xref: /openbmc/linux/fs/f2fs/super.c (revision 5c57132eaf5265937e46340bfbfb97ffb078c423)
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/quotaops.h>
26 #include <linux/f2fs_fs.h>
27 #include <linux/sysfs.h>
28 
29 #include "f2fs.h"
30 #include "node.h"
31 #include "segment.h"
32 #include "xattr.h"
33 #include "gc.h"
34 #include "trace.h"
35 
36 #define CREATE_TRACE_POINTS
37 #include <trace/events/f2fs.h>
38 
39 static struct kmem_cache *f2fs_inode_cachep;
40 
41 #ifdef CONFIG_F2FS_FAULT_INJECTION
42 
43 char *fault_name[FAULT_MAX] = {
44 	[FAULT_KMALLOC]		= "kmalloc",
45 	[FAULT_PAGE_ALLOC]	= "page alloc",
46 	[FAULT_ALLOC_NID]	= "alloc nid",
47 	[FAULT_ORPHAN]		= "orphan",
48 	[FAULT_BLOCK]		= "no more block",
49 	[FAULT_DIR_DEPTH]	= "too big dir depth",
50 	[FAULT_EVICT_INODE]	= "evict_inode fail",
51 	[FAULT_TRUNCATE]	= "truncate 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_heap,
86 	Opt_user_xattr,
87 	Opt_nouser_xattr,
88 	Opt_acl,
89 	Opt_noacl,
90 	Opt_active_logs,
91 	Opt_disable_ext_identify,
92 	Opt_inline_xattr,
93 	Opt_noinline_xattr,
94 	Opt_inline_data,
95 	Opt_inline_dentry,
96 	Opt_noinline_dentry,
97 	Opt_flush_merge,
98 	Opt_noflush_merge,
99 	Opt_nobarrier,
100 	Opt_fastboot,
101 	Opt_extent_cache,
102 	Opt_noextent_cache,
103 	Opt_noinline_data,
104 	Opt_data_flush,
105 	Opt_mode,
106 	Opt_io_size_bits,
107 	Opt_fault_injection,
108 	Opt_lazytime,
109 	Opt_nolazytime,
110 	Opt_usrquota,
111 	Opt_grpquota,
112 	Opt_prjquota,
113 	Opt_err,
114 };
115 
116 static match_table_t f2fs_tokens = {
117 	{Opt_gc_background, "background_gc=%s"},
118 	{Opt_disable_roll_forward, "disable_roll_forward"},
119 	{Opt_norecovery, "norecovery"},
120 	{Opt_discard, "discard"},
121 	{Opt_nodiscard, "nodiscard"},
122 	{Opt_noheap, "no_heap"},
123 	{Opt_heap, "heap"},
124 	{Opt_user_xattr, "user_xattr"},
125 	{Opt_nouser_xattr, "nouser_xattr"},
126 	{Opt_acl, "acl"},
127 	{Opt_noacl, "noacl"},
128 	{Opt_active_logs, "active_logs=%u"},
129 	{Opt_disable_ext_identify, "disable_ext_identify"},
130 	{Opt_inline_xattr, "inline_xattr"},
131 	{Opt_noinline_xattr, "noinline_xattr"},
132 	{Opt_inline_data, "inline_data"},
133 	{Opt_inline_dentry, "inline_dentry"},
134 	{Opt_noinline_dentry, "noinline_dentry"},
135 	{Opt_flush_merge, "flush_merge"},
136 	{Opt_noflush_merge, "noflush_merge"},
137 	{Opt_nobarrier, "nobarrier"},
138 	{Opt_fastboot, "fastboot"},
139 	{Opt_extent_cache, "extent_cache"},
140 	{Opt_noextent_cache, "noextent_cache"},
141 	{Opt_noinline_data, "noinline_data"},
142 	{Opt_data_flush, "data_flush"},
143 	{Opt_mode, "mode=%s"},
144 	{Opt_io_size_bits, "io_bits=%u"},
145 	{Opt_fault_injection, "fault_injection=%u"},
146 	{Opt_lazytime, "lazytime"},
147 	{Opt_nolazytime, "nolazytime"},
148 	{Opt_usrquota, "usrquota"},
149 	{Opt_grpquota, "grpquota"},
150 	{Opt_prjquota, "prjquota"},
151 	{Opt_err, NULL},
152 };
153 
154 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
155 {
156 	struct va_format vaf;
157 	va_list args;
158 
159 	va_start(args, fmt);
160 	vaf.fmt = fmt;
161 	vaf.va = &args;
162 	printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
163 	va_end(args);
164 }
165 
166 static void init_once(void *foo)
167 {
168 	struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
169 
170 	inode_init_once(&fi->vfs_inode);
171 }
172 
173 static int parse_options(struct super_block *sb, char *options)
174 {
175 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
176 	struct request_queue *q;
177 	substring_t args[MAX_OPT_ARGS];
178 	char *p, *name;
179 	int arg = 0;
180 
181 	if (!options)
182 		return 0;
183 
184 	while ((p = strsep(&options, ",")) != NULL) {
185 		int token;
186 		if (!*p)
187 			continue;
188 		/*
189 		 * Initialize args struct so we know whether arg was
190 		 * found; some options take optional arguments.
191 		 */
192 		args[0].to = args[0].from = NULL;
193 		token = match_token(p, f2fs_tokens, args);
194 
195 		switch (token) {
196 		case Opt_gc_background:
197 			name = match_strdup(&args[0]);
198 
199 			if (!name)
200 				return -ENOMEM;
201 			if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
202 				set_opt(sbi, BG_GC);
203 				clear_opt(sbi, FORCE_FG_GC);
204 			} else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
205 				clear_opt(sbi, BG_GC);
206 				clear_opt(sbi, FORCE_FG_GC);
207 			} else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
208 				set_opt(sbi, BG_GC);
209 				set_opt(sbi, FORCE_FG_GC);
210 			} else {
211 				kfree(name);
212 				return -EINVAL;
213 			}
214 			kfree(name);
215 			break;
216 		case Opt_disable_roll_forward:
217 			set_opt(sbi, DISABLE_ROLL_FORWARD);
218 			break;
219 		case Opt_norecovery:
220 			/* this option mounts f2fs with ro */
221 			set_opt(sbi, DISABLE_ROLL_FORWARD);
222 			if (!f2fs_readonly(sb))
223 				return -EINVAL;
224 			break;
225 		case Opt_discard:
226 			q = bdev_get_queue(sb->s_bdev);
227 			if (blk_queue_discard(q)) {
228 				set_opt(sbi, DISCARD);
229 			} else if (!f2fs_sb_mounted_blkzoned(sb)) {
230 				f2fs_msg(sb, KERN_WARNING,
231 					"mounting with \"discard\" option, but "
232 					"the device does not support discard");
233 			}
234 			break;
235 		case Opt_nodiscard:
236 			if (f2fs_sb_mounted_blkzoned(sb)) {
237 				f2fs_msg(sb, KERN_WARNING,
238 					"discard is required for zoned block devices");
239 				return -EINVAL;
240 			}
241 			clear_opt(sbi, DISCARD);
242 			break;
243 		case Opt_noheap:
244 			set_opt(sbi, NOHEAP);
245 			break;
246 		case Opt_heap:
247 			clear_opt(sbi, NOHEAP);
248 			break;
249 #ifdef CONFIG_F2FS_FS_XATTR
250 		case Opt_user_xattr:
251 			set_opt(sbi, XATTR_USER);
252 			break;
253 		case Opt_nouser_xattr:
254 			clear_opt(sbi, XATTR_USER);
255 			break;
256 		case Opt_inline_xattr:
257 			set_opt(sbi, INLINE_XATTR);
258 			break;
259 		case Opt_noinline_xattr:
260 			clear_opt(sbi, INLINE_XATTR);
261 			break;
262 #else
263 		case Opt_user_xattr:
264 			f2fs_msg(sb, KERN_INFO,
265 				"user_xattr options not supported");
266 			break;
267 		case Opt_nouser_xattr:
268 			f2fs_msg(sb, KERN_INFO,
269 				"nouser_xattr options not supported");
270 			break;
271 		case Opt_inline_xattr:
272 			f2fs_msg(sb, KERN_INFO,
273 				"inline_xattr options not supported");
274 			break;
275 		case Opt_noinline_xattr:
276 			f2fs_msg(sb, KERN_INFO,
277 				"noinline_xattr options not supported");
278 			break;
279 #endif
280 #ifdef CONFIG_F2FS_FS_POSIX_ACL
281 		case Opt_acl:
282 			set_opt(sbi, POSIX_ACL);
283 			break;
284 		case Opt_noacl:
285 			clear_opt(sbi, POSIX_ACL);
286 			break;
287 #else
288 		case Opt_acl:
289 			f2fs_msg(sb, KERN_INFO, "acl options not supported");
290 			break;
291 		case Opt_noacl:
292 			f2fs_msg(sb, KERN_INFO, "noacl options not supported");
293 			break;
294 #endif
295 		case Opt_active_logs:
296 			if (args->from && match_int(args, &arg))
297 				return -EINVAL;
298 			if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
299 				return -EINVAL;
300 			sbi->active_logs = arg;
301 			break;
302 		case Opt_disable_ext_identify:
303 			set_opt(sbi, DISABLE_EXT_IDENTIFY);
304 			break;
305 		case Opt_inline_data:
306 			set_opt(sbi, INLINE_DATA);
307 			break;
308 		case Opt_inline_dentry:
309 			set_opt(sbi, INLINE_DENTRY);
310 			break;
311 		case Opt_noinline_dentry:
312 			clear_opt(sbi, INLINE_DENTRY);
313 			break;
314 		case Opt_flush_merge:
315 			set_opt(sbi, FLUSH_MERGE);
316 			break;
317 		case Opt_noflush_merge:
318 			clear_opt(sbi, FLUSH_MERGE);
319 			break;
320 		case Opt_nobarrier:
321 			set_opt(sbi, NOBARRIER);
322 			break;
323 		case Opt_fastboot:
324 			set_opt(sbi, FASTBOOT);
325 			break;
326 		case Opt_extent_cache:
327 			set_opt(sbi, EXTENT_CACHE);
328 			break;
329 		case Opt_noextent_cache:
330 			clear_opt(sbi, EXTENT_CACHE);
331 			break;
332 		case Opt_noinline_data:
333 			clear_opt(sbi, INLINE_DATA);
334 			break;
335 		case Opt_data_flush:
336 			set_opt(sbi, DATA_FLUSH);
337 			break;
338 		case Opt_mode:
339 			name = match_strdup(&args[0]);
340 
341 			if (!name)
342 				return -ENOMEM;
343 			if (strlen(name) == 8 &&
344 					!strncmp(name, "adaptive", 8)) {
345 				if (f2fs_sb_mounted_blkzoned(sb)) {
346 					f2fs_msg(sb, KERN_WARNING,
347 						 "adaptive mode is not allowed with "
348 						 "zoned block device feature");
349 					kfree(name);
350 					return -EINVAL;
351 				}
352 				set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
353 			} else if (strlen(name) == 3 &&
354 					!strncmp(name, "lfs", 3)) {
355 				set_opt_mode(sbi, F2FS_MOUNT_LFS);
356 			} else {
357 				kfree(name);
358 				return -EINVAL;
359 			}
360 			kfree(name);
361 			break;
362 		case Opt_io_size_bits:
363 			if (args->from && match_int(args, &arg))
364 				return -EINVAL;
365 			if (arg > __ilog2_u32(BIO_MAX_PAGES)) {
366 				f2fs_msg(sb, KERN_WARNING,
367 					"Not support %d, larger than %d",
368 					1 << arg, BIO_MAX_PAGES);
369 				return -EINVAL;
370 			}
371 			sbi->write_io_size_bits = arg;
372 			break;
373 		case Opt_fault_injection:
374 			if (args->from && match_int(args, &arg))
375 				return -EINVAL;
376 #ifdef CONFIG_F2FS_FAULT_INJECTION
377 			f2fs_build_fault_attr(sbi, arg);
378 			set_opt(sbi, FAULT_INJECTION);
379 #else
380 			f2fs_msg(sb, KERN_INFO,
381 				"FAULT_INJECTION was not selected");
382 #endif
383 			break;
384 		case Opt_lazytime:
385 			sb->s_flags |= MS_LAZYTIME;
386 			break;
387 		case Opt_nolazytime:
388 			sb->s_flags &= ~MS_LAZYTIME;
389 			break;
390 #ifdef CONFIG_QUOTA
391 		case Opt_usrquota:
392 			set_opt(sbi, USRQUOTA);
393 			break;
394 		case Opt_grpquota:
395 			set_opt(sbi, GRPQUOTA);
396 			break;
397 		case Opt_prjquota:
398 			set_opt(sbi, PRJQUOTA);
399 			break;
400 #else
401 		case Opt_usrquota:
402 		case Opt_grpquota:
403 		case Opt_prjquota:
404 			f2fs_msg(sb, KERN_INFO,
405 					"quota operations not supported");
406 			break;
407 #endif
408 		default:
409 			f2fs_msg(sb, KERN_ERR,
410 				"Unrecognized mount option \"%s\" or missing value",
411 				p);
412 			return -EINVAL;
413 		}
414 	}
415 
416 	if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) {
417 		f2fs_msg(sb, KERN_ERR,
418 				"Should set mode=lfs with %uKB-sized IO",
419 				F2FS_IO_SIZE_KB(sbi));
420 		return -EINVAL;
421 	}
422 	return 0;
423 }
424 
425 static struct inode *f2fs_alloc_inode(struct super_block *sb)
426 {
427 	struct f2fs_inode_info *fi;
428 
429 	fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
430 	if (!fi)
431 		return NULL;
432 
433 	init_once((void *) fi);
434 
435 	/* Initialize f2fs-specific inode info */
436 	fi->vfs_inode.i_version = 1;
437 	atomic_set(&fi->dirty_pages, 0);
438 	fi->i_current_depth = 1;
439 	fi->i_advise = 0;
440 	init_rwsem(&fi->i_sem);
441 	INIT_LIST_HEAD(&fi->dirty_list);
442 	INIT_LIST_HEAD(&fi->gdirty_list);
443 	INIT_LIST_HEAD(&fi->inmem_pages);
444 	mutex_init(&fi->inmem_lock);
445 	init_rwsem(&fi->dio_rwsem[READ]);
446 	init_rwsem(&fi->dio_rwsem[WRITE]);
447 	init_rwsem(&fi->i_mmap_sem);
448 
449 #ifdef CONFIG_QUOTA
450 	memset(&fi->i_dquot, 0, sizeof(fi->i_dquot));
451 	fi->i_reserved_quota = 0;
452 #endif
453 	/* Will be used by directory only */
454 	fi->i_dir_level = F2FS_SB(sb)->dir_level;
455 
456 	return &fi->vfs_inode;
457 }
458 
459 static int f2fs_drop_inode(struct inode *inode)
460 {
461 	int ret;
462 	/*
463 	 * This is to avoid a deadlock condition like below.
464 	 * writeback_single_inode(inode)
465 	 *  - f2fs_write_data_page
466 	 *    - f2fs_gc -> iput -> evict
467 	 *       - inode_wait_for_writeback(inode)
468 	 */
469 	if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
470 		if (!inode->i_nlink && !is_bad_inode(inode)) {
471 			/* to avoid evict_inode call simultaneously */
472 			atomic_inc(&inode->i_count);
473 			spin_unlock(&inode->i_lock);
474 
475 			/* some remained atomic pages should discarded */
476 			if (f2fs_is_atomic_file(inode))
477 				drop_inmem_pages(inode);
478 
479 			/* should remain fi->extent_tree for writepage */
480 			f2fs_destroy_extent_node(inode);
481 
482 			sb_start_intwrite(inode->i_sb);
483 			f2fs_i_size_write(inode, 0);
484 
485 			if (F2FS_HAS_BLOCKS(inode))
486 				f2fs_truncate(inode);
487 
488 			sb_end_intwrite(inode->i_sb);
489 
490 			fscrypt_put_encryption_info(inode, NULL);
491 			spin_lock(&inode->i_lock);
492 			atomic_dec(&inode->i_count);
493 		}
494 		trace_f2fs_drop_inode(inode, 0);
495 		return 0;
496 	}
497 	ret = generic_drop_inode(inode);
498 	trace_f2fs_drop_inode(inode, ret);
499 	return ret;
500 }
501 
502 int f2fs_inode_dirtied(struct inode *inode, bool sync)
503 {
504 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
505 	int ret = 0;
506 
507 	spin_lock(&sbi->inode_lock[DIRTY_META]);
508 	if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
509 		ret = 1;
510 	} else {
511 		set_inode_flag(inode, FI_DIRTY_INODE);
512 		stat_inc_dirty_inode(sbi, DIRTY_META);
513 	}
514 	if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
515 		list_add_tail(&F2FS_I(inode)->gdirty_list,
516 				&sbi->inode_list[DIRTY_META]);
517 		inc_page_count(sbi, F2FS_DIRTY_IMETA);
518 	}
519 	spin_unlock(&sbi->inode_lock[DIRTY_META]);
520 	return ret;
521 }
522 
523 void f2fs_inode_synced(struct inode *inode)
524 {
525 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
526 
527 	spin_lock(&sbi->inode_lock[DIRTY_META]);
528 	if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
529 		spin_unlock(&sbi->inode_lock[DIRTY_META]);
530 		return;
531 	}
532 	if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
533 		list_del_init(&F2FS_I(inode)->gdirty_list);
534 		dec_page_count(sbi, F2FS_DIRTY_IMETA);
535 	}
536 	clear_inode_flag(inode, FI_DIRTY_INODE);
537 	clear_inode_flag(inode, FI_AUTO_RECOVER);
538 	stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
539 	spin_unlock(&sbi->inode_lock[DIRTY_META]);
540 }
541 
542 /*
543  * f2fs_dirty_inode() is called from __mark_inode_dirty()
544  *
545  * We should call set_dirty_inode to write the dirty inode through write_inode.
546  */
547 static void f2fs_dirty_inode(struct inode *inode, int flags)
548 {
549 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
550 
551 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
552 			inode->i_ino == F2FS_META_INO(sbi))
553 		return;
554 
555 	if (flags == I_DIRTY_TIME)
556 		return;
557 
558 	if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
559 		clear_inode_flag(inode, FI_AUTO_RECOVER);
560 
561 	f2fs_inode_dirtied(inode, false);
562 }
563 
564 static void f2fs_i_callback(struct rcu_head *head)
565 {
566 	struct inode *inode = container_of(head, struct inode, i_rcu);
567 	kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
568 }
569 
570 static void f2fs_destroy_inode(struct inode *inode)
571 {
572 	call_rcu(&inode->i_rcu, f2fs_i_callback);
573 }
574 
575 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
576 {
577 	percpu_counter_destroy(&sbi->alloc_valid_block_count);
578 	percpu_counter_destroy(&sbi->total_valid_inode_count);
579 }
580 
581 static void destroy_device_list(struct f2fs_sb_info *sbi)
582 {
583 	int i;
584 
585 	for (i = 0; i < sbi->s_ndevs; i++) {
586 		blkdev_put(FDEV(i).bdev, FMODE_EXCL);
587 #ifdef CONFIG_BLK_DEV_ZONED
588 		kfree(FDEV(i).blkz_type);
589 #endif
590 	}
591 	kfree(sbi->devs);
592 }
593 
594 static void f2fs_quota_off_umount(struct super_block *sb);
595 static void f2fs_put_super(struct super_block *sb)
596 {
597 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
598 	int i;
599 
600 	f2fs_quota_off_umount(sb);
601 
602 	/* prevent remaining shrinker jobs */
603 	mutex_lock(&sbi->umount_mutex);
604 
605 	/*
606 	 * We don't need to do checkpoint when superblock is clean.
607 	 * But, the previous checkpoint was not done by umount, it needs to do
608 	 * clean checkpoint again.
609 	 */
610 	if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
611 			!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
612 		struct cp_control cpc = {
613 			.reason = CP_UMOUNT,
614 		};
615 		write_checkpoint(sbi, &cpc);
616 	}
617 
618 	/* be sure to wait for any on-going discard commands */
619 	f2fs_wait_discard_bios(sbi);
620 
621 	if (f2fs_discard_en(sbi) && !sbi->discard_blks) {
622 		struct cp_control cpc = {
623 			.reason = CP_UMOUNT | CP_TRIMMED,
624 		};
625 		write_checkpoint(sbi, &cpc);
626 	}
627 
628 	/* write_checkpoint can update stat informaion */
629 	f2fs_destroy_stats(sbi);
630 
631 	/*
632 	 * normally superblock is clean, so we need to release this.
633 	 * In addition, EIO will skip do checkpoint, we need this as well.
634 	 */
635 	release_ino_entry(sbi, true);
636 
637 	f2fs_leave_shrinker(sbi);
638 	mutex_unlock(&sbi->umount_mutex);
639 
640 	/* our cp_error case, we can wait for any writeback page */
641 	f2fs_flush_merged_writes(sbi);
642 
643 	iput(sbi->node_inode);
644 	iput(sbi->meta_inode);
645 
646 	/* destroy f2fs internal modules */
647 	destroy_node_manager(sbi);
648 	destroy_segment_manager(sbi);
649 
650 	kfree(sbi->ckpt);
651 
652 	f2fs_exit_sysfs(sbi);
653 
654 	sb->s_fs_info = NULL;
655 	if (sbi->s_chksum_driver)
656 		crypto_free_shash(sbi->s_chksum_driver);
657 	kfree(sbi->raw_super);
658 
659 	destroy_device_list(sbi);
660 	mempool_destroy(sbi->write_io_dummy);
661 	destroy_percpu_info(sbi);
662 	for (i = 0; i < NR_PAGE_TYPE; i++)
663 		kfree(sbi->write_io[i]);
664 	kfree(sbi);
665 }
666 
667 int f2fs_sync_fs(struct super_block *sb, int sync)
668 {
669 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
670 	int err = 0;
671 
672 	trace_f2fs_sync_fs(sb, sync);
673 
674 	if (sync) {
675 		struct cp_control cpc;
676 
677 		cpc.reason = __get_cp_reason(sbi);
678 
679 		mutex_lock(&sbi->gc_mutex);
680 		err = write_checkpoint(sbi, &cpc);
681 		mutex_unlock(&sbi->gc_mutex);
682 	}
683 	f2fs_trace_ios(NULL, 1);
684 
685 	return err;
686 }
687 
688 static int f2fs_freeze(struct super_block *sb)
689 {
690 	if (f2fs_readonly(sb))
691 		return 0;
692 
693 	/* IO error happened before */
694 	if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
695 		return -EIO;
696 
697 	/* must be clean, since sync_filesystem() was already called */
698 	if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
699 		return -EINVAL;
700 	return 0;
701 }
702 
703 static int f2fs_unfreeze(struct super_block *sb)
704 {
705 	return 0;
706 }
707 
708 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
709 {
710 	struct super_block *sb = dentry->d_sb;
711 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
712 	u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
713 	block_t total_count, user_block_count, start_count, ovp_count;
714 	u64 avail_node_count;
715 
716 	total_count = le64_to_cpu(sbi->raw_super->block_count);
717 	user_block_count = sbi->user_block_count;
718 	start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
719 	ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
720 	buf->f_type = F2FS_SUPER_MAGIC;
721 	buf->f_bsize = sbi->blocksize;
722 
723 	buf->f_blocks = total_count - start_count;
724 	buf->f_bfree = user_block_count - valid_user_blocks(sbi) + ovp_count;
725 	buf->f_bavail = user_block_count - valid_user_blocks(sbi) -
726 						sbi->reserved_blocks;
727 
728 	avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
729 
730 	if (avail_node_count > user_block_count) {
731 		buf->f_files = user_block_count;
732 		buf->f_ffree = buf->f_bavail;
733 	} else {
734 		buf->f_files = avail_node_count;
735 		buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
736 					buf->f_bavail);
737 	}
738 
739 	buf->f_namelen = F2FS_NAME_LEN;
740 	buf->f_fsid.val[0] = (u32)id;
741 	buf->f_fsid.val[1] = (u32)(id >> 32);
742 
743 	return 0;
744 }
745 
746 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
747 {
748 	struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
749 
750 	if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
751 		if (test_opt(sbi, FORCE_FG_GC))
752 			seq_printf(seq, ",background_gc=%s", "sync");
753 		else
754 			seq_printf(seq, ",background_gc=%s", "on");
755 	} else {
756 		seq_printf(seq, ",background_gc=%s", "off");
757 	}
758 	if (test_opt(sbi, DISABLE_ROLL_FORWARD))
759 		seq_puts(seq, ",disable_roll_forward");
760 	if (test_opt(sbi, DISCARD))
761 		seq_puts(seq, ",discard");
762 	if (test_opt(sbi, NOHEAP))
763 		seq_puts(seq, ",no_heap");
764 	else
765 		seq_puts(seq, ",heap");
766 #ifdef CONFIG_F2FS_FS_XATTR
767 	if (test_opt(sbi, XATTR_USER))
768 		seq_puts(seq, ",user_xattr");
769 	else
770 		seq_puts(seq, ",nouser_xattr");
771 	if (test_opt(sbi, INLINE_XATTR))
772 		seq_puts(seq, ",inline_xattr");
773 	else
774 		seq_puts(seq, ",noinline_xattr");
775 #endif
776 #ifdef CONFIG_F2FS_FS_POSIX_ACL
777 	if (test_opt(sbi, POSIX_ACL))
778 		seq_puts(seq, ",acl");
779 	else
780 		seq_puts(seq, ",noacl");
781 #endif
782 	if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
783 		seq_puts(seq, ",disable_ext_identify");
784 	if (test_opt(sbi, INLINE_DATA))
785 		seq_puts(seq, ",inline_data");
786 	else
787 		seq_puts(seq, ",noinline_data");
788 	if (test_opt(sbi, INLINE_DENTRY))
789 		seq_puts(seq, ",inline_dentry");
790 	else
791 		seq_puts(seq, ",noinline_dentry");
792 	if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
793 		seq_puts(seq, ",flush_merge");
794 	if (test_opt(sbi, NOBARRIER))
795 		seq_puts(seq, ",nobarrier");
796 	if (test_opt(sbi, FASTBOOT))
797 		seq_puts(seq, ",fastboot");
798 	if (test_opt(sbi, EXTENT_CACHE))
799 		seq_puts(seq, ",extent_cache");
800 	else
801 		seq_puts(seq, ",noextent_cache");
802 	if (test_opt(sbi, DATA_FLUSH))
803 		seq_puts(seq, ",data_flush");
804 
805 	seq_puts(seq, ",mode=");
806 	if (test_opt(sbi, ADAPTIVE))
807 		seq_puts(seq, "adaptive");
808 	else if (test_opt(sbi, LFS))
809 		seq_puts(seq, "lfs");
810 	seq_printf(seq, ",active_logs=%u", sbi->active_logs);
811 	if (F2FS_IO_SIZE_BITS(sbi))
812 		seq_printf(seq, ",io_size=%uKB", F2FS_IO_SIZE_KB(sbi));
813 #ifdef CONFIG_F2FS_FAULT_INJECTION
814 	if (test_opt(sbi, FAULT_INJECTION))
815 		seq_printf(seq, ",fault_injection=%u",
816 				sbi->fault_info.inject_rate);
817 #endif
818 #ifdef CONFIG_QUOTA
819 	if (test_opt(sbi, USRQUOTA))
820 		seq_puts(seq, ",usrquota");
821 	if (test_opt(sbi, GRPQUOTA))
822 		seq_puts(seq, ",grpquota");
823 	if (test_opt(sbi, PRJQUOTA))
824 		seq_puts(seq, ",prjquota");
825 #endif
826 
827 	return 0;
828 }
829 
830 static void default_options(struct f2fs_sb_info *sbi)
831 {
832 	/* init some FS parameters */
833 	sbi->active_logs = NR_CURSEG_TYPE;
834 
835 	set_opt(sbi, BG_GC);
836 	set_opt(sbi, INLINE_XATTR);
837 	set_opt(sbi, INLINE_DATA);
838 	set_opt(sbi, INLINE_DENTRY);
839 	set_opt(sbi, EXTENT_CACHE);
840 	set_opt(sbi, NOHEAP);
841 	sbi->sb->s_flags |= MS_LAZYTIME;
842 	set_opt(sbi, FLUSH_MERGE);
843 	if (f2fs_sb_mounted_blkzoned(sbi->sb)) {
844 		set_opt_mode(sbi, F2FS_MOUNT_LFS);
845 		set_opt(sbi, DISCARD);
846 	} else {
847 		set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
848 	}
849 
850 #ifdef CONFIG_F2FS_FS_XATTR
851 	set_opt(sbi, XATTR_USER);
852 #endif
853 #ifdef CONFIG_F2FS_FS_POSIX_ACL
854 	set_opt(sbi, POSIX_ACL);
855 #endif
856 
857 #ifdef CONFIG_F2FS_FAULT_INJECTION
858 	f2fs_build_fault_attr(sbi, 0);
859 #endif
860 }
861 
862 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
863 {
864 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
865 	struct f2fs_mount_info org_mount_opt;
866 	unsigned long old_sb_flags;
867 	int err, active_logs;
868 	bool need_restart_gc = false;
869 	bool need_stop_gc = false;
870 	bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
871 #ifdef CONFIG_F2FS_FAULT_INJECTION
872 	struct f2fs_fault_info ffi = sbi->fault_info;
873 #endif
874 
875 	/*
876 	 * Save the old mount options in case we
877 	 * need to restore them.
878 	 */
879 	org_mount_opt = sbi->mount_opt;
880 	old_sb_flags = sb->s_flags;
881 	active_logs = sbi->active_logs;
882 
883 	/* recover superblocks we couldn't write due to previous RO mount */
884 	if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
885 		err = f2fs_commit_super(sbi, false);
886 		f2fs_msg(sb, KERN_INFO,
887 			"Try to recover all the superblocks, ret: %d", err);
888 		if (!err)
889 			clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
890 	}
891 
892 	default_options(sbi);
893 
894 	/* parse mount options */
895 	err = parse_options(sb, data);
896 	if (err)
897 		goto restore_opts;
898 
899 	/*
900 	 * Previous and new state of filesystem is RO,
901 	 * so skip checking GC and FLUSH_MERGE conditions.
902 	 */
903 	if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
904 		goto skip;
905 
906 	if (!f2fs_readonly(sb) && (*flags & MS_RDONLY)) {
907 		err = dquot_suspend(sb, -1);
908 		if (err < 0)
909 			goto restore_opts;
910 	} else {
911 		/* dquot_resume needs RW */
912 		sb->s_flags &= ~MS_RDONLY;
913 		dquot_resume(sb, -1);
914 	}
915 
916 	/* disallow enable/disable extent_cache dynamically */
917 	if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
918 		err = -EINVAL;
919 		f2fs_msg(sbi->sb, KERN_WARNING,
920 				"switch extent_cache option is not allowed");
921 		goto restore_opts;
922 	}
923 
924 	/*
925 	 * We stop the GC thread if FS is mounted as RO
926 	 * or if background_gc = off is passed in mount
927 	 * option. Also sync the filesystem.
928 	 */
929 	if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
930 		if (sbi->gc_thread) {
931 			stop_gc_thread(sbi);
932 			need_restart_gc = true;
933 		}
934 	} else if (!sbi->gc_thread) {
935 		err = start_gc_thread(sbi);
936 		if (err)
937 			goto restore_opts;
938 		need_stop_gc = true;
939 	}
940 
941 	if (*flags & MS_RDONLY) {
942 		writeback_inodes_sb(sb, WB_REASON_SYNC);
943 		sync_inodes_sb(sb);
944 
945 		set_sbi_flag(sbi, SBI_IS_DIRTY);
946 		set_sbi_flag(sbi, SBI_IS_CLOSE);
947 		f2fs_sync_fs(sb, 1);
948 		clear_sbi_flag(sbi, SBI_IS_CLOSE);
949 	}
950 
951 	/*
952 	 * We stop issue flush thread if FS is mounted as RO
953 	 * or if flush_merge is not passed in mount option.
954 	 */
955 	if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
956 		clear_opt(sbi, FLUSH_MERGE);
957 		destroy_flush_cmd_control(sbi, false);
958 	} else {
959 		err = create_flush_cmd_control(sbi);
960 		if (err)
961 			goto restore_gc;
962 	}
963 skip:
964 	/* Update the POSIXACL Flag */
965 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
966 		(test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
967 
968 	return 0;
969 restore_gc:
970 	if (need_restart_gc) {
971 		if (start_gc_thread(sbi))
972 			f2fs_msg(sbi->sb, KERN_WARNING,
973 				"background gc thread has stopped");
974 	} else if (need_stop_gc) {
975 		stop_gc_thread(sbi);
976 	}
977 restore_opts:
978 	sbi->mount_opt = org_mount_opt;
979 	sbi->active_logs = active_logs;
980 	sb->s_flags = old_sb_flags;
981 #ifdef CONFIG_F2FS_FAULT_INJECTION
982 	sbi->fault_info = ffi;
983 #endif
984 	return err;
985 }
986 
987 #ifdef CONFIG_QUOTA
988 /* Read data from quotafile */
989 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
990 			       size_t len, loff_t off)
991 {
992 	struct inode *inode = sb_dqopt(sb)->files[type];
993 	struct address_space *mapping = inode->i_mapping;
994 	block_t blkidx = F2FS_BYTES_TO_BLK(off);
995 	int offset = off & (sb->s_blocksize - 1);
996 	int tocopy;
997 	size_t toread;
998 	loff_t i_size = i_size_read(inode);
999 	struct page *page;
1000 	char *kaddr;
1001 
1002 	if (off > i_size)
1003 		return 0;
1004 
1005 	if (off + len > i_size)
1006 		len = i_size - off;
1007 	toread = len;
1008 	while (toread > 0) {
1009 		tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1010 repeat:
1011 		page = read_mapping_page(mapping, blkidx, NULL);
1012 		if (IS_ERR(page))
1013 			return PTR_ERR(page);
1014 
1015 		lock_page(page);
1016 
1017 		if (unlikely(page->mapping != mapping)) {
1018 			f2fs_put_page(page, 1);
1019 			goto repeat;
1020 		}
1021 		if (unlikely(!PageUptodate(page))) {
1022 			f2fs_put_page(page, 1);
1023 			return -EIO;
1024 		}
1025 
1026 		kaddr = kmap_atomic(page);
1027 		memcpy(data, kaddr + offset, tocopy);
1028 		kunmap_atomic(kaddr);
1029 		f2fs_put_page(page, 1);
1030 
1031 		offset = 0;
1032 		toread -= tocopy;
1033 		data += tocopy;
1034 		blkidx++;
1035 	}
1036 	return len;
1037 }
1038 
1039 /* Write to quotafile */
1040 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1041 				const char *data, size_t len, loff_t off)
1042 {
1043 	struct inode *inode = sb_dqopt(sb)->files[type];
1044 	struct address_space *mapping = inode->i_mapping;
1045 	const struct address_space_operations *a_ops = mapping->a_ops;
1046 	int offset = off & (sb->s_blocksize - 1);
1047 	size_t towrite = len;
1048 	struct page *page;
1049 	char *kaddr;
1050 	int err = 0;
1051 	int tocopy;
1052 
1053 	while (towrite > 0) {
1054 		tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1055 								towrite);
1056 
1057 		err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1058 							&page, NULL);
1059 		if (unlikely(err))
1060 			break;
1061 
1062 		kaddr = kmap_atomic(page);
1063 		memcpy(kaddr + offset, data, tocopy);
1064 		kunmap_atomic(kaddr);
1065 		flush_dcache_page(page);
1066 
1067 		a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1068 						page, NULL);
1069 		offset = 0;
1070 		towrite -= tocopy;
1071 		off += tocopy;
1072 		data += tocopy;
1073 		cond_resched();
1074 	}
1075 
1076 	if (len == towrite)
1077 		return err;
1078 	inode->i_version++;
1079 	inode->i_mtime = inode->i_ctime = current_time(inode);
1080 	f2fs_mark_inode_dirty_sync(inode, false);
1081 	return len - towrite;
1082 }
1083 
1084 static struct dquot **f2fs_get_dquots(struct inode *inode)
1085 {
1086 	return F2FS_I(inode)->i_dquot;
1087 }
1088 
1089 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1090 {
1091 	return &F2FS_I(inode)->i_reserved_quota;
1092 }
1093 
1094 static int f2fs_quota_sync(struct super_block *sb, int type)
1095 {
1096 	struct quota_info *dqopt = sb_dqopt(sb);
1097 	int cnt;
1098 	int ret;
1099 
1100 	ret = dquot_writeback_dquots(sb, type);
1101 	if (ret)
1102 		return ret;
1103 
1104 	/*
1105 	 * Now when everything is written we can discard the pagecache so
1106 	 * that userspace sees the changes.
1107 	 */
1108 	for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
1109 		if (type != -1 && cnt != type)
1110 			continue;
1111 		if (!sb_has_quota_active(sb, cnt))
1112 			continue;
1113 
1114 		ret = filemap_write_and_wait(dqopt->files[cnt]->i_mapping);
1115 		if (ret)
1116 			return ret;
1117 
1118 		inode_lock(dqopt->files[cnt]);
1119 		truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
1120 		inode_unlock(dqopt->files[cnt]);
1121 	}
1122 	return 0;
1123 }
1124 
1125 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
1126 							const struct path *path)
1127 {
1128 	struct inode *inode;
1129 	int err;
1130 
1131 	err = f2fs_quota_sync(sb, -1);
1132 	if (err)
1133 		return err;
1134 
1135 	err = dquot_quota_on(sb, type, format_id, path);
1136 	if (err)
1137 		return err;
1138 
1139 	inode = d_inode(path->dentry);
1140 
1141 	inode_lock(inode);
1142 	F2FS_I(inode)->i_flags |= FS_NOATIME_FL | FS_IMMUTABLE_FL;
1143 	inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1144 					S_NOATIME | S_IMMUTABLE);
1145 	inode_unlock(inode);
1146 	f2fs_mark_inode_dirty_sync(inode, false);
1147 
1148 	return 0;
1149 }
1150 
1151 static int f2fs_quota_off(struct super_block *sb, int type)
1152 {
1153 	struct inode *inode = sb_dqopt(sb)->files[type];
1154 	int err;
1155 
1156 	if (!inode || !igrab(inode))
1157 		return dquot_quota_off(sb, type);
1158 
1159 	f2fs_quota_sync(sb, -1);
1160 
1161 	err = dquot_quota_off(sb, type);
1162 	if (err)
1163 		goto out_put;
1164 
1165 	inode_lock(inode);
1166 	F2FS_I(inode)->i_flags &= ~(FS_NOATIME_FL | FS_IMMUTABLE_FL);
1167 	inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1168 	inode_unlock(inode);
1169 	f2fs_mark_inode_dirty_sync(inode, false);
1170 out_put:
1171 	iput(inode);
1172 	return err;
1173 }
1174 
1175 static void f2fs_quota_off_umount(struct super_block *sb)
1176 {
1177 	int type;
1178 
1179 	for (type = 0; type < MAXQUOTAS; type++)
1180 		f2fs_quota_off(sb, type);
1181 }
1182 
1183 int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
1184 {
1185 	*projid = F2FS_I(inode)->i_projid;
1186 	return 0;
1187 }
1188 
1189 static const struct dquot_operations f2fs_quota_operations = {
1190 	.get_reserved_space = f2fs_get_reserved_space,
1191 	.write_dquot	= dquot_commit,
1192 	.acquire_dquot	= dquot_acquire,
1193 	.release_dquot	= dquot_release,
1194 	.mark_dirty	= dquot_mark_dquot_dirty,
1195 	.write_info	= dquot_commit_info,
1196 	.alloc_dquot	= dquot_alloc,
1197 	.destroy_dquot	= dquot_destroy,
1198 	.get_projid	= f2fs_get_projid,
1199 	.get_next_id	= dquot_get_next_id,
1200 };
1201 
1202 static const struct quotactl_ops f2fs_quotactl_ops = {
1203 	.quota_on	= f2fs_quota_on,
1204 	.quota_off	= f2fs_quota_off,
1205 	.quota_sync	= f2fs_quota_sync,
1206 	.get_state	= dquot_get_state,
1207 	.set_info	= dquot_set_dqinfo,
1208 	.get_dqblk	= dquot_get_dqblk,
1209 	.set_dqblk	= dquot_set_dqblk,
1210 	.get_nextdqblk	= dquot_get_next_dqblk,
1211 };
1212 #else
1213 static inline void f2fs_quota_off_umount(struct super_block *sb)
1214 {
1215 }
1216 #endif
1217 
1218 static struct super_operations f2fs_sops = {
1219 	.alloc_inode	= f2fs_alloc_inode,
1220 	.drop_inode	= f2fs_drop_inode,
1221 	.destroy_inode	= f2fs_destroy_inode,
1222 	.write_inode	= f2fs_write_inode,
1223 	.dirty_inode	= f2fs_dirty_inode,
1224 	.show_options	= f2fs_show_options,
1225 #ifdef CONFIG_QUOTA
1226 	.quota_read	= f2fs_quota_read,
1227 	.quota_write	= f2fs_quota_write,
1228 	.get_dquots	= f2fs_get_dquots,
1229 #endif
1230 	.evict_inode	= f2fs_evict_inode,
1231 	.put_super	= f2fs_put_super,
1232 	.sync_fs	= f2fs_sync_fs,
1233 	.freeze_fs	= f2fs_freeze,
1234 	.unfreeze_fs	= f2fs_unfreeze,
1235 	.statfs		= f2fs_statfs,
1236 	.remount_fs	= f2fs_remount,
1237 };
1238 
1239 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1240 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1241 {
1242 	return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1243 				F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1244 				ctx, len, NULL);
1245 }
1246 
1247 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1248 							void *fs_data)
1249 {
1250 	return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1251 				F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1252 				ctx, len, fs_data, XATTR_CREATE);
1253 }
1254 
1255 static unsigned f2fs_max_namelen(struct inode *inode)
1256 {
1257 	return S_ISLNK(inode->i_mode) ?
1258 			inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1259 }
1260 
1261 static const struct fscrypt_operations f2fs_cryptops = {
1262 	.key_prefix	= "f2fs:",
1263 	.get_context	= f2fs_get_context,
1264 	.set_context	= f2fs_set_context,
1265 	.is_encrypted	= f2fs_encrypted_inode,
1266 	.empty_dir	= f2fs_empty_dir,
1267 	.max_namelen	= f2fs_max_namelen,
1268 };
1269 #else
1270 static const struct fscrypt_operations f2fs_cryptops = {
1271 	.is_encrypted	= f2fs_encrypted_inode,
1272 };
1273 #endif
1274 
1275 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1276 		u64 ino, u32 generation)
1277 {
1278 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1279 	struct inode *inode;
1280 
1281 	if (check_nid_range(sbi, ino))
1282 		return ERR_PTR(-ESTALE);
1283 
1284 	/*
1285 	 * f2fs_iget isn't quite right if the inode is currently unallocated!
1286 	 * However f2fs_iget currently does appropriate checks to handle stale
1287 	 * inodes so everything is OK.
1288 	 */
1289 	inode = f2fs_iget(sb, ino);
1290 	if (IS_ERR(inode))
1291 		return ERR_CAST(inode);
1292 	if (unlikely(generation && inode->i_generation != generation)) {
1293 		/* we didn't find the right inode.. */
1294 		iput(inode);
1295 		return ERR_PTR(-ESTALE);
1296 	}
1297 	return inode;
1298 }
1299 
1300 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1301 		int fh_len, int fh_type)
1302 {
1303 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1304 				    f2fs_nfs_get_inode);
1305 }
1306 
1307 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1308 		int fh_len, int fh_type)
1309 {
1310 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1311 				    f2fs_nfs_get_inode);
1312 }
1313 
1314 static const struct export_operations f2fs_export_ops = {
1315 	.fh_to_dentry = f2fs_fh_to_dentry,
1316 	.fh_to_parent = f2fs_fh_to_parent,
1317 	.get_parent = f2fs_get_parent,
1318 };
1319 
1320 static loff_t max_file_blocks(void)
1321 {
1322 	loff_t result = 0;
1323 	loff_t leaf_count = ADDRS_PER_BLOCK;
1324 
1325 	/*
1326 	 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
1327 	 * F2FS_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
1328 	 * space in inode.i_addr, it will be more safe to reassign
1329 	 * result as zero.
1330 	 */
1331 
1332 	/* two direct node blocks */
1333 	result += (leaf_count * 2);
1334 
1335 	/* two indirect node blocks */
1336 	leaf_count *= NIDS_PER_BLOCK;
1337 	result += (leaf_count * 2);
1338 
1339 	/* one double indirect node block */
1340 	leaf_count *= NIDS_PER_BLOCK;
1341 	result += leaf_count;
1342 
1343 	return result;
1344 }
1345 
1346 static int __f2fs_commit_super(struct buffer_head *bh,
1347 			struct f2fs_super_block *super)
1348 {
1349 	lock_buffer(bh);
1350 	if (super)
1351 		memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
1352 	set_buffer_uptodate(bh);
1353 	set_buffer_dirty(bh);
1354 	unlock_buffer(bh);
1355 
1356 	/* it's rare case, we can do fua all the time */
1357 	return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
1358 }
1359 
1360 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
1361 					struct buffer_head *bh)
1362 {
1363 	struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1364 					(bh->b_data + F2FS_SUPER_OFFSET);
1365 	struct super_block *sb = sbi->sb;
1366 	u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
1367 	u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
1368 	u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
1369 	u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
1370 	u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
1371 	u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
1372 	u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
1373 	u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
1374 	u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
1375 	u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
1376 	u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1377 	u32 segment_count = le32_to_cpu(raw_super->segment_count);
1378 	u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1379 	u64 main_end_blkaddr = main_blkaddr +
1380 				(segment_count_main << log_blocks_per_seg);
1381 	u64 seg_end_blkaddr = segment0_blkaddr +
1382 				(segment_count << log_blocks_per_seg);
1383 
1384 	if (segment0_blkaddr != cp_blkaddr) {
1385 		f2fs_msg(sb, KERN_INFO,
1386 			"Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1387 			segment0_blkaddr, cp_blkaddr);
1388 		return true;
1389 	}
1390 
1391 	if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
1392 							sit_blkaddr) {
1393 		f2fs_msg(sb, KERN_INFO,
1394 			"Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1395 			cp_blkaddr, sit_blkaddr,
1396 			segment_count_ckpt << log_blocks_per_seg);
1397 		return true;
1398 	}
1399 
1400 	if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
1401 							nat_blkaddr) {
1402 		f2fs_msg(sb, KERN_INFO,
1403 			"Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1404 			sit_blkaddr, nat_blkaddr,
1405 			segment_count_sit << log_blocks_per_seg);
1406 		return true;
1407 	}
1408 
1409 	if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
1410 							ssa_blkaddr) {
1411 		f2fs_msg(sb, KERN_INFO,
1412 			"Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1413 			nat_blkaddr, ssa_blkaddr,
1414 			segment_count_nat << log_blocks_per_seg);
1415 		return true;
1416 	}
1417 
1418 	if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
1419 							main_blkaddr) {
1420 		f2fs_msg(sb, KERN_INFO,
1421 			"Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1422 			ssa_blkaddr, main_blkaddr,
1423 			segment_count_ssa << log_blocks_per_seg);
1424 		return true;
1425 	}
1426 
1427 	if (main_end_blkaddr > seg_end_blkaddr) {
1428 		f2fs_msg(sb, KERN_INFO,
1429 			"Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1430 			main_blkaddr,
1431 			segment0_blkaddr +
1432 				(segment_count << log_blocks_per_seg),
1433 			segment_count_main << log_blocks_per_seg);
1434 		return true;
1435 	} else if (main_end_blkaddr < seg_end_blkaddr) {
1436 		int err = 0;
1437 		char *res;
1438 
1439 		/* fix in-memory information all the time */
1440 		raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
1441 				segment0_blkaddr) >> log_blocks_per_seg);
1442 
1443 		if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
1444 			set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1445 			res = "internally";
1446 		} else {
1447 			err = __f2fs_commit_super(bh, NULL);
1448 			res = err ? "failed" : "done";
1449 		}
1450 		f2fs_msg(sb, KERN_INFO,
1451 			"Fix alignment : %s, start(%u) end(%u) block(%u)",
1452 			res, main_blkaddr,
1453 			segment0_blkaddr +
1454 				(segment_count << log_blocks_per_seg),
1455 			segment_count_main << log_blocks_per_seg);
1456 		if (err)
1457 			return true;
1458 	}
1459 	return false;
1460 }
1461 
1462 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
1463 				struct buffer_head *bh)
1464 {
1465 	struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1466 					(bh->b_data + F2FS_SUPER_OFFSET);
1467 	struct super_block *sb = sbi->sb;
1468 	unsigned int blocksize;
1469 
1470 	if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
1471 		f2fs_msg(sb, KERN_INFO,
1472 			"Magic Mismatch, valid(0x%x) - read(0x%x)",
1473 			F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
1474 		return 1;
1475 	}
1476 
1477 	/* Currently, support only 4KB page cache size */
1478 	if (F2FS_BLKSIZE != PAGE_SIZE) {
1479 		f2fs_msg(sb, KERN_INFO,
1480 			"Invalid page_cache_size (%lu), supports only 4KB\n",
1481 			PAGE_SIZE);
1482 		return 1;
1483 	}
1484 
1485 	/* Currently, support only 4KB block size */
1486 	blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
1487 	if (blocksize != F2FS_BLKSIZE) {
1488 		f2fs_msg(sb, KERN_INFO,
1489 			"Invalid blocksize (%u), supports only 4KB\n",
1490 			blocksize);
1491 		return 1;
1492 	}
1493 
1494 	/* check log blocks per segment */
1495 	if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
1496 		f2fs_msg(sb, KERN_INFO,
1497 			"Invalid log blocks per segment (%u)\n",
1498 			le32_to_cpu(raw_super->log_blocks_per_seg));
1499 		return 1;
1500 	}
1501 
1502 	/* Currently, support 512/1024/2048/4096 bytes sector size */
1503 	if (le32_to_cpu(raw_super->log_sectorsize) >
1504 				F2FS_MAX_LOG_SECTOR_SIZE ||
1505 		le32_to_cpu(raw_super->log_sectorsize) <
1506 				F2FS_MIN_LOG_SECTOR_SIZE) {
1507 		f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
1508 			le32_to_cpu(raw_super->log_sectorsize));
1509 		return 1;
1510 	}
1511 	if (le32_to_cpu(raw_super->log_sectors_per_block) +
1512 		le32_to_cpu(raw_super->log_sectorsize) !=
1513 			F2FS_MAX_LOG_SECTOR_SIZE) {
1514 		f2fs_msg(sb, KERN_INFO,
1515 			"Invalid log sectors per block(%u) log sectorsize(%u)",
1516 			le32_to_cpu(raw_super->log_sectors_per_block),
1517 			le32_to_cpu(raw_super->log_sectorsize));
1518 		return 1;
1519 	}
1520 
1521 	/* check reserved ino info */
1522 	if (le32_to_cpu(raw_super->node_ino) != 1 ||
1523 		le32_to_cpu(raw_super->meta_ino) != 2 ||
1524 		le32_to_cpu(raw_super->root_ino) != 3) {
1525 		f2fs_msg(sb, KERN_INFO,
1526 			"Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1527 			le32_to_cpu(raw_super->node_ino),
1528 			le32_to_cpu(raw_super->meta_ino),
1529 			le32_to_cpu(raw_super->root_ino));
1530 		return 1;
1531 	}
1532 
1533 	if (le32_to_cpu(raw_super->segment_count) > F2FS_MAX_SEGMENT) {
1534 		f2fs_msg(sb, KERN_INFO,
1535 			"Invalid segment count (%u)",
1536 			le32_to_cpu(raw_super->segment_count));
1537 		return 1;
1538 	}
1539 
1540 	/* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1541 	if (sanity_check_area_boundary(sbi, bh))
1542 		return 1;
1543 
1544 	return 0;
1545 }
1546 
1547 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1548 {
1549 	unsigned int total, fsmeta;
1550 	struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1551 	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1552 	unsigned int ovp_segments, reserved_segments;
1553 	unsigned int main_segs, blocks_per_seg;
1554 	int i;
1555 
1556 	total = le32_to_cpu(raw_super->segment_count);
1557 	fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
1558 	fsmeta += le32_to_cpu(raw_super->segment_count_sit);
1559 	fsmeta += le32_to_cpu(raw_super->segment_count_nat);
1560 	fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
1561 	fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
1562 
1563 	if (unlikely(fsmeta >= total))
1564 		return 1;
1565 
1566 	ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
1567 	reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
1568 
1569 	if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
1570 			ovp_segments == 0 || reserved_segments == 0)) {
1571 		f2fs_msg(sbi->sb, KERN_ERR,
1572 			"Wrong layout: check mkfs.f2fs version");
1573 		return 1;
1574 	}
1575 
1576 	main_segs = le32_to_cpu(raw_super->segment_count_main);
1577 	blocks_per_seg = sbi->blocks_per_seg;
1578 
1579 	for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
1580 		if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
1581 			le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
1582 			return 1;
1583 	}
1584 	for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
1585 		if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
1586 			le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
1587 			return 1;
1588 	}
1589 
1590 	if (unlikely(f2fs_cp_error(sbi))) {
1591 		f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
1592 		return 1;
1593 	}
1594 	return 0;
1595 }
1596 
1597 static void init_sb_info(struct f2fs_sb_info *sbi)
1598 {
1599 	struct f2fs_super_block *raw_super = sbi->raw_super;
1600 	int i, j;
1601 
1602 	sbi->log_sectors_per_block =
1603 		le32_to_cpu(raw_super->log_sectors_per_block);
1604 	sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
1605 	sbi->blocksize = 1 << sbi->log_blocksize;
1606 	sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1607 	sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
1608 	sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1609 	sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1610 	sbi->total_sections = le32_to_cpu(raw_super->section_count);
1611 	sbi->total_node_count =
1612 		(le32_to_cpu(raw_super->segment_count_nat) / 2)
1613 			* sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
1614 	sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
1615 	sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
1616 	sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
1617 	sbi->cur_victim_sec = NULL_SECNO;
1618 	sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
1619 
1620 	sbi->dir_level = DEF_DIR_LEVEL;
1621 	sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
1622 	sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
1623 	clear_sbi_flag(sbi, SBI_NEED_FSCK);
1624 
1625 	for (i = 0; i < NR_COUNT_TYPE; i++)
1626 		atomic_set(&sbi->nr_pages[i], 0);
1627 
1628 	atomic_set(&sbi->wb_sync_req, 0);
1629 
1630 	INIT_LIST_HEAD(&sbi->s_list);
1631 	mutex_init(&sbi->umount_mutex);
1632 	for (i = 0; i < NR_PAGE_TYPE - 1; i++)
1633 		for (j = HOT; j < NR_TEMP_TYPE; j++)
1634 			mutex_init(&sbi->wio_mutex[i][j]);
1635 	spin_lock_init(&sbi->cp_lock);
1636 }
1637 
1638 static int init_percpu_info(struct f2fs_sb_info *sbi)
1639 {
1640 	int err;
1641 
1642 	err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
1643 	if (err)
1644 		return err;
1645 
1646 	return percpu_counter_init(&sbi->total_valid_inode_count, 0,
1647 								GFP_KERNEL);
1648 }
1649 
1650 #ifdef CONFIG_BLK_DEV_ZONED
1651 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
1652 {
1653 	struct block_device *bdev = FDEV(devi).bdev;
1654 	sector_t nr_sectors = bdev->bd_part->nr_sects;
1655 	sector_t sector = 0;
1656 	struct blk_zone *zones;
1657 	unsigned int i, nr_zones;
1658 	unsigned int n = 0;
1659 	int err = -EIO;
1660 
1661 	if (!f2fs_sb_mounted_blkzoned(sbi->sb))
1662 		return 0;
1663 
1664 	if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
1665 				SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
1666 		return -EINVAL;
1667 	sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
1668 	if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
1669 				__ilog2_u32(sbi->blocks_per_blkz))
1670 		return -EINVAL;
1671 	sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
1672 	FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
1673 					sbi->log_blocks_per_blkz;
1674 	if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
1675 		FDEV(devi).nr_blkz++;
1676 
1677 	FDEV(devi).blkz_type = kmalloc(FDEV(devi).nr_blkz, GFP_KERNEL);
1678 	if (!FDEV(devi).blkz_type)
1679 		return -ENOMEM;
1680 
1681 #define F2FS_REPORT_NR_ZONES   4096
1682 
1683 	zones = kcalloc(F2FS_REPORT_NR_ZONES, sizeof(struct blk_zone),
1684 			GFP_KERNEL);
1685 	if (!zones)
1686 		return -ENOMEM;
1687 
1688 	/* Get block zones type */
1689 	while (zones && sector < nr_sectors) {
1690 
1691 		nr_zones = F2FS_REPORT_NR_ZONES;
1692 		err = blkdev_report_zones(bdev, sector,
1693 					  zones, &nr_zones,
1694 					  GFP_KERNEL);
1695 		if (err)
1696 			break;
1697 		if (!nr_zones) {
1698 			err = -EIO;
1699 			break;
1700 		}
1701 
1702 		for (i = 0; i < nr_zones; i++) {
1703 			FDEV(devi).blkz_type[n] = zones[i].type;
1704 			sector += zones[i].len;
1705 			n++;
1706 		}
1707 	}
1708 
1709 	kfree(zones);
1710 
1711 	return err;
1712 }
1713 #endif
1714 
1715 /*
1716  * Read f2fs raw super block.
1717  * Because we have two copies of super block, so read both of them
1718  * to get the first valid one. If any one of them is broken, we pass
1719  * them recovery flag back to the caller.
1720  */
1721 static int read_raw_super_block(struct f2fs_sb_info *sbi,
1722 			struct f2fs_super_block **raw_super,
1723 			int *valid_super_block, int *recovery)
1724 {
1725 	struct super_block *sb = sbi->sb;
1726 	int block;
1727 	struct buffer_head *bh;
1728 	struct f2fs_super_block *super;
1729 	int err = 0;
1730 
1731 	super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
1732 	if (!super)
1733 		return -ENOMEM;
1734 
1735 	for (block = 0; block < 2; block++) {
1736 		bh = sb_bread(sb, block);
1737 		if (!bh) {
1738 			f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
1739 				block + 1);
1740 			err = -EIO;
1741 			continue;
1742 		}
1743 
1744 		/* sanity checking of raw super */
1745 		if (sanity_check_raw_super(sbi, bh)) {
1746 			f2fs_msg(sb, KERN_ERR,
1747 				"Can't find valid F2FS filesystem in %dth superblock",
1748 				block + 1);
1749 			err = -EINVAL;
1750 			brelse(bh);
1751 			continue;
1752 		}
1753 
1754 		if (!*raw_super) {
1755 			memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
1756 							sizeof(*super));
1757 			*valid_super_block = block;
1758 			*raw_super = super;
1759 		}
1760 		brelse(bh);
1761 	}
1762 
1763 	/* Fail to read any one of the superblocks*/
1764 	if (err < 0)
1765 		*recovery = 1;
1766 
1767 	/* No valid superblock */
1768 	if (!*raw_super)
1769 		kfree(super);
1770 	else
1771 		err = 0;
1772 
1773 	return err;
1774 }
1775 
1776 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
1777 {
1778 	struct buffer_head *bh;
1779 	int err;
1780 
1781 	if ((recover && f2fs_readonly(sbi->sb)) ||
1782 				bdev_read_only(sbi->sb->s_bdev)) {
1783 		set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1784 		return -EROFS;
1785 	}
1786 
1787 	/* write back-up superblock first */
1788 	bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1);
1789 	if (!bh)
1790 		return -EIO;
1791 	err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1792 	brelse(bh);
1793 
1794 	/* if we are in recovery path, skip writing valid superblock */
1795 	if (recover || err)
1796 		return err;
1797 
1798 	/* write current valid superblock */
1799 	bh = sb_getblk(sbi->sb, sbi->valid_super_block);
1800 	if (!bh)
1801 		return -EIO;
1802 	err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1803 	brelse(bh);
1804 	return err;
1805 }
1806 
1807 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
1808 {
1809 	struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1810 	unsigned int max_devices = MAX_DEVICES;
1811 	int i;
1812 
1813 	/* Initialize single device information */
1814 	if (!RDEV(0).path[0]) {
1815 		if (!bdev_is_zoned(sbi->sb->s_bdev))
1816 			return 0;
1817 		max_devices = 1;
1818 	}
1819 
1820 	/*
1821 	 * Initialize multiple devices information, or single
1822 	 * zoned block device information.
1823 	 */
1824 	sbi->devs = kcalloc(max_devices, sizeof(struct f2fs_dev_info),
1825 				GFP_KERNEL);
1826 	if (!sbi->devs)
1827 		return -ENOMEM;
1828 
1829 	for (i = 0; i < max_devices; i++) {
1830 
1831 		if (i > 0 && !RDEV(i).path[0])
1832 			break;
1833 
1834 		if (max_devices == 1) {
1835 			/* Single zoned block device mount */
1836 			FDEV(0).bdev =
1837 				blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
1838 					sbi->sb->s_mode, sbi->sb->s_type);
1839 		} else {
1840 			/* Multi-device mount */
1841 			memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
1842 			FDEV(i).total_segments =
1843 				le32_to_cpu(RDEV(i).total_segments);
1844 			if (i == 0) {
1845 				FDEV(i).start_blk = 0;
1846 				FDEV(i).end_blk = FDEV(i).start_blk +
1847 				    (FDEV(i).total_segments <<
1848 				    sbi->log_blocks_per_seg) - 1 +
1849 				    le32_to_cpu(raw_super->segment0_blkaddr);
1850 			} else {
1851 				FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
1852 				FDEV(i).end_blk = FDEV(i).start_blk +
1853 					(FDEV(i).total_segments <<
1854 					sbi->log_blocks_per_seg) - 1;
1855 			}
1856 			FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
1857 					sbi->sb->s_mode, sbi->sb->s_type);
1858 		}
1859 		if (IS_ERR(FDEV(i).bdev))
1860 			return PTR_ERR(FDEV(i).bdev);
1861 
1862 		/* to release errored devices */
1863 		sbi->s_ndevs = i + 1;
1864 
1865 #ifdef CONFIG_BLK_DEV_ZONED
1866 		if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
1867 				!f2fs_sb_mounted_blkzoned(sbi->sb)) {
1868 			f2fs_msg(sbi->sb, KERN_ERR,
1869 				"Zoned block device feature not enabled\n");
1870 			return -EINVAL;
1871 		}
1872 		if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
1873 			if (init_blkz_info(sbi, i)) {
1874 				f2fs_msg(sbi->sb, KERN_ERR,
1875 					"Failed to initialize F2FS blkzone information");
1876 				return -EINVAL;
1877 			}
1878 			if (max_devices == 1)
1879 				break;
1880 			f2fs_msg(sbi->sb, KERN_INFO,
1881 				"Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
1882 				i, FDEV(i).path,
1883 				FDEV(i).total_segments,
1884 				FDEV(i).start_blk, FDEV(i).end_blk,
1885 				bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
1886 				"Host-aware" : "Host-managed");
1887 			continue;
1888 		}
1889 #endif
1890 		f2fs_msg(sbi->sb, KERN_INFO,
1891 			"Mount Device [%2d]: %20s, %8u, %8x - %8x",
1892 				i, FDEV(i).path,
1893 				FDEV(i).total_segments,
1894 				FDEV(i).start_blk, FDEV(i).end_blk);
1895 	}
1896 	f2fs_msg(sbi->sb, KERN_INFO,
1897 			"IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
1898 	return 0;
1899 }
1900 
1901 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
1902 {
1903 	struct f2fs_sb_info *sbi;
1904 	struct f2fs_super_block *raw_super;
1905 	struct inode *root;
1906 	int err;
1907 	bool retry = true, need_fsck = false;
1908 	char *options = NULL;
1909 	int recovery, i, valid_super_block;
1910 	struct curseg_info *seg_i;
1911 
1912 try_onemore:
1913 	err = -EINVAL;
1914 	raw_super = NULL;
1915 	valid_super_block = -1;
1916 	recovery = 0;
1917 
1918 	/* allocate memory for f2fs-specific super block info */
1919 	sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
1920 	if (!sbi)
1921 		return -ENOMEM;
1922 
1923 	sbi->sb = sb;
1924 
1925 	/* Load the checksum driver */
1926 	sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
1927 	if (IS_ERR(sbi->s_chksum_driver)) {
1928 		f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
1929 		err = PTR_ERR(sbi->s_chksum_driver);
1930 		sbi->s_chksum_driver = NULL;
1931 		goto free_sbi;
1932 	}
1933 
1934 	/* set a block size */
1935 	if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
1936 		f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
1937 		goto free_sbi;
1938 	}
1939 
1940 	err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
1941 								&recovery);
1942 	if (err)
1943 		goto free_sbi;
1944 
1945 	sb->s_fs_info = sbi;
1946 	sbi->raw_super = raw_super;
1947 
1948 	/*
1949 	 * The BLKZONED feature indicates that the drive was formatted with
1950 	 * zone alignment optimization. This is optional for host-aware
1951 	 * devices, but mandatory for host-managed zoned block devices.
1952 	 */
1953 #ifndef CONFIG_BLK_DEV_ZONED
1954 	if (f2fs_sb_mounted_blkzoned(sb)) {
1955 		f2fs_msg(sb, KERN_ERR,
1956 			 "Zoned block device support is not enabled\n");
1957 		err = -EOPNOTSUPP;
1958 		goto free_sb_buf;
1959 	}
1960 #endif
1961 	default_options(sbi);
1962 	/* parse mount options */
1963 	options = kstrdup((const char *)data, GFP_KERNEL);
1964 	if (data && !options) {
1965 		err = -ENOMEM;
1966 		goto free_sb_buf;
1967 	}
1968 
1969 	err = parse_options(sb, options);
1970 	if (err)
1971 		goto free_options;
1972 
1973 	sbi->max_file_blocks = max_file_blocks();
1974 	sb->s_maxbytes = sbi->max_file_blocks <<
1975 				le32_to_cpu(raw_super->log_blocksize);
1976 	sb->s_max_links = F2FS_LINK_MAX;
1977 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1978 
1979 #ifdef CONFIG_QUOTA
1980 	sb->dq_op = &f2fs_quota_operations;
1981 	sb->s_qcop = &f2fs_quotactl_ops;
1982 	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
1983 #endif
1984 
1985 	sb->s_op = &f2fs_sops;
1986 	sb->s_cop = &f2fs_cryptops;
1987 	sb->s_xattr = f2fs_xattr_handlers;
1988 	sb->s_export_op = &f2fs_export_ops;
1989 	sb->s_magic = F2FS_SUPER_MAGIC;
1990 	sb->s_time_gran = 1;
1991 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1992 		(test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1993 	memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1994 
1995 	/* init f2fs-specific super block info */
1996 	sbi->valid_super_block = valid_super_block;
1997 	mutex_init(&sbi->gc_mutex);
1998 	mutex_init(&sbi->cp_mutex);
1999 	init_rwsem(&sbi->node_write);
2000 	init_rwsem(&sbi->node_change);
2001 
2002 	/* disallow all the data/node/meta page writes */
2003 	set_sbi_flag(sbi, SBI_POR_DOING);
2004 	spin_lock_init(&sbi->stat_lock);
2005 
2006 	for (i = 0; i < NR_PAGE_TYPE; i++) {
2007 		int n = (i == META) ? 1: NR_TEMP_TYPE;
2008 		int j;
2009 
2010 		sbi->write_io[i] = kmalloc(n * sizeof(struct f2fs_bio_info),
2011 								GFP_KERNEL);
2012 		if (!sbi->write_io[i]) {
2013 			err = -ENOMEM;
2014 			goto free_options;
2015 		}
2016 
2017 		for (j = HOT; j < n; j++) {
2018 			init_rwsem(&sbi->write_io[i][j].io_rwsem);
2019 			sbi->write_io[i][j].sbi = sbi;
2020 			sbi->write_io[i][j].bio = NULL;
2021 			spin_lock_init(&sbi->write_io[i][j].io_lock);
2022 			INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
2023 		}
2024 	}
2025 
2026 	init_rwsem(&sbi->cp_rwsem);
2027 	init_waitqueue_head(&sbi->cp_wait);
2028 	init_sb_info(sbi);
2029 
2030 	err = init_percpu_info(sbi);
2031 	if (err)
2032 		goto free_options;
2033 
2034 	if (F2FS_IO_SIZE(sbi) > 1) {
2035 		sbi->write_io_dummy =
2036 			mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
2037 		if (!sbi->write_io_dummy) {
2038 			err = -ENOMEM;
2039 			goto free_options;
2040 		}
2041 	}
2042 
2043 	/* get an inode for meta space */
2044 	sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
2045 	if (IS_ERR(sbi->meta_inode)) {
2046 		f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
2047 		err = PTR_ERR(sbi->meta_inode);
2048 		goto free_io_dummy;
2049 	}
2050 
2051 	err = get_valid_checkpoint(sbi);
2052 	if (err) {
2053 		f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
2054 		goto free_meta_inode;
2055 	}
2056 
2057 	/* Initialize device list */
2058 	err = f2fs_scan_devices(sbi);
2059 	if (err) {
2060 		f2fs_msg(sb, KERN_ERR, "Failed to find devices");
2061 		goto free_devices;
2062 	}
2063 
2064 	sbi->total_valid_node_count =
2065 				le32_to_cpu(sbi->ckpt->valid_node_count);
2066 	percpu_counter_set(&sbi->total_valid_inode_count,
2067 				le32_to_cpu(sbi->ckpt->valid_inode_count));
2068 	sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
2069 	sbi->total_valid_block_count =
2070 				le64_to_cpu(sbi->ckpt->valid_block_count);
2071 	sbi->last_valid_block_count = sbi->total_valid_block_count;
2072 	sbi->reserved_blocks = 0;
2073 
2074 	for (i = 0; i < NR_INODE_TYPE; i++) {
2075 		INIT_LIST_HEAD(&sbi->inode_list[i]);
2076 		spin_lock_init(&sbi->inode_lock[i]);
2077 	}
2078 
2079 	init_extent_cache_info(sbi);
2080 
2081 	init_ino_entry_info(sbi);
2082 
2083 	/* setup f2fs internal modules */
2084 	err = build_segment_manager(sbi);
2085 	if (err) {
2086 		f2fs_msg(sb, KERN_ERR,
2087 			"Failed to initialize F2FS segment manager");
2088 		goto free_sm;
2089 	}
2090 	err = build_node_manager(sbi);
2091 	if (err) {
2092 		f2fs_msg(sb, KERN_ERR,
2093 			"Failed to initialize F2FS node manager");
2094 		goto free_nm;
2095 	}
2096 
2097 	/* For write statistics */
2098 	if (sb->s_bdev->bd_part)
2099 		sbi->sectors_written_start =
2100 			(u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
2101 
2102 	/* Read accumulated write IO statistics if exists */
2103 	seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
2104 	if (__exist_node_summaries(sbi))
2105 		sbi->kbytes_written =
2106 			le64_to_cpu(seg_i->journal->info.kbytes_written);
2107 
2108 	build_gc_manager(sbi);
2109 
2110 	/* get an inode for node space */
2111 	sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
2112 	if (IS_ERR(sbi->node_inode)) {
2113 		f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
2114 		err = PTR_ERR(sbi->node_inode);
2115 		goto free_nm;
2116 	}
2117 
2118 	f2fs_join_shrinker(sbi);
2119 
2120 	err = f2fs_build_stats(sbi);
2121 	if (err)
2122 		goto free_nm;
2123 
2124 	/* if there are nt orphan nodes free them */
2125 	err = recover_orphan_inodes(sbi);
2126 	if (err)
2127 		goto free_node_inode;
2128 
2129 	/* read root inode and dentry */
2130 	root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
2131 	if (IS_ERR(root)) {
2132 		f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
2133 		err = PTR_ERR(root);
2134 		goto free_node_inode;
2135 	}
2136 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2137 		iput(root);
2138 		err = -EINVAL;
2139 		goto free_node_inode;
2140 	}
2141 
2142 	sb->s_root = d_make_root(root); /* allocate root dentry */
2143 	if (!sb->s_root) {
2144 		err = -ENOMEM;
2145 		goto free_root_inode;
2146 	}
2147 
2148 	err = f2fs_init_sysfs(sbi);
2149 	if (err)
2150 		goto free_root_inode;
2151 
2152 	/* recover fsynced data */
2153 	if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
2154 		/*
2155 		 * mount should be failed, when device has readonly mode, and
2156 		 * previous checkpoint was not done by clean system shutdown.
2157 		 */
2158 		if (bdev_read_only(sb->s_bdev) &&
2159 				!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
2160 			err = -EROFS;
2161 			goto free_sysfs;
2162 		}
2163 
2164 		if (need_fsck)
2165 			set_sbi_flag(sbi, SBI_NEED_FSCK);
2166 
2167 		if (!retry)
2168 			goto skip_recovery;
2169 
2170 		err = recover_fsync_data(sbi, false);
2171 		if (err < 0) {
2172 			need_fsck = true;
2173 			f2fs_msg(sb, KERN_ERR,
2174 				"Cannot recover all fsync data errno=%d", err);
2175 			goto free_sysfs;
2176 		}
2177 	} else {
2178 		err = recover_fsync_data(sbi, true);
2179 
2180 		if (!f2fs_readonly(sb) && err > 0) {
2181 			err = -EINVAL;
2182 			f2fs_msg(sb, KERN_ERR,
2183 				"Need to recover fsync data");
2184 			goto free_sysfs;
2185 		}
2186 	}
2187 skip_recovery:
2188 	/* recover_fsync_data() cleared this already */
2189 	clear_sbi_flag(sbi, SBI_POR_DOING);
2190 
2191 	/*
2192 	 * If filesystem is not mounted as read-only then
2193 	 * do start the gc_thread.
2194 	 */
2195 	if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
2196 		/* After POR, we can run background GC thread.*/
2197 		err = start_gc_thread(sbi);
2198 		if (err)
2199 			goto free_sysfs;
2200 	}
2201 	kfree(options);
2202 
2203 	/* recover broken superblock */
2204 	if (recovery) {
2205 		err = f2fs_commit_super(sbi, true);
2206 		f2fs_msg(sb, KERN_INFO,
2207 			"Try to recover %dth superblock, ret: %d",
2208 			sbi->valid_super_block ? 1 : 2, err);
2209 	}
2210 
2211 	f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx",
2212 				cur_cp_version(F2FS_CKPT(sbi)));
2213 	f2fs_update_time(sbi, CP_TIME);
2214 	f2fs_update_time(sbi, REQ_TIME);
2215 	return 0;
2216 
2217 free_sysfs:
2218 	f2fs_sync_inode_meta(sbi);
2219 	f2fs_exit_sysfs(sbi);
2220 free_root_inode:
2221 	dput(sb->s_root);
2222 	sb->s_root = NULL;
2223 free_node_inode:
2224 	truncate_inode_pages_final(NODE_MAPPING(sbi));
2225 	mutex_lock(&sbi->umount_mutex);
2226 	release_ino_entry(sbi, true);
2227 	f2fs_leave_shrinker(sbi);
2228 	/*
2229 	 * Some dirty meta pages can be produced by recover_orphan_inodes()
2230 	 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
2231 	 * followed by write_checkpoint() through f2fs_write_node_pages(), which
2232 	 * falls into an infinite loop in sync_meta_pages().
2233 	 */
2234 	truncate_inode_pages_final(META_MAPPING(sbi));
2235 	iput(sbi->node_inode);
2236 	mutex_unlock(&sbi->umount_mutex);
2237 	f2fs_destroy_stats(sbi);
2238 free_nm:
2239 	destroy_node_manager(sbi);
2240 free_sm:
2241 	destroy_segment_manager(sbi);
2242 free_devices:
2243 	destroy_device_list(sbi);
2244 	kfree(sbi->ckpt);
2245 free_meta_inode:
2246 	make_bad_inode(sbi->meta_inode);
2247 	iput(sbi->meta_inode);
2248 free_io_dummy:
2249 	mempool_destroy(sbi->write_io_dummy);
2250 free_options:
2251 	for (i = 0; i < NR_PAGE_TYPE; i++)
2252 		kfree(sbi->write_io[i]);
2253 	destroy_percpu_info(sbi);
2254 	kfree(options);
2255 free_sb_buf:
2256 	kfree(raw_super);
2257 free_sbi:
2258 	if (sbi->s_chksum_driver)
2259 		crypto_free_shash(sbi->s_chksum_driver);
2260 	kfree(sbi);
2261 
2262 	/* give only one another chance */
2263 	if (retry) {
2264 		retry = false;
2265 		shrink_dcache_sb(sb);
2266 		goto try_onemore;
2267 	}
2268 	return err;
2269 }
2270 
2271 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
2272 			const char *dev_name, void *data)
2273 {
2274 	return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
2275 }
2276 
2277 static void kill_f2fs_super(struct super_block *sb)
2278 {
2279 	if (sb->s_root) {
2280 		set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
2281 		stop_gc_thread(F2FS_SB(sb));
2282 		stop_discard_thread(F2FS_SB(sb));
2283 	}
2284 	kill_block_super(sb);
2285 }
2286 
2287 static struct file_system_type f2fs_fs_type = {
2288 	.owner		= THIS_MODULE,
2289 	.name		= "f2fs",
2290 	.mount		= f2fs_mount,
2291 	.kill_sb	= kill_f2fs_super,
2292 	.fs_flags	= FS_REQUIRES_DEV,
2293 };
2294 MODULE_ALIAS_FS("f2fs");
2295 
2296 static int __init init_inodecache(void)
2297 {
2298 	f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
2299 			sizeof(struct f2fs_inode_info), 0,
2300 			SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
2301 	if (!f2fs_inode_cachep)
2302 		return -ENOMEM;
2303 	return 0;
2304 }
2305 
2306 static void destroy_inodecache(void)
2307 {
2308 	/*
2309 	 * Make sure all delayed rcu free inodes are flushed before we
2310 	 * destroy cache.
2311 	 */
2312 	rcu_barrier();
2313 	kmem_cache_destroy(f2fs_inode_cachep);
2314 }
2315 
2316 static int __init init_f2fs_fs(void)
2317 {
2318 	int err;
2319 
2320 	f2fs_build_trace_ios();
2321 
2322 	err = init_inodecache();
2323 	if (err)
2324 		goto fail;
2325 	err = create_node_manager_caches();
2326 	if (err)
2327 		goto free_inodecache;
2328 	err = create_segment_manager_caches();
2329 	if (err)
2330 		goto free_node_manager_caches;
2331 	err = create_checkpoint_caches();
2332 	if (err)
2333 		goto free_segment_manager_caches;
2334 	err = create_extent_cache();
2335 	if (err)
2336 		goto free_checkpoint_caches;
2337 	err = f2fs_register_sysfs();
2338 	if (err)
2339 		goto free_extent_cache;
2340 	err = register_shrinker(&f2fs_shrinker_info);
2341 	if (err)
2342 		goto free_sysfs;
2343 	err = register_filesystem(&f2fs_fs_type);
2344 	if (err)
2345 		goto free_shrinker;
2346 	err = f2fs_create_root_stats();
2347 	if (err)
2348 		goto free_filesystem;
2349 	return 0;
2350 
2351 free_filesystem:
2352 	unregister_filesystem(&f2fs_fs_type);
2353 free_shrinker:
2354 	unregister_shrinker(&f2fs_shrinker_info);
2355 free_sysfs:
2356 	f2fs_unregister_sysfs();
2357 free_extent_cache:
2358 	destroy_extent_cache();
2359 free_checkpoint_caches:
2360 	destroy_checkpoint_caches();
2361 free_segment_manager_caches:
2362 	destroy_segment_manager_caches();
2363 free_node_manager_caches:
2364 	destroy_node_manager_caches();
2365 free_inodecache:
2366 	destroy_inodecache();
2367 fail:
2368 	return err;
2369 }
2370 
2371 static void __exit exit_f2fs_fs(void)
2372 {
2373 	f2fs_destroy_root_stats();
2374 	unregister_filesystem(&f2fs_fs_type);
2375 	unregister_shrinker(&f2fs_shrinker_info);
2376 	f2fs_unregister_sysfs();
2377 	destroy_extent_cache();
2378 	destroy_checkpoint_caches();
2379 	destroy_segment_manager_caches();
2380 	destroy_node_manager_caches();
2381 	destroy_inodecache();
2382 	f2fs_destroy_trace_ios();
2383 }
2384 
2385 module_init(init_f2fs_fs)
2386 module_exit(exit_f2fs_fs)
2387 
2388 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
2389 MODULE_DESCRIPTION("Flash Friendly File System");
2390 MODULE_LICENSE("GPL");
2391 
2392