xref: /openbmc/linux/fs/f2fs/super.c (revision a468f0ef516fda9c7d91bb550d458e853d76955e)
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 struct f2fs_fault_info f2fs_fault;
44 
45 char *fault_name[FAULT_MAX] = {
46 	[FAULT_KMALLOC]		= "kmalloc",
47 	[FAULT_PAGE_ALLOC]	= "page alloc",
48 	[FAULT_ALLOC_NID]	= "alloc nid",
49 	[FAULT_ORPHAN]		= "orphan",
50 	[FAULT_BLOCK]		= "no more block",
51 	[FAULT_DIR_DEPTH]	= "too big dir depth",
52 	[FAULT_EVICT_INODE]	= "evict_inode fail",
53 	[FAULT_IO]		= "IO error",
54 };
55 
56 static void f2fs_build_fault_attr(unsigned int rate)
57 {
58 	if (rate) {
59 		atomic_set(&f2fs_fault.inject_ops, 0);
60 		f2fs_fault.inject_rate = rate;
61 		f2fs_fault.inject_type = (1 << FAULT_MAX) - 1;
62 	} else {
63 		memset(&f2fs_fault, 0, sizeof(struct f2fs_fault_info));
64 	}
65 }
66 #endif
67 
68 /* f2fs-wide shrinker description */
69 static struct shrinker f2fs_shrinker_info = {
70 	.scan_objects = f2fs_shrink_scan,
71 	.count_objects = f2fs_shrink_count,
72 	.seeks = DEFAULT_SEEKS,
73 };
74 
75 enum {
76 	Opt_gc_background,
77 	Opt_disable_roll_forward,
78 	Opt_norecovery,
79 	Opt_discard,
80 	Opt_nodiscard,
81 	Opt_noheap,
82 	Opt_user_xattr,
83 	Opt_nouser_xattr,
84 	Opt_acl,
85 	Opt_noacl,
86 	Opt_active_logs,
87 	Opt_disable_ext_identify,
88 	Opt_inline_xattr,
89 	Opt_inline_data,
90 	Opt_inline_dentry,
91 	Opt_noinline_dentry,
92 	Opt_flush_merge,
93 	Opt_noflush_merge,
94 	Opt_nobarrier,
95 	Opt_fastboot,
96 	Opt_extent_cache,
97 	Opt_noextent_cache,
98 	Opt_noinline_data,
99 	Opt_data_flush,
100 	Opt_mode,
101 	Opt_fault_injection,
102 	Opt_lazytime,
103 	Opt_nolazytime,
104 	Opt_err,
105 };
106 
107 static match_table_t f2fs_tokens = {
108 	{Opt_gc_background, "background_gc=%s"},
109 	{Opt_disable_roll_forward, "disable_roll_forward"},
110 	{Opt_norecovery, "norecovery"},
111 	{Opt_discard, "discard"},
112 	{Opt_nodiscard, "nodiscard"},
113 	{Opt_noheap, "no_heap"},
114 	{Opt_user_xattr, "user_xattr"},
115 	{Opt_nouser_xattr, "nouser_xattr"},
116 	{Opt_acl, "acl"},
117 	{Opt_noacl, "noacl"},
118 	{Opt_active_logs, "active_logs=%u"},
119 	{Opt_disable_ext_identify, "disable_ext_identify"},
120 	{Opt_inline_xattr, "inline_xattr"},
121 	{Opt_inline_data, "inline_data"},
122 	{Opt_inline_dentry, "inline_dentry"},
123 	{Opt_noinline_dentry, "noinline_dentry"},
124 	{Opt_flush_merge, "flush_merge"},
125 	{Opt_noflush_merge, "noflush_merge"},
126 	{Opt_nobarrier, "nobarrier"},
127 	{Opt_fastboot, "fastboot"},
128 	{Opt_extent_cache, "extent_cache"},
129 	{Opt_noextent_cache, "noextent_cache"},
130 	{Opt_noinline_data, "noinline_data"},
131 	{Opt_data_flush, "data_flush"},
132 	{Opt_mode, "mode=%s"},
133 	{Opt_fault_injection, "fault_injection=%u"},
134 	{Opt_lazytime, "lazytime"},
135 	{Opt_nolazytime, "nolazytime"},
136 	{Opt_err, NULL},
137 };
138 
139 /* Sysfs support for f2fs */
140 enum {
141 	GC_THREAD,	/* struct f2fs_gc_thread */
142 	SM_INFO,	/* struct f2fs_sm_info */
143 	NM_INFO,	/* struct f2fs_nm_info */
144 	F2FS_SBI,	/* struct f2fs_sb_info */
145 #ifdef CONFIG_F2FS_FAULT_INJECTION
146 	FAULT_INFO_RATE,	/* struct f2fs_fault_info */
147 	FAULT_INFO_TYPE,	/* struct f2fs_fault_info */
148 #endif
149 };
150 
151 struct f2fs_attr {
152 	struct attribute attr;
153 	ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
154 	ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
155 			 const char *, size_t);
156 	int struct_type;
157 	int offset;
158 };
159 
160 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
161 {
162 	if (struct_type == GC_THREAD)
163 		return (unsigned char *)sbi->gc_thread;
164 	else if (struct_type == SM_INFO)
165 		return (unsigned char *)SM_I(sbi);
166 	else if (struct_type == NM_INFO)
167 		return (unsigned char *)NM_I(sbi);
168 	else if (struct_type == F2FS_SBI)
169 		return (unsigned char *)sbi;
170 #ifdef CONFIG_F2FS_FAULT_INJECTION
171 	else if (struct_type == FAULT_INFO_RATE ||
172 					struct_type == FAULT_INFO_TYPE)
173 		return (unsigned char *)&f2fs_fault;
174 #endif
175 	return NULL;
176 }
177 
178 static ssize_t lifetime_write_kbytes_show(struct f2fs_attr *a,
179 		struct f2fs_sb_info *sbi, char *buf)
180 {
181 	struct super_block *sb = sbi->sb;
182 
183 	if (!sb->s_bdev->bd_part)
184 		return snprintf(buf, PAGE_SIZE, "0\n");
185 
186 	return snprintf(buf, PAGE_SIZE, "%llu\n",
187 		(unsigned long long)(sbi->kbytes_written +
188 			BD_PART_WRITTEN(sbi)));
189 }
190 
191 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
192 			struct f2fs_sb_info *sbi, char *buf)
193 {
194 	unsigned char *ptr = NULL;
195 	unsigned int *ui;
196 
197 	ptr = __struct_ptr(sbi, a->struct_type);
198 	if (!ptr)
199 		return -EINVAL;
200 
201 	ui = (unsigned int *)(ptr + a->offset);
202 
203 	return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
204 }
205 
206 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
207 			struct f2fs_sb_info *sbi,
208 			const char *buf, size_t count)
209 {
210 	unsigned char *ptr;
211 	unsigned long t;
212 	unsigned int *ui;
213 	ssize_t ret;
214 
215 	ptr = __struct_ptr(sbi, a->struct_type);
216 	if (!ptr)
217 		return -EINVAL;
218 
219 	ui = (unsigned int *)(ptr + a->offset);
220 
221 	ret = kstrtoul(skip_spaces(buf), 0, &t);
222 	if (ret < 0)
223 		return ret;
224 #ifdef CONFIG_F2FS_FAULT_INJECTION
225 	if (a->struct_type == FAULT_INFO_TYPE && t >= (1 << FAULT_MAX))
226 		return -EINVAL;
227 #endif
228 	*ui = t;
229 	return count;
230 }
231 
232 static ssize_t f2fs_attr_show(struct kobject *kobj,
233 				struct attribute *attr, char *buf)
234 {
235 	struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
236 								s_kobj);
237 	struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
238 
239 	return a->show ? a->show(a, sbi, buf) : 0;
240 }
241 
242 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
243 						const char *buf, size_t len)
244 {
245 	struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
246 									s_kobj);
247 	struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
248 
249 	return a->store ? a->store(a, sbi, buf, len) : 0;
250 }
251 
252 static void f2fs_sb_release(struct kobject *kobj)
253 {
254 	struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
255 								s_kobj);
256 	complete(&sbi->s_kobj_unregister);
257 }
258 
259 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
260 static struct f2fs_attr f2fs_attr_##_name = {			\
261 	.attr = {.name = __stringify(_name), .mode = _mode },	\
262 	.show	= _show,					\
263 	.store	= _store,					\
264 	.struct_type = _struct_type,				\
265 	.offset = _offset					\
266 }
267 
268 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname)	\
269 	F2FS_ATTR_OFFSET(struct_type, name, 0644,		\
270 		f2fs_sbi_show, f2fs_sbi_store,			\
271 		offsetof(struct struct_name, elname))
272 
273 #define F2FS_GENERAL_RO_ATTR(name) \
274 static struct f2fs_attr f2fs_attr_##name = __ATTR(name, 0444, name##_show, NULL)
275 
276 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
277 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
278 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
279 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
280 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
281 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
282 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, batched_trim_sections, trim_sections);
283 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
284 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
285 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks);
286 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
287 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ra_nid_pages, ra_nid_pages);
288 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, dirty_nats_ratio, dirty_nats_ratio);
289 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
290 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
291 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, cp_interval, interval_time[CP_TIME]);
292 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, idle_interval, interval_time[REQ_TIME]);
293 #ifdef CONFIG_F2FS_FAULT_INJECTION
294 F2FS_RW_ATTR(FAULT_INFO_RATE, f2fs_fault_info, inject_rate, inject_rate);
295 F2FS_RW_ATTR(FAULT_INFO_TYPE, f2fs_fault_info, inject_type, inject_type);
296 #endif
297 F2FS_GENERAL_RO_ATTR(lifetime_write_kbytes);
298 
299 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
300 static struct attribute *f2fs_attrs[] = {
301 	ATTR_LIST(gc_min_sleep_time),
302 	ATTR_LIST(gc_max_sleep_time),
303 	ATTR_LIST(gc_no_gc_sleep_time),
304 	ATTR_LIST(gc_idle),
305 	ATTR_LIST(reclaim_segments),
306 	ATTR_LIST(max_small_discards),
307 	ATTR_LIST(batched_trim_sections),
308 	ATTR_LIST(ipu_policy),
309 	ATTR_LIST(min_ipu_util),
310 	ATTR_LIST(min_fsync_blocks),
311 	ATTR_LIST(max_victim_search),
312 	ATTR_LIST(dir_level),
313 	ATTR_LIST(ram_thresh),
314 	ATTR_LIST(ra_nid_pages),
315 	ATTR_LIST(dirty_nats_ratio),
316 	ATTR_LIST(cp_interval),
317 	ATTR_LIST(idle_interval),
318 	ATTR_LIST(lifetime_write_kbytes),
319 	NULL,
320 };
321 
322 static const struct sysfs_ops f2fs_attr_ops = {
323 	.show	= f2fs_attr_show,
324 	.store	= f2fs_attr_store,
325 };
326 
327 static struct kobj_type f2fs_ktype = {
328 	.default_attrs	= f2fs_attrs,
329 	.sysfs_ops	= &f2fs_attr_ops,
330 	.release	= f2fs_sb_release,
331 };
332 
333 #ifdef CONFIG_F2FS_FAULT_INJECTION
334 /* sysfs for f2fs fault injection */
335 static struct kobject f2fs_fault_inject;
336 
337 static struct attribute *f2fs_fault_attrs[] = {
338 	ATTR_LIST(inject_rate),
339 	ATTR_LIST(inject_type),
340 	NULL
341 };
342 
343 static struct kobj_type f2fs_fault_ktype = {
344 	.default_attrs	= f2fs_fault_attrs,
345 	.sysfs_ops	= &f2fs_attr_ops,
346 };
347 #endif
348 
349 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
350 {
351 	struct va_format vaf;
352 	va_list args;
353 
354 	va_start(args, fmt);
355 	vaf.fmt = fmt;
356 	vaf.va = &args;
357 	printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
358 	va_end(args);
359 }
360 
361 static void init_once(void *foo)
362 {
363 	struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
364 
365 	inode_init_once(&fi->vfs_inode);
366 }
367 
368 static int parse_options(struct super_block *sb, char *options)
369 {
370 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
371 	struct request_queue *q;
372 	substring_t args[MAX_OPT_ARGS];
373 	char *p, *name;
374 	int arg = 0;
375 
376 #ifdef CONFIG_F2FS_FAULT_INJECTION
377 	f2fs_build_fault_attr(0);
378 #endif
379 
380 	if (!options)
381 		return 0;
382 
383 	while ((p = strsep(&options, ",")) != NULL) {
384 		int token;
385 		if (!*p)
386 			continue;
387 		/*
388 		 * Initialize args struct so we know whether arg was
389 		 * found; some options take optional arguments.
390 		 */
391 		args[0].to = args[0].from = NULL;
392 		token = match_token(p, f2fs_tokens, args);
393 
394 		switch (token) {
395 		case Opt_gc_background:
396 			name = match_strdup(&args[0]);
397 
398 			if (!name)
399 				return -ENOMEM;
400 			if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
401 				set_opt(sbi, BG_GC);
402 				clear_opt(sbi, FORCE_FG_GC);
403 			} else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
404 				clear_opt(sbi, BG_GC);
405 				clear_opt(sbi, FORCE_FG_GC);
406 			} else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
407 				set_opt(sbi, BG_GC);
408 				set_opt(sbi, FORCE_FG_GC);
409 			} else {
410 				kfree(name);
411 				return -EINVAL;
412 			}
413 			kfree(name);
414 			break;
415 		case Opt_disable_roll_forward:
416 			set_opt(sbi, DISABLE_ROLL_FORWARD);
417 			break;
418 		case Opt_norecovery:
419 			/* this option mounts f2fs with ro */
420 			set_opt(sbi, DISABLE_ROLL_FORWARD);
421 			if (!f2fs_readonly(sb))
422 				return -EINVAL;
423 			break;
424 		case Opt_discard:
425 			q = bdev_get_queue(sb->s_bdev);
426 			if (blk_queue_discard(q)) {
427 				set_opt(sbi, DISCARD);
428 			} else {
429 				f2fs_msg(sb, KERN_WARNING,
430 					"mounting with \"discard\" option, but "
431 					"the device does not support discard");
432 			}
433 			break;
434 		case Opt_nodiscard:
435 			clear_opt(sbi, DISCARD);
436 		case Opt_noheap:
437 			set_opt(sbi, NOHEAP);
438 			break;
439 #ifdef CONFIG_F2FS_FS_XATTR
440 		case Opt_user_xattr:
441 			set_opt(sbi, XATTR_USER);
442 			break;
443 		case Opt_nouser_xattr:
444 			clear_opt(sbi, XATTR_USER);
445 			break;
446 		case Opt_inline_xattr:
447 			set_opt(sbi, INLINE_XATTR);
448 			break;
449 #else
450 		case Opt_user_xattr:
451 			f2fs_msg(sb, KERN_INFO,
452 				"user_xattr options not supported");
453 			break;
454 		case Opt_nouser_xattr:
455 			f2fs_msg(sb, KERN_INFO,
456 				"nouser_xattr options not supported");
457 			break;
458 		case Opt_inline_xattr:
459 			f2fs_msg(sb, KERN_INFO,
460 				"inline_xattr options not supported");
461 			break;
462 #endif
463 #ifdef CONFIG_F2FS_FS_POSIX_ACL
464 		case Opt_acl:
465 			set_opt(sbi, POSIX_ACL);
466 			break;
467 		case Opt_noacl:
468 			clear_opt(sbi, POSIX_ACL);
469 			break;
470 #else
471 		case Opt_acl:
472 			f2fs_msg(sb, KERN_INFO, "acl options not supported");
473 			break;
474 		case Opt_noacl:
475 			f2fs_msg(sb, KERN_INFO, "noacl options not supported");
476 			break;
477 #endif
478 		case Opt_active_logs:
479 			if (args->from && match_int(args, &arg))
480 				return -EINVAL;
481 			if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
482 				return -EINVAL;
483 			sbi->active_logs = arg;
484 			break;
485 		case Opt_disable_ext_identify:
486 			set_opt(sbi, DISABLE_EXT_IDENTIFY);
487 			break;
488 		case Opt_inline_data:
489 			set_opt(sbi, INLINE_DATA);
490 			break;
491 		case Opt_inline_dentry:
492 			set_opt(sbi, INLINE_DENTRY);
493 			break;
494 		case Opt_noinline_dentry:
495 			clear_opt(sbi, INLINE_DENTRY);
496 			break;
497 		case Opt_flush_merge:
498 			set_opt(sbi, FLUSH_MERGE);
499 			break;
500 		case Opt_noflush_merge:
501 			clear_opt(sbi, FLUSH_MERGE);
502 			break;
503 		case Opt_nobarrier:
504 			set_opt(sbi, NOBARRIER);
505 			break;
506 		case Opt_fastboot:
507 			set_opt(sbi, FASTBOOT);
508 			break;
509 		case Opt_extent_cache:
510 			set_opt(sbi, EXTENT_CACHE);
511 			break;
512 		case Opt_noextent_cache:
513 			clear_opt(sbi, EXTENT_CACHE);
514 			break;
515 		case Opt_noinline_data:
516 			clear_opt(sbi, INLINE_DATA);
517 			break;
518 		case Opt_data_flush:
519 			set_opt(sbi, DATA_FLUSH);
520 			break;
521 		case Opt_mode:
522 			name = match_strdup(&args[0]);
523 
524 			if (!name)
525 				return -ENOMEM;
526 			if (strlen(name) == 8 &&
527 					!strncmp(name, "adaptive", 8)) {
528 				set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
529 			} else if (strlen(name) == 3 &&
530 					!strncmp(name, "lfs", 3)) {
531 				set_opt_mode(sbi, F2FS_MOUNT_LFS);
532 			} else {
533 				kfree(name);
534 				return -EINVAL;
535 			}
536 			kfree(name);
537 			break;
538 		case Opt_fault_injection:
539 			if (args->from && match_int(args, &arg))
540 				return -EINVAL;
541 #ifdef CONFIG_F2FS_FAULT_INJECTION
542 			f2fs_build_fault_attr(arg);
543 #else
544 			f2fs_msg(sb, KERN_INFO,
545 				"FAULT_INJECTION was not selected");
546 #endif
547 			break;
548 		case Opt_lazytime:
549 			sb->s_flags |= MS_LAZYTIME;
550 			break;
551 		case Opt_nolazytime:
552 			sb->s_flags &= ~MS_LAZYTIME;
553 			break;
554 		default:
555 			f2fs_msg(sb, KERN_ERR,
556 				"Unrecognized mount option \"%s\" or missing value",
557 				p);
558 			return -EINVAL;
559 		}
560 	}
561 	return 0;
562 }
563 
564 static struct inode *f2fs_alloc_inode(struct super_block *sb)
565 {
566 	struct f2fs_inode_info *fi;
567 
568 	fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
569 	if (!fi)
570 		return NULL;
571 
572 	init_once((void *) fi);
573 
574 	if (percpu_counter_init(&fi->dirty_pages, 0, GFP_NOFS)) {
575 		kmem_cache_free(f2fs_inode_cachep, fi);
576 		return NULL;
577 	}
578 
579 	/* Initialize f2fs-specific inode info */
580 	fi->vfs_inode.i_version = 1;
581 	fi->i_current_depth = 1;
582 	fi->i_advise = 0;
583 	init_rwsem(&fi->i_sem);
584 	INIT_LIST_HEAD(&fi->dirty_list);
585 	INIT_LIST_HEAD(&fi->gdirty_list);
586 	INIT_LIST_HEAD(&fi->inmem_pages);
587 	mutex_init(&fi->inmem_lock);
588 	init_rwsem(&fi->dio_rwsem[READ]);
589 	init_rwsem(&fi->dio_rwsem[WRITE]);
590 
591 	/* Will be used by directory only */
592 	fi->i_dir_level = F2FS_SB(sb)->dir_level;
593 	return &fi->vfs_inode;
594 }
595 
596 static int f2fs_drop_inode(struct inode *inode)
597 {
598 	/*
599 	 * This is to avoid a deadlock condition like below.
600 	 * writeback_single_inode(inode)
601 	 *  - f2fs_write_data_page
602 	 *    - f2fs_gc -> iput -> evict
603 	 *       - inode_wait_for_writeback(inode)
604 	 */
605 	if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
606 		if (!inode->i_nlink && !is_bad_inode(inode)) {
607 			/* to avoid evict_inode call simultaneously */
608 			atomic_inc(&inode->i_count);
609 			spin_unlock(&inode->i_lock);
610 
611 			/* some remained atomic pages should discarded */
612 			if (f2fs_is_atomic_file(inode))
613 				drop_inmem_pages(inode);
614 
615 			/* should remain fi->extent_tree for writepage */
616 			f2fs_destroy_extent_node(inode);
617 
618 			sb_start_intwrite(inode->i_sb);
619 			f2fs_i_size_write(inode, 0);
620 
621 			if (F2FS_HAS_BLOCKS(inode))
622 				f2fs_truncate(inode);
623 
624 			sb_end_intwrite(inode->i_sb);
625 
626 			fscrypt_put_encryption_info(inode, NULL);
627 			spin_lock(&inode->i_lock);
628 			atomic_dec(&inode->i_count);
629 		}
630 		return 0;
631 	}
632 
633 	return generic_drop_inode(inode);
634 }
635 
636 int f2fs_inode_dirtied(struct inode *inode)
637 {
638 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
639 
640 	spin_lock(&sbi->inode_lock[DIRTY_META]);
641 	if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
642 		spin_unlock(&sbi->inode_lock[DIRTY_META]);
643 		return 1;
644 	}
645 
646 	set_inode_flag(inode, FI_DIRTY_INODE);
647 	list_add_tail(&F2FS_I(inode)->gdirty_list,
648 				&sbi->inode_list[DIRTY_META]);
649 	inc_page_count(sbi, F2FS_DIRTY_IMETA);
650 	stat_inc_dirty_inode(sbi, DIRTY_META);
651 	spin_unlock(&sbi->inode_lock[DIRTY_META]);
652 
653 	return 0;
654 }
655 
656 void f2fs_inode_synced(struct inode *inode)
657 {
658 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
659 
660 	spin_lock(&sbi->inode_lock[DIRTY_META]);
661 	if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
662 		spin_unlock(&sbi->inode_lock[DIRTY_META]);
663 		return;
664 	}
665 	list_del_init(&F2FS_I(inode)->gdirty_list);
666 	clear_inode_flag(inode, FI_DIRTY_INODE);
667 	clear_inode_flag(inode, FI_AUTO_RECOVER);
668 	dec_page_count(sbi, F2FS_DIRTY_IMETA);
669 	stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
670 	spin_unlock(&sbi->inode_lock[DIRTY_META]);
671 }
672 
673 /*
674  * f2fs_dirty_inode() is called from __mark_inode_dirty()
675  *
676  * We should call set_dirty_inode to write the dirty inode through write_inode.
677  */
678 static void f2fs_dirty_inode(struct inode *inode, int flags)
679 {
680 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
681 
682 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
683 			inode->i_ino == F2FS_META_INO(sbi))
684 		return;
685 
686 	if (flags == I_DIRTY_TIME)
687 		return;
688 
689 	if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
690 		clear_inode_flag(inode, FI_AUTO_RECOVER);
691 
692 	f2fs_inode_dirtied(inode);
693 }
694 
695 static void f2fs_i_callback(struct rcu_head *head)
696 {
697 	struct inode *inode = container_of(head, struct inode, i_rcu);
698 	kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
699 }
700 
701 static void f2fs_destroy_inode(struct inode *inode)
702 {
703 	percpu_counter_destroy(&F2FS_I(inode)->dirty_pages);
704 	call_rcu(&inode->i_rcu, f2fs_i_callback);
705 }
706 
707 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
708 {
709 	int i;
710 
711 	for (i = 0; i < NR_COUNT_TYPE; i++)
712 		percpu_counter_destroy(&sbi->nr_pages[i]);
713 	percpu_counter_destroy(&sbi->alloc_valid_block_count);
714 	percpu_counter_destroy(&sbi->total_valid_inode_count);
715 }
716 
717 static void f2fs_put_super(struct super_block *sb)
718 {
719 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
720 
721 	if (sbi->s_proc) {
722 		remove_proc_entry("segment_info", sbi->s_proc);
723 		remove_proc_entry("segment_bits", sbi->s_proc);
724 		remove_proc_entry(sb->s_id, f2fs_proc_root);
725 	}
726 	kobject_del(&sbi->s_kobj);
727 
728 	stop_gc_thread(sbi);
729 
730 	/* prevent remaining shrinker jobs */
731 	mutex_lock(&sbi->umount_mutex);
732 
733 	/*
734 	 * We don't need to do checkpoint when superblock is clean.
735 	 * But, the previous checkpoint was not done by umount, it needs to do
736 	 * clean checkpoint again.
737 	 */
738 	if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
739 			!is_set_ckpt_flags(F2FS_CKPT(sbi), CP_UMOUNT_FLAG)) {
740 		struct cp_control cpc = {
741 			.reason = CP_UMOUNT,
742 		};
743 		write_checkpoint(sbi, &cpc);
744 	}
745 
746 	/* write_checkpoint can update stat informaion */
747 	f2fs_destroy_stats(sbi);
748 
749 	/*
750 	 * normally superblock is clean, so we need to release this.
751 	 * In addition, EIO will skip do checkpoint, we need this as well.
752 	 */
753 	release_ino_entry(sbi, true);
754 	release_discard_addrs(sbi);
755 
756 	f2fs_leave_shrinker(sbi);
757 	mutex_unlock(&sbi->umount_mutex);
758 
759 	/* our cp_error case, we can wait for any writeback page */
760 	f2fs_flush_merged_bios(sbi);
761 
762 	iput(sbi->node_inode);
763 	iput(sbi->meta_inode);
764 
765 	/* destroy f2fs internal modules */
766 	destroy_node_manager(sbi);
767 	destroy_segment_manager(sbi);
768 
769 	kfree(sbi->ckpt);
770 	kobject_put(&sbi->s_kobj);
771 	wait_for_completion(&sbi->s_kobj_unregister);
772 
773 	sb->s_fs_info = NULL;
774 	if (sbi->s_chksum_driver)
775 		crypto_free_shash(sbi->s_chksum_driver);
776 	kfree(sbi->raw_super);
777 
778 	destroy_percpu_info(sbi);
779 	kfree(sbi);
780 }
781 
782 int f2fs_sync_fs(struct super_block *sb, int sync)
783 {
784 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
785 	int err = 0;
786 
787 	trace_f2fs_sync_fs(sb, sync);
788 
789 	if (sync) {
790 		struct cp_control cpc;
791 
792 		cpc.reason = __get_cp_reason(sbi);
793 
794 		mutex_lock(&sbi->gc_mutex);
795 		err = write_checkpoint(sbi, &cpc);
796 		mutex_unlock(&sbi->gc_mutex);
797 	}
798 	f2fs_trace_ios(NULL, 1);
799 
800 	return err;
801 }
802 
803 static int f2fs_freeze(struct super_block *sb)
804 {
805 	int err;
806 
807 	if (f2fs_readonly(sb))
808 		return 0;
809 
810 	err = f2fs_sync_fs(sb, 1);
811 	return err;
812 }
813 
814 static int f2fs_unfreeze(struct super_block *sb)
815 {
816 	return 0;
817 }
818 
819 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
820 {
821 	struct super_block *sb = dentry->d_sb;
822 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
823 	u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
824 	block_t total_count, user_block_count, start_count, ovp_count;
825 
826 	total_count = le64_to_cpu(sbi->raw_super->block_count);
827 	user_block_count = sbi->user_block_count;
828 	start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
829 	ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
830 	buf->f_type = F2FS_SUPER_MAGIC;
831 	buf->f_bsize = sbi->blocksize;
832 
833 	buf->f_blocks = total_count - start_count;
834 	buf->f_bfree = user_block_count - valid_user_blocks(sbi) + ovp_count;
835 	buf->f_bavail = user_block_count - valid_user_blocks(sbi);
836 
837 	buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
838 	buf->f_ffree = buf->f_files - valid_inode_count(sbi);
839 
840 	buf->f_namelen = F2FS_NAME_LEN;
841 	buf->f_fsid.val[0] = (u32)id;
842 	buf->f_fsid.val[1] = (u32)(id >> 32);
843 
844 	return 0;
845 }
846 
847 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
848 {
849 	struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
850 
851 	if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
852 		if (test_opt(sbi, FORCE_FG_GC))
853 			seq_printf(seq, ",background_gc=%s", "sync");
854 		else
855 			seq_printf(seq, ",background_gc=%s", "on");
856 	} else {
857 		seq_printf(seq, ",background_gc=%s", "off");
858 	}
859 	if (test_opt(sbi, DISABLE_ROLL_FORWARD))
860 		seq_puts(seq, ",disable_roll_forward");
861 	if (test_opt(sbi, DISCARD))
862 		seq_puts(seq, ",discard");
863 	if (test_opt(sbi, NOHEAP))
864 		seq_puts(seq, ",no_heap_alloc");
865 #ifdef CONFIG_F2FS_FS_XATTR
866 	if (test_opt(sbi, XATTR_USER))
867 		seq_puts(seq, ",user_xattr");
868 	else
869 		seq_puts(seq, ",nouser_xattr");
870 	if (test_opt(sbi, INLINE_XATTR))
871 		seq_puts(seq, ",inline_xattr");
872 #endif
873 #ifdef CONFIG_F2FS_FS_POSIX_ACL
874 	if (test_opt(sbi, POSIX_ACL))
875 		seq_puts(seq, ",acl");
876 	else
877 		seq_puts(seq, ",noacl");
878 #endif
879 	if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
880 		seq_puts(seq, ",disable_ext_identify");
881 	if (test_opt(sbi, INLINE_DATA))
882 		seq_puts(seq, ",inline_data");
883 	else
884 		seq_puts(seq, ",noinline_data");
885 	if (test_opt(sbi, INLINE_DENTRY))
886 		seq_puts(seq, ",inline_dentry");
887 	else
888 		seq_puts(seq, ",noinline_dentry");
889 	if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
890 		seq_puts(seq, ",flush_merge");
891 	if (test_opt(sbi, NOBARRIER))
892 		seq_puts(seq, ",nobarrier");
893 	if (test_opt(sbi, FASTBOOT))
894 		seq_puts(seq, ",fastboot");
895 	if (test_opt(sbi, EXTENT_CACHE))
896 		seq_puts(seq, ",extent_cache");
897 	else
898 		seq_puts(seq, ",noextent_cache");
899 	if (test_opt(sbi, DATA_FLUSH))
900 		seq_puts(seq, ",data_flush");
901 
902 	seq_puts(seq, ",mode=");
903 	if (test_opt(sbi, ADAPTIVE))
904 		seq_puts(seq, "adaptive");
905 	else if (test_opt(sbi, LFS))
906 		seq_puts(seq, "lfs");
907 	seq_printf(seq, ",active_logs=%u", sbi->active_logs);
908 
909 	return 0;
910 }
911 
912 static int segment_info_seq_show(struct seq_file *seq, void *offset)
913 {
914 	struct super_block *sb = seq->private;
915 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
916 	unsigned int total_segs =
917 			le32_to_cpu(sbi->raw_super->segment_count_main);
918 	int i;
919 
920 	seq_puts(seq, "format: segment_type|valid_blocks\n"
921 		"segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
922 
923 	for (i = 0; i < total_segs; i++) {
924 		struct seg_entry *se = get_seg_entry(sbi, i);
925 
926 		if ((i % 10) == 0)
927 			seq_printf(seq, "%-10d", i);
928 		seq_printf(seq, "%d|%-3u", se->type,
929 					get_valid_blocks(sbi, i, 1));
930 		if ((i % 10) == 9 || i == (total_segs - 1))
931 			seq_putc(seq, '\n');
932 		else
933 			seq_putc(seq, ' ');
934 	}
935 
936 	return 0;
937 }
938 
939 static int segment_bits_seq_show(struct seq_file *seq, void *offset)
940 {
941 	struct super_block *sb = seq->private;
942 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
943 	unsigned int total_segs =
944 			le32_to_cpu(sbi->raw_super->segment_count_main);
945 	int i, j;
946 
947 	seq_puts(seq, "format: segment_type|valid_blocks|bitmaps\n"
948 		"segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
949 
950 	for (i = 0; i < total_segs; i++) {
951 		struct seg_entry *se = get_seg_entry(sbi, i);
952 
953 		seq_printf(seq, "%-10d", i);
954 		seq_printf(seq, "%d|%-3u|", se->type,
955 					get_valid_blocks(sbi, i, 1));
956 		for (j = 0; j < SIT_VBLOCK_MAP_SIZE; j++)
957 			seq_printf(seq, "%x ", se->cur_valid_map[j]);
958 		seq_putc(seq, '\n');
959 	}
960 	return 0;
961 }
962 
963 #define F2FS_PROC_FILE_DEF(_name)					\
964 static int _name##_open_fs(struct inode *inode, struct file *file)	\
965 {									\
966 	return single_open(file, _name##_seq_show, PDE_DATA(inode));	\
967 }									\
968 									\
969 static const struct file_operations f2fs_seq_##_name##_fops = {		\
970 	.open = _name##_open_fs,					\
971 	.read = seq_read,						\
972 	.llseek = seq_lseek,						\
973 	.release = single_release,					\
974 };
975 
976 F2FS_PROC_FILE_DEF(segment_info);
977 F2FS_PROC_FILE_DEF(segment_bits);
978 
979 static void default_options(struct f2fs_sb_info *sbi)
980 {
981 	/* init some FS parameters */
982 	sbi->active_logs = NR_CURSEG_TYPE;
983 
984 	set_opt(sbi, BG_GC);
985 	set_opt(sbi, INLINE_DATA);
986 	set_opt(sbi, INLINE_DENTRY);
987 	set_opt(sbi, EXTENT_CACHE);
988 	sbi->sb->s_flags |= MS_LAZYTIME;
989 	set_opt(sbi, FLUSH_MERGE);
990 	if (f2fs_sb_mounted_hmsmr(sbi->sb)) {
991 		set_opt_mode(sbi, F2FS_MOUNT_LFS);
992 		set_opt(sbi, DISCARD);
993 	} else {
994 		set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
995 	}
996 
997 #ifdef CONFIG_F2FS_FS_XATTR
998 	set_opt(sbi, XATTR_USER);
999 #endif
1000 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1001 	set_opt(sbi, POSIX_ACL);
1002 #endif
1003 }
1004 
1005 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1006 {
1007 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1008 	struct f2fs_mount_info org_mount_opt;
1009 	int err, active_logs;
1010 	bool need_restart_gc = false;
1011 	bool need_stop_gc = false;
1012 	bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1013 
1014 	/*
1015 	 * Save the old mount options in case we
1016 	 * need to restore them.
1017 	 */
1018 	org_mount_opt = sbi->mount_opt;
1019 	active_logs = sbi->active_logs;
1020 
1021 	/* recover superblocks we couldn't write due to previous RO mount */
1022 	if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1023 		err = f2fs_commit_super(sbi, false);
1024 		f2fs_msg(sb, KERN_INFO,
1025 			"Try to recover all the superblocks, ret: %d", err);
1026 		if (!err)
1027 			clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1028 	}
1029 
1030 	sbi->mount_opt.opt = 0;
1031 	default_options(sbi);
1032 
1033 	/* parse mount options */
1034 	err = parse_options(sb, data);
1035 	if (err)
1036 		goto restore_opts;
1037 
1038 	/*
1039 	 * Previous and new state of filesystem is RO,
1040 	 * so skip checking GC and FLUSH_MERGE conditions.
1041 	 */
1042 	if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
1043 		goto skip;
1044 
1045 	/* disallow enable/disable extent_cache dynamically */
1046 	if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1047 		err = -EINVAL;
1048 		f2fs_msg(sbi->sb, KERN_WARNING,
1049 				"switch extent_cache option is not allowed");
1050 		goto restore_opts;
1051 	}
1052 
1053 	/*
1054 	 * We stop the GC thread if FS is mounted as RO
1055 	 * or if background_gc = off is passed in mount
1056 	 * option. Also sync the filesystem.
1057 	 */
1058 	if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
1059 		if (sbi->gc_thread) {
1060 			stop_gc_thread(sbi);
1061 			need_restart_gc = true;
1062 		}
1063 	} else if (!sbi->gc_thread) {
1064 		err = start_gc_thread(sbi);
1065 		if (err)
1066 			goto restore_opts;
1067 		need_stop_gc = true;
1068 	}
1069 
1070 	if (*flags & MS_RDONLY) {
1071 		writeback_inodes_sb(sb, WB_REASON_SYNC);
1072 		sync_inodes_sb(sb);
1073 
1074 		set_sbi_flag(sbi, SBI_IS_DIRTY);
1075 		set_sbi_flag(sbi, SBI_IS_CLOSE);
1076 		f2fs_sync_fs(sb, 1);
1077 		clear_sbi_flag(sbi, SBI_IS_CLOSE);
1078 	}
1079 
1080 	/*
1081 	 * We stop issue flush thread if FS is mounted as RO
1082 	 * or if flush_merge is not passed in mount option.
1083 	 */
1084 	if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1085 		destroy_flush_cmd_control(sbi);
1086 	} else if (!SM_I(sbi)->cmd_control_info) {
1087 		err = create_flush_cmd_control(sbi);
1088 		if (err)
1089 			goto restore_gc;
1090 	}
1091 skip:
1092 	/* Update the POSIXACL Flag */
1093 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1094 		(test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1095 
1096 	return 0;
1097 restore_gc:
1098 	if (need_restart_gc) {
1099 		if (start_gc_thread(sbi))
1100 			f2fs_msg(sbi->sb, KERN_WARNING,
1101 				"background gc thread has stopped");
1102 	} else if (need_stop_gc) {
1103 		stop_gc_thread(sbi);
1104 	}
1105 restore_opts:
1106 	sbi->mount_opt = org_mount_opt;
1107 	sbi->active_logs = active_logs;
1108 	return err;
1109 }
1110 
1111 static struct super_operations f2fs_sops = {
1112 	.alloc_inode	= f2fs_alloc_inode,
1113 	.drop_inode	= f2fs_drop_inode,
1114 	.destroy_inode	= f2fs_destroy_inode,
1115 	.write_inode	= f2fs_write_inode,
1116 	.dirty_inode	= f2fs_dirty_inode,
1117 	.show_options	= f2fs_show_options,
1118 	.evict_inode	= f2fs_evict_inode,
1119 	.put_super	= f2fs_put_super,
1120 	.sync_fs	= f2fs_sync_fs,
1121 	.freeze_fs	= f2fs_freeze,
1122 	.unfreeze_fs	= f2fs_unfreeze,
1123 	.statfs		= f2fs_statfs,
1124 	.remount_fs	= f2fs_remount,
1125 };
1126 
1127 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1128 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1129 {
1130 	return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1131 				F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1132 				ctx, len, NULL);
1133 }
1134 
1135 static int f2fs_key_prefix(struct inode *inode, u8 **key)
1136 {
1137 	*key = F2FS_I_SB(inode)->key_prefix;
1138 	return F2FS_I_SB(inode)->key_prefix_size;
1139 }
1140 
1141 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1142 							void *fs_data)
1143 {
1144 	return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1145 				F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1146 				ctx, len, fs_data, XATTR_CREATE);
1147 }
1148 
1149 static unsigned f2fs_max_namelen(struct inode *inode)
1150 {
1151 	return S_ISLNK(inode->i_mode) ?
1152 			inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1153 }
1154 
1155 static struct fscrypt_operations f2fs_cryptops = {
1156 	.get_context	= f2fs_get_context,
1157 	.key_prefix	= f2fs_key_prefix,
1158 	.set_context	= f2fs_set_context,
1159 	.is_encrypted	= f2fs_encrypted_inode,
1160 	.empty_dir	= f2fs_empty_dir,
1161 	.max_namelen	= f2fs_max_namelen,
1162 };
1163 #else
1164 static struct fscrypt_operations f2fs_cryptops = {
1165 	.is_encrypted	= f2fs_encrypted_inode,
1166 };
1167 #endif
1168 
1169 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1170 		u64 ino, u32 generation)
1171 {
1172 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1173 	struct inode *inode;
1174 
1175 	if (check_nid_range(sbi, ino))
1176 		return ERR_PTR(-ESTALE);
1177 
1178 	/*
1179 	 * f2fs_iget isn't quite right if the inode is currently unallocated!
1180 	 * However f2fs_iget currently does appropriate checks to handle stale
1181 	 * inodes so everything is OK.
1182 	 */
1183 	inode = f2fs_iget(sb, ino);
1184 	if (IS_ERR(inode))
1185 		return ERR_CAST(inode);
1186 	if (unlikely(generation && inode->i_generation != generation)) {
1187 		/* we didn't find the right inode.. */
1188 		iput(inode);
1189 		return ERR_PTR(-ESTALE);
1190 	}
1191 	return inode;
1192 }
1193 
1194 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1195 		int fh_len, int fh_type)
1196 {
1197 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1198 				    f2fs_nfs_get_inode);
1199 }
1200 
1201 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1202 		int fh_len, int fh_type)
1203 {
1204 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1205 				    f2fs_nfs_get_inode);
1206 }
1207 
1208 static const struct export_operations f2fs_export_ops = {
1209 	.fh_to_dentry = f2fs_fh_to_dentry,
1210 	.fh_to_parent = f2fs_fh_to_parent,
1211 	.get_parent = f2fs_get_parent,
1212 };
1213 
1214 static loff_t max_file_blocks(void)
1215 {
1216 	loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
1217 	loff_t leaf_count = ADDRS_PER_BLOCK;
1218 
1219 	/* two direct node blocks */
1220 	result += (leaf_count * 2);
1221 
1222 	/* two indirect node blocks */
1223 	leaf_count *= NIDS_PER_BLOCK;
1224 	result += (leaf_count * 2);
1225 
1226 	/* one double indirect node block */
1227 	leaf_count *= NIDS_PER_BLOCK;
1228 	result += leaf_count;
1229 
1230 	return result;
1231 }
1232 
1233 static int __f2fs_commit_super(struct buffer_head *bh,
1234 			struct f2fs_super_block *super)
1235 {
1236 	lock_buffer(bh);
1237 	if (super)
1238 		memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
1239 	set_buffer_uptodate(bh);
1240 	set_buffer_dirty(bh);
1241 	unlock_buffer(bh);
1242 
1243 	/* it's rare case, we can do fua all the time */
1244 	return __sync_dirty_buffer(bh, WRITE_FLUSH_FUA);
1245 }
1246 
1247 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
1248 					struct buffer_head *bh)
1249 {
1250 	struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1251 					(bh->b_data + F2FS_SUPER_OFFSET);
1252 	struct super_block *sb = sbi->sb;
1253 	u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
1254 	u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
1255 	u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
1256 	u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
1257 	u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
1258 	u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
1259 	u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
1260 	u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
1261 	u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
1262 	u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
1263 	u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1264 	u32 segment_count = le32_to_cpu(raw_super->segment_count);
1265 	u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1266 	u64 main_end_blkaddr = main_blkaddr +
1267 				(segment_count_main << log_blocks_per_seg);
1268 	u64 seg_end_blkaddr = segment0_blkaddr +
1269 				(segment_count << log_blocks_per_seg);
1270 
1271 	if (segment0_blkaddr != cp_blkaddr) {
1272 		f2fs_msg(sb, KERN_INFO,
1273 			"Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1274 			segment0_blkaddr, cp_blkaddr);
1275 		return true;
1276 	}
1277 
1278 	if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
1279 							sit_blkaddr) {
1280 		f2fs_msg(sb, KERN_INFO,
1281 			"Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1282 			cp_blkaddr, sit_blkaddr,
1283 			segment_count_ckpt << log_blocks_per_seg);
1284 		return true;
1285 	}
1286 
1287 	if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
1288 							nat_blkaddr) {
1289 		f2fs_msg(sb, KERN_INFO,
1290 			"Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1291 			sit_blkaddr, nat_blkaddr,
1292 			segment_count_sit << log_blocks_per_seg);
1293 		return true;
1294 	}
1295 
1296 	if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
1297 							ssa_blkaddr) {
1298 		f2fs_msg(sb, KERN_INFO,
1299 			"Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1300 			nat_blkaddr, ssa_blkaddr,
1301 			segment_count_nat << log_blocks_per_seg);
1302 		return true;
1303 	}
1304 
1305 	if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
1306 							main_blkaddr) {
1307 		f2fs_msg(sb, KERN_INFO,
1308 			"Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1309 			ssa_blkaddr, main_blkaddr,
1310 			segment_count_ssa << log_blocks_per_seg);
1311 		return true;
1312 	}
1313 
1314 	if (main_end_blkaddr > seg_end_blkaddr) {
1315 		f2fs_msg(sb, KERN_INFO,
1316 			"Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1317 			main_blkaddr,
1318 			segment0_blkaddr +
1319 				(segment_count << log_blocks_per_seg),
1320 			segment_count_main << log_blocks_per_seg);
1321 		return true;
1322 	} else if (main_end_blkaddr < seg_end_blkaddr) {
1323 		int err = 0;
1324 		char *res;
1325 
1326 		/* fix in-memory information all the time */
1327 		raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
1328 				segment0_blkaddr) >> log_blocks_per_seg);
1329 
1330 		if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
1331 			set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1332 			res = "internally";
1333 		} else {
1334 			err = __f2fs_commit_super(bh, NULL);
1335 			res = err ? "failed" : "done";
1336 		}
1337 		f2fs_msg(sb, KERN_INFO,
1338 			"Fix alignment : %s, start(%u) end(%u) block(%u)",
1339 			res, main_blkaddr,
1340 			segment0_blkaddr +
1341 				(segment_count << log_blocks_per_seg),
1342 			segment_count_main << log_blocks_per_seg);
1343 		if (err)
1344 			return true;
1345 	}
1346 	return false;
1347 }
1348 
1349 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
1350 				struct buffer_head *bh)
1351 {
1352 	struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1353 					(bh->b_data + F2FS_SUPER_OFFSET);
1354 	struct super_block *sb = sbi->sb;
1355 	unsigned int blocksize;
1356 
1357 	if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
1358 		f2fs_msg(sb, KERN_INFO,
1359 			"Magic Mismatch, valid(0x%x) - read(0x%x)",
1360 			F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
1361 		return 1;
1362 	}
1363 
1364 	/* Currently, support only 4KB page cache size */
1365 	if (F2FS_BLKSIZE != PAGE_SIZE) {
1366 		f2fs_msg(sb, KERN_INFO,
1367 			"Invalid page_cache_size (%lu), supports only 4KB\n",
1368 			PAGE_SIZE);
1369 		return 1;
1370 	}
1371 
1372 	/* Currently, support only 4KB block size */
1373 	blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
1374 	if (blocksize != F2FS_BLKSIZE) {
1375 		f2fs_msg(sb, KERN_INFO,
1376 			"Invalid blocksize (%u), supports only 4KB\n",
1377 			blocksize);
1378 		return 1;
1379 	}
1380 
1381 	/* check log blocks per segment */
1382 	if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
1383 		f2fs_msg(sb, KERN_INFO,
1384 			"Invalid log blocks per segment (%u)\n",
1385 			le32_to_cpu(raw_super->log_blocks_per_seg));
1386 		return 1;
1387 	}
1388 
1389 	/* Currently, support 512/1024/2048/4096 bytes sector size */
1390 	if (le32_to_cpu(raw_super->log_sectorsize) >
1391 				F2FS_MAX_LOG_SECTOR_SIZE ||
1392 		le32_to_cpu(raw_super->log_sectorsize) <
1393 				F2FS_MIN_LOG_SECTOR_SIZE) {
1394 		f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
1395 			le32_to_cpu(raw_super->log_sectorsize));
1396 		return 1;
1397 	}
1398 	if (le32_to_cpu(raw_super->log_sectors_per_block) +
1399 		le32_to_cpu(raw_super->log_sectorsize) !=
1400 			F2FS_MAX_LOG_SECTOR_SIZE) {
1401 		f2fs_msg(sb, KERN_INFO,
1402 			"Invalid log sectors per block(%u) log sectorsize(%u)",
1403 			le32_to_cpu(raw_super->log_sectors_per_block),
1404 			le32_to_cpu(raw_super->log_sectorsize));
1405 		return 1;
1406 	}
1407 
1408 	/* check reserved ino info */
1409 	if (le32_to_cpu(raw_super->node_ino) != 1 ||
1410 		le32_to_cpu(raw_super->meta_ino) != 2 ||
1411 		le32_to_cpu(raw_super->root_ino) != 3) {
1412 		f2fs_msg(sb, KERN_INFO,
1413 			"Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1414 			le32_to_cpu(raw_super->node_ino),
1415 			le32_to_cpu(raw_super->meta_ino),
1416 			le32_to_cpu(raw_super->root_ino));
1417 		return 1;
1418 	}
1419 
1420 	/* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1421 	if (sanity_check_area_boundary(sbi, bh))
1422 		return 1;
1423 
1424 	return 0;
1425 }
1426 
1427 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1428 {
1429 	unsigned int total, fsmeta;
1430 	struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1431 	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1432 
1433 	total = le32_to_cpu(raw_super->segment_count);
1434 	fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
1435 	fsmeta += le32_to_cpu(raw_super->segment_count_sit);
1436 	fsmeta += le32_to_cpu(raw_super->segment_count_nat);
1437 	fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
1438 	fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
1439 
1440 	if (unlikely(fsmeta >= total))
1441 		return 1;
1442 
1443 	if (unlikely(f2fs_cp_error(sbi))) {
1444 		f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
1445 		return 1;
1446 	}
1447 	return 0;
1448 }
1449 
1450 static void init_sb_info(struct f2fs_sb_info *sbi)
1451 {
1452 	struct f2fs_super_block *raw_super = sbi->raw_super;
1453 
1454 	sbi->log_sectors_per_block =
1455 		le32_to_cpu(raw_super->log_sectors_per_block);
1456 	sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
1457 	sbi->blocksize = 1 << sbi->log_blocksize;
1458 	sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1459 	sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
1460 	sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1461 	sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1462 	sbi->total_sections = le32_to_cpu(raw_super->section_count);
1463 	sbi->total_node_count =
1464 		(le32_to_cpu(raw_super->segment_count_nat) / 2)
1465 			* sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
1466 	sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
1467 	sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
1468 	sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
1469 	sbi->cur_victim_sec = NULL_SECNO;
1470 	sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
1471 
1472 	sbi->dir_level = DEF_DIR_LEVEL;
1473 	sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
1474 	sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
1475 	clear_sbi_flag(sbi, SBI_NEED_FSCK);
1476 
1477 	INIT_LIST_HEAD(&sbi->s_list);
1478 	mutex_init(&sbi->umount_mutex);
1479 	mutex_init(&sbi->wio_mutex[NODE]);
1480 	mutex_init(&sbi->wio_mutex[DATA]);
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->ckpt, 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 	mutex_lock(&sbi->umount_mutex);
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 #ifdef CONFIG_F2FS_FAULT_INJECTION
1994 	f2fs_fault_inject.kset = f2fs_kset;
1995 	f2fs_build_fault_attr(0);
1996 	err = kobject_init_and_add(&f2fs_fault_inject, &f2fs_fault_ktype,
1997 				NULL, "fault_injection");
1998 	if (err) {
1999 		f2fs_fault_inject.kset = NULL;
2000 		goto free_kset;
2001 	}
2002 #endif
2003 	err = register_shrinker(&f2fs_shrinker_info);
2004 	if (err)
2005 		goto free_kset;
2006 
2007 	err = register_filesystem(&f2fs_fs_type);
2008 	if (err)
2009 		goto free_shrinker;
2010 	err = f2fs_create_root_stats();
2011 	if (err)
2012 		goto free_filesystem;
2013 	f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
2014 	return 0;
2015 
2016 free_filesystem:
2017 	unregister_filesystem(&f2fs_fs_type);
2018 free_shrinker:
2019 	unregister_shrinker(&f2fs_shrinker_info);
2020 free_kset:
2021 #ifdef CONFIG_F2FS_FAULT_INJECTION
2022 	if (f2fs_fault_inject.kset)
2023 		kobject_put(&f2fs_fault_inject);
2024 #endif
2025 	kset_unregister(f2fs_kset);
2026 free_extent_cache:
2027 	destroy_extent_cache();
2028 free_checkpoint_caches:
2029 	destroy_checkpoint_caches();
2030 free_segment_manager_caches:
2031 	destroy_segment_manager_caches();
2032 free_node_manager_caches:
2033 	destroy_node_manager_caches();
2034 free_inodecache:
2035 	destroy_inodecache();
2036 fail:
2037 	return err;
2038 }
2039 
2040 static void __exit exit_f2fs_fs(void)
2041 {
2042 	remove_proc_entry("fs/f2fs", NULL);
2043 	f2fs_destroy_root_stats();
2044 	unregister_filesystem(&f2fs_fs_type);
2045 	unregister_shrinker(&f2fs_shrinker_info);
2046 #ifdef CONFIG_F2FS_FAULT_INJECTION
2047 	kobject_put(&f2fs_fault_inject);
2048 #endif
2049 	kset_unregister(f2fs_kset);
2050 	destroy_extent_cache();
2051 	destroy_checkpoint_caches();
2052 	destroy_segment_manager_caches();
2053 	destroy_node_manager_caches();
2054 	destroy_inodecache();
2055 	f2fs_destroy_trace_ios();
2056 }
2057 
2058 module_init(init_f2fs_fs)
2059 module_exit(exit_f2fs_fs)
2060 
2061 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
2062 MODULE_DESCRIPTION("Flash Friendly File System");
2063 MODULE_LICENSE("GPL");
2064