• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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 #include <linux/part_stat.h>
28 
29 #include "f2fs.h"
30 #include "node.h"
31 #include "segment.h"
32 #include "xattr.h"
33 #include "gc.h"
34 #include "trace.h"
35 
36 #define CREATE_TRACE_POINTS
37 #include <trace/events/f2fs.h>
38 
39 #ifdef CONFIG_F2FS_GRADING_SSR
40 #define SSR_DEFALT_SPACE_LIMIT      (5<<20) /* 5G default space limit */
41 #define SSR_DEFALT_WATERLINE        80      /* 80% default waterline */
42 #define SSR_HN_SAPCE_LIMIT_128G     (8<<20) /* 8G default sapce limit for 128G devices */
43 #define SSR_HN_WATERLINE_128G       80      /* 80% default hot node waterline for 128G devices */
44 #define SSR_WN_SAPCE_LIMIT_128G     (5<<20) /* 5G default warm node sapce limit for 128G devices */
45 #define SSR_WN_WATERLINE_128G       70      /* 70% default warm node waterline for 128G devices */
46 #define SSR_HD_SAPCE_LIMIT_128G     (8<<20) /* 8G default hot data sapce limit for 128G devices */
47 #define SSR_HD_WATERLINE_128G       65      /* 65% default hot data waterline for 128G devices */
48 #define SSR_WD_SAPCE_LIMIT_128G     (5<<20) /* 5G default warm data sapce limit for 128G devices */
49 #define SSR_WD_WATERLINE_128G       60      /* 60% default warm data waterline for 128G devices */
50 #endif
51 
52 static struct kmem_cache *f2fs_inode_cachep;
53 
54 #ifdef CONFIG_F2FS_FAULT_INJECTION
55 
56 const char *f2fs_fault_name[FAULT_MAX] = {
57 	[FAULT_KMALLOC]		= "kmalloc",
58 	[FAULT_KVMALLOC]	= "kvmalloc",
59 	[FAULT_PAGE_ALLOC]	= "page alloc",
60 	[FAULT_PAGE_GET]	= "page get",
61 	[FAULT_ALLOC_BIO]	= "alloc bio",
62 	[FAULT_ALLOC_NID]	= "alloc nid",
63 	[FAULT_ORPHAN]		= "orphan",
64 	[FAULT_BLOCK]		= "no more block",
65 	[FAULT_DIR_DEPTH]	= "too big dir depth",
66 	[FAULT_EVICT_INODE]	= "evict_inode fail",
67 	[FAULT_TRUNCATE]	= "truncate fail",
68 	[FAULT_READ_IO]		= "read IO error",
69 	[FAULT_CHECKPOINT]	= "checkpoint error",
70 	[FAULT_DISCARD]		= "discard error",
71 	[FAULT_WRITE_IO]	= "write IO error",
72 };
73 
f2fs_build_fault_attr(struct f2fs_sb_info * sbi,unsigned int rate,unsigned int type)74 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
75 							unsigned int type)
76 {
77 	struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
78 
79 	if (rate) {
80 		atomic_set(&ffi->inject_ops, 0);
81 		ffi->inject_rate = rate;
82 	}
83 
84 	if (type)
85 		ffi->inject_type = type;
86 
87 	if (!rate && !type)
88 		memset(ffi, 0, sizeof(struct f2fs_fault_info));
89 }
90 #endif
91 
92 /* f2fs-wide shrinker description */
93 static struct shrinker f2fs_shrinker_info = {
94 	.scan_objects = f2fs_shrink_scan,
95 	.count_objects = f2fs_shrink_count,
96 	.seeks = DEFAULT_SEEKS,
97 };
98 
99 enum {
100 	Opt_gc_background,
101 	Opt_disable_roll_forward,
102 	Opt_norecovery,
103 	Opt_discard,
104 	Opt_nodiscard,
105 	Opt_noheap,
106 	Opt_heap,
107 	Opt_user_xattr,
108 	Opt_nouser_xattr,
109 	Opt_acl,
110 	Opt_noacl,
111 	Opt_active_logs,
112 	Opt_disable_ext_identify,
113 	Opt_inline_xattr,
114 	Opt_noinline_xattr,
115 	Opt_inline_xattr_size,
116 	Opt_inline_data,
117 	Opt_inline_dentry,
118 	Opt_noinline_dentry,
119 	Opt_flush_merge,
120 	Opt_noflush_merge,
121 	Opt_nobarrier,
122 	Opt_fastboot,
123 	Opt_extent_cache,
124 	Opt_noextent_cache,
125 	Opt_noinline_data,
126 	Opt_data_flush,
127 	Opt_reserve_root,
128 	Opt_resgid,
129 	Opt_resuid,
130 	Opt_mode,
131 	Opt_io_size_bits,
132 	Opt_fault_injection,
133 	Opt_fault_type,
134 	Opt_lazytime,
135 	Opt_nolazytime,
136 	Opt_quota,
137 	Opt_noquota,
138 	Opt_usrquota,
139 	Opt_grpquota,
140 	Opt_prjquota,
141 	Opt_usrjquota,
142 	Opt_grpjquota,
143 	Opt_prjjquota,
144 	Opt_offusrjquota,
145 	Opt_offgrpjquota,
146 	Opt_offprjjquota,
147 	Opt_jqfmt_vfsold,
148 	Opt_jqfmt_vfsv0,
149 	Opt_jqfmt_vfsv1,
150 	Opt_whint,
151 	Opt_alloc,
152 	Opt_fsync,
153 	Opt_test_dummy_encryption,
154 	Opt_inlinecrypt,
155 	Opt_checkpoint_disable,
156 	Opt_checkpoint_disable_cap,
157 	Opt_checkpoint_disable_cap_perc,
158 	Opt_checkpoint_enable,
159 	Opt_compress_algorithm,
160 	Opt_compress_log_size,
161 	Opt_compress_extension,
162 	Opt_atgc,
163 	Opt_gc_merge,
164 	Opt_nogc_merge,
165 	Opt_err,
166 };
167 
168 static match_table_t f2fs_tokens = {
169 	{Opt_gc_background, "background_gc=%s"},
170 	{Opt_disable_roll_forward, "disable_roll_forward"},
171 	{Opt_norecovery, "norecovery"},
172 	{Opt_discard, "discard"},
173 	{Opt_nodiscard, "nodiscard"},
174 	{Opt_noheap, "no_heap"},
175 	{Opt_heap, "heap"},
176 	{Opt_user_xattr, "user_xattr"},
177 	{Opt_nouser_xattr, "nouser_xattr"},
178 	{Opt_acl, "acl"},
179 	{Opt_noacl, "noacl"},
180 	{Opt_active_logs, "active_logs=%u"},
181 	{Opt_disable_ext_identify, "disable_ext_identify"},
182 	{Opt_inline_xattr, "inline_xattr"},
183 	{Opt_noinline_xattr, "noinline_xattr"},
184 	{Opt_inline_xattr_size, "inline_xattr_size=%u"},
185 	{Opt_inline_data, "inline_data"},
186 	{Opt_inline_dentry, "inline_dentry"},
187 	{Opt_noinline_dentry, "noinline_dentry"},
188 	{Opt_flush_merge, "flush_merge"},
189 	{Opt_noflush_merge, "noflush_merge"},
190 	{Opt_nobarrier, "nobarrier"},
191 	{Opt_fastboot, "fastboot"},
192 	{Opt_extent_cache, "extent_cache"},
193 	{Opt_noextent_cache, "noextent_cache"},
194 	{Opt_noinline_data, "noinline_data"},
195 	{Opt_data_flush, "data_flush"},
196 	{Opt_reserve_root, "reserve_root=%u"},
197 	{Opt_resgid, "resgid=%u"},
198 	{Opt_resuid, "resuid=%u"},
199 	{Opt_mode, "mode=%s"},
200 	{Opt_io_size_bits, "io_bits=%u"},
201 	{Opt_fault_injection, "fault_injection=%u"},
202 	{Opt_fault_type, "fault_type=%u"},
203 	{Opt_lazytime, "lazytime"},
204 	{Opt_nolazytime, "nolazytime"},
205 	{Opt_quota, "quota"},
206 	{Opt_noquota, "noquota"},
207 	{Opt_usrquota, "usrquota"},
208 	{Opt_grpquota, "grpquota"},
209 	{Opt_prjquota, "prjquota"},
210 	{Opt_usrjquota, "usrjquota=%s"},
211 	{Opt_grpjquota, "grpjquota=%s"},
212 	{Opt_prjjquota, "prjjquota=%s"},
213 	{Opt_offusrjquota, "usrjquota="},
214 	{Opt_offgrpjquota, "grpjquota="},
215 	{Opt_offprjjquota, "prjjquota="},
216 	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
217 	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
218 	{Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
219 	{Opt_whint, "whint_mode=%s"},
220 	{Opt_alloc, "alloc_mode=%s"},
221 	{Opt_fsync, "fsync_mode=%s"},
222 	{Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
223 	{Opt_test_dummy_encryption, "test_dummy_encryption"},
224 	{Opt_inlinecrypt, "inlinecrypt"},
225 	{Opt_checkpoint_disable, "checkpoint=disable"},
226 	{Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
227 	{Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
228 	{Opt_checkpoint_enable, "checkpoint=enable"},
229 	{Opt_compress_algorithm, "compress_algorithm=%s"},
230 	{Opt_compress_log_size, "compress_log_size=%u"},
231 	{Opt_compress_extension, "compress_extension=%s"},
232 	{Opt_atgc, "atgc"},
233 	{Opt_gc_merge, "gc_merge"},
234 	{Opt_nogc_merge, "nogc_merge"},
235 	{Opt_err, NULL},
236 };
237 
f2fs_printk(struct f2fs_sb_info * sbi,const char * fmt,...)238 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
239 {
240 	struct va_format vaf;
241 	va_list args;
242 	int level;
243 
244 	va_start(args, fmt);
245 
246 	level = printk_get_level(fmt);
247 	vaf.fmt = printk_skip_level(fmt);
248 	vaf.va = &args;
249 	printk("%c%cF2FS-fs (%s): %pV\n",
250 	       KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
251 
252 	va_end(args);
253 }
254 
255 #ifdef CONFIG_UNICODE
256 static const struct f2fs_sb_encodings {
257 	__u16 magic;
258 	char *name;
259 	char *version;
260 } f2fs_sb_encoding_map[] = {
261 	{F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"},
262 };
263 
f2fs_sb_read_encoding(const struct f2fs_super_block * sb,const struct f2fs_sb_encodings ** encoding,__u16 * flags)264 static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb,
265 				 const struct f2fs_sb_encodings **encoding,
266 				 __u16 *flags)
267 {
268 	__u16 magic = le16_to_cpu(sb->s_encoding);
269 	int i;
270 
271 	for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
272 		if (magic == f2fs_sb_encoding_map[i].magic)
273 			break;
274 
275 	if (i >= ARRAY_SIZE(f2fs_sb_encoding_map))
276 		return -EINVAL;
277 
278 	*encoding = &f2fs_sb_encoding_map[i];
279 	*flags = le16_to_cpu(sb->s_encoding_flags);
280 
281 	return 0;
282 }
283 #endif
284 
limit_reserve_root(struct f2fs_sb_info * sbi)285 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
286 {
287 	block_t limit = min((sbi->user_block_count >> 3),
288 			sbi->user_block_count - sbi->reserved_blocks);
289 
290 	/* limit is 12.5% */
291 	if (test_opt(sbi, RESERVE_ROOT) &&
292 			F2FS_OPTION(sbi).root_reserved_blocks > limit) {
293 		F2FS_OPTION(sbi).root_reserved_blocks = limit;
294 		f2fs_info(sbi, "Reduce reserved blocks for root = %u",
295 			  F2FS_OPTION(sbi).root_reserved_blocks);
296 	}
297 	if (!test_opt(sbi, RESERVE_ROOT) &&
298 		(!uid_eq(F2FS_OPTION(sbi).s_resuid,
299 				make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
300 		!gid_eq(F2FS_OPTION(sbi).s_resgid,
301 				make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
302 		f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
303 			  from_kuid_munged(&init_user_ns,
304 					   F2FS_OPTION(sbi).s_resuid),
305 			  from_kgid_munged(&init_user_ns,
306 					   F2FS_OPTION(sbi).s_resgid));
307 }
308 
adjust_reserved_segment(struct f2fs_sb_info * sbi)309 static inline int adjust_reserved_segment(struct f2fs_sb_info *sbi)
310 {
311 	unsigned int sec_blks = sbi->blocks_per_seg * sbi->segs_per_sec;
312 	unsigned int avg_vblocks;
313 	unsigned int wanted_reserved_segments;
314 	block_t avail_user_block_count;
315 
316 	if (!F2FS_IO_ALIGNED(sbi))
317 		return 0;
318 
319 	/* average valid block count in section in worst case */
320 	avg_vblocks = sec_blks / F2FS_IO_SIZE(sbi);
321 
322 	/*
323 	 * we need enough free space when migrating one section in worst case
324 	 */
325 	wanted_reserved_segments = (F2FS_IO_SIZE(sbi) / avg_vblocks) *
326 						reserved_segments(sbi);
327 	wanted_reserved_segments -= reserved_segments(sbi);
328 
329 	avail_user_block_count = sbi->user_block_count -
330 				sbi->current_reserved_blocks -
331 				F2FS_OPTION(sbi).root_reserved_blocks;
332 
333 	if (wanted_reserved_segments * sbi->blocks_per_seg >
334 					avail_user_block_count) {
335 		f2fs_err(sbi, "IO align feature can't grab additional reserved segment: %u, available segments: %u",
336 			wanted_reserved_segments,
337 			avail_user_block_count >> sbi->log_blocks_per_seg);
338 		return -ENOSPC;
339 	}
340 
341 	SM_I(sbi)->additional_reserved_segments = wanted_reserved_segments;
342 
343 	f2fs_info(sbi, "IO align feature needs additional reserved segment: %u",
344 			 wanted_reserved_segments);
345 
346 	return 0;
347 }
348 
adjust_unusable_cap_perc(struct f2fs_sb_info * sbi)349 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
350 {
351 	if (!F2FS_OPTION(sbi).unusable_cap_perc)
352 		return;
353 
354 	if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
355 		F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
356 	else
357 		F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
358 					F2FS_OPTION(sbi).unusable_cap_perc;
359 
360 	f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
361 			F2FS_OPTION(sbi).unusable_cap,
362 			F2FS_OPTION(sbi).unusable_cap_perc);
363 }
364 
init_once(void * foo)365 static void init_once(void *foo)
366 {
367 	struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
368 
369 	inode_init_once(&fi->vfs_inode);
370 }
371 
372 #ifdef CONFIG_QUOTA
373 static const char * const quotatypes[] = INITQFNAMES;
374 #define QTYPE2NAME(t) (quotatypes[t])
f2fs_set_qf_name(struct super_block * sb,int qtype,substring_t * args)375 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
376 							substring_t *args)
377 {
378 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
379 	char *qname;
380 	int ret = -EINVAL;
381 
382 	if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
383 		f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
384 		return -EINVAL;
385 	}
386 	if (f2fs_sb_has_quota_ino(sbi)) {
387 		f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
388 		return 0;
389 	}
390 
391 	qname = match_strdup(args);
392 	if (!qname) {
393 		f2fs_err(sbi, "Not enough memory for storing quotafile name");
394 		return -ENOMEM;
395 	}
396 	if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
397 		if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
398 			ret = 0;
399 		else
400 			f2fs_err(sbi, "%s quota file already specified",
401 				 QTYPE2NAME(qtype));
402 		goto errout;
403 	}
404 	if (strchr(qname, '/')) {
405 		f2fs_err(sbi, "quotafile must be on filesystem root");
406 		goto errout;
407 	}
408 	F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
409 	set_opt(sbi, QUOTA);
410 	return 0;
411 errout:
412 	kfree(qname);
413 	return ret;
414 }
415 
f2fs_clear_qf_name(struct super_block * sb,int qtype)416 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
417 {
418 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
419 
420 	if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
421 		f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
422 		return -EINVAL;
423 	}
424 	kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
425 	F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
426 	return 0;
427 }
428 
f2fs_check_quota_options(struct f2fs_sb_info * sbi)429 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
430 {
431 	/*
432 	 * We do the test below only for project quotas. 'usrquota' and
433 	 * 'grpquota' mount options are allowed even without quota feature
434 	 * to support legacy quotas in quota files.
435 	 */
436 	if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
437 		f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
438 		return -1;
439 	}
440 	if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
441 			F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
442 			F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
443 		if (test_opt(sbi, USRQUOTA) &&
444 				F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
445 			clear_opt(sbi, USRQUOTA);
446 
447 		if (test_opt(sbi, GRPQUOTA) &&
448 				F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
449 			clear_opt(sbi, GRPQUOTA);
450 
451 		if (test_opt(sbi, PRJQUOTA) &&
452 				F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
453 			clear_opt(sbi, PRJQUOTA);
454 
455 		if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
456 				test_opt(sbi, PRJQUOTA)) {
457 			f2fs_err(sbi, "old and new quota format mixing");
458 			return -1;
459 		}
460 
461 		if (!F2FS_OPTION(sbi).s_jquota_fmt) {
462 			f2fs_err(sbi, "journaled quota format not specified");
463 			return -1;
464 		}
465 	}
466 
467 	if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
468 		f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
469 		F2FS_OPTION(sbi).s_jquota_fmt = 0;
470 	}
471 	return 0;
472 }
473 #endif
474 
f2fs_set_test_dummy_encryption(struct super_block * sb,const char * opt,const substring_t * arg,bool is_remount)475 static int f2fs_set_test_dummy_encryption(struct super_block *sb,
476 					  const char *opt,
477 					  const substring_t *arg,
478 					  bool is_remount)
479 {
480 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
481 #ifdef CONFIG_FS_ENCRYPTION
482 	int err;
483 
484 	if (!f2fs_sb_has_encrypt(sbi)) {
485 		f2fs_err(sbi, "Encrypt feature is off");
486 		return -EINVAL;
487 	}
488 
489 	/*
490 	 * This mount option is just for testing, and it's not worthwhile to
491 	 * implement the extra complexity (e.g. RCU protection) that would be
492 	 * needed to allow it to be set or changed during remount.  We do allow
493 	 * it to be specified during remount, but only if there is no change.
494 	 */
495 	if (is_remount && !F2FS_OPTION(sbi).dummy_enc_policy.policy) {
496 		f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
497 		return -EINVAL;
498 	}
499 	err = fscrypt_set_test_dummy_encryption(
500 		sb, arg->from, &F2FS_OPTION(sbi).dummy_enc_policy);
501 	if (err) {
502 		if (err == -EEXIST)
503 			f2fs_warn(sbi,
504 				  "Can't change test_dummy_encryption on remount");
505 		else if (err == -EINVAL)
506 			f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
507 				  opt);
508 		else
509 			f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
510 				  opt, err);
511 		return -EINVAL;
512 	}
513 	f2fs_warn(sbi, "Test dummy encryption mode enabled");
514 #else
515 	f2fs_warn(sbi, "Test dummy encryption mount option ignored");
516 #endif
517 	return 0;
518 }
519 
parse_options(struct super_block * sb,char * options,bool is_remount)520 static int parse_options(struct super_block *sb, char *options, bool is_remount)
521 {
522 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
523 	substring_t args[MAX_OPT_ARGS];
524 #ifdef CONFIG_F2FS_FS_COMPRESSION
525 	unsigned char (*ext)[F2FS_EXTENSION_LEN];
526 	int ext_cnt;
527 #endif
528 	char *p, *name;
529 	int arg = 0;
530 	kuid_t uid;
531 	kgid_t gid;
532 	int ret;
533 
534 	if (!options)
535 		return 0;
536 
537 	while ((p = strsep(&options, ",")) != NULL) {
538 		int token;
539 		if (!*p)
540 			continue;
541 		/*
542 		 * Initialize args struct so we know whether arg was
543 		 * found; some options take optional arguments.
544 		 */
545 		args[0].to = args[0].from = NULL;
546 		token = match_token(p, f2fs_tokens, args);
547 
548 		switch (token) {
549 		case Opt_gc_background:
550 			name = match_strdup(&args[0]);
551 
552 			if (!name)
553 				return -ENOMEM;
554 			if (!strcmp(name, "on")) {
555 				F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
556 			} else if (!strcmp(name, "off")) {
557 				F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
558 			} else if (!strcmp(name, "sync")) {
559 				F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
560 			} else {
561 				kfree(name);
562 				return -EINVAL;
563 			}
564 			kfree(name);
565 			break;
566 		case Opt_disable_roll_forward:
567 			set_opt(sbi, DISABLE_ROLL_FORWARD);
568 			break;
569 		case Opt_norecovery:
570 			/* this option mounts f2fs with ro */
571 			set_opt(sbi, NORECOVERY);
572 			if (!f2fs_readonly(sb))
573 				return -EINVAL;
574 			break;
575 		case Opt_discard:
576 			set_opt(sbi, DISCARD);
577 			break;
578 		case Opt_nodiscard:
579 			if (f2fs_sb_has_blkzoned(sbi)) {
580 				f2fs_warn(sbi, "discard is required for zoned block devices");
581 				return -EINVAL;
582 			}
583 			clear_opt(sbi, DISCARD);
584 			break;
585 		case Opt_noheap:
586 			set_opt(sbi, NOHEAP);
587 			break;
588 		case Opt_heap:
589 			clear_opt(sbi, NOHEAP);
590 			break;
591 #ifdef CONFIG_F2FS_FS_XATTR
592 		case Opt_user_xattr:
593 			set_opt(sbi, XATTR_USER);
594 			break;
595 		case Opt_nouser_xattr:
596 			clear_opt(sbi, XATTR_USER);
597 			break;
598 		case Opt_inline_xattr:
599 			set_opt(sbi, INLINE_XATTR);
600 			break;
601 		case Opt_noinline_xattr:
602 			clear_opt(sbi, INLINE_XATTR);
603 			break;
604 		case Opt_inline_xattr_size:
605 			if (args->from && match_int(args, &arg))
606 				return -EINVAL;
607 			set_opt(sbi, INLINE_XATTR_SIZE);
608 			F2FS_OPTION(sbi).inline_xattr_size = arg;
609 			break;
610 #else
611 		case Opt_user_xattr:
612 			f2fs_info(sbi, "user_xattr options not supported");
613 			break;
614 		case Opt_nouser_xattr:
615 			f2fs_info(sbi, "nouser_xattr options not supported");
616 			break;
617 		case Opt_inline_xattr:
618 			f2fs_info(sbi, "inline_xattr options not supported");
619 			break;
620 		case Opt_noinline_xattr:
621 			f2fs_info(sbi, "noinline_xattr options not supported");
622 			break;
623 #endif
624 #ifdef CONFIG_F2FS_FS_POSIX_ACL
625 		case Opt_acl:
626 			set_opt(sbi, POSIX_ACL);
627 			break;
628 		case Opt_noacl:
629 			clear_opt(sbi, POSIX_ACL);
630 			break;
631 #else
632 		case Opt_acl:
633 			f2fs_info(sbi, "acl options not supported");
634 			break;
635 		case Opt_noacl:
636 			f2fs_info(sbi, "noacl options not supported");
637 			break;
638 #endif
639 		case Opt_active_logs:
640 			if (args->from && match_int(args, &arg))
641 				return -EINVAL;
642 			if (arg != 2 && arg != 4 &&
643 				arg != NR_CURSEG_PERSIST_TYPE)
644 				return -EINVAL;
645 			F2FS_OPTION(sbi).active_logs = arg;
646 			break;
647 		case Opt_disable_ext_identify:
648 			set_opt(sbi, DISABLE_EXT_IDENTIFY);
649 			break;
650 		case Opt_inline_data:
651 			set_opt(sbi, INLINE_DATA);
652 			break;
653 		case Opt_inline_dentry:
654 			set_opt(sbi, INLINE_DENTRY);
655 			break;
656 		case Opt_noinline_dentry:
657 			clear_opt(sbi, INLINE_DENTRY);
658 			break;
659 		case Opt_flush_merge:
660 			set_opt(sbi, FLUSH_MERGE);
661 			break;
662 		case Opt_noflush_merge:
663 			clear_opt(sbi, FLUSH_MERGE);
664 			break;
665 		case Opt_nobarrier:
666 			set_opt(sbi, NOBARRIER);
667 			break;
668 		case Opt_fastboot:
669 			set_opt(sbi, FASTBOOT);
670 			break;
671 		case Opt_extent_cache:
672 			set_opt(sbi, EXTENT_CACHE);
673 			break;
674 		case Opt_noextent_cache:
675 			clear_opt(sbi, EXTENT_CACHE);
676 			break;
677 		case Opt_noinline_data:
678 			clear_opt(sbi, INLINE_DATA);
679 			break;
680 		case Opt_data_flush:
681 			set_opt(sbi, DATA_FLUSH);
682 			break;
683 		case Opt_reserve_root:
684 			if (args->from && match_int(args, &arg))
685 				return -EINVAL;
686 			if (test_opt(sbi, RESERVE_ROOT)) {
687 				f2fs_info(sbi, "Preserve previous reserve_root=%u",
688 					  F2FS_OPTION(sbi).root_reserved_blocks);
689 			} else {
690 				F2FS_OPTION(sbi).root_reserved_blocks = arg;
691 				set_opt(sbi, RESERVE_ROOT);
692 			}
693 			break;
694 		case Opt_resuid:
695 			if (args->from && match_int(args, &arg))
696 				return -EINVAL;
697 			uid = make_kuid(current_user_ns(), arg);
698 			if (!uid_valid(uid)) {
699 				f2fs_err(sbi, "Invalid uid value %d", arg);
700 				return -EINVAL;
701 			}
702 			F2FS_OPTION(sbi).s_resuid = uid;
703 			break;
704 		case Opt_resgid:
705 			if (args->from && match_int(args, &arg))
706 				return -EINVAL;
707 			gid = make_kgid(current_user_ns(), arg);
708 			if (!gid_valid(gid)) {
709 				f2fs_err(sbi, "Invalid gid value %d", arg);
710 				return -EINVAL;
711 			}
712 			F2FS_OPTION(sbi).s_resgid = gid;
713 			break;
714 		case Opt_mode:
715 			name = match_strdup(&args[0]);
716 
717 			if (!name)
718 				return -ENOMEM;
719 			if (!strcmp(name, "adaptive")) {
720 				if (f2fs_sb_has_blkzoned(sbi)) {
721 					f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
722 					kfree(name);
723 					return -EINVAL;
724 				}
725 				F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
726 			} else if (!strcmp(name, "lfs")) {
727 				F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
728 			} else {
729 				kfree(name);
730 				return -EINVAL;
731 			}
732 			kfree(name);
733 			break;
734 		case Opt_io_size_bits:
735 			if (args->from && match_int(args, &arg))
736 				return -EINVAL;
737 			if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_PAGES)) {
738 				f2fs_warn(sbi, "Not support %d, larger than %d",
739 					  1 << arg, BIO_MAX_PAGES);
740 				return -EINVAL;
741 			}
742 			F2FS_OPTION(sbi).write_io_size_bits = arg;
743 			break;
744 #ifdef CONFIG_F2FS_FAULT_INJECTION
745 		case Opt_fault_injection:
746 			if (args->from && match_int(args, &arg))
747 				return -EINVAL;
748 			f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
749 			set_opt(sbi, FAULT_INJECTION);
750 			break;
751 
752 		case Opt_fault_type:
753 			if (args->from && match_int(args, &arg))
754 				return -EINVAL;
755 			f2fs_build_fault_attr(sbi, 0, arg);
756 			set_opt(sbi, FAULT_INJECTION);
757 			break;
758 #else
759 		case Opt_fault_injection:
760 			f2fs_info(sbi, "fault_injection options not supported");
761 			break;
762 
763 		case Opt_fault_type:
764 			f2fs_info(sbi, "fault_type options not supported");
765 			break;
766 #endif
767 		case Opt_lazytime:
768 			sb->s_flags |= SB_LAZYTIME;
769 			break;
770 		case Opt_nolazytime:
771 			sb->s_flags &= ~SB_LAZYTIME;
772 			break;
773 #ifdef CONFIG_QUOTA
774 		case Opt_quota:
775 		case Opt_usrquota:
776 			set_opt(sbi, USRQUOTA);
777 			break;
778 		case Opt_grpquota:
779 			set_opt(sbi, GRPQUOTA);
780 			break;
781 		case Opt_prjquota:
782 			set_opt(sbi, PRJQUOTA);
783 			break;
784 		case Opt_usrjquota:
785 			ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
786 			if (ret)
787 				return ret;
788 			break;
789 		case Opt_grpjquota:
790 			ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
791 			if (ret)
792 				return ret;
793 			break;
794 		case Opt_prjjquota:
795 			ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
796 			if (ret)
797 				return ret;
798 			break;
799 		case Opt_offusrjquota:
800 			ret = f2fs_clear_qf_name(sb, USRQUOTA);
801 			if (ret)
802 				return ret;
803 			break;
804 		case Opt_offgrpjquota:
805 			ret = f2fs_clear_qf_name(sb, GRPQUOTA);
806 			if (ret)
807 				return ret;
808 			break;
809 		case Opt_offprjjquota:
810 			ret = f2fs_clear_qf_name(sb, PRJQUOTA);
811 			if (ret)
812 				return ret;
813 			break;
814 		case Opt_jqfmt_vfsold:
815 			F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
816 			break;
817 		case Opt_jqfmt_vfsv0:
818 			F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
819 			break;
820 		case Opt_jqfmt_vfsv1:
821 			F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
822 			break;
823 		case Opt_noquota:
824 			clear_opt(sbi, QUOTA);
825 			clear_opt(sbi, USRQUOTA);
826 			clear_opt(sbi, GRPQUOTA);
827 			clear_opt(sbi, PRJQUOTA);
828 			break;
829 #else
830 		case Opt_quota:
831 		case Opt_usrquota:
832 		case Opt_grpquota:
833 		case Opt_prjquota:
834 		case Opt_usrjquota:
835 		case Opt_grpjquota:
836 		case Opt_prjjquota:
837 		case Opt_offusrjquota:
838 		case Opt_offgrpjquota:
839 		case Opt_offprjjquota:
840 		case Opt_jqfmt_vfsold:
841 		case Opt_jqfmt_vfsv0:
842 		case Opt_jqfmt_vfsv1:
843 		case Opt_noquota:
844 			f2fs_info(sbi, "quota operations not supported");
845 			break;
846 #endif
847 		case Opt_whint:
848 			name = match_strdup(&args[0]);
849 			if (!name)
850 				return -ENOMEM;
851 			if (!strcmp(name, "user-based")) {
852 				F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
853 			} else if (!strcmp(name, "off")) {
854 				F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
855 			} else if (!strcmp(name, "fs-based")) {
856 				F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
857 			} else {
858 				kfree(name);
859 				return -EINVAL;
860 			}
861 			kfree(name);
862 			break;
863 		case Opt_alloc:
864 			name = match_strdup(&args[0]);
865 			if (!name)
866 				return -ENOMEM;
867 
868 			if (!strcmp(name, "default")) {
869 				F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
870 			} else if (!strcmp(name, "reuse")) {
871 				F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
872 			} else {
873 				kfree(name);
874 				return -EINVAL;
875 			}
876 			kfree(name);
877 			break;
878 		case Opt_fsync:
879 			name = match_strdup(&args[0]);
880 			if (!name)
881 				return -ENOMEM;
882 			if (!strcmp(name, "posix")) {
883 				F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
884 			} else if (!strcmp(name, "strict")) {
885 				F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
886 			} else if (!strcmp(name, "nobarrier")) {
887 				F2FS_OPTION(sbi).fsync_mode =
888 							FSYNC_MODE_NOBARRIER;
889 			} else {
890 				kfree(name);
891 				return -EINVAL;
892 			}
893 			kfree(name);
894 			break;
895 		case Opt_test_dummy_encryption:
896 			ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
897 							     is_remount);
898 			if (ret)
899 				return ret;
900 			break;
901 		case Opt_inlinecrypt:
902 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
903 			sb->s_flags |= SB_INLINECRYPT;
904 #else
905 			f2fs_info(sbi, "inline encryption not supported");
906 #endif
907 			break;
908 		case Opt_checkpoint_disable_cap_perc:
909 			if (args->from && match_int(args, &arg))
910 				return -EINVAL;
911 			if (arg < 0 || arg > 100)
912 				return -EINVAL;
913 			F2FS_OPTION(sbi).unusable_cap_perc = arg;
914 			set_opt(sbi, DISABLE_CHECKPOINT);
915 			break;
916 		case Opt_checkpoint_disable_cap:
917 			if (args->from && match_int(args, &arg))
918 				return -EINVAL;
919 			F2FS_OPTION(sbi).unusable_cap = arg;
920 			set_opt(sbi, DISABLE_CHECKPOINT);
921 			break;
922 		case Opt_checkpoint_disable:
923 			set_opt(sbi, DISABLE_CHECKPOINT);
924 			break;
925 		case Opt_checkpoint_enable:
926 			clear_opt(sbi, DISABLE_CHECKPOINT);
927 			break;
928 #ifdef CONFIG_F2FS_FS_COMPRESSION
929 		case Opt_compress_algorithm:
930 			if (!f2fs_sb_has_compression(sbi)) {
931 				f2fs_info(sbi, "Image doesn't support compression");
932 				break;
933 			}
934 			name = match_strdup(&args[0]);
935 			if (!name)
936 				return -ENOMEM;
937 			if (!strcmp(name, "lzo")) {
938 				F2FS_OPTION(sbi).compress_algorithm =
939 								COMPRESS_LZO;
940 			} else if (!strcmp(name, "lz4")) {
941 				F2FS_OPTION(sbi).compress_algorithm =
942 								COMPRESS_LZ4;
943 			} else if (!strcmp(name, "zstd")) {
944 				F2FS_OPTION(sbi).compress_algorithm =
945 								COMPRESS_ZSTD;
946 			} else if (!strcmp(name, "lzo-rle")) {
947 				F2FS_OPTION(sbi).compress_algorithm =
948 								COMPRESS_LZORLE;
949 			} else {
950 				kfree(name);
951 				return -EINVAL;
952 			}
953 			kfree(name);
954 			break;
955 		case Opt_compress_log_size:
956 			if (!f2fs_sb_has_compression(sbi)) {
957 				f2fs_info(sbi, "Image doesn't support compression");
958 				break;
959 			}
960 			if (args->from && match_int(args, &arg))
961 				return -EINVAL;
962 			if (arg < MIN_COMPRESS_LOG_SIZE ||
963 				arg > MAX_COMPRESS_LOG_SIZE) {
964 				f2fs_err(sbi,
965 					"Compress cluster log size is out of range");
966 				return -EINVAL;
967 			}
968 			F2FS_OPTION(sbi).compress_log_size = arg;
969 			break;
970 		case Opt_compress_extension:
971 			if (!f2fs_sb_has_compression(sbi)) {
972 				f2fs_info(sbi, "Image doesn't support compression");
973 				break;
974 			}
975 			name = match_strdup(&args[0]);
976 			if (!name)
977 				return -ENOMEM;
978 
979 			ext = F2FS_OPTION(sbi).extensions;
980 			ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
981 
982 			if (strlen(name) >= F2FS_EXTENSION_LEN ||
983 				ext_cnt >= COMPRESS_EXT_NUM) {
984 				f2fs_err(sbi,
985 					"invalid extension length/number");
986 				kfree(name);
987 				return -EINVAL;
988 			}
989 
990 			strcpy(ext[ext_cnt], name);
991 			F2FS_OPTION(sbi).compress_ext_cnt++;
992 			kfree(name);
993 			break;
994 #else
995 		case Opt_compress_algorithm:
996 		case Opt_compress_log_size:
997 		case Opt_compress_extension:
998 			f2fs_info(sbi, "compression options not supported");
999 			break;
1000 #endif
1001 		case Opt_atgc:
1002 			set_opt(sbi, ATGC);
1003 			break;
1004 		case Opt_gc_merge:
1005 			set_opt(sbi, GC_MERGE);
1006 			break;
1007 		case Opt_nogc_merge:
1008 			clear_opt(sbi, GC_MERGE);
1009 			break;
1010 		default:
1011 			f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
1012 				 p);
1013 			return -EINVAL;
1014 		}
1015 	}
1016 #ifdef CONFIG_QUOTA
1017 	if (f2fs_check_quota_options(sbi))
1018 		return -EINVAL;
1019 #else
1020 	if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
1021 		f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1022 		return -EINVAL;
1023 	}
1024 	if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
1025 		f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1026 		return -EINVAL;
1027 	}
1028 #endif
1029 #ifndef CONFIG_UNICODE
1030 	if (f2fs_sb_has_casefold(sbi)) {
1031 		f2fs_err(sbi,
1032 			"Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
1033 		return -EINVAL;
1034 	}
1035 #endif
1036 	/*
1037 	 * The BLKZONED feature indicates that the drive was formatted with
1038 	 * zone alignment optimization. This is optional for host-aware
1039 	 * devices, but mandatory for host-managed zoned block devices.
1040 	 */
1041 #ifndef CONFIG_BLK_DEV_ZONED
1042 	if (f2fs_sb_has_blkzoned(sbi)) {
1043 		f2fs_err(sbi, "Zoned block device support is not enabled");
1044 		return -EINVAL;
1045 	}
1046 #endif
1047 
1048 	if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) {
1049 		f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
1050 			 F2FS_IO_SIZE_KB(sbi));
1051 		return -EINVAL;
1052 	}
1053 
1054 	if (test_opt(sbi, INLINE_XATTR_SIZE)) {
1055 		int min_size, max_size;
1056 
1057 		if (!f2fs_sb_has_extra_attr(sbi) ||
1058 			!f2fs_sb_has_flexible_inline_xattr(sbi)) {
1059 			f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
1060 			return -EINVAL;
1061 		}
1062 		if (!test_opt(sbi, INLINE_XATTR)) {
1063 			f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
1064 			return -EINVAL;
1065 		}
1066 
1067 		min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
1068 		max_size = MAX_INLINE_XATTR_SIZE;
1069 
1070 		if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
1071 				F2FS_OPTION(sbi).inline_xattr_size > max_size) {
1072 			f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
1073 				 min_size, max_size);
1074 			return -EINVAL;
1075 		}
1076 	}
1077 
1078 	if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) {
1079 		f2fs_err(sbi, "LFS not compatible with checkpoint=disable\n");
1080 		return -EINVAL;
1081 	}
1082 
1083 	/* Not pass down write hints if the number of active logs is lesser
1084 	 * than NR_CURSEG_PERSIST_TYPE.
1085 	 */
1086 	if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_PERSIST_TYPE)
1087 		F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1088 	return 0;
1089 }
1090 
f2fs_alloc_inode(struct super_block * sb)1091 static struct inode *f2fs_alloc_inode(struct super_block *sb)
1092 {
1093 	struct f2fs_inode_info *fi;
1094 
1095 	fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
1096 	if (!fi)
1097 		return NULL;
1098 
1099 	init_once((void *) fi);
1100 
1101 	/* Initialize f2fs-specific inode info */
1102 	atomic_set(&fi->dirty_pages, 0);
1103 	atomic_set(&fi->i_compr_blocks, 0);
1104 	init_rwsem(&fi->i_sem);
1105 	spin_lock_init(&fi->i_size_lock);
1106 	INIT_LIST_HEAD(&fi->dirty_list);
1107 	INIT_LIST_HEAD(&fi->gdirty_list);
1108 	INIT_LIST_HEAD(&fi->inmem_ilist);
1109 	INIT_LIST_HEAD(&fi->inmem_pages);
1110 	mutex_init(&fi->inmem_lock);
1111 	init_rwsem(&fi->i_gc_rwsem[READ]);
1112 	init_rwsem(&fi->i_gc_rwsem[WRITE]);
1113 	init_rwsem(&fi->i_mmap_sem);
1114 	init_rwsem(&fi->i_xattr_sem);
1115 
1116 	/* Will be used by directory only */
1117 	fi->i_dir_level = F2FS_SB(sb)->dir_level;
1118 
1119 	fi->ra_offset = -1;
1120 
1121 	return &fi->vfs_inode;
1122 }
1123 
f2fs_drop_inode(struct inode * inode)1124 static int f2fs_drop_inode(struct inode *inode)
1125 {
1126 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1127 	int ret;
1128 
1129 	/*
1130 	 * during filesystem shutdown, if checkpoint is disabled,
1131 	 * drop useless meta/node dirty pages.
1132 	 */
1133 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1134 		if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1135 			inode->i_ino == F2FS_META_INO(sbi)) {
1136 			trace_f2fs_drop_inode(inode, 1);
1137 			return 1;
1138 		}
1139 	}
1140 
1141 	/*
1142 	 * This is to avoid a deadlock condition like below.
1143 	 * writeback_single_inode(inode)
1144 	 *  - f2fs_write_data_page
1145 	 *    - f2fs_gc -> iput -> evict
1146 	 *       - inode_wait_for_writeback(inode)
1147 	 */
1148 	if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1149 		if (!inode->i_nlink && !is_bad_inode(inode)) {
1150 			/* to avoid evict_inode call simultaneously */
1151 			atomic_inc(&inode->i_count);
1152 			spin_unlock(&inode->i_lock);
1153 
1154 			/* some remained atomic pages should discarded */
1155 			if (f2fs_is_atomic_file(inode))
1156 				f2fs_drop_inmem_pages(inode);
1157 
1158 			/* should remain fi->extent_tree for writepage */
1159 			f2fs_destroy_extent_node(inode);
1160 
1161 			sb_start_intwrite(inode->i_sb);
1162 			f2fs_i_size_write(inode, 0);
1163 
1164 			f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1165 					inode, NULL, 0, DATA);
1166 			truncate_inode_pages_final(inode->i_mapping);
1167 
1168 			if (F2FS_HAS_BLOCKS(inode))
1169 				f2fs_truncate(inode);
1170 
1171 			sb_end_intwrite(inode->i_sb);
1172 
1173 			spin_lock(&inode->i_lock);
1174 			atomic_dec(&inode->i_count);
1175 		}
1176 		trace_f2fs_drop_inode(inode, 0);
1177 		return 0;
1178 	}
1179 	ret = generic_drop_inode(inode);
1180 	if (!ret)
1181 		ret = fscrypt_drop_inode(inode);
1182 	trace_f2fs_drop_inode(inode, ret);
1183 	return ret;
1184 }
1185 
f2fs_inode_dirtied(struct inode * inode,bool sync)1186 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1187 {
1188 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1189 	int ret = 0;
1190 
1191 	spin_lock(&sbi->inode_lock[DIRTY_META]);
1192 	if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1193 		ret = 1;
1194 	} else {
1195 		set_inode_flag(inode, FI_DIRTY_INODE);
1196 		stat_inc_dirty_inode(sbi, DIRTY_META);
1197 	}
1198 	if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1199 		list_add_tail(&F2FS_I(inode)->gdirty_list,
1200 				&sbi->inode_list[DIRTY_META]);
1201 		inc_page_count(sbi, F2FS_DIRTY_IMETA);
1202 	}
1203 	spin_unlock(&sbi->inode_lock[DIRTY_META]);
1204 	return ret;
1205 }
1206 
f2fs_inode_synced(struct inode * inode)1207 void f2fs_inode_synced(struct inode *inode)
1208 {
1209 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1210 
1211 	spin_lock(&sbi->inode_lock[DIRTY_META]);
1212 	if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1213 		spin_unlock(&sbi->inode_lock[DIRTY_META]);
1214 		return;
1215 	}
1216 	if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1217 		list_del_init(&F2FS_I(inode)->gdirty_list);
1218 		dec_page_count(sbi, F2FS_DIRTY_IMETA);
1219 	}
1220 	clear_inode_flag(inode, FI_DIRTY_INODE);
1221 	clear_inode_flag(inode, FI_AUTO_RECOVER);
1222 	stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1223 	spin_unlock(&sbi->inode_lock[DIRTY_META]);
1224 }
1225 
1226 /*
1227  * f2fs_dirty_inode() is called from __mark_inode_dirty()
1228  *
1229  * We should call set_dirty_inode to write the dirty inode through write_inode.
1230  */
f2fs_dirty_inode(struct inode * inode,int flags)1231 static void f2fs_dirty_inode(struct inode *inode, int flags)
1232 {
1233 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1234 
1235 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1236 			inode->i_ino == F2FS_META_INO(sbi))
1237 		return;
1238 
1239 	if (flags == I_DIRTY_TIME)
1240 		return;
1241 
1242 	if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1243 		clear_inode_flag(inode, FI_AUTO_RECOVER);
1244 
1245 	f2fs_inode_dirtied(inode, false);
1246 }
1247 
f2fs_free_inode(struct inode * inode)1248 static void f2fs_free_inode(struct inode *inode)
1249 {
1250 	fscrypt_free_inode(inode);
1251 	kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1252 }
1253 
destroy_percpu_info(struct f2fs_sb_info * sbi)1254 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1255 {
1256 	percpu_counter_destroy(&sbi->alloc_valid_block_count);
1257 	percpu_counter_destroy(&sbi->total_valid_inode_count);
1258 }
1259 
destroy_device_list(struct f2fs_sb_info * sbi)1260 static void destroy_device_list(struct f2fs_sb_info *sbi)
1261 {
1262 	int i;
1263 
1264 	for (i = 0; i < sbi->s_ndevs; i++) {
1265 		blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1266 #ifdef CONFIG_BLK_DEV_ZONED
1267 		kvfree(FDEV(i).blkz_seq);
1268 #endif
1269 	}
1270 	kvfree(sbi->devs);
1271 }
1272 
f2fs_put_super(struct super_block * sb)1273 static void f2fs_put_super(struct super_block *sb)
1274 {
1275 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1276 	int i;
1277 	bool dropped;
1278 
1279 	/* unregister procfs/sysfs entries in advance to avoid race case */
1280 	f2fs_unregister_sysfs(sbi);
1281 
1282 	f2fs_quota_off_umount(sb);
1283 
1284 	/* prevent remaining shrinker jobs */
1285 	mutex_lock(&sbi->umount_mutex);
1286 
1287 	/*
1288 	 * We don't need to do checkpoint when superblock is clean.
1289 	 * But, the previous checkpoint was not done by umount, it needs to do
1290 	 * clean checkpoint again.
1291 	 */
1292 	if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1293 			!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1294 		struct cp_control cpc = {
1295 			.reason = CP_UMOUNT,
1296 		};
1297 		f2fs_write_checkpoint(sbi, &cpc);
1298 	}
1299 
1300 	/* be sure to wait for any on-going discard commands */
1301 	dropped = f2fs_issue_discard_timeout(sbi);
1302 
1303 	if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1304 					!sbi->discard_blks && !dropped) {
1305 		struct cp_control cpc = {
1306 			.reason = CP_UMOUNT | CP_TRIMMED,
1307 		};
1308 		f2fs_write_checkpoint(sbi, &cpc);
1309 	}
1310 
1311 	/*
1312 	 * normally superblock is clean, so we need to release this.
1313 	 * In addition, EIO will skip do checkpoint, we need this as well.
1314 	 */
1315 	f2fs_release_ino_entry(sbi, true);
1316 
1317 	f2fs_leave_shrinker(sbi);
1318 	mutex_unlock(&sbi->umount_mutex);
1319 
1320 	/* our cp_error case, we can wait for any writeback page */
1321 	f2fs_flush_merged_writes(sbi);
1322 
1323 	f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1324 
1325 	f2fs_bug_on(sbi, sbi->fsync_node_num);
1326 
1327 	iput(sbi->node_inode);
1328 	sbi->node_inode = NULL;
1329 
1330 	iput(sbi->meta_inode);
1331 	sbi->meta_inode = NULL;
1332 
1333 	/*
1334 	 * iput() can update stat information, if f2fs_write_checkpoint()
1335 	 * above failed with error.
1336 	 */
1337 	f2fs_destroy_stats(sbi);
1338 
1339 	/* destroy f2fs internal modules */
1340 	f2fs_destroy_node_manager(sbi);
1341 	f2fs_destroy_segment_manager(sbi);
1342 
1343 	f2fs_destroy_post_read_wq(sbi);
1344 
1345 	kvfree(sbi->ckpt);
1346 
1347 	sb->s_fs_info = NULL;
1348 	if (sbi->s_chksum_driver)
1349 		crypto_free_shash(sbi->s_chksum_driver);
1350 	kfree(sbi->raw_super);
1351 
1352 	destroy_device_list(sbi);
1353 	f2fs_destroy_page_array_cache(sbi);
1354 	f2fs_destroy_xattr_caches(sbi);
1355 	mempool_destroy(sbi->write_io_dummy);
1356 #ifdef CONFIG_QUOTA
1357 	for (i = 0; i < MAXQUOTAS; i++)
1358 		kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1359 #endif
1360 	fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
1361 	destroy_percpu_info(sbi);
1362 	for (i = 0; i < NR_PAGE_TYPE; i++)
1363 		kvfree(sbi->write_io[i]);
1364 #ifdef CONFIG_UNICODE
1365 	utf8_unload(sb->s_encoding);
1366 #endif
1367 	kfree(sbi);
1368 }
1369 
f2fs_sync_fs(struct super_block * sb,int sync)1370 int f2fs_sync_fs(struct super_block *sb, int sync)
1371 {
1372 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1373 	int err = 0;
1374 
1375 	if (unlikely(f2fs_cp_error(sbi)))
1376 		return 0;
1377 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1378 		return 0;
1379 
1380 	trace_f2fs_sync_fs(sb, sync);
1381 
1382 	if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1383 		return -EAGAIN;
1384 
1385 	if (sync) {
1386 		struct cp_control cpc;
1387 
1388 		cpc.reason = __get_cp_reason(sbi);
1389 
1390 		down_write(&sbi->gc_lock);
1391 		err = f2fs_write_checkpoint(sbi, &cpc);
1392 		up_write(&sbi->gc_lock);
1393 	}
1394 	f2fs_trace_ios(NULL, 1);
1395 
1396 	return err;
1397 }
1398 
f2fs_freeze(struct super_block * sb)1399 static int f2fs_freeze(struct super_block *sb)
1400 {
1401 	if (f2fs_readonly(sb))
1402 		return 0;
1403 
1404 	/* IO error happened before */
1405 	if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1406 		return -EIO;
1407 
1408 	/* must be clean, since sync_filesystem() was already called */
1409 	if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1410 		return -EINVAL;
1411 	return 0;
1412 }
1413 
f2fs_unfreeze(struct super_block * sb)1414 static int f2fs_unfreeze(struct super_block *sb)
1415 {
1416 	return 0;
1417 }
1418 
1419 #ifdef CONFIG_QUOTA
f2fs_statfs_project(struct super_block * sb,kprojid_t projid,struct kstatfs * buf)1420 static int f2fs_statfs_project(struct super_block *sb,
1421 				kprojid_t projid, struct kstatfs *buf)
1422 {
1423 	struct kqid qid;
1424 	struct dquot *dquot;
1425 	u64 limit;
1426 	u64 curblock;
1427 
1428 	qid = make_kqid_projid(projid);
1429 	dquot = dqget(sb, qid);
1430 	if (IS_ERR(dquot))
1431 		return PTR_ERR(dquot);
1432 	spin_lock(&dquot->dq_dqb_lock);
1433 
1434 	limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1435 					dquot->dq_dqb.dqb_bhardlimit);
1436 	if (limit)
1437 		limit >>= sb->s_blocksize_bits;
1438 
1439 	if (limit && buf->f_blocks > limit) {
1440 		curblock = (dquot->dq_dqb.dqb_curspace +
1441 			    dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1442 		buf->f_blocks = limit;
1443 		buf->f_bfree = buf->f_bavail =
1444 			(buf->f_blocks > curblock) ?
1445 			 (buf->f_blocks - curblock) : 0;
1446 	}
1447 
1448 	limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1449 					dquot->dq_dqb.dqb_ihardlimit);
1450 
1451 	if (limit && buf->f_files > limit) {
1452 		buf->f_files = limit;
1453 		buf->f_ffree =
1454 			(buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1455 			 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1456 	}
1457 
1458 	spin_unlock(&dquot->dq_dqb_lock);
1459 	dqput(dquot);
1460 	return 0;
1461 }
1462 #endif
1463 
f2fs_statfs(struct dentry * dentry,struct kstatfs * buf)1464 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1465 {
1466 	struct super_block *sb = dentry->d_sb;
1467 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1468 	u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1469 	block_t total_count, user_block_count, start_count;
1470 	u64 avail_node_count;
1471 
1472 	total_count = le64_to_cpu(sbi->raw_super->block_count);
1473 	user_block_count = sbi->user_block_count;
1474 	start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1475 	buf->f_type = F2FS_SUPER_MAGIC;
1476 	buf->f_bsize = sbi->blocksize;
1477 
1478 	buf->f_blocks = total_count - start_count;
1479 	buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1480 						sbi->current_reserved_blocks;
1481 
1482 	spin_lock(&sbi->stat_lock);
1483 	if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1484 		buf->f_bfree = 0;
1485 	else
1486 		buf->f_bfree -= sbi->unusable_block_count;
1487 	spin_unlock(&sbi->stat_lock);
1488 
1489 	if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1490 		buf->f_bavail = buf->f_bfree -
1491 				F2FS_OPTION(sbi).root_reserved_blocks;
1492 	else
1493 		buf->f_bavail = 0;
1494 
1495 	avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1496 
1497 	if (avail_node_count > user_block_count) {
1498 		buf->f_files = user_block_count;
1499 		buf->f_ffree = buf->f_bavail;
1500 	} else {
1501 		buf->f_files = avail_node_count;
1502 		buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1503 					buf->f_bavail);
1504 	}
1505 
1506 	buf->f_namelen = F2FS_NAME_LEN;
1507 	buf->f_fsid    = u64_to_fsid(id);
1508 
1509 #ifdef CONFIG_QUOTA
1510 	if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1511 			sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1512 		f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1513 	}
1514 #endif
1515 	return 0;
1516 }
1517 
f2fs_show_quota_options(struct seq_file * seq,struct super_block * sb)1518 static inline void f2fs_show_quota_options(struct seq_file *seq,
1519 					   struct super_block *sb)
1520 {
1521 #ifdef CONFIG_QUOTA
1522 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1523 
1524 	if (F2FS_OPTION(sbi).s_jquota_fmt) {
1525 		char *fmtname = "";
1526 
1527 		switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1528 		case QFMT_VFS_OLD:
1529 			fmtname = "vfsold";
1530 			break;
1531 		case QFMT_VFS_V0:
1532 			fmtname = "vfsv0";
1533 			break;
1534 		case QFMT_VFS_V1:
1535 			fmtname = "vfsv1";
1536 			break;
1537 		}
1538 		seq_printf(seq, ",jqfmt=%s", fmtname);
1539 	}
1540 
1541 	if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1542 		seq_show_option(seq, "usrjquota",
1543 			F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1544 
1545 	if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1546 		seq_show_option(seq, "grpjquota",
1547 			F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1548 
1549 	if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1550 		seq_show_option(seq, "prjjquota",
1551 			F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1552 #endif
1553 }
1554 
f2fs_show_compress_options(struct seq_file * seq,struct super_block * sb)1555 static inline void f2fs_show_compress_options(struct seq_file *seq,
1556 							struct super_block *sb)
1557 {
1558 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1559 	char *algtype = "";
1560 	int i;
1561 
1562 	if (!f2fs_sb_has_compression(sbi))
1563 		return;
1564 
1565 	switch (F2FS_OPTION(sbi).compress_algorithm) {
1566 	case COMPRESS_LZO:
1567 		algtype = "lzo";
1568 		break;
1569 	case COMPRESS_LZ4:
1570 		algtype = "lz4";
1571 		break;
1572 	case COMPRESS_ZSTD:
1573 		algtype = "zstd";
1574 		break;
1575 	case COMPRESS_LZORLE:
1576 		algtype = "lzo-rle";
1577 		break;
1578 	}
1579 	seq_printf(seq, ",compress_algorithm=%s", algtype);
1580 
1581 	seq_printf(seq, ",compress_log_size=%u",
1582 			F2FS_OPTION(sbi).compress_log_size);
1583 
1584 	for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1585 		seq_printf(seq, ",compress_extension=%s",
1586 			F2FS_OPTION(sbi).extensions[i]);
1587 	}
1588 }
1589 
f2fs_show_options(struct seq_file * seq,struct dentry * root)1590 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1591 {
1592 	struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1593 
1594 	if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1595 		seq_printf(seq, ",background_gc=%s", "sync");
1596 	else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1597 		seq_printf(seq, ",background_gc=%s", "on");
1598 	else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1599 		seq_printf(seq, ",background_gc=%s", "off");
1600 
1601 	if (test_opt(sbi, GC_MERGE))
1602 		seq_puts(seq, ",gc_merge");
1603 
1604 	if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1605 		seq_puts(seq, ",disable_roll_forward");
1606 	if (test_opt(sbi, NORECOVERY))
1607 		seq_puts(seq, ",norecovery");
1608 	if (test_opt(sbi, DISCARD))
1609 		seq_puts(seq, ",discard");
1610 	else
1611 		seq_puts(seq, ",nodiscard");
1612 	if (test_opt(sbi, NOHEAP))
1613 		seq_puts(seq, ",no_heap");
1614 	else
1615 		seq_puts(seq, ",heap");
1616 #ifdef CONFIG_F2FS_FS_XATTR
1617 	if (test_opt(sbi, XATTR_USER))
1618 		seq_puts(seq, ",user_xattr");
1619 	else
1620 		seq_puts(seq, ",nouser_xattr");
1621 	if (test_opt(sbi, INLINE_XATTR))
1622 		seq_puts(seq, ",inline_xattr");
1623 	else
1624 		seq_puts(seq, ",noinline_xattr");
1625 	if (test_opt(sbi, INLINE_XATTR_SIZE))
1626 		seq_printf(seq, ",inline_xattr_size=%u",
1627 					F2FS_OPTION(sbi).inline_xattr_size);
1628 #endif
1629 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1630 	if (test_opt(sbi, POSIX_ACL))
1631 		seq_puts(seq, ",acl");
1632 	else
1633 		seq_puts(seq, ",noacl");
1634 #endif
1635 	if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1636 		seq_puts(seq, ",disable_ext_identify");
1637 	if (test_opt(sbi, INLINE_DATA))
1638 		seq_puts(seq, ",inline_data");
1639 	else
1640 		seq_puts(seq, ",noinline_data");
1641 	if (test_opt(sbi, INLINE_DENTRY))
1642 		seq_puts(seq, ",inline_dentry");
1643 	else
1644 		seq_puts(seq, ",noinline_dentry");
1645 	if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1646 		seq_puts(seq, ",flush_merge");
1647 	if (test_opt(sbi, NOBARRIER))
1648 		seq_puts(seq, ",nobarrier");
1649 	if (test_opt(sbi, FASTBOOT))
1650 		seq_puts(seq, ",fastboot");
1651 	if (test_opt(sbi, EXTENT_CACHE))
1652 		seq_puts(seq, ",extent_cache");
1653 	else
1654 		seq_puts(seq, ",noextent_cache");
1655 	if (test_opt(sbi, DATA_FLUSH))
1656 		seq_puts(seq, ",data_flush");
1657 
1658 	seq_puts(seq, ",mode=");
1659 	if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
1660 		seq_puts(seq, "adaptive");
1661 	else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
1662 		seq_puts(seq, "lfs");
1663 	seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1664 	if (test_opt(sbi, RESERVE_ROOT))
1665 		seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1666 				F2FS_OPTION(sbi).root_reserved_blocks,
1667 				from_kuid_munged(&init_user_ns,
1668 					F2FS_OPTION(sbi).s_resuid),
1669 				from_kgid_munged(&init_user_ns,
1670 					F2FS_OPTION(sbi).s_resgid));
1671 	if (F2FS_IO_SIZE_BITS(sbi))
1672 		seq_printf(seq, ",io_bits=%u",
1673 				F2FS_OPTION(sbi).write_io_size_bits);
1674 #ifdef CONFIG_F2FS_FAULT_INJECTION
1675 	if (test_opt(sbi, FAULT_INJECTION)) {
1676 		seq_printf(seq, ",fault_injection=%u",
1677 				F2FS_OPTION(sbi).fault_info.inject_rate);
1678 		seq_printf(seq, ",fault_type=%u",
1679 				F2FS_OPTION(sbi).fault_info.inject_type);
1680 	}
1681 #endif
1682 #ifdef CONFIG_QUOTA
1683 	if (test_opt(sbi, QUOTA))
1684 		seq_puts(seq, ",quota");
1685 	if (test_opt(sbi, USRQUOTA))
1686 		seq_puts(seq, ",usrquota");
1687 	if (test_opt(sbi, GRPQUOTA))
1688 		seq_puts(seq, ",grpquota");
1689 	if (test_opt(sbi, PRJQUOTA))
1690 		seq_puts(seq, ",prjquota");
1691 #endif
1692 	f2fs_show_quota_options(seq, sbi->sb);
1693 	if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1694 		seq_printf(seq, ",whint_mode=%s", "user-based");
1695 	else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1696 		seq_printf(seq, ",whint_mode=%s", "fs-based");
1697 
1698 	fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
1699 
1700 	if (sbi->sb->s_flags & SB_INLINECRYPT)
1701 		seq_puts(seq, ",inlinecrypt");
1702 
1703 	if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1704 		seq_printf(seq, ",alloc_mode=%s", "default");
1705 	else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1706 		seq_printf(seq, ",alloc_mode=%s", "reuse");
1707 
1708 	if (test_opt(sbi, DISABLE_CHECKPOINT))
1709 		seq_printf(seq, ",checkpoint=disable:%u",
1710 				F2FS_OPTION(sbi).unusable_cap);
1711 	if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1712 		seq_printf(seq, ",fsync_mode=%s", "posix");
1713 	else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1714 		seq_printf(seq, ",fsync_mode=%s", "strict");
1715 	else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1716 		seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1717 
1718 #ifdef CONFIG_F2FS_FS_COMPRESSION
1719 	f2fs_show_compress_options(seq, sbi->sb);
1720 #endif
1721 
1722 	if (test_opt(sbi, ATGC))
1723 		seq_puts(seq, ",atgc");
1724 	return 0;
1725 }
1726 
default_options(struct f2fs_sb_info * sbi)1727 static void default_options(struct f2fs_sb_info *sbi)
1728 {
1729 	/* init some FS parameters */
1730 	F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
1731 	F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1732 	F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1733 	F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1734 	F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1735 	F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1736 	F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1737 	F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
1738 	F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
1739 	F2FS_OPTION(sbi).compress_ext_cnt = 0;
1740 	F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
1741 
1742 	sbi->sb->s_flags &= ~SB_INLINECRYPT;
1743 
1744 	set_opt(sbi, INLINE_XATTR);
1745 	set_opt(sbi, INLINE_DATA);
1746 	set_opt(sbi, INLINE_DENTRY);
1747 	set_opt(sbi, EXTENT_CACHE);
1748 	set_opt(sbi, NOHEAP);
1749 	clear_opt(sbi, DISABLE_CHECKPOINT);
1750 	F2FS_OPTION(sbi).unusable_cap = 0;
1751 	sbi->sb->s_flags |= SB_LAZYTIME;
1752 	set_opt(sbi, FLUSH_MERGE);
1753 	set_opt(sbi, DISCARD);
1754 	if (f2fs_sb_has_blkzoned(sbi))
1755 		F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
1756 	else
1757 		F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
1758 
1759 #ifdef CONFIG_F2FS_FS_XATTR
1760 	set_opt(sbi, XATTR_USER);
1761 #endif
1762 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1763 	set_opt(sbi, POSIX_ACL);
1764 #endif
1765 
1766 	f2fs_build_fault_attr(sbi, 0, 0);
1767 }
1768 
1769 #ifdef CONFIG_QUOTA
1770 static int f2fs_enable_quotas(struct super_block *sb);
1771 #endif
1772 
f2fs_disable_checkpoint(struct f2fs_sb_info * sbi)1773 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1774 {
1775 	unsigned int s_flags = sbi->sb->s_flags;
1776 	struct cp_control cpc;
1777 	int err = 0;
1778 	int ret;
1779 	block_t unusable;
1780 
1781 	if (s_flags & SB_RDONLY) {
1782 		f2fs_err(sbi, "checkpoint=disable on readonly fs");
1783 		return -EINVAL;
1784 	}
1785 	sbi->sb->s_flags |= SB_ACTIVE;
1786 
1787 	f2fs_update_time(sbi, DISABLE_TIME);
1788 
1789 	while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1790 		down_write(&sbi->gc_lock);
1791 		err = f2fs_gc(sbi, true, false, false, NULL_SEGNO);
1792 		if (err == -ENODATA) {
1793 			err = 0;
1794 			break;
1795 		}
1796 		if (err && err != -EAGAIN)
1797 			break;
1798 	}
1799 
1800 	ret = sync_filesystem(sbi->sb);
1801 	if (ret || err) {
1802 		err = ret ? ret: err;
1803 		goto restore_flag;
1804 	}
1805 
1806 	unusable = f2fs_get_unusable_blocks(sbi);
1807 	if (f2fs_disable_cp_again(sbi, unusable)) {
1808 		err = -EAGAIN;
1809 		goto restore_flag;
1810 	}
1811 
1812 	down_write(&sbi->gc_lock);
1813 	cpc.reason = CP_PAUSE;
1814 	set_sbi_flag(sbi, SBI_CP_DISABLED);
1815 	err = f2fs_write_checkpoint(sbi, &cpc);
1816 	if (err)
1817 		goto out_unlock;
1818 
1819 	spin_lock(&sbi->stat_lock);
1820 	sbi->unusable_block_count = unusable;
1821 	spin_unlock(&sbi->stat_lock);
1822 
1823 out_unlock:
1824 	up_write(&sbi->gc_lock);
1825 restore_flag:
1826 	sbi->sb->s_flags = s_flags;	/* Restore SB_RDONLY status */
1827 	return err;
1828 }
1829 
f2fs_enable_checkpoint(struct f2fs_sb_info * sbi)1830 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1831 {
1832 	int retry = DEFAULT_RETRY_IO_COUNT;
1833 
1834 	/* we should flush all the data to keep data consistency */
1835 	do {
1836 		sync_inodes_sb(sbi->sb);
1837 		cond_resched();
1838 		congestion_wait(BLK_RW_ASYNC, DEFAULT_IO_TIMEOUT);
1839 	} while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--);
1840 
1841 	if (unlikely(retry < 0))
1842 		f2fs_warn(sbi, "checkpoint=enable has some unwritten data.");
1843 
1844 	down_write(&sbi->gc_lock);
1845 	f2fs_dirty_to_prefree(sbi);
1846 
1847 	clear_sbi_flag(sbi, SBI_CP_DISABLED);
1848 	set_sbi_flag(sbi, SBI_IS_DIRTY);
1849 	up_write(&sbi->gc_lock);
1850 
1851 	f2fs_sync_fs(sbi->sb, 1);
1852 }
1853 
f2fs_remount(struct super_block * sb,int * flags,char * data)1854 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1855 {
1856 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1857 	struct f2fs_mount_info org_mount_opt;
1858 	unsigned long old_sb_flags;
1859 	int err;
1860 	bool need_restart_gc = false;
1861 	bool need_stop_gc = false;
1862 	bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1863 	bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1864 	bool no_io_align = !F2FS_IO_ALIGNED(sbi);
1865 	bool no_atgc = !test_opt(sbi, ATGC);
1866 	bool checkpoint_changed;
1867 #ifdef CONFIG_QUOTA
1868 	int i, j;
1869 #endif
1870 
1871 	/*
1872 	 * Save the old mount options in case we
1873 	 * need to restore them.
1874 	 */
1875 	org_mount_opt = sbi->mount_opt;
1876 	old_sb_flags = sb->s_flags;
1877 
1878 #ifdef CONFIG_QUOTA
1879 	org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1880 	for (i = 0; i < MAXQUOTAS; i++) {
1881 		if (F2FS_OPTION(sbi).s_qf_names[i]) {
1882 			org_mount_opt.s_qf_names[i] =
1883 				kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1884 				GFP_KERNEL);
1885 			if (!org_mount_opt.s_qf_names[i]) {
1886 				for (j = 0; j < i; j++)
1887 					kfree(org_mount_opt.s_qf_names[j]);
1888 				return -ENOMEM;
1889 			}
1890 		} else {
1891 			org_mount_opt.s_qf_names[i] = NULL;
1892 		}
1893 	}
1894 #endif
1895 
1896 	/* recover superblocks we couldn't write due to previous RO mount */
1897 	if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1898 		err = f2fs_commit_super(sbi, false);
1899 		f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
1900 			  err);
1901 		if (!err)
1902 			clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1903 	}
1904 
1905 	default_options(sbi);
1906 
1907 	/* parse mount options */
1908 	err = parse_options(sb, data, true);
1909 	if (err)
1910 		goto restore_opts;
1911 	checkpoint_changed =
1912 			disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1913 
1914 	/*
1915 	 * Previous and new state of filesystem is RO,
1916 	 * so skip checking GC and FLUSH_MERGE conditions.
1917 	 */
1918 	if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1919 		goto skip;
1920 
1921 #ifdef CONFIG_QUOTA
1922 	if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1923 		err = dquot_suspend(sb, -1);
1924 		if (err < 0)
1925 			goto restore_opts;
1926 	} else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1927 		/* dquot_resume needs RW */
1928 		sb->s_flags &= ~SB_RDONLY;
1929 		if (sb_any_quota_suspended(sb)) {
1930 			dquot_resume(sb, -1);
1931 		} else if (f2fs_sb_has_quota_ino(sbi)) {
1932 			err = f2fs_enable_quotas(sb);
1933 			if (err)
1934 				goto restore_opts;
1935 		}
1936 	}
1937 #endif
1938 	/* disallow enable atgc dynamically */
1939 	if (no_atgc == !!test_opt(sbi, ATGC)) {
1940 		err = -EINVAL;
1941 		f2fs_warn(sbi, "switch atgc option is not allowed");
1942 		goto restore_opts;
1943 	}
1944 
1945 	/* disallow enable/disable extent_cache dynamically */
1946 	if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1947 		err = -EINVAL;
1948 		f2fs_warn(sbi, "switch extent_cache option is not allowed");
1949 		goto restore_opts;
1950 	}
1951 
1952 	if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
1953 		err = -EINVAL;
1954 		f2fs_warn(sbi, "switch io_bits option is not allowed");
1955 		goto restore_opts;
1956 	}
1957 
1958 	if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1959 		err = -EINVAL;
1960 		f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
1961 		goto restore_opts;
1962 	}
1963 
1964 	/*
1965 	 * We stop the GC thread if FS is mounted as RO
1966 	 * or if background_gc = off is passed in mount
1967 	 * option. Also sync the filesystem.
1968 	 */
1969 	if ((*flags & SB_RDONLY) ||
1970 			(F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF &&
1971 			!test_opt(sbi, GC_MERGE))) {
1972 		if (sbi->gc_thread) {
1973 			f2fs_stop_gc_thread(sbi);
1974 			need_restart_gc = true;
1975 		}
1976 	} else if (!sbi->gc_thread) {
1977 		err = f2fs_start_gc_thread(sbi);
1978 		if (err)
1979 			goto restore_opts;
1980 		need_stop_gc = true;
1981 	}
1982 
1983 	if (*flags & SB_RDONLY ||
1984 		F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1985 		writeback_inodes_sb(sb, WB_REASON_SYNC);
1986 		sync_inodes_sb(sb);
1987 
1988 		set_sbi_flag(sbi, SBI_IS_DIRTY);
1989 		set_sbi_flag(sbi, SBI_IS_CLOSE);
1990 		f2fs_sync_fs(sb, 1);
1991 		clear_sbi_flag(sbi, SBI_IS_CLOSE);
1992 	}
1993 
1994 	if (checkpoint_changed) {
1995 		if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1996 			err = f2fs_disable_checkpoint(sbi);
1997 			if (err)
1998 				goto restore_gc;
1999 		} else {
2000 			f2fs_enable_checkpoint(sbi);
2001 		}
2002 	}
2003 
2004 	/*
2005 	 * We stop issue flush thread if FS is mounted as RO
2006 	 * or if flush_merge is not passed in mount option.
2007 	 */
2008 	if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2009 		clear_opt(sbi, FLUSH_MERGE);
2010 		f2fs_destroy_flush_cmd_control(sbi, false);
2011 	} else {
2012 		err = f2fs_create_flush_cmd_control(sbi);
2013 		if (err)
2014 			goto restore_gc;
2015 	}
2016 skip:
2017 #ifdef CONFIG_QUOTA
2018 	/* Release old quota file names */
2019 	for (i = 0; i < MAXQUOTAS; i++)
2020 		kfree(org_mount_opt.s_qf_names[i]);
2021 #endif
2022 	/* Update the POSIXACL Flag */
2023 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2024 		(test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2025 
2026 	limit_reserve_root(sbi);
2027 	adjust_unusable_cap_perc(sbi);
2028 	*flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
2029 	return 0;
2030 restore_gc:
2031 	if (need_restart_gc) {
2032 		if (f2fs_start_gc_thread(sbi))
2033 			f2fs_warn(sbi, "background gc thread has stopped");
2034 	} else if (need_stop_gc) {
2035 		f2fs_stop_gc_thread(sbi);
2036 	}
2037 restore_opts:
2038 #ifdef CONFIG_QUOTA
2039 	F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
2040 	for (i = 0; i < MAXQUOTAS; i++) {
2041 		kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2042 		F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
2043 	}
2044 #endif
2045 	sbi->mount_opt = org_mount_opt;
2046 	sb->s_flags = old_sb_flags;
2047 	return err;
2048 }
2049 
2050 #ifdef CONFIG_QUOTA
2051 /* Read data from quotafile */
f2fs_quota_read(struct super_block * sb,int type,char * data,size_t len,loff_t off)2052 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
2053 			       size_t len, loff_t off)
2054 {
2055 	struct inode *inode = sb_dqopt(sb)->files[type];
2056 	struct address_space *mapping = inode->i_mapping;
2057 	block_t blkidx = F2FS_BYTES_TO_BLK(off);
2058 	int offset = off & (sb->s_blocksize - 1);
2059 	int tocopy;
2060 	size_t toread;
2061 	loff_t i_size = i_size_read(inode);
2062 	struct page *page;
2063 
2064 	if (off > i_size)
2065 		return 0;
2066 
2067 	if (off + len > i_size)
2068 		len = i_size - off;
2069 	toread = len;
2070 	while (toread > 0) {
2071 		tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
2072 repeat:
2073 		page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
2074 		if (IS_ERR(page)) {
2075 			if (PTR_ERR(page) == -ENOMEM) {
2076 				congestion_wait(BLK_RW_ASYNC,
2077 						DEFAULT_IO_TIMEOUT);
2078 				goto repeat;
2079 			}
2080 			set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2081 			return PTR_ERR(page);
2082 		}
2083 
2084 		lock_page(page);
2085 
2086 		if (unlikely(page->mapping != mapping)) {
2087 			f2fs_put_page(page, 1);
2088 			goto repeat;
2089 		}
2090 		if (unlikely(!PageUptodate(page))) {
2091 			f2fs_put_page(page, 1);
2092 			set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2093 			return -EIO;
2094 		}
2095 
2096 		memcpy_from_page(data, page, offset, tocopy);
2097 		f2fs_put_page(page, 1);
2098 
2099 		offset = 0;
2100 		toread -= tocopy;
2101 		data += tocopy;
2102 		blkidx++;
2103 	}
2104 	return len;
2105 }
2106 
2107 /* Write to quotafile */
f2fs_quota_write(struct super_block * sb,int type,const char * data,size_t len,loff_t off)2108 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2109 				const char *data, size_t len, loff_t off)
2110 {
2111 	struct inode *inode = sb_dqopt(sb)->files[type];
2112 	struct address_space *mapping = inode->i_mapping;
2113 	const struct address_space_operations *a_ops = mapping->a_ops;
2114 	int offset = off & (sb->s_blocksize - 1);
2115 	size_t towrite = len;
2116 	struct page *page;
2117 	void *fsdata = NULL;
2118 	int err = 0;
2119 	int tocopy;
2120 
2121 	while (towrite > 0) {
2122 		tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2123 								towrite);
2124 retry:
2125 		err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
2126 							&page, &fsdata);
2127 		if (unlikely(err)) {
2128 			if (err == -ENOMEM) {
2129 				congestion_wait(BLK_RW_ASYNC,
2130 						DEFAULT_IO_TIMEOUT);
2131 				goto retry;
2132 			}
2133 			set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2134 			break;
2135 		}
2136 
2137 		memcpy_to_page(page, offset, data, tocopy);
2138 
2139 		a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2140 						page, fsdata);
2141 		offset = 0;
2142 		towrite -= tocopy;
2143 		off += tocopy;
2144 		data += tocopy;
2145 		cond_resched();
2146 	}
2147 
2148 	if (len == towrite)
2149 		return err;
2150 	inode->i_mtime = inode->i_ctime = current_time(inode);
2151 	f2fs_mark_inode_dirty_sync(inode, false);
2152 	return len - towrite;
2153 }
2154 
f2fs_get_dquots(struct inode * inode)2155 static struct dquot **f2fs_get_dquots(struct inode *inode)
2156 {
2157 	return F2FS_I(inode)->i_dquot;
2158 }
2159 
f2fs_get_reserved_space(struct inode * inode)2160 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2161 {
2162 	return &F2FS_I(inode)->i_reserved_quota;
2163 }
2164 
f2fs_quota_on_mount(struct f2fs_sb_info * sbi,int type)2165 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2166 {
2167 	if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2168 		f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2169 		return 0;
2170 	}
2171 
2172 	return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2173 					F2FS_OPTION(sbi).s_jquota_fmt, type);
2174 }
2175 
f2fs_enable_quota_files(struct f2fs_sb_info * sbi,bool rdonly)2176 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2177 {
2178 	int enabled = 0;
2179 	int i, err;
2180 
2181 	if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2182 		err = f2fs_enable_quotas(sbi->sb);
2183 		if (err) {
2184 			f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2185 			return 0;
2186 		}
2187 		return 1;
2188 	}
2189 
2190 	for (i = 0; i < MAXQUOTAS; i++) {
2191 		if (F2FS_OPTION(sbi).s_qf_names[i]) {
2192 			err = f2fs_quota_on_mount(sbi, i);
2193 			if (!err) {
2194 				enabled = 1;
2195 				continue;
2196 			}
2197 			f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2198 				 err, i);
2199 		}
2200 	}
2201 	return enabled;
2202 }
2203 
f2fs_quota_enable(struct super_block * sb,int type,int format_id,unsigned int flags)2204 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2205 			     unsigned int flags)
2206 {
2207 	struct inode *qf_inode;
2208 	unsigned long qf_inum;
2209 	int err;
2210 
2211 	BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2212 
2213 	qf_inum = f2fs_qf_ino(sb, type);
2214 	if (!qf_inum)
2215 		return -EPERM;
2216 
2217 	qf_inode = f2fs_iget(sb, qf_inum);
2218 	if (IS_ERR(qf_inode)) {
2219 		f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2220 		return PTR_ERR(qf_inode);
2221 	}
2222 
2223 	/* Don't account quota for quota files to avoid recursion */
2224 	qf_inode->i_flags |= S_NOQUOTA;
2225 	err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2226 	iput(qf_inode);
2227 	return err;
2228 }
2229 
f2fs_enable_quotas(struct super_block * sb)2230 static int f2fs_enable_quotas(struct super_block *sb)
2231 {
2232 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2233 	int type, err = 0;
2234 	unsigned long qf_inum;
2235 	bool quota_mopt[MAXQUOTAS] = {
2236 		test_opt(sbi, USRQUOTA),
2237 		test_opt(sbi, GRPQUOTA),
2238 		test_opt(sbi, PRJQUOTA),
2239 	};
2240 
2241 	if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2242 		f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2243 		return 0;
2244 	}
2245 
2246 	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2247 
2248 	for (type = 0; type < MAXQUOTAS; type++) {
2249 		qf_inum = f2fs_qf_ino(sb, type);
2250 		if (qf_inum) {
2251 			err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2252 				DQUOT_USAGE_ENABLED |
2253 				(quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2254 			if (err) {
2255 				f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2256 					 type, err);
2257 				for (type--; type >= 0; type--)
2258 					dquot_quota_off(sb, type);
2259 				set_sbi_flag(F2FS_SB(sb),
2260 						SBI_QUOTA_NEED_REPAIR);
2261 				return err;
2262 			}
2263 		}
2264 	}
2265 	return 0;
2266 }
2267 
f2fs_quota_sync_file(struct f2fs_sb_info * sbi,int type)2268 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type)
2269 {
2270 	struct quota_info *dqopt = sb_dqopt(sbi->sb);
2271 	struct address_space *mapping = dqopt->files[type]->i_mapping;
2272 	int ret = 0;
2273 
2274 	ret = dquot_writeback_dquots(sbi->sb, type);
2275 	if (ret)
2276 		goto out;
2277 
2278 	ret = filemap_fdatawrite(mapping);
2279 	if (ret)
2280 		goto out;
2281 
2282 	/* if we are using journalled quota */
2283 	if (is_journalled_quota(sbi))
2284 		goto out;
2285 
2286 	ret = filemap_fdatawait(mapping);
2287 
2288 	truncate_inode_pages(&dqopt->files[type]->i_data, 0);
2289 out:
2290 	if (ret)
2291 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2292 	return ret;
2293 }
2294 
f2fs_quota_sync(struct super_block * sb,int type)2295 int f2fs_quota_sync(struct super_block *sb, int type)
2296 {
2297 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2298 	struct quota_info *dqopt = sb_dqopt(sb);
2299 	int cnt;
2300 	int ret = 0;
2301 
2302 	/*
2303 	 * Now when everything is written we can discard the pagecache so
2304 	 * that userspace sees the changes.
2305 	 */
2306 	for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2307 
2308 		if (type != -1 && cnt != type)
2309 			continue;
2310 
2311 		if (!sb_has_quota_active(sb, cnt))
2312 			continue;
2313 
2314 		if (!f2fs_sb_has_quota_ino(sbi))
2315 			inode_lock(dqopt->files[cnt]);
2316 
2317 		/*
2318 		 * do_quotactl
2319 		 *  f2fs_quota_sync
2320 		 *  down_read(quota_sem)
2321 		 *  dquot_writeback_dquots()
2322 		 *  f2fs_dquot_commit
2323 		 *			      block_operation
2324 		 *			      down_read(quota_sem)
2325 		 */
2326 		f2fs_lock_op(sbi);
2327 		down_read(&sbi->quota_sem);
2328 
2329 		ret = f2fs_quota_sync_file(sbi, cnt);
2330 
2331 		up_read(&sbi->quota_sem);
2332 		f2fs_unlock_op(sbi);
2333 
2334 		if (!f2fs_sb_has_quota_ino(sbi))
2335 			inode_unlock(dqopt->files[cnt]);
2336 
2337 		if (ret)
2338 			break;
2339 	}
2340 	return ret;
2341 }
2342 
f2fs_quota_on(struct super_block * sb,int type,int format_id,const struct path * path)2343 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2344 							const struct path *path)
2345 {
2346 	struct inode *inode;
2347 	int err;
2348 
2349 	/* if quota sysfile exists, deny enabling quota with specific file */
2350 	if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2351 		f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2352 		return -EBUSY;
2353 	}
2354 
2355 	err = f2fs_quota_sync(sb, type);
2356 	if (err)
2357 		return err;
2358 
2359 	err = dquot_quota_on(sb, type, format_id, path);
2360 	if (err)
2361 		return err;
2362 
2363 	inode = d_inode(path->dentry);
2364 
2365 	inode_lock(inode);
2366 	F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2367 	f2fs_set_inode_flags(inode);
2368 	inode_unlock(inode);
2369 	f2fs_mark_inode_dirty_sync(inode, false);
2370 
2371 	return 0;
2372 }
2373 
__f2fs_quota_off(struct super_block * sb,int type)2374 static int __f2fs_quota_off(struct super_block *sb, int type)
2375 {
2376 	struct inode *inode = sb_dqopt(sb)->files[type];
2377 	int err;
2378 
2379 	if (!inode || !igrab(inode))
2380 		return dquot_quota_off(sb, type);
2381 
2382 	err = f2fs_quota_sync(sb, type);
2383 	if (err)
2384 		goto out_put;
2385 
2386 	err = dquot_quota_off(sb, type);
2387 	if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2388 		goto out_put;
2389 
2390 	inode_lock(inode);
2391 	F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2392 	f2fs_set_inode_flags(inode);
2393 	inode_unlock(inode);
2394 	f2fs_mark_inode_dirty_sync(inode, false);
2395 out_put:
2396 	iput(inode);
2397 	return err;
2398 }
2399 
f2fs_quota_off(struct super_block * sb,int type)2400 static int f2fs_quota_off(struct super_block *sb, int type)
2401 {
2402 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2403 	int err;
2404 
2405 	err = __f2fs_quota_off(sb, type);
2406 
2407 	/*
2408 	 * quotactl can shutdown journalled quota, result in inconsistence
2409 	 * between quota record and fs data by following updates, tag the
2410 	 * flag to let fsck be aware of it.
2411 	 */
2412 	if (is_journalled_quota(sbi))
2413 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2414 	return err;
2415 }
2416 
f2fs_quota_off_umount(struct super_block * sb)2417 void f2fs_quota_off_umount(struct super_block *sb)
2418 {
2419 	int type;
2420 	int err;
2421 
2422 	for (type = 0; type < MAXQUOTAS; type++) {
2423 		err = __f2fs_quota_off(sb, type);
2424 		if (err) {
2425 			int ret = dquot_quota_off(sb, type);
2426 
2427 			f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2428 				 type, err, ret);
2429 			set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2430 		}
2431 	}
2432 	/*
2433 	 * In case of checkpoint=disable, we must flush quota blocks.
2434 	 * This can cause NULL exception for node_inode in end_io, since
2435 	 * put_super already dropped it.
2436 	 */
2437 	sync_filesystem(sb);
2438 }
2439 
f2fs_truncate_quota_inode_pages(struct super_block * sb)2440 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2441 {
2442 	struct quota_info *dqopt = sb_dqopt(sb);
2443 	int type;
2444 
2445 	for (type = 0; type < MAXQUOTAS; type++) {
2446 		if (!dqopt->files[type])
2447 			continue;
2448 		f2fs_inode_synced(dqopt->files[type]);
2449 	}
2450 }
2451 
f2fs_dquot_commit(struct dquot * dquot)2452 static int f2fs_dquot_commit(struct dquot *dquot)
2453 {
2454 	struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2455 	int ret;
2456 
2457 	down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
2458 	ret = dquot_commit(dquot);
2459 	if (ret < 0)
2460 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2461 	up_read(&sbi->quota_sem);
2462 	return ret;
2463 }
2464 
f2fs_dquot_acquire(struct dquot * dquot)2465 static int f2fs_dquot_acquire(struct dquot *dquot)
2466 {
2467 	struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2468 	int ret;
2469 
2470 	down_read(&sbi->quota_sem);
2471 	ret = dquot_acquire(dquot);
2472 	if (ret < 0)
2473 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2474 	up_read(&sbi->quota_sem);
2475 	return ret;
2476 }
2477 
f2fs_dquot_release(struct dquot * dquot)2478 static int f2fs_dquot_release(struct dquot *dquot)
2479 {
2480 	struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2481 	int ret = dquot_release(dquot);
2482 
2483 	if (ret < 0)
2484 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2485 	return ret;
2486 }
2487 
f2fs_dquot_mark_dquot_dirty(struct dquot * dquot)2488 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2489 {
2490 	struct super_block *sb = dquot->dq_sb;
2491 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2492 	int ret = dquot_mark_dquot_dirty(dquot);
2493 
2494 	/* if we are using journalled quota */
2495 	if (is_journalled_quota(sbi))
2496 		set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2497 
2498 	return ret;
2499 }
2500 
f2fs_dquot_commit_info(struct super_block * sb,int type)2501 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2502 {
2503 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2504 	int ret = dquot_commit_info(sb, type);
2505 
2506 	if (ret < 0)
2507 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2508 	return ret;
2509 }
2510 
f2fs_get_projid(struct inode * inode,kprojid_t * projid)2511 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2512 {
2513 	*projid = F2FS_I(inode)->i_projid;
2514 	return 0;
2515 }
2516 
2517 static const struct dquot_operations f2fs_quota_operations = {
2518 	.get_reserved_space = f2fs_get_reserved_space,
2519 	.write_dquot	= f2fs_dquot_commit,
2520 	.acquire_dquot	= f2fs_dquot_acquire,
2521 	.release_dquot	= f2fs_dquot_release,
2522 	.mark_dirty	= f2fs_dquot_mark_dquot_dirty,
2523 	.write_info	= f2fs_dquot_commit_info,
2524 	.alloc_dquot	= dquot_alloc,
2525 	.destroy_dquot	= dquot_destroy,
2526 	.get_projid	= f2fs_get_projid,
2527 	.get_next_id	= dquot_get_next_id,
2528 };
2529 
2530 static const struct quotactl_ops f2fs_quotactl_ops = {
2531 	.quota_on	= f2fs_quota_on,
2532 	.quota_off	= f2fs_quota_off,
2533 	.quota_sync	= f2fs_quota_sync,
2534 	.get_state	= dquot_get_state,
2535 	.set_info	= dquot_set_dqinfo,
2536 	.get_dqblk	= dquot_get_dqblk,
2537 	.set_dqblk	= dquot_set_dqblk,
2538 	.get_nextdqblk	= dquot_get_next_dqblk,
2539 };
2540 #else
f2fs_quota_sync(struct super_block * sb,int type)2541 int f2fs_quota_sync(struct super_block *sb, int type)
2542 {
2543 	return 0;
2544 }
2545 
f2fs_quota_off_umount(struct super_block * sb)2546 void f2fs_quota_off_umount(struct super_block *sb)
2547 {
2548 }
2549 #endif
2550 
2551 static const struct super_operations f2fs_sops = {
2552 	.alloc_inode	= f2fs_alloc_inode,
2553 	.free_inode	= f2fs_free_inode,
2554 	.drop_inode	= f2fs_drop_inode,
2555 	.write_inode	= f2fs_write_inode,
2556 	.dirty_inode	= f2fs_dirty_inode,
2557 	.show_options	= f2fs_show_options,
2558 #ifdef CONFIG_QUOTA
2559 	.quota_read	= f2fs_quota_read,
2560 	.quota_write	= f2fs_quota_write,
2561 	.get_dquots	= f2fs_get_dquots,
2562 #endif
2563 	.evict_inode	= f2fs_evict_inode,
2564 	.put_super	= f2fs_put_super,
2565 	.sync_fs	= f2fs_sync_fs,
2566 	.freeze_fs	= f2fs_freeze,
2567 	.unfreeze_fs	= f2fs_unfreeze,
2568 	.statfs		= f2fs_statfs,
2569 	.remount_fs	= f2fs_remount,
2570 };
2571 
2572 #ifdef CONFIG_FS_ENCRYPTION
f2fs_get_context(struct inode * inode,void * ctx,size_t len)2573 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2574 {
2575 	return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2576 				F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2577 				ctx, len, NULL);
2578 }
2579 
f2fs_set_context(struct inode * inode,const void * ctx,size_t len,void * fs_data)2580 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2581 							void *fs_data)
2582 {
2583 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2584 
2585 	/*
2586 	 * Encrypting the root directory is not allowed because fsck
2587 	 * expects lost+found directory to exist and remain unencrypted
2588 	 * if LOST_FOUND feature is enabled.
2589 	 *
2590 	 */
2591 	if (f2fs_sb_has_lost_found(sbi) &&
2592 			inode->i_ino == F2FS_ROOT_INO(sbi))
2593 		return -EPERM;
2594 
2595 	return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2596 				F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2597 				ctx, len, fs_data, XATTR_CREATE);
2598 }
2599 
f2fs_get_dummy_policy(struct super_block * sb)2600 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb)
2601 {
2602 	return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy;
2603 }
2604 
f2fs_has_stable_inodes(struct super_block * sb)2605 static bool f2fs_has_stable_inodes(struct super_block *sb)
2606 {
2607 	return true;
2608 }
2609 
f2fs_get_ino_and_lblk_bits(struct super_block * sb,int * ino_bits_ret,int * lblk_bits_ret)2610 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
2611 				       int *ino_bits_ret, int *lblk_bits_ret)
2612 {
2613 	*ino_bits_ret = 8 * sizeof(nid_t);
2614 	*lblk_bits_ret = 8 * sizeof(block_t);
2615 }
2616 
f2fs_get_num_devices(struct super_block * sb)2617 static int f2fs_get_num_devices(struct super_block *sb)
2618 {
2619 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2620 
2621 	if (f2fs_is_multi_device(sbi))
2622 		return sbi->s_ndevs;
2623 	return 1;
2624 }
2625 
f2fs_get_devices(struct super_block * sb,struct request_queue ** devs)2626 static void f2fs_get_devices(struct super_block *sb,
2627 			     struct request_queue **devs)
2628 {
2629 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2630 	int i;
2631 
2632 	for (i = 0; i < sbi->s_ndevs; i++)
2633 		devs[i] = bdev_get_queue(FDEV(i).bdev);
2634 }
2635 
2636 static const struct fscrypt_operations f2fs_cryptops = {
2637 	.key_prefix		= "f2fs:",
2638 	.get_context		= f2fs_get_context,
2639 	.set_context		= f2fs_set_context,
2640 	.get_dummy_policy	= f2fs_get_dummy_policy,
2641 	.empty_dir		= f2fs_empty_dir,
2642 	.max_namelen		= F2FS_NAME_LEN,
2643 	.has_stable_inodes	= f2fs_has_stable_inodes,
2644 	.get_ino_and_lblk_bits	= f2fs_get_ino_and_lblk_bits,
2645 	.get_num_devices	= f2fs_get_num_devices,
2646 	.get_devices		= f2fs_get_devices,
2647 };
2648 #endif
2649 
f2fs_nfs_get_inode(struct super_block * sb,u64 ino,u32 generation)2650 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2651 		u64 ino, u32 generation)
2652 {
2653 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2654 	struct inode *inode;
2655 
2656 	if (f2fs_check_nid_range(sbi, ino))
2657 		return ERR_PTR(-ESTALE);
2658 
2659 	/*
2660 	 * f2fs_iget isn't quite right if the inode is currently unallocated!
2661 	 * However f2fs_iget currently does appropriate checks to handle stale
2662 	 * inodes so everything is OK.
2663 	 */
2664 	inode = f2fs_iget(sb, ino);
2665 	if (IS_ERR(inode))
2666 		return ERR_CAST(inode);
2667 	if (unlikely(generation && inode->i_generation != generation)) {
2668 		/* we didn't find the right inode.. */
2669 		iput(inode);
2670 		return ERR_PTR(-ESTALE);
2671 	}
2672 	return inode;
2673 }
2674 
f2fs_fh_to_dentry(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)2675 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2676 		int fh_len, int fh_type)
2677 {
2678 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2679 				    f2fs_nfs_get_inode);
2680 }
2681 
f2fs_fh_to_parent(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)2682 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2683 		int fh_len, int fh_type)
2684 {
2685 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2686 				    f2fs_nfs_get_inode);
2687 }
2688 
2689 static const struct export_operations f2fs_export_ops = {
2690 	.fh_to_dentry = f2fs_fh_to_dentry,
2691 	.fh_to_parent = f2fs_fh_to_parent,
2692 	.get_parent = f2fs_get_parent,
2693 };
2694 
max_file_blocks(void)2695 static loff_t max_file_blocks(void)
2696 {
2697 	loff_t result = 0;
2698 	loff_t leaf_count = DEF_ADDRS_PER_BLOCK;
2699 
2700 	/*
2701 	 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2702 	 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2703 	 * space in inode.i_addr, it will be more safe to reassign
2704 	 * result as zero.
2705 	 */
2706 
2707 	/* two direct node blocks */
2708 	result += (leaf_count * 2);
2709 
2710 	/* two indirect node blocks */
2711 	leaf_count *= NIDS_PER_BLOCK;
2712 	result += (leaf_count * 2);
2713 
2714 	/* one double indirect node block */
2715 	leaf_count *= NIDS_PER_BLOCK;
2716 	result += leaf_count;
2717 
2718 	return result;
2719 }
2720 
__f2fs_commit_super(struct buffer_head * bh,struct f2fs_super_block * super)2721 static int __f2fs_commit_super(struct buffer_head *bh,
2722 			struct f2fs_super_block *super)
2723 {
2724 	lock_buffer(bh);
2725 	if (super)
2726 		memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2727 	set_buffer_dirty(bh);
2728 	unlock_buffer(bh);
2729 
2730 	/* it's rare case, we can do fua all the time */
2731 	return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2732 }
2733 
sanity_check_area_boundary(struct f2fs_sb_info * sbi,struct buffer_head * bh)2734 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2735 					struct buffer_head *bh)
2736 {
2737 	struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2738 					(bh->b_data + F2FS_SUPER_OFFSET);
2739 	struct super_block *sb = sbi->sb;
2740 	u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2741 	u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2742 	u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2743 	u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2744 	u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2745 	u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2746 	u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2747 	u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2748 	u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2749 	u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2750 	u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2751 	u32 segment_count = le32_to_cpu(raw_super->segment_count);
2752 	u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2753 	u64 main_end_blkaddr = main_blkaddr +
2754 				(segment_count_main << log_blocks_per_seg);
2755 	u64 seg_end_blkaddr = segment0_blkaddr +
2756 				(segment_count << log_blocks_per_seg);
2757 
2758 	if (segment0_blkaddr != cp_blkaddr) {
2759 		f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2760 			  segment0_blkaddr, cp_blkaddr);
2761 		return true;
2762 	}
2763 
2764 	if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2765 							sit_blkaddr) {
2766 		f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2767 			  cp_blkaddr, sit_blkaddr,
2768 			  segment_count_ckpt << log_blocks_per_seg);
2769 		return true;
2770 	}
2771 
2772 	if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2773 							nat_blkaddr) {
2774 		f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2775 			  sit_blkaddr, nat_blkaddr,
2776 			  segment_count_sit << log_blocks_per_seg);
2777 		return true;
2778 	}
2779 
2780 	if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2781 							ssa_blkaddr) {
2782 		f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2783 			  nat_blkaddr, ssa_blkaddr,
2784 			  segment_count_nat << log_blocks_per_seg);
2785 		return true;
2786 	}
2787 
2788 	if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2789 							main_blkaddr) {
2790 		f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2791 			  ssa_blkaddr, main_blkaddr,
2792 			  segment_count_ssa << log_blocks_per_seg);
2793 		return true;
2794 	}
2795 
2796 	if (main_end_blkaddr > seg_end_blkaddr) {
2797 		f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)",
2798 			  main_blkaddr, seg_end_blkaddr,
2799 			  segment_count_main << log_blocks_per_seg);
2800 		return true;
2801 	} else if (main_end_blkaddr < seg_end_blkaddr) {
2802 		int err = 0;
2803 		char *res;
2804 
2805 		/* fix in-memory information all the time */
2806 		raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2807 				segment0_blkaddr) >> log_blocks_per_seg);
2808 
2809 		if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2810 			set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2811 			res = "internally";
2812 		} else {
2813 			err = __f2fs_commit_super(bh, NULL);
2814 			res = err ? "failed" : "done";
2815 		}
2816 		f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)",
2817 			  res, main_blkaddr, seg_end_blkaddr,
2818 			  segment_count_main << log_blocks_per_seg);
2819 		if (err)
2820 			return true;
2821 	}
2822 	return false;
2823 }
2824 
sanity_check_raw_super(struct f2fs_sb_info * sbi,struct buffer_head * bh)2825 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2826 				struct buffer_head *bh)
2827 {
2828 	block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main;
2829 	block_t total_sections, blocks_per_seg;
2830 	struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2831 					(bh->b_data + F2FS_SUPER_OFFSET);
2832 	size_t crc_offset = 0;
2833 	__u32 crc = 0;
2834 
2835 	if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
2836 		f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
2837 			  F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2838 		return -EINVAL;
2839 	}
2840 
2841 	/* Check checksum_offset and crc in superblock */
2842 	if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2843 		crc_offset = le32_to_cpu(raw_super->checksum_offset);
2844 		if (crc_offset !=
2845 			offsetof(struct f2fs_super_block, crc)) {
2846 			f2fs_info(sbi, "Invalid SB checksum offset: %zu",
2847 				  crc_offset);
2848 			return -EFSCORRUPTED;
2849 		}
2850 		crc = le32_to_cpu(raw_super->crc);
2851 		if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2852 			f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
2853 			return -EFSCORRUPTED;
2854 		}
2855 	}
2856 
2857 	/* Currently, support only 4KB page cache size */
2858 	if (F2FS_BLKSIZE != PAGE_SIZE) {
2859 		f2fs_info(sbi, "Invalid page_cache_size (%lu), supports only 4KB",
2860 			  PAGE_SIZE);
2861 		return -EFSCORRUPTED;
2862 	}
2863 
2864 	/* Currently, support only 4KB block size */
2865 	if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) {
2866 		f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u",
2867 			  le32_to_cpu(raw_super->log_blocksize),
2868 			  F2FS_BLKSIZE_BITS);
2869 		return -EFSCORRUPTED;
2870 	}
2871 
2872 	/* check log blocks per segment */
2873 	if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2874 		f2fs_info(sbi, "Invalid log blocks per segment (%u)",
2875 			  le32_to_cpu(raw_super->log_blocks_per_seg));
2876 		return -EFSCORRUPTED;
2877 	}
2878 
2879 	/* Currently, support 512/1024/2048/4096 bytes sector size */
2880 	if (le32_to_cpu(raw_super->log_sectorsize) >
2881 				F2FS_MAX_LOG_SECTOR_SIZE ||
2882 		le32_to_cpu(raw_super->log_sectorsize) <
2883 				F2FS_MIN_LOG_SECTOR_SIZE) {
2884 		f2fs_info(sbi, "Invalid log sectorsize (%u)",
2885 			  le32_to_cpu(raw_super->log_sectorsize));
2886 		return -EFSCORRUPTED;
2887 	}
2888 	if (le32_to_cpu(raw_super->log_sectors_per_block) +
2889 		le32_to_cpu(raw_super->log_sectorsize) !=
2890 			F2FS_MAX_LOG_SECTOR_SIZE) {
2891 		f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
2892 			  le32_to_cpu(raw_super->log_sectors_per_block),
2893 			  le32_to_cpu(raw_super->log_sectorsize));
2894 		return -EFSCORRUPTED;
2895 	}
2896 
2897 	segment_count = le32_to_cpu(raw_super->segment_count);
2898 	segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2899 	segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2900 	secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2901 	total_sections = le32_to_cpu(raw_super->section_count);
2902 
2903 	/* blocks_per_seg should be 512, given the above check */
2904 	blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2905 
2906 	if (segment_count > F2FS_MAX_SEGMENT ||
2907 				segment_count < F2FS_MIN_SEGMENTS) {
2908 		f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
2909 		return -EFSCORRUPTED;
2910 	}
2911 
2912 	if (total_sections > segment_count_main || total_sections < 1 ||
2913 			segs_per_sec > segment_count || !segs_per_sec) {
2914 		f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
2915 			  segment_count, total_sections, segs_per_sec);
2916 		return -EFSCORRUPTED;
2917 	}
2918 
2919 	if (segment_count_main != total_sections * segs_per_sec) {
2920 		f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)",
2921 			  segment_count_main, total_sections, segs_per_sec);
2922 		return -EFSCORRUPTED;
2923 	}
2924 
2925 	if ((segment_count / segs_per_sec) < total_sections) {
2926 		f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
2927 			  segment_count, segs_per_sec, total_sections);
2928 		return -EFSCORRUPTED;
2929 	}
2930 
2931 	if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2932 		f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
2933 			  segment_count, le64_to_cpu(raw_super->block_count));
2934 		return -EFSCORRUPTED;
2935 	}
2936 
2937 	if (RDEV(0).path[0]) {
2938 		block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
2939 		int i = 1;
2940 
2941 		while (i < MAX_DEVICES && RDEV(i).path[0]) {
2942 			dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
2943 			i++;
2944 		}
2945 		if (segment_count != dev_seg_count) {
2946 			f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
2947 					segment_count, dev_seg_count);
2948 			return -EFSCORRUPTED;
2949 		}
2950 	} else {
2951 		if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
2952 					!bdev_is_zoned(sbi->sb->s_bdev)) {
2953 			f2fs_info(sbi, "Zoned block device path is missing");
2954 			return -EFSCORRUPTED;
2955 		}
2956 	}
2957 
2958 	if (secs_per_zone > total_sections || !secs_per_zone) {
2959 		f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
2960 			  secs_per_zone, total_sections);
2961 		return -EFSCORRUPTED;
2962 	}
2963 	if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2964 			raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2965 			(le32_to_cpu(raw_super->extension_count) +
2966 			raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2967 		f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
2968 			  le32_to_cpu(raw_super->extension_count),
2969 			  raw_super->hot_ext_count,
2970 			  F2FS_MAX_EXTENSION);
2971 		return -EFSCORRUPTED;
2972 	}
2973 
2974 	if (le32_to_cpu(raw_super->cp_payload) >=
2975 				(blocks_per_seg - F2FS_CP_PACKS -
2976 				NR_CURSEG_PERSIST_TYPE)) {
2977 		f2fs_info(sbi, "Insane cp_payload (%u >= %u)",
2978 			  le32_to_cpu(raw_super->cp_payload),
2979 			  blocks_per_seg - F2FS_CP_PACKS -
2980 			  NR_CURSEG_PERSIST_TYPE);
2981 		return -EFSCORRUPTED;
2982 	}
2983 
2984 	/* check reserved ino info */
2985 	if (le32_to_cpu(raw_super->node_ino) != 1 ||
2986 		le32_to_cpu(raw_super->meta_ino) != 2 ||
2987 		le32_to_cpu(raw_super->root_ino) != 3) {
2988 		f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2989 			  le32_to_cpu(raw_super->node_ino),
2990 			  le32_to_cpu(raw_super->meta_ino),
2991 			  le32_to_cpu(raw_super->root_ino));
2992 		return -EFSCORRUPTED;
2993 	}
2994 
2995 	/* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2996 	if (sanity_check_area_boundary(sbi, bh))
2997 		return -EFSCORRUPTED;
2998 
2999 	return 0;
3000 }
3001 
f2fs_sanity_check_ckpt(struct f2fs_sb_info * sbi)3002 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
3003 {
3004 	unsigned int total, fsmeta;
3005 	struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3006 	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3007 	unsigned int ovp_segments, reserved_segments;
3008 	unsigned int main_segs, blocks_per_seg;
3009 	unsigned int sit_segs, nat_segs;
3010 	unsigned int sit_bitmap_size, nat_bitmap_size;
3011 	unsigned int log_blocks_per_seg;
3012 	unsigned int segment_count_main;
3013 	unsigned int cp_pack_start_sum, cp_payload;
3014 	block_t user_block_count, valid_user_blocks;
3015 	block_t avail_node_count, valid_node_count;
3016 	unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks;
3017 	int i, j;
3018 
3019 	total = le32_to_cpu(raw_super->segment_count);
3020 	fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
3021 	sit_segs = le32_to_cpu(raw_super->segment_count_sit);
3022 	fsmeta += sit_segs;
3023 	nat_segs = le32_to_cpu(raw_super->segment_count_nat);
3024 	fsmeta += nat_segs;
3025 	fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
3026 	fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
3027 
3028 	if (unlikely(fsmeta >= total))
3029 		return 1;
3030 
3031 	ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3032 	reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3033 
3034 	if (unlikely(fsmeta < F2FS_MIN_META_SEGMENTS ||
3035 			ovp_segments == 0 || reserved_segments == 0)) {
3036 		f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
3037 		return 1;
3038 	}
3039 
3040 	user_block_count = le64_to_cpu(ckpt->user_block_count);
3041 	segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3042 	log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3043 	if (!user_block_count || user_block_count >=
3044 			segment_count_main << log_blocks_per_seg) {
3045 		f2fs_err(sbi, "Wrong user_block_count: %u",
3046 			 user_block_count);
3047 		return 1;
3048 	}
3049 
3050 	valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
3051 	if (valid_user_blocks > user_block_count) {
3052 		f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
3053 			 valid_user_blocks, user_block_count);
3054 		return 1;
3055 	}
3056 
3057 	valid_node_count = le32_to_cpu(ckpt->valid_node_count);
3058 	avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
3059 	if (valid_node_count > avail_node_count) {
3060 		f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
3061 			 valid_node_count, avail_node_count);
3062 		return 1;
3063 	}
3064 
3065 	main_segs = le32_to_cpu(raw_super->segment_count_main);
3066 	blocks_per_seg = sbi->blocks_per_seg;
3067 
3068 	for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3069 		if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
3070 			le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
3071 			return 1;
3072 		for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
3073 			if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3074 				le32_to_cpu(ckpt->cur_node_segno[j])) {
3075 				f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
3076 					 i, j,
3077 					 le32_to_cpu(ckpt->cur_node_segno[i]));
3078 				return 1;
3079 			}
3080 		}
3081 	}
3082 	for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
3083 		if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
3084 			le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
3085 			return 1;
3086 		for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
3087 			if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
3088 				le32_to_cpu(ckpt->cur_data_segno[j])) {
3089 				f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
3090 					 i, j,
3091 					 le32_to_cpu(ckpt->cur_data_segno[i]));
3092 				return 1;
3093 			}
3094 		}
3095 	}
3096 	for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3097 		for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
3098 			if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3099 				le32_to_cpu(ckpt->cur_data_segno[j])) {
3100 				f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3101 					 i, j,
3102 					 le32_to_cpu(ckpt->cur_node_segno[i]));
3103 				return 1;
3104 			}
3105 		}
3106 	}
3107 
3108 	sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3109 	nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3110 
3111 	if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3112 		nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3113 		f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3114 			 sit_bitmap_size, nat_bitmap_size);
3115 		return 1;
3116 	}
3117 
3118 	cp_pack_start_sum = __start_sum_addr(sbi);
3119 	cp_payload = __cp_payload(sbi);
3120 	if (cp_pack_start_sum < cp_payload + 1 ||
3121 		cp_pack_start_sum > blocks_per_seg - 1 -
3122 			NR_CURSEG_PERSIST_TYPE) {
3123 		f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3124 			 cp_pack_start_sum);
3125 		return 1;
3126 	}
3127 
3128 	if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3129 		le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3130 		f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3131 			  "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3132 			  "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3133 			  le32_to_cpu(ckpt->checksum_offset));
3134 		return 1;
3135 	}
3136 
3137 	nat_blocks = nat_segs << log_blocks_per_seg;
3138 	nat_bits_bytes = nat_blocks / BITS_PER_BYTE;
3139 	nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
3140 	if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) &&
3141 		(cp_payload + F2FS_CP_PACKS +
3142 		NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) {
3143 		f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)",
3144 			  cp_payload, nat_bits_blocks);
3145 		return 1;
3146 	}
3147 
3148 	if (unlikely(f2fs_cp_error(sbi))) {
3149 		f2fs_err(sbi, "A bug case: need to run fsck");
3150 		return 1;
3151 	}
3152 	return 0;
3153 }
3154 
init_sb_info(struct f2fs_sb_info * sbi)3155 static void init_sb_info(struct f2fs_sb_info *sbi)
3156 {
3157 	struct f2fs_super_block *raw_super = sbi->raw_super;
3158 	int i;
3159 
3160 	sbi->log_sectors_per_block =
3161 		le32_to_cpu(raw_super->log_sectors_per_block);
3162 	sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3163 	sbi->blocksize = 1 << sbi->log_blocksize;
3164 	sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3165 	sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
3166 	sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3167 	sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3168 	sbi->total_sections = le32_to_cpu(raw_super->section_count);
3169 	sbi->total_node_count =
3170 		(le32_to_cpu(raw_super->segment_count_nat) / 2)
3171 			* sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
3172 	sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
3173 	sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
3174 	sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
3175 	sbi->cur_victim_sec = NULL_SECNO;
3176 	sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3177 	sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3178 	sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3179 	sbi->migration_granularity = sbi->segs_per_sec;
3180 
3181 	sbi->dir_level = DEF_DIR_LEVEL;
3182 	sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3183 	sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3184 	sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3185 	sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3186 	sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3187 	sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3188 				DEF_UMOUNT_DISCARD_TIMEOUT;
3189 	clear_sbi_flag(sbi, SBI_NEED_FSCK);
3190 
3191 	for (i = 0; i < NR_COUNT_TYPE; i++)
3192 		atomic_set(&sbi->nr_pages[i], 0);
3193 
3194 	for (i = 0; i < META; i++)
3195 		atomic_set(&sbi->wb_sync_req[i], 0);
3196 
3197 	INIT_LIST_HEAD(&sbi->s_list);
3198 	mutex_init(&sbi->umount_mutex);
3199 	init_rwsem(&sbi->io_order_lock);
3200 	spin_lock_init(&sbi->cp_lock);
3201 
3202 	sbi->dirty_device = 0;
3203 	spin_lock_init(&sbi->dev_lock);
3204 
3205 	init_rwsem(&sbi->sb_lock);
3206 	init_rwsem(&sbi->pin_sem);
3207 }
3208 
init_percpu_info(struct f2fs_sb_info * sbi)3209 static int init_percpu_info(struct f2fs_sb_info *sbi)
3210 {
3211 	int err;
3212 
3213 	err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3214 	if (err)
3215 		return err;
3216 
3217 	err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3218 								GFP_KERNEL);
3219 	if (err)
3220 		percpu_counter_destroy(&sbi->alloc_valid_block_count);
3221 
3222 	return err;
3223 }
3224 
3225 #ifdef CONFIG_BLK_DEV_ZONED
3226 
3227 struct f2fs_report_zones_args {
3228 	struct f2fs_sb_info *sbi;
3229 	struct f2fs_dev_info *dev;
3230 };
3231 
f2fs_report_zone_cb(struct blk_zone * zone,unsigned int idx,void * data)3232 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3233 			      void *data)
3234 {
3235 	struct f2fs_report_zones_args *rz_args = data;
3236 	block_t unusable_blocks = (zone->len - zone->capacity) >>
3237 					F2FS_LOG_SECTORS_PER_BLOCK;
3238 
3239 	if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3240 		return 0;
3241 
3242 	set_bit(idx, rz_args->dev->blkz_seq);
3243 	if (!rz_args->sbi->unusable_blocks_per_sec) {
3244 		rz_args->sbi->unusable_blocks_per_sec = unusable_blocks;
3245 		return 0;
3246 	}
3247 	if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) {
3248 		f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n");
3249 		return -EINVAL;
3250 	}
3251 	return 0;
3252 }
3253 
init_blkz_info(struct f2fs_sb_info * sbi,int devi)3254 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3255 {
3256 	struct block_device *bdev = FDEV(devi).bdev;
3257 	sector_t nr_sectors = bdev->bd_part->nr_sects;
3258 	struct f2fs_report_zones_args rep_zone_arg;
3259 	int ret;
3260 
3261 	if (!f2fs_sb_has_blkzoned(sbi))
3262 		return 0;
3263 
3264 	if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3265 				SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
3266 		return -EINVAL;
3267 	sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
3268 	if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
3269 				__ilog2_u32(sbi->blocks_per_blkz))
3270 		return -EINVAL;
3271 	sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
3272 	FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
3273 					sbi->log_blocks_per_blkz;
3274 	if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
3275 		FDEV(devi).nr_blkz++;
3276 
3277 	FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3278 					BITS_TO_LONGS(FDEV(devi).nr_blkz)
3279 					* sizeof(unsigned long),
3280 					GFP_KERNEL);
3281 	if (!FDEV(devi).blkz_seq)
3282 		return -ENOMEM;
3283 
3284 	rep_zone_arg.sbi = sbi;
3285 	rep_zone_arg.dev = &FDEV(devi);
3286 
3287 	ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3288 				  &rep_zone_arg);
3289 	if (ret < 0)
3290 		return ret;
3291 	return 0;
3292 }
3293 #endif
3294 
3295 /*
3296  * Read f2fs raw super block.
3297  * Because we have two copies of super block, so read both of them
3298  * to get the first valid one. If any one of them is broken, we pass
3299  * them recovery flag back to the caller.
3300  */
read_raw_super_block(struct f2fs_sb_info * sbi,struct f2fs_super_block ** raw_super,int * valid_super_block,int * recovery)3301 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3302 			struct f2fs_super_block **raw_super,
3303 			int *valid_super_block, int *recovery)
3304 {
3305 	struct super_block *sb = sbi->sb;
3306 	int block;
3307 	struct buffer_head *bh;
3308 	struct f2fs_super_block *super;
3309 	int err = 0;
3310 
3311 	super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3312 	if (!super)
3313 		return -ENOMEM;
3314 
3315 	for (block = 0; block < 2; block++) {
3316 		bh = sb_bread(sb, block);
3317 		if (!bh) {
3318 			f2fs_err(sbi, "Unable to read %dth superblock",
3319 				 block + 1);
3320 			err = -EIO;
3321 			*recovery = 1;
3322 			continue;
3323 		}
3324 
3325 		/* sanity checking of raw super */
3326 		err = sanity_check_raw_super(sbi, bh);
3327 		if (err) {
3328 			f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3329 				 block + 1);
3330 			brelse(bh);
3331 			*recovery = 1;
3332 			continue;
3333 		}
3334 
3335 		if (!*raw_super) {
3336 			memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3337 							sizeof(*super));
3338 			*valid_super_block = block;
3339 			*raw_super = super;
3340 		}
3341 		brelse(bh);
3342 	}
3343 
3344 	/* No valid superblock */
3345 	if (!*raw_super)
3346 		kfree(super);
3347 	else
3348 		err = 0;
3349 
3350 	return err;
3351 }
3352 
f2fs_commit_super(struct f2fs_sb_info * sbi,bool recover)3353 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3354 {
3355 	struct buffer_head *bh;
3356 	__u32 crc = 0;
3357 	int err;
3358 
3359 	if ((recover && f2fs_readonly(sbi->sb)) ||
3360 				bdev_read_only(sbi->sb->s_bdev)) {
3361 		set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3362 		return -EROFS;
3363 	}
3364 
3365 	/* we should update superblock crc here */
3366 	if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3367 		crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3368 				offsetof(struct f2fs_super_block, crc));
3369 		F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3370 	}
3371 
3372 	/* write back-up superblock first */
3373 	bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3374 	if (!bh)
3375 		return -EIO;
3376 	err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3377 	brelse(bh);
3378 
3379 	/* if we are in recovery path, skip writing valid superblock */
3380 	if (recover || err)
3381 		return err;
3382 
3383 	/* write current valid superblock */
3384 	bh = sb_bread(sbi->sb, sbi->valid_super_block);
3385 	if (!bh)
3386 		return -EIO;
3387 	err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3388 	brelse(bh);
3389 	return err;
3390 }
3391 
f2fs_scan_devices(struct f2fs_sb_info * sbi)3392 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3393 {
3394 	struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3395 	unsigned int max_devices = MAX_DEVICES;
3396 	int i;
3397 
3398 	/* Initialize single device information */
3399 	if (!RDEV(0).path[0]) {
3400 		if (!bdev_is_zoned(sbi->sb->s_bdev))
3401 			return 0;
3402 		max_devices = 1;
3403 	}
3404 
3405 	/*
3406 	 * Initialize multiple devices information, or single
3407 	 * zoned block device information.
3408 	 */
3409 	sbi->devs = f2fs_kzalloc(sbi,
3410 				 array_size(max_devices,
3411 					    sizeof(struct f2fs_dev_info)),
3412 				 GFP_KERNEL);
3413 	if (!sbi->devs)
3414 		return -ENOMEM;
3415 
3416 	for (i = 0; i < max_devices; i++) {
3417 
3418 		if (i > 0 && !RDEV(i).path[0])
3419 			break;
3420 
3421 		if (max_devices == 1) {
3422 			/* Single zoned block device mount */
3423 			FDEV(0).bdev =
3424 				blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3425 					sbi->sb->s_mode, sbi->sb->s_type);
3426 		} else {
3427 			/* Multi-device mount */
3428 			memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3429 			FDEV(i).total_segments =
3430 				le32_to_cpu(RDEV(i).total_segments);
3431 			if (i == 0) {
3432 				FDEV(i).start_blk = 0;
3433 				FDEV(i).end_blk = FDEV(i).start_blk +
3434 				    (FDEV(i).total_segments <<
3435 				    sbi->log_blocks_per_seg) - 1 +
3436 				    le32_to_cpu(raw_super->segment0_blkaddr);
3437 			} else {
3438 				FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3439 				FDEV(i).end_blk = FDEV(i).start_blk +
3440 					(FDEV(i).total_segments <<
3441 					sbi->log_blocks_per_seg) - 1;
3442 			}
3443 			FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3444 					sbi->sb->s_mode, sbi->sb->s_type);
3445 		}
3446 		if (IS_ERR(FDEV(i).bdev))
3447 			return PTR_ERR(FDEV(i).bdev);
3448 
3449 		/* to release errored devices */
3450 		sbi->s_ndevs = i + 1;
3451 
3452 #ifdef CONFIG_BLK_DEV_ZONED
3453 		if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3454 				!f2fs_sb_has_blkzoned(sbi)) {
3455 			f2fs_err(sbi, "Zoned block device feature not enabled\n");
3456 			return -EINVAL;
3457 		}
3458 		if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3459 			if (init_blkz_info(sbi, i)) {
3460 				f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3461 				return -EINVAL;
3462 			}
3463 			if (max_devices == 1)
3464 				break;
3465 			f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3466 				  i, FDEV(i).path,
3467 				  FDEV(i).total_segments,
3468 				  FDEV(i).start_blk, FDEV(i).end_blk,
3469 				  bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3470 				  "Host-aware" : "Host-managed");
3471 			continue;
3472 		}
3473 #endif
3474 		f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3475 			  i, FDEV(i).path,
3476 			  FDEV(i).total_segments,
3477 			  FDEV(i).start_blk, FDEV(i).end_blk);
3478 	}
3479 	f2fs_info(sbi,
3480 		  "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3481 	return 0;
3482 }
3483 
f2fs_setup_casefold(struct f2fs_sb_info * sbi)3484 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
3485 {
3486 #ifdef CONFIG_UNICODE
3487 	if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
3488 		const struct f2fs_sb_encodings *encoding_info;
3489 		struct unicode_map *encoding;
3490 		__u16 encoding_flags;
3491 
3492 		if (f2fs_sb_has_encrypt(sbi)) {
3493 			f2fs_err(sbi,
3494 				"Can't mount with encoding and encryption");
3495 			return -EINVAL;
3496 		}
3497 
3498 		if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info,
3499 					  &encoding_flags)) {
3500 			f2fs_err(sbi,
3501 				 "Encoding requested by superblock is unknown");
3502 			return -EINVAL;
3503 		}
3504 
3505 		encoding = utf8_load(encoding_info->version);
3506 		if (IS_ERR(encoding)) {
3507 			f2fs_err(sbi,
3508 				 "can't mount with superblock charset: %s-%s "
3509 				 "not supported by the kernel. flags: 0x%x.",
3510 				 encoding_info->name, encoding_info->version,
3511 				 encoding_flags);
3512 			return PTR_ERR(encoding);
3513 		}
3514 		f2fs_info(sbi, "Using encoding defined by superblock: "
3515 			 "%s-%s with flags 0x%hx", encoding_info->name,
3516 			 encoding_info->version?:"\b", encoding_flags);
3517 
3518 		sbi->sb->s_encoding = encoding;
3519 		sbi->sb->s_encoding_flags = encoding_flags;
3520 		sbi->sb->s_d_op = &f2fs_dentry_ops;
3521 	}
3522 #else
3523 	if (f2fs_sb_has_casefold(sbi)) {
3524 		f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
3525 		return -EINVAL;
3526 	}
3527 #endif
3528 	return 0;
3529 }
3530 
f2fs_tuning_parameters(struct f2fs_sb_info * sbi)3531 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3532 {
3533 	struct f2fs_sm_info *sm_i = SM_I(sbi);
3534 
3535 	/* adjust parameters according to the volume size */
3536 	if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3537 		F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3538 		sm_i->dcc_info->discard_granularity = 1;
3539 		sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3540 	}
3541 
3542 	sbi->readdir_ra = 1;
3543 }
3544 
3545 #ifdef CONFIG_F2FS_GRADING_SSR
f2fs_init_grading_ssr(struct f2fs_sb_info * sbi)3546 static void f2fs_init_grading_ssr(struct f2fs_sb_info *sbi)
3547 {
3548 	u32 total_blocks = le64_to_cpu(sbi->raw_super->block_count) >> 18;
3549 
3550 	if (total_blocks > 64) { /* 64G */
3551 		sbi->hot_cold_params.hot_data_lower_limit = SSR_HD_SAPCE_LIMIT_128G;
3552 		sbi->hot_cold_params.hot_data_waterline = SSR_HD_WATERLINE_128G;
3553 		sbi->hot_cold_params.warm_data_lower_limit = SSR_WD_SAPCE_LIMIT_128G;
3554 		sbi->hot_cold_params.warm_data_waterline = SSR_WD_WATERLINE_128G;
3555 		sbi->hot_cold_params.hot_node_lower_limit = SSR_HD_SAPCE_LIMIT_128G;
3556 		sbi->hot_cold_params.hot_node_waterline = SSR_HN_WATERLINE_128G;
3557 		sbi->hot_cold_params.warm_node_lower_limit = SSR_WN_SAPCE_LIMIT_128G;
3558 		sbi->hot_cold_params.warm_node_waterline = SSR_WN_WATERLINE_128G;
3559 		sbi->hot_cold_params.enable = GRADING_SSR_OFF;
3560 	} else {
3561 		sbi->hot_cold_params.hot_data_lower_limit = SSR_DEFALT_SPACE_LIMIT;
3562 		sbi->hot_cold_params.hot_data_waterline = SSR_DEFALT_WATERLINE;
3563 		sbi->hot_cold_params.warm_data_lower_limit = SSR_DEFALT_SPACE_LIMIT;
3564 		sbi->hot_cold_params.warm_data_waterline = SSR_DEFALT_WATERLINE;
3565 		sbi->hot_cold_params.hot_node_lower_limit = SSR_DEFALT_SPACE_LIMIT;
3566 		sbi->hot_cold_params.hot_node_waterline = SSR_DEFALT_WATERLINE;
3567 		sbi->hot_cold_params.warm_node_lower_limit = SSR_DEFALT_SPACE_LIMIT;
3568 		sbi->hot_cold_params.warm_node_waterline = SSR_DEFALT_WATERLINE;
3569 		sbi->hot_cold_params.enable = GRADING_SSR_OFF;
3570 	}
3571 }
3572 #endif
3573 
f2fs_fill_super(struct super_block * sb,void * data,int silent)3574 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3575 {
3576 	struct f2fs_sb_info *sbi;
3577 	struct f2fs_super_block *raw_super;
3578 	struct inode *root;
3579 	int err;
3580 	bool skip_recovery = false, need_fsck = false;
3581 	char *options = NULL;
3582 	int recovery, i, valid_super_block;
3583 	struct curseg_info *seg_i;
3584 	int retry_cnt = 1;
3585 
3586 try_onemore:
3587 	err = -EINVAL;
3588 	raw_super = NULL;
3589 	valid_super_block = -1;
3590 	recovery = 0;
3591 
3592 	/* allocate memory for f2fs-specific super block info */
3593 	sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3594 	if (!sbi)
3595 		return -ENOMEM;
3596 
3597 	sbi->sb = sb;
3598 
3599 	/* Load the checksum driver */
3600 	sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3601 	if (IS_ERR(sbi->s_chksum_driver)) {
3602 		f2fs_err(sbi, "Cannot load crc32 driver.");
3603 		err = PTR_ERR(sbi->s_chksum_driver);
3604 		sbi->s_chksum_driver = NULL;
3605 		goto free_sbi;
3606 	}
3607 
3608 	/* set a block size */
3609 	if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3610 		f2fs_err(sbi, "unable to set blocksize");
3611 		goto free_sbi;
3612 	}
3613 
3614 	err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3615 								&recovery);
3616 	if (err)
3617 		goto free_sbi;
3618 
3619 	sb->s_fs_info = sbi;
3620 	sbi->raw_super = raw_super;
3621 
3622 	/* precompute checksum seed for metadata */
3623 	if (f2fs_sb_has_inode_chksum(sbi))
3624 		sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3625 						sizeof(raw_super->uuid));
3626 
3627 	default_options(sbi);
3628 	/* parse mount options */
3629 	options = kstrdup((const char *)data, GFP_KERNEL);
3630 	if (data && !options) {
3631 		err = -ENOMEM;
3632 		goto free_sb_buf;
3633 	}
3634 
3635 	err = parse_options(sb, options, false);
3636 	if (err)
3637 		goto free_options;
3638 
3639 	sbi->max_file_blocks = max_file_blocks();
3640 	sb->s_maxbytes = sbi->max_file_blocks <<
3641 				le32_to_cpu(raw_super->log_blocksize);
3642 	sb->s_max_links = F2FS_LINK_MAX;
3643 
3644 	err = f2fs_setup_casefold(sbi);
3645 	if (err)
3646 		goto free_options;
3647 
3648 #ifdef CONFIG_QUOTA
3649 	sb->dq_op = &f2fs_quota_operations;
3650 	sb->s_qcop = &f2fs_quotactl_ops;
3651 	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3652 
3653 	if (f2fs_sb_has_quota_ino(sbi)) {
3654 		for (i = 0; i < MAXQUOTAS; i++) {
3655 			if (f2fs_qf_ino(sbi->sb, i))
3656 				sbi->nquota_files++;
3657 		}
3658 	}
3659 #endif
3660 
3661 	sb->s_op = &f2fs_sops;
3662 #ifdef CONFIG_FS_ENCRYPTION
3663 	sb->s_cop = &f2fs_cryptops;
3664 #endif
3665 #ifdef CONFIG_FS_VERITY
3666 	sb->s_vop = &f2fs_verityops;
3667 #endif
3668 	sb->s_xattr = f2fs_xattr_handlers;
3669 	sb->s_export_op = &f2fs_export_ops;
3670 	sb->s_magic = F2FS_SUPER_MAGIC;
3671 	sb->s_time_gran = 1;
3672 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3673 		(test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3674 	memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3675 	sb->s_iflags |= SB_I_CGROUPWB;
3676 
3677 	/* init f2fs-specific super block info */
3678 	sbi->valid_super_block = valid_super_block;
3679 	init_rwsem(&sbi->gc_lock);
3680 	mutex_init(&sbi->writepages);
3681 	mutex_init(&sbi->cp_mutex);
3682 	init_rwsem(&sbi->node_write);
3683 	init_rwsem(&sbi->node_change);
3684 
3685 	/* disallow all the data/node/meta page writes */
3686 	set_sbi_flag(sbi, SBI_POR_DOING);
3687 	spin_lock_init(&sbi->stat_lock);
3688 
3689 	/* init iostat info */
3690 	spin_lock_init(&sbi->iostat_lock);
3691 	sbi->iostat_enable = false;
3692 	sbi->iostat_period_ms = DEFAULT_IOSTAT_PERIOD_MS;
3693 
3694 	for (i = 0; i < NR_PAGE_TYPE; i++) {
3695 		int n = (i == META) ? 1: NR_TEMP_TYPE;
3696 		int j;
3697 
3698 		sbi->write_io[i] =
3699 			f2fs_kmalloc(sbi,
3700 				     array_size(n,
3701 						sizeof(struct f2fs_bio_info)),
3702 				     GFP_KERNEL);
3703 		if (!sbi->write_io[i]) {
3704 			err = -ENOMEM;
3705 			goto free_bio_info;
3706 		}
3707 
3708 		for (j = HOT; j < n; j++) {
3709 			init_rwsem(&sbi->write_io[i][j].io_rwsem);
3710 			sbi->write_io[i][j].sbi = sbi;
3711 			sbi->write_io[i][j].bio = NULL;
3712 			spin_lock_init(&sbi->write_io[i][j].io_lock);
3713 			INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3714 			INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list);
3715 			init_rwsem(&sbi->write_io[i][j].bio_list_lock);
3716 		}
3717 	}
3718 
3719 	init_rwsem(&sbi->cp_rwsem);
3720 	init_rwsem(&sbi->quota_sem);
3721 	init_waitqueue_head(&sbi->cp_wait);
3722 	init_sb_info(sbi);
3723 
3724 	err = init_percpu_info(sbi);
3725 	if (err)
3726 		goto free_bio_info;
3727 
3728 	if (F2FS_IO_ALIGNED(sbi)) {
3729 		sbi->write_io_dummy =
3730 			mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3731 		if (!sbi->write_io_dummy) {
3732 			err = -ENOMEM;
3733 			goto free_percpu;
3734 		}
3735 	}
3736 
3737 	/* init per sbi slab cache */
3738 	err = f2fs_init_xattr_caches(sbi);
3739 	if (err)
3740 		goto free_io_dummy;
3741 	err = f2fs_init_page_array_cache(sbi);
3742 	if (err)
3743 		goto free_xattr_cache;
3744 
3745 	/* get an inode for meta space */
3746 	sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3747 	if (IS_ERR(sbi->meta_inode)) {
3748 		f2fs_err(sbi, "Failed to read F2FS meta data inode");
3749 		err = PTR_ERR(sbi->meta_inode);
3750 		goto free_page_array_cache;
3751 	}
3752 
3753 	err = f2fs_get_valid_checkpoint(sbi);
3754 	if (err) {
3755 		f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
3756 		goto free_meta_inode;
3757 	}
3758 
3759 	if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3760 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3761 	if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3762 		set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3763 		sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3764 	}
3765 
3766 	if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
3767 		set_sbi_flag(sbi, SBI_NEED_FSCK);
3768 
3769 	/* Initialize device list */
3770 	err = f2fs_scan_devices(sbi);
3771 	if (err) {
3772 		f2fs_err(sbi, "Failed to find devices");
3773 		goto free_devices;
3774 	}
3775 
3776 	err = f2fs_init_post_read_wq(sbi);
3777 	if (err) {
3778 		f2fs_err(sbi, "Failed to initialize post read workqueue");
3779 		goto free_devices;
3780 	}
3781 
3782 	sbi->total_valid_node_count =
3783 				le32_to_cpu(sbi->ckpt->valid_node_count);
3784 	percpu_counter_set(&sbi->total_valid_inode_count,
3785 				le32_to_cpu(sbi->ckpt->valid_inode_count));
3786 	sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3787 	sbi->total_valid_block_count =
3788 				le64_to_cpu(sbi->ckpt->valid_block_count);
3789 	sbi->last_valid_block_count = sbi->total_valid_block_count;
3790 	sbi->reserved_blocks = 0;
3791 	sbi->current_reserved_blocks = 0;
3792 	limit_reserve_root(sbi);
3793 	adjust_unusable_cap_perc(sbi);
3794 
3795 	for (i = 0; i < NR_INODE_TYPE; i++) {
3796 		INIT_LIST_HEAD(&sbi->inode_list[i]);
3797 		spin_lock_init(&sbi->inode_lock[i]);
3798 	}
3799 	mutex_init(&sbi->flush_lock);
3800 
3801 	f2fs_init_extent_cache_info(sbi);
3802 
3803 	f2fs_init_ino_entry_info(sbi);
3804 
3805 	f2fs_init_fsync_node_info(sbi);
3806 
3807 	/* setup f2fs internal modules */
3808 	err = f2fs_build_segment_manager(sbi);
3809 	if (err) {
3810 		f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
3811 			 err);
3812 		goto free_sm;
3813 	}
3814 	err = f2fs_build_node_manager(sbi);
3815 	if (err) {
3816 		f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
3817 			 err);
3818 		goto free_nm;
3819 	}
3820 
3821 	err = adjust_reserved_segment(sbi);
3822 	if (err)
3823 		goto free_nm;
3824 
3825 	/* For write statistics */
3826 	if (sb->s_bdev->bd_part)
3827 		sbi->sectors_written_start =
3828 			(u64)part_stat_read(sb->s_bdev->bd_part,
3829 					    sectors[STAT_WRITE]);
3830 
3831 	/* Read accumulated write IO statistics if exists */
3832 	seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3833 	if (__exist_node_summaries(sbi))
3834 		sbi->kbytes_written =
3835 			le64_to_cpu(seg_i->journal->info.kbytes_written);
3836 
3837 	f2fs_build_gc_manager(sbi);
3838 
3839 	err = f2fs_build_stats(sbi);
3840 	if (err)
3841 		goto free_nm;
3842 
3843 	/* get an inode for node space */
3844 	sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3845 	if (IS_ERR(sbi->node_inode)) {
3846 		f2fs_err(sbi, "Failed to read node inode");
3847 		err = PTR_ERR(sbi->node_inode);
3848 		goto free_stats;
3849 	}
3850 
3851 	/* read root inode and dentry */
3852 	root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3853 	if (IS_ERR(root)) {
3854 		f2fs_err(sbi, "Failed to read root inode");
3855 		err = PTR_ERR(root);
3856 		goto free_node_inode;
3857 	}
3858 	if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3859 			!root->i_size || !root->i_nlink) {
3860 		iput(root);
3861 		err = -EINVAL;
3862 		goto free_node_inode;
3863 	}
3864 
3865 	sb->s_root = d_make_root(root); /* allocate root dentry */
3866 	if (!sb->s_root) {
3867 		err = -ENOMEM;
3868 		goto free_node_inode;
3869 	}
3870 #ifdef CONFIG_F2FS_GRADING_SSR
3871 	f2fs_init_grading_ssr(sbi);
3872 #endif
3873 	err = f2fs_register_sysfs(sbi);
3874 	if (err)
3875 		goto free_root_inode;
3876 
3877 #ifdef CONFIG_QUOTA
3878 	/* Enable quota usage during mount */
3879 	if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3880 		err = f2fs_enable_quotas(sb);
3881 		if (err)
3882 			f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
3883 	}
3884 #endif
3885 	/* if there are any orphan inodes, free them */
3886 	err = f2fs_recover_orphan_inodes(sbi);
3887 	if (err)
3888 		goto free_meta;
3889 
3890 	if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3891 		goto reset_checkpoint;
3892 
3893 	/* recover fsynced data */
3894 	if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
3895 			!test_opt(sbi, NORECOVERY)) {
3896 		/*
3897 		 * mount should be failed, when device has readonly mode, and
3898 		 * previous checkpoint was not done by clean system shutdown.
3899 		 */
3900 		if (f2fs_hw_is_readonly(sbi)) {
3901 			if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3902 				err = -EROFS;
3903 				f2fs_err(sbi, "Need to recover fsync data, but write access unavailable");
3904 				goto free_meta;
3905 			}
3906 			f2fs_info(sbi, "write access unavailable, skipping recovery");
3907 			goto reset_checkpoint;
3908 		}
3909 
3910 		if (need_fsck)
3911 			set_sbi_flag(sbi, SBI_NEED_FSCK);
3912 
3913 		if (skip_recovery)
3914 			goto reset_checkpoint;
3915 
3916 		err = f2fs_recover_fsync_data(sbi, false);
3917 		if (err < 0) {
3918 			if (err != -ENOMEM)
3919 				skip_recovery = true;
3920 			need_fsck = true;
3921 			f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
3922 				 err);
3923 			goto free_meta;
3924 		}
3925 	} else {
3926 		err = f2fs_recover_fsync_data(sbi, true);
3927 
3928 		if (!f2fs_readonly(sb) && err > 0) {
3929 			err = -EINVAL;
3930 			f2fs_err(sbi, "Need to recover fsync data");
3931 			goto free_meta;
3932 		}
3933 	}
3934 
3935 	/*
3936 	 * If the f2fs is not readonly and fsync data recovery succeeds,
3937 	 * check zoned block devices' write pointer consistency.
3938 	 */
3939 	if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
3940 		err = f2fs_check_write_pointer(sbi);
3941 		if (err)
3942 			goto free_meta;
3943 	}
3944 
3945 reset_checkpoint:
3946 	f2fs_init_inmem_curseg(sbi);
3947 
3948 	/* f2fs_recover_fsync_data() cleared this already */
3949 	clear_sbi_flag(sbi, SBI_POR_DOING);
3950 
3951 	if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3952 		err = f2fs_disable_checkpoint(sbi);
3953 		if (err)
3954 			goto sync_free_meta;
3955 	} else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3956 		f2fs_enable_checkpoint(sbi);
3957 	}
3958 
3959 	/*
3960 	 * If filesystem is not mounted as read-only then
3961 	 * do start the gc_thread.
3962 	 */
3963 	if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF ||
3964 		test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) {
3965 		/* After POR, we can run background GC thread.*/
3966 		err = f2fs_start_gc_thread(sbi);
3967 		if (err)
3968 			goto sync_free_meta;
3969 	}
3970 	kvfree(options);
3971 
3972 	/* recover broken superblock */
3973 	if (recovery) {
3974 		err = f2fs_commit_super(sbi, true);
3975 		f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
3976 			  sbi->valid_super_block ? 1 : 2, err);
3977 	}
3978 
3979 	f2fs_join_shrinker(sbi);
3980 
3981 	f2fs_tuning_parameters(sbi);
3982 
3983 	f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
3984 		    cur_cp_version(F2FS_CKPT(sbi)));
3985 	f2fs_update_time(sbi, CP_TIME);
3986 	f2fs_update_time(sbi, REQ_TIME);
3987 	clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3988 	return 0;
3989 
3990 sync_free_meta:
3991 	/* safe to flush all the data */
3992 	sync_filesystem(sbi->sb);
3993 	retry_cnt = 0;
3994 
3995 free_meta:
3996 #ifdef CONFIG_QUOTA
3997 	f2fs_truncate_quota_inode_pages(sb);
3998 	if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3999 		f2fs_quota_off_umount(sbi->sb);
4000 #endif
4001 	/*
4002 	 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
4003 	 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
4004 	 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
4005 	 * falls into an infinite loop in f2fs_sync_meta_pages().
4006 	 */
4007 	truncate_inode_pages_final(META_MAPPING(sbi));
4008 	/* evict some inodes being cached by GC */
4009 	evict_inodes(sb);
4010 	f2fs_unregister_sysfs(sbi);
4011 free_root_inode:
4012 	dput(sb->s_root);
4013 	sb->s_root = NULL;
4014 free_node_inode:
4015 	f2fs_release_ino_entry(sbi, true);
4016 	truncate_inode_pages_final(NODE_MAPPING(sbi));
4017 	iput(sbi->node_inode);
4018 	sbi->node_inode = NULL;
4019 free_stats:
4020 	f2fs_destroy_stats(sbi);
4021 free_nm:
4022 	f2fs_destroy_node_manager(sbi);
4023 free_sm:
4024 	f2fs_destroy_segment_manager(sbi);
4025 	f2fs_destroy_post_read_wq(sbi);
4026 free_devices:
4027 	destroy_device_list(sbi);
4028 	kvfree(sbi->ckpt);
4029 free_meta_inode:
4030 	make_bad_inode(sbi->meta_inode);
4031 	iput(sbi->meta_inode);
4032 	sbi->meta_inode = NULL;
4033 free_page_array_cache:
4034 	f2fs_destroy_page_array_cache(sbi);
4035 free_xattr_cache:
4036 	f2fs_destroy_xattr_caches(sbi);
4037 free_io_dummy:
4038 	mempool_destroy(sbi->write_io_dummy);
4039 free_percpu:
4040 	destroy_percpu_info(sbi);
4041 free_bio_info:
4042 	for (i = 0; i < NR_PAGE_TYPE; i++)
4043 		kvfree(sbi->write_io[i]);
4044 
4045 #ifdef CONFIG_UNICODE
4046 	utf8_unload(sb->s_encoding);
4047 	sb->s_encoding = NULL;
4048 #endif
4049 free_options:
4050 #ifdef CONFIG_QUOTA
4051 	for (i = 0; i < MAXQUOTAS; i++)
4052 		kfree(F2FS_OPTION(sbi).s_qf_names[i]);
4053 #endif
4054 	fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
4055 	kvfree(options);
4056 free_sb_buf:
4057 	kfree(raw_super);
4058 free_sbi:
4059 	if (sbi->s_chksum_driver)
4060 		crypto_free_shash(sbi->s_chksum_driver);
4061 	kfree(sbi);
4062 
4063 	/* give only one another chance */
4064 	if (retry_cnt > 0 && skip_recovery) {
4065 		retry_cnt--;
4066 		shrink_dcache_sb(sb);
4067 		goto try_onemore;
4068 	}
4069 	return err;
4070 }
4071 
f2fs_mount(struct file_system_type * fs_type,int flags,const char * dev_name,void * data)4072 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
4073 			const char *dev_name, void *data)
4074 {
4075 	return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
4076 }
4077 
kill_f2fs_super(struct super_block * sb)4078 static void kill_f2fs_super(struct super_block *sb)
4079 {
4080 	if (sb->s_root) {
4081 		struct f2fs_sb_info *sbi = F2FS_SB(sb);
4082 
4083 		set_sbi_flag(sbi, SBI_IS_CLOSE);
4084 		f2fs_stop_gc_thread(sbi);
4085 		f2fs_stop_discard_thread(sbi);
4086 
4087 		if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
4088 				!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4089 			struct cp_control cpc = {
4090 				.reason = CP_UMOUNT,
4091 			};
4092 			f2fs_write_checkpoint(sbi, &cpc);
4093 		}
4094 
4095 		if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
4096 			sb->s_flags &= ~SB_RDONLY;
4097 	}
4098 	kill_block_super(sb);
4099 }
4100 
4101 static struct file_system_type f2fs_fs_type = {
4102 	.owner		= THIS_MODULE,
4103 	.name		= "f2fs",
4104 	.mount		= f2fs_mount,
4105 	.kill_sb	= kill_f2fs_super,
4106 	.fs_flags	= FS_REQUIRES_DEV,
4107 };
4108 MODULE_ALIAS_FS("f2fs");
4109 
init_inodecache(void)4110 static int __init init_inodecache(void)
4111 {
4112 	f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
4113 			sizeof(struct f2fs_inode_info), 0,
4114 			SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
4115 	if (!f2fs_inode_cachep)
4116 		return -ENOMEM;
4117 	return 0;
4118 }
4119 
destroy_inodecache(void)4120 static void destroy_inodecache(void)
4121 {
4122 	/*
4123 	 * Make sure all delayed rcu free inodes are flushed before we
4124 	 * destroy cache.
4125 	 */
4126 	rcu_barrier();
4127 	kmem_cache_destroy(f2fs_inode_cachep);
4128 }
4129 
init_f2fs_fs(void)4130 static int __init init_f2fs_fs(void)
4131 {
4132 	int err;
4133 
4134 	if (PAGE_SIZE != F2FS_BLKSIZE) {
4135 		printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
4136 				PAGE_SIZE, F2FS_BLKSIZE);
4137 		return -EINVAL;
4138 	}
4139 
4140 	f2fs_build_trace_ios();
4141 
4142 	err = init_inodecache();
4143 	if (err)
4144 		goto fail;
4145 	err = f2fs_create_node_manager_caches();
4146 	if (err)
4147 		goto free_inodecache;
4148 	err = f2fs_create_segment_manager_caches();
4149 	if (err)
4150 		goto free_node_manager_caches;
4151 	err = f2fs_create_checkpoint_caches();
4152 	if (err)
4153 		goto free_segment_manager_caches;
4154 	err = f2fs_create_recovery_cache();
4155 	if (err)
4156 		goto free_checkpoint_caches;
4157 	err = f2fs_create_extent_cache();
4158 	if (err)
4159 		goto free_recovery_cache;
4160 	err = f2fs_create_garbage_collection_cache();
4161 	if (err)
4162 		goto free_extent_cache;
4163 	err = f2fs_init_sysfs();
4164 	if (err)
4165 		goto free_garbage_collection_cache;
4166 	err = register_shrinker(&f2fs_shrinker_info);
4167 	if (err)
4168 		goto free_sysfs;
4169 	err = register_filesystem(&f2fs_fs_type);
4170 	if (err)
4171 		goto free_shrinker;
4172 	f2fs_create_root_stats();
4173 	err = f2fs_init_post_read_processing();
4174 	if (err)
4175 		goto free_root_stats;
4176 	err = f2fs_init_bio_entry_cache();
4177 	if (err)
4178 		goto free_post_read;
4179 	err = f2fs_init_bioset();
4180 	if (err)
4181 		goto free_bio_enrty_cache;
4182 	err = f2fs_init_compress_mempool();
4183 	if (err)
4184 		goto free_bioset;
4185 	err = f2fs_init_compress_cache();
4186 	if (err)
4187 		goto free_compress_mempool;
4188 	return 0;
4189 free_compress_mempool:
4190 	f2fs_destroy_compress_mempool();
4191 free_bioset:
4192 	f2fs_destroy_bioset();
4193 free_bio_enrty_cache:
4194 	f2fs_destroy_bio_entry_cache();
4195 free_post_read:
4196 	f2fs_destroy_post_read_processing();
4197 free_root_stats:
4198 	f2fs_destroy_root_stats();
4199 	unregister_filesystem(&f2fs_fs_type);
4200 free_shrinker:
4201 	unregister_shrinker(&f2fs_shrinker_info);
4202 free_sysfs:
4203 	f2fs_exit_sysfs();
4204 free_garbage_collection_cache:
4205 	f2fs_destroy_garbage_collection_cache();
4206 free_extent_cache:
4207 	f2fs_destroy_extent_cache();
4208 free_recovery_cache:
4209 	f2fs_destroy_recovery_cache();
4210 free_checkpoint_caches:
4211 	f2fs_destroy_checkpoint_caches();
4212 free_segment_manager_caches:
4213 	f2fs_destroy_segment_manager_caches();
4214 free_node_manager_caches:
4215 	f2fs_destroy_node_manager_caches();
4216 free_inodecache:
4217 	destroy_inodecache();
4218 fail:
4219 	return err;
4220 }
4221 
exit_f2fs_fs(void)4222 static void __exit exit_f2fs_fs(void)
4223 {
4224 	f2fs_destroy_compress_cache();
4225 	f2fs_destroy_compress_mempool();
4226 	f2fs_destroy_bioset();
4227 	f2fs_destroy_bio_entry_cache();
4228 	f2fs_destroy_post_read_processing();
4229 	f2fs_destroy_root_stats();
4230 	unregister_filesystem(&f2fs_fs_type);
4231 	unregister_shrinker(&f2fs_shrinker_info);
4232 	f2fs_exit_sysfs();
4233 	f2fs_destroy_garbage_collection_cache();
4234 	f2fs_destroy_extent_cache();
4235 	f2fs_destroy_recovery_cache();
4236 	f2fs_destroy_checkpoint_caches();
4237 	f2fs_destroy_segment_manager_caches();
4238 	f2fs_destroy_node_manager_caches();
4239 	destroy_inodecache();
4240 	f2fs_destroy_trace_ios();
4241 }
4242 
4243 module_init(init_f2fs_fs)
4244 module_exit(exit_f2fs_fs)
4245 
4246 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
4247 MODULE_DESCRIPTION("Flash Friendly File System");
4248 MODULE_LICENSE("GPL");
4249 MODULE_SOFTDEP("pre: crc32");
4250 
4251