xref: /linux/fs/f2fs/super.c (revision 80d443e8876602be2c130f79c4de81e12e2a700d)
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 if (!f2fs_sb_mounted_blkzoned(sb)) {
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 			if (f2fs_sb_mounted_blkzoned(sb)) {
423 				f2fs_msg(sb, KERN_WARNING,
424 					"discard is required for zoned block devices");
425 				return -EINVAL;
426 			}
427 			clear_opt(sbi, DISCARD);
428 			break;
429 		case Opt_noheap:
430 			set_opt(sbi, NOHEAP);
431 			break;
432 #ifdef CONFIG_F2FS_FS_XATTR
433 		case Opt_user_xattr:
434 			set_opt(sbi, XATTR_USER);
435 			break;
436 		case Opt_nouser_xattr:
437 			clear_opt(sbi, XATTR_USER);
438 			break;
439 		case Opt_inline_xattr:
440 			set_opt(sbi, INLINE_XATTR);
441 			break;
442 #else
443 		case Opt_user_xattr:
444 			f2fs_msg(sb, KERN_INFO,
445 				"user_xattr options not supported");
446 			break;
447 		case Opt_nouser_xattr:
448 			f2fs_msg(sb, KERN_INFO,
449 				"nouser_xattr options not supported");
450 			break;
451 		case Opt_inline_xattr:
452 			f2fs_msg(sb, KERN_INFO,
453 				"inline_xattr options not supported");
454 			break;
455 #endif
456 #ifdef CONFIG_F2FS_FS_POSIX_ACL
457 		case Opt_acl:
458 			set_opt(sbi, POSIX_ACL);
459 			break;
460 		case Opt_noacl:
461 			clear_opt(sbi, POSIX_ACL);
462 			break;
463 #else
464 		case Opt_acl:
465 			f2fs_msg(sb, KERN_INFO, "acl options not supported");
466 			break;
467 		case Opt_noacl:
468 			f2fs_msg(sb, KERN_INFO, "noacl options not supported");
469 			break;
470 #endif
471 		case Opt_active_logs:
472 			if (args->from && match_int(args, &arg))
473 				return -EINVAL;
474 			if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
475 				return -EINVAL;
476 			sbi->active_logs = arg;
477 			break;
478 		case Opt_disable_ext_identify:
479 			set_opt(sbi, DISABLE_EXT_IDENTIFY);
480 			break;
481 		case Opt_inline_data:
482 			set_opt(sbi, INLINE_DATA);
483 			break;
484 		case Opt_inline_dentry:
485 			set_opt(sbi, INLINE_DENTRY);
486 			break;
487 		case Opt_noinline_dentry:
488 			clear_opt(sbi, INLINE_DENTRY);
489 			break;
490 		case Opt_flush_merge:
491 			set_opt(sbi, FLUSH_MERGE);
492 			break;
493 		case Opt_noflush_merge:
494 			clear_opt(sbi, FLUSH_MERGE);
495 			break;
496 		case Opt_nobarrier:
497 			set_opt(sbi, NOBARRIER);
498 			break;
499 		case Opt_fastboot:
500 			set_opt(sbi, FASTBOOT);
501 			break;
502 		case Opt_extent_cache:
503 			set_opt(sbi, EXTENT_CACHE);
504 			break;
505 		case Opt_noextent_cache:
506 			clear_opt(sbi, EXTENT_CACHE);
507 			break;
508 		case Opt_noinline_data:
509 			clear_opt(sbi, INLINE_DATA);
510 			break;
511 		case Opt_data_flush:
512 			set_opt(sbi, DATA_FLUSH);
513 			break;
514 		case Opt_mode:
515 			name = match_strdup(&args[0]);
516 
517 			if (!name)
518 				return -ENOMEM;
519 			if (strlen(name) == 8 &&
520 					!strncmp(name, "adaptive", 8)) {
521 				if (f2fs_sb_mounted_blkzoned(sb)) {
522 					f2fs_msg(sb, KERN_WARNING,
523 						 "adaptive mode is not allowed with "
524 						 "zoned block device feature");
525 					kfree(name);
526 					return -EINVAL;
527 				}
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(sbi, 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 	/* Initialize f2fs-specific inode info */
575 	fi->vfs_inode.i_version = 1;
576 	atomic_set(&fi->dirty_pages, 0);
577 	fi->i_current_depth = 1;
578 	fi->i_advise = 0;
579 	init_rwsem(&fi->i_sem);
580 	INIT_LIST_HEAD(&fi->dirty_list);
581 	INIT_LIST_HEAD(&fi->gdirty_list);
582 	INIT_LIST_HEAD(&fi->inmem_pages);
583 	mutex_init(&fi->inmem_lock);
584 	init_rwsem(&fi->dio_rwsem[READ]);
585 	init_rwsem(&fi->dio_rwsem[WRITE]);
586 
587 	/* Will be used by directory only */
588 	fi->i_dir_level = F2FS_SB(sb)->dir_level;
589 	return &fi->vfs_inode;
590 }
591 
592 static int f2fs_drop_inode(struct inode *inode)
593 {
594 	/*
595 	 * This is to avoid a deadlock condition like below.
596 	 * writeback_single_inode(inode)
597 	 *  - f2fs_write_data_page
598 	 *    - f2fs_gc -> iput -> evict
599 	 *       - inode_wait_for_writeback(inode)
600 	 */
601 	if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
602 		if (!inode->i_nlink && !is_bad_inode(inode)) {
603 			/* to avoid evict_inode call simultaneously */
604 			atomic_inc(&inode->i_count);
605 			spin_unlock(&inode->i_lock);
606 
607 			/* some remained atomic pages should discarded */
608 			if (f2fs_is_atomic_file(inode))
609 				drop_inmem_pages(inode);
610 
611 			/* should remain fi->extent_tree for writepage */
612 			f2fs_destroy_extent_node(inode);
613 
614 			sb_start_intwrite(inode->i_sb);
615 			f2fs_i_size_write(inode, 0);
616 
617 			if (F2FS_HAS_BLOCKS(inode))
618 				f2fs_truncate(inode);
619 
620 			sb_end_intwrite(inode->i_sb);
621 
622 			fscrypt_put_encryption_info(inode, NULL);
623 			spin_lock(&inode->i_lock);
624 			atomic_dec(&inode->i_count);
625 		}
626 		return 0;
627 	}
628 
629 	return generic_drop_inode(inode);
630 }
631 
632 int f2fs_inode_dirtied(struct inode *inode, bool sync)
633 {
634 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
635 	int ret = 0;
636 
637 	spin_lock(&sbi->inode_lock[DIRTY_META]);
638 	if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
639 		ret = 1;
640 	} else {
641 		set_inode_flag(inode, FI_DIRTY_INODE);
642 		stat_inc_dirty_inode(sbi, DIRTY_META);
643 	}
644 	if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
645 		list_add_tail(&F2FS_I(inode)->gdirty_list,
646 				&sbi->inode_list[DIRTY_META]);
647 		inc_page_count(sbi, F2FS_DIRTY_IMETA);
648 	}
649 	spin_unlock(&sbi->inode_lock[DIRTY_META]);
650 	return ret;
651 }
652 
653 void f2fs_inode_synced(struct inode *inode)
654 {
655 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
656 
657 	spin_lock(&sbi->inode_lock[DIRTY_META]);
658 	if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
659 		spin_unlock(&sbi->inode_lock[DIRTY_META]);
660 		return;
661 	}
662 	if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
663 		list_del_init(&F2FS_I(inode)->gdirty_list);
664 		dec_page_count(sbi, F2FS_DIRTY_IMETA);
665 	}
666 	clear_inode_flag(inode, FI_DIRTY_INODE);
667 	clear_inode_flag(inode, FI_AUTO_RECOVER);
668 	stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
669 	spin_unlock(&sbi->inode_lock[DIRTY_META]);
670 }
671 
672 /*
673  * f2fs_dirty_inode() is called from __mark_inode_dirty()
674  *
675  * We should call set_dirty_inode to write the dirty inode through write_inode.
676  */
677 static void f2fs_dirty_inode(struct inode *inode, int flags)
678 {
679 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
680 
681 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
682 			inode->i_ino == F2FS_META_INO(sbi))
683 		return;
684 
685 	if (flags == I_DIRTY_TIME)
686 		return;
687 
688 	if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
689 		clear_inode_flag(inode, FI_AUTO_RECOVER);
690 
691 	f2fs_inode_dirtied(inode, false);
692 }
693 
694 static void f2fs_i_callback(struct rcu_head *head)
695 {
696 	struct inode *inode = container_of(head, struct inode, i_rcu);
697 	kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
698 }
699 
700 static void f2fs_destroy_inode(struct inode *inode)
701 {
702 	call_rcu(&inode->i_rcu, f2fs_i_callback);
703 }
704 
705 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
706 {
707 	percpu_counter_destroy(&sbi->alloc_valid_block_count);
708 	percpu_counter_destroy(&sbi->total_valid_inode_count);
709 }
710 
711 static void destroy_device_list(struct f2fs_sb_info *sbi)
712 {
713 	int i;
714 
715 	for (i = 0; i < sbi->s_ndevs; i++) {
716 		blkdev_put(FDEV(i).bdev, FMODE_EXCL);
717 #ifdef CONFIG_BLK_DEV_ZONED
718 		kfree(FDEV(i).blkz_type);
719 #endif
720 	}
721 	kfree(sbi->devs);
722 }
723 
724 static void f2fs_put_super(struct super_block *sb)
725 {
726 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
727 
728 	if (sbi->s_proc) {
729 		remove_proc_entry("segment_info", sbi->s_proc);
730 		remove_proc_entry("segment_bits", sbi->s_proc);
731 		remove_proc_entry(sb->s_id, f2fs_proc_root);
732 	}
733 	kobject_del(&sbi->s_kobj);
734 
735 	stop_gc_thread(sbi);
736 
737 	/* prevent remaining shrinker jobs */
738 	mutex_lock(&sbi->umount_mutex);
739 
740 	/*
741 	 * We don't need to do checkpoint when superblock is clean.
742 	 * But, the previous checkpoint was not done by umount, it needs to do
743 	 * clean checkpoint again.
744 	 */
745 	if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
746 			!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
747 		struct cp_control cpc = {
748 			.reason = CP_UMOUNT,
749 		};
750 		write_checkpoint(sbi, &cpc);
751 	}
752 
753 	/* write_checkpoint can update stat informaion */
754 	f2fs_destroy_stats(sbi);
755 
756 	/*
757 	 * normally superblock is clean, so we need to release this.
758 	 * In addition, EIO will skip do checkpoint, we need this as well.
759 	 */
760 	release_ino_entry(sbi, true);
761 
762 	f2fs_leave_shrinker(sbi);
763 	mutex_unlock(&sbi->umount_mutex);
764 
765 	/* our cp_error case, we can wait for any writeback page */
766 	f2fs_flush_merged_bios(sbi);
767 
768 	iput(sbi->node_inode);
769 	iput(sbi->meta_inode);
770 
771 	/* destroy f2fs internal modules */
772 	destroy_node_manager(sbi);
773 	destroy_segment_manager(sbi);
774 
775 	kfree(sbi->ckpt);
776 	kobject_put(&sbi->s_kobj);
777 	wait_for_completion(&sbi->s_kobj_unregister);
778 
779 	sb->s_fs_info = NULL;
780 	if (sbi->s_chksum_driver)
781 		crypto_free_shash(sbi->s_chksum_driver);
782 	kfree(sbi->raw_super);
783 
784 	destroy_device_list(sbi);
785 
786 	destroy_percpu_info(sbi);
787 	kfree(sbi);
788 }
789 
790 int f2fs_sync_fs(struct super_block *sb, int sync)
791 {
792 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
793 	int err = 0;
794 
795 	trace_f2fs_sync_fs(sb, sync);
796 
797 	if (sync) {
798 		struct cp_control cpc;
799 
800 		cpc.reason = __get_cp_reason(sbi);
801 
802 		mutex_lock(&sbi->gc_mutex);
803 		err = write_checkpoint(sbi, &cpc);
804 		mutex_unlock(&sbi->gc_mutex);
805 	}
806 	f2fs_trace_ios(NULL, 1);
807 
808 	return err;
809 }
810 
811 static int f2fs_freeze(struct super_block *sb)
812 {
813 	if (f2fs_readonly(sb))
814 		return 0;
815 
816 	/* IO error happened before */
817 	if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
818 		return -EIO;
819 
820 	/* must be clean, since sync_filesystem() was already called */
821 	if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
822 		return -EINVAL;
823 	return 0;
824 }
825 
826 static int f2fs_unfreeze(struct super_block *sb)
827 {
828 	return 0;
829 }
830 
831 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
832 {
833 	struct super_block *sb = dentry->d_sb;
834 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
835 	u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
836 	block_t total_count, user_block_count, start_count, ovp_count;
837 
838 	total_count = le64_to_cpu(sbi->raw_super->block_count);
839 	user_block_count = sbi->user_block_count;
840 	start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
841 	ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
842 	buf->f_type = F2FS_SUPER_MAGIC;
843 	buf->f_bsize = sbi->blocksize;
844 
845 	buf->f_blocks = total_count - start_count;
846 	buf->f_bfree = user_block_count - valid_user_blocks(sbi) + ovp_count;
847 	buf->f_bavail = user_block_count - valid_user_blocks(sbi);
848 
849 	buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
850 	buf->f_ffree = min(buf->f_files - valid_node_count(sbi),
851 							buf->f_bavail);
852 
853 	buf->f_namelen = F2FS_NAME_LEN;
854 	buf->f_fsid.val[0] = (u32)id;
855 	buf->f_fsid.val[1] = (u32)(id >> 32);
856 
857 	return 0;
858 }
859 
860 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
861 {
862 	struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
863 
864 	if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
865 		if (test_opt(sbi, FORCE_FG_GC))
866 			seq_printf(seq, ",background_gc=%s", "sync");
867 		else
868 			seq_printf(seq, ",background_gc=%s", "on");
869 	} else {
870 		seq_printf(seq, ",background_gc=%s", "off");
871 	}
872 	if (test_opt(sbi, DISABLE_ROLL_FORWARD))
873 		seq_puts(seq, ",disable_roll_forward");
874 	if (test_opt(sbi, DISCARD))
875 		seq_puts(seq, ",discard");
876 	if (test_opt(sbi, NOHEAP))
877 		seq_puts(seq, ",no_heap_alloc");
878 #ifdef CONFIG_F2FS_FS_XATTR
879 	if (test_opt(sbi, XATTR_USER))
880 		seq_puts(seq, ",user_xattr");
881 	else
882 		seq_puts(seq, ",nouser_xattr");
883 	if (test_opt(sbi, INLINE_XATTR))
884 		seq_puts(seq, ",inline_xattr");
885 #endif
886 #ifdef CONFIG_F2FS_FS_POSIX_ACL
887 	if (test_opt(sbi, POSIX_ACL))
888 		seq_puts(seq, ",acl");
889 	else
890 		seq_puts(seq, ",noacl");
891 #endif
892 	if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
893 		seq_puts(seq, ",disable_ext_identify");
894 	if (test_opt(sbi, INLINE_DATA))
895 		seq_puts(seq, ",inline_data");
896 	else
897 		seq_puts(seq, ",noinline_data");
898 	if (test_opt(sbi, INLINE_DENTRY))
899 		seq_puts(seq, ",inline_dentry");
900 	else
901 		seq_puts(seq, ",noinline_dentry");
902 	if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
903 		seq_puts(seq, ",flush_merge");
904 	if (test_opt(sbi, NOBARRIER))
905 		seq_puts(seq, ",nobarrier");
906 	if (test_opt(sbi, FASTBOOT))
907 		seq_puts(seq, ",fastboot");
908 	if (test_opt(sbi, EXTENT_CACHE))
909 		seq_puts(seq, ",extent_cache");
910 	else
911 		seq_puts(seq, ",noextent_cache");
912 	if (test_opt(sbi, DATA_FLUSH))
913 		seq_puts(seq, ",data_flush");
914 
915 	seq_puts(seq, ",mode=");
916 	if (test_opt(sbi, ADAPTIVE))
917 		seq_puts(seq, "adaptive");
918 	else if (test_opt(sbi, LFS))
919 		seq_puts(seq, "lfs");
920 	seq_printf(seq, ",active_logs=%u", sbi->active_logs);
921 
922 	return 0;
923 }
924 
925 static int segment_info_seq_show(struct seq_file *seq, void *offset)
926 {
927 	struct super_block *sb = seq->private;
928 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
929 	unsigned int total_segs =
930 			le32_to_cpu(sbi->raw_super->segment_count_main);
931 	int i;
932 
933 	seq_puts(seq, "format: segment_type|valid_blocks\n"
934 		"segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
935 
936 	for (i = 0; i < total_segs; i++) {
937 		struct seg_entry *se = get_seg_entry(sbi, i);
938 
939 		if ((i % 10) == 0)
940 			seq_printf(seq, "%-10d", i);
941 		seq_printf(seq, "%d|%-3u", se->type,
942 					get_valid_blocks(sbi, i, 1));
943 		if ((i % 10) == 9 || i == (total_segs - 1))
944 			seq_putc(seq, '\n');
945 		else
946 			seq_putc(seq, ' ');
947 	}
948 
949 	return 0;
950 }
951 
952 static int segment_bits_seq_show(struct seq_file *seq, void *offset)
953 {
954 	struct super_block *sb = seq->private;
955 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
956 	unsigned int total_segs =
957 			le32_to_cpu(sbi->raw_super->segment_count_main);
958 	int i, j;
959 
960 	seq_puts(seq, "format: segment_type|valid_blocks|bitmaps\n"
961 		"segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
962 
963 	for (i = 0; i < total_segs; i++) {
964 		struct seg_entry *se = get_seg_entry(sbi, i);
965 
966 		seq_printf(seq, "%-10d", i);
967 		seq_printf(seq, "%d|%-3u|", se->type,
968 					get_valid_blocks(sbi, i, 1));
969 		for (j = 0; j < SIT_VBLOCK_MAP_SIZE; j++)
970 			seq_printf(seq, " %.2x", se->cur_valid_map[j]);
971 		seq_putc(seq, '\n');
972 	}
973 	return 0;
974 }
975 
976 #define F2FS_PROC_FILE_DEF(_name)					\
977 static int _name##_open_fs(struct inode *inode, struct file *file)	\
978 {									\
979 	return single_open(file, _name##_seq_show, PDE_DATA(inode));	\
980 }									\
981 									\
982 static const struct file_operations f2fs_seq_##_name##_fops = {		\
983 	.open = _name##_open_fs,					\
984 	.read = seq_read,						\
985 	.llseek = seq_lseek,						\
986 	.release = single_release,					\
987 };
988 
989 F2FS_PROC_FILE_DEF(segment_info);
990 F2FS_PROC_FILE_DEF(segment_bits);
991 
992 static void default_options(struct f2fs_sb_info *sbi)
993 {
994 	/* init some FS parameters */
995 	sbi->active_logs = NR_CURSEG_TYPE;
996 
997 	set_opt(sbi, BG_GC);
998 	set_opt(sbi, INLINE_DATA);
999 	set_opt(sbi, INLINE_DENTRY);
1000 	set_opt(sbi, EXTENT_CACHE);
1001 	sbi->sb->s_flags |= MS_LAZYTIME;
1002 	set_opt(sbi, FLUSH_MERGE);
1003 	if (f2fs_sb_mounted_blkzoned(sbi->sb)) {
1004 		set_opt_mode(sbi, F2FS_MOUNT_LFS);
1005 		set_opt(sbi, DISCARD);
1006 	} else {
1007 		set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1008 	}
1009 
1010 #ifdef CONFIG_F2FS_FS_XATTR
1011 	set_opt(sbi, XATTR_USER);
1012 #endif
1013 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1014 	set_opt(sbi, POSIX_ACL);
1015 #endif
1016 
1017 #ifdef CONFIG_F2FS_FAULT_INJECTION
1018 	f2fs_build_fault_attr(sbi, 0);
1019 #endif
1020 }
1021 
1022 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1023 {
1024 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1025 	struct f2fs_mount_info org_mount_opt;
1026 	int err, active_logs;
1027 	bool need_restart_gc = false;
1028 	bool need_stop_gc = false;
1029 	bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1030 #ifdef CONFIG_F2FS_FAULT_INJECTION
1031 	struct f2fs_fault_info ffi = sbi->fault_info;
1032 #endif
1033 
1034 	/*
1035 	 * Save the old mount options in case we
1036 	 * need to restore them.
1037 	 */
1038 	org_mount_opt = sbi->mount_opt;
1039 	active_logs = sbi->active_logs;
1040 
1041 	/* recover superblocks we couldn't write due to previous RO mount */
1042 	if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1043 		err = f2fs_commit_super(sbi, false);
1044 		f2fs_msg(sb, KERN_INFO,
1045 			"Try to recover all the superblocks, ret: %d", err);
1046 		if (!err)
1047 			clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1048 	}
1049 
1050 	sbi->mount_opt.opt = 0;
1051 	default_options(sbi);
1052 
1053 	/* parse mount options */
1054 	err = parse_options(sb, data);
1055 	if (err)
1056 		goto restore_opts;
1057 
1058 	/*
1059 	 * Previous and new state of filesystem is RO,
1060 	 * so skip checking GC and FLUSH_MERGE conditions.
1061 	 */
1062 	if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
1063 		goto skip;
1064 
1065 	/* disallow enable/disable extent_cache dynamically */
1066 	if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1067 		err = -EINVAL;
1068 		f2fs_msg(sbi->sb, KERN_WARNING,
1069 				"switch extent_cache option is not allowed");
1070 		goto restore_opts;
1071 	}
1072 
1073 	/*
1074 	 * We stop the GC thread if FS is mounted as RO
1075 	 * or if background_gc = off is passed in mount
1076 	 * option. Also sync the filesystem.
1077 	 */
1078 	if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
1079 		if (sbi->gc_thread) {
1080 			stop_gc_thread(sbi);
1081 			need_restart_gc = true;
1082 		}
1083 	} else if (!sbi->gc_thread) {
1084 		err = start_gc_thread(sbi);
1085 		if (err)
1086 			goto restore_opts;
1087 		need_stop_gc = true;
1088 	}
1089 
1090 	if (*flags & MS_RDONLY) {
1091 		writeback_inodes_sb(sb, WB_REASON_SYNC);
1092 		sync_inodes_sb(sb);
1093 
1094 		set_sbi_flag(sbi, SBI_IS_DIRTY);
1095 		set_sbi_flag(sbi, SBI_IS_CLOSE);
1096 		f2fs_sync_fs(sb, 1);
1097 		clear_sbi_flag(sbi, SBI_IS_CLOSE);
1098 	}
1099 
1100 	/*
1101 	 * We stop issue flush thread if FS is mounted as RO
1102 	 * or if flush_merge is not passed in mount option.
1103 	 */
1104 	if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1105 		clear_opt(sbi, FLUSH_MERGE);
1106 		destroy_flush_cmd_control(sbi, false);
1107 	} else {
1108 		err = create_flush_cmd_control(sbi);
1109 		if (err)
1110 			goto restore_gc;
1111 	}
1112 skip:
1113 	/* Update the POSIXACL Flag */
1114 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1115 		(test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1116 
1117 	return 0;
1118 restore_gc:
1119 	if (need_restart_gc) {
1120 		if (start_gc_thread(sbi))
1121 			f2fs_msg(sbi->sb, KERN_WARNING,
1122 				"background gc thread has stopped");
1123 	} else if (need_stop_gc) {
1124 		stop_gc_thread(sbi);
1125 	}
1126 restore_opts:
1127 	sbi->mount_opt = org_mount_opt;
1128 	sbi->active_logs = active_logs;
1129 #ifdef CONFIG_F2FS_FAULT_INJECTION
1130 	sbi->fault_info = ffi;
1131 #endif
1132 	return err;
1133 }
1134 
1135 static struct super_operations f2fs_sops = {
1136 	.alloc_inode	= f2fs_alloc_inode,
1137 	.drop_inode	= f2fs_drop_inode,
1138 	.destroy_inode	= f2fs_destroy_inode,
1139 	.write_inode	= f2fs_write_inode,
1140 	.dirty_inode	= f2fs_dirty_inode,
1141 	.show_options	= f2fs_show_options,
1142 	.evict_inode	= f2fs_evict_inode,
1143 	.put_super	= f2fs_put_super,
1144 	.sync_fs	= f2fs_sync_fs,
1145 	.freeze_fs	= f2fs_freeze,
1146 	.unfreeze_fs	= f2fs_unfreeze,
1147 	.statfs		= f2fs_statfs,
1148 	.remount_fs	= f2fs_remount,
1149 };
1150 
1151 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1152 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1153 {
1154 	return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1155 				F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1156 				ctx, len, NULL);
1157 }
1158 
1159 static int f2fs_key_prefix(struct inode *inode, u8 **key)
1160 {
1161 	*key = F2FS_I_SB(inode)->key_prefix;
1162 	return F2FS_I_SB(inode)->key_prefix_size;
1163 }
1164 
1165 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1166 							void *fs_data)
1167 {
1168 	return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1169 				F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1170 				ctx, len, fs_data, XATTR_CREATE);
1171 }
1172 
1173 static unsigned f2fs_max_namelen(struct inode *inode)
1174 {
1175 	return S_ISLNK(inode->i_mode) ?
1176 			inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1177 }
1178 
1179 static struct fscrypt_operations f2fs_cryptops = {
1180 	.get_context	= f2fs_get_context,
1181 	.key_prefix	= f2fs_key_prefix,
1182 	.set_context	= f2fs_set_context,
1183 	.is_encrypted	= f2fs_encrypted_inode,
1184 	.empty_dir	= f2fs_empty_dir,
1185 	.max_namelen	= f2fs_max_namelen,
1186 };
1187 #else
1188 static struct fscrypt_operations f2fs_cryptops = {
1189 	.is_encrypted	= f2fs_encrypted_inode,
1190 };
1191 #endif
1192 
1193 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1194 		u64 ino, u32 generation)
1195 {
1196 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1197 	struct inode *inode;
1198 
1199 	if (check_nid_range(sbi, ino))
1200 		return ERR_PTR(-ESTALE);
1201 
1202 	/*
1203 	 * f2fs_iget isn't quite right if the inode is currently unallocated!
1204 	 * However f2fs_iget currently does appropriate checks to handle stale
1205 	 * inodes so everything is OK.
1206 	 */
1207 	inode = f2fs_iget(sb, ino);
1208 	if (IS_ERR(inode))
1209 		return ERR_CAST(inode);
1210 	if (unlikely(generation && inode->i_generation != generation)) {
1211 		/* we didn't find the right inode.. */
1212 		iput(inode);
1213 		return ERR_PTR(-ESTALE);
1214 	}
1215 	return inode;
1216 }
1217 
1218 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1219 		int fh_len, int fh_type)
1220 {
1221 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1222 				    f2fs_nfs_get_inode);
1223 }
1224 
1225 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1226 		int fh_len, int fh_type)
1227 {
1228 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1229 				    f2fs_nfs_get_inode);
1230 }
1231 
1232 static const struct export_operations f2fs_export_ops = {
1233 	.fh_to_dentry = f2fs_fh_to_dentry,
1234 	.fh_to_parent = f2fs_fh_to_parent,
1235 	.get_parent = f2fs_get_parent,
1236 };
1237 
1238 static loff_t max_file_blocks(void)
1239 {
1240 	loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
1241 	loff_t leaf_count = ADDRS_PER_BLOCK;
1242 
1243 	/* two direct node blocks */
1244 	result += (leaf_count * 2);
1245 
1246 	/* two indirect node blocks */
1247 	leaf_count *= NIDS_PER_BLOCK;
1248 	result += (leaf_count * 2);
1249 
1250 	/* one double indirect node block */
1251 	leaf_count *= NIDS_PER_BLOCK;
1252 	result += leaf_count;
1253 
1254 	return result;
1255 }
1256 
1257 static int __f2fs_commit_super(struct buffer_head *bh,
1258 			struct f2fs_super_block *super)
1259 {
1260 	lock_buffer(bh);
1261 	if (super)
1262 		memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
1263 	set_buffer_uptodate(bh);
1264 	set_buffer_dirty(bh);
1265 	unlock_buffer(bh);
1266 
1267 	/* it's rare case, we can do fua all the time */
1268 	return __sync_dirty_buffer(bh, REQ_PREFLUSH | REQ_FUA);
1269 }
1270 
1271 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
1272 					struct buffer_head *bh)
1273 {
1274 	struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1275 					(bh->b_data + F2FS_SUPER_OFFSET);
1276 	struct super_block *sb = sbi->sb;
1277 	u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
1278 	u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
1279 	u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
1280 	u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
1281 	u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
1282 	u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
1283 	u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
1284 	u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
1285 	u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
1286 	u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
1287 	u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1288 	u32 segment_count = le32_to_cpu(raw_super->segment_count);
1289 	u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1290 	u64 main_end_blkaddr = main_blkaddr +
1291 				(segment_count_main << log_blocks_per_seg);
1292 	u64 seg_end_blkaddr = segment0_blkaddr +
1293 				(segment_count << log_blocks_per_seg);
1294 
1295 	if (segment0_blkaddr != cp_blkaddr) {
1296 		f2fs_msg(sb, KERN_INFO,
1297 			"Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1298 			segment0_blkaddr, cp_blkaddr);
1299 		return true;
1300 	}
1301 
1302 	if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
1303 							sit_blkaddr) {
1304 		f2fs_msg(sb, KERN_INFO,
1305 			"Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1306 			cp_blkaddr, sit_blkaddr,
1307 			segment_count_ckpt << log_blocks_per_seg);
1308 		return true;
1309 	}
1310 
1311 	if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
1312 							nat_blkaddr) {
1313 		f2fs_msg(sb, KERN_INFO,
1314 			"Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1315 			sit_blkaddr, nat_blkaddr,
1316 			segment_count_sit << log_blocks_per_seg);
1317 		return true;
1318 	}
1319 
1320 	if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
1321 							ssa_blkaddr) {
1322 		f2fs_msg(sb, KERN_INFO,
1323 			"Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1324 			nat_blkaddr, ssa_blkaddr,
1325 			segment_count_nat << log_blocks_per_seg);
1326 		return true;
1327 	}
1328 
1329 	if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
1330 							main_blkaddr) {
1331 		f2fs_msg(sb, KERN_INFO,
1332 			"Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1333 			ssa_blkaddr, main_blkaddr,
1334 			segment_count_ssa << log_blocks_per_seg);
1335 		return true;
1336 	}
1337 
1338 	if (main_end_blkaddr > seg_end_blkaddr) {
1339 		f2fs_msg(sb, KERN_INFO,
1340 			"Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1341 			main_blkaddr,
1342 			segment0_blkaddr +
1343 				(segment_count << log_blocks_per_seg),
1344 			segment_count_main << log_blocks_per_seg);
1345 		return true;
1346 	} else if (main_end_blkaddr < seg_end_blkaddr) {
1347 		int err = 0;
1348 		char *res;
1349 
1350 		/* fix in-memory information all the time */
1351 		raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
1352 				segment0_blkaddr) >> log_blocks_per_seg);
1353 
1354 		if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
1355 			set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1356 			res = "internally";
1357 		} else {
1358 			err = __f2fs_commit_super(bh, NULL);
1359 			res = err ? "failed" : "done";
1360 		}
1361 		f2fs_msg(sb, KERN_INFO,
1362 			"Fix alignment : %s, start(%u) end(%u) block(%u)",
1363 			res, main_blkaddr,
1364 			segment0_blkaddr +
1365 				(segment_count << log_blocks_per_seg),
1366 			segment_count_main << log_blocks_per_seg);
1367 		if (err)
1368 			return true;
1369 	}
1370 	return false;
1371 }
1372 
1373 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
1374 				struct buffer_head *bh)
1375 {
1376 	struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1377 					(bh->b_data + F2FS_SUPER_OFFSET);
1378 	struct super_block *sb = sbi->sb;
1379 	unsigned int blocksize;
1380 
1381 	if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
1382 		f2fs_msg(sb, KERN_INFO,
1383 			"Magic Mismatch, valid(0x%x) - read(0x%x)",
1384 			F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
1385 		return 1;
1386 	}
1387 
1388 	/* Currently, support only 4KB page cache size */
1389 	if (F2FS_BLKSIZE != PAGE_SIZE) {
1390 		f2fs_msg(sb, KERN_INFO,
1391 			"Invalid page_cache_size (%lu), supports only 4KB\n",
1392 			PAGE_SIZE);
1393 		return 1;
1394 	}
1395 
1396 	/* Currently, support only 4KB block size */
1397 	blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
1398 	if (blocksize != F2FS_BLKSIZE) {
1399 		f2fs_msg(sb, KERN_INFO,
1400 			"Invalid blocksize (%u), supports only 4KB\n",
1401 			blocksize);
1402 		return 1;
1403 	}
1404 
1405 	/* check log blocks per segment */
1406 	if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
1407 		f2fs_msg(sb, KERN_INFO,
1408 			"Invalid log blocks per segment (%u)\n",
1409 			le32_to_cpu(raw_super->log_blocks_per_seg));
1410 		return 1;
1411 	}
1412 
1413 	/* Currently, support 512/1024/2048/4096 bytes sector size */
1414 	if (le32_to_cpu(raw_super->log_sectorsize) >
1415 				F2FS_MAX_LOG_SECTOR_SIZE ||
1416 		le32_to_cpu(raw_super->log_sectorsize) <
1417 				F2FS_MIN_LOG_SECTOR_SIZE) {
1418 		f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
1419 			le32_to_cpu(raw_super->log_sectorsize));
1420 		return 1;
1421 	}
1422 	if (le32_to_cpu(raw_super->log_sectors_per_block) +
1423 		le32_to_cpu(raw_super->log_sectorsize) !=
1424 			F2FS_MAX_LOG_SECTOR_SIZE) {
1425 		f2fs_msg(sb, KERN_INFO,
1426 			"Invalid log sectors per block(%u) log sectorsize(%u)",
1427 			le32_to_cpu(raw_super->log_sectors_per_block),
1428 			le32_to_cpu(raw_super->log_sectorsize));
1429 		return 1;
1430 	}
1431 
1432 	/* check reserved ino info */
1433 	if (le32_to_cpu(raw_super->node_ino) != 1 ||
1434 		le32_to_cpu(raw_super->meta_ino) != 2 ||
1435 		le32_to_cpu(raw_super->root_ino) != 3) {
1436 		f2fs_msg(sb, KERN_INFO,
1437 			"Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1438 			le32_to_cpu(raw_super->node_ino),
1439 			le32_to_cpu(raw_super->meta_ino),
1440 			le32_to_cpu(raw_super->root_ino));
1441 		return 1;
1442 	}
1443 
1444 	/* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1445 	if (sanity_check_area_boundary(sbi, bh))
1446 		return 1;
1447 
1448 	return 0;
1449 }
1450 
1451 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1452 {
1453 	unsigned int total, fsmeta;
1454 	struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1455 	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1456 	unsigned int ovp_segments, reserved_segments;
1457 
1458 	total = le32_to_cpu(raw_super->segment_count);
1459 	fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
1460 	fsmeta += le32_to_cpu(raw_super->segment_count_sit);
1461 	fsmeta += le32_to_cpu(raw_super->segment_count_nat);
1462 	fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
1463 	fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
1464 
1465 	if (unlikely(fsmeta >= total))
1466 		return 1;
1467 
1468 	ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
1469 	reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
1470 
1471 	if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
1472 			ovp_segments == 0 || reserved_segments == 0)) {
1473 		f2fs_msg(sbi->sb, KERN_ERR,
1474 			"Wrong layout: check mkfs.f2fs version");
1475 		return 1;
1476 	}
1477 
1478 	if (unlikely(f2fs_cp_error(sbi))) {
1479 		f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
1480 		return 1;
1481 	}
1482 	return 0;
1483 }
1484 
1485 static void init_sb_info(struct f2fs_sb_info *sbi)
1486 {
1487 	struct f2fs_super_block *raw_super = sbi->raw_super;
1488 	int i;
1489 
1490 	sbi->log_sectors_per_block =
1491 		le32_to_cpu(raw_super->log_sectors_per_block);
1492 	sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
1493 	sbi->blocksize = 1 << sbi->log_blocksize;
1494 	sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1495 	sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
1496 	sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1497 	sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1498 	sbi->total_sections = le32_to_cpu(raw_super->section_count);
1499 	sbi->total_node_count =
1500 		(le32_to_cpu(raw_super->segment_count_nat) / 2)
1501 			* sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
1502 	sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
1503 	sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
1504 	sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
1505 	sbi->cur_victim_sec = NULL_SECNO;
1506 	sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
1507 
1508 	sbi->dir_level = DEF_DIR_LEVEL;
1509 	sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
1510 	sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
1511 	clear_sbi_flag(sbi, SBI_NEED_FSCK);
1512 
1513 	for (i = 0; i < NR_COUNT_TYPE; i++)
1514 		atomic_set(&sbi->nr_pages[i], 0);
1515 
1516 	INIT_LIST_HEAD(&sbi->s_list);
1517 	mutex_init(&sbi->umount_mutex);
1518 	mutex_init(&sbi->wio_mutex[NODE]);
1519 	mutex_init(&sbi->wio_mutex[DATA]);
1520 	spin_lock_init(&sbi->cp_lock);
1521 
1522 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1523 	memcpy(sbi->key_prefix, F2FS_KEY_DESC_PREFIX,
1524 				F2FS_KEY_DESC_PREFIX_SIZE);
1525 	sbi->key_prefix_size = F2FS_KEY_DESC_PREFIX_SIZE;
1526 #endif
1527 }
1528 
1529 static int init_percpu_info(struct f2fs_sb_info *sbi)
1530 {
1531 	int err;
1532 
1533 	err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
1534 	if (err)
1535 		return err;
1536 
1537 	return percpu_counter_init(&sbi->total_valid_inode_count, 0,
1538 								GFP_KERNEL);
1539 }
1540 
1541 #ifdef CONFIG_BLK_DEV_ZONED
1542 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
1543 {
1544 	struct block_device *bdev = FDEV(devi).bdev;
1545 	sector_t nr_sectors = bdev->bd_part->nr_sects;
1546 	sector_t sector = 0;
1547 	struct blk_zone *zones;
1548 	unsigned int i, nr_zones;
1549 	unsigned int n = 0;
1550 	int err = -EIO;
1551 
1552 	if (!f2fs_sb_mounted_blkzoned(sbi->sb))
1553 		return 0;
1554 
1555 	if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
1556 				SECTOR_TO_BLOCK(bdev_zone_size(bdev)))
1557 		return -EINVAL;
1558 	sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_size(bdev));
1559 	if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
1560 				__ilog2_u32(sbi->blocks_per_blkz))
1561 		return -EINVAL;
1562 	sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
1563 	FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
1564 					sbi->log_blocks_per_blkz;
1565 	if (nr_sectors & (bdev_zone_size(bdev) - 1))
1566 		FDEV(devi).nr_blkz++;
1567 
1568 	FDEV(devi).blkz_type = kmalloc(FDEV(devi).nr_blkz, GFP_KERNEL);
1569 	if (!FDEV(devi).blkz_type)
1570 		return -ENOMEM;
1571 
1572 #define F2FS_REPORT_NR_ZONES   4096
1573 
1574 	zones = kcalloc(F2FS_REPORT_NR_ZONES, sizeof(struct blk_zone),
1575 			GFP_KERNEL);
1576 	if (!zones)
1577 		return -ENOMEM;
1578 
1579 	/* Get block zones type */
1580 	while (zones && sector < nr_sectors) {
1581 
1582 		nr_zones = F2FS_REPORT_NR_ZONES;
1583 		err = blkdev_report_zones(bdev, sector,
1584 					  zones, &nr_zones,
1585 					  GFP_KERNEL);
1586 		if (err)
1587 			break;
1588 		if (!nr_zones) {
1589 			err = -EIO;
1590 			break;
1591 		}
1592 
1593 		for (i = 0; i < nr_zones; i++) {
1594 			FDEV(devi).blkz_type[n] = zones[i].type;
1595 			sector += zones[i].len;
1596 			n++;
1597 		}
1598 	}
1599 
1600 	kfree(zones);
1601 
1602 	return err;
1603 }
1604 #endif
1605 
1606 /*
1607  * Read f2fs raw super block.
1608  * Because we have two copies of super block, so read both of them
1609  * to get the first valid one. If any one of them is broken, we pass
1610  * them recovery flag back to the caller.
1611  */
1612 static int read_raw_super_block(struct f2fs_sb_info *sbi,
1613 			struct f2fs_super_block **raw_super,
1614 			int *valid_super_block, int *recovery)
1615 {
1616 	struct super_block *sb = sbi->sb;
1617 	int block;
1618 	struct buffer_head *bh;
1619 	struct f2fs_super_block *super;
1620 	int err = 0;
1621 
1622 	super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
1623 	if (!super)
1624 		return -ENOMEM;
1625 
1626 	for (block = 0; block < 2; block++) {
1627 		bh = sb_bread(sb, block);
1628 		if (!bh) {
1629 			f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
1630 				block + 1);
1631 			err = -EIO;
1632 			continue;
1633 		}
1634 
1635 		/* sanity checking of raw super */
1636 		if (sanity_check_raw_super(sbi, bh)) {
1637 			f2fs_msg(sb, KERN_ERR,
1638 				"Can't find valid F2FS filesystem in %dth superblock",
1639 				block + 1);
1640 			err = -EINVAL;
1641 			brelse(bh);
1642 			continue;
1643 		}
1644 
1645 		if (!*raw_super) {
1646 			memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
1647 							sizeof(*super));
1648 			*valid_super_block = block;
1649 			*raw_super = super;
1650 		}
1651 		brelse(bh);
1652 	}
1653 
1654 	/* Fail to read any one of the superblocks*/
1655 	if (err < 0)
1656 		*recovery = 1;
1657 
1658 	/* No valid superblock */
1659 	if (!*raw_super)
1660 		kfree(super);
1661 	else
1662 		err = 0;
1663 
1664 	return err;
1665 }
1666 
1667 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
1668 {
1669 	struct buffer_head *bh;
1670 	int err;
1671 
1672 	if ((recover && f2fs_readonly(sbi->sb)) ||
1673 				bdev_read_only(sbi->sb->s_bdev)) {
1674 		set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1675 		return -EROFS;
1676 	}
1677 
1678 	/* write back-up superblock first */
1679 	bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1);
1680 	if (!bh)
1681 		return -EIO;
1682 	err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1683 	brelse(bh);
1684 
1685 	/* if we are in recovery path, skip writing valid superblock */
1686 	if (recover || err)
1687 		return err;
1688 
1689 	/* write current valid superblock */
1690 	bh = sb_getblk(sbi->sb, sbi->valid_super_block);
1691 	if (!bh)
1692 		return -EIO;
1693 	err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1694 	brelse(bh);
1695 	return err;
1696 }
1697 
1698 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
1699 {
1700 	struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1701 	int i;
1702 
1703 	for (i = 0; i < MAX_DEVICES; i++) {
1704 		if (!RDEV(i).path[0])
1705 			return 0;
1706 
1707 		if (i == 0) {
1708 			sbi->devs = kzalloc(sizeof(struct f2fs_dev_info) *
1709 						MAX_DEVICES, GFP_KERNEL);
1710 			if (!sbi->devs)
1711 				return -ENOMEM;
1712 		}
1713 
1714 		memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
1715 		FDEV(i).total_segments = le32_to_cpu(RDEV(i).total_segments);
1716 		if (i == 0) {
1717 			FDEV(i).start_blk = 0;
1718 			FDEV(i).end_blk = FDEV(i).start_blk +
1719 				(FDEV(i).total_segments <<
1720 				sbi->log_blocks_per_seg) - 1 +
1721 				le32_to_cpu(raw_super->segment0_blkaddr);
1722 		} else {
1723 			FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
1724 			FDEV(i).end_blk = FDEV(i).start_blk +
1725 				(FDEV(i).total_segments <<
1726 				sbi->log_blocks_per_seg) - 1;
1727 		}
1728 
1729 		FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
1730 					sbi->sb->s_mode, sbi->sb->s_type);
1731 		if (IS_ERR(FDEV(i).bdev))
1732 			return PTR_ERR(FDEV(i).bdev);
1733 
1734 		/* to release errored devices */
1735 		sbi->s_ndevs = i + 1;
1736 
1737 #ifdef CONFIG_BLK_DEV_ZONED
1738 		if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
1739 				!f2fs_sb_mounted_blkzoned(sbi->sb)) {
1740 			f2fs_msg(sbi->sb, KERN_ERR,
1741 				"Zoned block device feature not enabled\n");
1742 			return -EINVAL;
1743 		}
1744 		if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
1745 			if (init_blkz_info(sbi, i)) {
1746 				f2fs_msg(sbi->sb, KERN_ERR,
1747 					"Failed to initialize F2FS blkzone information");
1748 				return -EINVAL;
1749 			}
1750 			f2fs_msg(sbi->sb, KERN_INFO,
1751 				"Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
1752 				i, FDEV(i).path,
1753 				FDEV(i).total_segments,
1754 				FDEV(i).start_blk, FDEV(i).end_blk,
1755 				bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
1756 				"Host-aware" : "Host-managed");
1757 			continue;
1758 		}
1759 #endif
1760 		f2fs_msg(sbi->sb, KERN_INFO,
1761 			"Mount Device [%2d]: %20s, %8u, %8x - %8x",
1762 				i, FDEV(i).path,
1763 				FDEV(i).total_segments,
1764 				FDEV(i).start_blk, FDEV(i).end_blk);
1765 	}
1766 	return 0;
1767 }
1768 
1769 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
1770 {
1771 	struct f2fs_sb_info *sbi;
1772 	struct f2fs_super_block *raw_super;
1773 	struct inode *root;
1774 	int err;
1775 	bool retry = true, need_fsck = false;
1776 	char *options = NULL;
1777 	int recovery, i, valid_super_block;
1778 	struct curseg_info *seg_i;
1779 
1780 try_onemore:
1781 	err = -EINVAL;
1782 	raw_super = NULL;
1783 	valid_super_block = -1;
1784 	recovery = 0;
1785 
1786 	/* allocate memory for f2fs-specific super block info */
1787 	sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
1788 	if (!sbi)
1789 		return -ENOMEM;
1790 
1791 	sbi->sb = sb;
1792 
1793 	/* Load the checksum driver */
1794 	sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
1795 	if (IS_ERR(sbi->s_chksum_driver)) {
1796 		f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
1797 		err = PTR_ERR(sbi->s_chksum_driver);
1798 		sbi->s_chksum_driver = NULL;
1799 		goto free_sbi;
1800 	}
1801 
1802 	/* set a block size */
1803 	if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
1804 		f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
1805 		goto free_sbi;
1806 	}
1807 
1808 	err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
1809 								&recovery);
1810 	if (err)
1811 		goto free_sbi;
1812 
1813 	sb->s_fs_info = sbi;
1814 	sbi->raw_super = raw_super;
1815 
1816 	/*
1817 	 * The BLKZONED feature indicates that the drive was formatted with
1818 	 * zone alignment optimization. This is optional for host-aware
1819 	 * devices, but mandatory for host-managed zoned block devices.
1820 	 */
1821 #ifndef CONFIG_BLK_DEV_ZONED
1822 	if (f2fs_sb_mounted_blkzoned(sb)) {
1823 		f2fs_msg(sb, KERN_ERR,
1824 			 "Zoned block device support is not enabled\n");
1825 		goto free_sb_buf;
1826 	}
1827 #endif
1828 	default_options(sbi);
1829 	/* parse mount options */
1830 	options = kstrdup((const char *)data, GFP_KERNEL);
1831 	if (data && !options) {
1832 		err = -ENOMEM;
1833 		goto free_sb_buf;
1834 	}
1835 
1836 	err = parse_options(sb, options);
1837 	if (err)
1838 		goto free_options;
1839 
1840 	sbi->max_file_blocks = max_file_blocks();
1841 	sb->s_maxbytes = sbi->max_file_blocks <<
1842 				le32_to_cpu(raw_super->log_blocksize);
1843 	sb->s_max_links = F2FS_LINK_MAX;
1844 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1845 
1846 	sb->s_op = &f2fs_sops;
1847 	sb->s_cop = &f2fs_cryptops;
1848 	sb->s_xattr = f2fs_xattr_handlers;
1849 	sb->s_export_op = &f2fs_export_ops;
1850 	sb->s_magic = F2FS_SUPER_MAGIC;
1851 	sb->s_time_gran = 1;
1852 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1853 		(test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1854 	memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1855 
1856 	/* init f2fs-specific super block info */
1857 	sbi->valid_super_block = valid_super_block;
1858 	mutex_init(&sbi->gc_mutex);
1859 	mutex_init(&sbi->cp_mutex);
1860 	init_rwsem(&sbi->node_write);
1861 
1862 	/* disallow all the data/node/meta page writes */
1863 	set_sbi_flag(sbi, SBI_POR_DOING);
1864 	spin_lock_init(&sbi->stat_lock);
1865 
1866 	init_rwsem(&sbi->read_io.io_rwsem);
1867 	sbi->read_io.sbi = sbi;
1868 	sbi->read_io.bio = NULL;
1869 	for (i = 0; i < NR_PAGE_TYPE; i++) {
1870 		init_rwsem(&sbi->write_io[i].io_rwsem);
1871 		sbi->write_io[i].sbi = sbi;
1872 		sbi->write_io[i].bio = NULL;
1873 	}
1874 
1875 	init_rwsem(&sbi->cp_rwsem);
1876 	init_waitqueue_head(&sbi->cp_wait);
1877 	init_sb_info(sbi);
1878 
1879 	err = init_percpu_info(sbi);
1880 	if (err)
1881 		goto free_options;
1882 
1883 	/* get an inode for meta space */
1884 	sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1885 	if (IS_ERR(sbi->meta_inode)) {
1886 		f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1887 		err = PTR_ERR(sbi->meta_inode);
1888 		goto free_options;
1889 	}
1890 
1891 	err = get_valid_checkpoint(sbi);
1892 	if (err) {
1893 		f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1894 		goto free_meta_inode;
1895 	}
1896 
1897 	/* Initialize device list */
1898 	err = f2fs_scan_devices(sbi);
1899 	if (err) {
1900 		f2fs_msg(sb, KERN_ERR, "Failed to find devices");
1901 		goto free_devices;
1902 	}
1903 
1904 	sbi->total_valid_node_count =
1905 				le32_to_cpu(sbi->ckpt->valid_node_count);
1906 	percpu_counter_set(&sbi->total_valid_inode_count,
1907 				le32_to_cpu(sbi->ckpt->valid_inode_count));
1908 	sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1909 	sbi->total_valid_block_count =
1910 				le64_to_cpu(sbi->ckpt->valid_block_count);
1911 	sbi->last_valid_block_count = sbi->total_valid_block_count;
1912 
1913 	for (i = 0; i < NR_INODE_TYPE; i++) {
1914 		INIT_LIST_HEAD(&sbi->inode_list[i]);
1915 		spin_lock_init(&sbi->inode_lock[i]);
1916 	}
1917 
1918 	init_extent_cache_info(sbi);
1919 
1920 	init_ino_entry_info(sbi);
1921 
1922 	/* setup f2fs internal modules */
1923 	err = build_segment_manager(sbi);
1924 	if (err) {
1925 		f2fs_msg(sb, KERN_ERR,
1926 			"Failed to initialize F2FS segment manager");
1927 		goto free_sm;
1928 	}
1929 	err = build_node_manager(sbi);
1930 	if (err) {
1931 		f2fs_msg(sb, KERN_ERR,
1932 			"Failed to initialize F2FS node manager");
1933 		goto free_nm;
1934 	}
1935 
1936 	/* For write statistics */
1937 	if (sb->s_bdev->bd_part)
1938 		sbi->sectors_written_start =
1939 			(u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
1940 
1941 	/* Read accumulated write IO statistics if exists */
1942 	seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
1943 	if (__exist_node_summaries(sbi))
1944 		sbi->kbytes_written =
1945 			le64_to_cpu(seg_i->journal->info.kbytes_written);
1946 
1947 	build_gc_manager(sbi);
1948 
1949 	/* get an inode for node space */
1950 	sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1951 	if (IS_ERR(sbi->node_inode)) {
1952 		f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1953 		err = PTR_ERR(sbi->node_inode);
1954 		goto free_nm;
1955 	}
1956 
1957 	f2fs_join_shrinker(sbi);
1958 
1959 	/* if there are nt orphan nodes free them */
1960 	err = recover_orphan_inodes(sbi);
1961 	if (err)
1962 		goto free_node_inode;
1963 
1964 	/* read root inode and dentry */
1965 	root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1966 	if (IS_ERR(root)) {
1967 		f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1968 		err = PTR_ERR(root);
1969 		goto free_node_inode;
1970 	}
1971 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1972 		iput(root);
1973 		err = -EINVAL;
1974 		goto free_node_inode;
1975 	}
1976 
1977 	sb->s_root = d_make_root(root); /* allocate root dentry */
1978 	if (!sb->s_root) {
1979 		err = -ENOMEM;
1980 		goto free_root_inode;
1981 	}
1982 
1983 	err = f2fs_build_stats(sbi);
1984 	if (err)
1985 		goto free_root_inode;
1986 
1987 	if (f2fs_proc_root)
1988 		sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1989 
1990 	if (sbi->s_proc) {
1991 		proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1992 				 &f2fs_seq_segment_info_fops, sb);
1993 		proc_create_data("segment_bits", S_IRUGO, sbi->s_proc,
1994 				 &f2fs_seq_segment_bits_fops, sb);
1995 	}
1996 
1997 	sbi->s_kobj.kset = f2fs_kset;
1998 	init_completion(&sbi->s_kobj_unregister);
1999 	err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
2000 							"%s", sb->s_id);
2001 	if (err)
2002 		goto free_proc;
2003 
2004 	/* recover fsynced data */
2005 	if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
2006 		/*
2007 		 * mount should be failed, when device has readonly mode, and
2008 		 * previous checkpoint was not done by clean system shutdown.
2009 		 */
2010 		if (bdev_read_only(sb->s_bdev) &&
2011 				!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
2012 			err = -EROFS;
2013 			goto free_kobj;
2014 		}
2015 
2016 		if (need_fsck)
2017 			set_sbi_flag(sbi, SBI_NEED_FSCK);
2018 
2019 		if (!retry)
2020 			goto skip_recovery;
2021 
2022 		err = recover_fsync_data(sbi, false);
2023 		if (err < 0) {
2024 			need_fsck = true;
2025 			f2fs_msg(sb, KERN_ERR,
2026 				"Cannot recover all fsync data errno=%d", err);
2027 			goto free_kobj;
2028 		}
2029 	} else {
2030 		err = recover_fsync_data(sbi, true);
2031 
2032 		if (!f2fs_readonly(sb) && err > 0) {
2033 			err = -EINVAL;
2034 			f2fs_msg(sb, KERN_ERR,
2035 				"Need to recover fsync data");
2036 			goto free_kobj;
2037 		}
2038 	}
2039 skip_recovery:
2040 	/* recover_fsync_data() cleared this already */
2041 	clear_sbi_flag(sbi, SBI_POR_DOING);
2042 
2043 	/*
2044 	 * If filesystem is not mounted as read-only then
2045 	 * do start the gc_thread.
2046 	 */
2047 	if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
2048 		/* After POR, we can run background GC thread.*/
2049 		err = start_gc_thread(sbi);
2050 		if (err)
2051 			goto free_kobj;
2052 	}
2053 	kfree(options);
2054 
2055 	/* recover broken superblock */
2056 	if (recovery) {
2057 		err = f2fs_commit_super(sbi, true);
2058 		f2fs_msg(sb, KERN_INFO,
2059 			"Try to recover %dth superblock, ret: %d",
2060 			sbi->valid_super_block ? 1 : 2, err);
2061 	}
2062 
2063 	f2fs_update_time(sbi, CP_TIME);
2064 	f2fs_update_time(sbi, REQ_TIME);
2065 	return 0;
2066 
2067 free_kobj:
2068 	f2fs_sync_inode_meta(sbi);
2069 	kobject_del(&sbi->s_kobj);
2070 	kobject_put(&sbi->s_kobj);
2071 	wait_for_completion(&sbi->s_kobj_unregister);
2072 free_proc:
2073 	if (sbi->s_proc) {
2074 		remove_proc_entry("segment_info", sbi->s_proc);
2075 		remove_proc_entry("segment_bits", sbi->s_proc);
2076 		remove_proc_entry(sb->s_id, f2fs_proc_root);
2077 	}
2078 	f2fs_destroy_stats(sbi);
2079 free_root_inode:
2080 	dput(sb->s_root);
2081 	sb->s_root = NULL;
2082 free_node_inode:
2083 	truncate_inode_pages_final(NODE_MAPPING(sbi));
2084 	mutex_lock(&sbi->umount_mutex);
2085 	release_ino_entry(sbi, true);
2086 	f2fs_leave_shrinker(sbi);
2087 	/*
2088 	 * Some dirty meta pages can be produced by recover_orphan_inodes()
2089 	 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
2090 	 * followed by write_checkpoint() through f2fs_write_node_pages(), which
2091 	 * falls into an infinite loop in sync_meta_pages().
2092 	 */
2093 	truncate_inode_pages_final(META_MAPPING(sbi));
2094 	iput(sbi->node_inode);
2095 	mutex_unlock(&sbi->umount_mutex);
2096 free_nm:
2097 	destroy_node_manager(sbi);
2098 free_sm:
2099 	destroy_segment_manager(sbi);
2100 free_devices:
2101 	destroy_device_list(sbi);
2102 	kfree(sbi->ckpt);
2103 free_meta_inode:
2104 	make_bad_inode(sbi->meta_inode);
2105 	iput(sbi->meta_inode);
2106 free_options:
2107 	destroy_percpu_info(sbi);
2108 	kfree(options);
2109 free_sb_buf:
2110 	kfree(raw_super);
2111 free_sbi:
2112 	if (sbi->s_chksum_driver)
2113 		crypto_free_shash(sbi->s_chksum_driver);
2114 	kfree(sbi);
2115 
2116 	/* give only one another chance */
2117 	if (retry) {
2118 		retry = false;
2119 		shrink_dcache_sb(sb);
2120 		goto try_onemore;
2121 	}
2122 	return err;
2123 }
2124 
2125 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
2126 			const char *dev_name, void *data)
2127 {
2128 	return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
2129 }
2130 
2131 static void kill_f2fs_super(struct super_block *sb)
2132 {
2133 	if (sb->s_root)
2134 		set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
2135 	kill_block_super(sb);
2136 }
2137 
2138 static struct file_system_type f2fs_fs_type = {
2139 	.owner		= THIS_MODULE,
2140 	.name		= "f2fs",
2141 	.mount		= f2fs_mount,
2142 	.kill_sb	= kill_f2fs_super,
2143 	.fs_flags	= FS_REQUIRES_DEV,
2144 };
2145 MODULE_ALIAS_FS("f2fs");
2146 
2147 static int __init init_inodecache(void)
2148 {
2149 	f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
2150 			sizeof(struct f2fs_inode_info), 0,
2151 			SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
2152 	if (!f2fs_inode_cachep)
2153 		return -ENOMEM;
2154 	return 0;
2155 }
2156 
2157 static void destroy_inodecache(void)
2158 {
2159 	/*
2160 	 * Make sure all delayed rcu free inodes are flushed before we
2161 	 * destroy cache.
2162 	 */
2163 	rcu_barrier();
2164 	kmem_cache_destroy(f2fs_inode_cachep);
2165 }
2166 
2167 static int __init init_f2fs_fs(void)
2168 {
2169 	int err;
2170 
2171 	f2fs_build_trace_ios();
2172 
2173 	err = init_inodecache();
2174 	if (err)
2175 		goto fail;
2176 	err = create_node_manager_caches();
2177 	if (err)
2178 		goto free_inodecache;
2179 	err = create_segment_manager_caches();
2180 	if (err)
2181 		goto free_node_manager_caches;
2182 	err = create_checkpoint_caches();
2183 	if (err)
2184 		goto free_segment_manager_caches;
2185 	err = create_extent_cache();
2186 	if (err)
2187 		goto free_checkpoint_caches;
2188 	f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
2189 	if (!f2fs_kset) {
2190 		err = -ENOMEM;
2191 		goto free_extent_cache;
2192 	}
2193 	err = register_shrinker(&f2fs_shrinker_info);
2194 	if (err)
2195 		goto free_kset;
2196 
2197 	err = register_filesystem(&f2fs_fs_type);
2198 	if (err)
2199 		goto free_shrinker;
2200 	err = f2fs_create_root_stats();
2201 	if (err)
2202 		goto free_filesystem;
2203 	f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
2204 	return 0;
2205 
2206 free_filesystem:
2207 	unregister_filesystem(&f2fs_fs_type);
2208 free_shrinker:
2209 	unregister_shrinker(&f2fs_shrinker_info);
2210 free_kset:
2211 	kset_unregister(f2fs_kset);
2212 free_extent_cache:
2213 	destroy_extent_cache();
2214 free_checkpoint_caches:
2215 	destroy_checkpoint_caches();
2216 free_segment_manager_caches:
2217 	destroy_segment_manager_caches();
2218 free_node_manager_caches:
2219 	destroy_node_manager_caches();
2220 free_inodecache:
2221 	destroy_inodecache();
2222 fail:
2223 	return err;
2224 }
2225 
2226 static void __exit exit_f2fs_fs(void)
2227 {
2228 	remove_proc_entry("fs/f2fs", NULL);
2229 	f2fs_destroy_root_stats();
2230 	unregister_filesystem(&f2fs_fs_type);
2231 	unregister_shrinker(&f2fs_shrinker_info);
2232 	kset_unregister(f2fs_kset);
2233 	destroy_extent_cache();
2234 	destroy_checkpoint_caches();
2235 	destroy_segment_manager_caches();
2236 	destroy_node_manager_caches();
2237 	destroy_inodecache();
2238 	f2fs_destroy_trace_ios();
2239 }
2240 
2241 module_init(init_f2fs_fs)
2242 module_exit(exit_f2fs_fs)
2243 
2244 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
2245 MODULE_DESCRIPTION("Flash Friendly File System");
2246 MODULE_LICENSE("GPL");
2247 
2248