xref: /linux/fs/f2fs/super.c (revision c0e79fd89749b0cda1c72049e2772dd2eeada86f)
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 #include <linux/unicode.h>
27 
28 #include "f2fs.h"
29 #include "node.h"
30 #include "segment.h"
31 #include "xattr.h"
32 #include "gc.h"
33 #include "trace.h"
34 
35 #define CREATE_TRACE_POINTS
36 #include <trace/events/f2fs.h>
37 
38 static struct kmem_cache *f2fs_inode_cachep;
39 
40 #ifdef CONFIG_F2FS_FAULT_INJECTION
41 
42 const char *f2fs_fault_name[FAULT_MAX] = {
43 	[FAULT_KMALLOC]		= "kmalloc",
44 	[FAULT_KVMALLOC]	= "kvmalloc",
45 	[FAULT_PAGE_ALLOC]	= "page alloc",
46 	[FAULT_PAGE_GET]	= "page get",
47 	[FAULT_ALLOC_BIO]	= "alloc bio",
48 	[FAULT_ALLOC_NID]	= "alloc nid",
49 	[FAULT_ORPHAN]		= "orphan",
50 	[FAULT_BLOCK]		= "no more block",
51 	[FAULT_DIR_DEPTH]	= "too big dir depth",
52 	[FAULT_EVICT_INODE]	= "evict_inode fail",
53 	[FAULT_TRUNCATE]	= "truncate fail",
54 	[FAULT_READ_IO]		= "read IO error",
55 	[FAULT_CHECKPOINT]	= "checkpoint error",
56 	[FAULT_DISCARD]		= "discard error",
57 	[FAULT_WRITE_IO]	= "write IO error",
58 };
59 
60 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
61 							unsigned int type)
62 {
63 	struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
64 
65 	if (rate) {
66 		atomic_set(&ffi->inject_ops, 0);
67 		ffi->inject_rate = rate;
68 	}
69 
70 	if (type)
71 		ffi->inject_type = type;
72 
73 	if (!rate && !type)
74 		memset(ffi, 0, sizeof(struct f2fs_fault_info));
75 }
76 #endif
77 
78 /* f2fs-wide shrinker description */
79 static struct shrinker f2fs_shrinker_info = {
80 	.scan_objects = f2fs_shrink_scan,
81 	.count_objects = f2fs_shrink_count,
82 	.seeks = DEFAULT_SEEKS,
83 };
84 
85 enum {
86 	Opt_gc_background,
87 	Opt_disable_roll_forward,
88 	Opt_norecovery,
89 	Opt_discard,
90 	Opt_nodiscard,
91 	Opt_noheap,
92 	Opt_heap,
93 	Opt_user_xattr,
94 	Opt_nouser_xattr,
95 	Opt_acl,
96 	Opt_noacl,
97 	Opt_active_logs,
98 	Opt_disable_ext_identify,
99 	Opt_inline_xattr,
100 	Opt_noinline_xattr,
101 	Opt_inline_xattr_size,
102 	Opt_inline_data,
103 	Opt_inline_dentry,
104 	Opt_noinline_dentry,
105 	Opt_flush_merge,
106 	Opt_noflush_merge,
107 	Opt_nobarrier,
108 	Opt_fastboot,
109 	Opt_extent_cache,
110 	Opt_noextent_cache,
111 	Opt_noinline_data,
112 	Opt_data_flush,
113 	Opt_reserve_root,
114 	Opt_resgid,
115 	Opt_resuid,
116 	Opt_mode,
117 	Opt_io_size_bits,
118 	Opt_fault_injection,
119 	Opt_fault_type,
120 	Opt_lazytime,
121 	Opt_nolazytime,
122 	Opt_quota,
123 	Opt_noquota,
124 	Opt_usrquota,
125 	Opt_grpquota,
126 	Opt_prjquota,
127 	Opt_usrjquota,
128 	Opt_grpjquota,
129 	Opt_prjjquota,
130 	Opt_offusrjquota,
131 	Opt_offgrpjquota,
132 	Opt_offprjjquota,
133 	Opt_jqfmt_vfsold,
134 	Opt_jqfmt_vfsv0,
135 	Opt_jqfmt_vfsv1,
136 	Opt_whint,
137 	Opt_alloc,
138 	Opt_fsync,
139 	Opt_test_dummy_encryption,
140 	Opt_checkpoint_disable,
141 	Opt_checkpoint_disable_cap,
142 	Opt_checkpoint_disable_cap_perc,
143 	Opt_checkpoint_enable,
144 	Opt_err,
145 };
146 
147 static match_table_t f2fs_tokens = {
148 	{Opt_gc_background, "background_gc=%s"},
149 	{Opt_disable_roll_forward, "disable_roll_forward"},
150 	{Opt_norecovery, "norecovery"},
151 	{Opt_discard, "discard"},
152 	{Opt_nodiscard, "nodiscard"},
153 	{Opt_noheap, "no_heap"},
154 	{Opt_heap, "heap"},
155 	{Opt_user_xattr, "user_xattr"},
156 	{Opt_nouser_xattr, "nouser_xattr"},
157 	{Opt_acl, "acl"},
158 	{Opt_noacl, "noacl"},
159 	{Opt_active_logs, "active_logs=%u"},
160 	{Opt_disable_ext_identify, "disable_ext_identify"},
161 	{Opt_inline_xattr, "inline_xattr"},
162 	{Opt_noinline_xattr, "noinline_xattr"},
163 	{Opt_inline_xattr_size, "inline_xattr_size=%u"},
164 	{Opt_inline_data, "inline_data"},
165 	{Opt_inline_dentry, "inline_dentry"},
166 	{Opt_noinline_dentry, "noinline_dentry"},
167 	{Opt_flush_merge, "flush_merge"},
168 	{Opt_noflush_merge, "noflush_merge"},
169 	{Opt_nobarrier, "nobarrier"},
170 	{Opt_fastboot, "fastboot"},
171 	{Opt_extent_cache, "extent_cache"},
172 	{Opt_noextent_cache, "noextent_cache"},
173 	{Opt_noinline_data, "noinline_data"},
174 	{Opt_data_flush, "data_flush"},
175 	{Opt_reserve_root, "reserve_root=%u"},
176 	{Opt_resgid, "resgid=%u"},
177 	{Opt_resuid, "resuid=%u"},
178 	{Opt_mode, "mode=%s"},
179 	{Opt_io_size_bits, "io_bits=%u"},
180 	{Opt_fault_injection, "fault_injection=%u"},
181 	{Opt_fault_type, "fault_type=%u"},
182 	{Opt_lazytime, "lazytime"},
183 	{Opt_nolazytime, "nolazytime"},
184 	{Opt_quota, "quota"},
185 	{Opt_noquota, "noquota"},
186 	{Opt_usrquota, "usrquota"},
187 	{Opt_grpquota, "grpquota"},
188 	{Opt_prjquota, "prjquota"},
189 	{Opt_usrjquota, "usrjquota=%s"},
190 	{Opt_grpjquota, "grpjquota=%s"},
191 	{Opt_prjjquota, "prjjquota=%s"},
192 	{Opt_offusrjquota, "usrjquota="},
193 	{Opt_offgrpjquota, "grpjquota="},
194 	{Opt_offprjjquota, "prjjquota="},
195 	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
196 	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
197 	{Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
198 	{Opt_whint, "whint_mode=%s"},
199 	{Opt_alloc, "alloc_mode=%s"},
200 	{Opt_fsync, "fsync_mode=%s"},
201 	{Opt_test_dummy_encryption, "test_dummy_encryption"},
202 	{Opt_checkpoint_disable, "checkpoint=disable"},
203 	{Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
204 	{Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
205 	{Opt_checkpoint_enable, "checkpoint=enable"},
206 	{Opt_err, NULL},
207 };
208 
209 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
210 {
211 	struct va_format vaf;
212 	va_list args;
213 	int level;
214 
215 	va_start(args, fmt);
216 
217 	level = printk_get_level(fmt);
218 	vaf.fmt = printk_skip_level(fmt);
219 	vaf.va = &args;
220 	printk("%c%cF2FS-fs (%s): %pV\n",
221 	       KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
222 
223 	va_end(args);
224 }
225 
226 #ifdef CONFIG_UNICODE
227 static const struct f2fs_sb_encodings {
228 	__u16 magic;
229 	char *name;
230 	char *version;
231 } f2fs_sb_encoding_map[] = {
232 	{F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"},
233 };
234 
235 static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb,
236 				 const struct f2fs_sb_encodings **encoding,
237 				 __u16 *flags)
238 {
239 	__u16 magic = le16_to_cpu(sb->s_encoding);
240 	int i;
241 
242 	for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
243 		if (magic == f2fs_sb_encoding_map[i].magic)
244 			break;
245 
246 	if (i >= ARRAY_SIZE(f2fs_sb_encoding_map))
247 		return -EINVAL;
248 
249 	*encoding = &f2fs_sb_encoding_map[i];
250 	*flags = le16_to_cpu(sb->s_encoding_flags);
251 
252 	return 0;
253 }
254 #endif
255 
256 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
257 {
258 	block_t limit = min((sbi->user_block_count << 1) / 1000,
259 			sbi->user_block_count - sbi->reserved_blocks);
260 
261 	/* limit is 0.2% */
262 	if (test_opt(sbi, RESERVE_ROOT) &&
263 			F2FS_OPTION(sbi).root_reserved_blocks > limit) {
264 		F2FS_OPTION(sbi).root_reserved_blocks = limit;
265 		f2fs_info(sbi, "Reduce reserved blocks for root = %u",
266 			  F2FS_OPTION(sbi).root_reserved_blocks);
267 	}
268 	if (!test_opt(sbi, RESERVE_ROOT) &&
269 		(!uid_eq(F2FS_OPTION(sbi).s_resuid,
270 				make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
271 		!gid_eq(F2FS_OPTION(sbi).s_resgid,
272 				make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
273 		f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
274 			  from_kuid_munged(&init_user_ns,
275 					   F2FS_OPTION(sbi).s_resuid),
276 			  from_kgid_munged(&init_user_ns,
277 					   F2FS_OPTION(sbi).s_resgid));
278 }
279 
280 static void init_once(void *foo)
281 {
282 	struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
283 
284 	inode_init_once(&fi->vfs_inode);
285 }
286 
287 #ifdef CONFIG_QUOTA
288 static const char * const quotatypes[] = INITQFNAMES;
289 #define QTYPE2NAME(t) (quotatypes[t])
290 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
291 							substring_t *args)
292 {
293 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
294 	char *qname;
295 	int ret = -EINVAL;
296 
297 	if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
298 		f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
299 		return -EINVAL;
300 	}
301 	if (f2fs_sb_has_quota_ino(sbi)) {
302 		f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
303 		return 0;
304 	}
305 
306 	qname = match_strdup(args);
307 	if (!qname) {
308 		f2fs_err(sbi, "Not enough memory for storing quotafile name");
309 		return -ENOMEM;
310 	}
311 	if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
312 		if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
313 			ret = 0;
314 		else
315 			f2fs_err(sbi, "%s quota file already specified",
316 				 QTYPE2NAME(qtype));
317 		goto errout;
318 	}
319 	if (strchr(qname, '/')) {
320 		f2fs_err(sbi, "quotafile must be on filesystem root");
321 		goto errout;
322 	}
323 	F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
324 	set_opt(sbi, QUOTA);
325 	return 0;
326 errout:
327 	kvfree(qname);
328 	return ret;
329 }
330 
331 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
332 {
333 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
334 
335 	if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
336 		f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
337 		return -EINVAL;
338 	}
339 	kvfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
340 	F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
341 	return 0;
342 }
343 
344 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
345 {
346 	/*
347 	 * We do the test below only for project quotas. 'usrquota' and
348 	 * 'grpquota' mount options are allowed even without quota feature
349 	 * to support legacy quotas in quota files.
350 	 */
351 	if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
352 		f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
353 		return -1;
354 	}
355 	if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
356 			F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
357 			F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
358 		if (test_opt(sbi, USRQUOTA) &&
359 				F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
360 			clear_opt(sbi, USRQUOTA);
361 
362 		if (test_opt(sbi, GRPQUOTA) &&
363 				F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
364 			clear_opt(sbi, GRPQUOTA);
365 
366 		if (test_opt(sbi, PRJQUOTA) &&
367 				F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
368 			clear_opt(sbi, PRJQUOTA);
369 
370 		if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
371 				test_opt(sbi, PRJQUOTA)) {
372 			f2fs_err(sbi, "old and new quota format mixing");
373 			return -1;
374 		}
375 
376 		if (!F2FS_OPTION(sbi).s_jquota_fmt) {
377 			f2fs_err(sbi, "journaled quota format not specified");
378 			return -1;
379 		}
380 	}
381 
382 	if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
383 		f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
384 		F2FS_OPTION(sbi).s_jquota_fmt = 0;
385 	}
386 	return 0;
387 }
388 #endif
389 
390 static int parse_options(struct super_block *sb, char *options)
391 {
392 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
393 	substring_t args[MAX_OPT_ARGS];
394 	char *p, *name;
395 	int arg = 0;
396 	kuid_t uid;
397 	kgid_t gid;
398 #ifdef CONFIG_QUOTA
399 	int ret;
400 #endif
401 
402 	if (!options)
403 		return 0;
404 
405 	while ((p = strsep(&options, ",")) != NULL) {
406 		int token;
407 		if (!*p)
408 			continue;
409 		/*
410 		 * Initialize args struct so we know whether arg was
411 		 * found; some options take optional arguments.
412 		 */
413 		args[0].to = args[0].from = NULL;
414 		token = match_token(p, f2fs_tokens, args);
415 
416 		switch (token) {
417 		case Opt_gc_background:
418 			name = match_strdup(&args[0]);
419 
420 			if (!name)
421 				return -ENOMEM;
422 			if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
423 				set_opt(sbi, BG_GC);
424 				clear_opt(sbi, FORCE_FG_GC);
425 			} else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
426 				clear_opt(sbi, BG_GC);
427 				clear_opt(sbi, FORCE_FG_GC);
428 			} else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
429 				set_opt(sbi, BG_GC);
430 				set_opt(sbi, FORCE_FG_GC);
431 			} else {
432 				kvfree(name);
433 				return -EINVAL;
434 			}
435 			kvfree(name);
436 			break;
437 		case Opt_disable_roll_forward:
438 			set_opt(sbi, DISABLE_ROLL_FORWARD);
439 			break;
440 		case Opt_norecovery:
441 			/* this option mounts f2fs with ro */
442 			set_opt(sbi, DISABLE_ROLL_FORWARD);
443 			if (!f2fs_readonly(sb))
444 				return -EINVAL;
445 			break;
446 		case Opt_discard:
447 			set_opt(sbi, DISCARD);
448 			break;
449 		case Opt_nodiscard:
450 			if (f2fs_sb_has_blkzoned(sbi)) {
451 				f2fs_warn(sbi, "discard is required for zoned block devices");
452 				return -EINVAL;
453 			}
454 			clear_opt(sbi, DISCARD);
455 			break;
456 		case Opt_noheap:
457 			set_opt(sbi, NOHEAP);
458 			break;
459 		case Opt_heap:
460 			clear_opt(sbi, NOHEAP);
461 			break;
462 #ifdef CONFIG_F2FS_FS_XATTR
463 		case Opt_user_xattr:
464 			set_opt(sbi, XATTR_USER);
465 			break;
466 		case Opt_nouser_xattr:
467 			clear_opt(sbi, XATTR_USER);
468 			break;
469 		case Opt_inline_xattr:
470 			set_opt(sbi, INLINE_XATTR);
471 			break;
472 		case Opt_noinline_xattr:
473 			clear_opt(sbi, INLINE_XATTR);
474 			break;
475 		case Opt_inline_xattr_size:
476 			if (args->from && match_int(args, &arg))
477 				return -EINVAL;
478 			set_opt(sbi, INLINE_XATTR_SIZE);
479 			F2FS_OPTION(sbi).inline_xattr_size = arg;
480 			break;
481 #else
482 		case Opt_user_xattr:
483 			f2fs_info(sbi, "user_xattr options not supported");
484 			break;
485 		case Opt_nouser_xattr:
486 			f2fs_info(sbi, "nouser_xattr options not supported");
487 			break;
488 		case Opt_inline_xattr:
489 			f2fs_info(sbi, "inline_xattr options not supported");
490 			break;
491 		case Opt_noinline_xattr:
492 			f2fs_info(sbi, "noinline_xattr options not supported");
493 			break;
494 #endif
495 #ifdef CONFIG_F2FS_FS_POSIX_ACL
496 		case Opt_acl:
497 			set_opt(sbi, POSIX_ACL);
498 			break;
499 		case Opt_noacl:
500 			clear_opt(sbi, POSIX_ACL);
501 			break;
502 #else
503 		case Opt_acl:
504 			f2fs_info(sbi, "acl options not supported");
505 			break;
506 		case Opt_noacl:
507 			f2fs_info(sbi, "noacl options not supported");
508 			break;
509 #endif
510 		case Opt_active_logs:
511 			if (args->from && match_int(args, &arg))
512 				return -EINVAL;
513 			if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
514 				return -EINVAL;
515 			F2FS_OPTION(sbi).active_logs = arg;
516 			break;
517 		case Opt_disable_ext_identify:
518 			set_opt(sbi, DISABLE_EXT_IDENTIFY);
519 			break;
520 		case Opt_inline_data:
521 			set_opt(sbi, INLINE_DATA);
522 			break;
523 		case Opt_inline_dentry:
524 			set_opt(sbi, INLINE_DENTRY);
525 			break;
526 		case Opt_noinline_dentry:
527 			clear_opt(sbi, INLINE_DENTRY);
528 			break;
529 		case Opt_flush_merge:
530 			set_opt(sbi, FLUSH_MERGE);
531 			break;
532 		case Opt_noflush_merge:
533 			clear_opt(sbi, FLUSH_MERGE);
534 			break;
535 		case Opt_nobarrier:
536 			set_opt(sbi, NOBARRIER);
537 			break;
538 		case Opt_fastboot:
539 			set_opt(sbi, FASTBOOT);
540 			break;
541 		case Opt_extent_cache:
542 			set_opt(sbi, EXTENT_CACHE);
543 			break;
544 		case Opt_noextent_cache:
545 			clear_opt(sbi, EXTENT_CACHE);
546 			break;
547 		case Opt_noinline_data:
548 			clear_opt(sbi, INLINE_DATA);
549 			break;
550 		case Opt_data_flush:
551 			set_opt(sbi, DATA_FLUSH);
552 			break;
553 		case Opt_reserve_root:
554 			if (args->from && match_int(args, &arg))
555 				return -EINVAL;
556 			if (test_opt(sbi, RESERVE_ROOT)) {
557 				f2fs_info(sbi, "Preserve previous reserve_root=%u",
558 					  F2FS_OPTION(sbi).root_reserved_blocks);
559 			} else {
560 				F2FS_OPTION(sbi).root_reserved_blocks = arg;
561 				set_opt(sbi, RESERVE_ROOT);
562 			}
563 			break;
564 		case Opt_resuid:
565 			if (args->from && match_int(args, &arg))
566 				return -EINVAL;
567 			uid = make_kuid(current_user_ns(), arg);
568 			if (!uid_valid(uid)) {
569 				f2fs_err(sbi, "Invalid uid value %d", arg);
570 				return -EINVAL;
571 			}
572 			F2FS_OPTION(sbi).s_resuid = uid;
573 			break;
574 		case Opt_resgid:
575 			if (args->from && match_int(args, &arg))
576 				return -EINVAL;
577 			gid = make_kgid(current_user_ns(), arg);
578 			if (!gid_valid(gid)) {
579 				f2fs_err(sbi, "Invalid gid value %d", arg);
580 				return -EINVAL;
581 			}
582 			F2FS_OPTION(sbi).s_resgid = gid;
583 			break;
584 		case Opt_mode:
585 			name = match_strdup(&args[0]);
586 
587 			if (!name)
588 				return -ENOMEM;
589 			if (strlen(name) == 8 &&
590 					!strncmp(name, "adaptive", 8)) {
591 				if (f2fs_sb_has_blkzoned(sbi)) {
592 					f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
593 					kvfree(name);
594 					return -EINVAL;
595 				}
596 				set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
597 			} else if (strlen(name) == 3 &&
598 					!strncmp(name, "lfs", 3)) {
599 				set_opt_mode(sbi, F2FS_MOUNT_LFS);
600 			} else {
601 				kvfree(name);
602 				return -EINVAL;
603 			}
604 			kvfree(name);
605 			break;
606 		case Opt_io_size_bits:
607 			if (args->from && match_int(args, &arg))
608 				return -EINVAL;
609 			if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_PAGES)) {
610 				f2fs_warn(sbi, "Not support %d, larger than %d",
611 					  1 << arg, BIO_MAX_PAGES);
612 				return -EINVAL;
613 			}
614 			F2FS_OPTION(sbi).write_io_size_bits = arg;
615 			break;
616 #ifdef CONFIG_F2FS_FAULT_INJECTION
617 		case Opt_fault_injection:
618 			if (args->from && match_int(args, &arg))
619 				return -EINVAL;
620 			f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
621 			set_opt(sbi, FAULT_INJECTION);
622 			break;
623 
624 		case Opt_fault_type:
625 			if (args->from && match_int(args, &arg))
626 				return -EINVAL;
627 			f2fs_build_fault_attr(sbi, 0, arg);
628 			set_opt(sbi, FAULT_INJECTION);
629 			break;
630 #else
631 		case Opt_fault_injection:
632 			f2fs_info(sbi, "fault_injection options not supported");
633 			break;
634 
635 		case Opt_fault_type:
636 			f2fs_info(sbi, "fault_type options not supported");
637 			break;
638 #endif
639 		case Opt_lazytime:
640 			sb->s_flags |= SB_LAZYTIME;
641 			break;
642 		case Opt_nolazytime:
643 			sb->s_flags &= ~SB_LAZYTIME;
644 			break;
645 #ifdef CONFIG_QUOTA
646 		case Opt_quota:
647 		case Opt_usrquota:
648 			set_opt(sbi, USRQUOTA);
649 			break;
650 		case Opt_grpquota:
651 			set_opt(sbi, GRPQUOTA);
652 			break;
653 		case Opt_prjquota:
654 			set_opt(sbi, PRJQUOTA);
655 			break;
656 		case Opt_usrjquota:
657 			ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
658 			if (ret)
659 				return ret;
660 			break;
661 		case Opt_grpjquota:
662 			ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
663 			if (ret)
664 				return ret;
665 			break;
666 		case Opt_prjjquota:
667 			ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
668 			if (ret)
669 				return ret;
670 			break;
671 		case Opt_offusrjquota:
672 			ret = f2fs_clear_qf_name(sb, USRQUOTA);
673 			if (ret)
674 				return ret;
675 			break;
676 		case Opt_offgrpjquota:
677 			ret = f2fs_clear_qf_name(sb, GRPQUOTA);
678 			if (ret)
679 				return ret;
680 			break;
681 		case Opt_offprjjquota:
682 			ret = f2fs_clear_qf_name(sb, PRJQUOTA);
683 			if (ret)
684 				return ret;
685 			break;
686 		case Opt_jqfmt_vfsold:
687 			F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
688 			break;
689 		case Opt_jqfmt_vfsv0:
690 			F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
691 			break;
692 		case Opt_jqfmt_vfsv1:
693 			F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
694 			break;
695 		case Opt_noquota:
696 			clear_opt(sbi, QUOTA);
697 			clear_opt(sbi, USRQUOTA);
698 			clear_opt(sbi, GRPQUOTA);
699 			clear_opt(sbi, PRJQUOTA);
700 			break;
701 #else
702 		case Opt_quota:
703 		case Opt_usrquota:
704 		case Opt_grpquota:
705 		case Opt_prjquota:
706 		case Opt_usrjquota:
707 		case Opt_grpjquota:
708 		case Opt_prjjquota:
709 		case Opt_offusrjquota:
710 		case Opt_offgrpjquota:
711 		case Opt_offprjjquota:
712 		case Opt_jqfmt_vfsold:
713 		case Opt_jqfmt_vfsv0:
714 		case Opt_jqfmt_vfsv1:
715 		case Opt_noquota:
716 			f2fs_info(sbi, "quota operations not supported");
717 			break;
718 #endif
719 		case Opt_whint:
720 			name = match_strdup(&args[0]);
721 			if (!name)
722 				return -ENOMEM;
723 			if (strlen(name) == 10 &&
724 					!strncmp(name, "user-based", 10)) {
725 				F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
726 			} else if (strlen(name) == 3 &&
727 					!strncmp(name, "off", 3)) {
728 				F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
729 			} else if (strlen(name) == 8 &&
730 					!strncmp(name, "fs-based", 8)) {
731 				F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
732 			} else {
733 				kvfree(name);
734 				return -EINVAL;
735 			}
736 			kvfree(name);
737 			break;
738 		case Opt_alloc:
739 			name = match_strdup(&args[0]);
740 			if (!name)
741 				return -ENOMEM;
742 
743 			if (strlen(name) == 7 &&
744 					!strncmp(name, "default", 7)) {
745 				F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
746 			} else if (strlen(name) == 5 &&
747 					!strncmp(name, "reuse", 5)) {
748 				F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
749 			} else {
750 				kvfree(name);
751 				return -EINVAL;
752 			}
753 			kvfree(name);
754 			break;
755 		case Opt_fsync:
756 			name = match_strdup(&args[0]);
757 			if (!name)
758 				return -ENOMEM;
759 			if (strlen(name) == 5 &&
760 					!strncmp(name, "posix", 5)) {
761 				F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
762 			} else if (strlen(name) == 6 &&
763 					!strncmp(name, "strict", 6)) {
764 				F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
765 			} else if (strlen(name) == 9 &&
766 					!strncmp(name, "nobarrier", 9)) {
767 				F2FS_OPTION(sbi).fsync_mode =
768 							FSYNC_MODE_NOBARRIER;
769 			} else {
770 				kvfree(name);
771 				return -EINVAL;
772 			}
773 			kvfree(name);
774 			break;
775 		case Opt_test_dummy_encryption:
776 #ifdef CONFIG_FS_ENCRYPTION
777 			if (!f2fs_sb_has_encrypt(sbi)) {
778 				f2fs_err(sbi, "Encrypt feature is off");
779 				return -EINVAL;
780 			}
781 
782 			F2FS_OPTION(sbi).test_dummy_encryption = true;
783 			f2fs_info(sbi, "Test dummy encryption mode enabled");
784 #else
785 			f2fs_info(sbi, "Test dummy encryption mount option ignored");
786 #endif
787 			break;
788 		case Opt_checkpoint_disable_cap_perc:
789 			if (args->from && match_int(args, &arg))
790 				return -EINVAL;
791 			if (arg < 0 || arg > 100)
792 				return -EINVAL;
793 			if (arg == 100)
794 				F2FS_OPTION(sbi).unusable_cap =
795 					sbi->user_block_count;
796 			else
797 				F2FS_OPTION(sbi).unusable_cap =
798 					(sbi->user_block_count / 100) *	arg;
799 			set_opt(sbi, DISABLE_CHECKPOINT);
800 			break;
801 		case Opt_checkpoint_disable_cap:
802 			if (args->from && match_int(args, &arg))
803 				return -EINVAL;
804 			F2FS_OPTION(sbi).unusable_cap = arg;
805 			set_opt(sbi, DISABLE_CHECKPOINT);
806 			break;
807 		case Opt_checkpoint_disable:
808 			set_opt(sbi, DISABLE_CHECKPOINT);
809 			break;
810 		case Opt_checkpoint_enable:
811 			clear_opt(sbi, DISABLE_CHECKPOINT);
812 			break;
813 		default:
814 			f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
815 				 p);
816 			return -EINVAL;
817 		}
818 	}
819 #ifdef CONFIG_QUOTA
820 	if (f2fs_check_quota_options(sbi))
821 		return -EINVAL;
822 #else
823 	if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
824 		f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
825 		return -EINVAL;
826 	}
827 	if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
828 		f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
829 		return -EINVAL;
830 	}
831 #endif
832 #ifndef CONFIG_UNICODE
833 	if (f2fs_sb_has_casefold(sbi)) {
834 		f2fs_err(sbi,
835 			"Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
836 		return -EINVAL;
837 	}
838 #endif
839 
840 	if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) {
841 		f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
842 			 F2FS_IO_SIZE_KB(sbi));
843 		return -EINVAL;
844 	}
845 
846 	if (test_opt(sbi, INLINE_XATTR_SIZE)) {
847 		int min_size, max_size;
848 
849 		if (!f2fs_sb_has_extra_attr(sbi) ||
850 			!f2fs_sb_has_flexible_inline_xattr(sbi)) {
851 			f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
852 			return -EINVAL;
853 		}
854 		if (!test_opt(sbi, INLINE_XATTR)) {
855 			f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
856 			return -EINVAL;
857 		}
858 
859 		min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
860 		max_size = MAX_INLINE_XATTR_SIZE;
861 
862 		if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
863 				F2FS_OPTION(sbi).inline_xattr_size > max_size) {
864 			f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
865 				 min_size, max_size);
866 			return -EINVAL;
867 		}
868 	}
869 
870 	if (test_opt(sbi, DISABLE_CHECKPOINT) && test_opt(sbi, LFS)) {
871 		f2fs_err(sbi, "LFS not compatible with checkpoint=disable\n");
872 		return -EINVAL;
873 	}
874 
875 	/* Not pass down write hints if the number of active logs is lesser
876 	 * than NR_CURSEG_TYPE.
877 	 */
878 	if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
879 		F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
880 	return 0;
881 }
882 
883 static struct inode *f2fs_alloc_inode(struct super_block *sb)
884 {
885 	struct f2fs_inode_info *fi;
886 
887 	fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
888 	if (!fi)
889 		return NULL;
890 
891 	init_once((void *) fi);
892 
893 	/* Initialize f2fs-specific inode info */
894 	atomic_set(&fi->dirty_pages, 0);
895 	init_rwsem(&fi->i_sem);
896 	INIT_LIST_HEAD(&fi->dirty_list);
897 	INIT_LIST_HEAD(&fi->gdirty_list);
898 	INIT_LIST_HEAD(&fi->inmem_ilist);
899 	INIT_LIST_HEAD(&fi->inmem_pages);
900 	mutex_init(&fi->inmem_lock);
901 	init_rwsem(&fi->i_gc_rwsem[READ]);
902 	init_rwsem(&fi->i_gc_rwsem[WRITE]);
903 	init_rwsem(&fi->i_mmap_sem);
904 	init_rwsem(&fi->i_xattr_sem);
905 
906 	/* Will be used by directory only */
907 	fi->i_dir_level = F2FS_SB(sb)->dir_level;
908 
909 	return &fi->vfs_inode;
910 }
911 
912 static int f2fs_drop_inode(struct inode *inode)
913 {
914 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
915 	int ret;
916 
917 	/*
918 	 * during filesystem shutdown, if checkpoint is disabled,
919 	 * drop useless meta/node dirty pages.
920 	 */
921 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
922 		if (inode->i_ino == F2FS_NODE_INO(sbi) ||
923 			inode->i_ino == F2FS_META_INO(sbi)) {
924 			trace_f2fs_drop_inode(inode, 1);
925 			return 1;
926 		}
927 	}
928 
929 	/*
930 	 * This is to avoid a deadlock condition like below.
931 	 * writeback_single_inode(inode)
932 	 *  - f2fs_write_data_page
933 	 *    - f2fs_gc -> iput -> evict
934 	 *       - inode_wait_for_writeback(inode)
935 	 */
936 	if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
937 		if (!inode->i_nlink && !is_bad_inode(inode)) {
938 			/* to avoid evict_inode call simultaneously */
939 			atomic_inc(&inode->i_count);
940 			spin_unlock(&inode->i_lock);
941 
942 			/* some remained atomic pages should discarded */
943 			if (f2fs_is_atomic_file(inode))
944 				f2fs_drop_inmem_pages(inode);
945 
946 			/* should remain fi->extent_tree for writepage */
947 			f2fs_destroy_extent_node(inode);
948 
949 			sb_start_intwrite(inode->i_sb);
950 			f2fs_i_size_write(inode, 0);
951 
952 			f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
953 					inode, NULL, 0, DATA);
954 			truncate_inode_pages_final(inode->i_mapping);
955 
956 			if (F2FS_HAS_BLOCKS(inode))
957 				f2fs_truncate(inode);
958 
959 			sb_end_intwrite(inode->i_sb);
960 
961 			spin_lock(&inode->i_lock);
962 			atomic_dec(&inode->i_count);
963 		}
964 		trace_f2fs_drop_inode(inode, 0);
965 		return 0;
966 	}
967 	ret = generic_drop_inode(inode);
968 	if (!ret)
969 		ret = fscrypt_drop_inode(inode);
970 	trace_f2fs_drop_inode(inode, ret);
971 	return ret;
972 }
973 
974 int f2fs_inode_dirtied(struct inode *inode, bool sync)
975 {
976 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
977 	int ret = 0;
978 
979 	spin_lock(&sbi->inode_lock[DIRTY_META]);
980 	if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
981 		ret = 1;
982 	} else {
983 		set_inode_flag(inode, FI_DIRTY_INODE);
984 		stat_inc_dirty_inode(sbi, DIRTY_META);
985 	}
986 	if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
987 		list_add_tail(&F2FS_I(inode)->gdirty_list,
988 				&sbi->inode_list[DIRTY_META]);
989 		inc_page_count(sbi, F2FS_DIRTY_IMETA);
990 	}
991 	spin_unlock(&sbi->inode_lock[DIRTY_META]);
992 	return ret;
993 }
994 
995 void f2fs_inode_synced(struct inode *inode)
996 {
997 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
998 
999 	spin_lock(&sbi->inode_lock[DIRTY_META]);
1000 	if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1001 		spin_unlock(&sbi->inode_lock[DIRTY_META]);
1002 		return;
1003 	}
1004 	if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1005 		list_del_init(&F2FS_I(inode)->gdirty_list);
1006 		dec_page_count(sbi, F2FS_DIRTY_IMETA);
1007 	}
1008 	clear_inode_flag(inode, FI_DIRTY_INODE);
1009 	clear_inode_flag(inode, FI_AUTO_RECOVER);
1010 	stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1011 	spin_unlock(&sbi->inode_lock[DIRTY_META]);
1012 }
1013 
1014 /*
1015  * f2fs_dirty_inode() is called from __mark_inode_dirty()
1016  *
1017  * We should call set_dirty_inode to write the dirty inode through write_inode.
1018  */
1019 static void f2fs_dirty_inode(struct inode *inode, int flags)
1020 {
1021 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1022 
1023 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1024 			inode->i_ino == F2FS_META_INO(sbi))
1025 		return;
1026 
1027 	if (flags == I_DIRTY_TIME)
1028 		return;
1029 
1030 	if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1031 		clear_inode_flag(inode, FI_AUTO_RECOVER);
1032 
1033 	f2fs_inode_dirtied(inode, false);
1034 }
1035 
1036 static void f2fs_free_inode(struct inode *inode)
1037 {
1038 	fscrypt_free_inode(inode);
1039 	kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1040 }
1041 
1042 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1043 {
1044 	percpu_counter_destroy(&sbi->alloc_valid_block_count);
1045 	percpu_counter_destroy(&sbi->total_valid_inode_count);
1046 }
1047 
1048 static void destroy_device_list(struct f2fs_sb_info *sbi)
1049 {
1050 	int i;
1051 
1052 	for (i = 0; i < sbi->s_ndevs; i++) {
1053 		blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1054 #ifdef CONFIG_BLK_DEV_ZONED
1055 		kvfree(FDEV(i).blkz_seq);
1056 #endif
1057 	}
1058 	kvfree(sbi->devs);
1059 }
1060 
1061 static void f2fs_put_super(struct super_block *sb)
1062 {
1063 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1064 	int i;
1065 	bool dropped;
1066 
1067 	f2fs_quota_off_umount(sb);
1068 
1069 	/* prevent remaining shrinker jobs */
1070 	mutex_lock(&sbi->umount_mutex);
1071 
1072 	/*
1073 	 * We don't need to do checkpoint when superblock is clean.
1074 	 * But, the previous checkpoint was not done by umount, it needs to do
1075 	 * clean checkpoint again.
1076 	 */
1077 	if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1078 			!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1079 		struct cp_control cpc = {
1080 			.reason = CP_UMOUNT,
1081 		};
1082 		f2fs_write_checkpoint(sbi, &cpc);
1083 	}
1084 
1085 	/* be sure to wait for any on-going discard commands */
1086 	dropped = f2fs_issue_discard_timeout(sbi);
1087 
1088 	if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1089 					!sbi->discard_blks && !dropped) {
1090 		struct cp_control cpc = {
1091 			.reason = CP_UMOUNT | CP_TRIMMED,
1092 		};
1093 		f2fs_write_checkpoint(sbi, &cpc);
1094 	}
1095 
1096 	/*
1097 	 * normally superblock is clean, so we need to release this.
1098 	 * In addition, EIO will skip do checkpoint, we need this as well.
1099 	 */
1100 	f2fs_release_ino_entry(sbi, true);
1101 
1102 	f2fs_leave_shrinker(sbi);
1103 	mutex_unlock(&sbi->umount_mutex);
1104 
1105 	/* our cp_error case, we can wait for any writeback page */
1106 	f2fs_flush_merged_writes(sbi);
1107 
1108 	f2fs_wait_on_all_pages_writeback(sbi);
1109 
1110 	f2fs_bug_on(sbi, sbi->fsync_node_num);
1111 
1112 	iput(sbi->node_inode);
1113 	sbi->node_inode = NULL;
1114 
1115 	iput(sbi->meta_inode);
1116 	sbi->meta_inode = NULL;
1117 
1118 	/*
1119 	 * iput() can update stat information, if f2fs_write_checkpoint()
1120 	 * above failed with error.
1121 	 */
1122 	f2fs_destroy_stats(sbi);
1123 
1124 	/* destroy f2fs internal modules */
1125 	f2fs_destroy_node_manager(sbi);
1126 	f2fs_destroy_segment_manager(sbi);
1127 
1128 	kvfree(sbi->ckpt);
1129 
1130 	f2fs_unregister_sysfs(sbi);
1131 
1132 	sb->s_fs_info = NULL;
1133 	if (sbi->s_chksum_driver)
1134 		crypto_free_shash(sbi->s_chksum_driver);
1135 	kvfree(sbi->raw_super);
1136 
1137 	destroy_device_list(sbi);
1138 	mempool_destroy(sbi->write_io_dummy);
1139 #ifdef CONFIG_QUOTA
1140 	for (i = 0; i < MAXQUOTAS; i++)
1141 		kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1142 #endif
1143 	destroy_percpu_info(sbi);
1144 	for (i = 0; i < NR_PAGE_TYPE; i++)
1145 		kvfree(sbi->write_io[i]);
1146 #ifdef CONFIG_UNICODE
1147 	utf8_unload(sbi->s_encoding);
1148 #endif
1149 	kvfree(sbi);
1150 }
1151 
1152 int f2fs_sync_fs(struct super_block *sb, int sync)
1153 {
1154 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1155 	int err = 0;
1156 
1157 	if (unlikely(f2fs_cp_error(sbi)))
1158 		return 0;
1159 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1160 		return 0;
1161 
1162 	trace_f2fs_sync_fs(sb, sync);
1163 
1164 	if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1165 		return -EAGAIN;
1166 
1167 	if (sync) {
1168 		struct cp_control cpc;
1169 
1170 		cpc.reason = __get_cp_reason(sbi);
1171 
1172 		mutex_lock(&sbi->gc_mutex);
1173 		err = f2fs_write_checkpoint(sbi, &cpc);
1174 		mutex_unlock(&sbi->gc_mutex);
1175 	}
1176 	f2fs_trace_ios(NULL, 1);
1177 
1178 	return err;
1179 }
1180 
1181 static int f2fs_freeze(struct super_block *sb)
1182 {
1183 	if (f2fs_readonly(sb))
1184 		return 0;
1185 
1186 	/* IO error happened before */
1187 	if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1188 		return -EIO;
1189 
1190 	/* must be clean, since sync_filesystem() was already called */
1191 	if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1192 		return -EINVAL;
1193 	return 0;
1194 }
1195 
1196 static int f2fs_unfreeze(struct super_block *sb)
1197 {
1198 	return 0;
1199 }
1200 
1201 #ifdef CONFIG_QUOTA
1202 static int f2fs_statfs_project(struct super_block *sb,
1203 				kprojid_t projid, struct kstatfs *buf)
1204 {
1205 	struct kqid qid;
1206 	struct dquot *dquot;
1207 	u64 limit;
1208 	u64 curblock;
1209 
1210 	qid = make_kqid_projid(projid);
1211 	dquot = dqget(sb, qid);
1212 	if (IS_ERR(dquot))
1213 		return PTR_ERR(dquot);
1214 	spin_lock(&dquot->dq_dqb_lock);
1215 
1216 	limit = 0;
1217 	if (dquot->dq_dqb.dqb_bsoftlimit)
1218 		limit = dquot->dq_dqb.dqb_bsoftlimit;
1219 	if (dquot->dq_dqb.dqb_bhardlimit &&
1220 			(!limit || dquot->dq_dqb.dqb_bhardlimit < limit))
1221 		limit = dquot->dq_dqb.dqb_bhardlimit;
1222 
1223 	if (limit && buf->f_blocks > limit) {
1224 		curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1225 		buf->f_blocks = limit;
1226 		buf->f_bfree = buf->f_bavail =
1227 			(buf->f_blocks > curblock) ?
1228 			 (buf->f_blocks - curblock) : 0;
1229 	}
1230 
1231 	limit = 0;
1232 	if (dquot->dq_dqb.dqb_isoftlimit)
1233 		limit = dquot->dq_dqb.dqb_isoftlimit;
1234 	if (dquot->dq_dqb.dqb_ihardlimit &&
1235 			(!limit || dquot->dq_dqb.dqb_ihardlimit < limit))
1236 		limit = dquot->dq_dqb.dqb_ihardlimit;
1237 
1238 	if (limit && buf->f_files > limit) {
1239 		buf->f_files = limit;
1240 		buf->f_ffree =
1241 			(buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1242 			 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1243 	}
1244 
1245 	spin_unlock(&dquot->dq_dqb_lock);
1246 	dqput(dquot);
1247 	return 0;
1248 }
1249 #endif
1250 
1251 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1252 {
1253 	struct super_block *sb = dentry->d_sb;
1254 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1255 	u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1256 	block_t total_count, user_block_count, start_count;
1257 	u64 avail_node_count;
1258 
1259 	total_count = le64_to_cpu(sbi->raw_super->block_count);
1260 	user_block_count = sbi->user_block_count;
1261 	start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1262 	buf->f_type = F2FS_SUPER_MAGIC;
1263 	buf->f_bsize = sbi->blocksize;
1264 
1265 	buf->f_blocks = total_count - start_count;
1266 	buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1267 						sbi->current_reserved_blocks;
1268 
1269 	spin_lock(&sbi->stat_lock);
1270 	if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1271 		buf->f_bfree = 0;
1272 	else
1273 		buf->f_bfree -= sbi->unusable_block_count;
1274 	spin_unlock(&sbi->stat_lock);
1275 
1276 	if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1277 		buf->f_bavail = buf->f_bfree -
1278 				F2FS_OPTION(sbi).root_reserved_blocks;
1279 	else
1280 		buf->f_bavail = 0;
1281 
1282 	avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1283 
1284 	if (avail_node_count > user_block_count) {
1285 		buf->f_files = user_block_count;
1286 		buf->f_ffree = buf->f_bavail;
1287 	} else {
1288 		buf->f_files = avail_node_count;
1289 		buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1290 					buf->f_bavail);
1291 	}
1292 
1293 	buf->f_namelen = F2FS_NAME_LEN;
1294 	buf->f_fsid.val[0] = (u32)id;
1295 	buf->f_fsid.val[1] = (u32)(id >> 32);
1296 
1297 #ifdef CONFIG_QUOTA
1298 	if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1299 			sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1300 		f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1301 	}
1302 #endif
1303 	return 0;
1304 }
1305 
1306 static inline void f2fs_show_quota_options(struct seq_file *seq,
1307 					   struct super_block *sb)
1308 {
1309 #ifdef CONFIG_QUOTA
1310 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1311 
1312 	if (F2FS_OPTION(sbi).s_jquota_fmt) {
1313 		char *fmtname = "";
1314 
1315 		switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1316 		case QFMT_VFS_OLD:
1317 			fmtname = "vfsold";
1318 			break;
1319 		case QFMT_VFS_V0:
1320 			fmtname = "vfsv0";
1321 			break;
1322 		case QFMT_VFS_V1:
1323 			fmtname = "vfsv1";
1324 			break;
1325 		}
1326 		seq_printf(seq, ",jqfmt=%s", fmtname);
1327 	}
1328 
1329 	if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1330 		seq_show_option(seq, "usrjquota",
1331 			F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1332 
1333 	if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1334 		seq_show_option(seq, "grpjquota",
1335 			F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1336 
1337 	if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1338 		seq_show_option(seq, "prjjquota",
1339 			F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1340 #endif
1341 }
1342 
1343 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1344 {
1345 	struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1346 
1347 	if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1348 		if (test_opt(sbi, FORCE_FG_GC))
1349 			seq_printf(seq, ",background_gc=%s", "sync");
1350 		else
1351 			seq_printf(seq, ",background_gc=%s", "on");
1352 	} else {
1353 		seq_printf(seq, ",background_gc=%s", "off");
1354 	}
1355 	if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1356 		seq_puts(seq, ",disable_roll_forward");
1357 	if (test_opt(sbi, DISCARD))
1358 		seq_puts(seq, ",discard");
1359 	else
1360 		seq_puts(seq, ",nodiscard");
1361 	if (test_opt(sbi, NOHEAP))
1362 		seq_puts(seq, ",no_heap");
1363 	else
1364 		seq_puts(seq, ",heap");
1365 #ifdef CONFIG_F2FS_FS_XATTR
1366 	if (test_opt(sbi, XATTR_USER))
1367 		seq_puts(seq, ",user_xattr");
1368 	else
1369 		seq_puts(seq, ",nouser_xattr");
1370 	if (test_opt(sbi, INLINE_XATTR))
1371 		seq_puts(seq, ",inline_xattr");
1372 	else
1373 		seq_puts(seq, ",noinline_xattr");
1374 	if (test_opt(sbi, INLINE_XATTR_SIZE))
1375 		seq_printf(seq, ",inline_xattr_size=%u",
1376 					F2FS_OPTION(sbi).inline_xattr_size);
1377 #endif
1378 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1379 	if (test_opt(sbi, POSIX_ACL))
1380 		seq_puts(seq, ",acl");
1381 	else
1382 		seq_puts(seq, ",noacl");
1383 #endif
1384 	if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1385 		seq_puts(seq, ",disable_ext_identify");
1386 	if (test_opt(sbi, INLINE_DATA))
1387 		seq_puts(seq, ",inline_data");
1388 	else
1389 		seq_puts(seq, ",noinline_data");
1390 	if (test_opt(sbi, INLINE_DENTRY))
1391 		seq_puts(seq, ",inline_dentry");
1392 	else
1393 		seq_puts(seq, ",noinline_dentry");
1394 	if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1395 		seq_puts(seq, ",flush_merge");
1396 	if (test_opt(sbi, NOBARRIER))
1397 		seq_puts(seq, ",nobarrier");
1398 	if (test_opt(sbi, FASTBOOT))
1399 		seq_puts(seq, ",fastboot");
1400 	if (test_opt(sbi, EXTENT_CACHE))
1401 		seq_puts(seq, ",extent_cache");
1402 	else
1403 		seq_puts(seq, ",noextent_cache");
1404 	if (test_opt(sbi, DATA_FLUSH))
1405 		seq_puts(seq, ",data_flush");
1406 
1407 	seq_puts(seq, ",mode=");
1408 	if (test_opt(sbi, ADAPTIVE))
1409 		seq_puts(seq, "adaptive");
1410 	else if (test_opt(sbi, LFS))
1411 		seq_puts(seq, "lfs");
1412 	seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1413 	if (test_opt(sbi, RESERVE_ROOT))
1414 		seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1415 				F2FS_OPTION(sbi).root_reserved_blocks,
1416 				from_kuid_munged(&init_user_ns,
1417 					F2FS_OPTION(sbi).s_resuid),
1418 				from_kgid_munged(&init_user_ns,
1419 					F2FS_OPTION(sbi).s_resgid));
1420 	if (F2FS_IO_SIZE_BITS(sbi))
1421 		seq_printf(seq, ",io_bits=%u",
1422 				F2FS_OPTION(sbi).write_io_size_bits);
1423 #ifdef CONFIG_F2FS_FAULT_INJECTION
1424 	if (test_opt(sbi, FAULT_INJECTION)) {
1425 		seq_printf(seq, ",fault_injection=%u",
1426 				F2FS_OPTION(sbi).fault_info.inject_rate);
1427 		seq_printf(seq, ",fault_type=%u",
1428 				F2FS_OPTION(sbi).fault_info.inject_type);
1429 	}
1430 #endif
1431 #ifdef CONFIG_QUOTA
1432 	if (test_opt(sbi, QUOTA))
1433 		seq_puts(seq, ",quota");
1434 	if (test_opt(sbi, USRQUOTA))
1435 		seq_puts(seq, ",usrquota");
1436 	if (test_opt(sbi, GRPQUOTA))
1437 		seq_puts(seq, ",grpquota");
1438 	if (test_opt(sbi, PRJQUOTA))
1439 		seq_puts(seq, ",prjquota");
1440 #endif
1441 	f2fs_show_quota_options(seq, sbi->sb);
1442 	if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1443 		seq_printf(seq, ",whint_mode=%s", "user-based");
1444 	else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1445 		seq_printf(seq, ",whint_mode=%s", "fs-based");
1446 #ifdef CONFIG_FS_ENCRYPTION
1447 	if (F2FS_OPTION(sbi).test_dummy_encryption)
1448 		seq_puts(seq, ",test_dummy_encryption");
1449 #endif
1450 
1451 	if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1452 		seq_printf(seq, ",alloc_mode=%s", "default");
1453 	else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1454 		seq_printf(seq, ",alloc_mode=%s", "reuse");
1455 
1456 	if (test_opt(sbi, DISABLE_CHECKPOINT))
1457 		seq_printf(seq, ",checkpoint=disable:%u",
1458 				F2FS_OPTION(sbi).unusable_cap);
1459 	if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1460 		seq_printf(seq, ",fsync_mode=%s", "posix");
1461 	else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1462 		seq_printf(seq, ",fsync_mode=%s", "strict");
1463 	else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1464 		seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1465 	return 0;
1466 }
1467 
1468 static void default_options(struct f2fs_sb_info *sbi)
1469 {
1470 	/* init some FS parameters */
1471 	F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1472 	F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1473 	F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1474 	F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1475 	F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1476 	F2FS_OPTION(sbi).test_dummy_encryption = false;
1477 	F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1478 	F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1479 
1480 	set_opt(sbi, BG_GC);
1481 	set_opt(sbi, INLINE_XATTR);
1482 	set_opt(sbi, INLINE_DATA);
1483 	set_opt(sbi, INLINE_DENTRY);
1484 	set_opt(sbi, EXTENT_CACHE);
1485 	set_opt(sbi, NOHEAP);
1486 	clear_opt(sbi, DISABLE_CHECKPOINT);
1487 	F2FS_OPTION(sbi).unusable_cap = 0;
1488 	sbi->sb->s_flags |= SB_LAZYTIME;
1489 	set_opt(sbi, FLUSH_MERGE);
1490 	set_opt(sbi, DISCARD);
1491 	if (f2fs_sb_has_blkzoned(sbi))
1492 		set_opt_mode(sbi, F2FS_MOUNT_LFS);
1493 	else
1494 		set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1495 
1496 #ifdef CONFIG_F2FS_FS_XATTR
1497 	set_opt(sbi, XATTR_USER);
1498 #endif
1499 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1500 	set_opt(sbi, POSIX_ACL);
1501 #endif
1502 
1503 	f2fs_build_fault_attr(sbi, 0, 0);
1504 }
1505 
1506 #ifdef CONFIG_QUOTA
1507 static int f2fs_enable_quotas(struct super_block *sb);
1508 #endif
1509 
1510 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1511 {
1512 	unsigned int s_flags = sbi->sb->s_flags;
1513 	struct cp_control cpc;
1514 	int err = 0;
1515 	int ret;
1516 	block_t unusable;
1517 
1518 	if (s_flags & SB_RDONLY) {
1519 		f2fs_err(sbi, "checkpoint=disable on readonly fs");
1520 		return -EINVAL;
1521 	}
1522 	sbi->sb->s_flags |= SB_ACTIVE;
1523 
1524 	f2fs_update_time(sbi, DISABLE_TIME);
1525 
1526 	while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1527 		mutex_lock(&sbi->gc_mutex);
1528 		err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1529 		if (err == -ENODATA) {
1530 			err = 0;
1531 			break;
1532 		}
1533 		if (err && err != -EAGAIN)
1534 			break;
1535 	}
1536 
1537 	ret = sync_filesystem(sbi->sb);
1538 	if (ret || err) {
1539 		err = ret ? ret: err;
1540 		goto restore_flag;
1541 	}
1542 
1543 	unusable = f2fs_get_unusable_blocks(sbi);
1544 	if (f2fs_disable_cp_again(sbi, unusable)) {
1545 		err = -EAGAIN;
1546 		goto restore_flag;
1547 	}
1548 
1549 	mutex_lock(&sbi->gc_mutex);
1550 	cpc.reason = CP_PAUSE;
1551 	set_sbi_flag(sbi, SBI_CP_DISABLED);
1552 	err = f2fs_write_checkpoint(sbi, &cpc);
1553 	if (err)
1554 		goto out_unlock;
1555 
1556 	spin_lock(&sbi->stat_lock);
1557 	sbi->unusable_block_count = unusable;
1558 	spin_unlock(&sbi->stat_lock);
1559 
1560 out_unlock:
1561 	mutex_unlock(&sbi->gc_mutex);
1562 restore_flag:
1563 	sbi->sb->s_flags = s_flags;	/* Restore MS_RDONLY status */
1564 	return err;
1565 }
1566 
1567 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1568 {
1569 	mutex_lock(&sbi->gc_mutex);
1570 	f2fs_dirty_to_prefree(sbi);
1571 
1572 	clear_sbi_flag(sbi, SBI_CP_DISABLED);
1573 	set_sbi_flag(sbi, SBI_IS_DIRTY);
1574 	mutex_unlock(&sbi->gc_mutex);
1575 
1576 	f2fs_sync_fs(sbi->sb, 1);
1577 }
1578 
1579 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1580 {
1581 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1582 	struct f2fs_mount_info org_mount_opt;
1583 	unsigned long old_sb_flags;
1584 	int err;
1585 	bool need_restart_gc = false;
1586 	bool need_stop_gc = false;
1587 	bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1588 	bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1589 	bool no_io_align = !F2FS_IO_ALIGNED(sbi);
1590 	bool checkpoint_changed;
1591 #ifdef CONFIG_QUOTA
1592 	int i, j;
1593 #endif
1594 
1595 	/*
1596 	 * Save the old mount options in case we
1597 	 * need to restore them.
1598 	 */
1599 	org_mount_opt = sbi->mount_opt;
1600 	old_sb_flags = sb->s_flags;
1601 
1602 #ifdef CONFIG_QUOTA
1603 	org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1604 	for (i = 0; i < MAXQUOTAS; i++) {
1605 		if (F2FS_OPTION(sbi).s_qf_names[i]) {
1606 			org_mount_opt.s_qf_names[i] =
1607 				kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1608 				GFP_KERNEL);
1609 			if (!org_mount_opt.s_qf_names[i]) {
1610 				for (j = 0; j < i; j++)
1611 					kvfree(org_mount_opt.s_qf_names[j]);
1612 				return -ENOMEM;
1613 			}
1614 		} else {
1615 			org_mount_opt.s_qf_names[i] = NULL;
1616 		}
1617 	}
1618 #endif
1619 
1620 	/* recover superblocks we couldn't write due to previous RO mount */
1621 	if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1622 		err = f2fs_commit_super(sbi, false);
1623 		f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
1624 			  err);
1625 		if (!err)
1626 			clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1627 	}
1628 
1629 	default_options(sbi);
1630 
1631 	/* parse mount options */
1632 	err = parse_options(sb, data);
1633 	if (err)
1634 		goto restore_opts;
1635 	checkpoint_changed =
1636 			disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1637 
1638 	/*
1639 	 * Previous and new state of filesystem is RO,
1640 	 * so skip checking GC and FLUSH_MERGE conditions.
1641 	 */
1642 	if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1643 		goto skip;
1644 
1645 #ifdef CONFIG_QUOTA
1646 	if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1647 		err = dquot_suspend(sb, -1);
1648 		if (err < 0)
1649 			goto restore_opts;
1650 	} else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1651 		/* dquot_resume needs RW */
1652 		sb->s_flags &= ~SB_RDONLY;
1653 		if (sb_any_quota_suspended(sb)) {
1654 			dquot_resume(sb, -1);
1655 		} else if (f2fs_sb_has_quota_ino(sbi)) {
1656 			err = f2fs_enable_quotas(sb);
1657 			if (err)
1658 				goto restore_opts;
1659 		}
1660 	}
1661 #endif
1662 	/* disallow enable/disable extent_cache dynamically */
1663 	if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1664 		err = -EINVAL;
1665 		f2fs_warn(sbi, "switch extent_cache option is not allowed");
1666 		goto restore_opts;
1667 	}
1668 
1669 	if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
1670 		err = -EINVAL;
1671 		f2fs_warn(sbi, "switch io_bits option is not allowed");
1672 		goto restore_opts;
1673 	}
1674 
1675 	if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1676 		err = -EINVAL;
1677 		f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
1678 		goto restore_opts;
1679 	}
1680 
1681 	/*
1682 	 * We stop the GC thread if FS is mounted as RO
1683 	 * or if background_gc = off is passed in mount
1684 	 * option. Also sync the filesystem.
1685 	 */
1686 	if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) {
1687 		if (sbi->gc_thread) {
1688 			f2fs_stop_gc_thread(sbi);
1689 			need_restart_gc = true;
1690 		}
1691 	} else if (!sbi->gc_thread) {
1692 		err = f2fs_start_gc_thread(sbi);
1693 		if (err)
1694 			goto restore_opts;
1695 		need_stop_gc = true;
1696 	}
1697 
1698 	if (*flags & SB_RDONLY ||
1699 		F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1700 		writeback_inodes_sb(sb, WB_REASON_SYNC);
1701 		sync_inodes_sb(sb);
1702 
1703 		set_sbi_flag(sbi, SBI_IS_DIRTY);
1704 		set_sbi_flag(sbi, SBI_IS_CLOSE);
1705 		f2fs_sync_fs(sb, 1);
1706 		clear_sbi_flag(sbi, SBI_IS_CLOSE);
1707 	}
1708 
1709 	if (checkpoint_changed) {
1710 		if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1711 			err = f2fs_disable_checkpoint(sbi);
1712 			if (err)
1713 				goto restore_gc;
1714 		} else {
1715 			f2fs_enable_checkpoint(sbi);
1716 		}
1717 	}
1718 
1719 	/*
1720 	 * We stop issue flush thread if FS is mounted as RO
1721 	 * or if flush_merge is not passed in mount option.
1722 	 */
1723 	if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1724 		clear_opt(sbi, FLUSH_MERGE);
1725 		f2fs_destroy_flush_cmd_control(sbi, false);
1726 	} else {
1727 		err = f2fs_create_flush_cmd_control(sbi);
1728 		if (err)
1729 			goto restore_gc;
1730 	}
1731 skip:
1732 #ifdef CONFIG_QUOTA
1733 	/* Release old quota file names */
1734 	for (i = 0; i < MAXQUOTAS; i++)
1735 		kvfree(org_mount_opt.s_qf_names[i]);
1736 #endif
1737 	/* Update the POSIXACL Flag */
1738 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1739 		(test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1740 
1741 	limit_reserve_root(sbi);
1742 	*flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1743 	return 0;
1744 restore_gc:
1745 	if (need_restart_gc) {
1746 		if (f2fs_start_gc_thread(sbi))
1747 			f2fs_warn(sbi, "background gc thread has stopped");
1748 	} else if (need_stop_gc) {
1749 		f2fs_stop_gc_thread(sbi);
1750 	}
1751 restore_opts:
1752 #ifdef CONFIG_QUOTA
1753 	F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1754 	for (i = 0; i < MAXQUOTAS; i++) {
1755 		kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1756 		F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1757 	}
1758 #endif
1759 	sbi->mount_opt = org_mount_opt;
1760 	sb->s_flags = old_sb_flags;
1761 	return err;
1762 }
1763 
1764 #ifdef CONFIG_QUOTA
1765 /* Read data from quotafile */
1766 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1767 			       size_t len, loff_t off)
1768 {
1769 	struct inode *inode = sb_dqopt(sb)->files[type];
1770 	struct address_space *mapping = inode->i_mapping;
1771 	block_t blkidx = F2FS_BYTES_TO_BLK(off);
1772 	int offset = off & (sb->s_blocksize - 1);
1773 	int tocopy;
1774 	size_t toread;
1775 	loff_t i_size = i_size_read(inode);
1776 	struct page *page;
1777 	char *kaddr;
1778 
1779 	if (off > i_size)
1780 		return 0;
1781 
1782 	if (off + len > i_size)
1783 		len = i_size - off;
1784 	toread = len;
1785 	while (toread > 0) {
1786 		tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1787 repeat:
1788 		page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1789 		if (IS_ERR(page)) {
1790 			if (PTR_ERR(page) == -ENOMEM) {
1791 				congestion_wait(BLK_RW_ASYNC, HZ/50);
1792 				goto repeat;
1793 			}
1794 			set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1795 			return PTR_ERR(page);
1796 		}
1797 
1798 		lock_page(page);
1799 
1800 		if (unlikely(page->mapping != mapping)) {
1801 			f2fs_put_page(page, 1);
1802 			goto repeat;
1803 		}
1804 		if (unlikely(!PageUptodate(page))) {
1805 			f2fs_put_page(page, 1);
1806 			set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1807 			return -EIO;
1808 		}
1809 
1810 		kaddr = kmap_atomic(page);
1811 		memcpy(data, kaddr + offset, tocopy);
1812 		kunmap_atomic(kaddr);
1813 		f2fs_put_page(page, 1);
1814 
1815 		offset = 0;
1816 		toread -= tocopy;
1817 		data += tocopy;
1818 		blkidx++;
1819 	}
1820 	return len;
1821 }
1822 
1823 /* Write to quotafile */
1824 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1825 				const char *data, size_t len, loff_t off)
1826 {
1827 	struct inode *inode = sb_dqopt(sb)->files[type];
1828 	struct address_space *mapping = inode->i_mapping;
1829 	const struct address_space_operations *a_ops = mapping->a_ops;
1830 	int offset = off & (sb->s_blocksize - 1);
1831 	size_t towrite = len;
1832 	struct page *page;
1833 	char *kaddr;
1834 	int err = 0;
1835 	int tocopy;
1836 
1837 	while (towrite > 0) {
1838 		tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1839 								towrite);
1840 retry:
1841 		err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1842 							&page, NULL);
1843 		if (unlikely(err)) {
1844 			if (err == -ENOMEM) {
1845 				congestion_wait(BLK_RW_ASYNC, HZ/50);
1846 				goto retry;
1847 			}
1848 			set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1849 			break;
1850 		}
1851 
1852 		kaddr = kmap_atomic(page);
1853 		memcpy(kaddr + offset, data, tocopy);
1854 		kunmap_atomic(kaddr);
1855 		flush_dcache_page(page);
1856 
1857 		a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1858 						page, NULL);
1859 		offset = 0;
1860 		towrite -= tocopy;
1861 		off += tocopy;
1862 		data += tocopy;
1863 		cond_resched();
1864 	}
1865 
1866 	if (len == towrite)
1867 		return err;
1868 	inode->i_mtime = inode->i_ctime = current_time(inode);
1869 	f2fs_mark_inode_dirty_sync(inode, false);
1870 	return len - towrite;
1871 }
1872 
1873 static struct dquot **f2fs_get_dquots(struct inode *inode)
1874 {
1875 	return F2FS_I(inode)->i_dquot;
1876 }
1877 
1878 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1879 {
1880 	return &F2FS_I(inode)->i_reserved_quota;
1881 }
1882 
1883 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1884 {
1885 	if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
1886 		f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
1887 		return 0;
1888 	}
1889 
1890 	return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
1891 					F2FS_OPTION(sbi).s_jquota_fmt, type);
1892 }
1893 
1894 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1895 {
1896 	int enabled = 0;
1897 	int i, err;
1898 
1899 	if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
1900 		err = f2fs_enable_quotas(sbi->sb);
1901 		if (err) {
1902 			f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
1903 			return 0;
1904 		}
1905 		return 1;
1906 	}
1907 
1908 	for (i = 0; i < MAXQUOTAS; i++) {
1909 		if (F2FS_OPTION(sbi).s_qf_names[i]) {
1910 			err = f2fs_quota_on_mount(sbi, i);
1911 			if (!err) {
1912 				enabled = 1;
1913 				continue;
1914 			}
1915 			f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
1916 				 err, i);
1917 		}
1918 	}
1919 	return enabled;
1920 }
1921 
1922 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
1923 			     unsigned int flags)
1924 {
1925 	struct inode *qf_inode;
1926 	unsigned long qf_inum;
1927 	int err;
1928 
1929 	BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
1930 
1931 	qf_inum = f2fs_qf_ino(sb, type);
1932 	if (!qf_inum)
1933 		return -EPERM;
1934 
1935 	qf_inode = f2fs_iget(sb, qf_inum);
1936 	if (IS_ERR(qf_inode)) {
1937 		f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
1938 		return PTR_ERR(qf_inode);
1939 	}
1940 
1941 	/* Don't account quota for quota files to avoid recursion */
1942 	qf_inode->i_flags |= S_NOQUOTA;
1943 	err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
1944 	iput(qf_inode);
1945 	return err;
1946 }
1947 
1948 static int f2fs_enable_quotas(struct super_block *sb)
1949 {
1950 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1951 	int type, err = 0;
1952 	unsigned long qf_inum;
1953 	bool quota_mopt[MAXQUOTAS] = {
1954 		test_opt(sbi, USRQUOTA),
1955 		test_opt(sbi, GRPQUOTA),
1956 		test_opt(sbi, PRJQUOTA),
1957 	};
1958 
1959 	if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
1960 		f2fs_err(sbi, "quota file may be corrupted, skip loading it");
1961 		return 0;
1962 	}
1963 
1964 	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
1965 
1966 	for (type = 0; type < MAXQUOTAS; type++) {
1967 		qf_inum = f2fs_qf_ino(sb, type);
1968 		if (qf_inum) {
1969 			err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
1970 				DQUOT_USAGE_ENABLED |
1971 				(quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
1972 			if (err) {
1973 				f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
1974 					 type, err);
1975 				for (type--; type >= 0; type--)
1976 					dquot_quota_off(sb, type);
1977 				set_sbi_flag(F2FS_SB(sb),
1978 						SBI_QUOTA_NEED_REPAIR);
1979 				return err;
1980 			}
1981 		}
1982 	}
1983 	return 0;
1984 }
1985 
1986 int f2fs_quota_sync(struct super_block *sb, int type)
1987 {
1988 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1989 	struct quota_info *dqopt = sb_dqopt(sb);
1990 	int cnt;
1991 	int ret;
1992 
1993 	/*
1994 	 * do_quotactl
1995 	 *  f2fs_quota_sync
1996 	 *  down_read(quota_sem)
1997 	 *  dquot_writeback_dquots()
1998 	 *  f2fs_dquot_commit
1999 	 *                            block_operation
2000 	 *                            down_read(quota_sem)
2001 	 */
2002 	f2fs_lock_op(sbi);
2003 
2004 	down_read(&sbi->quota_sem);
2005 	ret = dquot_writeback_dquots(sb, type);
2006 	if (ret)
2007 		goto out;
2008 
2009 	/*
2010 	 * Now when everything is written we can discard the pagecache so
2011 	 * that userspace sees the changes.
2012 	 */
2013 	for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2014 		struct address_space *mapping;
2015 
2016 		if (type != -1 && cnt != type)
2017 			continue;
2018 		if (!sb_has_quota_active(sb, cnt))
2019 			continue;
2020 
2021 		mapping = dqopt->files[cnt]->i_mapping;
2022 
2023 		ret = filemap_fdatawrite(mapping);
2024 		if (ret)
2025 			goto out;
2026 
2027 		/* if we are using journalled quota */
2028 		if (is_journalled_quota(sbi))
2029 			continue;
2030 
2031 		ret = filemap_fdatawait(mapping);
2032 		if (ret)
2033 			set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2034 
2035 		inode_lock(dqopt->files[cnt]);
2036 		truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
2037 		inode_unlock(dqopt->files[cnt]);
2038 	}
2039 out:
2040 	if (ret)
2041 		set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2042 	up_read(&sbi->quota_sem);
2043 	f2fs_unlock_op(sbi);
2044 	return ret;
2045 }
2046 
2047 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2048 							const struct path *path)
2049 {
2050 	struct inode *inode;
2051 	int err;
2052 
2053 	/* if quota sysfile exists, deny enabling quota with specific file */
2054 	if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2055 		f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2056 		return -EBUSY;
2057 	}
2058 
2059 	err = f2fs_quota_sync(sb, type);
2060 	if (err)
2061 		return err;
2062 
2063 	err = dquot_quota_on(sb, type, format_id, path);
2064 	if (err)
2065 		return err;
2066 
2067 	inode = d_inode(path->dentry);
2068 
2069 	inode_lock(inode);
2070 	F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2071 	f2fs_set_inode_flags(inode);
2072 	inode_unlock(inode);
2073 	f2fs_mark_inode_dirty_sync(inode, false);
2074 
2075 	return 0;
2076 }
2077 
2078 static int __f2fs_quota_off(struct super_block *sb, int type)
2079 {
2080 	struct inode *inode = sb_dqopt(sb)->files[type];
2081 	int err;
2082 
2083 	if (!inode || !igrab(inode))
2084 		return dquot_quota_off(sb, type);
2085 
2086 	err = f2fs_quota_sync(sb, type);
2087 	if (err)
2088 		goto out_put;
2089 
2090 	err = dquot_quota_off(sb, type);
2091 	if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2092 		goto out_put;
2093 
2094 	inode_lock(inode);
2095 	F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2096 	f2fs_set_inode_flags(inode);
2097 	inode_unlock(inode);
2098 	f2fs_mark_inode_dirty_sync(inode, false);
2099 out_put:
2100 	iput(inode);
2101 	return err;
2102 }
2103 
2104 static int f2fs_quota_off(struct super_block *sb, int type)
2105 {
2106 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2107 	int err;
2108 
2109 	err = __f2fs_quota_off(sb, type);
2110 
2111 	/*
2112 	 * quotactl can shutdown journalled quota, result in inconsistence
2113 	 * between quota record and fs data by following updates, tag the
2114 	 * flag to let fsck be aware of it.
2115 	 */
2116 	if (is_journalled_quota(sbi))
2117 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2118 	return err;
2119 }
2120 
2121 void f2fs_quota_off_umount(struct super_block *sb)
2122 {
2123 	int type;
2124 	int err;
2125 
2126 	for (type = 0; type < MAXQUOTAS; type++) {
2127 		err = __f2fs_quota_off(sb, type);
2128 		if (err) {
2129 			int ret = dquot_quota_off(sb, type);
2130 
2131 			f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2132 				 type, err, ret);
2133 			set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2134 		}
2135 	}
2136 	/*
2137 	 * In case of checkpoint=disable, we must flush quota blocks.
2138 	 * This can cause NULL exception for node_inode in end_io, since
2139 	 * put_super already dropped it.
2140 	 */
2141 	sync_filesystem(sb);
2142 }
2143 
2144 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2145 {
2146 	struct quota_info *dqopt = sb_dqopt(sb);
2147 	int type;
2148 
2149 	for (type = 0; type < MAXQUOTAS; type++) {
2150 		if (!dqopt->files[type])
2151 			continue;
2152 		f2fs_inode_synced(dqopt->files[type]);
2153 	}
2154 }
2155 
2156 static int f2fs_dquot_commit(struct dquot *dquot)
2157 {
2158 	struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2159 	int ret;
2160 
2161 	down_read(&sbi->quota_sem);
2162 	ret = dquot_commit(dquot);
2163 	if (ret < 0)
2164 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2165 	up_read(&sbi->quota_sem);
2166 	return ret;
2167 }
2168 
2169 static int f2fs_dquot_acquire(struct dquot *dquot)
2170 {
2171 	struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2172 	int ret;
2173 
2174 	down_read(&sbi->quota_sem);
2175 	ret = dquot_acquire(dquot);
2176 	if (ret < 0)
2177 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2178 	up_read(&sbi->quota_sem);
2179 	return ret;
2180 }
2181 
2182 static int f2fs_dquot_release(struct dquot *dquot)
2183 {
2184 	struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2185 	int ret;
2186 
2187 	down_read(&sbi->quota_sem);
2188 	ret = dquot_release(dquot);
2189 	if (ret < 0)
2190 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2191 	up_read(&sbi->quota_sem);
2192 	return ret;
2193 }
2194 
2195 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2196 {
2197 	struct super_block *sb = dquot->dq_sb;
2198 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2199 	int ret;
2200 
2201 	down_read(&sbi->quota_sem);
2202 	ret = dquot_mark_dquot_dirty(dquot);
2203 
2204 	/* if we are using journalled quota */
2205 	if (is_journalled_quota(sbi))
2206 		set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2207 
2208 	up_read(&sbi->quota_sem);
2209 	return ret;
2210 }
2211 
2212 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2213 {
2214 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2215 	int ret;
2216 
2217 	down_read(&sbi->quota_sem);
2218 	ret = dquot_commit_info(sb, type);
2219 	if (ret < 0)
2220 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2221 	up_read(&sbi->quota_sem);
2222 	return ret;
2223 }
2224 
2225 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2226 {
2227 	*projid = F2FS_I(inode)->i_projid;
2228 	return 0;
2229 }
2230 
2231 static const struct dquot_operations f2fs_quota_operations = {
2232 	.get_reserved_space = f2fs_get_reserved_space,
2233 	.write_dquot	= f2fs_dquot_commit,
2234 	.acquire_dquot	= f2fs_dquot_acquire,
2235 	.release_dquot	= f2fs_dquot_release,
2236 	.mark_dirty	= f2fs_dquot_mark_dquot_dirty,
2237 	.write_info	= f2fs_dquot_commit_info,
2238 	.alloc_dquot	= dquot_alloc,
2239 	.destroy_dquot	= dquot_destroy,
2240 	.get_projid	= f2fs_get_projid,
2241 	.get_next_id	= dquot_get_next_id,
2242 };
2243 
2244 static const struct quotactl_ops f2fs_quotactl_ops = {
2245 	.quota_on	= f2fs_quota_on,
2246 	.quota_off	= f2fs_quota_off,
2247 	.quota_sync	= f2fs_quota_sync,
2248 	.get_state	= dquot_get_state,
2249 	.set_info	= dquot_set_dqinfo,
2250 	.get_dqblk	= dquot_get_dqblk,
2251 	.set_dqblk	= dquot_set_dqblk,
2252 	.get_nextdqblk	= dquot_get_next_dqblk,
2253 };
2254 #else
2255 int f2fs_quota_sync(struct super_block *sb, int type)
2256 {
2257 	return 0;
2258 }
2259 
2260 void f2fs_quota_off_umount(struct super_block *sb)
2261 {
2262 }
2263 #endif
2264 
2265 static const struct super_operations f2fs_sops = {
2266 	.alloc_inode	= f2fs_alloc_inode,
2267 	.free_inode	= f2fs_free_inode,
2268 	.drop_inode	= f2fs_drop_inode,
2269 	.write_inode	= f2fs_write_inode,
2270 	.dirty_inode	= f2fs_dirty_inode,
2271 	.show_options	= f2fs_show_options,
2272 #ifdef CONFIG_QUOTA
2273 	.quota_read	= f2fs_quota_read,
2274 	.quota_write	= f2fs_quota_write,
2275 	.get_dquots	= f2fs_get_dquots,
2276 #endif
2277 	.evict_inode	= f2fs_evict_inode,
2278 	.put_super	= f2fs_put_super,
2279 	.sync_fs	= f2fs_sync_fs,
2280 	.freeze_fs	= f2fs_freeze,
2281 	.unfreeze_fs	= f2fs_unfreeze,
2282 	.statfs		= f2fs_statfs,
2283 	.remount_fs	= f2fs_remount,
2284 };
2285 
2286 #ifdef CONFIG_FS_ENCRYPTION
2287 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2288 {
2289 	return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2290 				F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2291 				ctx, len, NULL);
2292 }
2293 
2294 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2295 							void *fs_data)
2296 {
2297 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2298 
2299 	/*
2300 	 * Encrypting the root directory is not allowed because fsck
2301 	 * expects lost+found directory to exist and remain unencrypted
2302 	 * if LOST_FOUND feature is enabled.
2303 	 *
2304 	 */
2305 	if (f2fs_sb_has_lost_found(sbi) &&
2306 			inode->i_ino == F2FS_ROOT_INO(sbi))
2307 		return -EPERM;
2308 
2309 	return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2310 				F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2311 				ctx, len, fs_data, XATTR_CREATE);
2312 }
2313 
2314 static bool f2fs_dummy_context(struct inode *inode)
2315 {
2316 	return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
2317 }
2318 
2319 static bool f2fs_has_stable_inodes(struct super_block *sb)
2320 {
2321 	return true;
2322 }
2323 
2324 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
2325 				       int *ino_bits_ret, int *lblk_bits_ret)
2326 {
2327 	*ino_bits_ret = 8 * sizeof(nid_t);
2328 	*lblk_bits_ret = 8 * sizeof(block_t);
2329 }
2330 
2331 static const struct fscrypt_operations f2fs_cryptops = {
2332 	.key_prefix		= "f2fs:",
2333 	.get_context		= f2fs_get_context,
2334 	.set_context		= f2fs_set_context,
2335 	.dummy_context		= f2fs_dummy_context,
2336 	.empty_dir		= f2fs_empty_dir,
2337 	.max_namelen		= F2FS_NAME_LEN,
2338 	.has_stable_inodes	= f2fs_has_stable_inodes,
2339 	.get_ino_and_lblk_bits	= f2fs_get_ino_and_lblk_bits,
2340 };
2341 #endif
2342 
2343 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2344 		u64 ino, u32 generation)
2345 {
2346 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2347 	struct inode *inode;
2348 
2349 	if (f2fs_check_nid_range(sbi, ino))
2350 		return ERR_PTR(-ESTALE);
2351 
2352 	/*
2353 	 * f2fs_iget isn't quite right if the inode is currently unallocated!
2354 	 * However f2fs_iget currently does appropriate checks to handle stale
2355 	 * inodes so everything is OK.
2356 	 */
2357 	inode = f2fs_iget(sb, ino);
2358 	if (IS_ERR(inode))
2359 		return ERR_CAST(inode);
2360 	if (unlikely(generation && inode->i_generation != generation)) {
2361 		/* we didn't find the right inode.. */
2362 		iput(inode);
2363 		return ERR_PTR(-ESTALE);
2364 	}
2365 	return inode;
2366 }
2367 
2368 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2369 		int fh_len, int fh_type)
2370 {
2371 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2372 				    f2fs_nfs_get_inode);
2373 }
2374 
2375 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2376 		int fh_len, int fh_type)
2377 {
2378 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2379 				    f2fs_nfs_get_inode);
2380 }
2381 
2382 static const struct export_operations f2fs_export_ops = {
2383 	.fh_to_dentry = f2fs_fh_to_dentry,
2384 	.fh_to_parent = f2fs_fh_to_parent,
2385 	.get_parent = f2fs_get_parent,
2386 };
2387 
2388 static loff_t max_file_blocks(void)
2389 {
2390 	loff_t result = 0;
2391 	loff_t leaf_count = DEF_ADDRS_PER_BLOCK;
2392 
2393 	/*
2394 	 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2395 	 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2396 	 * space in inode.i_addr, it will be more safe to reassign
2397 	 * result as zero.
2398 	 */
2399 
2400 	/* two direct node blocks */
2401 	result += (leaf_count * 2);
2402 
2403 	/* two indirect node blocks */
2404 	leaf_count *= NIDS_PER_BLOCK;
2405 	result += (leaf_count * 2);
2406 
2407 	/* one double indirect node block */
2408 	leaf_count *= NIDS_PER_BLOCK;
2409 	result += leaf_count;
2410 
2411 	return result;
2412 }
2413 
2414 static int __f2fs_commit_super(struct buffer_head *bh,
2415 			struct f2fs_super_block *super)
2416 {
2417 	lock_buffer(bh);
2418 	if (super)
2419 		memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2420 	set_buffer_dirty(bh);
2421 	unlock_buffer(bh);
2422 
2423 	/* it's rare case, we can do fua all the time */
2424 	return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2425 }
2426 
2427 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2428 					struct buffer_head *bh)
2429 {
2430 	struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2431 					(bh->b_data + F2FS_SUPER_OFFSET);
2432 	struct super_block *sb = sbi->sb;
2433 	u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2434 	u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2435 	u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2436 	u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2437 	u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2438 	u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2439 	u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2440 	u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2441 	u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2442 	u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2443 	u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2444 	u32 segment_count = le32_to_cpu(raw_super->segment_count);
2445 	u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2446 	u64 main_end_blkaddr = main_blkaddr +
2447 				(segment_count_main << log_blocks_per_seg);
2448 	u64 seg_end_blkaddr = segment0_blkaddr +
2449 				(segment_count << log_blocks_per_seg);
2450 
2451 	if (segment0_blkaddr != cp_blkaddr) {
2452 		f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2453 			  segment0_blkaddr, cp_blkaddr);
2454 		return true;
2455 	}
2456 
2457 	if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2458 							sit_blkaddr) {
2459 		f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2460 			  cp_blkaddr, sit_blkaddr,
2461 			  segment_count_ckpt << log_blocks_per_seg);
2462 		return true;
2463 	}
2464 
2465 	if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2466 							nat_blkaddr) {
2467 		f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2468 			  sit_blkaddr, nat_blkaddr,
2469 			  segment_count_sit << log_blocks_per_seg);
2470 		return true;
2471 	}
2472 
2473 	if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2474 							ssa_blkaddr) {
2475 		f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2476 			  nat_blkaddr, ssa_blkaddr,
2477 			  segment_count_nat << log_blocks_per_seg);
2478 		return true;
2479 	}
2480 
2481 	if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2482 							main_blkaddr) {
2483 		f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2484 			  ssa_blkaddr, main_blkaddr,
2485 			  segment_count_ssa << log_blocks_per_seg);
2486 		return true;
2487 	}
2488 
2489 	if (main_end_blkaddr > seg_end_blkaddr) {
2490 		f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2491 			  main_blkaddr,
2492 			  segment0_blkaddr +
2493 			  (segment_count << log_blocks_per_seg),
2494 			  segment_count_main << log_blocks_per_seg);
2495 		return true;
2496 	} else if (main_end_blkaddr < seg_end_blkaddr) {
2497 		int err = 0;
2498 		char *res;
2499 
2500 		/* fix in-memory information all the time */
2501 		raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2502 				segment0_blkaddr) >> log_blocks_per_seg);
2503 
2504 		if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2505 			set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2506 			res = "internally";
2507 		} else {
2508 			err = __f2fs_commit_super(bh, NULL);
2509 			res = err ? "failed" : "done";
2510 		}
2511 		f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%u) block(%u)",
2512 			  res, main_blkaddr,
2513 			  segment0_blkaddr +
2514 			  (segment_count << log_blocks_per_seg),
2515 			  segment_count_main << log_blocks_per_seg);
2516 		if (err)
2517 			return true;
2518 	}
2519 	return false;
2520 }
2521 
2522 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2523 				struct buffer_head *bh)
2524 {
2525 	block_t segment_count, segs_per_sec, secs_per_zone;
2526 	block_t total_sections, blocks_per_seg;
2527 	struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2528 					(bh->b_data + F2FS_SUPER_OFFSET);
2529 	unsigned int blocksize;
2530 	size_t crc_offset = 0;
2531 	__u32 crc = 0;
2532 
2533 	if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
2534 		f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
2535 			  F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2536 		return -EINVAL;
2537 	}
2538 
2539 	/* Check checksum_offset and crc in superblock */
2540 	if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2541 		crc_offset = le32_to_cpu(raw_super->checksum_offset);
2542 		if (crc_offset !=
2543 			offsetof(struct f2fs_super_block, crc)) {
2544 			f2fs_info(sbi, "Invalid SB checksum offset: %zu",
2545 				  crc_offset);
2546 			return -EFSCORRUPTED;
2547 		}
2548 		crc = le32_to_cpu(raw_super->crc);
2549 		if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2550 			f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
2551 			return -EFSCORRUPTED;
2552 		}
2553 	}
2554 
2555 	/* Currently, support only 4KB page cache size */
2556 	if (F2FS_BLKSIZE != PAGE_SIZE) {
2557 		f2fs_info(sbi, "Invalid page_cache_size (%lu), supports only 4KB",
2558 			  PAGE_SIZE);
2559 		return -EFSCORRUPTED;
2560 	}
2561 
2562 	/* Currently, support only 4KB block size */
2563 	blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2564 	if (blocksize != F2FS_BLKSIZE) {
2565 		f2fs_info(sbi, "Invalid blocksize (%u), supports only 4KB",
2566 			  blocksize);
2567 		return -EFSCORRUPTED;
2568 	}
2569 
2570 	/* check log blocks per segment */
2571 	if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2572 		f2fs_info(sbi, "Invalid log blocks per segment (%u)",
2573 			  le32_to_cpu(raw_super->log_blocks_per_seg));
2574 		return -EFSCORRUPTED;
2575 	}
2576 
2577 	/* Currently, support 512/1024/2048/4096 bytes sector size */
2578 	if (le32_to_cpu(raw_super->log_sectorsize) >
2579 				F2FS_MAX_LOG_SECTOR_SIZE ||
2580 		le32_to_cpu(raw_super->log_sectorsize) <
2581 				F2FS_MIN_LOG_SECTOR_SIZE) {
2582 		f2fs_info(sbi, "Invalid log sectorsize (%u)",
2583 			  le32_to_cpu(raw_super->log_sectorsize));
2584 		return -EFSCORRUPTED;
2585 	}
2586 	if (le32_to_cpu(raw_super->log_sectors_per_block) +
2587 		le32_to_cpu(raw_super->log_sectorsize) !=
2588 			F2FS_MAX_LOG_SECTOR_SIZE) {
2589 		f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
2590 			  le32_to_cpu(raw_super->log_sectors_per_block),
2591 			  le32_to_cpu(raw_super->log_sectorsize));
2592 		return -EFSCORRUPTED;
2593 	}
2594 
2595 	segment_count = le32_to_cpu(raw_super->segment_count);
2596 	segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2597 	secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2598 	total_sections = le32_to_cpu(raw_super->section_count);
2599 
2600 	/* blocks_per_seg should be 512, given the above check */
2601 	blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2602 
2603 	if (segment_count > F2FS_MAX_SEGMENT ||
2604 				segment_count < F2FS_MIN_SEGMENTS) {
2605 		f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
2606 		return -EFSCORRUPTED;
2607 	}
2608 
2609 	if (total_sections > segment_count ||
2610 			total_sections < F2FS_MIN_SEGMENTS ||
2611 			segs_per_sec > segment_count || !segs_per_sec) {
2612 		f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
2613 			  segment_count, total_sections, segs_per_sec);
2614 		return -EFSCORRUPTED;
2615 	}
2616 
2617 	if ((segment_count / segs_per_sec) < total_sections) {
2618 		f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
2619 			  segment_count, segs_per_sec, total_sections);
2620 		return -EFSCORRUPTED;
2621 	}
2622 
2623 	if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2624 		f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
2625 			  segment_count, le64_to_cpu(raw_super->block_count));
2626 		return -EFSCORRUPTED;
2627 	}
2628 
2629 	if (RDEV(0).path[0]) {
2630 		block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
2631 		int i = 1;
2632 
2633 		while (i < MAX_DEVICES && RDEV(i).path[0]) {
2634 			dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
2635 			i++;
2636 		}
2637 		if (segment_count != dev_seg_count) {
2638 			f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
2639 					segment_count, dev_seg_count);
2640 			return -EFSCORRUPTED;
2641 		}
2642 	}
2643 
2644 	if (secs_per_zone > total_sections || !secs_per_zone) {
2645 		f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
2646 			  secs_per_zone, total_sections);
2647 		return -EFSCORRUPTED;
2648 	}
2649 	if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2650 			raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2651 			(le32_to_cpu(raw_super->extension_count) +
2652 			raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2653 		f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
2654 			  le32_to_cpu(raw_super->extension_count),
2655 			  raw_super->hot_ext_count,
2656 			  F2FS_MAX_EXTENSION);
2657 		return -EFSCORRUPTED;
2658 	}
2659 
2660 	if (le32_to_cpu(raw_super->cp_payload) >
2661 				(blocks_per_seg - F2FS_CP_PACKS)) {
2662 		f2fs_info(sbi, "Insane cp_payload (%u > %u)",
2663 			  le32_to_cpu(raw_super->cp_payload),
2664 			  blocks_per_seg - F2FS_CP_PACKS);
2665 		return -EFSCORRUPTED;
2666 	}
2667 
2668 	/* check reserved ino info */
2669 	if (le32_to_cpu(raw_super->node_ino) != 1 ||
2670 		le32_to_cpu(raw_super->meta_ino) != 2 ||
2671 		le32_to_cpu(raw_super->root_ino) != 3) {
2672 		f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2673 			  le32_to_cpu(raw_super->node_ino),
2674 			  le32_to_cpu(raw_super->meta_ino),
2675 			  le32_to_cpu(raw_super->root_ino));
2676 		return -EFSCORRUPTED;
2677 	}
2678 
2679 	/* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2680 	if (sanity_check_area_boundary(sbi, bh))
2681 		return -EFSCORRUPTED;
2682 
2683 	return 0;
2684 }
2685 
2686 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2687 {
2688 	unsigned int total, fsmeta;
2689 	struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2690 	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2691 	unsigned int ovp_segments, reserved_segments;
2692 	unsigned int main_segs, blocks_per_seg;
2693 	unsigned int sit_segs, nat_segs;
2694 	unsigned int sit_bitmap_size, nat_bitmap_size;
2695 	unsigned int log_blocks_per_seg;
2696 	unsigned int segment_count_main;
2697 	unsigned int cp_pack_start_sum, cp_payload;
2698 	block_t user_block_count, valid_user_blocks;
2699 	block_t avail_node_count, valid_node_count;
2700 	int i, j;
2701 
2702 	total = le32_to_cpu(raw_super->segment_count);
2703 	fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2704 	sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2705 	fsmeta += sit_segs;
2706 	nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2707 	fsmeta += nat_segs;
2708 	fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2709 	fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2710 
2711 	if (unlikely(fsmeta >= total))
2712 		return 1;
2713 
2714 	ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2715 	reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2716 
2717 	if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2718 			ovp_segments == 0 || reserved_segments == 0)) {
2719 		f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
2720 		return 1;
2721 	}
2722 
2723 	user_block_count = le64_to_cpu(ckpt->user_block_count);
2724 	segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2725 	log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2726 	if (!user_block_count || user_block_count >=
2727 			segment_count_main << log_blocks_per_seg) {
2728 		f2fs_err(sbi, "Wrong user_block_count: %u",
2729 			 user_block_count);
2730 		return 1;
2731 	}
2732 
2733 	valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
2734 	if (valid_user_blocks > user_block_count) {
2735 		f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
2736 			 valid_user_blocks, user_block_count);
2737 		return 1;
2738 	}
2739 
2740 	valid_node_count = le32_to_cpu(ckpt->valid_node_count);
2741 	avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
2742 	if (valid_node_count > avail_node_count) {
2743 		f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
2744 			 valid_node_count, avail_node_count);
2745 		return 1;
2746 	}
2747 
2748 	main_segs = le32_to_cpu(raw_super->segment_count_main);
2749 	blocks_per_seg = sbi->blocks_per_seg;
2750 
2751 	for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2752 		if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2753 			le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2754 			return 1;
2755 		for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2756 			if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2757 				le32_to_cpu(ckpt->cur_node_segno[j])) {
2758 				f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
2759 					 i, j,
2760 					 le32_to_cpu(ckpt->cur_node_segno[i]));
2761 				return 1;
2762 			}
2763 		}
2764 	}
2765 	for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2766 		if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2767 			le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2768 			return 1;
2769 		for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2770 			if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2771 				le32_to_cpu(ckpt->cur_data_segno[j])) {
2772 				f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
2773 					 i, j,
2774 					 le32_to_cpu(ckpt->cur_data_segno[i]));
2775 				return 1;
2776 			}
2777 		}
2778 	}
2779 	for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2780 		for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
2781 			if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2782 				le32_to_cpu(ckpt->cur_data_segno[j])) {
2783 				f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
2784 					 i, j,
2785 					 le32_to_cpu(ckpt->cur_node_segno[i]));
2786 				return 1;
2787 			}
2788 		}
2789 	}
2790 
2791 	sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2792 	nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2793 
2794 	if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2795 		nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2796 		f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
2797 			 sit_bitmap_size, nat_bitmap_size);
2798 		return 1;
2799 	}
2800 
2801 	cp_pack_start_sum = __start_sum_addr(sbi);
2802 	cp_payload = __cp_payload(sbi);
2803 	if (cp_pack_start_sum < cp_payload + 1 ||
2804 		cp_pack_start_sum > blocks_per_seg - 1 -
2805 			NR_CURSEG_TYPE) {
2806 		f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
2807 			 cp_pack_start_sum);
2808 		return 1;
2809 	}
2810 
2811 	if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
2812 		le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
2813 		f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
2814 			  "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
2815 			  "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
2816 			  le32_to_cpu(ckpt->checksum_offset));
2817 		return 1;
2818 	}
2819 
2820 	if (unlikely(f2fs_cp_error(sbi))) {
2821 		f2fs_err(sbi, "A bug case: need to run fsck");
2822 		return 1;
2823 	}
2824 	return 0;
2825 }
2826 
2827 static void init_sb_info(struct f2fs_sb_info *sbi)
2828 {
2829 	struct f2fs_super_block *raw_super = sbi->raw_super;
2830 	int i;
2831 
2832 	sbi->log_sectors_per_block =
2833 		le32_to_cpu(raw_super->log_sectors_per_block);
2834 	sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2835 	sbi->blocksize = 1 << sbi->log_blocksize;
2836 	sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2837 	sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2838 	sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2839 	sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2840 	sbi->total_sections = le32_to_cpu(raw_super->section_count);
2841 	sbi->total_node_count =
2842 		(le32_to_cpu(raw_super->segment_count_nat) / 2)
2843 			* sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2844 	sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2845 	sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2846 	sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2847 	sbi->cur_victim_sec = NULL_SECNO;
2848 	sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
2849 	sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
2850 	sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2851 	sbi->migration_granularity = sbi->segs_per_sec;
2852 
2853 	sbi->dir_level = DEF_DIR_LEVEL;
2854 	sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2855 	sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2856 	sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
2857 	sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
2858 	sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
2859 	sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
2860 				DEF_UMOUNT_DISCARD_TIMEOUT;
2861 	clear_sbi_flag(sbi, SBI_NEED_FSCK);
2862 
2863 	for (i = 0; i < NR_COUNT_TYPE; i++)
2864 		atomic_set(&sbi->nr_pages[i], 0);
2865 
2866 	for (i = 0; i < META; i++)
2867 		atomic_set(&sbi->wb_sync_req[i], 0);
2868 
2869 	INIT_LIST_HEAD(&sbi->s_list);
2870 	mutex_init(&sbi->umount_mutex);
2871 	init_rwsem(&sbi->io_order_lock);
2872 	spin_lock_init(&sbi->cp_lock);
2873 
2874 	sbi->dirty_device = 0;
2875 	spin_lock_init(&sbi->dev_lock);
2876 
2877 	init_rwsem(&sbi->sb_lock);
2878 	init_rwsem(&sbi->pin_sem);
2879 }
2880 
2881 static int init_percpu_info(struct f2fs_sb_info *sbi)
2882 {
2883 	int err;
2884 
2885 	err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2886 	if (err)
2887 		return err;
2888 
2889 	err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
2890 								GFP_KERNEL);
2891 	if (err)
2892 		percpu_counter_destroy(&sbi->alloc_valid_block_count);
2893 
2894 	return err;
2895 }
2896 
2897 #ifdef CONFIG_BLK_DEV_ZONED
2898 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
2899 			       void *data)
2900 {
2901 	struct f2fs_dev_info *dev = data;
2902 
2903 	if (zone->type != BLK_ZONE_TYPE_CONVENTIONAL)
2904 		set_bit(idx, dev->blkz_seq);
2905 	return 0;
2906 }
2907 
2908 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2909 {
2910 	struct block_device *bdev = FDEV(devi).bdev;
2911 	sector_t nr_sectors = bdev->bd_part->nr_sects;
2912 	int ret;
2913 
2914 	if (!f2fs_sb_has_blkzoned(sbi))
2915 		return 0;
2916 
2917 	if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2918 				SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2919 		return -EINVAL;
2920 	sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2921 	if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2922 				__ilog2_u32(sbi->blocks_per_blkz))
2923 		return -EINVAL;
2924 	sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2925 	FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2926 					sbi->log_blocks_per_blkz;
2927 	if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2928 		FDEV(devi).nr_blkz++;
2929 
2930 	FDEV(devi).blkz_seq = f2fs_kzalloc(sbi,
2931 					BITS_TO_LONGS(FDEV(devi).nr_blkz)
2932 					* sizeof(unsigned long),
2933 					GFP_KERNEL);
2934 	if (!FDEV(devi).blkz_seq)
2935 		return -ENOMEM;
2936 
2937 	/* Get block zones type */
2938 	ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
2939 				  &FDEV(devi));
2940 	if (ret < 0)
2941 		return ret;
2942 
2943 	return 0;
2944 }
2945 #endif
2946 
2947 /*
2948  * Read f2fs raw super block.
2949  * Because we have two copies of super block, so read both of them
2950  * to get the first valid one. If any one of them is broken, we pass
2951  * them recovery flag back to the caller.
2952  */
2953 static int read_raw_super_block(struct f2fs_sb_info *sbi,
2954 			struct f2fs_super_block **raw_super,
2955 			int *valid_super_block, int *recovery)
2956 {
2957 	struct super_block *sb = sbi->sb;
2958 	int block;
2959 	struct buffer_head *bh;
2960 	struct f2fs_super_block *super;
2961 	int err = 0;
2962 
2963 	super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2964 	if (!super)
2965 		return -ENOMEM;
2966 
2967 	for (block = 0; block < 2; block++) {
2968 		bh = sb_bread(sb, block);
2969 		if (!bh) {
2970 			f2fs_err(sbi, "Unable to read %dth superblock",
2971 				 block + 1);
2972 			err = -EIO;
2973 			*recovery = 1;
2974 			continue;
2975 		}
2976 
2977 		/* sanity checking of raw super */
2978 		err = sanity_check_raw_super(sbi, bh);
2979 		if (err) {
2980 			f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
2981 				 block + 1);
2982 			brelse(bh);
2983 			*recovery = 1;
2984 			continue;
2985 		}
2986 
2987 		if (!*raw_super) {
2988 			memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2989 							sizeof(*super));
2990 			*valid_super_block = block;
2991 			*raw_super = super;
2992 		}
2993 		brelse(bh);
2994 	}
2995 
2996 	/* No valid superblock */
2997 	if (!*raw_super)
2998 		kvfree(super);
2999 	else
3000 		err = 0;
3001 
3002 	return err;
3003 }
3004 
3005 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3006 {
3007 	struct buffer_head *bh;
3008 	__u32 crc = 0;
3009 	int err;
3010 
3011 	if ((recover && f2fs_readonly(sbi->sb)) ||
3012 				bdev_read_only(sbi->sb->s_bdev)) {
3013 		set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3014 		return -EROFS;
3015 	}
3016 
3017 	/* we should update superblock crc here */
3018 	if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3019 		crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3020 				offsetof(struct f2fs_super_block, crc));
3021 		F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3022 	}
3023 
3024 	/* write back-up superblock first */
3025 	bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3026 	if (!bh)
3027 		return -EIO;
3028 	err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3029 	brelse(bh);
3030 
3031 	/* if we are in recovery path, skip writing valid superblock */
3032 	if (recover || err)
3033 		return err;
3034 
3035 	/* write current valid superblock */
3036 	bh = sb_bread(sbi->sb, sbi->valid_super_block);
3037 	if (!bh)
3038 		return -EIO;
3039 	err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3040 	brelse(bh);
3041 	return err;
3042 }
3043 
3044 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3045 {
3046 	struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3047 	unsigned int max_devices = MAX_DEVICES;
3048 	int i;
3049 
3050 	/* Initialize single device information */
3051 	if (!RDEV(0).path[0]) {
3052 		if (!bdev_is_zoned(sbi->sb->s_bdev))
3053 			return 0;
3054 		max_devices = 1;
3055 	}
3056 
3057 	/*
3058 	 * Initialize multiple devices information, or single
3059 	 * zoned block device information.
3060 	 */
3061 	sbi->devs = f2fs_kzalloc(sbi,
3062 				 array_size(max_devices,
3063 					    sizeof(struct f2fs_dev_info)),
3064 				 GFP_KERNEL);
3065 	if (!sbi->devs)
3066 		return -ENOMEM;
3067 
3068 	for (i = 0; i < max_devices; i++) {
3069 
3070 		if (i > 0 && !RDEV(i).path[0])
3071 			break;
3072 
3073 		if (max_devices == 1) {
3074 			/* Single zoned block device mount */
3075 			FDEV(0).bdev =
3076 				blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3077 					sbi->sb->s_mode, sbi->sb->s_type);
3078 		} else {
3079 			/* Multi-device mount */
3080 			memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3081 			FDEV(i).total_segments =
3082 				le32_to_cpu(RDEV(i).total_segments);
3083 			if (i == 0) {
3084 				FDEV(i).start_blk = 0;
3085 				FDEV(i).end_blk = FDEV(i).start_blk +
3086 				    (FDEV(i).total_segments <<
3087 				    sbi->log_blocks_per_seg) - 1 +
3088 				    le32_to_cpu(raw_super->segment0_blkaddr);
3089 			} else {
3090 				FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3091 				FDEV(i).end_blk = FDEV(i).start_blk +
3092 					(FDEV(i).total_segments <<
3093 					sbi->log_blocks_per_seg) - 1;
3094 			}
3095 			FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3096 					sbi->sb->s_mode, sbi->sb->s_type);
3097 		}
3098 		if (IS_ERR(FDEV(i).bdev))
3099 			return PTR_ERR(FDEV(i).bdev);
3100 
3101 		/* to release errored devices */
3102 		sbi->s_ndevs = i + 1;
3103 
3104 #ifdef CONFIG_BLK_DEV_ZONED
3105 		if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3106 				!f2fs_sb_has_blkzoned(sbi)) {
3107 			f2fs_err(sbi, "Zoned block device feature not enabled\n");
3108 			return -EINVAL;
3109 		}
3110 		if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3111 			if (init_blkz_info(sbi, i)) {
3112 				f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3113 				return -EINVAL;
3114 			}
3115 			if (max_devices == 1)
3116 				break;
3117 			f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3118 				  i, FDEV(i).path,
3119 				  FDEV(i).total_segments,
3120 				  FDEV(i).start_blk, FDEV(i).end_blk,
3121 				  bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3122 				  "Host-aware" : "Host-managed");
3123 			continue;
3124 		}
3125 #endif
3126 		f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3127 			  i, FDEV(i).path,
3128 			  FDEV(i).total_segments,
3129 			  FDEV(i).start_blk, FDEV(i).end_blk);
3130 	}
3131 	f2fs_info(sbi,
3132 		  "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3133 	return 0;
3134 }
3135 
3136 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
3137 {
3138 #ifdef CONFIG_UNICODE
3139 	if (f2fs_sb_has_casefold(sbi) && !sbi->s_encoding) {
3140 		const struct f2fs_sb_encodings *encoding_info;
3141 		struct unicode_map *encoding;
3142 		__u16 encoding_flags;
3143 
3144 		if (f2fs_sb_has_encrypt(sbi)) {
3145 			f2fs_err(sbi,
3146 				"Can't mount with encoding and encryption");
3147 			return -EINVAL;
3148 		}
3149 
3150 		if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info,
3151 					  &encoding_flags)) {
3152 			f2fs_err(sbi,
3153 				 "Encoding requested by superblock is unknown");
3154 			return -EINVAL;
3155 		}
3156 
3157 		encoding = utf8_load(encoding_info->version);
3158 		if (IS_ERR(encoding)) {
3159 			f2fs_err(sbi,
3160 				 "can't mount with superblock charset: %s-%s "
3161 				 "not supported by the kernel. flags: 0x%x.",
3162 				 encoding_info->name, encoding_info->version,
3163 				 encoding_flags);
3164 			return PTR_ERR(encoding);
3165 		}
3166 		f2fs_info(sbi, "Using encoding defined by superblock: "
3167 			 "%s-%s with flags 0x%hx", encoding_info->name,
3168 			 encoding_info->version?:"\b", encoding_flags);
3169 
3170 		sbi->s_encoding = encoding;
3171 		sbi->s_encoding_flags = encoding_flags;
3172 		sbi->sb->s_d_op = &f2fs_dentry_ops;
3173 	}
3174 #else
3175 	if (f2fs_sb_has_casefold(sbi)) {
3176 		f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
3177 		return -EINVAL;
3178 	}
3179 #endif
3180 	return 0;
3181 }
3182 
3183 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3184 {
3185 	struct f2fs_sm_info *sm_i = SM_I(sbi);
3186 
3187 	/* adjust parameters according to the volume size */
3188 	if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3189 		F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3190 		sm_i->dcc_info->discard_granularity = 1;
3191 		sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3192 	}
3193 
3194 	sbi->readdir_ra = 1;
3195 }
3196 
3197 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3198 {
3199 	struct f2fs_sb_info *sbi;
3200 	struct f2fs_super_block *raw_super;
3201 	struct inode *root;
3202 	int err;
3203 	bool skip_recovery = false, need_fsck = false;
3204 	char *options = NULL;
3205 	int recovery, i, valid_super_block;
3206 	struct curseg_info *seg_i;
3207 	int retry_cnt = 1;
3208 
3209 try_onemore:
3210 	err = -EINVAL;
3211 	raw_super = NULL;
3212 	valid_super_block = -1;
3213 	recovery = 0;
3214 
3215 	/* allocate memory for f2fs-specific super block info */
3216 	sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3217 	if (!sbi)
3218 		return -ENOMEM;
3219 
3220 	sbi->sb = sb;
3221 
3222 	/* Load the checksum driver */
3223 	sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3224 	if (IS_ERR(sbi->s_chksum_driver)) {
3225 		f2fs_err(sbi, "Cannot load crc32 driver.");
3226 		err = PTR_ERR(sbi->s_chksum_driver);
3227 		sbi->s_chksum_driver = NULL;
3228 		goto free_sbi;
3229 	}
3230 
3231 	/* set a block size */
3232 	if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3233 		f2fs_err(sbi, "unable to set blocksize");
3234 		goto free_sbi;
3235 	}
3236 
3237 	err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3238 								&recovery);
3239 	if (err)
3240 		goto free_sbi;
3241 
3242 	sb->s_fs_info = sbi;
3243 	sbi->raw_super = raw_super;
3244 
3245 	/* precompute checksum seed for metadata */
3246 	if (f2fs_sb_has_inode_chksum(sbi))
3247 		sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3248 						sizeof(raw_super->uuid));
3249 
3250 	/*
3251 	 * The BLKZONED feature indicates that the drive was formatted with
3252 	 * zone alignment optimization. This is optional for host-aware
3253 	 * devices, but mandatory for host-managed zoned block devices.
3254 	 */
3255 #ifndef CONFIG_BLK_DEV_ZONED
3256 	if (f2fs_sb_has_blkzoned(sbi)) {
3257 		f2fs_err(sbi, "Zoned block device support is not enabled");
3258 		err = -EOPNOTSUPP;
3259 		goto free_sb_buf;
3260 	}
3261 #endif
3262 	default_options(sbi);
3263 	/* parse mount options */
3264 	options = kstrdup((const char *)data, GFP_KERNEL);
3265 	if (data && !options) {
3266 		err = -ENOMEM;
3267 		goto free_sb_buf;
3268 	}
3269 
3270 	err = parse_options(sb, options);
3271 	if (err)
3272 		goto free_options;
3273 
3274 	sbi->max_file_blocks = max_file_blocks();
3275 	sb->s_maxbytes = sbi->max_file_blocks <<
3276 				le32_to_cpu(raw_super->log_blocksize);
3277 	sb->s_max_links = F2FS_LINK_MAX;
3278 
3279 	err = f2fs_setup_casefold(sbi);
3280 	if (err)
3281 		goto free_options;
3282 
3283 #ifdef CONFIG_QUOTA
3284 	sb->dq_op = &f2fs_quota_operations;
3285 	sb->s_qcop = &f2fs_quotactl_ops;
3286 	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3287 
3288 	if (f2fs_sb_has_quota_ino(sbi)) {
3289 		for (i = 0; i < MAXQUOTAS; i++) {
3290 			if (f2fs_qf_ino(sbi->sb, i))
3291 				sbi->nquota_files++;
3292 		}
3293 	}
3294 #endif
3295 
3296 	sb->s_op = &f2fs_sops;
3297 #ifdef CONFIG_FS_ENCRYPTION
3298 	sb->s_cop = &f2fs_cryptops;
3299 #endif
3300 #ifdef CONFIG_FS_VERITY
3301 	sb->s_vop = &f2fs_verityops;
3302 #endif
3303 	sb->s_xattr = f2fs_xattr_handlers;
3304 	sb->s_export_op = &f2fs_export_ops;
3305 	sb->s_magic = F2FS_SUPER_MAGIC;
3306 	sb->s_time_gran = 1;
3307 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3308 		(test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3309 	memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3310 	sb->s_iflags |= SB_I_CGROUPWB;
3311 
3312 	/* init f2fs-specific super block info */
3313 	sbi->valid_super_block = valid_super_block;
3314 	mutex_init(&sbi->gc_mutex);
3315 	mutex_init(&sbi->writepages);
3316 	mutex_init(&sbi->cp_mutex);
3317 	mutex_init(&sbi->resize_mutex);
3318 	init_rwsem(&sbi->node_write);
3319 	init_rwsem(&sbi->node_change);
3320 
3321 	/* disallow all the data/node/meta page writes */
3322 	set_sbi_flag(sbi, SBI_POR_DOING);
3323 	spin_lock_init(&sbi->stat_lock);
3324 
3325 	/* init iostat info */
3326 	spin_lock_init(&sbi->iostat_lock);
3327 	sbi->iostat_enable = false;
3328 
3329 	for (i = 0; i < NR_PAGE_TYPE; i++) {
3330 		int n = (i == META) ? 1: NR_TEMP_TYPE;
3331 		int j;
3332 
3333 		sbi->write_io[i] =
3334 			f2fs_kmalloc(sbi,
3335 				     array_size(n,
3336 						sizeof(struct f2fs_bio_info)),
3337 				     GFP_KERNEL);
3338 		if (!sbi->write_io[i]) {
3339 			err = -ENOMEM;
3340 			goto free_bio_info;
3341 		}
3342 
3343 		for (j = HOT; j < n; j++) {
3344 			init_rwsem(&sbi->write_io[i][j].io_rwsem);
3345 			sbi->write_io[i][j].sbi = sbi;
3346 			sbi->write_io[i][j].bio = NULL;
3347 			spin_lock_init(&sbi->write_io[i][j].io_lock);
3348 			INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3349 			INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list);
3350 			init_rwsem(&sbi->write_io[i][j].bio_list_lock);
3351 		}
3352 	}
3353 
3354 	init_rwsem(&sbi->cp_rwsem);
3355 	init_rwsem(&sbi->quota_sem);
3356 	init_waitqueue_head(&sbi->cp_wait);
3357 	init_sb_info(sbi);
3358 
3359 	err = init_percpu_info(sbi);
3360 	if (err)
3361 		goto free_bio_info;
3362 
3363 	if (F2FS_IO_ALIGNED(sbi)) {
3364 		sbi->write_io_dummy =
3365 			mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3366 		if (!sbi->write_io_dummy) {
3367 			err = -ENOMEM;
3368 			goto free_percpu;
3369 		}
3370 	}
3371 
3372 	/* get an inode for meta space */
3373 	sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3374 	if (IS_ERR(sbi->meta_inode)) {
3375 		f2fs_err(sbi, "Failed to read F2FS meta data inode");
3376 		err = PTR_ERR(sbi->meta_inode);
3377 		goto free_io_dummy;
3378 	}
3379 
3380 	err = f2fs_get_valid_checkpoint(sbi);
3381 	if (err) {
3382 		f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
3383 		goto free_meta_inode;
3384 	}
3385 
3386 	if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3387 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3388 	if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3389 		set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3390 		sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3391 	}
3392 
3393 	if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
3394 		set_sbi_flag(sbi, SBI_NEED_FSCK);
3395 
3396 	/* Initialize device list */
3397 	err = f2fs_scan_devices(sbi);
3398 	if (err) {
3399 		f2fs_err(sbi, "Failed to find devices");
3400 		goto free_devices;
3401 	}
3402 
3403 	sbi->total_valid_node_count =
3404 				le32_to_cpu(sbi->ckpt->valid_node_count);
3405 	percpu_counter_set(&sbi->total_valid_inode_count,
3406 				le32_to_cpu(sbi->ckpt->valid_inode_count));
3407 	sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3408 	sbi->total_valid_block_count =
3409 				le64_to_cpu(sbi->ckpt->valid_block_count);
3410 	sbi->last_valid_block_count = sbi->total_valid_block_count;
3411 	sbi->reserved_blocks = 0;
3412 	sbi->current_reserved_blocks = 0;
3413 	limit_reserve_root(sbi);
3414 
3415 	for (i = 0; i < NR_INODE_TYPE; i++) {
3416 		INIT_LIST_HEAD(&sbi->inode_list[i]);
3417 		spin_lock_init(&sbi->inode_lock[i]);
3418 	}
3419 	mutex_init(&sbi->flush_lock);
3420 
3421 	f2fs_init_extent_cache_info(sbi);
3422 
3423 	f2fs_init_ino_entry_info(sbi);
3424 
3425 	f2fs_init_fsync_node_info(sbi);
3426 
3427 	/* setup f2fs internal modules */
3428 	err = f2fs_build_segment_manager(sbi);
3429 	if (err) {
3430 		f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
3431 			 err);
3432 		goto free_sm;
3433 	}
3434 	err = f2fs_build_node_manager(sbi);
3435 	if (err) {
3436 		f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
3437 			 err);
3438 		goto free_nm;
3439 	}
3440 
3441 	/* For write statistics */
3442 	if (sb->s_bdev->bd_part)
3443 		sbi->sectors_written_start =
3444 			(u64)part_stat_read(sb->s_bdev->bd_part,
3445 					    sectors[STAT_WRITE]);
3446 
3447 	/* Read accumulated write IO statistics if exists */
3448 	seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3449 	if (__exist_node_summaries(sbi))
3450 		sbi->kbytes_written =
3451 			le64_to_cpu(seg_i->journal->info.kbytes_written);
3452 
3453 	f2fs_build_gc_manager(sbi);
3454 
3455 	err = f2fs_build_stats(sbi);
3456 	if (err)
3457 		goto free_nm;
3458 
3459 	/* get an inode for node space */
3460 	sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3461 	if (IS_ERR(sbi->node_inode)) {
3462 		f2fs_err(sbi, "Failed to read node inode");
3463 		err = PTR_ERR(sbi->node_inode);
3464 		goto free_stats;
3465 	}
3466 
3467 	/* read root inode and dentry */
3468 	root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3469 	if (IS_ERR(root)) {
3470 		f2fs_err(sbi, "Failed to read root inode");
3471 		err = PTR_ERR(root);
3472 		goto free_node_inode;
3473 	}
3474 	if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3475 			!root->i_size || !root->i_nlink) {
3476 		iput(root);
3477 		err = -EINVAL;
3478 		goto free_node_inode;
3479 	}
3480 
3481 	sb->s_root = d_make_root(root); /* allocate root dentry */
3482 	if (!sb->s_root) {
3483 		err = -ENOMEM;
3484 		goto free_node_inode;
3485 	}
3486 
3487 	err = f2fs_register_sysfs(sbi);
3488 	if (err)
3489 		goto free_root_inode;
3490 
3491 #ifdef CONFIG_QUOTA
3492 	/* Enable quota usage during mount */
3493 	if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3494 		err = f2fs_enable_quotas(sb);
3495 		if (err)
3496 			f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
3497 	}
3498 #endif
3499 	/* if there are nt orphan nodes free them */
3500 	err = f2fs_recover_orphan_inodes(sbi);
3501 	if (err)
3502 		goto free_meta;
3503 
3504 	if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3505 		goto reset_checkpoint;
3506 
3507 	/* recover fsynced data */
3508 	if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
3509 		/*
3510 		 * mount should be failed, when device has readonly mode, and
3511 		 * previous checkpoint was not done by clean system shutdown.
3512 		 */
3513 		if (f2fs_hw_is_readonly(sbi)) {
3514 			if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3515 				err = -EROFS;
3516 				f2fs_err(sbi, "Need to recover fsync data, but write access unavailable");
3517 				goto free_meta;
3518 			}
3519 			f2fs_info(sbi, "write access unavailable, skipping recovery");
3520 			goto reset_checkpoint;
3521 		}
3522 
3523 		if (need_fsck)
3524 			set_sbi_flag(sbi, SBI_NEED_FSCK);
3525 
3526 		if (skip_recovery)
3527 			goto reset_checkpoint;
3528 
3529 		err = f2fs_recover_fsync_data(sbi, false);
3530 		if (err < 0) {
3531 			if (err != -ENOMEM)
3532 				skip_recovery = true;
3533 			need_fsck = true;
3534 			f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
3535 				 err);
3536 			goto free_meta;
3537 		}
3538 	} else {
3539 		err = f2fs_recover_fsync_data(sbi, true);
3540 
3541 		if (!f2fs_readonly(sb) && err > 0) {
3542 			err = -EINVAL;
3543 			f2fs_err(sbi, "Need to recover fsync data");
3544 			goto free_meta;
3545 		}
3546 	}
3547 reset_checkpoint:
3548 	/* f2fs_recover_fsync_data() cleared this already */
3549 	clear_sbi_flag(sbi, SBI_POR_DOING);
3550 
3551 	if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3552 		err = f2fs_disable_checkpoint(sbi);
3553 		if (err)
3554 			goto sync_free_meta;
3555 	} else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3556 		f2fs_enable_checkpoint(sbi);
3557 	}
3558 
3559 	/*
3560 	 * If filesystem is not mounted as read-only then
3561 	 * do start the gc_thread.
3562 	 */
3563 	if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
3564 		/* After POR, we can run background GC thread.*/
3565 		err = f2fs_start_gc_thread(sbi);
3566 		if (err)
3567 			goto sync_free_meta;
3568 	}
3569 	kvfree(options);
3570 
3571 	/* recover broken superblock */
3572 	if (recovery) {
3573 		err = f2fs_commit_super(sbi, true);
3574 		f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
3575 			  sbi->valid_super_block ? 1 : 2, err);
3576 	}
3577 
3578 	f2fs_join_shrinker(sbi);
3579 
3580 	f2fs_tuning_parameters(sbi);
3581 
3582 	f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
3583 		    cur_cp_version(F2FS_CKPT(sbi)));
3584 	f2fs_update_time(sbi, CP_TIME);
3585 	f2fs_update_time(sbi, REQ_TIME);
3586 	clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3587 	return 0;
3588 
3589 sync_free_meta:
3590 	/* safe to flush all the data */
3591 	sync_filesystem(sbi->sb);
3592 	retry_cnt = 0;
3593 
3594 free_meta:
3595 #ifdef CONFIG_QUOTA
3596 	f2fs_truncate_quota_inode_pages(sb);
3597 	if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3598 		f2fs_quota_off_umount(sbi->sb);
3599 #endif
3600 	/*
3601 	 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3602 	 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3603 	 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3604 	 * falls into an infinite loop in f2fs_sync_meta_pages().
3605 	 */
3606 	truncate_inode_pages_final(META_MAPPING(sbi));
3607 	/* evict some inodes being cached by GC */
3608 	evict_inodes(sb);
3609 	f2fs_unregister_sysfs(sbi);
3610 free_root_inode:
3611 	dput(sb->s_root);
3612 	sb->s_root = NULL;
3613 free_node_inode:
3614 	f2fs_release_ino_entry(sbi, true);
3615 	truncate_inode_pages_final(NODE_MAPPING(sbi));
3616 	iput(sbi->node_inode);
3617 	sbi->node_inode = NULL;
3618 free_stats:
3619 	f2fs_destroy_stats(sbi);
3620 free_nm:
3621 	f2fs_destroy_node_manager(sbi);
3622 free_sm:
3623 	f2fs_destroy_segment_manager(sbi);
3624 free_devices:
3625 	destroy_device_list(sbi);
3626 	kvfree(sbi->ckpt);
3627 free_meta_inode:
3628 	make_bad_inode(sbi->meta_inode);
3629 	iput(sbi->meta_inode);
3630 	sbi->meta_inode = NULL;
3631 free_io_dummy:
3632 	mempool_destroy(sbi->write_io_dummy);
3633 free_percpu:
3634 	destroy_percpu_info(sbi);
3635 free_bio_info:
3636 	for (i = 0; i < NR_PAGE_TYPE; i++)
3637 		kvfree(sbi->write_io[i]);
3638 
3639 #ifdef CONFIG_UNICODE
3640 	utf8_unload(sbi->s_encoding);
3641 #endif
3642 free_options:
3643 #ifdef CONFIG_QUOTA
3644 	for (i = 0; i < MAXQUOTAS; i++)
3645 		kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
3646 #endif
3647 	kvfree(options);
3648 free_sb_buf:
3649 	kvfree(raw_super);
3650 free_sbi:
3651 	if (sbi->s_chksum_driver)
3652 		crypto_free_shash(sbi->s_chksum_driver);
3653 	kvfree(sbi);
3654 
3655 	/* give only one another chance */
3656 	if (retry_cnt > 0 && skip_recovery) {
3657 		retry_cnt--;
3658 		shrink_dcache_sb(sb);
3659 		goto try_onemore;
3660 	}
3661 	return err;
3662 }
3663 
3664 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3665 			const char *dev_name, void *data)
3666 {
3667 	return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3668 }
3669 
3670 static void kill_f2fs_super(struct super_block *sb)
3671 {
3672 	if (sb->s_root) {
3673 		struct f2fs_sb_info *sbi = F2FS_SB(sb);
3674 
3675 		set_sbi_flag(sbi, SBI_IS_CLOSE);
3676 		f2fs_stop_gc_thread(sbi);
3677 		f2fs_stop_discard_thread(sbi);
3678 
3679 		if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3680 				!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3681 			struct cp_control cpc = {
3682 				.reason = CP_UMOUNT,
3683 			};
3684 			f2fs_write_checkpoint(sbi, &cpc);
3685 		}
3686 
3687 		if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3688 			sb->s_flags &= ~SB_RDONLY;
3689 	}
3690 	kill_block_super(sb);
3691 }
3692 
3693 static struct file_system_type f2fs_fs_type = {
3694 	.owner		= THIS_MODULE,
3695 	.name		= "f2fs",
3696 	.mount		= f2fs_mount,
3697 	.kill_sb	= kill_f2fs_super,
3698 	.fs_flags	= FS_REQUIRES_DEV,
3699 };
3700 MODULE_ALIAS_FS("f2fs");
3701 
3702 static int __init init_inodecache(void)
3703 {
3704 	f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3705 			sizeof(struct f2fs_inode_info), 0,
3706 			SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3707 	if (!f2fs_inode_cachep)
3708 		return -ENOMEM;
3709 	return 0;
3710 }
3711 
3712 static void destroy_inodecache(void)
3713 {
3714 	/*
3715 	 * Make sure all delayed rcu free inodes are flushed before we
3716 	 * destroy cache.
3717 	 */
3718 	rcu_barrier();
3719 	kmem_cache_destroy(f2fs_inode_cachep);
3720 }
3721 
3722 static int __init init_f2fs_fs(void)
3723 {
3724 	int err;
3725 
3726 	if (PAGE_SIZE != F2FS_BLKSIZE) {
3727 		printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3728 				PAGE_SIZE, F2FS_BLKSIZE);
3729 		return -EINVAL;
3730 	}
3731 
3732 	f2fs_build_trace_ios();
3733 
3734 	err = init_inodecache();
3735 	if (err)
3736 		goto fail;
3737 	err = f2fs_create_node_manager_caches();
3738 	if (err)
3739 		goto free_inodecache;
3740 	err = f2fs_create_segment_manager_caches();
3741 	if (err)
3742 		goto free_node_manager_caches;
3743 	err = f2fs_create_checkpoint_caches();
3744 	if (err)
3745 		goto free_segment_manager_caches;
3746 	err = f2fs_create_extent_cache();
3747 	if (err)
3748 		goto free_checkpoint_caches;
3749 	err = f2fs_init_sysfs();
3750 	if (err)
3751 		goto free_extent_cache;
3752 	err = register_shrinker(&f2fs_shrinker_info);
3753 	if (err)
3754 		goto free_sysfs;
3755 	err = register_filesystem(&f2fs_fs_type);
3756 	if (err)
3757 		goto free_shrinker;
3758 	f2fs_create_root_stats();
3759 	err = f2fs_init_post_read_processing();
3760 	if (err)
3761 		goto free_root_stats;
3762 	err = f2fs_init_bio_entry_cache();
3763 	if (err)
3764 		goto free_post_read;
3765 	return 0;
3766 
3767 free_post_read:
3768 	f2fs_destroy_post_read_processing();
3769 free_root_stats:
3770 	f2fs_destroy_root_stats();
3771 	unregister_filesystem(&f2fs_fs_type);
3772 free_shrinker:
3773 	unregister_shrinker(&f2fs_shrinker_info);
3774 free_sysfs:
3775 	f2fs_exit_sysfs();
3776 free_extent_cache:
3777 	f2fs_destroy_extent_cache();
3778 free_checkpoint_caches:
3779 	f2fs_destroy_checkpoint_caches();
3780 free_segment_manager_caches:
3781 	f2fs_destroy_segment_manager_caches();
3782 free_node_manager_caches:
3783 	f2fs_destroy_node_manager_caches();
3784 free_inodecache:
3785 	destroy_inodecache();
3786 fail:
3787 	return err;
3788 }
3789 
3790 static void __exit exit_f2fs_fs(void)
3791 {
3792 	f2fs_destroy_bio_entry_cache();
3793 	f2fs_destroy_post_read_processing();
3794 	f2fs_destroy_root_stats();
3795 	unregister_filesystem(&f2fs_fs_type);
3796 	unregister_shrinker(&f2fs_shrinker_info);
3797 	f2fs_exit_sysfs();
3798 	f2fs_destroy_extent_cache();
3799 	f2fs_destroy_checkpoint_caches();
3800 	f2fs_destroy_segment_manager_caches();
3801 	f2fs_destroy_node_manager_caches();
3802 	destroy_inodecache();
3803 	f2fs_destroy_trace_ios();
3804 }
3805 
3806 module_init(init_f2fs_fs)
3807 module_exit(exit_f2fs_fs)
3808 
3809 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3810 MODULE_DESCRIPTION("Flash Friendly File System");
3811 MODULE_LICENSE("GPL");
3812 
3813