xref: /linux/fs/f2fs/super.c (revision 73b0140bf0fe9df90fb267c00673c4b9bf285430)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * fs/f2fs/super.c
4  *
5  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6  *             http://www.samsung.com/
7  */
8 #include <linux/module.h>
9 #include <linux/init.h>
10 #include <linux/fs.h>
11 #include <linux/statfs.h>
12 #include <linux/buffer_head.h>
13 #include <linux/backing-dev.h>
14 #include <linux/kthread.h>
15 #include <linux/parser.h>
16 #include <linux/mount.h>
17 #include <linux/seq_file.h>
18 #include <linux/proc_fs.h>
19 #include <linux/random.h>
20 #include <linux/exportfs.h>
21 #include <linux/blkdev.h>
22 #include <linux/quotaops.h>
23 #include <linux/f2fs_fs.h>
24 #include <linux/sysfs.h>
25 #include <linux/quota.h>
26 
27 #include "f2fs.h"
28 #include "node.h"
29 #include "segment.h"
30 #include "xattr.h"
31 #include "gc.h"
32 #include "trace.h"
33 
34 #define CREATE_TRACE_POINTS
35 #include <trace/events/f2fs.h>
36 
37 static struct kmem_cache *f2fs_inode_cachep;
38 
39 #ifdef CONFIG_F2FS_FAULT_INJECTION
40 
41 const char *f2fs_fault_name[FAULT_MAX] = {
42 	[FAULT_KMALLOC]		= "kmalloc",
43 	[FAULT_KVMALLOC]	= "kvmalloc",
44 	[FAULT_PAGE_ALLOC]	= "page alloc",
45 	[FAULT_PAGE_GET]	= "page get",
46 	[FAULT_ALLOC_BIO]	= "alloc bio",
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_TRUNCATE]	= "truncate fail",
53 	[FAULT_READ_IO]		= "read IO error",
54 	[FAULT_CHECKPOINT]	= "checkpoint error",
55 	[FAULT_DISCARD]		= "discard error",
56 	[FAULT_WRITE_IO]	= "write IO error",
57 };
58 
59 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
60 							unsigned int type)
61 {
62 	struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
63 
64 	if (rate) {
65 		atomic_set(&ffi->inject_ops, 0);
66 		ffi->inject_rate = rate;
67 	}
68 
69 	if (type)
70 		ffi->inject_type = type;
71 
72 	if (!rate && !type)
73 		memset(ffi, 0, sizeof(struct f2fs_fault_info));
74 }
75 #endif
76 
77 /* f2fs-wide shrinker description */
78 static struct shrinker f2fs_shrinker_info = {
79 	.scan_objects = f2fs_shrink_scan,
80 	.count_objects = f2fs_shrink_count,
81 	.seeks = DEFAULT_SEEKS,
82 };
83 
84 enum {
85 	Opt_gc_background,
86 	Opt_disable_roll_forward,
87 	Opt_norecovery,
88 	Opt_discard,
89 	Opt_nodiscard,
90 	Opt_noheap,
91 	Opt_heap,
92 	Opt_user_xattr,
93 	Opt_nouser_xattr,
94 	Opt_acl,
95 	Opt_noacl,
96 	Opt_active_logs,
97 	Opt_disable_ext_identify,
98 	Opt_inline_xattr,
99 	Opt_noinline_xattr,
100 	Opt_inline_xattr_size,
101 	Opt_inline_data,
102 	Opt_inline_dentry,
103 	Opt_noinline_dentry,
104 	Opt_flush_merge,
105 	Opt_noflush_merge,
106 	Opt_nobarrier,
107 	Opt_fastboot,
108 	Opt_extent_cache,
109 	Opt_noextent_cache,
110 	Opt_noinline_data,
111 	Opt_data_flush,
112 	Opt_reserve_root,
113 	Opt_resgid,
114 	Opt_resuid,
115 	Opt_mode,
116 	Opt_io_size_bits,
117 	Opt_fault_injection,
118 	Opt_fault_type,
119 	Opt_lazytime,
120 	Opt_nolazytime,
121 	Opt_quota,
122 	Opt_noquota,
123 	Opt_usrquota,
124 	Opt_grpquota,
125 	Opt_prjquota,
126 	Opt_usrjquota,
127 	Opt_grpjquota,
128 	Opt_prjjquota,
129 	Opt_offusrjquota,
130 	Opt_offgrpjquota,
131 	Opt_offprjjquota,
132 	Opt_jqfmt_vfsold,
133 	Opt_jqfmt_vfsv0,
134 	Opt_jqfmt_vfsv1,
135 	Opt_whint,
136 	Opt_alloc,
137 	Opt_fsync,
138 	Opt_test_dummy_encryption,
139 	Opt_checkpoint,
140 	Opt_err,
141 };
142 
143 static match_table_t f2fs_tokens = {
144 	{Opt_gc_background, "background_gc=%s"},
145 	{Opt_disable_roll_forward, "disable_roll_forward"},
146 	{Opt_norecovery, "norecovery"},
147 	{Opt_discard, "discard"},
148 	{Opt_nodiscard, "nodiscard"},
149 	{Opt_noheap, "no_heap"},
150 	{Opt_heap, "heap"},
151 	{Opt_user_xattr, "user_xattr"},
152 	{Opt_nouser_xattr, "nouser_xattr"},
153 	{Opt_acl, "acl"},
154 	{Opt_noacl, "noacl"},
155 	{Opt_active_logs, "active_logs=%u"},
156 	{Opt_disable_ext_identify, "disable_ext_identify"},
157 	{Opt_inline_xattr, "inline_xattr"},
158 	{Opt_noinline_xattr, "noinline_xattr"},
159 	{Opt_inline_xattr_size, "inline_xattr_size=%u"},
160 	{Opt_inline_data, "inline_data"},
161 	{Opt_inline_dentry, "inline_dentry"},
162 	{Opt_noinline_dentry, "noinline_dentry"},
163 	{Opt_flush_merge, "flush_merge"},
164 	{Opt_noflush_merge, "noflush_merge"},
165 	{Opt_nobarrier, "nobarrier"},
166 	{Opt_fastboot, "fastboot"},
167 	{Opt_extent_cache, "extent_cache"},
168 	{Opt_noextent_cache, "noextent_cache"},
169 	{Opt_noinline_data, "noinline_data"},
170 	{Opt_data_flush, "data_flush"},
171 	{Opt_reserve_root, "reserve_root=%u"},
172 	{Opt_resgid, "resgid=%u"},
173 	{Opt_resuid, "resuid=%u"},
174 	{Opt_mode, "mode=%s"},
175 	{Opt_io_size_bits, "io_bits=%u"},
176 	{Opt_fault_injection, "fault_injection=%u"},
177 	{Opt_fault_type, "fault_type=%u"},
178 	{Opt_lazytime, "lazytime"},
179 	{Opt_nolazytime, "nolazytime"},
180 	{Opt_quota, "quota"},
181 	{Opt_noquota, "noquota"},
182 	{Opt_usrquota, "usrquota"},
183 	{Opt_grpquota, "grpquota"},
184 	{Opt_prjquota, "prjquota"},
185 	{Opt_usrjquota, "usrjquota=%s"},
186 	{Opt_grpjquota, "grpjquota=%s"},
187 	{Opt_prjjquota, "prjjquota=%s"},
188 	{Opt_offusrjquota, "usrjquota="},
189 	{Opt_offgrpjquota, "grpjquota="},
190 	{Opt_offprjjquota, "prjjquota="},
191 	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
192 	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
193 	{Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
194 	{Opt_whint, "whint_mode=%s"},
195 	{Opt_alloc, "alloc_mode=%s"},
196 	{Opt_fsync, "fsync_mode=%s"},
197 	{Opt_test_dummy_encryption, "test_dummy_encryption"},
198 	{Opt_checkpoint, "checkpoint=%s"},
199 	{Opt_err, NULL},
200 };
201 
202 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
203 {
204 	struct va_format vaf;
205 	va_list args;
206 
207 	va_start(args, fmt);
208 	vaf.fmt = fmt;
209 	vaf.va = &args;
210 	printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
211 	va_end(args);
212 }
213 
214 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
215 {
216 	block_t limit = (sbi->user_block_count << 1) / 1000;
217 
218 	/* limit is 0.2% */
219 	if (test_opt(sbi, RESERVE_ROOT) &&
220 			F2FS_OPTION(sbi).root_reserved_blocks > limit) {
221 		F2FS_OPTION(sbi).root_reserved_blocks = limit;
222 		f2fs_msg(sbi->sb, KERN_INFO,
223 			"Reduce reserved blocks for root = %u",
224 			F2FS_OPTION(sbi).root_reserved_blocks);
225 	}
226 	if (!test_opt(sbi, RESERVE_ROOT) &&
227 		(!uid_eq(F2FS_OPTION(sbi).s_resuid,
228 				make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
229 		!gid_eq(F2FS_OPTION(sbi).s_resgid,
230 				make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
231 		f2fs_msg(sbi->sb, KERN_INFO,
232 			"Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
233 				from_kuid_munged(&init_user_ns,
234 					F2FS_OPTION(sbi).s_resuid),
235 				from_kgid_munged(&init_user_ns,
236 					F2FS_OPTION(sbi).s_resgid));
237 }
238 
239 static void init_once(void *foo)
240 {
241 	struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
242 
243 	inode_init_once(&fi->vfs_inode);
244 }
245 
246 #ifdef CONFIG_QUOTA
247 static const char * const quotatypes[] = INITQFNAMES;
248 #define QTYPE2NAME(t) (quotatypes[t])
249 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
250 							substring_t *args)
251 {
252 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
253 	char *qname;
254 	int ret = -EINVAL;
255 
256 	if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
257 		f2fs_msg(sb, KERN_ERR,
258 			"Cannot change journaled "
259 			"quota options when quota turned on");
260 		return -EINVAL;
261 	}
262 	if (f2fs_sb_has_quota_ino(sbi)) {
263 		f2fs_msg(sb, KERN_INFO,
264 			"QUOTA feature is enabled, so ignore qf_name");
265 		return 0;
266 	}
267 
268 	qname = match_strdup(args);
269 	if (!qname) {
270 		f2fs_msg(sb, KERN_ERR,
271 			"Not enough memory for storing quotafile name");
272 		return -ENOMEM;
273 	}
274 	if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
275 		if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
276 			ret = 0;
277 		else
278 			f2fs_msg(sb, KERN_ERR,
279 				 "%s quota file already specified",
280 				 QTYPE2NAME(qtype));
281 		goto errout;
282 	}
283 	if (strchr(qname, '/')) {
284 		f2fs_msg(sb, KERN_ERR,
285 			"quotafile must be on filesystem root");
286 		goto errout;
287 	}
288 	F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
289 	set_opt(sbi, QUOTA);
290 	return 0;
291 errout:
292 	kvfree(qname);
293 	return ret;
294 }
295 
296 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
297 {
298 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
299 
300 	if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
301 		f2fs_msg(sb, KERN_ERR, "Cannot change journaled quota options"
302 			" when quota turned on");
303 		return -EINVAL;
304 	}
305 	kvfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
306 	F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
307 	return 0;
308 }
309 
310 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
311 {
312 	/*
313 	 * We do the test below only for project quotas. 'usrquota' and
314 	 * 'grpquota' mount options are allowed even without quota feature
315 	 * to support legacy quotas in quota files.
316 	 */
317 	if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
318 		f2fs_msg(sbi->sb, KERN_ERR, "Project quota feature not enabled. "
319 			 "Cannot enable project quota enforcement.");
320 		return -1;
321 	}
322 	if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
323 			F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
324 			F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
325 		if (test_opt(sbi, USRQUOTA) &&
326 				F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
327 			clear_opt(sbi, USRQUOTA);
328 
329 		if (test_opt(sbi, GRPQUOTA) &&
330 				F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
331 			clear_opt(sbi, GRPQUOTA);
332 
333 		if (test_opt(sbi, PRJQUOTA) &&
334 				F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
335 			clear_opt(sbi, PRJQUOTA);
336 
337 		if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
338 				test_opt(sbi, PRJQUOTA)) {
339 			f2fs_msg(sbi->sb, KERN_ERR, "old and new quota "
340 					"format mixing");
341 			return -1;
342 		}
343 
344 		if (!F2FS_OPTION(sbi).s_jquota_fmt) {
345 			f2fs_msg(sbi->sb, KERN_ERR, "journaled quota format "
346 					"not specified");
347 			return -1;
348 		}
349 	}
350 
351 	if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
352 		f2fs_msg(sbi->sb, KERN_INFO,
353 			"QUOTA feature is enabled, so ignore jquota_fmt");
354 		F2FS_OPTION(sbi).s_jquota_fmt = 0;
355 	}
356 	return 0;
357 }
358 #endif
359 
360 static int parse_options(struct super_block *sb, char *options)
361 {
362 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
363 	substring_t args[MAX_OPT_ARGS];
364 	char *p, *name;
365 	int arg = 0;
366 	kuid_t uid;
367 	kgid_t gid;
368 #ifdef CONFIG_QUOTA
369 	int ret;
370 #endif
371 
372 	if (!options)
373 		return 0;
374 
375 	while ((p = strsep(&options, ",")) != NULL) {
376 		int token;
377 		if (!*p)
378 			continue;
379 		/*
380 		 * Initialize args struct so we know whether arg was
381 		 * found; some options take optional arguments.
382 		 */
383 		args[0].to = args[0].from = NULL;
384 		token = match_token(p, f2fs_tokens, args);
385 
386 		switch (token) {
387 		case Opt_gc_background:
388 			name = match_strdup(&args[0]);
389 
390 			if (!name)
391 				return -ENOMEM;
392 			if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
393 				set_opt(sbi, BG_GC);
394 				clear_opt(sbi, FORCE_FG_GC);
395 			} else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
396 				clear_opt(sbi, BG_GC);
397 				clear_opt(sbi, FORCE_FG_GC);
398 			} else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
399 				set_opt(sbi, BG_GC);
400 				set_opt(sbi, FORCE_FG_GC);
401 			} else {
402 				kvfree(name);
403 				return -EINVAL;
404 			}
405 			kvfree(name);
406 			break;
407 		case Opt_disable_roll_forward:
408 			set_opt(sbi, DISABLE_ROLL_FORWARD);
409 			break;
410 		case Opt_norecovery:
411 			/* this option mounts f2fs with ro */
412 			set_opt(sbi, DISABLE_ROLL_FORWARD);
413 			if (!f2fs_readonly(sb))
414 				return -EINVAL;
415 			break;
416 		case Opt_discard:
417 			set_opt(sbi, DISCARD);
418 			break;
419 		case Opt_nodiscard:
420 			if (f2fs_sb_has_blkzoned(sbi)) {
421 				f2fs_msg(sb, KERN_WARNING,
422 					"discard is required for zoned block devices");
423 				return -EINVAL;
424 			}
425 			clear_opt(sbi, DISCARD);
426 			break;
427 		case Opt_noheap:
428 			set_opt(sbi, NOHEAP);
429 			break;
430 		case Opt_heap:
431 			clear_opt(sbi, NOHEAP);
432 			break;
433 #ifdef CONFIG_F2FS_FS_XATTR
434 		case Opt_user_xattr:
435 			set_opt(sbi, XATTR_USER);
436 			break;
437 		case Opt_nouser_xattr:
438 			clear_opt(sbi, XATTR_USER);
439 			break;
440 		case Opt_inline_xattr:
441 			set_opt(sbi, INLINE_XATTR);
442 			break;
443 		case Opt_noinline_xattr:
444 			clear_opt(sbi, INLINE_XATTR);
445 			break;
446 		case Opt_inline_xattr_size:
447 			if (args->from && match_int(args, &arg))
448 				return -EINVAL;
449 			set_opt(sbi, INLINE_XATTR_SIZE);
450 			F2FS_OPTION(sbi).inline_xattr_size = arg;
451 			break;
452 #else
453 		case Opt_user_xattr:
454 			f2fs_msg(sb, KERN_INFO,
455 				"user_xattr options not supported");
456 			break;
457 		case Opt_nouser_xattr:
458 			f2fs_msg(sb, KERN_INFO,
459 				"nouser_xattr options not supported");
460 			break;
461 		case Opt_inline_xattr:
462 			f2fs_msg(sb, KERN_INFO,
463 				"inline_xattr options not supported");
464 			break;
465 		case Opt_noinline_xattr:
466 			f2fs_msg(sb, KERN_INFO,
467 				"noinline_xattr options not supported");
468 			break;
469 #endif
470 #ifdef CONFIG_F2FS_FS_POSIX_ACL
471 		case Opt_acl:
472 			set_opt(sbi, POSIX_ACL);
473 			break;
474 		case Opt_noacl:
475 			clear_opt(sbi, POSIX_ACL);
476 			break;
477 #else
478 		case Opt_acl:
479 			f2fs_msg(sb, KERN_INFO, "acl options not supported");
480 			break;
481 		case Opt_noacl:
482 			f2fs_msg(sb, KERN_INFO, "noacl options not supported");
483 			break;
484 #endif
485 		case Opt_active_logs:
486 			if (args->from && match_int(args, &arg))
487 				return -EINVAL;
488 			if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
489 				return -EINVAL;
490 			F2FS_OPTION(sbi).active_logs = arg;
491 			break;
492 		case Opt_disable_ext_identify:
493 			set_opt(sbi, DISABLE_EXT_IDENTIFY);
494 			break;
495 		case Opt_inline_data:
496 			set_opt(sbi, INLINE_DATA);
497 			break;
498 		case Opt_inline_dentry:
499 			set_opt(sbi, INLINE_DENTRY);
500 			break;
501 		case Opt_noinline_dentry:
502 			clear_opt(sbi, INLINE_DENTRY);
503 			break;
504 		case Opt_flush_merge:
505 			set_opt(sbi, FLUSH_MERGE);
506 			break;
507 		case Opt_noflush_merge:
508 			clear_opt(sbi, FLUSH_MERGE);
509 			break;
510 		case Opt_nobarrier:
511 			set_opt(sbi, NOBARRIER);
512 			break;
513 		case Opt_fastboot:
514 			set_opt(sbi, FASTBOOT);
515 			break;
516 		case Opt_extent_cache:
517 			set_opt(sbi, EXTENT_CACHE);
518 			break;
519 		case Opt_noextent_cache:
520 			clear_opt(sbi, EXTENT_CACHE);
521 			break;
522 		case Opt_noinline_data:
523 			clear_opt(sbi, INLINE_DATA);
524 			break;
525 		case Opt_data_flush:
526 			set_opt(sbi, DATA_FLUSH);
527 			break;
528 		case Opt_reserve_root:
529 			if (args->from && match_int(args, &arg))
530 				return -EINVAL;
531 			if (test_opt(sbi, RESERVE_ROOT)) {
532 				f2fs_msg(sb, KERN_INFO,
533 					"Preserve previous reserve_root=%u",
534 					F2FS_OPTION(sbi).root_reserved_blocks);
535 			} else {
536 				F2FS_OPTION(sbi).root_reserved_blocks = arg;
537 				set_opt(sbi, RESERVE_ROOT);
538 			}
539 			break;
540 		case Opt_resuid:
541 			if (args->from && match_int(args, &arg))
542 				return -EINVAL;
543 			uid = make_kuid(current_user_ns(), arg);
544 			if (!uid_valid(uid)) {
545 				f2fs_msg(sb, KERN_ERR,
546 					"Invalid uid value %d", arg);
547 				return -EINVAL;
548 			}
549 			F2FS_OPTION(sbi).s_resuid = uid;
550 			break;
551 		case Opt_resgid:
552 			if (args->from && match_int(args, &arg))
553 				return -EINVAL;
554 			gid = make_kgid(current_user_ns(), arg);
555 			if (!gid_valid(gid)) {
556 				f2fs_msg(sb, KERN_ERR,
557 					"Invalid gid value %d", arg);
558 				return -EINVAL;
559 			}
560 			F2FS_OPTION(sbi).s_resgid = gid;
561 			break;
562 		case Opt_mode:
563 			name = match_strdup(&args[0]);
564 
565 			if (!name)
566 				return -ENOMEM;
567 			if (strlen(name) == 8 &&
568 					!strncmp(name, "adaptive", 8)) {
569 				if (f2fs_sb_has_blkzoned(sbi)) {
570 					f2fs_msg(sb, KERN_WARNING,
571 						 "adaptive mode is not allowed with "
572 						 "zoned block device feature");
573 					kvfree(name);
574 					return -EINVAL;
575 				}
576 				set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
577 			} else if (strlen(name) == 3 &&
578 					!strncmp(name, "lfs", 3)) {
579 				set_opt_mode(sbi, F2FS_MOUNT_LFS);
580 			} else {
581 				kvfree(name);
582 				return -EINVAL;
583 			}
584 			kvfree(name);
585 			break;
586 		case Opt_io_size_bits:
587 			if (args->from && match_int(args, &arg))
588 				return -EINVAL;
589 			if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_PAGES)) {
590 				f2fs_msg(sb, KERN_WARNING,
591 					"Not support %d, larger than %d",
592 					1 << arg, BIO_MAX_PAGES);
593 				return -EINVAL;
594 			}
595 			F2FS_OPTION(sbi).write_io_size_bits = arg;
596 			break;
597 #ifdef CONFIG_F2FS_FAULT_INJECTION
598 		case Opt_fault_injection:
599 			if (args->from && match_int(args, &arg))
600 				return -EINVAL;
601 			f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
602 			set_opt(sbi, FAULT_INJECTION);
603 			break;
604 
605 		case Opt_fault_type:
606 			if (args->from && match_int(args, &arg))
607 				return -EINVAL;
608 			f2fs_build_fault_attr(sbi, 0, arg);
609 			set_opt(sbi, FAULT_INJECTION);
610 			break;
611 #else
612 		case Opt_fault_injection:
613 			f2fs_msg(sb, KERN_INFO,
614 				"fault_injection options not supported");
615 			break;
616 
617 		case Opt_fault_type:
618 			f2fs_msg(sb, KERN_INFO,
619 				"fault_type options not supported");
620 			break;
621 #endif
622 		case Opt_lazytime:
623 			sb->s_flags |= SB_LAZYTIME;
624 			break;
625 		case Opt_nolazytime:
626 			sb->s_flags &= ~SB_LAZYTIME;
627 			break;
628 #ifdef CONFIG_QUOTA
629 		case Opt_quota:
630 		case Opt_usrquota:
631 			set_opt(sbi, USRQUOTA);
632 			break;
633 		case Opt_grpquota:
634 			set_opt(sbi, GRPQUOTA);
635 			break;
636 		case Opt_prjquota:
637 			set_opt(sbi, PRJQUOTA);
638 			break;
639 		case Opt_usrjquota:
640 			ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
641 			if (ret)
642 				return ret;
643 			break;
644 		case Opt_grpjquota:
645 			ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
646 			if (ret)
647 				return ret;
648 			break;
649 		case Opt_prjjquota:
650 			ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
651 			if (ret)
652 				return ret;
653 			break;
654 		case Opt_offusrjquota:
655 			ret = f2fs_clear_qf_name(sb, USRQUOTA);
656 			if (ret)
657 				return ret;
658 			break;
659 		case Opt_offgrpjquota:
660 			ret = f2fs_clear_qf_name(sb, GRPQUOTA);
661 			if (ret)
662 				return ret;
663 			break;
664 		case Opt_offprjjquota:
665 			ret = f2fs_clear_qf_name(sb, PRJQUOTA);
666 			if (ret)
667 				return ret;
668 			break;
669 		case Opt_jqfmt_vfsold:
670 			F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
671 			break;
672 		case Opt_jqfmt_vfsv0:
673 			F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
674 			break;
675 		case Opt_jqfmt_vfsv1:
676 			F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
677 			break;
678 		case Opt_noquota:
679 			clear_opt(sbi, QUOTA);
680 			clear_opt(sbi, USRQUOTA);
681 			clear_opt(sbi, GRPQUOTA);
682 			clear_opt(sbi, PRJQUOTA);
683 			break;
684 #else
685 		case Opt_quota:
686 		case Opt_usrquota:
687 		case Opt_grpquota:
688 		case Opt_prjquota:
689 		case Opt_usrjquota:
690 		case Opt_grpjquota:
691 		case Opt_prjjquota:
692 		case Opt_offusrjquota:
693 		case Opt_offgrpjquota:
694 		case Opt_offprjjquota:
695 		case Opt_jqfmt_vfsold:
696 		case Opt_jqfmt_vfsv0:
697 		case Opt_jqfmt_vfsv1:
698 		case Opt_noquota:
699 			f2fs_msg(sb, KERN_INFO,
700 					"quota operations not supported");
701 			break;
702 #endif
703 		case Opt_whint:
704 			name = match_strdup(&args[0]);
705 			if (!name)
706 				return -ENOMEM;
707 			if (strlen(name) == 10 &&
708 					!strncmp(name, "user-based", 10)) {
709 				F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
710 			} else if (strlen(name) == 3 &&
711 					!strncmp(name, "off", 3)) {
712 				F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
713 			} else if (strlen(name) == 8 &&
714 					!strncmp(name, "fs-based", 8)) {
715 				F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
716 			} else {
717 				kvfree(name);
718 				return -EINVAL;
719 			}
720 			kvfree(name);
721 			break;
722 		case Opt_alloc:
723 			name = match_strdup(&args[0]);
724 			if (!name)
725 				return -ENOMEM;
726 
727 			if (strlen(name) == 7 &&
728 					!strncmp(name, "default", 7)) {
729 				F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
730 			} else if (strlen(name) == 5 &&
731 					!strncmp(name, "reuse", 5)) {
732 				F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
733 			} else {
734 				kvfree(name);
735 				return -EINVAL;
736 			}
737 			kvfree(name);
738 			break;
739 		case Opt_fsync:
740 			name = match_strdup(&args[0]);
741 			if (!name)
742 				return -ENOMEM;
743 			if (strlen(name) == 5 &&
744 					!strncmp(name, "posix", 5)) {
745 				F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
746 			} else if (strlen(name) == 6 &&
747 					!strncmp(name, "strict", 6)) {
748 				F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
749 			} else if (strlen(name) == 9 &&
750 					!strncmp(name, "nobarrier", 9)) {
751 				F2FS_OPTION(sbi).fsync_mode =
752 							FSYNC_MODE_NOBARRIER;
753 			} else {
754 				kvfree(name);
755 				return -EINVAL;
756 			}
757 			kvfree(name);
758 			break;
759 		case Opt_test_dummy_encryption:
760 #ifdef CONFIG_FS_ENCRYPTION
761 			if (!f2fs_sb_has_encrypt(sbi)) {
762 				f2fs_msg(sb, KERN_ERR, "Encrypt feature is off");
763 				return -EINVAL;
764 			}
765 
766 			F2FS_OPTION(sbi).test_dummy_encryption = true;
767 			f2fs_msg(sb, KERN_INFO,
768 					"Test dummy encryption mode enabled");
769 #else
770 			f2fs_msg(sb, KERN_INFO,
771 					"Test dummy encryption mount option ignored");
772 #endif
773 			break;
774 		case Opt_checkpoint:
775 			name = match_strdup(&args[0]);
776 			if (!name)
777 				return -ENOMEM;
778 
779 			if (strlen(name) == 6 &&
780 					!strncmp(name, "enable", 6)) {
781 				clear_opt(sbi, DISABLE_CHECKPOINT);
782 			} else if (strlen(name) == 7 &&
783 					!strncmp(name, "disable", 7)) {
784 				set_opt(sbi, DISABLE_CHECKPOINT);
785 			} else {
786 				kvfree(name);
787 				return -EINVAL;
788 			}
789 			kvfree(name);
790 			break;
791 		default:
792 			f2fs_msg(sb, KERN_ERR,
793 				"Unrecognized mount option \"%s\" or missing value",
794 				p);
795 			return -EINVAL;
796 		}
797 	}
798 #ifdef CONFIG_QUOTA
799 	if (f2fs_check_quota_options(sbi))
800 		return -EINVAL;
801 #else
802 	if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
803 		f2fs_msg(sbi->sb, KERN_INFO,
804 			 "Filesystem with quota feature cannot be mounted RDWR "
805 			 "without CONFIG_QUOTA");
806 		return -EINVAL;
807 	}
808 	if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
809 		f2fs_msg(sb, KERN_ERR,
810 			"Filesystem with project quota feature cannot be "
811 			"mounted RDWR without CONFIG_QUOTA");
812 		return -EINVAL;
813 	}
814 #endif
815 
816 	if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) {
817 		f2fs_msg(sb, KERN_ERR,
818 				"Should set mode=lfs with %uKB-sized IO",
819 				F2FS_IO_SIZE_KB(sbi));
820 		return -EINVAL;
821 	}
822 
823 	if (test_opt(sbi, INLINE_XATTR_SIZE)) {
824 		int min_size, max_size;
825 
826 		if (!f2fs_sb_has_extra_attr(sbi) ||
827 			!f2fs_sb_has_flexible_inline_xattr(sbi)) {
828 			f2fs_msg(sb, KERN_ERR,
829 					"extra_attr or flexible_inline_xattr "
830 					"feature is off");
831 			return -EINVAL;
832 		}
833 		if (!test_opt(sbi, INLINE_XATTR)) {
834 			f2fs_msg(sb, KERN_ERR,
835 					"inline_xattr_size option should be "
836 					"set with inline_xattr option");
837 			return -EINVAL;
838 		}
839 
840 		min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
841 		max_size = MAX_INLINE_XATTR_SIZE;
842 
843 		if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
844 				F2FS_OPTION(sbi).inline_xattr_size > max_size) {
845 			f2fs_msg(sb, KERN_ERR,
846 				"inline xattr size is out of range: %d ~ %d",
847 				min_size, max_size);
848 			return -EINVAL;
849 		}
850 	}
851 
852 	if (test_opt(sbi, DISABLE_CHECKPOINT) && test_opt(sbi, LFS)) {
853 		f2fs_msg(sb, KERN_ERR,
854 				"LFS not compatible with checkpoint=disable\n");
855 		return -EINVAL;
856 	}
857 
858 	/* Not pass down write hints if the number of active logs is lesser
859 	 * than NR_CURSEG_TYPE.
860 	 */
861 	if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
862 		F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
863 	return 0;
864 }
865 
866 static struct inode *f2fs_alloc_inode(struct super_block *sb)
867 {
868 	struct f2fs_inode_info *fi;
869 
870 	fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
871 	if (!fi)
872 		return NULL;
873 
874 	init_once((void *) fi);
875 
876 	/* Initialize f2fs-specific inode info */
877 	atomic_set(&fi->dirty_pages, 0);
878 	init_rwsem(&fi->i_sem);
879 	INIT_LIST_HEAD(&fi->dirty_list);
880 	INIT_LIST_HEAD(&fi->gdirty_list);
881 	INIT_LIST_HEAD(&fi->inmem_ilist);
882 	INIT_LIST_HEAD(&fi->inmem_pages);
883 	mutex_init(&fi->inmem_lock);
884 	init_rwsem(&fi->i_gc_rwsem[READ]);
885 	init_rwsem(&fi->i_gc_rwsem[WRITE]);
886 	init_rwsem(&fi->i_mmap_sem);
887 	init_rwsem(&fi->i_xattr_sem);
888 
889 	/* Will be used by directory only */
890 	fi->i_dir_level = F2FS_SB(sb)->dir_level;
891 
892 	return &fi->vfs_inode;
893 }
894 
895 static int f2fs_drop_inode(struct inode *inode)
896 {
897 	int ret;
898 	/*
899 	 * This is to avoid a deadlock condition like below.
900 	 * writeback_single_inode(inode)
901 	 *  - f2fs_write_data_page
902 	 *    - f2fs_gc -> iput -> evict
903 	 *       - inode_wait_for_writeback(inode)
904 	 */
905 	if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
906 		if (!inode->i_nlink && !is_bad_inode(inode)) {
907 			/* to avoid evict_inode call simultaneously */
908 			atomic_inc(&inode->i_count);
909 			spin_unlock(&inode->i_lock);
910 
911 			/* some remained atomic pages should discarded */
912 			if (f2fs_is_atomic_file(inode))
913 				f2fs_drop_inmem_pages(inode);
914 
915 			/* should remain fi->extent_tree for writepage */
916 			f2fs_destroy_extent_node(inode);
917 
918 			sb_start_intwrite(inode->i_sb);
919 			f2fs_i_size_write(inode, 0);
920 
921 			f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
922 					inode, NULL, 0, DATA);
923 			truncate_inode_pages_final(inode->i_mapping);
924 
925 			if (F2FS_HAS_BLOCKS(inode))
926 				f2fs_truncate(inode);
927 
928 			sb_end_intwrite(inode->i_sb);
929 
930 			spin_lock(&inode->i_lock);
931 			atomic_dec(&inode->i_count);
932 		}
933 		trace_f2fs_drop_inode(inode, 0);
934 		return 0;
935 	}
936 	ret = generic_drop_inode(inode);
937 	trace_f2fs_drop_inode(inode, ret);
938 	return ret;
939 }
940 
941 int f2fs_inode_dirtied(struct inode *inode, bool sync)
942 {
943 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
944 	int ret = 0;
945 
946 	spin_lock(&sbi->inode_lock[DIRTY_META]);
947 	if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
948 		ret = 1;
949 	} else {
950 		set_inode_flag(inode, FI_DIRTY_INODE);
951 		stat_inc_dirty_inode(sbi, DIRTY_META);
952 	}
953 	if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
954 		list_add_tail(&F2FS_I(inode)->gdirty_list,
955 				&sbi->inode_list[DIRTY_META]);
956 		inc_page_count(sbi, F2FS_DIRTY_IMETA);
957 	}
958 	spin_unlock(&sbi->inode_lock[DIRTY_META]);
959 	return ret;
960 }
961 
962 void f2fs_inode_synced(struct inode *inode)
963 {
964 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
965 
966 	spin_lock(&sbi->inode_lock[DIRTY_META]);
967 	if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
968 		spin_unlock(&sbi->inode_lock[DIRTY_META]);
969 		return;
970 	}
971 	if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
972 		list_del_init(&F2FS_I(inode)->gdirty_list);
973 		dec_page_count(sbi, F2FS_DIRTY_IMETA);
974 	}
975 	clear_inode_flag(inode, FI_DIRTY_INODE);
976 	clear_inode_flag(inode, FI_AUTO_RECOVER);
977 	stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
978 	spin_unlock(&sbi->inode_lock[DIRTY_META]);
979 }
980 
981 /*
982  * f2fs_dirty_inode() is called from __mark_inode_dirty()
983  *
984  * We should call set_dirty_inode to write the dirty inode through write_inode.
985  */
986 static void f2fs_dirty_inode(struct inode *inode, int flags)
987 {
988 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
989 
990 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
991 			inode->i_ino == F2FS_META_INO(sbi))
992 		return;
993 
994 	if (flags == I_DIRTY_TIME)
995 		return;
996 
997 	if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
998 		clear_inode_flag(inode, FI_AUTO_RECOVER);
999 
1000 	f2fs_inode_dirtied(inode, false);
1001 }
1002 
1003 static void f2fs_free_inode(struct inode *inode)
1004 {
1005 	fscrypt_free_inode(inode);
1006 	kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1007 }
1008 
1009 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1010 {
1011 	percpu_counter_destroy(&sbi->alloc_valid_block_count);
1012 	percpu_counter_destroy(&sbi->total_valid_inode_count);
1013 }
1014 
1015 static void destroy_device_list(struct f2fs_sb_info *sbi)
1016 {
1017 	int i;
1018 
1019 	for (i = 0; i < sbi->s_ndevs; i++) {
1020 		blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1021 #ifdef CONFIG_BLK_DEV_ZONED
1022 		kvfree(FDEV(i).blkz_type);
1023 #endif
1024 	}
1025 	kvfree(sbi->devs);
1026 }
1027 
1028 static void f2fs_put_super(struct super_block *sb)
1029 {
1030 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1031 	int i;
1032 	bool dropped;
1033 
1034 	f2fs_quota_off_umount(sb);
1035 
1036 	/* prevent remaining shrinker jobs */
1037 	mutex_lock(&sbi->umount_mutex);
1038 
1039 	/*
1040 	 * We don't need to do checkpoint when superblock is clean.
1041 	 * But, the previous checkpoint was not done by umount, it needs to do
1042 	 * clean checkpoint again.
1043 	 */
1044 	if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1045 			!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1046 		struct cp_control cpc = {
1047 			.reason = CP_UMOUNT,
1048 		};
1049 		f2fs_write_checkpoint(sbi, &cpc);
1050 	}
1051 
1052 	/* be sure to wait for any on-going discard commands */
1053 	dropped = f2fs_issue_discard_timeout(sbi);
1054 
1055 	if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1056 					!sbi->discard_blks && !dropped) {
1057 		struct cp_control cpc = {
1058 			.reason = CP_UMOUNT | CP_TRIMMED,
1059 		};
1060 		f2fs_write_checkpoint(sbi, &cpc);
1061 	}
1062 
1063 	/*
1064 	 * normally superblock is clean, so we need to release this.
1065 	 * In addition, EIO will skip do checkpoint, we need this as well.
1066 	 */
1067 	f2fs_release_ino_entry(sbi, true);
1068 
1069 	f2fs_leave_shrinker(sbi);
1070 	mutex_unlock(&sbi->umount_mutex);
1071 
1072 	/* our cp_error case, we can wait for any writeback page */
1073 	f2fs_flush_merged_writes(sbi);
1074 
1075 	f2fs_wait_on_all_pages_writeback(sbi);
1076 
1077 	f2fs_bug_on(sbi, sbi->fsync_node_num);
1078 
1079 	iput(sbi->node_inode);
1080 	sbi->node_inode = NULL;
1081 
1082 	iput(sbi->meta_inode);
1083 	sbi->meta_inode = NULL;
1084 
1085 	/*
1086 	 * iput() can update stat information, if f2fs_write_checkpoint()
1087 	 * above failed with error.
1088 	 */
1089 	f2fs_destroy_stats(sbi);
1090 
1091 	/* destroy f2fs internal modules */
1092 	f2fs_destroy_node_manager(sbi);
1093 	f2fs_destroy_segment_manager(sbi);
1094 
1095 	kvfree(sbi->ckpt);
1096 
1097 	f2fs_unregister_sysfs(sbi);
1098 
1099 	sb->s_fs_info = NULL;
1100 	if (sbi->s_chksum_driver)
1101 		crypto_free_shash(sbi->s_chksum_driver);
1102 	kvfree(sbi->raw_super);
1103 
1104 	destroy_device_list(sbi);
1105 	mempool_destroy(sbi->write_io_dummy);
1106 #ifdef CONFIG_QUOTA
1107 	for (i = 0; i < MAXQUOTAS; i++)
1108 		kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1109 #endif
1110 	destroy_percpu_info(sbi);
1111 	for (i = 0; i < NR_PAGE_TYPE; i++)
1112 		kvfree(sbi->write_io[i]);
1113 	kvfree(sbi);
1114 }
1115 
1116 int f2fs_sync_fs(struct super_block *sb, int sync)
1117 {
1118 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1119 	int err = 0;
1120 
1121 	if (unlikely(f2fs_cp_error(sbi)))
1122 		return 0;
1123 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1124 		return 0;
1125 
1126 	trace_f2fs_sync_fs(sb, sync);
1127 
1128 	if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1129 		return -EAGAIN;
1130 
1131 	if (sync) {
1132 		struct cp_control cpc;
1133 
1134 		cpc.reason = __get_cp_reason(sbi);
1135 
1136 		mutex_lock(&sbi->gc_mutex);
1137 		err = f2fs_write_checkpoint(sbi, &cpc);
1138 		mutex_unlock(&sbi->gc_mutex);
1139 	}
1140 	f2fs_trace_ios(NULL, 1);
1141 
1142 	return err;
1143 }
1144 
1145 static int f2fs_freeze(struct super_block *sb)
1146 {
1147 	if (f2fs_readonly(sb))
1148 		return 0;
1149 
1150 	/* IO error happened before */
1151 	if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1152 		return -EIO;
1153 
1154 	/* must be clean, since sync_filesystem() was already called */
1155 	if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1156 		return -EINVAL;
1157 	return 0;
1158 }
1159 
1160 static int f2fs_unfreeze(struct super_block *sb)
1161 {
1162 	return 0;
1163 }
1164 
1165 #ifdef CONFIG_QUOTA
1166 static int f2fs_statfs_project(struct super_block *sb,
1167 				kprojid_t projid, struct kstatfs *buf)
1168 {
1169 	struct kqid qid;
1170 	struct dquot *dquot;
1171 	u64 limit;
1172 	u64 curblock;
1173 
1174 	qid = make_kqid_projid(projid);
1175 	dquot = dqget(sb, qid);
1176 	if (IS_ERR(dquot))
1177 		return PTR_ERR(dquot);
1178 	spin_lock(&dquot->dq_dqb_lock);
1179 
1180 	limit = (dquot->dq_dqb.dqb_bsoftlimit ?
1181 		 dquot->dq_dqb.dqb_bsoftlimit :
1182 		 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
1183 	if (limit && buf->f_blocks > limit) {
1184 		curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1185 		buf->f_blocks = limit;
1186 		buf->f_bfree = buf->f_bavail =
1187 			(buf->f_blocks > curblock) ?
1188 			 (buf->f_blocks - curblock) : 0;
1189 	}
1190 
1191 	limit = dquot->dq_dqb.dqb_isoftlimit ?
1192 		dquot->dq_dqb.dqb_isoftlimit :
1193 		dquot->dq_dqb.dqb_ihardlimit;
1194 	if (limit && buf->f_files > limit) {
1195 		buf->f_files = limit;
1196 		buf->f_ffree =
1197 			(buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1198 			 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1199 	}
1200 
1201 	spin_unlock(&dquot->dq_dqb_lock);
1202 	dqput(dquot);
1203 	return 0;
1204 }
1205 #endif
1206 
1207 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1208 {
1209 	struct super_block *sb = dentry->d_sb;
1210 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1211 	u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1212 	block_t total_count, user_block_count, start_count;
1213 	u64 avail_node_count;
1214 
1215 	total_count = le64_to_cpu(sbi->raw_super->block_count);
1216 	user_block_count = sbi->user_block_count;
1217 	start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1218 	buf->f_type = F2FS_SUPER_MAGIC;
1219 	buf->f_bsize = sbi->blocksize;
1220 
1221 	buf->f_blocks = total_count - start_count;
1222 	buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1223 						sbi->current_reserved_blocks;
1224 	if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1225 		buf->f_bfree = 0;
1226 	else
1227 		buf->f_bfree -= sbi->unusable_block_count;
1228 
1229 	if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1230 		buf->f_bavail = buf->f_bfree -
1231 				F2FS_OPTION(sbi).root_reserved_blocks;
1232 	else
1233 		buf->f_bavail = 0;
1234 
1235 	avail_node_count = sbi->total_node_count - sbi->nquota_files -
1236 						F2FS_RESERVED_NODE_NUM;
1237 
1238 	if (avail_node_count > user_block_count) {
1239 		buf->f_files = user_block_count;
1240 		buf->f_ffree = buf->f_bavail;
1241 	} else {
1242 		buf->f_files = avail_node_count;
1243 		buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1244 					buf->f_bavail);
1245 	}
1246 
1247 	buf->f_namelen = F2FS_NAME_LEN;
1248 	buf->f_fsid.val[0] = (u32)id;
1249 	buf->f_fsid.val[1] = (u32)(id >> 32);
1250 
1251 #ifdef CONFIG_QUOTA
1252 	if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1253 			sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1254 		f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1255 	}
1256 #endif
1257 	return 0;
1258 }
1259 
1260 static inline void f2fs_show_quota_options(struct seq_file *seq,
1261 					   struct super_block *sb)
1262 {
1263 #ifdef CONFIG_QUOTA
1264 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1265 
1266 	if (F2FS_OPTION(sbi).s_jquota_fmt) {
1267 		char *fmtname = "";
1268 
1269 		switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1270 		case QFMT_VFS_OLD:
1271 			fmtname = "vfsold";
1272 			break;
1273 		case QFMT_VFS_V0:
1274 			fmtname = "vfsv0";
1275 			break;
1276 		case QFMT_VFS_V1:
1277 			fmtname = "vfsv1";
1278 			break;
1279 		}
1280 		seq_printf(seq, ",jqfmt=%s", fmtname);
1281 	}
1282 
1283 	if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1284 		seq_show_option(seq, "usrjquota",
1285 			F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1286 
1287 	if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1288 		seq_show_option(seq, "grpjquota",
1289 			F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1290 
1291 	if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1292 		seq_show_option(seq, "prjjquota",
1293 			F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1294 #endif
1295 }
1296 
1297 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1298 {
1299 	struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1300 
1301 	if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1302 		if (test_opt(sbi, FORCE_FG_GC))
1303 			seq_printf(seq, ",background_gc=%s", "sync");
1304 		else
1305 			seq_printf(seq, ",background_gc=%s", "on");
1306 	} else {
1307 		seq_printf(seq, ",background_gc=%s", "off");
1308 	}
1309 	if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1310 		seq_puts(seq, ",disable_roll_forward");
1311 	if (test_opt(sbi, DISCARD))
1312 		seq_puts(seq, ",discard");
1313 	if (test_opt(sbi, NOHEAP))
1314 		seq_puts(seq, ",no_heap");
1315 	else
1316 		seq_puts(seq, ",heap");
1317 #ifdef CONFIG_F2FS_FS_XATTR
1318 	if (test_opt(sbi, XATTR_USER))
1319 		seq_puts(seq, ",user_xattr");
1320 	else
1321 		seq_puts(seq, ",nouser_xattr");
1322 	if (test_opt(sbi, INLINE_XATTR))
1323 		seq_puts(seq, ",inline_xattr");
1324 	else
1325 		seq_puts(seq, ",noinline_xattr");
1326 	if (test_opt(sbi, INLINE_XATTR_SIZE))
1327 		seq_printf(seq, ",inline_xattr_size=%u",
1328 					F2FS_OPTION(sbi).inline_xattr_size);
1329 #endif
1330 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1331 	if (test_opt(sbi, POSIX_ACL))
1332 		seq_puts(seq, ",acl");
1333 	else
1334 		seq_puts(seq, ",noacl");
1335 #endif
1336 	if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1337 		seq_puts(seq, ",disable_ext_identify");
1338 	if (test_opt(sbi, INLINE_DATA))
1339 		seq_puts(seq, ",inline_data");
1340 	else
1341 		seq_puts(seq, ",noinline_data");
1342 	if (test_opt(sbi, INLINE_DENTRY))
1343 		seq_puts(seq, ",inline_dentry");
1344 	else
1345 		seq_puts(seq, ",noinline_dentry");
1346 	if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1347 		seq_puts(seq, ",flush_merge");
1348 	if (test_opt(sbi, NOBARRIER))
1349 		seq_puts(seq, ",nobarrier");
1350 	if (test_opt(sbi, FASTBOOT))
1351 		seq_puts(seq, ",fastboot");
1352 	if (test_opt(sbi, EXTENT_CACHE))
1353 		seq_puts(seq, ",extent_cache");
1354 	else
1355 		seq_puts(seq, ",noextent_cache");
1356 	if (test_opt(sbi, DATA_FLUSH))
1357 		seq_puts(seq, ",data_flush");
1358 
1359 	seq_puts(seq, ",mode=");
1360 	if (test_opt(sbi, ADAPTIVE))
1361 		seq_puts(seq, "adaptive");
1362 	else if (test_opt(sbi, LFS))
1363 		seq_puts(seq, "lfs");
1364 	seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1365 	if (test_opt(sbi, RESERVE_ROOT))
1366 		seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1367 				F2FS_OPTION(sbi).root_reserved_blocks,
1368 				from_kuid_munged(&init_user_ns,
1369 					F2FS_OPTION(sbi).s_resuid),
1370 				from_kgid_munged(&init_user_ns,
1371 					F2FS_OPTION(sbi).s_resgid));
1372 	if (F2FS_IO_SIZE_BITS(sbi))
1373 		seq_printf(seq, ",io_bits=%u",
1374 				F2FS_OPTION(sbi).write_io_size_bits);
1375 #ifdef CONFIG_F2FS_FAULT_INJECTION
1376 	if (test_opt(sbi, FAULT_INJECTION)) {
1377 		seq_printf(seq, ",fault_injection=%u",
1378 				F2FS_OPTION(sbi).fault_info.inject_rate);
1379 		seq_printf(seq, ",fault_type=%u",
1380 				F2FS_OPTION(sbi).fault_info.inject_type);
1381 	}
1382 #endif
1383 #ifdef CONFIG_QUOTA
1384 	if (test_opt(sbi, QUOTA))
1385 		seq_puts(seq, ",quota");
1386 	if (test_opt(sbi, USRQUOTA))
1387 		seq_puts(seq, ",usrquota");
1388 	if (test_opt(sbi, GRPQUOTA))
1389 		seq_puts(seq, ",grpquota");
1390 	if (test_opt(sbi, PRJQUOTA))
1391 		seq_puts(seq, ",prjquota");
1392 #endif
1393 	f2fs_show_quota_options(seq, sbi->sb);
1394 	if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1395 		seq_printf(seq, ",whint_mode=%s", "user-based");
1396 	else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1397 		seq_printf(seq, ",whint_mode=%s", "fs-based");
1398 #ifdef CONFIG_FS_ENCRYPTION
1399 	if (F2FS_OPTION(sbi).test_dummy_encryption)
1400 		seq_puts(seq, ",test_dummy_encryption");
1401 #endif
1402 
1403 	if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1404 		seq_printf(seq, ",alloc_mode=%s", "default");
1405 	else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1406 		seq_printf(seq, ",alloc_mode=%s", "reuse");
1407 
1408 	if (test_opt(sbi, DISABLE_CHECKPOINT))
1409 		seq_puts(seq, ",checkpoint=disable");
1410 
1411 	if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1412 		seq_printf(seq, ",fsync_mode=%s", "posix");
1413 	else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1414 		seq_printf(seq, ",fsync_mode=%s", "strict");
1415 	else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1416 		seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1417 	return 0;
1418 }
1419 
1420 static void default_options(struct f2fs_sb_info *sbi)
1421 {
1422 	/* init some FS parameters */
1423 	F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1424 	F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1425 	F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1426 	F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1427 	F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1428 	F2FS_OPTION(sbi).test_dummy_encryption = false;
1429 	F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1430 	F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1431 
1432 	set_opt(sbi, BG_GC);
1433 	set_opt(sbi, INLINE_XATTR);
1434 	set_opt(sbi, INLINE_DATA);
1435 	set_opt(sbi, INLINE_DENTRY);
1436 	set_opt(sbi, EXTENT_CACHE);
1437 	set_opt(sbi, NOHEAP);
1438 	clear_opt(sbi, DISABLE_CHECKPOINT);
1439 	sbi->sb->s_flags |= SB_LAZYTIME;
1440 	set_opt(sbi, FLUSH_MERGE);
1441 	set_opt(sbi, DISCARD);
1442 	if (f2fs_sb_has_blkzoned(sbi))
1443 		set_opt_mode(sbi, F2FS_MOUNT_LFS);
1444 	else
1445 		set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1446 
1447 #ifdef CONFIG_F2FS_FS_XATTR
1448 	set_opt(sbi, XATTR_USER);
1449 #endif
1450 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1451 	set_opt(sbi, POSIX_ACL);
1452 #endif
1453 
1454 	f2fs_build_fault_attr(sbi, 0, 0);
1455 }
1456 
1457 #ifdef CONFIG_QUOTA
1458 static int f2fs_enable_quotas(struct super_block *sb);
1459 #endif
1460 
1461 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1462 {
1463 	unsigned int s_flags = sbi->sb->s_flags;
1464 	struct cp_control cpc;
1465 	int err = 0;
1466 	int ret;
1467 
1468 	if (s_flags & SB_RDONLY) {
1469 		f2fs_msg(sbi->sb, KERN_ERR,
1470 				"checkpoint=disable on readonly fs");
1471 		return -EINVAL;
1472 	}
1473 	sbi->sb->s_flags |= SB_ACTIVE;
1474 
1475 	f2fs_update_time(sbi, DISABLE_TIME);
1476 
1477 	while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1478 		mutex_lock(&sbi->gc_mutex);
1479 		err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1480 		if (err == -ENODATA) {
1481 			err = 0;
1482 			break;
1483 		}
1484 		if (err && err != -EAGAIN)
1485 			break;
1486 	}
1487 
1488 	ret = sync_filesystem(sbi->sb);
1489 	if (ret || err) {
1490 		err = ret ? ret: err;
1491 		goto restore_flag;
1492 	}
1493 
1494 	if (f2fs_disable_cp_again(sbi)) {
1495 		err = -EAGAIN;
1496 		goto restore_flag;
1497 	}
1498 
1499 	mutex_lock(&sbi->gc_mutex);
1500 	cpc.reason = CP_PAUSE;
1501 	set_sbi_flag(sbi, SBI_CP_DISABLED);
1502 	f2fs_write_checkpoint(sbi, &cpc);
1503 
1504 	sbi->unusable_block_count = 0;
1505 	mutex_unlock(&sbi->gc_mutex);
1506 restore_flag:
1507 	sbi->sb->s_flags = s_flags;	/* Restore MS_RDONLY status */
1508 	return err;
1509 }
1510 
1511 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1512 {
1513 	mutex_lock(&sbi->gc_mutex);
1514 	f2fs_dirty_to_prefree(sbi);
1515 
1516 	clear_sbi_flag(sbi, SBI_CP_DISABLED);
1517 	set_sbi_flag(sbi, SBI_IS_DIRTY);
1518 	mutex_unlock(&sbi->gc_mutex);
1519 
1520 	f2fs_sync_fs(sbi->sb, 1);
1521 }
1522 
1523 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1524 {
1525 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1526 	struct f2fs_mount_info org_mount_opt;
1527 	unsigned long old_sb_flags;
1528 	int err;
1529 	bool need_restart_gc = false;
1530 	bool need_stop_gc = false;
1531 	bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1532 	bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1533 	bool checkpoint_changed;
1534 #ifdef CONFIG_QUOTA
1535 	int i, j;
1536 #endif
1537 
1538 	/*
1539 	 * Save the old mount options in case we
1540 	 * need to restore them.
1541 	 */
1542 	org_mount_opt = sbi->mount_opt;
1543 	old_sb_flags = sb->s_flags;
1544 
1545 #ifdef CONFIG_QUOTA
1546 	org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1547 	for (i = 0; i < MAXQUOTAS; i++) {
1548 		if (F2FS_OPTION(sbi).s_qf_names[i]) {
1549 			org_mount_opt.s_qf_names[i] =
1550 				kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1551 				GFP_KERNEL);
1552 			if (!org_mount_opt.s_qf_names[i]) {
1553 				for (j = 0; j < i; j++)
1554 					kvfree(org_mount_opt.s_qf_names[j]);
1555 				return -ENOMEM;
1556 			}
1557 		} else {
1558 			org_mount_opt.s_qf_names[i] = NULL;
1559 		}
1560 	}
1561 #endif
1562 
1563 	/* recover superblocks we couldn't write due to previous RO mount */
1564 	if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1565 		err = f2fs_commit_super(sbi, false);
1566 		f2fs_msg(sb, KERN_INFO,
1567 			"Try to recover all the superblocks, ret: %d", err);
1568 		if (!err)
1569 			clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1570 	}
1571 
1572 	default_options(sbi);
1573 
1574 	/* parse mount options */
1575 	err = parse_options(sb, data);
1576 	if (err)
1577 		goto restore_opts;
1578 	checkpoint_changed =
1579 			disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1580 
1581 	/*
1582 	 * Previous and new state of filesystem is RO,
1583 	 * so skip checking GC and FLUSH_MERGE conditions.
1584 	 */
1585 	if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1586 		goto skip;
1587 
1588 #ifdef CONFIG_QUOTA
1589 	if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1590 		err = dquot_suspend(sb, -1);
1591 		if (err < 0)
1592 			goto restore_opts;
1593 	} else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1594 		/* dquot_resume needs RW */
1595 		sb->s_flags &= ~SB_RDONLY;
1596 		if (sb_any_quota_suspended(sb)) {
1597 			dquot_resume(sb, -1);
1598 		} else if (f2fs_sb_has_quota_ino(sbi)) {
1599 			err = f2fs_enable_quotas(sb);
1600 			if (err)
1601 				goto restore_opts;
1602 		}
1603 	}
1604 #endif
1605 	/* disallow enable/disable extent_cache dynamically */
1606 	if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1607 		err = -EINVAL;
1608 		f2fs_msg(sbi->sb, KERN_WARNING,
1609 				"switch extent_cache option is not allowed");
1610 		goto restore_opts;
1611 	}
1612 
1613 	if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1614 		err = -EINVAL;
1615 		f2fs_msg(sbi->sb, KERN_WARNING,
1616 			"disabling checkpoint not compatible with read-only");
1617 		goto restore_opts;
1618 	}
1619 
1620 	/*
1621 	 * We stop the GC thread if FS is mounted as RO
1622 	 * or if background_gc = off is passed in mount
1623 	 * option. Also sync the filesystem.
1624 	 */
1625 	if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) {
1626 		if (sbi->gc_thread) {
1627 			f2fs_stop_gc_thread(sbi);
1628 			need_restart_gc = true;
1629 		}
1630 	} else if (!sbi->gc_thread) {
1631 		err = f2fs_start_gc_thread(sbi);
1632 		if (err)
1633 			goto restore_opts;
1634 		need_stop_gc = true;
1635 	}
1636 
1637 	if (*flags & SB_RDONLY ||
1638 		F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1639 		writeback_inodes_sb(sb, WB_REASON_SYNC);
1640 		sync_inodes_sb(sb);
1641 
1642 		set_sbi_flag(sbi, SBI_IS_DIRTY);
1643 		set_sbi_flag(sbi, SBI_IS_CLOSE);
1644 		f2fs_sync_fs(sb, 1);
1645 		clear_sbi_flag(sbi, SBI_IS_CLOSE);
1646 	}
1647 
1648 	if (checkpoint_changed) {
1649 		if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1650 			err = f2fs_disable_checkpoint(sbi);
1651 			if (err)
1652 				goto restore_gc;
1653 		} else {
1654 			f2fs_enable_checkpoint(sbi);
1655 		}
1656 	}
1657 
1658 	/*
1659 	 * We stop issue flush thread if FS is mounted as RO
1660 	 * or if flush_merge is not passed in mount option.
1661 	 */
1662 	if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1663 		clear_opt(sbi, FLUSH_MERGE);
1664 		f2fs_destroy_flush_cmd_control(sbi, false);
1665 	} else {
1666 		err = f2fs_create_flush_cmd_control(sbi);
1667 		if (err)
1668 			goto restore_gc;
1669 	}
1670 skip:
1671 #ifdef CONFIG_QUOTA
1672 	/* Release old quota file names */
1673 	for (i = 0; i < MAXQUOTAS; i++)
1674 		kvfree(org_mount_opt.s_qf_names[i]);
1675 #endif
1676 	/* Update the POSIXACL Flag */
1677 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1678 		(test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1679 
1680 	limit_reserve_root(sbi);
1681 	*flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1682 	return 0;
1683 restore_gc:
1684 	if (need_restart_gc) {
1685 		if (f2fs_start_gc_thread(sbi))
1686 			f2fs_msg(sbi->sb, KERN_WARNING,
1687 				"background gc thread has stopped");
1688 	} else if (need_stop_gc) {
1689 		f2fs_stop_gc_thread(sbi);
1690 	}
1691 restore_opts:
1692 #ifdef CONFIG_QUOTA
1693 	F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1694 	for (i = 0; i < MAXQUOTAS; i++) {
1695 		kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1696 		F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1697 	}
1698 #endif
1699 	sbi->mount_opt = org_mount_opt;
1700 	sb->s_flags = old_sb_flags;
1701 	return err;
1702 }
1703 
1704 #ifdef CONFIG_QUOTA
1705 /* Read data from quotafile */
1706 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1707 			       size_t len, loff_t off)
1708 {
1709 	struct inode *inode = sb_dqopt(sb)->files[type];
1710 	struct address_space *mapping = inode->i_mapping;
1711 	block_t blkidx = F2FS_BYTES_TO_BLK(off);
1712 	int offset = off & (sb->s_blocksize - 1);
1713 	int tocopy;
1714 	size_t toread;
1715 	loff_t i_size = i_size_read(inode);
1716 	struct page *page;
1717 	char *kaddr;
1718 
1719 	if (off > i_size)
1720 		return 0;
1721 
1722 	if (off + len > i_size)
1723 		len = i_size - off;
1724 	toread = len;
1725 	while (toread > 0) {
1726 		tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1727 repeat:
1728 		page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1729 		if (IS_ERR(page)) {
1730 			if (PTR_ERR(page) == -ENOMEM) {
1731 				congestion_wait(BLK_RW_ASYNC, HZ/50);
1732 				goto repeat;
1733 			}
1734 			set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1735 			return PTR_ERR(page);
1736 		}
1737 
1738 		lock_page(page);
1739 
1740 		if (unlikely(page->mapping != mapping)) {
1741 			f2fs_put_page(page, 1);
1742 			goto repeat;
1743 		}
1744 		if (unlikely(!PageUptodate(page))) {
1745 			f2fs_put_page(page, 1);
1746 			set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1747 			return -EIO;
1748 		}
1749 
1750 		kaddr = kmap_atomic(page);
1751 		memcpy(data, kaddr + offset, tocopy);
1752 		kunmap_atomic(kaddr);
1753 		f2fs_put_page(page, 1);
1754 
1755 		offset = 0;
1756 		toread -= tocopy;
1757 		data += tocopy;
1758 		blkidx++;
1759 	}
1760 	return len;
1761 }
1762 
1763 /* Write to quotafile */
1764 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1765 				const char *data, size_t len, loff_t off)
1766 {
1767 	struct inode *inode = sb_dqopt(sb)->files[type];
1768 	struct address_space *mapping = inode->i_mapping;
1769 	const struct address_space_operations *a_ops = mapping->a_ops;
1770 	int offset = off & (sb->s_blocksize - 1);
1771 	size_t towrite = len;
1772 	struct page *page;
1773 	char *kaddr;
1774 	int err = 0;
1775 	int tocopy;
1776 
1777 	while (towrite > 0) {
1778 		tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1779 								towrite);
1780 retry:
1781 		err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1782 							&page, NULL);
1783 		if (unlikely(err)) {
1784 			if (err == -ENOMEM) {
1785 				congestion_wait(BLK_RW_ASYNC, HZ/50);
1786 				goto retry;
1787 			}
1788 			set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1789 			break;
1790 		}
1791 
1792 		kaddr = kmap_atomic(page);
1793 		memcpy(kaddr + offset, data, tocopy);
1794 		kunmap_atomic(kaddr);
1795 		flush_dcache_page(page);
1796 
1797 		a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1798 						page, NULL);
1799 		offset = 0;
1800 		towrite -= tocopy;
1801 		off += tocopy;
1802 		data += tocopy;
1803 		cond_resched();
1804 	}
1805 
1806 	if (len == towrite)
1807 		return err;
1808 	inode->i_mtime = inode->i_ctime = current_time(inode);
1809 	f2fs_mark_inode_dirty_sync(inode, false);
1810 	return len - towrite;
1811 }
1812 
1813 static struct dquot **f2fs_get_dquots(struct inode *inode)
1814 {
1815 	return F2FS_I(inode)->i_dquot;
1816 }
1817 
1818 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1819 {
1820 	return &F2FS_I(inode)->i_reserved_quota;
1821 }
1822 
1823 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1824 {
1825 	if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
1826 		f2fs_msg(sbi->sb, KERN_ERR,
1827 			"quota sysfile may be corrupted, skip loading it");
1828 		return 0;
1829 	}
1830 
1831 	return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
1832 					F2FS_OPTION(sbi).s_jquota_fmt, type);
1833 }
1834 
1835 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1836 {
1837 	int enabled = 0;
1838 	int i, err;
1839 
1840 	if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
1841 		err = f2fs_enable_quotas(sbi->sb);
1842 		if (err) {
1843 			f2fs_msg(sbi->sb, KERN_ERR,
1844 					"Cannot turn on quota_ino: %d", err);
1845 			return 0;
1846 		}
1847 		return 1;
1848 	}
1849 
1850 	for (i = 0; i < MAXQUOTAS; i++) {
1851 		if (F2FS_OPTION(sbi).s_qf_names[i]) {
1852 			err = f2fs_quota_on_mount(sbi, i);
1853 			if (!err) {
1854 				enabled = 1;
1855 				continue;
1856 			}
1857 			f2fs_msg(sbi->sb, KERN_ERR,
1858 				"Cannot turn on quotas: %d on %d", err, i);
1859 		}
1860 	}
1861 	return enabled;
1862 }
1863 
1864 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
1865 			     unsigned int flags)
1866 {
1867 	struct inode *qf_inode;
1868 	unsigned long qf_inum;
1869 	int err;
1870 
1871 	BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
1872 
1873 	qf_inum = f2fs_qf_ino(sb, type);
1874 	if (!qf_inum)
1875 		return -EPERM;
1876 
1877 	qf_inode = f2fs_iget(sb, qf_inum);
1878 	if (IS_ERR(qf_inode)) {
1879 		f2fs_msg(sb, KERN_ERR,
1880 			"Bad quota inode %u:%lu", type, qf_inum);
1881 		return PTR_ERR(qf_inode);
1882 	}
1883 
1884 	/* Don't account quota for quota files to avoid recursion */
1885 	qf_inode->i_flags |= S_NOQUOTA;
1886 	err = dquot_enable(qf_inode, type, format_id, flags);
1887 	iput(qf_inode);
1888 	return err;
1889 }
1890 
1891 static int f2fs_enable_quotas(struct super_block *sb)
1892 {
1893 	int type, err = 0;
1894 	unsigned long qf_inum;
1895 	bool quota_mopt[MAXQUOTAS] = {
1896 		test_opt(F2FS_SB(sb), USRQUOTA),
1897 		test_opt(F2FS_SB(sb), GRPQUOTA),
1898 		test_opt(F2FS_SB(sb), PRJQUOTA),
1899 	};
1900 
1901 	if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
1902 		f2fs_msg(sb, KERN_ERR,
1903 			"quota file may be corrupted, skip loading it");
1904 		return 0;
1905 	}
1906 
1907 	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
1908 
1909 	for (type = 0; type < MAXQUOTAS; type++) {
1910 		qf_inum = f2fs_qf_ino(sb, type);
1911 		if (qf_inum) {
1912 			err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
1913 				DQUOT_USAGE_ENABLED |
1914 				(quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
1915 			if (err) {
1916 				f2fs_msg(sb, KERN_ERR,
1917 					"Failed to enable quota tracking "
1918 					"(type=%d, err=%d). Please run "
1919 					"fsck to fix.", type, err);
1920 				for (type--; type >= 0; type--)
1921 					dquot_quota_off(sb, type);
1922 				set_sbi_flag(F2FS_SB(sb),
1923 						SBI_QUOTA_NEED_REPAIR);
1924 				return err;
1925 			}
1926 		}
1927 	}
1928 	return 0;
1929 }
1930 
1931 int f2fs_quota_sync(struct super_block *sb, int type)
1932 {
1933 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1934 	struct quota_info *dqopt = sb_dqopt(sb);
1935 	int cnt;
1936 	int ret;
1937 
1938 	ret = dquot_writeback_dquots(sb, type);
1939 	if (ret)
1940 		goto out;
1941 
1942 	/*
1943 	 * Now when everything is written we can discard the pagecache so
1944 	 * that userspace sees the changes.
1945 	 */
1946 	for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
1947 		struct address_space *mapping;
1948 
1949 		if (type != -1 && cnt != type)
1950 			continue;
1951 		if (!sb_has_quota_active(sb, cnt))
1952 			continue;
1953 
1954 		mapping = dqopt->files[cnt]->i_mapping;
1955 
1956 		ret = filemap_fdatawrite(mapping);
1957 		if (ret)
1958 			goto out;
1959 
1960 		/* if we are using journalled quota */
1961 		if (is_journalled_quota(sbi))
1962 			continue;
1963 
1964 		ret = filemap_fdatawait(mapping);
1965 		if (ret)
1966 			set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1967 
1968 		inode_lock(dqopt->files[cnt]);
1969 		truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
1970 		inode_unlock(dqopt->files[cnt]);
1971 	}
1972 out:
1973 	if (ret)
1974 		set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1975 	return ret;
1976 }
1977 
1978 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
1979 							const struct path *path)
1980 {
1981 	struct inode *inode;
1982 	int err;
1983 
1984 	err = f2fs_quota_sync(sb, type);
1985 	if (err)
1986 		return err;
1987 
1988 	err = dquot_quota_on(sb, type, format_id, path);
1989 	if (err)
1990 		return err;
1991 
1992 	inode = d_inode(path->dentry);
1993 
1994 	inode_lock(inode);
1995 	F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
1996 	f2fs_set_inode_flags(inode);
1997 	inode_unlock(inode);
1998 	f2fs_mark_inode_dirty_sync(inode, false);
1999 
2000 	return 0;
2001 }
2002 
2003 static int f2fs_quota_off(struct super_block *sb, int type)
2004 {
2005 	struct inode *inode = sb_dqopt(sb)->files[type];
2006 	int err;
2007 
2008 	if (!inode || !igrab(inode))
2009 		return dquot_quota_off(sb, type);
2010 
2011 	err = f2fs_quota_sync(sb, type);
2012 	if (err)
2013 		goto out_put;
2014 
2015 	err = dquot_quota_off(sb, type);
2016 	if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2017 		goto out_put;
2018 
2019 	inode_lock(inode);
2020 	F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2021 	f2fs_set_inode_flags(inode);
2022 	inode_unlock(inode);
2023 	f2fs_mark_inode_dirty_sync(inode, false);
2024 out_put:
2025 	iput(inode);
2026 	return err;
2027 }
2028 
2029 void f2fs_quota_off_umount(struct super_block *sb)
2030 {
2031 	int type;
2032 	int err;
2033 
2034 	for (type = 0; type < MAXQUOTAS; type++) {
2035 		err = f2fs_quota_off(sb, type);
2036 		if (err) {
2037 			int ret = dquot_quota_off(sb, type);
2038 
2039 			f2fs_msg(sb, KERN_ERR,
2040 				"Fail to turn off disk quota "
2041 				"(type: %d, err: %d, ret:%d), Please "
2042 				"run fsck to fix it.", type, err, ret);
2043 			set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2044 		}
2045 	}
2046 	/*
2047 	 * In case of checkpoint=disable, we must flush quota blocks.
2048 	 * This can cause NULL exception for node_inode in end_io, since
2049 	 * put_super already dropped it.
2050 	 */
2051 	sync_filesystem(sb);
2052 }
2053 
2054 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2055 {
2056 	struct quota_info *dqopt = sb_dqopt(sb);
2057 	int type;
2058 
2059 	for (type = 0; type < MAXQUOTAS; type++) {
2060 		if (!dqopt->files[type])
2061 			continue;
2062 		f2fs_inode_synced(dqopt->files[type]);
2063 	}
2064 }
2065 
2066 static int f2fs_dquot_commit(struct dquot *dquot)
2067 {
2068 	int ret;
2069 
2070 	ret = dquot_commit(dquot);
2071 	if (ret < 0)
2072 		set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2073 	return ret;
2074 }
2075 
2076 static int f2fs_dquot_acquire(struct dquot *dquot)
2077 {
2078 	int ret;
2079 
2080 	ret = dquot_acquire(dquot);
2081 	if (ret < 0)
2082 		set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2083 
2084 	return ret;
2085 }
2086 
2087 static int f2fs_dquot_release(struct dquot *dquot)
2088 {
2089 	int ret;
2090 
2091 	ret = dquot_release(dquot);
2092 	if (ret < 0)
2093 		set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2094 	return ret;
2095 }
2096 
2097 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2098 {
2099 	struct super_block *sb = dquot->dq_sb;
2100 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2101 	int ret;
2102 
2103 	ret = dquot_mark_dquot_dirty(dquot);
2104 
2105 	/* if we are using journalled quota */
2106 	if (is_journalled_quota(sbi))
2107 		set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2108 
2109 	return ret;
2110 }
2111 
2112 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2113 {
2114 	int ret;
2115 
2116 	ret = dquot_commit_info(sb, type);
2117 	if (ret < 0)
2118 		set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2119 	return ret;
2120 }
2121 
2122 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2123 {
2124 	*projid = F2FS_I(inode)->i_projid;
2125 	return 0;
2126 }
2127 
2128 static const struct dquot_operations f2fs_quota_operations = {
2129 	.get_reserved_space = f2fs_get_reserved_space,
2130 	.write_dquot	= f2fs_dquot_commit,
2131 	.acquire_dquot	= f2fs_dquot_acquire,
2132 	.release_dquot	= f2fs_dquot_release,
2133 	.mark_dirty	= f2fs_dquot_mark_dquot_dirty,
2134 	.write_info	= f2fs_dquot_commit_info,
2135 	.alloc_dquot	= dquot_alloc,
2136 	.destroy_dquot	= dquot_destroy,
2137 	.get_projid	= f2fs_get_projid,
2138 	.get_next_id	= dquot_get_next_id,
2139 };
2140 
2141 static const struct quotactl_ops f2fs_quotactl_ops = {
2142 	.quota_on	= f2fs_quota_on,
2143 	.quota_off	= f2fs_quota_off,
2144 	.quota_sync	= f2fs_quota_sync,
2145 	.get_state	= dquot_get_state,
2146 	.set_info	= dquot_set_dqinfo,
2147 	.get_dqblk	= dquot_get_dqblk,
2148 	.set_dqblk	= dquot_set_dqblk,
2149 	.get_nextdqblk	= dquot_get_next_dqblk,
2150 };
2151 #else
2152 int f2fs_quota_sync(struct super_block *sb, int type)
2153 {
2154 	return 0;
2155 }
2156 
2157 void f2fs_quota_off_umount(struct super_block *sb)
2158 {
2159 }
2160 #endif
2161 
2162 static const struct super_operations f2fs_sops = {
2163 	.alloc_inode	= f2fs_alloc_inode,
2164 	.free_inode	= f2fs_free_inode,
2165 	.drop_inode	= f2fs_drop_inode,
2166 	.write_inode	= f2fs_write_inode,
2167 	.dirty_inode	= f2fs_dirty_inode,
2168 	.show_options	= f2fs_show_options,
2169 #ifdef CONFIG_QUOTA
2170 	.quota_read	= f2fs_quota_read,
2171 	.quota_write	= f2fs_quota_write,
2172 	.get_dquots	= f2fs_get_dquots,
2173 #endif
2174 	.evict_inode	= f2fs_evict_inode,
2175 	.put_super	= f2fs_put_super,
2176 	.sync_fs	= f2fs_sync_fs,
2177 	.freeze_fs	= f2fs_freeze,
2178 	.unfreeze_fs	= f2fs_unfreeze,
2179 	.statfs		= f2fs_statfs,
2180 	.remount_fs	= f2fs_remount,
2181 };
2182 
2183 #ifdef CONFIG_FS_ENCRYPTION
2184 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2185 {
2186 	return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2187 				F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2188 				ctx, len, NULL);
2189 }
2190 
2191 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2192 							void *fs_data)
2193 {
2194 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2195 
2196 	/*
2197 	 * Encrypting the root directory is not allowed because fsck
2198 	 * expects lost+found directory to exist and remain unencrypted
2199 	 * if LOST_FOUND feature is enabled.
2200 	 *
2201 	 */
2202 	if (f2fs_sb_has_lost_found(sbi) &&
2203 			inode->i_ino == F2FS_ROOT_INO(sbi))
2204 		return -EPERM;
2205 
2206 	return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2207 				F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2208 				ctx, len, fs_data, XATTR_CREATE);
2209 }
2210 
2211 static bool f2fs_dummy_context(struct inode *inode)
2212 {
2213 	return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
2214 }
2215 
2216 static const struct fscrypt_operations f2fs_cryptops = {
2217 	.key_prefix	= "f2fs:",
2218 	.get_context	= f2fs_get_context,
2219 	.set_context	= f2fs_set_context,
2220 	.dummy_context	= f2fs_dummy_context,
2221 	.empty_dir	= f2fs_empty_dir,
2222 	.max_namelen	= F2FS_NAME_LEN,
2223 };
2224 #endif
2225 
2226 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2227 		u64 ino, u32 generation)
2228 {
2229 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2230 	struct inode *inode;
2231 
2232 	if (f2fs_check_nid_range(sbi, ino))
2233 		return ERR_PTR(-ESTALE);
2234 
2235 	/*
2236 	 * f2fs_iget isn't quite right if the inode is currently unallocated!
2237 	 * However f2fs_iget currently does appropriate checks to handle stale
2238 	 * inodes so everything is OK.
2239 	 */
2240 	inode = f2fs_iget(sb, ino);
2241 	if (IS_ERR(inode))
2242 		return ERR_CAST(inode);
2243 	if (unlikely(generation && inode->i_generation != generation)) {
2244 		/* we didn't find the right inode.. */
2245 		iput(inode);
2246 		return ERR_PTR(-ESTALE);
2247 	}
2248 	return inode;
2249 }
2250 
2251 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2252 		int fh_len, int fh_type)
2253 {
2254 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2255 				    f2fs_nfs_get_inode);
2256 }
2257 
2258 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2259 		int fh_len, int fh_type)
2260 {
2261 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2262 				    f2fs_nfs_get_inode);
2263 }
2264 
2265 static const struct export_operations f2fs_export_ops = {
2266 	.fh_to_dentry = f2fs_fh_to_dentry,
2267 	.fh_to_parent = f2fs_fh_to_parent,
2268 	.get_parent = f2fs_get_parent,
2269 };
2270 
2271 static loff_t max_file_blocks(void)
2272 {
2273 	loff_t result = 0;
2274 	loff_t leaf_count = ADDRS_PER_BLOCK;
2275 
2276 	/*
2277 	 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2278 	 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2279 	 * space in inode.i_addr, it will be more safe to reassign
2280 	 * result as zero.
2281 	 */
2282 
2283 	/* two direct node blocks */
2284 	result += (leaf_count * 2);
2285 
2286 	/* two indirect node blocks */
2287 	leaf_count *= NIDS_PER_BLOCK;
2288 	result += (leaf_count * 2);
2289 
2290 	/* one double indirect node block */
2291 	leaf_count *= NIDS_PER_BLOCK;
2292 	result += leaf_count;
2293 
2294 	return result;
2295 }
2296 
2297 static int __f2fs_commit_super(struct buffer_head *bh,
2298 			struct f2fs_super_block *super)
2299 {
2300 	lock_buffer(bh);
2301 	if (super)
2302 		memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2303 	set_buffer_dirty(bh);
2304 	unlock_buffer(bh);
2305 
2306 	/* it's rare case, we can do fua all the time */
2307 	return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2308 }
2309 
2310 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2311 					struct buffer_head *bh)
2312 {
2313 	struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2314 					(bh->b_data + F2FS_SUPER_OFFSET);
2315 	struct super_block *sb = sbi->sb;
2316 	u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2317 	u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2318 	u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2319 	u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2320 	u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2321 	u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2322 	u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2323 	u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2324 	u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2325 	u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2326 	u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2327 	u32 segment_count = le32_to_cpu(raw_super->segment_count);
2328 	u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2329 	u64 main_end_blkaddr = main_blkaddr +
2330 				(segment_count_main << log_blocks_per_seg);
2331 	u64 seg_end_blkaddr = segment0_blkaddr +
2332 				(segment_count << log_blocks_per_seg);
2333 
2334 	if (segment0_blkaddr != cp_blkaddr) {
2335 		f2fs_msg(sb, KERN_INFO,
2336 			"Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2337 			segment0_blkaddr, cp_blkaddr);
2338 		return true;
2339 	}
2340 
2341 	if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2342 							sit_blkaddr) {
2343 		f2fs_msg(sb, KERN_INFO,
2344 			"Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2345 			cp_blkaddr, sit_blkaddr,
2346 			segment_count_ckpt << log_blocks_per_seg);
2347 		return true;
2348 	}
2349 
2350 	if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2351 							nat_blkaddr) {
2352 		f2fs_msg(sb, KERN_INFO,
2353 			"Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2354 			sit_blkaddr, nat_blkaddr,
2355 			segment_count_sit << log_blocks_per_seg);
2356 		return true;
2357 	}
2358 
2359 	if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2360 							ssa_blkaddr) {
2361 		f2fs_msg(sb, KERN_INFO,
2362 			"Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2363 			nat_blkaddr, ssa_blkaddr,
2364 			segment_count_nat << log_blocks_per_seg);
2365 		return true;
2366 	}
2367 
2368 	if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2369 							main_blkaddr) {
2370 		f2fs_msg(sb, KERN_INFO,
2371 			"Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2372 			ssa_blkaddr, main_blkaddr,
2373 			segment_count_ssa << log_blocks_per_seg);
2374 		return true;
2375 	}
2376 
2377 	if (main_end_blkaddr > seg_end_blkaddr) {
2378 		f2fs_msg(sb, KERN_INFO,
2379 			"Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2380 			main_blkaddr,
2381 			segment0_blkaddr +
2382 				(segment_count << log_blocks_per_seg),
2383 			segment_count_main << log_blocks_per_seg);
2384 		return true;
2385 	} else if (main_end_blkaddr < seg_end_blkaddr) {
2386 		int err = 0;
2387 		char *res;
2388 
2389 		/* fix in-memory information all the time */
2390 		raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2391 				segment0_blkaddr) >> log_blocks_per_seg);
2392 
2393 		if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2394 			set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2395 			res = "internally";
2396 		} else {
2397 			err = __f2fs_commit_super(bh, NULL);
2398 			res = err ? "failed" : "done";
2399 		}
2400 		f2fs_msg(sb, KERN_INFO,
2401 			"Fix alignment : %s, start(%u) end(%u) block(%u)",
2402 			res, main_blkaddr,
2403 			segment0_blkaddr +
2404 				(segment_count << log_blocks_per_seg),
2405 			segment_count_main << log_blocks_per_seg);
2406 		if (err)
2407 			return true;
2408 	}
2409 	return false;
2410 }
2411 
2412 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2413 				struct buffer_head *bh)
2414 {
2415 	block_t segment_count, segs_per_sec, secs_per_zone;
2416 	block_t total_sections, blocks_per_seg;
2417 	struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2418 					(bh->b_data + F2FS_SUPER_OFFSET);
2419 	struct super_block *sb = sbi->sb;
2420 	unsigned int blocksize;
2421 	size_t crc_offset = 0;
2422 	__u32 crc = 0;
2423 
2424 	/* Check checksum_offset and crc in superblock */
2425 	if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2426 		crc_offset = le32_to_cpu(raw_super->checksum_offset);
2427 		if (crc_offset !=
2428 			offsetof(struct f2fs_super_block, crc)) {
2429 			f2fs_msg(sb, KERN_INFO,
2430 				"Invalid SB checksum offset: %zu",
2431 				crc_offset);
2432 			return 1;
2433 		}
2434 		crc = le32_to_cpu(raw_super->crc);
2435 		if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2436 			f2fs_msg(sb, KERN_INFO,
2437 				"Invalid SB checksum value: %u", crc);
2438 			return 1;
2439 		}
2440 	}
2441 
2442 	if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
2443 		f2fs_msg(sb, KERN_INFO,
2444 			"Magic Mismatch, valid(0x%x) - read(0x%x)",
2445 			F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2446 		return 1;
2447 	}
2448 
2449 	/* Currently, support only 4KB page cache size */
2450 	if (F2FS_BLKSIZE != PAGE_SIZE) {
2451 		f2fs_msg(sb, KERN_INFO,
2452 			"Invalid page_cache_size (%lu), supports only 4KB\n",
2453 			PAGE_SIZE);
2454 		return 1;
2455 	}
2456 
2457 	/* Currently, support only 4KB block size */
2458 	blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2459 	if (blocksize != F2FS_BLKSIZE) {
2460 		f2fs_msg(sb, KERN_INFO,
2461 			"Invalid blocksize (%u), supports only 4KB\n",
2462 			blocksize);
2463 		return 1;
2464 	}
2465 
2466 	/* check log blocks per segment */
2467 	if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2468 		f2fs_msg(sb, KERN_INFO,
2469 			"Invalid log blocks per segment (%u)\n",
2470 			le32_to_cpu(raw_super->log_blocks_per_seg));
2471 		return 1;
2472 	}
2473 
2474 	/* Currently, support 512/1024/2048/4096 bytes sector size */
2475 	if (le32_to_cpu(raw_super->log_sectorsize) >
2476 				F2FS_MAX_LOG_SECTOR_SIZE ||
2477 		le32_to_cpu(raw_super->log_sectorsize) <
2478 				F2FS_MIN_LOG_SECTOR_SIZE) {
2479 		f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
2480 			le32_to_cpu(raw_super->log_sectorsize));
2481 		return 1;
2482 	}
2483 	if (le32_to_cpu(raw_super->log_sectors_per_block) +
2484 		le32_to_cpu(raw_super->log_sectorsize) !=
2485 			F2FS_MAX_LOG_SECTOR_SIZE) {
2486 		f2fs_msg(sb, KERN_INFO,
2487 			"Invalid log sectors per block(%u) log sectorsize(%u)",
2488 			le32_to_cpu(raw_super->log_sectors_per_block),
2489 			le32_to_cpu(raw_super->log_sectorsize));
2490 		return 1;
2491 	}
2492 
2493 	segment_count = le32_to_cpu(raw_super->segment_count);
2494 	segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2495 	secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2496 	total_sections = le32_to_cpu(raw_super->section_count);
2497 
2498 	/* blocks_per_seg should be 512, given the above check */
2499 	blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2500 
2501 	if (segment_count > F2FS_MAX_SEGMENT ||
2502 				segment_count < F2FS_MIN_SEGMENTS) {
2503 		f2fs_msg(sb, KERN_INFO,
2504 			"Invalid segment count (%u)",
2505 			segment_count);
2506 		return 1;
2507 	}
2508 
2509 	if (total_sections > segment_count ||
2510 			total_sections < F2FS_MIN_SEGMENTS ||
2511 			segs_per_sec > segment_count || !segs_per_sec) {
2512 		f2fs_msg(sb, KERN_INFO,
2513 			"Invalid segment/section count (%u, %u x %u)",
2514 			segment_count, total_sections, segs_per_sec);
2515 		return 1;
2516 	}
2517 
2518 	if ((segment_count / segs_per_sec) < total_sections) {
2519 		f2fs_msg(sb, KERN_INFO,
2520 			"Small segment_count (%u < %u * %u)",
2521 			segment_count, segs_per_sec, total_sections);
2522 		return 1;
2523 	}
2524 
2525 	if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2526 		f2fs_msg(sb, KERN_INFO,
2527 			"Wrong segment_count / block_count (%u > %llu)",
2528 			segment_count, le64_to_cpu(raw_super->block_count));
2529 		return 1;
2530 	}
2531 
2532 	if (secs_per_zone > total_sections || !secs_per_zone) {
2533 		f2fs_msg(sb, KERN_INFO,
2534 			"Wrong secs_per_zone / total_sections (%u, %u)",
2535 			secs_per_zone, total_sections);
2536 		return 1;
2537 	}
2538 	if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2539 			raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2540 			(le32_to_cpu(raw_super->extension_count) +
2541 			raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2542 		f2fs_msg(sb, KERN_INFO,
2543 			"Corrupted extension count (%u + %u > %u)",
2544 			le32_to_cpu(raw_super->extension_count),
2545 			raw_super->hot_ext_count,
2546 			F2FS_MAX_EXTENSION);
2547 		return 1;
2548 	}
2549 
2550 	if (le32_to_cpu(raw_super->cp_payload) >
2551 				(blocks_per_seg - F2FS_CP_PACKS)) {
2552 		f2fs_msg(sb, KERN_INFO,
2553 			"Insane cp_payload (%u > %u)",
2554 			le32_to_cpu(raw_super->cp_payload),
2555 			blocks_per_seg - F2FS_CP_PACKS);
2556 		return 1;
2557 	}
2558 
2559 	/* check reserved ino info */
2560 	if (le32_to_cpu(raw_super->node_ino) != 1 ||
2561 		le32_to_cpu(raw_super->meta_ino) != 2 ||
2562 		le32_to_cpu(raw_super->root_ino) != 3) {
2563 		f2fs_msg(sb, KERN_INFO,
2564 			"Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2565 			le32_to_cpu(raw_super->node_ino),
2566 			le32_to_cpu(raw_super->meta_ino),
2567 			le32_to_cpu(raw_super->root_ino));
2568 		return 1;
2569 	}
2570 
2571 	/* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2572 	if (sanity_check_area_boundary(sbi, bh))
2573 		return 1;
2574 
2575 	return 0;
2576 }
2577 
2578 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2579 {
2580 	unsigned int total, fsmeta;
2581 	struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2582 	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2583 	unsigned int ovp_segments, reserved_segments;
2584 	unsigned int main_segs, blocks_per_seg;
2585 	unsigned int sit_segs, nat_segs;
2586 	unsigned int sit_bitmap_size, nat_bitmap_size;
2587 	unsigned int log_blocks_per_seg;
2588 	unsigned int segment_count_main;
2589 	unsigned int cp_pack_start_sum, cp_payload;
2590 	block_t user_block_count;
2591 	int i, j;
2592 
2593 	total = le32_to_cpu(raw_super->segment_count);
2594 	fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2595 	sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2596 	fsmeta += sit_segs;
2597 	nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2598 	fsmeta += nat_segs;
2599 	fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2600 	fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2601 
2602 	if (unlikely(fsmeta >= total))
2603 		return 1;
2604 
2605 	ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2606 	reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2607 
2608 	if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2609 			ovp_segments == 0 || reserved_segments == 0)) {
2610 		f2fs_msg(sbi->sb, KERN_ERR,
2611 			"Wrong layout: check mkfs.f2fs version");
2612 		return 1;
2613 	}
2614 
2615 	user_block_count = le64_to_cpu(ckpt->user_block_count);
2616 	segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2617 	log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2618 	if (!user_block_count || user_block_count >=
2619 			segment_count_main << log_blocks_per_seg) {
2620 		f2fs_msg(sbi->sb, KERN_ERR,
2621 			"Wrong user_block_count: %u", user_block_count);
2622 		return 1;
2623 	}
2624 
2625 	main_segs = le32_to_cpu(raw_super->segment_count_main);
2626 	blocks_per_seg = sbi->blocks_per_seg;
2627 
2628 	for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2629 		if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2630 			le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2631 			return 1;
2632 		for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2633 			if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2634 				le32_to_cpu(ckpt->cur_node_segno[j])) {
2635 				f2fs_msg(sbi->sb, KERN_ERR,
2636 					"Node segment (%u, %u) has the same "
2637 					"segno: %u", i, j,
2638 					le32_to_cpu(ckpt->cur_node_segno[i]));
2639 				return 1;
2640 			}
2641 		}
2642 	}
2643 	for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2644 		if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2645 			le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2646 			return 1;
2647 		for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2648 			if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2649 				le32_to_cpu(ckpt->cur_data_segno[j])) {
2650 				f2fs_msg(sbi->sb, KERN_ERR,
2651 					"Data segment (%u, %u) has the same "
2652 					"segno: %u", i, j,
2653 					le32_to_cpu(ckpt->cur_data_segno[i]));
2654 				return 1;
2655 			}
2656 		}
2657 	}
2658 	for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2659 		for (j = i; j < NR_CURSEG_DATA_TYPE; j++) {
2660 			if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2661 				le32_to_cpu(ckpt->cur_data_segno[j])) {
2662 				f2fs_msg(sbi->sb, KERN_ERR,
2663 					"Data segment (%u) and Data segment (%u)"
2664 					" has the same segno: %u", i, j,
2665 					le32_to_cpu(ckpt->cur_node_segno[i]));
2666 				return 1;
2667 			}
2668 		}
2669 	}
2670 
2671 	sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2672 	nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2673 
2674 	if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2675 		nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2676 		f2fs_msg(sbi->sb, KERN_ERR,
2677 			"Wrong bitmap size: sit: %u, nat:%u",
2678 			sit_bitmap_size, nat_bitmap_size);
2679 		return 1;
2680 	}
2681 
2682 	cp_pack_start_sum = __start_sum_addr(sbi);
2683 	cp_payload = __cp_payload(sbi);
2684 	if (cp_pack_start_sum < cp_payload + 1 ||
2685 		cp_pack_start_sum > blocks_per_seg - 1 -
2686 			NR_CURSEG_TYPE) {
2687 		f2fs_msg(sbi->sb, KERN_ERR,
2688 			"Wrong cp_pack_start_sum: %u",
2689 			cp_pack_start_sum);
2690 		return 1;
2691 	}
2692 
2693 	if (unlikely(f2fs_cp_error(sbi))) {
2694 		f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
2695 		return 1;
2696 	}
2697 	return 0;
2698 }
2699 
2700 static void init_sb_info(struct f2fs_sb_info *sbi)
2701 {
2702 	struct f2fs_super_block *raw_super = sbi->raw_super;
2703 	int i;
2704 
2705 	sbi->log_sectors_per_block =
2706 		le32_to_cpu(raw_super->log_sectors_per_block);
2707 	sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2708 	sbi->blocksize = 1 << sbi->log_blocksize;
2709 	sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2710 	sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2711 	sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2712 	sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2713 	sbi->total_sections = le32_to_cpu(raw_super->section_count);
2714 	sbi->total_node_count =
2715 		(le32_to_cpu(raw_super->segment_count_nat) / 2)
2716 			* sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2717 	sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2718 	sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2719 	sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2720 	sbi->cur_victim_sec = NULL_SECNO;
2721 	sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
2722 	sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
2723 	sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2724 	sbi->migration_granularity = sbi->segs_per_sec;
2725 
2726 	sbi->dir_level = DEF_DIR_LEVEL;
2727 	sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2728 	sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2729 	sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
2730 	sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
2731 	sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
2732 	sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
2733 				DEF_UMOUNT_DISCARD_TIMEOUT;
2734 	clear_sbi_flag(sbi, SBI_NEED_FSCK);
2735 
2736 	for (i = 0; i < NR_COUNT_TYPE; i++)
2737 		atomic_set(&sbi->nr_pages[i], 0);
2738 
2739 	for (i = 0; i < META; i++)
2740 		atomic_set(&sbi->wb_sync_req[i], 0);
2741 
2742 	INIT_LIST_HEAD(&sbi->s_list);
2743 	mutex_init(&sbi->umount_mutex);
2744 	init_rwsem(&sbi->io_order_lock);
2745 	spin_lock_init(&sbi->cp_lock);
2746 
2747 	sbi->dirty_device = 0;
2748 	spin_lock_init(&sbi->dev_lock);
2749 
2750 	init_rwsem(&sbi->sb_lock);
2751 }
2752 
2753 static int init_percpu_info(struct f2fs_sb_info *sbi)
2754 {
2755 	int err;
2756 
2757 	err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2758 	if (err)
2759 		return err;
2760 
2761 	err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
2762 								GFP_KERNEL);
2763 	if (err)
2764 		percpu_counter_destroy(&sbi->alloc_valid_block_count);
2765 
2766 	return err;
2767 }
2768 
2769 #ifdef CONFIG_BLK_DEV_ZONED
2770 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2771 {
2772 	struct block_device *bdev = FDEV(devi).bdev;
2773 	sector_t nr_sectors = bdev->bd_part->nr_sects;
2774 	sector_t sector = 0;
2775 	struct blk_zone *zones;
2776 	unsigned int i, nr_zones;
2777 	unsigned int n = 0;
2778 	int err = -EIO;
2779 
2780 	if (!f2fs_sb_has_blkzoned(sbi))
2781 		return 0;
2782 
2783 	if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2784 				SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2785 		return -EINVAL;
2786 	sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2787 	if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2788 				__ilog2_u32(sbi->blocks_per_blkz))
2789 		return -EINVAL;
2790 	sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2791 	FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2792 					sbi->log_blocks_per_blkz;
2793 	if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2794 		FDEV(devi).nr_blkz++;
2795 
2796 	FDEV(devi).blkz_type = f2fs_kmalloc(sbi, FDEV(devi).nr_blkz,
2797 								GFP_KERNEL);
2798 	if (!FDEV(devi).blkz_type)
2799 		return -ENOMEM;
2800 
2801 #define F2FS_REPORT_NR_ZONES   4096
2802 
2803 	zones = f2fs_kzalloc(sbi,
2804 			     array_size(F2FS_REPORT_NR_ZONES,
2805 					sizeof(struct blk_zone)),
2806 			     GFP_KERNEL);
2807 	if (!zones)
2808 		return -ENOMEM;
2809 
2810 	/* Get block zones type */
2811 	while (zones && sector < nr_sectors) {
2812 
2813 		nr_zones = F2FS_REPORT_NR_ZONES;
2814 		err = blkdev_report_zones(bdev, sector,
2815 					  zones, &nr_zones,
2816 					  GFP_KERNEL);
2817 		if (err)
2818 			break;
2819 		if (!nr_zones) {
2820 			err = -EIO;
2821 			break;
2822 		}
2823 
2824 		for (i = 0; i < nr_zones; i++) {
2825 			FDEV(devi).blkz_type[n] = zones[i].type;
2826 			sector += zones[i].len;
2827 			n++;
2828 		}
2829 	}
2830 
2831 	kvfree(zones);
2832 
2833 	return err;
2834 }
2835 #endif
2836 
2837 /*
2838  * Read f2fs raw super block.
2839  * Because we have two copies of super block, so read both of them
2840  * to get the first valid one. If any one of them is broken, we pass
2841  * them recovery flag back to the caller.
2842  */
2843 static int read_raw_super_block(struct f2fs_sb_info *sbi,
2844 			struct f2fs_super_block **raw_super,
2845 			int *valid_super_block, int *recovery)
2846 {
2847 	struct super_block *sb = sbi->sb;
2848 	int block;
2849 	struct buffer_head *bh;
2850 	struct f2fs_super_block *super;
2851 	int err = 0;
2852 
2853 	super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2854 	if (!super)
2855 		return -ENOMEM;
2856 
2857 	for (block = 0; block < 2; block++) {
2858 		bh = sb_bread(sb, block);
2859 		if (!bh) {
2860 			f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
2861 				block + 1);
2862 			err = -EIO;
2863 			continue;
2864 		}
2865 
2866 		/* sanity checking of raw super */
2867 		if (sanity_check_raw_super(sbi, bh)) {
2868 			f2fs_msg(sb, KERN_ERR,
2869 				"Can't find valid F2FS filesystem in %dth superblock",
2870 				block + 1);
2871 			err = -EINVAL;
2872 			brelse(bh);
2873 			continue;
2874 		}
2875 
2876 		if (!*raw_super) {
2877 			memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2878 							sizeof(*super));
2879 			*valid_super_block = block;
2880 			*raw_super = super;
2881 		}
2882 		brelse(bh);
2883 	}
2884 
2885 	/* Fail to read any one of the superblocks*/
2886 	if (err < 0)
2887 		*recovery = 1;
2888 
2889 	/* No valid superblock */
2890 	if (!*raw_super)
2891 		kvfree(super);
2892 	else
2893 		err = 0;
2894 
2895 	return err;
2896 }
2897 
2898 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
2899 {
2900 	struct buffer_head *bh;
2901 	__u32 crc = 0;
2902 	int err;
2903 
2904 	if ((recover && f2fs_readonly(sbi->sb)) ||
2905 				bdev_read_only(sbi->sb->s_bdev)) {
2906 		set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2907 		return -EROFS;
2908 	}
2909 
2910 	/* we should update superblock crc here */
2911 	if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
2912 		crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
2913 				offsetof(struct f2fs_super_block, crc));
2914 		F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
2915 	}
2916 
2917 	/* write back-up superblock first */
2918 	bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
2919 	if (!bh)
2920 		return -EIO;
2921 	err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2922 	brelse(bh);
2923 
2924 	/* if we are in recovery path, skip writing valid superblock */
2925 	if (recover || err)
2926 		return err;
2927 
2928 	/* write current valid superblock */
2929 	bh = sb_bread(sbi->sb, sbi->valid_super_block);
2930 	if (!bh)
2931 		return -EIO;
2932 	err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2933 	brelse(bh);
2934 	return err;
2935 }
2936 
2937 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
2938 {
2939 	struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2940 	unsigned int max_devices = MAX_DEVICES;
2941 	int i;
2942 
2943 	/* Initialize single device information */
2944 	if (!RDEV(0).path[0]) {
2945 		if (!bdev_is_zoned(sbi->sb->s_bdev))
2946 			return 0;
2947 		max_devices = 1;
2948 	}
2949 
2950 	/*
2951 	 * Initialize multiple devices information, or single
2952 	 * zoned block device information.
2953 	 */
2954 	sbi->devs = f2fs_kzalloc(sbi,
2955 				 array_size(max_devices,
2956 					    sizeof(struct f2fs_dev_info)),
2957 				 GFP_KERNEL);
2958 	if (!sbi->devs)
2959 		return -ENOMEM;
2960 
2961 	for (i = 0; i < max_devices; i++) {
2962 
2963 		if (i > 0 && !RDEV(i).path[0])
2964 			break;
2965 
2966 		if (max_devices == 1) {
2967 			/* Single zoned block device mount */
2968 			FDEV(0).bdev =
2969 				blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
2970 					sbi->sb->s_mode, sbi->sb->s_type);
2971 		} else {
2972 			/* Multi-device mount */
2973 			memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
2974 			FDEV(i).total_segments =
2975 				le32_to_cpu(RDEV(i).total_segments);
2976 			if (i == 0) {
2977 				FDEV(i).start_blk = 0;
2978 				FDEV(i).end_blk = FDEV(i).start_blk +
2979 				    (FDEV(i).total_segments <<
2980 				    sbi->log_blocks_per_seg) - 1 +
2981 				    le32_to_cpu(raw_super->segment0_blkaddr);
2982 			} else {
2983 				FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
2984 				FDEV(i).end_blk = FDEV(i).start_blk +
2985 					(FDEV(i).total_segments <<
2986 					sbi->log_blocks_per_seg) - 1;
2987 			}
2988 			FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
2989 					sbi->sb->s_mode, sbi->sb->s_type);
2990 		}
2991 		if (IS_ERR(FDEV(i).bdev))
2992 			return PTR_ERR(FDEV(i).bdev);
2993 
2994 		/* to release errored devices */
2995 		sbi->s_ndevs = i + 1;
2996 
2997 #ifdef CONFIG_BLK_DEV_ZONED
2998 		if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
2999 				!f2fs_sb_has_blkzoned(sbi)) {
3000 			f2fs_msg(sbi->sb, KERN_ERR,
3001 				"Zoned block device feature not enabled\n");
3002 			return -EINVAL;
3003 		}
3004 		if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3005 			if (init_blkz_info(sbi, i)) {
3006 				f2fs_msg(sbi->sb, KERN_ERR,
3007 					"Failed to initialize F2FS blkzone information");
3008 				return -EINVAL;
3009 			}
3010 			if (max_devices == 1)
3011 				break;
3012 			f2fs_msg(sbi->sb, KERN_INFO,
3013 				"Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3014 				i, FDEV(i).path,
3015 				FDEV(i).total_segments,
3016 				FDEV(i).start_blk, FDEV(i).end_blk,
3017 				bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3018 				"Host-aware" : "Host-managed");
3019 			continue;
3020 		}
3021 #endif
3022 		f2fs_msg(sbi->sb, KERN_INFO,
3023 			"Mount Device [%2d]: %20s, %8u, %8x - %8x",
3024 				i, FDEV(i).path,
3025 				FDEV(i).total_segments,
3026 				FDEV(i).start_blk, FDEV(i).end_blk);
3027 	}
3028 	f2fs_msg(sbi->sb, KERN_INFO,
3029 			"IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3030 	return 0;
3031 }
3032 
3033 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3034 {
3035 	struct f2fs_sm_info *sm_i = SM_I(sbi);
3036 
3037 	/* adjust parameters according to the volume size */
3038 	if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3039 		F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3040 		sm_i->dcc_info->discard_granularity = 1;
3041 		sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3042 	}
3043 
3044 	sbi->readdir_ra = 1;
3045 }
3046 
3047 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3048 {
3049 	struct f2fs_sb_info *sbi;
3050 	struct f2fs_super_block *raw_super;
3051 	struct inode *root;
3052 	int err;
3053 	bool skip_recovery = false, need_fsck = false;
3054 	char *options = NULL;
3055 	int recovery, i, valid_super_block;
3056 	struct curseg_info *seg_i;
3057 	int retry_cnt = 1;
3058 
3059 try_onemore:
3060 	err = -EINVAL;
3061 	raw_super = NULL;
3062 	valid_super_block = -1;
3063 	recovery = 0;
3064 
3065 	/* allocate memory for f2fs-specific super block info */
3066 	sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3067 	if (!sbi)
3068 		return -ENOMEM;
3069 
3070 	sbi->sb = sb;
3071 
3072 	/* Load the checksum driver */
3073 	sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3074 	if (IS_ERR(sbi->s_chksum_driver)) {
3075 		f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
3076 		err = PTR_ERR(sbi->s_chksum_driver);
3077 		sbi->s_chksum_driver = NULL;
3078 		goto free_sbi;
3079 	}
3080 
3081 	/* set a block size */
3082 	if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3083 		f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
3084 		goto free_sbi;
3085 	}
3086 
3087 	err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3088 								&recovery);
3089 	if (err)
3090 		goto free_sbi;
3091 
3092 	sb->s_fs_info = sbi;
3093 	sbi->raw_super = raw_super;
3094 
3095 	/* precompute checksum seed for metadata */
3096 	if (f2fs_sb_has_inode_chksum(sbi))
3097 		sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3098 						sizeof(raw_super->uuid));
3099 
3100 	/*
3101 	 * The BLKZONED feature indicates that the drive was formatted with
3102 	 * zone alignment optimization. This is optional for host-aware
3103 	 * devices, but mandatory for host-managed zoned block devices.
3104 	 */
3105 #ifndef CONFIG_BLK_DEV_ZONED
3106 	if (f2fs_sb_has_blkzoned(sbi)) {
3107 		f2fs_msg(sb, KERN_ERR,
3108 			 "Zoned block device support is not enabled\n");
3109 		err = -EOPNOTSUPP;
3110 		goto free_sb_buf;
3111 	}
3112 #endif
3113 	default_options(sbi);
3114 	/* parse mount options */
3115 	options = kstrdup((const char *)data, GFP_KERNEL);
3116 	if (data && !options) {
3117 		err = -ENOMEM;
3118 		goto free_sb_buf;
3119 	}
3120 
3121 	err = parse_options(sb, options);
3122 	if (err)
3123 		goto free_options;
3124 
3125 	sbi->max_file_blocks = max_file_blocks();
3126 	sb->s_maxbytes = sbi->max_file_blocks <<
3127 				le32_to_cpu(raw_super->log_blocksize);
3128 	sb->s_max_links = F2FS_LINK_MAX;
3129 
3130 #ifdef CONFIG_QUOTA
3131 	sb->dq_op = &f2fs_quota_operations;
3132 	if (f2fs_sb_has_quota_ino(sbi))
3133 		sb->s_qcop = &dquot_quotactl_sysfile_ops;
3134 	else
3135 		sb->s_qcop = &f2fs_quotactl_ops;
3136 	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3137 
3138 	if (f2fs_sb_has_quota_ino(sbi)) {
3139 		for (i = 0; i < MAXQUOTAS; i++) {
3140 			if (f2fs_qf_ino(sbi->sb, i))
3141 				sbi->nquota_files++;
3142 		}
3143 	}
3144 #endif
3145 
3146 	sb->s_op = &f2fs_sops;
3147 #ifdef CONFIG_FS_ENCRYPTION
3148 	sb->s_cop = &f2fs_cryptops;
3149 #endif
3150 	sb->s_xattr = f2fs_xattr_handlers;
3151 	sb->s_export_op = &f2fs_export_ops;
3152 	sb->s_magic = F2FS_SUPER_MAGIC;
3153 	sb->s_time_gran = 1;
3154 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3155 		(test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3156 	memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3157 	sb->s_iflags |= SB_I_CGROUPWB;
3158 
3159 	/* init f2fs-specific super block info */
3160 	sbi->valid_super_block = valid_super_block;
3161 	mutex_init(&sbi->gc_mutex);
3162 	mutex_init(&sbi->writepages);
3163 	mutex_init(&sbi->cp_mutex);
3164 	init_rwsem(&sbi->node_write);
3165 	init_rwsem(&sbi->node_change);
3166 
3167 	/* disallow all the data/node/meta page writes */
3168 	set_sbi_flag(sbi, SBI_POR_DOING);
3169 	spin_lock_init(&sbi->stat_lock);
3170 
3171 	/* init iostat info */
3172 	spin_lock_init(&sbi->iostat_lock);
3173 	sbi->iostat_enable = false;
3174 
3175 	for (i = 0; i < NR_PAGE_TYPE; i++) {
3176 		int n = (i == META) ? 1: NR_TEMP_TYPE;
3177 		int j;
3178 
3179 		sbi->write_io[i] =
3180 			f2fs_kmalloc(sbi,
3181 				     array_size(n,
3182 						sizeof(struct f2fs_bio_info)),
3183 				     GFP_KERNEL);
3184 		if (!sbi->write_io[i]) {
3185 			err = -ENOMEM;
3186 			goto free_bio_info;
3187 		}
3188 
3189 		for (j = HOT; j < n; j++) {
3190 			init_rwsem(&sbi->write_io[i][j].io_rwsem);
3191 			sbi->write_io[i][j].sbi = sbi;
3192 			sbi->write_io[i][j].bio = NULL;
3193 			spin_lock_init(&sbi->write_io[i][j].io_lock);
3194 			INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3195 		}
3196 	}
3197 
3198 	init_rwsem(&sbi->cp_rwsem);
3199 	init_waitqueue_head(&sbi->cp_wait);
3200 	init_sb_info(sbi);
3201 
3202 	err = init_percpu_info(sbi);
3203 	if (err)
3204 		goto free_bio_info;
3205 
3206 	if (F2FS_IO_SIZE(sbi) > 1) {
3207 		sbi->write_io_dummy =
3208 			mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3209 		if (!sbi->write_io_dummy) {
3210 			err = -ENOMEM;
3211 			goto free_percpu;
3212 		}
3213 	}
3214 
3215 	/* get an inode for meta space */
3216 	sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3217 	if (IS_ERR(sbi->meta_inode)) {
3218 		f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
3219 		err = PTR_ERR(sbi->meta_inode);
3220 		goto free_io_dummy;
3221 	}
3222 
3223 	err = f2fs_get_valid_checkpoint(sbi);
3224 	if (err) {
3225 		f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
3226 		goto free_meta_inode;
3227 	}
3228 
3229 	if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3230 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3231 	if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3232 		set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3233 		sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3234 	}
3235 
3236 	/* Initialize device list */
3237 	err = f2fs_scan_devices(sbi);
3238 	if (err) {
3239 		f2fs_msg(sb, KERN_ERR, "Failed to find devices");
3240 		goto free_devices;
3241 	}
3242 
3243 	sbi->total_valid_node_count =
3244 				le32_to_cpu(sbi->ckpt->valid_node_count);
3245 	percpu_counter_set(&sbi->total_valid_inode_count,
3246 				le32_to_cpu(sbi->ckpt->valid_inode_count));
3247 	sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3248 	sbi->total_valid_block_count =
3249 				le64_to_cpu(sbi->ckpt->valid_block_count);
3250 	sbi->last_valid_block_count = sbi->total_valid_block_count;
3251 	sbi->reserved_blocks = 0;
3252 	sbi->current_reserved_blocks = 0;
3253 	limit_reserve_root(sbi);
3254 
3255 	for (i = 0; i < NR_INODE_TYPE; i++) {
3256 		INIT_LIST_HEAD(&sbi->inode_list[i]);
3257 		spin_lock_init(&sbi->inode_lock[i]);
3258 	}
3259 
3260 	f2fs_init_extent_cache_info(sbi);
3261 
3262 	f2fs_init_ino_entry_info(sbi);
3263 
3264 	f2fs_init_fsync_node_info(sbi);
3265 
3266 	/* setup f2fs internal modules */
3267 	err = f2fs_build_segment_manager(sbi);
3268 	if (err) {
3269 		f2fs_msg(sb, KERN_ERR,
3270 			"Failed to initialize F2FS segment manager");
3271 		goto free_sm;
3272 	}
3273 	err = f2fs_build_node_manager(sbi);
3274 	if (err) {
3275 		f2fs_msg(sb, KERN_ERR,
3276 			"Failed to initialize F2FS node manager");
3277 		goto free_nm;
3278 	}
3279 
3280 	/* For write statistics */
3281 	if (sb->s_bdev->bd_part)
3282 		sbi->sectors_written_start =
3283 			(u64)part_stat_read(sb->s_bdev->bd_part,
3284 					    sectors[STAT_WRITE]);
3285 
3286 	/* Read accumulated write IO statistics if exists */
3287 	seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3288 	if (__exist_node_summaries(sbi))
3289 		sbi->kbytes_written =
3290 			le64_to_cpu(seg_i->journal->info.kbytes_written);
3291 
3292 	f2fs_build_gc_manager(sbi);
3293 
3294 	err = f2fs_build_stats(sbi);
3295 	if (err)
3296 		goto free_nm;
3297 
3298 	/* get an inode for node space */
3299 	sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3300 	if (IS_ERR(sbi->node_inode)) {
3301 		f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
3302 		err = PTR_ERR(sbi->node_inode);
3303 		goto free_stats;
3304 	}
3305 
3306 	/* read root inode and dentry */
3307 	root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3308 	if (IS_ERR(root)) {
3309 		f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
3310 		err = PTR_ERR(root);
3311 		goto free_node_inode;
3312 	}
3313 	if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3314 			!root->i_size || !root->i_nlink) {
3315 		iput(root);
3316 		err = -EINVAL;
3317 		goto free_node_inode;
3318 	}
3319 
3320 	sb->s_root = d_make_root(root); /* allocate root dentry */
3321 	if (!sb->s_root) {
3322 		err = -ENOMEM;
3323 		goto free_node_inode;
3324 	}
3325 
3326 	err = f2fs_register_sysfs(sbi);
3327 	if (err)
3328 		goto free_root_inode;
3329 
3330 #ifdef CONFIG_QUOTA
3331 	/* Enable quota usage during mount */
3332 	if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3333 		err = f2fs_enable_quotas(sb);
3334 		if (err)
3335 			f2fs_msg(sb, KERN_ERR,
3336 				"Cannot turn on quotas: error %d", err);
3337 	}
3338 #endif
3339 	/* if there are nt orphan nodes free them */
3340 	err = f2fs_recover_orphan_inodes(sbi);
3341 	if (err)
3342 		goto free_meta;
3343 
3344 	if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3345 		goto reset_checkpoint;
3346 
3347 	/* recover fsynced data */
3348 	if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
3349 		/*
3350 		 * mount should be failed, when device has readonly mode, and
3351 		 * previous checkpoint was not done by clean system shutdown.
3352 		 */
3353 		if (bdev_read_only(sb->s_bdev) &&
3354 				!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3355 			err = -EROFS;
3356 			goto free_meta;
3357 		}
3358 
3359 		if (need_fsck)
3360 			set_sbi_flag(sbi, SBI_NEED_FSCK);
3361 
3362 		if (skip_recovery)
3363 			goto reset_checkpoint;
3364 
3365 		err = f2fs_recover_fsync_data(sbi, false);
3366 		if (err < 0) {
3367 			if (err != -ENOMEM)
3368 				skip_recovery = true;
3369 			need_fsck = true;
3370 			f2fs_msg(sb, KERN_ERR,
3371 				"Cannot recover all fsync data errno=%d", err);
3372 			goto free_meta;
3373 		}
3374 	} else {
3375 		err = f2fs_recover_fsync_data(sbi, true);
3376 
3377 		if (!f2fs_readonly(sb) && err > 0) {
3378 			err = -EINVAL;
3379 			f2fs_msg(sb, KERN_ERR,
3380 				"Need to recover fsync data");
3381 			goto free_meta;
3382 		}
3383 	}
3384 reset_checkpoint:
3385 	/* f2fs_recover_fsync_data() cleared this already */
3386 	clear_sbi_flag(sbi, SBI_POR_DOING);
3387 
3388 	if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3389 		err = f2fs_disable_checkpoint(sbi);
3390 		if (err)
3391 			goto sync_free_meta;
3392 	} else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3393 		f2fs_enable_checkpoint(sbi);
3394 	}
3395 
3396 	/*
3397 	 * If filesystem is not mounted as read-only then
3398 	 * do start the gc_thread.
3399 	 */
3400 	if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
3401 		/* After POR, we can run background GC thread.*/
3402 		err = f2fs_start_gc_thread(sbi);
3403 		if (err)
3404 			goto sync_free_meta;
3405 	}
3406 	kvfree(options);
3407 
3408 	/* recover broken superblock */
3409 	if (recovery) {
3410 		err = f2fs_commit_super(sbi, true);
3411 		f2fs_msg(sb, KERN_INFO,
3412 			"Try to recover %dth superblock, ret: %d",
3413 			sbi->valid_super_block ? 1 : 2, err);
3414 	}
3415 
3416 	f2fs_join_shrinker(sbi);
3417 
3418 	f2fs_tuning_parameters(sbi);
3419 
3420 	f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx",
3421 				cur_cp_version(F2FS_CKPT(sbi)));
3422 	f2fs_update_time(sbi, CP_TIME);
3423 	f2fs_update_time(sbi, REQ_TIME);
3424 	clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3425 	return 0;
3426 
3427 sync_free_meta:
3428 	/* safe to flush all the data */
3429 	sync_filesystem(sbi->sb);
3430 	retry_cnt = 0;
3431 
3432 free_meta:
3433 #ifdef CONFIG_QUOTA
3434 	f2fs_truncate_quota_inode_pages(sb);
3435 	if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3436 		f2fs_quota_off_umount(sbi->sb);
3437 #endif
3438 	/*
3439 	 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3440 	 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3441 	 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3442 	 * falls into an infinite loop in f2fs_sync_meta_pages().
3443 	 */
3444 	truncate_inode_pages_final(META_MAPPING(sbi));
3445 	/* evict some inodes being cached by GC */
3446 	evict_inodes(sb);
3447 	f2fs_unregister_sysfs(sbi);
3448 free_root_inode:
3449 	dput(sb->s_root);
3450 	sb->s_root = NULL;
3451 free_node_inode:
3452 	f2fs_release_ino_entry(sbi, true);
3453 	truncate_inode_pages_final(NODE_MAPPING(sbi));
3454 	iput(sbi->node_inode);
3455 	sbi->node_inode = NULL;
3456 free_stats:
3457 	f2fs_destroy_stats(sbi);
3458 free_nm:
3459 	f2fs_destroy_node_manager(sbi);
3460 free_sm:
3461 	f2fs_destroy_segment_manager(sbi);
3462 free_devices:
3463 	destroy_device_list(sbi);
3464 	kvfree(sbi->ckpt);
3465 free_meta_inode:
3466 	make_bad_inode(sbi->meta_inode);
3467 	iput(sbi->meta_inode);
3468 	sbi->meta_inode = NULL;
3469 free_io_dummy:
3470 	mempool_destroy(sbi->write_io_dummy);
3471 free_percpu:
3472 	destroy_percpu_info(sbi);
3473 free_bio_info:
3474 	for (i = 0; i < NR_PAGE_TYPE; i++)
3475 		kvfree(sbi->write_io[i]);
3476 free_options:
3477 #ifdef CONFIG_QUOTA
3478 	for (i = 0; i < MAXQUOTAS; i++)
3479 		kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
3480 #endif
3481 	kvfree(options);
3482 free_sb_buf:
3483 	kvfree(raw_super);
3484 free_sbi:
3485 	if (sbi->s_chksum_driver)
3486 		crypto_free_shash(sbi->s_chksum_driver);
3487 	kvfree(sbi);
3488 
3489 	/* give only one another chance */
3490 	if (retry_cnt > 0 && skip_recovery) {
3491 		retry_cnt--;
3492 		shrink_dcache_sb(sb);
3493 		goto try_onemore;
3494 	}
3495 	return err;
3496 }
3497 
3498 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3499 			const char *dev_name, void *data)
3500 {
3501 	return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3502 }
3503 
3504 static void kill_f2fs_super(struct super_block *sb)
3505 {
3506 	if (sb->s_root) {
3507 		struct f2fs_sb_info *sbi = F2FS_SB(sb);
3508 
3509 		set_sbi_flag(sbi, SBI_IS_CLOSE);
3510 		f2fs_stop_gc_thread(sbi);
3511 		f2fs_stop_discard_thread(sbi);
3512 
3513 		if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3514 				!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3515 			struct cp_control cpc = {
3516 				.reason = CP_UMOUNT,
3517 			};
3518 			f2fs_write_checkpoint(sbi, &cpc);
3519 		}
3520 
3521 		if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3522 			sb->s_flags &= ~SB_RDONLY;
3523 	}
3524 	kill_block_super(sb);
3525 }
3526 
3527 static struct file_system_type f2fs_fs_type = {
3528 	.owner		= THIS_MODULE,
3529 	.name		= "f2fs",
3530 	.mount		= f2fs_mount,
3531 	.kill_sb	= kill_f2fs_super,
3532 	.fs_flags	= FS_REQUIRES_DEV,
3533 };
3534 MODULE_ALIAS_FS("f2fs");
3535 
3536 static int __init init_inodecache(void)
3537 {
3538 	f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3539 			sizeof(struct f2fs_inode_info), 0,
3540 			SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3541 	if (!f2fs_inode_cachep)
3542 		return -ENOMEM;
3543 	return 0;
3544 }
3545 
3546 static void destroy_inodecache(void)
3547 {
3548 	/*
3549 	 * Make sure all delayed rcu free inodes are flushed before we
3550 	 * destroy cache.
3551 	 */
3552 	rcu_barrier();
3553 	kmem_cache_destroy(f2fs_inode_cachep);
3554 }
3555 
3556 static int __init init_f2fs_fs(void)
3557 {
3558 	int err;
3559 
3560 	if (PAGE_SIZE != F2FS_BLKSIZE) {
3561 		printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3562 				PAGE_SIZE, F2FS_BLKSIZE);
3563 		return -EINVAL;
3564 	}
3565 
3566 	f2fs_build_trace_ios();
3567 
3568 	err = init_inodecache();
3569 	if (err)
3570 		goto fail;
3571 	err = f2fs_create_node_manager_caches();
3572 	if (err)
3573 		goto free_inodecache;
3574 	err = f2fs_create_segment_manager_caches();
3575 	if (err)
3576 		goto free_node_manager_caches;
3577 	err = f2fs_create_checkpoint_caches();
3578 	if (err)
3579 		goto free_segment_manager_caches;
3580 	err = f2fs_create_extent_cache();
3581 	if (err)
3582 		goto free_checkpoint_caches;
3583 	err = f2fs_init_sysfs();
3584 	if (err)
3585 		goto free_extent_cache;
3586 	err = register_shrinker(&f2fs_shrinker_info);
3587 	if (err)
3588 		goto free_sysfs;
3589 	err = register_filesystem(&f2fs_fs_type);
3590 	if (err)
3591 		goto free_shrinker;
3592 	f2fs_create_root_stats();
3593 	err = f2fs_init_post_read_processing();
3594 	if (err)
3595 		goto free_root_stats;
3596 	return 0;
3597 
3598 free_root_stats:
3599 	f2fs_destroy_root_stats();
3600 	unregister_filesystem(&f2fs_fs_type);
3601 free_shrinker:
3602 	unregister_shrinker(&f2fs_shrinker_info);
3603 free_sysfs:
3604 	f2fs_exit_sysfs();
3605 free_extent_cache:
3606 	f2fs_destroy_extent_cache();
3607 free_checkpoint_caches:
3608 	f2fs_destroy_checkpoint_caches();
3609 free_segment_manager_caches:
3610 	f2fs_destroy_segment_manager_caches();
3611 free_node_manager_caches:
3612 	f2fs_destroy_node_manager_caches();
3613 free_inodecache:
3614 	destroy_inodecache();
3615 fail:
3616 	return err;
3617 }
3618 
3619 static void __exit exit_f2fs_fs(void)
3620 {
3621 	f2fs_destroy_post_read_processing();
3622 	f2fs_destroy_root_stats();
3623 	unregister_filesystem(&f2fs_fs_type);
3624 	unregister_shrinker(&f2fs_shrinker_info);
3625 	f2fs_exit_sysfs();
3626 	f2fs_destroy_extent_cache();
3627 	f2fs_destroy_checkpoint_caches();
3628 	f2fs_destroy_segment_manager_caches();
3629 	f2fs_destroy_node_manager_caches();
3630 	destroy_inodecache();
3631 	f2fs_destroy_trace_ios();
3632 }
3633 
3634 module_init(init_f2fs_fs)
3635 module_exit(exit_f2fs_fs)
3636 
3637 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3638 MODULE_DESCRIPTION("Flash Friendly File System");
3639 MODULE_LICENSE("GPL");
3640 
3641