				F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1145 				ctx, len, fs_data, XATTR_CREATE);
1146 }
1147 
1148 static unsigned f2fs_max_namelen(struct inode *inode)
1149 {
1150 	return S_ISLNK(inode->i_mode) ?
1151 			inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1152 }
1153 
1154 static struct fscrypt_operations f2fs_cryptops = {
1155 	.get_context	= f2fs_get_context,
1156 	.key_prefix	= f2fs_key_prefix,
1157 	.set_context	= f2fs_set_context,
1158 	.is_encrypted	= f2fs_encrypted_inode,
1159 	.empty_dir	= f2fs_empty_dir,
1160 	.max_namelen	= f2fs_max_namelen,
1161 };
1162 #else
1163 static struct fscrypt_operations f2fs_cryptops = {
1164 	.is_encrypted	= f2fs_encrypted_inode,
1165 };
1166 #endif
1167 
1168 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1169 		u64 ino, u32 generation)
1170 {
1171 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1172 	struct inode *inode;
1173 
1174 	if (check_nid_range(sbi, ino))
1175 		return ERR_PTR(-ESTALE);
1176 
1177 	/*
1178 	 * f2fs_iget isn't quite right if the inode is currently unallocated!
1179 	 * However f2fs_iget currently does appropriate checks to handle stale
1180 	 * inodes so everything is OK.
1181 	 */
1182 	inode = f2fs_iget(sb, ino);
1183 	if (IS_ERR(inode))
1184 		return ERR_CAST(inode);
1185 	if (unlikely(generation && inode->i_generation != generation)) {
1186 		/* we didn't find the right inode.. */
1187 		iput(inode);
1188 		return ERR_PTR(-ESTALE);
1189 	}
1190 	return inode;
1191 }
1192 
1193 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1194 		int fh_len, int fh_type)
1195 {
1196 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1197 				    f2fs_nfs_get_inode);
1198 }
1199 
1200 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1201 		int fh_len, int fh_type)
1202 {
1203 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1204 				    f2fs_nfs_get_inode);
1205 }
1206 
1207 static const struct export_operations f2fs_export_ops = {
1208 	.fh_to_dentry = f2fs_fh_to_dentry,
1209 	.fh_to_parent = f2fs_fh_to_parent,
1210 	.get_parent = f2fs_get_parent,
1211 };
1212 
1213 static loff_t max_file_blocks(void)
1214 {
1215 	loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
1216 	loff_t leaf_count = ADDRS_PER_BLOCK;
1217 
1218 	/* two direct node blocks */
1219 	result += (leaf_count * 2);
1220 
1221 	/* two indirect node blocks */
1222 	leaf_count *= NIDS_PER_BLOCK;
1223 	result += (leaf_count * 2);
1224 
1225 	/* one double indirect node block */
1226 	leaf_count *= NIDS_PER_BLOCK;
1227 	result += leaf_count;
1228 
1229 	return result;
1230 }
1231 
1232 static int __f2fs_commit_super(struct buffer_head *bh,
1233 			struct f2fs_super_block *super)
1234 {
1235 	lock_buffer(bh);
1236 	if (super)
1237 		memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
1238 	set_buffer_uptodate(bh);
1239 	set_buffer_dirty(bh);
1240 	unlock_buffer(bh);
1241 
1242 	/* it's rare case, we can do fua all the time */
1243 	return __sync_dirty_buffer(bh, WRITE_FLUSH_FUA);
1244 }
1245 
1246 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
1247 					struct buffer_head *bh)
1248 {
1249 	struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1250 					(bh->b_data + F2FS_SUPER_OFFSET);
1251 	struct super_block *sb = sbi->sb;
1252 	u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
1253 	u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
1254 	u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
1255 	u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
1256 	u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
1257 	u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
1258 	u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
1259 	u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
1260 	u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
1261 	u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
1262 	u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1263 	u32 segment_count = le32_to_cpu(raw_super->segment_count);
1264 	u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1265 	u64 main_end_blkaddr = main_blkaddr +
1266 				(segment_count_main << log_blocks_per_seg);
1267 	u64 seg_end_blkaddr = segment0_blkaddr +
1268 				(segment_count << log_blocks_per_seg);
1269 
1270 	if (segment0_blkaddr != cp_blkaddr) {
1271 		f2fs_msg(sb, KERN_INFO,
1272 			"Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1273 			segment0_blkaddr, cp_blkaddr);
1274 		return true;
1275 	}
1276 
1277 	if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
1278 							sit_blkaddr) {
1279 		f2fs_msg(sb, KERN_INFO,
1280 			"Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1281 			cp_blkaddr, sit_blkaddr,
1282 			segment_count_ckpt << log_blocks_per_seg);
1283 		return true;
1284 	}
1285 
1286 	if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
1287 							nat_blkaddr) {
1288 		f2fs_msg(sb, KERN_INFO,
1289 			"Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1290 			sit_blkaddr, nat_blkaddr,
1291 			segment_count_sit << log_blocks_per_seg);
1292 		return true;
1293 	}
1294 
1295 	if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
1296 							ssa_blkaddr) {
1297 		f2fs_msg(sb, KERN_INFO,
1298 			"Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1299 			nat_blkaddr, ssa_blkaddr,
1300 			segment_count_nat << log_blocks_per_seg);
1301 		return true;
1302 	}
1303 
1304 	if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
1305 							main_blkaddr) {
1306 		f2fs_msg(sb, KERN_INFO,
1307 			"Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1308 			ssa_blkaddr, main_blkaddr,
1309 			segment_count_ssa << log_blocks_per_seg);
1310 		return true;
1311 	}
1312 
1313 	if (main_end_blkaddr > seg_end_blkaddr) {
1314 		f2fs_msg(sb, KERN_INFO,
1315 			"Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1316 			main_blkaddr,
1317 			segment0_blkaddr +
1318 				(segment_count << log_blocks_per_seg),
1319 			segment_count_main << log_blocks_per_seg);
1320 		return true;
1321 	} else if (main_end_blkaddr < seg_end_blkaddr) {
1322 		int err = 0;
1323 		char *res;
1324 
1325 		/* fix in-memory information all the time */
1326 		raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
1327 				segment0_blkaddr) >> log_blocks_per_seg);
1328 
1329 		if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
1330 			set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1331 			res = "internally";
1332 		} else {
1333 			err = __f2fs_commit_super(bh, NULL);
1334 			res = err ? "failed" : "done";
1335 		}
1336 		f2fs_msg(sb, KERN_INFO,
1337 			"Fix alignment : %s, start(%u) end(%u) block(%u)",
1338 			res, main_blkaddr,
1339 			segment0_blkaddr +
1340 				(segment_count << log_blocks_per_seg),
1341 			segment_count_main << log_blocks_per_seg);
1342 		if (err)
1343 			return true;
1344 	}
1345 	return false;
1346 }
1347 
1348 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
1349 				struct buffer_head *bh)
1350 {
1351 	struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1352 					(bh->b_data + F2FS_SUPER_OFFSET);
1353 	struct super_block *sb = sbi->sb;
1354 	unsigned int blocksize;
1355 
1356 	if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
1357 		f2fs_msg(sb, KERN_INFO,
1358 			"Magic Mismatch, valid(0x%x) - read(0x%x)",
1359 			F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
1360 		return 1;
1361 	}
1362 
1363 	/* Currently, support only 4KB page cache size */
1364 	if (F2FS_BLKSIZE != PAGE_SIZE) {
1365 		f2fs_msg(sb, KERN_INFO,
1366 			"Invalid page_cache_size (%lu), supports only 4KB\n",
1367 			PAGE_SIZE);
1368 		return 1;
1369 	}
1370 
1371 	/* Currently, support only 4KB block size */
1372 	blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
1373 	if (blocksize != F2FS_BLKSIZE) {
1374 		f2fs_msg(sb, KERN_INFO,
1375 			"Invalid blocksize (%u), supports only 4KB\n",
1376 			blocksize);
1377 		return 1;
1378 	}
1379 
1380 	/* check log blocks per segment */
1381 	if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
1382 		f2fs_msg(sb, KERN_INFO,
1383 			"Invalid log blocks per segment (%u)\n",
1384 			le32_to_cpu(raw_super->log_blocks_per_seg));
1385 		return 1;
1386 	}
1387 
1388 	/* Currently, support 512/1024/2048/4096 bytes sector size */
1389 	if (le32_to_cpu(raw_super->log_sectorsize) >
1390 				F2FS_MAX_LOG_SECTOR_SIZE ||
1391 		le32_to_cpu(raw_super->log_sectorsize) <
1392 				F2FS_MIN_LOG_SECTOR_SIZE) {
1393 		f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
1394 			le32_to_cpu(raw_super->log_sectorsize));
1395 		return 1;
1396 	}
1397 	if (le32_to_cpu(raw_super->log_sectors_per_block) +
1398 		le32_to_cpu(raw_super->log_sectorsize) !=
1399 			F2FS_MAX_LOG_SECTOR_SIZE) {
1400 		f2fs_msg(sb, KERN_INFO,
1401 			"Invalid log sectors per block(%u) log sectorsize(%u)",
1402 			le32_to_cpu(raw_super->log_sectors_per_block),
1403 			le32_to_cpu(raw_super->log_sectorsize));
1404 		return 1;
1405 	}
1406 
1407 	/* check reserved ino info */
1408 	if (le32_to_cpu(raw_super->node_ino) != 1 ||
1409 		le32_to_cpu(raw_super->meta_ino) != 2 ||
1410 		le32_to_cpu(raw_super->root_ino) != 3) {
1411 		f2fs_msg(sb, KERN_INFO,
1412 			"Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1413 			le32_to_cpu(raw_super->node_ino),
1414 			le32_to_cpu(raw_super->meta_ino),
1415 			le32_to_cpu(raw_super->root_ino));
1416 		return 1;
1417 	}
1418 
1419 	/* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1420 	if (sanity_check_area_boundary(sbi, bh))
1421 		return 1;
1422 
1423 	return 0;
1424 }
1425 
1426 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1427 {
1428 	unsigned int total, fsmeta;
1429 	struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1430 	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1431 
1432 	total = le32_to_cpu(raw_super->segment_count);
1433 	fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
1434 	fsmeta += le32_to_cpu(raw_super->segment_count_sit);
1435 	fsmeta += le32_to_cpu(raw_super->segment_count_nat);
1436 	fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
1437 	fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
1438 
1439 	if (unlikely(fsmeta >= total))
1440 		return 1;
1441 
1442 	if (unlikely(f2fs_cp_error(sbi))) {
1443 		f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
1444 		return 1;
1445 	}
1446 	return 0;
1447 }
1448 
1449 static void init_sb_info(struct f2fs_sb_info *sbi)
1450 {
1451 	struct f2fs_super_block *raw_super = sbi->raw_super;
1452 
1453 	sbi->log_sectors_per_block =
1454 		le32_to_cpu(raw_super->log_sectors_per_block);
1455 	sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
1456 	sbi->blocksize = 1 << sbi->log_blocksize;
1457 	sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1458 	sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
1459 	sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1460 	sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1461 	sbi->total_sections = le32_to_cpu(raw_super->section_count);
1462 	sbi->total_node_count =
1463 		(le32_to_cpu(raw_super->segment_count_nat) / 2)
1464 			* sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
1465 	sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
1466 	sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
1467 	sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
1468 	sbi->cur_victim_sec = NULL_SECNO;
1469 	sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
1470 
1471 	sbi->dir_level = DEF_DIR_LEVEL;
1472 	sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
1473 	sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
1474 	clear_sbi_flag(sbi, SBI_NEED_FSCK);
1475 
1476 	INIT_LIST_HEAD(&sbi->s_list);
1477 	mutex_init(&sbi->umount_mutex);
1478 	mutex_init(&sbi->wio_mutex[NODE]);
1479 	mutex_init(&sbi->wio_mutex[DATA]);
1480 	spin_lock_init(&sbi->cp_lock);
1481 
1482 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1483 	memcpy(sbi->key_prefix, F2FS_KEY_DESC_PREFIX,
1484 				F2FS_KEY_DESC_PREFIX_SIZE);
1485 	sbi->key_prefix_size = F2FS_KEY_DESC_PREFIX_SIZE;
1486 #endif
1487 }
1488 
1489 static int init_percpu_info(struct f2fs_sb_info *sbi)
1490 {
1491 	int i, err;
1492 
1493 	for (i = 0; i < NR_COUNT_TYPE; i++) {
1494 		err = percpu_counter_init(&sbi->nr_pages[i], 0, GFP_KERNEL);
1495 		if (err)
1496 			return err;
1497 	}
1498 
1499 	err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
1500 	if (err)
1501 		return err;
1502 
1503 	return percpu_counter_init(&sbi->total_valid_inode_count, 0,
1504 								GFP_KERNEL);
1505 }
1506 
1507 /*
1508  * Read f2fs raw super block.
1509  * Because we have two copies of super block, so read both of them
1510  * to get the first valid one. If any one of them is broken, we pass
1511  * them recovery flag back to the caller.
1512  */
1513 static int read_raw_super_block(struct f2fs_sb_info *sbi,
1514 			struct f2fs_super_block **raw_super,
1515 			int *valid_super_block, int *recovery)
1516 {
1517 	struct super_block *sb = sbi->sb;
1518 	int block;
1519 	struct buffer_head *bh;
1520 	struct f2fs_super_block *super;
1521 	int err = 0;
1522 
1523 	super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
1524 	if (!super)
1525 		return -ENOMEM;
1526 
1527 	for (block = 0; block < 2; block++) {
1528 		bh = sb_bread(sb, block);
1529 		if (!bh) {
1530 			f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
1531 				block + 1);
1532 			err = -EIO;
1533 			continue;
1534 		}
1535 
1536 		/* sanity checking of raw super */
1537 		if (sanity_check_raw_super(sbi, bh)) {
1538 			f2fs_msg(sb, KERN_ERR,
1539 				"Can't find valid F2FS filesystem in %dth superblock",
1540 				block + 1);
1541 			err = -EINVAL;
1542 			brelse(bh);
1543 			continue;
1544 		}
1545 
1546 		if (!*raw_super) {
1547 			memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
1548 							sizeof(*super));
1549 			*valid_super_block = block;
1550 			*raw_super = super;
1551 		}
1552 		brelse(bh);
1553 	}
1554 
1555 	/* Fail to read any one of the superblocks*/
1556 	if (err < 0)
1557 		*recovery = 1;
1558 
1559 	/* No valid superblock */
1560 	if (!*raw_super)
1561 		kfree(super);
1562 	else
1563 		err = 0;
1564 
1565 	return err;
1566 }
1567 
1568 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
1569 {
1570 	struct buffer_head *bh;
1571 	int err;
1572 
1573 	if ((recover && f2fs_readonly(sbi->sb)) ||
1574 				bdev_read_only(sbi->sb->s_bdev)) {
1575 		set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1576 		return -EROFS;
1577 	}
1578 
1579 	/* write back-up superblock first */
1580 	bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1);
1581 	if (!bh)
1582 		return -EIO;
1583 	err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1584 	brelse(bh);
1585 
1586 	/* if we are in recovery path, skip writing valid superblock */
1587 	if (recover || err)
1588 		return err;
1589 
1590 	/* write current valid superblock */
1591 	bh = sb_getblk(sbi->sb, sbi->valid_super_block);
1592 	if (!bh)
1593 		return -EIO;
1594 	err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1595 	brelse(bh);
1596 	return err;
1597 }
1598 
1599 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
1600 {
1601 	struct f2fs_sb_info *sbi;
1602 	struct f2fs_super_block *raw_super;
1603 	struct inode *root;
1604 	int err;
1605 	bool retry = true, need_fsck = false;
1606 	char *options = NULL;
1607 	int recovery, i, valid_super_block;
1608 	struct curseg_info *seg_i;
1609 
1610 try_onemore:
1611 	err = -EINVAL;
1612 	raw_super = NULL;
1613 	valid_super_block = -1;
1614 	recovery = 0;
1615 
1616 	/* allocate memory for f2fs-specific super block info */
1617 	sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
1618 	if (!sbi)
1619 		return -ENOMEM;
1620 
1621 	sbi->sb = sb;
1622 
1623 	/* Load the checksum driver */
1624 	sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
1625 	if (IS_ERR(sbi->s_chksum_driver)) {
1626 		f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
1627 		err = PTR_ERR(sbi->s_chksum_driver);
1628 		sbi->s_chksum_driver = NULL;
1629 		goto free_sbi;
1630 	}
1631 
1632 	/* set a block size */
1633 	if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
1634 		f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
1635 		goto free_sbi;
1636 	}
1637 
1638 	err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
1639 								&recovery);
1640 	if (err)
1641 		goto free_sbi;
1642 
1643 	sb->s_fs_info = sbi;
1644 	sbi->raw_super = raw_super;
1645 
1646 	default_options(sbi);
1647 	/* parse mount options */
1648 	options = kstrdup((const char *)data, GFP_KERNEL);
1649 	if (data && !options) {
1650 		err = -ENOMEM;
1651 		goto free_sb_buf;
1652 	}
1653 
1654 	err = parse_options(sb, options);
1655 	if (err)
1656 		goto free_options;
1657 
1658 	sbi->max_file_blocks = max_file_blocks();
1659 	sb->s_maxbytes = sbi->max_file_blocks <<
1660 				le32_to_cpu(raw_super->log_blocksize);
1661 	sb->s_max_links = F2FS_LINK_MAX;
1662 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1663 
1664 	sb->s_op = &f2fs_sops;
1665 	sb->s_cop = &f2fs_cryptops;
1666 	sb->s_xattr = f2fs_xattr_handlers;
1667 	sb->s_export_op = &f2fs_export_ops;
1668 	sb->s_magic = F2FS_SUPER_MAGIC;
1669 	sb->s_time_gran = 1;
1670 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1671 		(test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1672 	memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1673 
1674 	/* init f2fs-specific super block info */
1675 	sbi->valid_super_block = valid_super_block;
1676 	mutex_init(&sbi->gc_mutex);
1677 	mutex_init(&sbi->cp_mutex);
1678 	init_rwsem(&sbi->node_write);
1679 
1680 	/* disallow all the data/node/meta page writes */
1681 	set_sbi_flag(sbi, SBI_POR_DOING);
1682 	spin_lock_init(&sbi->stat_lock);
1683 
1684 	init_rwsem(&sbi->read_io.io_rwsem);
1685 	sbi->read_io.sbi = sbi;
1686 	sbi->read_io.bio = NULL;
1687 	for (i = 0; i < NR_PAGE_TYPE; i++) {
1688 		init_rwsem(&sbi->write_io[i].io_rwsem);
1689 		sbi->write_io[i].sbi = sbi;
1690 		sbi->write_io[i].bio = NULL;
1691 	}
1692 
1693 	init_rwsem(&sbi->cp_rwsem);
1694 	init_waitqueue_head(&sbi->cp_wait);
1695 	init_sb_info(sbi);
1696 
1697 	err = init_percpu_info(sbi);
1698 	if (err)
1699 		goto free_options;
1700 
1701 	/* get an inode for meta space */
1702 	sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1703 	if (IS_ERR(sbi->meta_inode)) {
1704 		f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1705 		err = PTR_ERR(sbi->meta_inode);
1706 		goto free_options;
1707 	}
1708 
1709 	err = get_valid_checkpoint(sbi);
1710 	if (err) {
1711 		f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1712 		goto free_meta_inode;
1713 	}
1714 
1715 	sbi->total_valid_node_count =
1716 				le32_to_cpu(sbi->ckpt->valid_node_count);
1717 	percpu_counter_set(&sbi->total_valid_inode_count,
1718 				le32_to_cpu(sbi->ckpt->valid_inode_count));
1719 	sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1720 	sbi->total_valid_block_count =
1721 				le64_to_cpu(sbi->ckpt->valid_block_count);
1722 	sbi->last_valid_block_count = sbi->total_valid_block_count;
1723 
1724 	for (i = 0; i < NR_INODE_TYPE; i++) {
1725 		INIT_LIST_HEAD(&sbi->inode_list[i]);
1726 		spin_lock_init(&sbi->inode_lock[i]);
1727 	}
1728 
1729 	init_extent_cache_info(sbi);
1730 
1731 	init_ino_entry_info(sbi);
1732 
1733 	/* setup f2fs internal modules */
1734 	err = build_segment_manager(sbi);
1735 	if (err) {
1736 		f2fs_msg(sb, KERN_ERR,
1737 			"Failed to initialize F2FS segment manager");
1738 		goto free_sm;
1739 	}
1740 	err = build_node_manager(sbi);
1741 	if (err) {
1742 		f2fs_msg(sb, KERN_ERR,
1743 			"Failed to initialize F2FS node manager");
1744 		goto free_nm;
1745 	}
1746 
1747 	/* For write statistics */
1748 	if (sb->s_bdev->bd_part)
1749 		sbi->sectors_written_start =
1750 			(u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
1751 
1752 	/* Read accumulated write IO statistics if exists */
1753 	seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
1754 	if (__exist_node_summaries(sbi))
1755 		sbi->kbytes_written =
1756 			le64_to_cpu(seg_i->journal->info.kbytes_written);
1757 
1758 	build_gc_manager(sbi);
1759 
1760 	/* get an inode for node space */
1761 	sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1762 	if (IS_ERR(sbi->node_inode)) {
1763 		f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1764 		err = PTR_ERR(sbi->node_inode);
1765 		goto free_nm;
1766 	}
1767 
1768 	f2fs_join_shrinker(sbi);
1769 
1770 	/* if there are nt orphan nodes free them */
1771 	err = recover_orphan_inodes(sbi);
1772 	if (err)
1773 		goto free_node_inode;
1774 
1775 	/* read root inode and dentry */
1776 	root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1777 	if (IS_ERR(root)) {
1778 		f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1779 		err = PTR_ERR(root);
1780 		goto free_node_inode;
1781 	}
1782 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1783 		iput(root);
1784 		err = -EINVAL;
1785 		goto free_node_inode;
1786 	}
1787 
1788 	sb->s_root = d_make_root(root); /* allocate root dentry */
1789 	if (!sb->s_root) {
1790 		err = -ENOMEM;
1791 		goto free_root_inode;
1792 	}
1793 
1794 	err = f2fs_build_stats(sbi);
1795 	if (err)
1796 		goto free_root_inode;
1797 
1798 	if (f2fs_proc_root)
1799 		sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1800 
1801 	if (sbi->s_proc) {
1802 		proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1803 				 &f2fs_seq_segment_info_fops, sb);
1804 		proc_create_data("segment_bits", S_IRUGO, sbi->s_proc,
1805 				 &f2fs_seq_segment_bits_fops, sb);
1806 	}
1807 
1808 	sbi->s_kobj.kset = f2fs_kset;
1809 	init_completion(&sbi->s_kobj_unregister);
1810 	err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1811 							"%s", sb->s_id);
1812 	if (err)
1813 		goto free_proc;
1814 
1815 	/* recover fsynced data */
1816 	if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1817 		/*
1818 		 * mount should be failed, when device has readonly mode, and
1819 		 * previous checkpoint was not done by clean system shutdown.
1820 		 */
1821 		if (bdev_read_only(sb->s_bdev) &&
1822 				!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
1823 			err = -EROFS;
1824 			goto free_kobj;
1825 		}
1826 
1827 		if (need_fsck)
1828 			set_sbi_flag(sbi, SBI_NEED_FSCK);
1829 
1830 		if (!retry)
1831 			goto skip_recovery;
1832 
1833 		err = recover_fsync_data(sbi, false);
1834 		if (err < 0) {
1835 			need_fsck = true;
1836 			f2fs_msg(sb, KERN_ERR,
1837 				"Cannot recover all fsync data errno=%d", err);
1838 			goto free_kobj;
1839 		}
1840 	} else {
1841 		err = recover_fsync_data(sbi, true);
1842 
1843 		if (!f2fs_readonly(sb) && err > 0) {
1844 			err = -EINVAL;
1845 			f2fs_msg(sb, KERN_ERR,
1846 				"Need to recover fsync data");
1847 			goto free_kobj;
1848 		}
1849 	}
1850 skip_recovery:
1851 	/* recover_fsync_data() cleared this already */
1852 	clear_sbi_flag(sbi, SBI_POR_DOING);
1853 
1854 	/*
1855 	 * If filesystem is not mounted as read-only then
1856 	 * do start the gc_thread.
1857 	 */
1858 	if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
1859 		/* After POR, we can run background GC thread.*/
1860 		err = start_gc_thread(sbi);
1861 		if (err)
1862 			goto free_kobj;
1863 	}
1864 	kfree(options);
1865 
1866 	/* recover broken superblock */
1867 	if (recovery) {
1868 		err = f2fs_commit_super(sbi, true);
1869 		f2fs_msg(sb, KERN_INFO,
1870 			"Try to recover %dth superblock, ret: %d",
1871 			sbi->valid_super_block ? 1 : 2, err);
1872 	}
1873 
1874 	f2fs_update_time(sbi, CP_TIME);
1875 	f2fs_update_time(sbi, REQ_TIME);
1876 	return 0;
1877 
1878 free_kobj:
1879 	f2fs_sync_inode_meta(sbi);
1880 	kobject_del(&sbi->s_kobj);
1881 	kobject_put(&sbi->s_kobj);
1882 	wait_for_completion(&sbi->s_kobj_unregister);
1883 free_proc:
1884 	if (sbi->s_proc) {
1885 		remove_proc_entry("segment_info", sbi->s_proc);
1886 		remove_proc_entry("segment_bits", sbi->s_proc);
1887 		remove_proc_entry(sb->s_id, f2fs_proc_root);
1888 	}
1889 	f2fs_destroy_stats(sbi);
1890 free_root_inode:
1891 	dput(sb->s_root);
1892 	sb->s_root = NULL;
1893 free_node_inode:
1894 	truncate_inode_pages_final(NODE_MAPPING(sbi));
1895 	mutex_lock(&sbi->umount_mutex);
1896 	release_ino_entry(sbi, true);
1897 	f2fs_leave_shrinker(sbi);
1898 	iput(sbi->node_inode);
1899 	mutex_unlock(&sbi->umount_mutex);
1900 free_nm:
1901 	destroy_node_manager(sbi);
1902 free_sm:
1903 	destroy_segment_manager(sbi);
1904 	kfree(sbi->ckpt);
1905 free_meta_inode:
1906 	make_bad_inode(sbi->meta_inode);
1907 	iput(sbi->meta_inode);
1908 free_options:
1909 	destroy_percpu_info(sbi);
1910 	kfree(options);
1911 free_sb_buf:
1912 	kfree(raw_super);
1913 free_sbi:
1914 	if (sbi->s_chksum_driver)
1915 		crypto_free_shash(sbi->s_chksum_driver);
1916 	kfree(sbi);
1917 
1918 	/* give only one another chance */
1919 	if (retry) {
1920 		retry = false;
1921 		shrink_dcache_sb(sb);
1922 		goto try_onemore;
1923 	}
1924 	return err;
1925 }
1926 
1927 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1928 			const char *dev_name, void *data)
1929 {
1930 	return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1931 }
1932 
1933 static void kill_f2fs_super(struct super_block *sb)
1934 {
1935 	if (sb->s_root)
1936 		set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
1937 	kill_block_super(sb);
1938 }
1939 
1940 static struct file_system_type f2fs_fs_type = {
1941 	.owner		= THIS_MODULE,
1942 	.name		= "f2fs",
1943 	.mount		= f2fs_mount,
1944 	.kill_sb	= kill_f2fs_super,
1945 	.fs_flags	= FS_REQUIRES_DEV,
1946 };
1947 MODULE_ALIAS_FS("f2fs");
1948 
1949 static int __init init_inodecache(void)
1950 {
1951 	f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
1952 			sizeof(struct f2fs_inode_info), 0,
1953 			SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
1954 	if (!f2fs_inode_cachep)
1955 		return -ENOMEM;
1956 	return 0;
1957 }
1958 
1959 static void destroy_inodecache(void)
1960 {
1961 	/*
1962 	 * Make sure all delayed rcu free inodes are flushed before we
1963 	 * destroy cache.
1964 	 */
1965 	rcu_barrier();
1966 	kmem_cache_destroy(f2fs_inode_cachep);
1967 }
1968 
1969 static int __init init_f2fs_fs(void)
1970 {
1971 	int err;
1972 
1973 	f2fs_build_trace_ios();
1974 
1975 	err = init_inodecache();
1976 	if (err)
1977 		goto fail;
1978 	err = create_node_manager_caches();
1979 	if (err)
1980 		goto free_inodecache;
1981 	err = create_segment_manager_caches();
1982 	if (err)
1983 		goto free_node_manager_caches;
1984 	err = create_checkpoint_caches();
1985 	if (err)
1986 		goto free_segment_manager_caches;
1987 	err = create_extent_cache();
1988 	if (err)
1989 		goto free_checkpoint_caches;
1990 	f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1991 	if (!f2fs_kset) {
1992 		err = -ENOMEM;
1993 		goto free_extent_cache;
1994 	}
1995 	err = register_shrinker(&f2fs_shrinker_info);
1996 	if (err)
1997 		goto free_kset;
1998 
1999 	err = register_filesystem(&f2fs_fs_type);
2000 	if (err)
2001 		goto free_shrinker;
2002 	err = f2fs_create_root_stats();
2003 	if (err)
2004 		goto free_filesystem;
2005 	f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
2006 	return 0;
2007 
2008 free_filesystem:
2009 	unregister_filesystem(&f2fs_fs_type);
2010 free_shrinker:
2011 	unregister_shrinker(&f2fs_shrinker_info);
2012 free_kset:
2013 	kset_unregister(f2fs_kset);
2014 free_extent_cache:
2015 	destroy_extent_cache();
2016 free_checkpoint_caches:
2017 	destroy_checkpoint_caches();
2018 free_segment_manager_caches:
2019 	destroy_segment_manager_caches();
2020 free_node_manager_caches:
2021 	destroy_node_manager_caches();
2022 free_inodecache:
2023 	destroy_inodecache();
2024 fail:
2025 	return err;
2026 }
2027 
2028 static void __exit exit_f2fs_fs(void)
2029 {
2030 	remove_proc_entry("fs/f2fs", NULL);
2031 	f2fs_destroy_root_stats();
2032 	unregister_filesystem(&f2fs_fs_type);
2033 	unregister_shrinker(&f2fs_shrinker_info);
2034 	kset_unregister(f2fs_kset);
2035 	destroy_extent_cache();
2036 	destroy_checkpoint_caches();
2037 	destroy_segment_manager_caches();
2038 	destroy_node_manager_caches();
2039 	destroy_inodecache();
2040 	f2fs_destroy_trace_ios();
2041 }
2042 
2043 module_init(init_f2fs_fs)
2044 module_exit(exit_f2fs_fs)
2045 
2046 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
2047 MODULE_DESCRIPTION("Flash Friendly File System");
2048 MODULE_LICENSE("GPL");
2049