Revert "tty: hvc: Fix data abort due to race in hvc_open"
[linux/fpc-iii.git] / fs / f2fs / super.c
blob4696c9cb47a5f381ab9b09cc95f575418cd55e50
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * fs/f2fs/super.c
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>
29 #include "f2fs.h"
30 #include "node.h"
31 #include "segment.h"
32 #include "xattr.h"
33 #include "gc.h"
34 #include "trace.h"
36 #define CREATE_TRACE_POINTS
37 #include <trace/events/f2fs.h>
39 static struct kmem_cache *f2fs_inode_cachep;
41 #ifdef CONFIG_F2FS_FAULT_INJECTION
43 const char *f2fs_fault_name[FAULT_MAX] = {
44 [FAULT_KMALLOC] = "kmalloc",
45 [FAULT_KVMALLOC] = "kvmalloc",
46 [FAULT_PAGE_ALLOC] = "page alloc",
47 [FAULT_PAGE_GET] = "page get",
48 [FAULT_ALLOC_BIO] = "alloc bio",
49 [FAULT_ALLOC_NID] = "alloc nid",
50 [FAULT_ORPHAN] = "orphan",
51 [FAULT_BLOCK] = "no more block",
52 [FAULT_DIR_DEPTH] = "too big dir depth",
53 [FAULT_EVICT_INODE] = "evict_inode fail",
54 [FAULT_TRUNCATE] = "truncate fail",
55 [FAULT_READ_IO] = "read IO error",
56 [FAULT_CHECKPOINT] = "checkpoint error",
57 [FAULT_DISCARD] = "discard error",
58 [FAULT_WRITE_IO] = "write IO error",
61 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
62 unsigned int type)
64 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
66 if (rate) {
67 atomic_set(&ffi->inject_ops, 0);
68 ffi->inject_rate = rate;
71 if (type)
72 ffi->inject_type = type;
74 if (!rate && !type)
75 memset(ffi, 0, sizeof(struct f2fs_fault_info));
77 #endif
79 /* f2fs-wide shrinker description */
80 static struct shrinker f2fs_shrinker_info = {
81 .scan_objects = f2fs_shrink_scan,
82 .count_objects = f2fs_shrink_count,
83 .seeks = DEFAULT_SEEKS,
86 enum {
87 Opt_gc_background,
88 Opt_disable_roll_forward,
89 Opt_norecovery,
90 Opt_discard,
91 Opt_nodiscard,
92 Opt_noheap,
93 Opt_heap,
94 Opt_user_xattr,
95 Opt_nouser_xattr,
96 Opt_acl,
97 Opt_noacl,
98 Opt_active_logs,
99 Opt_disable_ext_identify,
100 Opt_inline_xattr,
101 Opt_noinline_xattr,
102 Opt_inline_xattr_size,
103 Opt_inline_data,
104 Opt_inline_dentry,
105 Opt_noinline_dentry,
106 Opt_flush_merge,
107 Opt_noflush_merge,
108 Opt_nobarrier,
109 Opt_fastboot,
110 Opt_extent_cache,
111 Opt_noextent_cache,
112 Opt_noinline_data,
113 Opt_data_flush,
114 Opt_reserve_root,
115 Opt_resgid,
116 Opt_resuid,
117 Opt_mode,
118 Opt_io_size_bits,
119 Opt_fault_injection,
120 Opt_fault_type,
121 Opt_lazytime,
122 Opt_nolazytime,
123 Opt_quota,
124 Opt_noquota,
125 Opt_usrquota,
126 Opt_grpquota,
127 Opt_prjquota,
128 Opt_usrjquota,
129 Opt_grpjquota,
130 Opt_prjjquota,
131 Opt_offusrjquota,
132 Opt_offgrpjquota,
133 Opt_offprjjquota,
134 Opt_jqfmt_vfsold,
135 Opt_jqfmt_vfsv0,
136 Opt_jqfmt_vfsv1,
137 Opt_whint,
138 Opt_alloc,
139 Opt_fsync,
140 Opt_test_dummy_encryption,
141 Opt_checkpoint_disable,
142 Opt_checkpoint_disable_cap,
143 Opt_checkpoint_disable_cap_perc,
144 Opt_checkpoint_enable,
145 Opt_compress_algorithm,
146 Opt_compress_log_size,
147 Opt_compress_extension,
148 Opt_err,
151 static match_table_t f2fs_tokens = {
152 {Opt_gc_background, "background_gc=%s"},
153 {Opt_disable_roll_forward, "disable_roll_forward"},
154 {Opt_norecovery, "norecovery"},
155 {Opt_discard, "discard"},
156 {Opt_nodiscard, "nodiscard"},
157 {Opt_noheap, "no_heap"},
158 {Opt_heap, "heap"},
159 {Opt_user_xattr, "user_xattr"},
160 {Opt_nouser_xattr, "nouser_xattr"},
161 {Opt_acl, "acl"},
162 {Opt_noacl, "noacl"},
163 {Opt_active_logs, "active_logs=%u"},
164 {Opt_disable_ext_identify, "disable_ext_identify"},
165 {Opt_inline_xattr, "inline_xattr"},
166 {Opt_noinline_xattr, "noinline_xattr"},
167 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
168 {Opt_inline_data, "inline_data"},
169 {Opt_inline_dentry, "inline_dentry"},
170 {Opt_noinline_dentry, "noinline_dentry"},
171 {Opt_flush_merge, "flush_merge"},
172 {Opt_noflush_merge, "noflush_merge"},
173 {Opt_nobarrier, "nobarrier"},
174 {Opt_fastboot, "fastboot"},
175 {Opt_extent_cache, "extent_cache"},
176 {Opt_noextent_cache, "noextent_cache"},
177 {Opt_noinline_data, "noinline_data"},
178 {Opt_data_flush, "data_flush"},
179 {Opt_reserve_root, "reserve_root=%u"},
180 {Opt_resgid, "resgid=%u"},
181 {Opt_resuid, "resuid=%u"},
182 {Opt_mode, "mode=%s"},
183 {Opt_io_size_bits, "io_bits=%u"},
184 {Opt_fault_injection, "fault_injection=%u"},
185 {Opt_fault_type, "fault_type=%u"},
186 {Opt_lazytime, "lazytime"},
187 {Opt_nolazytime, "nolazytime"},
188 {Opt_quota, "quota"},
189 {Opt_noquota, "noquota"},
190 {Opt_usrquota, "usrquota"},
191 {Opt_grpquota, "grpquota"},
192 {Opt_prjquota, "prjquota"},
193 {Opt_usrjquota, "usrjquota=%s"},
194 {Opt_grpjquota, "grpjquota=%s"},
195 {Opt_prjjquota, "prjjquota=%s"},
196 {Opt_offusrjquota, "usrjquota="},
197 {Opt_offgrpjquota, "grpjquota="},
198 {Opt_offprjjquota, "prjjquota="},
199 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
200 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
201 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
202 {Opt_whint, "whint_mode=%s"},
203 {Opt_alloc, "alloc_mode=%s"},
204 {Opt_fsync, "fsync_mode=%s"},
205 {Opt_test_dummy_encryption, "test_dummy_encryption"},
206 {Opt_checkpoint_disable, "checkpoint=disable"},
207 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
208 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
209 {Opt_checkpoint_enable, "checkpoint=enable"},
210 {Opt_compress_algorithm, "compress_algorithm=%s"},
211 {Opt_compress_log_size, "compress_log_size=%u"},
212 {Opt_compress_extension, "compress_extension=%s"},
213 {Opt_err, NULL},
216 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
218 struct va_format vaf;
219 va_list args;
220 int level;
222 va_start(args, fmt);
224 level = printk_get_level(fmt);
225 vaf.fmt = printk_skip_level(fmt);
226 vaf.va = &args;
227 printk("%c%cF2FS-fs (%s): %pV\n",
228 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
230 va_end(args);
233 #ifdef CONFIG_UNICODE
234 static const struct f2fs_sb_encodings {
235 __u16 magic;
236 char *name;
237 char *version;
238 } f2fs_sb_encoding_map[] = {
239 {F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"},
242 static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb,
243 const struct f2fs_sb_encodings **encoding,
244 __u16 *flags)
246 __u16 magic = le16_to_cpu(sb->s_encoding);
247 int i;
249 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
250 if (magic == f2fs_sb_encoding_map[i].magic)
251 break;
253 if (i >= ARRAY_SIZE(f2fs_sb_encoding_map))
254 return -EINVAL;
256 *encoding = &f2fs_sb_encoding_map[i];
257 *flags = le16_to_cpu(sb->s_encoding_flags);
259 return 0;
261 #endif
263 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
265 block_t limit = min((sbi->user_block_count << 1) / 1000,
266 sbi->user_block_count - sbi->reserved_blocks);
268 /* limit is 0.2% */
269 if (test_opt(sbi, RESERVE_ROOT) &&
270 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
271 F2FS_OPTION(sbi).root_reserved_blocks = limit;
272 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
273 F2FS_OPTION(sbi).root_reserved_blocks);
275 if (!test_opt(sbi, RESERVE_ROOT) &&
276 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
277 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
278 !gid_eq(F2FS_OPTION(sbi).s_resgid,
279 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
280 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
281 from_kuid_munged(&init_user_ns,
282 F2FS_OPTION(sbi).s_resuid),
283 from_kgid_munged(&init_user_ns,
284 F2FS_OPTION(sbi).s_resgid));
287 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
289 if (!F2FS_OPTION(sbi).unusable_cap_perc)
290 return;
292 if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
293 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
294 else
295 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
296 F2FS_OPTION(sbi).unusable_cap_perc;
298 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
299 F2FS_OPTION(sbi).unusable_cap,
300 F2FS_OPTION(sbi).unusable_cap_perc);
303 static void init_once(void *foo)
305 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
307 inode_init_once(&fi->vfs_inode);
310 #ifdef CONFIG_QUOTA
311 static const char * const quotatypes[] = INITQFNAMES;
312 #define QTYPE2NAME(t) (quotatypes[t])
313 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
314 substring_t *args)
316 struct f2fs_sb_info *sbi = F2FS_SB(sb);
317 char *qname;
318 int ret = -EINVAL;
320 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
321 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
322 return -EINVAL;
324 if (f2fs_sb_has_quota_ino(sbi)) {
325 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
326 return 0;
329 qname = match_strdup(args);
330 if (!qname) {
331 f2fs_err(sbi, "Not enough memory for storing quotafile name");
332 return -ENOMEM;
334 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
335 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
336 ret = 0;
337 else
338 f2fs_err(sbi, "%s quota file already specified",
339 QTYPE2NAME(qtype));
340 goto errout;
342 if (strchr(qname, '/')) {
343 f2fs_err(sbi, "quotafile must be on filesystem root");
344 goto errout;
346 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
347 set_opt(sbi, QUOTA);
348 return 0;
349 errout:
350 kvfree(qname);
351 return ret;
354 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
356 struct f2fs_sb_info *sbi = F2FS_SB(sb);
358 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
359 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
360 return -EINVAL;
362 kvfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
363 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
364 return 0;
367 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
370 * We do the test below only for project quotas. 'usrquota' and
371 * 'grpquota' mount options are allowed even without quota feature
372 * to support legacy quotas in quota files.
374 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
375 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
376 return -1;
378 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
379 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
380 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
381 if (test_opt(sbi, USRQUOTA) &&
382 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
383 clear_opt(sbi, USRQUOTA);
385 if (test_opt(sbi, GRPQUOTA) &&
386 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
387 clear_opt(sbi, GRPQUOTA);
389 if (test_opt(sbi, PRJQUOTA) &&
390 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
391 clear_opt(sbi, PRJQUOTA);
393 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
394 test_opt(sbi, PRJQUOTA)) {
395 f2fs_err(sbi, "old and new quota format mixing");
396 return -1;
399 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
400 f2fs_err(sbi, "journaled quota format not specified");
401 return -1;
405 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
406 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
407 F2FS_OPTION(sbi).s_jquota_fmt = 0;
409 return 0;
411 #endif
413 static int parse_options(struct super_block *sb, char *options)
415 struct f2fs_sb_info *sbi = F2FS_SB(sb);
416 substring_t args[MAX_OPT_ARGS];
417 unsigned char (*ext)[F2FS_EXTENSION_LEN];
418 char *p, *name;
419 int arg = 0, ext_cnt;
420 kuid_t uid;
421 kgid_t gid;
422 #ifdef CONFIG_QUOTA
423 int ret;
424 #endif
426 if (!options)
427 return 0;
429 while ((p = strsep(&options, ",")) != NULL) {
430 int token;
431 if (!*p)
432 continue;
434 * Initialize args struct so we know whether arg was
435 * found; some options take optional arguments.
437 args[0].to = args[0].from = NULL;
438 token = match_token(p, f2fs_tokens, args);
440 switch (token) {
441 case Opt_gc_background:
442 name = match_strdup(&args[0]);
444 if (!name)
445 return -ENOMEM;
446 if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
447 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
448 } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
449 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
450 } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
451 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
452 } else {
453 kvfree(name);
454 return -EINVAL;
456 kvfree(name);
457 break;
458 case Opt_disable_roll_forward:
459 set_opt(sbi, DISABLE_ROLL_FORWARD);
460 break;
461 case Opt_norecovery:
462 /* this option mounts f2fs with ro */
463 set_opt(sbi, NORECOVERY);
464 if (!f2fs_readonly(sb))
465 return -EINVAL;
466 break;
467 case Opt_discard:
468 set_opt(sbi, DISCARD);
469 break;
470 case Opt_nodiscard:
471 if (f2fs_sb_has_blkzoned(sbi)) {
472 f2fs_warn(sbi, "discard is required for zoned block devices");
473 return -EINVAL;
475 clear_opt(sbi, DISCARD);
476 break;
477 case Opt_noheap:
478 set_opt(sbi, NOHEAP);
479 break;
480 case Opt_heap:
481 clear_opt(sbi, NOHEAP);
482 break;
483 #ifdef CONFIG_F2FS_FS_XATTR
484 case Opt_user_xattr:
485 set_opt(sbi, XATTR_USER);
486 break;
487 case Opt_nouser_xattr:
488 clear_opt(sbi, XATTR_USER);
489 break;
490 case Opt_inline_xattr:
491 set_opt(sbi, INLINE_XATTR);
492 break;
493 case Opt_noinline_xattr:
494 clear_opt(sbi, INLINE_XATTR);
495 break;
496 case Opt_inline_xattr_size:
497 if (args->from && match_int(args, &arg))
498 return -EINVAL;
499 set_opt(sbi, INLINE_XATTR_SIZE);
500 F2FS_OPTION(sbi).inline_xattr_size = arg;
501 break;
502 #else
503 case Opt_user_xattr:
504 f2fs_info(sbi, "user_xattr options not supported");
505 break;
506 case Opt_nouser_xattr:
507 f2fs_info(sbi, "nouser_xattr options not supported");
508 break;
509 case Opt_inline_xattr:
510 f2fs_info(sbi, "inline_xattr options not supported");
511 break;
512 case Opt_noinline_xattr:
513 f2fs_info(sbi, "noinline_xattr options not supported");
514 break;
515 #endif
516 #ifdef CONFIG_F2FS_FS_POSIX_ACL
517 case Opt_acl:
518 set_opt(sbi, POSIX_ACL);
519 break;
520 case Opt_noacl:
521 clear_opt(sbi, POSIX_ACL);
522 break;
523 #else
524 case Opt_acl:
525 f2fs_info(sbi, "acl options not supported");
526 break;
527 case Opt_noacl:
528 f2fs_info(sbi, "noacl options not supported");
529 break;
530 #endif
531 case Opt_active_logs:
532 if (args->from && match_int(args, &arg))
533 return -EINVAL;
534 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
535 return -EINVAL;
536 F2FS_OPTION(sbi).active_logs = arg;
537 break;
538 case Opt_disable_ext_identify:
539 set_opt(sbi, DISABLE_EXT_IDENTIFY);
540 break;
541 case Opt_inline_data:
542 set_opt(sbi, INLINE_DATA);
543 break;
544 case Opt_inline_dentry:
545 set_opt(sbi, INLINE_DENTRY);
546 break;
547 case Opt_noinline_dentry:
548 clear_opt(sbi, INLINE_DENTRY);
549 break;
550 case Opt_flush_merge:
551 set_opt(sbi, FLUSH_MERGE);
552 break;
553 case Opt_noflush_merge:
554 clear_opt(sbi, FLUSH_MERGE);
555 break;
556 case Opt_nobarrier:
557 set_opt(sbi, NOBARRIER);
558 break;
559 case Opt_fastboot:
560 set_opt(sbi, FASTBOOT);
561 break;
562 case Opt_extent_cache:
563 set_opt(sbi, EXTENT_CACHE);
564 break;
565 case Opt_noextent_cache:
566 clear_opt(sbi, EXTENT_CACHE);
567 break;
568 case Opt_noinline_data:
569 clear_opt(sbi, INLINE_DATA);
570 break;
571 case Opt_data_flush:
572 set_opt(sbi, DATA_FLUSH);
573 break;
574 case Opt_reserve_root:
575 if (args->from && match_int(args, &arg))
576 return -EINVAL;
577 if (test_opt(sbi, RESERVE_ROOT)) {
578 f2fs_info(sbi, "Preserve previous reserve_root=%u",
579 F2FS_OPTION(sbi).root_reserved_blocks);
580 } else {
581 F2FS_OPTION(sbi).root_reserved_blocks = arg;
582 set_opt(sbi, RESERVE_ROOT);
584 break;
585 case Opt_resuid:
586 if (args->from && match_int(args, &arg))
587 return -EINVAL;
588 uid = make_kuid(current_user_ns(), arg);
589 if (!uid_valid(uid)) {
590 f2fs_err(sbi, "Invalid uid value %d", arg);
591 return -EINVAL;
593 F2FS_OPTION(sbi).s_resuid = uid;
594 break;
595 case Opt_resgid:
596 if (args->from && match_int(args, &arg))
597 return -EINVAL;
598 gid = make_kgid(current_user_ns(), arg);
599 if (!gid_valid(gid)) {
600 f2fs_err(sbi, "Invalid gid value %d", arg);
601 return -EINVAL;
603 F2FS_OPTION(sbi).s_resgid = gid;
604 break;
605 case Opt_mode:
606 name = match_strdup(&args[0]);
608 if (!name)
609 return -ENOMEM;
610 if (strlen(name) == 8 &&
611 !strncmp(name, "adaptive", 8)) {
612 if (f2fs_sb_has_blkzoned(sbi)) {
613 f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
614 kvfree(name);
615 return -EINVAL;
617 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
618 } else if (strlen(name) == 3 &&
619 !strncmp(name, "lfs", 3)) {
620 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
621 } else {
622 kvfree(name);
623 return -EINVAL;
625 kvfree(name);
626 break;
627 case Opt_io_size_bits:
628 if (args->from && match_int(args, &arg))
629 return -EINVAL;
630 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_PAGES)) {
631 f2fs_warn(sbi, "Not support %d, larger than %d",
632 1 << arg, BIO_MAX_PAGES);
633 return -EINVAL;
635 F2FS_OPTION(sbi).write_io_size_bits = arg;
636 break;
637 #ifdef CONFIG_F2FS_FAULT_INJECTION
638 case Opt_fault_injection:
639 if (args->from && match_int(args, &arg))
640 return -EINVAL;
641 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
642 set_opt(sbi, FAULT_INJECTION);
643 break;
645 case Opt_fault_type:
646 if (args->from && match_int(args, &arg))
647 return -EINVAL;
648 f2fs_build_fault_attr(sbi, 0, arg);
649 set_opt(sbi, FAULT_INJECTION);
650 break;
651 #else
652 case Opt_fault_injection:
653 f2fs_info(sbi, "fault_injection options not supported");
654 break;
656 case Opt_fault_type:
657 f2fs_info(sbi, "fault_type options not supported");
658 break;
659 #endif
660 case Opt_lazytime:
661 sb->s_flags |= SB_LAZYTIME;
662 break;
663 case Opt_nolazytime:
664 sb->s_flags &= ~SB_LAZYTIME;
665 break;
666 #ifdef CONFIG_QUOTA
667 case Opt_quota:
668 case Opt_usrquota:
669 set_opt(sbi, USRQUOTA);
670 break;
671 case Opt_grpquota:
672 set_opt(sbi, GRPQUOTA);
673 break;
674 case Opt_prjquota:
675 set_opt(sbi, PRJQUOTA);
676 break;
677 case Opt_usrjquota:
678 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
679 if (ret)
680 return ret;
681 break;
682 case Opt_grpjquota:
683 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
684 if (ret)
685 return ret;
686 break;
687 case Opt_prjjquota:
688 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
689 if (ret)
690 return ret;
691 break;
692 case Opt_offusrjquota:
693 ret = f2fs_clear_qf_name(sb, USRQUOTA);
694 if (ret)
695 return ret;
696 break;
697 case Opt_offgrpjquota:
698 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
699 if (ret)
700 return ret;
701 break;
702 case Opt_offprjjquota:
703 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
704 if (ret)
705 return ret;
706 break;
707 case Opt_jqfmt_vfsold:
708 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
709 break;
710 case Opt_jqfmt_vfsv0:
711 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
712 break;
713 case Opt_jqfmt_vfsv1:
714 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
715 break;
716 case Opt_noquota:
717 clear_opt(sbi, QUOTA);
718 clear_opt(sbi, USRQUOTA);
719 clear_opt(sbi, GRPQUOTA);
720 clear_opt(sbi, PRJQUOTA);
721 break;
722 #else
723 case Opt_quota:
724 case Opt_usrquota:
725 case Opt_grpquota:
726 case Opt_prjquota:
727 case Opt_usrjquota:
728 case Opt_grpjquota:
729 case Opt_prjjquota:
730 case Opt_offusrjquota:
731 case Opt_offgrpjquota:
732 case Opt_offprjjquota:
733 case Opt_jqfmt_vfsold:
734 case Opt_jqfmt_vfsv0:
735 case Opt_jqfmt_vfsv1:
736 case Opt_noquota:
737 f2fs_info(sbi, "quota operations not supported");
738 break;
739 #endif
740 case Opt_whint:
741 name = match_strdup(&args[0]);
742 if (!name)
743 return -ENOMEM;
744 if (strlen(name) == 10 &&
745 !strncmp(name, "user-based", 10)) {
746 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
747 } else if (strlen(name) == 3 &&
748 !strncmp(name, "off", 3)) {
749 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
750 } else if (strlen(name) == 8 &&
751 !strncmp(name, "fs-based", 8)) {
752 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
753 } else {
754 kvfree(name);
755 return -EINVAL;
757 kvfree(name);
758 break;
759 case Opt_alloc:
760 name = match_strdup(&args[0]);
761 if (!name)
762 return -ENOMEM;
764 if (strlen(name) == 7 &&
765 !strncmp(name, "default", 7)) {
766 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
767 } else if (strlen(name) == 5 &&
768 !strncmp(name, "reuse", 5)) {
769 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
770 } else {
771 kvfree(name);
772 return -EINVAL;
774 kvfree(name);
775 break;
776 case Opt_fsync:
777 name = match_strdup(&args[0]);
778 if (!name)
779 return -ENOMEM;
780 if (strlen(name) == 5 &&
781 !strncmp(name, "posix", 5)) {
782 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
783 } else if (strlen(name) == 6 &&
784 !strncmp(name, "strict", 6)) {
785 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
786 } else if (strlen(name) == 9 &&
787 !strncmp(name, "nobarrier", 9)) {
788 F2FS_OPTION(sbi).fsync_mode =
789 FSYNC_MODE_NOBARRIER;
790 } else {
791 kvfree(name);
792 return -EINVAL;
794 kvfree(name);
795 break;
796 case Opt_test_dummy_encryption:
797 #ifdef CONFIG_FS_ENCRYPTION
798 if (!f2fs_sb_has_encrypt(sbi)) {
799 f2fs_err(sbi, "Encrypt feature is off");
800 return -EINVAL;
803 F2FS_OPTION(sbi).test_dummy_encryption = true;
804 f2fs_info(sbi, "Test dummy encryption mode enabled");
805 #else
806 f2fs_info(sbi, "Test dummy encryption mount option ignored");
807 #endif
808 break;
809 case Opt_checkpoint_disable_cap_perc:
810 if (args->from && match_int(args, &arg))
811 return -EINVAL;
812 if (arg < 0 || arg > 100)
813 return -EINVAL;
814 F2FS_OPTION(sbi).unusable_cap_perc = arg;
815 set_opt(sbi, DISABLE_CHECKPOINT);
816 break;
817 case Opt_checkpoint_disable_cap:
818 if (args->from && match_int(args, &arg))
819 return -EINVAL;
820 F2FS_OPTION(sbi).unusable_cap = arg;
821 set_opt(sbi, DISABLE_CHECKPOINT);
822 break;
823 case Opt_checkpoint_disable:
824 set_opt(sbi, DISABLE_CHECKPOINT);
825 break;
826 case Opt_checkpoint_enable:
827 clear_opt(sbi, DISABLE_CHECKPOINT);
828 break;
829 case Opt_compress_algorithm:
830 if (!f2fs_sb_has_compression(sbi)) {
831 f2fs_err(sbi, "Compression feature if off");
832 return -EINVAL;
834 name = match_strdup(&args[0]);
835 if (!name)
836 return -ENOMEM;
837 if (strlen(name) == 3 && !strcmp(name, "lzo")) {
838 F2FS_OPTION(sbi).compress_algorithm =
839 COMPRESS_LZO;
840 } else if (strlen(name) == 3 &&
841 !strcmp(name, "lz4")) {
842 F2FS_OPTION(sbi).compress_algorithm =
843 COMPRESS_LZ4;
844 } else if (strlen(name) == 4 &&
845 !strcmp(name, "zstd")) {
846 F2FS_OPTION(sbi).compress_algorithm =
847 COMPRESS_ZSTD;
848 } else {
849 kfree(name);
850 return -EINVAL;
852 kfree(name);
853 break;
854 case Opt_compress_log_size:
855 if (!f2fs_sb_has_compression(sbi)) {
856 f2fs_err(sbi, "Compression feature is off");
857 return -EINVAL;
859 if (args->from && match_int(args, &arg))
860 return -EINVAL;
861 if (arg < MIN_COMPRESS_LOG_SIZE ||
862 arg > MAX_COMPRESS_LOG_SIZE) {
863 f2fs_err(sbi,
864 "Compress cluster log size is out of range");
865 return -EINVAL;
867 F2FS_OPTION(sbi).compress_log_size = arg;
868 break;
869 case Opt_compress_extension:
870 if (!f2fs_sb_has_compression(sbi)) {
871 f2fs_err(sbi, "Compression feature is off");
872 return -EINVAL;
874 name = match_strdup(&args[0]);
875 if (!name)
876 return -ENOMEM;
878 ext = F2FS_OPTION(sbi).extensions;
879 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
881 if (strlen(name) >= F2FS_EXTENSION_LEN ||
882 ext_cnt >= COMPRESS_EXT_NUM) {
883 f2fs_err(sbi,
884 "invalid extension length/number");
885 kfree(name);
886 return -EINVAL;
889 strcpy(ext[ext_cnt], name);
890 F2FS_OPTION(sbi).compress_ext_cnt++;
891 kfree(name);
892 break;
893 default:
894 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
896 return -EINVAL;
899 #ifdef CONFIG_QUOTA
900 if (f2fs_check_quota_options(sbi))
901 return -EINVAL;
902 #else
903 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
904 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
905 return -EINVAL;
907 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
908 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
909 return -EINVAL;
911 #endif
912 #ifndef CONFIG_UNICODE
913 if (f2fs_sb_has_casefold(sbi)) {
914 f2fs_err(sbi,
915 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
916 return -EINVAL;
918 #endif
920 if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) {
921 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
922 F2FS_IO_SIZE_KB(sbi));
923 return -EINVAL;
926 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
927 int min_size, max_size;
929 if (!f2fs_sb_has_extra_attr(sbi) ||
930 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
931 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
932 return -EINVAL;
934 if (!test_opt(sbi, INLINE_XATTR)) {
935 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
936 return -EINVAL;
939 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
940 max_size = MAX_INLINE_XATTR_SIZE;
942 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
943 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
944 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
945 min_size, max_size);
946 return -EINVAL;
950 if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) {
951 f2fs_err(sbi, "LFS not compatible with checkpoint=disable\n");
952 return -EINVAL;
955 /* Not pass down write hints if the number of active logs is lesser
956 * than NR_CURSEG_TYPE.
958 if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
959 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
960 return 0;
963 static struct inode *f2fs_alloc_inode(struct super_block *sb)
965 struct f2fs_inode_info *fi;
967 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
968 if (!fi)
969 return NULL;
971 init_once((void *) fi);
973 /* Initialize f2fs-specific inode info */
974 atomic_set(&fi->dirty_pages, 0);
975 init_rwsem(&fi->i_sem);
976 spin_lock_init(&fi->i_size_lock);
977 INIT_LIST_HEAD(&fi->dirty_list);
978 INIT_LIST_HEAD(&fi->gdirty_list);
979 INIT_LIST_HEAD(&fi->inmem_ilist);
980 INIT_LIST_HEAD(&fi->inmem_pages);
981 mutex_init(&fi->inmem_lock);
982 init_rwsem(&fi->i_gc_rwsem[READ]);
983 init_rwsem(&fi->i_gc_rwsem[WRITE]);
984 init_rwsem(&fi->i_mmap_sem);
985 init_rwsem(&fi->i_xattr_sem);
987 /* Will be used by directory only */
988 fi->i_dir_level = F2FS_SB(sb)->dir_level;
990 return &fi->vfs_inode;
993 static int f2fs_drop_inode(struct inode *inode)
995 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
996 int ret;
999 * during filesystem shutdown, if checkpoint is disabled,
1000 * drop useless meta/node dirty pages.
1002 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1003 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1004 inode->i_ino == F2FS_META_INO(sbi)) {
1005 trace_f2fs_drop_inode(inode, 1);
1006 return 1;
1011 * This is to avoid a deadlock condition like below.
1012 * writeback_single_inode(inode)
1013 * - f2fs_write_data_page
1014 * - f2fs_gc -> iput -> evict
1015 * - inode_wait_for_writeback(inode)
1017 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1018 if (!inode->i_nlink && !is_bad_inode(inode)) {
1019 /* to avoid evict_inode call simultaneously */
1020 atomic_inc(&inode->i_count);
1021 spin_unlock(&inode->i_lock);
1023 /* some remained atomic pages should discarded */
1024 if (f2fs_is_atomic_file(inode))
1025 f2fs_drop_inmem_pages(inode);
1027 /* should remain fi->extent_tree for writepage */
1028 f2fs_destroy_extent_node(inode);
1030 sb_start_intwrite(inode->i_sb);
1031 f2fs_i_size_write(inode, 0);
1033 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1034 inode, NULL, 0, DATA);
1035 truncate_inode_pages_final(inode->i_mapping);
1037 if (F2FS_HAS_BLOCKS(inode))
1038 f2fs_truncate(inode);
1040 sb_end_intwrite(inode->i_sb);
1042 spin_lock(&inode->i_lock);
1043 atomic_dec(&inode->i_count);
1045 trace_f2fs_drop_inode(inode, 0);
1046 return 0;
1048 ret = generic_drop_inode(inode);
1049 if (!ret)
1050 ret = fscrypt_drop_inode(inode);
1051 trace_f2fs_drop_inode(inode, ret);
1052 return ret;
1055 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1057 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1058 int ret = 0;
1060 spin_lock(&sbi->inode_lock[DIRTY_META]);
1061 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1062 ret = 1;
1063 } else {
1064 set_inode_flag(inode, FI_DIRTY_INODE);
1065 stat_inc_dirty_inode(sbi, DIRTY_META);
1067 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1068 list_add_tail(&F2FS_I(inode)->gdirty_list,
1069 &sbi->inode_list[DIRTY_META]);
1070 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1072 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1073 return ret;
1076 void f2fs_inode_synced(struct inode *inode)
1078 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1080 spin_lock(&sbi->inode_lock[DIRTY_META]);
1081 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1082 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1083 return;
1085 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1086 list_del_init(&F2FS_I(inode)->gdirty_list);
1087 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1089 clear_inode_flag(inode, FI_DIRTY_INODE);
1090 clear_inode_flag(inode, FI_AUTO_RECOVER);
1091 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1092 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1096 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1098 * We should call set_dirty_inode to write the dirty inode through write_inode.
1100 static void f2fs_dirty_inode(struct inode *inode, int flags)
1102 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1104 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1105 inode->i_ino == F2FS_META_INO(sbi))
1106 return;
1108 if (flags == I_DIRTY_TIME)
1109 return;
1111 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1112 clear_inode_flag(inode, FI_AUTO_RECOVER);
1114 f2fs_inode_dirtied(inode, false);
1117 static void f2fs_free_inode(struct inode *inode)
1119 fscrypt_free_inode(inode);
1120 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1123 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1125 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1126 percpu_counter_destroy(&sbi->total_valid_inode_count);
1129 static void destroy_device_list(struct f2fs_sb_info *sbi)
1131 int i;
1133 for (i = 0; i < sbi->s_ndevs; i++) {
1134 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1135 #ifdef CONFIG_BLK_DEV_ZONED
1136 kvfree(FDEV(i).blkz_seq);
1137 #endif
1139 kvfree(sbi->devs);
1142 static void f2fs_put_super(struct super_block *sb)
1144 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1145 int i;
1146 bool dropped;
1148 f2fs_quota_off_umount(sb);
1150 /* prevent remaining shrinker jobs */
1151 mutex_lock(&sbi->umount_mutex);
1154 * We don't need to do checkpoint when superblock is clean.
1155 * But, the previous checkpoint was not done by umount, it needs to do
1156 * clean checkpoint again.
1158 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1159 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1160 struct cp_control cpc = {
1161 .reason = CP_UMOUNT,
1163 f2fs_write_checkpoint(sbi, &cpc);
1166 /* be sure to wait for any on-going discard commands */
1167 dropped = f2fs_issue_discard_timeout(sbi);
1169 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1170 !sbi->discard_blks && !dropped) {
1171 struct cp_control cpc = {
1172 .reason = CP_UMOUNT | CP_TRIMMED,
1174 f2fs_write_checkpoint(sbi, &cpc);
1178 * normally superblock is clean, so we need to release this.
1179 * In addition, EIO will skip do checkpoint, we need this as well.
1181 f2fs_release_ino_entry(sbi, true);
1183 f2fs_leave_shrinker(sbi);
1184 mutex_unlock(&sbi->umount_mutex);
1186 /* our cp_error case, we can wait for any writeback page */
1187 f2fs_flush_merged_writes(sbi);
1189 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1191 f2fs_bug_on(sbi, sbi->fsync_node_num);
1193 iput(sbi->node_inode);
1194 sbi->node_inode = NULL;
1196 iput(sbi->meta_inode);
1197 sbi->meta_inode = NULL;
1200 * iput() can update stat information, if f2fs_write_checkpoint()
1201 * above failed with error.
1203 f2fs_destroy_stats(sbi);
1205 /* destroy f2fs internal modules */
1206 f2fs_destroy_node_manager(sbi);
1207 f2fs_destroy_segment_manager(sbi);
1209 f2fs_destroy_post_read_wq(sbi);
1211 kvfree(sbi->ckpt);
1213 f2fs_unregister_sysfs(sbi);
1215 sb->s_fs_info = NULL;
1216 if (sbi->s_chksum_driver)
1217 crypto_free_shash(sbi->s_chksum_driver);
1218 kvfree(sbi->raw_super);
1220 destroy_device_list(sbi);
1221 f2fs_destroy_xattr_caches(sbi);
1222 mempool_destroy(sbi->write_io_dummy);
1223 #ifdef CONFIG_QUOTA
1224 for (i = 0; i < MAXQUOTAS; i++)
1225 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1226 #endif
1227 destroy_percpu_info(sbi);
1228 for (i = 0; i < NR_PAGE_TYPE; i++)
1229 kvfree(sbi->write_io[i]);
1230 #ifdef CONFIG_UNICODE
1231 utf8_unload(sbi->s_encoding);
1232 #endif
1233 kvfree(sbi);
1236 int f2fs_sync_fs(struct super_block *sb, int sync)
1238 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1239 int err = 0;
1241 if (unlikely(f2fs_cp_error(sbi)))
1242 return 0;
1243 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1244 return 0;
1246 trace_f2fs_sync_fs(sb, sync);
1248 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1249 return -EAGAIN;
1251 if (sync) {
1252 struct cp_control cpc;
1254 cpc.reason = __get_cp_reason(sbi);
1256 down_write(&sbi->gc_lock);
1257 err = f2fs_write_checkpoint(sbi, &cpc);
1258 up_write(&sbi->gc_lock);
1260 f2fs_trace_ios(NULL, 1);
1262 return err;
1265 static int f2fs_freeze(struct super_block *sb)
1267 if (f2fs_readonly(sb))
1268 return 0;
1270 /* IO error happened before */
1271 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1272 return -EIO;
1274 /* must be clean, since sync_filesystem() was already called */
1275 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1276 return -EINVAL;
1277 return 0;
1280 static int f2fs_unfreeze(struct super_block *sb)
1282 return 0;
1285 #ifdef CONFIG_QUOTA
1286 static int f2fs_statfs_project(struct super_block *sb,
1287 kprojid_t projid, struct kstatfs *buf)
1289 struct kqid qid;
1290 struct dquot *dquot;
1291 u64 limit;
1292 u64 curblock;
1294 qid = make_kqid_projid(projid);
1295 dquot = dqget(sb, qid);
1296 if (IS_ERR(dquot))
1297 return PTR_ERR(dquot);
1298 spin_lock(&dquot->dq_dqb_lock);
1300 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1301 dquot->dq_dqb.dqb_bhardlimit);
1302 if (limit)
1303 limit >>= sb->s_blocksize_bits;
1305 if (limit && buf->f_blocks > limit) {
1306 curblock = (dquot->dq_dqb.dqb_curspace +
1307 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1308 buf->f_blocks = limit;
1309 buf->f_bfree = buf->f_bavail =
1310 (buf->f_blocks > curblock) ?
1311 (buf->f_blocks - curblock) : 0;
1314 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1315 dquot->dq_dqb.dqb_ihardlimit);
1317 if (limit && buf->f_files > limit) {
1318 buf->f_files = limit;
1319 buf->f_ffree =
1320 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1321 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1324 spin_unlock(&dquot->dq_dqb_lock);
1325 dqput(dquot);
1326 return 0;
1328 #endif
1330 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1332 struct super_block *sb = dentry->d_sb;
1333 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1334 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1335 block_t total_count, user_block_count, start_count;
1336 u64 avail_node_count;
1338 total_count = le64_to_cpu(sbi->raw_super->block_count);
1339 user_block_count = sbi->user_block_count;
1340 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1341 buf->f_type = F2FS_SUPER_MAGIC;
1342 buf->f_bsize = sbi->blocksize;
1344 buf->f_blocks = total_count - start_count;
1345 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1346 sbi->current_reserved_blocks;
1348 spin_lock(&sbi->stat_lock);
1349 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1350 buf->f_bfree = 0;
1351 else
1352 buf->f_bfree -= sbi->unusable_block_count;
1353 spin_unlock(&sbi->stat_lock);
1355 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1356 buf->f_bavail = buf->f_bfree -
1357 F2FS_OPTION(sbi).root_reserved_blocks;
1358 else
1359 buf->f_bavail = 0;
1361 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1363 if (avail_node_count > user_block_count) {
1364 buf->f_files = user_block_count;
1365 buf->f_ffree = buf->f_bavail;
1366 } else {
1367 buf->f_files = avail_node_count;
1368 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1369 buf->f_bavail);
1372 buf->f_namelen = F2FS_NAME_LEN;
1373 buf->f_fsid.val[0] = (u32)id;
1374 buf->f_fsid.val[1] = (u32)(id >> 32);
1376 #ifdef CONFIG_QUOTA
1377 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1378 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1379 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1381 #endif
1382 return 0;
1385 static inline void f2fs_show_quota_options(struct seq_file *seq,
1386 struct super_block *sb)
1388 #ifdef CONFIG_QUOTA
1389 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1391 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1392 char *fmtname = "";
1394 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1395 case QFMT_VFS_OLD:
1396 fmtname = "vfsold";
1397 break;
1398 case QFMT_VFS_V0:
1399 fmtname = "vfsv0";
1400 break;
1401 case QFMT_VFS_V1:
1402 fmtname = "vfsv1";
1403 break;
1405 seq_printf(seq, ",jqfmt=%s", fmtname);
1408 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1409 seq_show_option(seq, "usrjquota",
1410 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1412 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1413 seq_show_option(seq, "grpjquota",
1414 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1416 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1417 seq_show_option(seq, "prjjquota",
1418 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1419 #endif
1422 static inline void f2fs_show_compress_options(struct seq_file *seq,
1423 struct super_block *sb)
1425 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1426 char *algtype = "";
1427 int i;
1429 if (!f2fs_sb_has_compression(sbi))
1430 return;
1432 switch (F2FS_OPTION(sbi).compress_algorithm) {
1433 case COMPRESS_LZO:
1434 algtype = "lzo";
1435 break;
1436 case COMPRESS_LZ4:
1437 algtype = "lz4";
1438 break;
1439 case COMPRESS_ZSTD:
1440 algtype = "zstd";
1441 break;
1443 seq_printf(seq, ",compress_algorithm=%s", algtype);
1445 seq_printf(seq, ",compress_log_size=%u",
1446 F2FS_OPTION(sbi).compress_log_size);
1448 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1449 seq_printf(seq, ",compress_extension=%s",
1450 F2FS_OPTION(sbi).extensions[i]);
1454 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1456 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1458 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1459 seq_printf(seq, ",background_gc=%s", "sync");
1460 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1461 seq_printf(seq, ",background_gc=%s", "on");
1462 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1463 seq_printf(seq, ",background_gc=%s", "off");
1465 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1466 seq_puts(seq, ",disable_roll_forward");
1467 if (test_opt(sbi, NORECOVERY))
1468 seq_puts(seq, ",norecovery");
1469 if (test_opt(sbi, DISCARD))
1470 seq_puts(seq, ",discard");
1471 else
1472 seq_puts(seq, ",nodiscard");
1473 if (test_opt(sbi, NOHEAP))
1474 seq_puts(seq, ",no_heap");
1475 else
1476 seq_puts(seq, ",heap");
1477 #ifdef CONFIG_F2FS_FS_XATTR
1478 if (test_opt(sbi, XATTR_USER))
1479 seq_puts(seq, ",user_xattr");
1480 else
1481 seq_puts(seq, ",nouser_xattr");
1482 if (test_opt(sbi, INLINE_XATTR))
1483 seq_puts(seq, ",inline_xattr");
1484 else
1485 seq_puts(seq, ",noinline_xattr");
1486 if (test_opt(sbi, INLINE_XATTR_SIZE))
1487 seq_printf(seq, ",inline_xattr_size=%u",
1488 F2FS_OPTION(sbi).inline_xattr_size);
1489 #endif
1490 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1491 if (test_opt(sbi, POSIX_ACL))
1492 seq_puts(seq, ",acl");
1493 else
1494 seq_puts(seq, ",noacl");
1495 #endif
1496 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1497 seq_puts(seq, ",disable_ext_identify");
1498 if (test_opt(sbi, INLINE_DATA))
1499 seq_puts(seq, ",inline_data");
1500 else
1501 seq_puts(seq, ",noinline_data");
1502 if (test_opt(sbi, INLINE_DENTRY))
1503 seq_puts(seq, ",inline_dentry");
1504 else
1505 seq_puts(seq, ",noinline_dentry");
1506 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1507 seq_puts(seq, ",flush_merge");
1508 if (test_opt(sbi, NOBARRIER))
1509 seq_puts(seq, ",nobarrier");
1510 if (test_opt(sbi, FASTBOOT))
1511 seq_puts(seq, ",fastboot");
1512 if (test_opt(sbi, EXTENT_CACHE))
1513 seq_puts(seq, ",extent_cache");
1514 else
1515 seq_puts(seq, ",noextent_cache");
1516 if (test_opt(sbi, DATA_FLUSH))
1517 seq_puts(seq, ",data_flush");
1519 seq_puts(seq, ",mode=");
1520 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
1521 seq_puts(seq, "adaptive");
1522 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
1523 seq_puts(seq, "lfs");
1524 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1525 if (test_opt(sbi, RESERVE_ROOT))
1526 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1527 F2FS_OPTION(sbi).root_reserved_blocks,
1528 from_kuid_munged(&init_user_ns,
1529 F2FS_OPTION(sbi).s_resuid),
1530 from_kgid_munged(&init_user_ns,
1531 F2FS_OPTION(sbi).s_resgid));
1532 if (F2FS_IO_SIZE_BITS(sbi))
1533 seq_printf(seq, ",io_bits=%u",
1534 F2FS_OPTION(sbi).write_io_size_bits);
1535 #ifdef CONFIG_F2FS_FAULT_INJECTION
1536 if (test_opt(sbi, FAULT_INJECTION)) {
1537 seq_printf(seq, ",fault_injection=%u",
1538 F2FS_OPTION(sbi).fault_info.inject_rate);
1539 seq_printf(seq, ",fault_type=%u",
1540 F2FS_OPTION(sbi).fault_info.inject_type);
1542 #endif
1543 #ifdef CONFIG_QUOTA
1544 if (test_opt(sbi, QUOTA))
1545 seq_puts(seq, ",quota");
1546 if (test_opt(sbi, USRQUOTA))
1547 seq_puts(seq, ",usrquota");
1548 if (test_opt(sbi, GRPQUOTA))
1549 seq_puts(seq, ",grpquota");
1550 if (test_opt(sbi, PRJQUOTA))
1551 seq_puts(seq, ",prjquota");
1552 #endif
1553 f2fs_show_quota_options(seq, sbi->sb);
1554 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1555 seq_printf(seq, ",whint_mode=%s", "user-based");
1556 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1557 seq_printf(seq, ",whint_mode=%s", "fs-based");
1558 #ifdef CONFIG_FS_ENCRYPTION
1559 if (F2FS_OPTION(sbi).test_dummy_encryption)
1560 seq_puts(seq, ",test_dummy_encryption");
1561 #endif
1563 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1564 seq_printf(seq, ",alloc_mode=%s", "default");
1565 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1566 seq_printf(seq, ",alloc_mode=%s", "reuse");
1568 if (test_opt(sbi, DISABLE_CHECKPOINT))
1569 seq_printf(seq, ",checkpoint=disable:%u",
1570 F2FS_OPTION(sbi).unusable_cap);
1571 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1572 seq_printf(seq, ",fsync_mode=%s", "posix");
1573 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1574 seq_printf(seq, ",fsync_mode=%s", "strict");
1575 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1576 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1578 f2fs_show_compress_options(seq, sbi->sb);
1579 return 0;
1582 static void default_options(struct f2fs_sb_info *sbi)
1584 /* init some FS parameters */
1585 F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1586 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1587 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1588 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1589 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1590 F2FS_OPTION(sbi).test_dummy_encryption = false;
1591 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1592 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1593 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
1594 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
1595 F2FS_OPTION(sbi).compress_ext_cnt = 0;
1596 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
1598 set_opt(sbi, INLINE_XATTR);
1599 set_opt(sbi, INLINE_DATA);
1600 set_opt(sbi, INLINE_DENTRY);
1601 set_opt(sbi, EXTENT_CACHE);
1602 set_opt(sbi, NOHEAP);
1603 clear_opt(sbi, DISABLE_CHECKPOINT);
1604 F2FS_OPTION(sbi).unusable_cap = 0;
1605 sbi->sb->s_flags |= SB_LAZYTIME;
1606 set_opt(sbi, FLUSH_MERGE);
1607 set_opt(sbi, DISCARD);
1608 if (f2fs_sb_has_blkzoned(sbi))
1609 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
1610 else
1611 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
1613 #ifdef CONFIG_F2FS_FS_XATTR
1614 set_opt(sbi, XATTR_USER);
1615 #endif
1616 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1617 set_opt(sbi, POSIX_ACL);
1618 #endif
1620 f2fs_build_fault_attr(sbi, 0, 0);
1623 #ifdef CONFIG_QUOTA
1624 static int f2fs_enable_quotas(struct super_block *sb);
1625 #endif
1627 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1629 unsigned int s_flags = sbi->sb->s_flags;
1630 struct cp_control cpc;
1631 int err = 0;
1632 int ret;
1633 block_t unusable;
1635 if (s_flags & SB_RDONLY) {
1636 f2fs_err(sbi, "checkpoint=disable on readonly fs");
1637 return -EINVAL;
1639 sbi->sb->s_flags |= SB_ACTIVE;
1641 f2fs_update_time(sbi, DISABLE_TIME);
1643 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1644 down_write(&sbi->gc_lock);
1645 err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1646 if (err == -ENODATA) {
1647 err = 0;
1648 break;
1650 if (err && err != -EAGAIN)
1651 break;
1654 ret = sync_filesystem(sbi->sb);
1655 if (ret || err) {
1656 err = ret ? ret: err;
1657 goto restore_flag;
1660 unusable = f2fs_get_unusable_blocks(sbi);
1661 if (f2fs_disable_cp_again(sbi, unusable)) {
1662 err = -EAGAIN;
1663 goto restore_flag;
1666 down_write(&sbi->gc_lock);
1667 cpc.reason = CP_PAUSE;
1668 set_sbi_flag(sbi, SBI_CP_DISABLED);
1669 err = f2fs_write_checkpoint(sbi, &cpc);
1670 if (err)
1671 goto out_unlock;
1673 spin_lock(&sbi->stat_lock);
1674 sbi->unusable_block_count = unusable;
1675 spin_unlock(&sbi->stat_lock);
1677 out_unlock:
1678 up_write(&sbi->gc_lock);
1679 restore_flag:
1680 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */
1681 return err;
1684 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1686 down_write(&sbi->gc_lock);
1687 f2fs_dirty_to_prefree(sbi);
1689 clear_sbi_flag(sbi, SBI_CP_DISABLED);
1690 set_sbi_flag(sbi, SBI_IS_DIRTY);
1691 up_write(&sbi->gc_lock);
1693 f2fs_sync_fs(sbi->sb, 1);
1696 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1698 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1699 struct f2fs_mount_info org_mount_opt;
1700 unsigned long old_sb_flags;
1701 int err;
1702 bool need_restart_gc = false;
1703 bool need_stop_gc = false;
1704 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1705 bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1706 bool no_io_align = !F2FS_IO_ALIGNED(sbi);
1707 bool checkpoint_changed;
1708 #ifdef CONFIG_QUOTA
1709 int i, j;
1710 #endif
1713 * Save the old mount options in case we
1714 * need to restore them.
1716 org_mount_opt = sbi->mount_opt;
1717 old_sb_flags = sb->s_flags;
1719 #ifdef CONFIG_QUOTA
1720 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1721 for (i = 0; i < MAXQUOTAS; i++) {
1722 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1723 org_mount_opt.s_qf_names[i] =
1724 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1725 GFP_KERNEL);
1726 if (!org_mount_opt.s_qf_names[i]) {
1727 for (j = 0; j < i; j++)
1728 kvfree(org_mount_opt.s_qf_names[j]);
1729 return -ENOMEM;
1731 } else {
1732 org_mount_opt.s_qf_names[i] = NULL;
1735 #endif
1737 /* recover superblocks we couldn't write due to previous RO mount */
1738 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1739 err = f2fs_commit_super(sbi, false);
1740 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
1741 err);
1742 if (!err)
1743 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1746 default_options(sbi);
1748 /* parse mount options */
1749 err = parse_options(sb, data);
1750 if (err)
1751 goto restore_opts;
1752 checkpoint_changed =
1753 disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1756 * Previous and new state of filesystem is RO,
1757 * so skip checking GC and FLUSH_MERGE conditions.
1759 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1760 goto skip;
1762 #ifdef CONFIG_QUOTA
1763 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1764 err = dquot_suspend(sb, -1);
1765 if (err < 0)
1766 goto restore_opts;
1767 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1768 /* dquot_resume needs RW */
1769 sb->s_flags &= ~SB_RDONLY;
1770 if (sb_any_quota_suspended(sb)) {
1771 dquot_resume(sb, -1);
1772 } else if (f2fs_sb_has_quota_ino(sbi)) {
1773 err = f2fs_enable_quotas(sb);
1774 if (err)
1775 goto restore_opts;
1778 #endif
1779 /* disallow enable/disable extent_cache dynamically */
1780 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1781 err = -EINVAL;
1782 f2fs_warn(sbi, "switch extent_cache option is not allowed");
1783 goto restore_opts;
1786 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
1787 err = -EINVAL;
1788 f2fs_warn(sbi, "switch io_bits option is not allowed");
1789 goto restore_opts;
1792 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1793 err = -EINVAL;
1794 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
1795 goto restore_opts;
1799 * We stop the GC thread if FS is mounted as RO
1800 * or if background_gc = off is passed in mount
1801 * option. Also sync the filesystem.
1803 if ((*flags & SB_RDONLY) ||
1804 F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) {
1805 if (sbi->gc_thread) {
1806 f2fs_stop_gc_thread(sbi);
1807 need_restart_gc = true;
1809 } else if (!sbi->gc_thread) {
1810 err = f2fs_start_gc_thread(sbi);
1811 if (err)
1812 goto restore_opts;
1813 need_stop_gc = true;
1816 if (*flags & SB_RDONLY ||
1817 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1818 writeback_inodes_sb(sb, WB_REASON_SYNC);
1819 sync_inodes_sb(sb);
1821 set_sbi_flag(sbi, SBI_IS_DIRTY);
1822 set_sbi_flag(sbi, SBI_IS_CLOSE);
1823 f2fs_sync_fs(sb, 1);
1824 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1827 if (checkpoint_changed) {
1828 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1829 err = f2fs_disable_checkpoint(sbi);
1830 if (err)
1831 goto restore_gc;
1832 } else {
1833 f2fs_enable_checkpoint(sbi);
1838 * We stop issue flush thread if FS is mounted as RO
1839 * or if flush_merge is not passed in mount option.
1841 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1842 clear_opt(sbi, FLUSH_MERGE);
1843 f2fs_destroy_flush_cmd_control(sbi, false);
1844 } else {
1845 err = f2fs_create_flush_cmd_control(sbi);
1846 if (err)
1847 goto restore_gc;
1849 skip:
1850 #ifdef CONFIG_QUOTA
1851 /* Release old quota file names */
1852 for (i = 0; i < MAXQUOTAS; i++)
1853 kvfree(org_mount_opt.s_qf_names[i]);
1854 #endif
1855 /* Update the POSIXACL Flag */
1856 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1857 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1859 limit_reserve_root(sbi);
1860 adjust_unusable_cap_perc(sbi);
1861 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1862 return 0;
1863 restore_gc:
1864 if (need_restart_gc) {
1865 if (f2fs_start_gc_thread(sbi))
1866 f2fs_warn(sbi, "background gc thread has stopped");
1867 } else if (need_stop_gc) {
1868 f2fs_stop_gc_thread(sbi);
1870 restore_opts:
1871 #ifdef CONFIG_QUOTA
1872 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1873 for (i = 0; i < MAXQUOTAS; i++) {
1874 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1875 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1877 #endif
1878 sbi->mount_opt = org_mount_opt;
1879 sb->s_flags = old_sb_flags;
1880 return err;
1883 #ifdef CONFIG_QUOTA
1884 /* Read data from quotafile */
1885 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1886 size_t len, loff_t off)
1888 struct inode *inode = sb_dqopt(sb)->files[type];
1889 struct address_space *mapping = inode->i_mapping;
1890 block_t blkidx = F2FS_BYTES_TO_BLK(off);
1891 int offset = off & (sb->s_blocksize - 1);
1892 int tocopy;
1893 size_t toread;
1894 loff_t i_size = i_size_read(inode);
1895 struct page *page;
1896 char *kaddr;
1898 if (off > i_size)
1899 return 0;
1901 if (off + len > i_size)
1902 len = i_size - off;
1903 toread = len;
1904 while (toread > 0) {
1905 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1906 repeat:
1907 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1908 if (IS_ERR(page)) {
1909 if (PTR_ERR(page) == -ENOMEM) {
1910 congestion_wait(BLK_RW_ASYNC,
1911 DEFAULT_IO_TIMEOUT);
1912 goto repeat;
1914 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1915 return PTR_ERR(page);
1918 lock_page(page);
1920 if (unlikely(page->mapping != mapping)) {
1921 f2fs_put_page(page, 1);
1922 goto repeat;
1924 if (unlikely(!PageUptodate(page))) {
1925 f2fs_put_page(page, 1);
1926 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1927 return -EIO;
1930 kaddr = kmap_atomic(page);
1931 memcpy(data, kaddr + offset, tocopy);
1932 kunmap_atomic(kaddr);
1933 f2fs_put_page(page, 1);
1935 offset = 0;
1936 toread -= tocopy;
1937 data += tocopy;
1938 blkidx++;
1940 return len;
1943 /* Write to quotafile */
1944 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1945 const char *data, size_t len, loff_t off)
1947 struct inode *inode = sb_dqopt(sb)->files[type];
1948 struct address_space *mapping = inode->i_mapping;
1949 const struct address_space_operations *a_ops = mapping->a_ops;
1950 int offset = off & (sb->s_blocksize - 1);
1951 size_t towrite = len;
1952 struct page *page;
1953 void *fsdata = NULL;
1954 char *kaddr;
1955 int err = 0;
1956 int tocopy;
1958 while (towrite > 0) {
1959 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1960 towrite);
1961 retry:
1962 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1963 &page, &fsdata);
1964 if (unlikely(err)) {
1965 if (err == -ENOMEM) {
1966 congestion_wait(BLK_RW_ASYNC,
1967 DEFAULT_IO_TIMEOUT);
1968 goto retry;
1970 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1971 break;
1974 kaddr = kmap_atomic(page);
1975 memcpy(kaddr + offset, data, tocopy);
1976 kunmap_atomic(kaddr);
1977 flush_dcache_page(page);
1979 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1980 page, fsdata);
1981 offset = 0;
1982 towrite -= tocopy;
1983 off += tocopy;
1984 data += tocopy;
1985 cond_resched();
1988 if (len == towrite)
1989 return err;
1990 inode->i_mtime = inode->i_ctime = current_time(inode);
1991 f2fs_mark_inode_dirty_sync(inode, false);
1992 return len - towrite;
1995 static struct dquot **f2fs_get_dquots(struct inode *inode)
1997 return F2FS_I(inode)->i_dquot;
2000 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2002 return &F2FS_I(inode)->i_reserved_quota;
2005 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2007 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2008 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2009 return 0;
2012 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2013 F2FS_OPTION(sbi).s_jquota_fmt, type);
2016 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2018 int enabled = 0;
2019 int i, err;
2021 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2022 err = f2fs_enable_quotas(sbi->sb);
2023 if (err) {
2024 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2025 return 0;
2027 return 1;
2030 for (i = 0; i < MAXQUOTAS; i++) {
2031 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2032 err = f2fs_quota_on_mount(sbi, i);
2033 if (!err) {
2034 enabled = 1;
2035 continue;
2037 f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2038 err, i);
2041 return enabled;
2044 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2045 unsigned int flags)
2047 struct inode *qf_inode;
2048 unsigned long qf_inum;
2049 int err;
2051 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2053 qf_inum = f2fs_qf_ino(sb, type);
2054 if (!qf_inum)
2055 return -EPERM;
2057 qf_inode = f2fs_iget(sb, qf_inum);
2058 if (IS_ERR(qf_inode)) {
2059 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2060 return PTR_ERR(qf_inode);
2063 /* Don't account quota for quota files to avoid recursion */
2064 qf_inode->i_flags |= S_NOQUOTA;
2065 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2066 iput(qf_inode);
2067 return err;
2070 static int f2fs_enable_quotas(struct super_block *sb)
2072 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2073 int type, err = 0;
2074 unsigned long qf_inum;
2075 bool quota_mopt[MAXQUOTAS] = {
2076 test_opt(sbi, USRQUOTA),
2077 test_opt(sbi, GRPQUOTA),
2078 test_opt(sbi, PRJQUOTA),
2081 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2082 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2083 return 0;
2086 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2088 for (type = 0; type < MAXQUOTAS; type++) {
2089 qf_inum = f2fs_qf_ino(sb, type);
2090 if (qf_inum) {
2091 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2092 DQUOT_USAGE_ENABLED |
2093 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2094 if (err) {
2095 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2096 type, err);
2097 for (type--; type >= 0; type--)
2098 dquot_quota_off(sb, type);
2099 set_sbi_flag(F2FS_SB(sb),
2100 SBI_QUOTA_NEED_REPAIR);
2101 return err;
2105 return 0;
2108 int f2fs_quota_sync(struct super_block *sb, int type)
2110 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2111 struct quota_info *dqopt = sb_dqopt(sb);
2112 int cnt;
2113 int ret;
2116 * do_quotactl
2117 * f2fs_quota_sync
2118 * down_read(quota_sem)
2119 * dquot_writeback_dquots()
2120 * f2fs_dquot_commit
2121 * block_operation
2122 * down_read(quota_sem)
2124 f2fs_lock_op(sbi);
2126 down_read(&sbi->quota_sem);
2127 ret = dquot_writeback_dquots(sb, type);
2128 if (ret)
2129 goto out;
2132 * Now when everything is written we can discard the pagecache so
2133 * that userspace sees the changes.
2135 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2136 struct address_space *mapping;
2138 if (type != -1 && cnt != type)
2139 continue;
2140 if (!sb_has_quota_active(sb, cnt))
2141 continue;
2143 mapping = dqopt->files[cnt]->i_mapping;
2145 ret = filemap_fdatawrite(mapping);
2146 if (ret)
2147 goto out;
2149 /* if we are using journalled quota */
2150 if (is_journalled_quota(sbi))
2151 continue;
2153 ret = filemap_fdatawait(mapping);
2154 if (ret)
2155 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2157 inode_lock(dqopt->files[cnt]);
2158 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
2159 inode_unlock(dqopt->files[cnt]);
2161 out:
2162 if (ret)
2163 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2164 up_read(&sbi->quota_sem);
2165 f2fs_unlock_op(sbi);
2166 return ret;
2169 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2170 const struct path *path)
2172 struct inode *inode;
2173 int err;
2175 /* if quota sysfile exists, deny enabling quota with specific file */
2176 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2177 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2178 return -EBUSY;
2181 err = f2fs_quota_sync(sb, type);
2182 if (err)
2183 return err;
2185 err = dquot_quota_on(sb, type, format_id, path);
2186 if (err)
2187 return err;
2189 inode = d_inode(path->dentry);
2191 inode_lock(inode);
2192 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2193 f2fs_set_inode_flags(inode);
2194 inode_unlock(inode);
2195 f2fs_mark_inode_dirty_sync(inode, false);
2197 return 0;
2200 static int __f2fs_quota_off(struct super_block *sb, int type)
2202 struct inode *inode = sb_dqopt(sb)->files[type];
2203 int err;
2205 if (!inode || !igrab(inode))
2206 return dquot_quota_off(sb, type);
2208 err = f2fs_quota_sync(sb, type);
2209 if (err)
2210 goto out_put;
2212 err = dquot_quota_off(sb, type);
2213 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2214 goto out_put;
2216 inode_lock(inode);
2217 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2218 f2fs_set_inode_flags(inode);
2219 inode_unlock(inode);
2220 f2fs_mark_inode_dirty_sync(inode, false);
2221 out_put:
2222 iput(inode);
2223 return err;
2226 static int f2fs_quota_off(struct super_block *sb, int type)
2228 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2229 int err;
2231 err = __f2fs_quota_off(sb, type);
2234 * quotactl can shutdown journalled quota, result in inconsistence
2235 * between quota record and fs data by following updates, tag the
2236 * flag to let fsck be aware of it.
2238 if (is_journalled_quota(sbi))
2239 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2240 return err;
2243 void f2fs_quota_off_umount(struct super_block *sb)
2245 int type;
2246 int err;
2248 for (type = 0; type < MAXQUOTAS; type++) {
2249 err = __f2fs_quota_off(sb, type);
2250 if (err) {
2251 int ret = dquot_quota_off(sb, type);
2253 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2254 type, err, ret);
2255 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2259 * In case of checkpoint=disable, we must flush quota blocks.
2260 * This can cause NULL exception for node_inode in end_io, since
2261 * put_super already dropped it.
2263 sync_filesystem(sb);
2266 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2268 struct quota_info *dqopt = sb_dqopt(sb);
2269 int type;
2271 for (type = 0; type < MAXQUOTAS; type++) {
2272 if (!dqopt->files[type])
2273 continue;
2274 f2fs_inode_synced(dqopt->files[type]);
2278 static int f2fs_dquot_commit(struct dquot *dquot)
2280 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2281 int ret;
2283 down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
2284 ret = dquot_commit(dquot);
2285 if (ret < 0)
2286 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2287 up_read(&sbi->quota_sem);
2288 return ret;
2291 static int f2fs_dquot_acquire(struct dquot *dquot)
2293 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2294 int ret;
2296 down_read(&sbi->quota_sem);
2297 ret = dquot_acquire(dquot);
2298 if (ret < 0)
2299 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2300 up_read(&sbi->quota_sem);
2301 return ret;
2304 static int f2fs_dquot_release(struct dquot *dquot)
2306 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2307 int ret = dquot_release(dquot);
2309 if (ret < 0)
2310 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2311 return ret;
2314 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2316 struct super_block *sb = dquot->dq_sb;
2317 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2318 int ret = dquot_mark_dquot_dirty(dquot);
2320 /* if we are using journalled quota */
2321 if (is_journalled_quota(sbi))
2322 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2324 return ret;
2327 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2329 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2330 int ret = dquot_commit_info(sb, type);
2332 if (ret < 0)
2333 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2334 return ret;
2337 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2339 *projid = F2FS_I(inode)->i_projid;
2340 return 0;
2343 static const struct dquot_operations f2fs_quota_operations = {
2344 .get_reserved_space = f2fs_get_reserved_space,
2345 .write_dquot = f2fs_dquot_commit,
2346 .acquire_dquot = f2fs_dquot_acquire,
2347 .release_dquot = f2fs_dquot_release,
2348 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
2349 .write_info = f2fs_dquot_commit_info,
2350 .alloc_dquot = dquot_alloc,
2351 .destroy_dquot = dquot_destroy,
2352 .get_projid = f2fs_get_projid,
2353 .get_next_id = dquot_get_next_id,
2356 static const struct quotactl_ops f2fs_quotactl_ops = {
2357 .quota_on = f2fs_quota_on,
2358 .quota_off = f2fs_quota_off,
2359 .quota_sync = f2fs_quota_sync,
2360 .get_state = dquot_get_state,
2361 .set_info = dquot_set_dqinfo,
2362 .get_dqblk = dquot_get_dqblk,
2363 .set_dqblk = dquot_set_dqblk,
2364 .get_nextdqblk = dquot_get_next_dqblk,
2366 #else
2367 int f2fs_quota_sync(struct super_block *sb, int type)
2369 return 0;
2372 void f2fs_quota_off_umount(struct super_block *sb)
2375 #endif
2377 static const struct super_operations f2fs_sops = {
2378 .alloc_inode = f2fs_alloc_inode,
2379 .free_inode = f2fs_free_inode,
2380 .drop_inode = f2fs_drop_inode,
2381 .write_inode = f2fs_write_inode,
2382 .dirty_inode = f2fs_dirty_inode,
2383 .show_options = f2fs_show_options,
2384 #ifdef CONFIG_QUOTA
2385 .quota_read = f2fs_quota_read,
2386 .quota_write = f2fs_quota_write,
2387 .get_dquots = f2fs_get_dquots,
2388 #endif
2389 .evict_inode = f2fs_evict_inode,
2390 .put_super = f2fs_put_super,
2391 .sync_fs = f2fs_sync_fs,
2392 .freeze_fs = f2fs_freeze,
2393 .unfreeze_fs = f2fs_unfreeze,
2394 .statfs = f2fs_statfs,
2395 .remount_fs = f2fs_remount,
2398 #ifdef CONFIG_FS_ENCRYPTION
2399 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2401 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2402 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2403 ctx, len, NULL);
2406 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2407 void *fs_data)
2409 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2412 * Encrypting the root directory is not allowed because fsck
2413 * expects lost+found directory to exist and remain unencrypted
2414 * if LOST_FOUND feature is enabled.
2417 if (f2fs_sb_has_lost_found(sbi) &&
2418 inode->i_ino == F2FS_ROOT_INO(sbi))
2419 return -EPERM;
2421 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2422 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2423 ctx, len, fs_data, XATTR_CREATE);
2426 static bool f2fs_dummy_context(struct inode *inode)
2428 return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
2431 static bool f2fs_has_stable_inodes(struct super_block *sb)
2433 return true;
2436 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
2437 int *ino_bits_ret, int *lblk_bits_ret)
2439 *ino_bits_ret = 8 * sizeof(nid_t);
2440 *lblk_bits_ret = 8 * sizeof(block_t);
2443 static const struct fscrypt_operations f2fs_cryptops = {
2444 .key_prefix = "f2fs:",
2445 .get_context = f2fs_get_context,
2446 .set_context = f2fs_set_context,
2447 .dummy_context = f2fs_dummy_context,
2448 .empty_dir = f2fs_empty_dir,
2449 .max_namelen = F2FS_NAME_LEN,
2450 .has_stable_inodes = f2fs_has_stable_inodes,
2451 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits,
2453 #endif
2455 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2456 u64 ino, u32 generation)
2458 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2459 struct inode *inode;
2461 if (f2fs_check_nid_range(sbi, ino))
2462 return ERR_PTR(-ESTALE);
2465 * f2fs_iget isn't quite right if the inode is currently unallocated!
2466 * However f2fs_iget currently does appropriate checks to handle stale
2467 * inodes so everything is OK.
2469 inode = f2fs_iget(sb, ino);
2470 if (IS_ERR(inode))
2471 return ERR_CAST(inode);
2472 if (unlikely(generation && inode->i_generation != generation)) {
2473 /* we didn't find the right inode.. */
2474 iput(inode);
2475 return ERR_PTR(-ESTALE);
2477 return inode;
2480 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2481 int fh_len, int fh_type)
2483 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2484 f2fs_nfs_get_inode);
2487 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2488 int fh_len, int fh_type)
2490 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2491 f2fs_nfs_get_inode);
2494 static const struct export_operations f2fs_export_ops = {
2495 .fh_to_dentry = f2fs_fh_to_dentry,
2496 .fh_to_parent = f2fs_fh_to_parent,
2497 .get_parent = f2fs_get_parent,
2500 static loff_t max_file_blocks(void)
2502 loff_t result = 0;
2503 loff_t leaf_count = DEF_ADDRS_PER_BLOCK;
2506 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2507 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2508 * space in inode.i_addr, it will be more safe to reassign
2509 * result as zero.
2512 /* two direct node blocks */
2513 result += (leaf_count * 2);
2515 /* two indirect node blocks */
2516 leaf_count *= NIDS_PER_BLOCK;
2517 result += (leaf_count * 2);
2519 /* one double indirect node block */
2520 leaf_count *= NIDS_PER_BLOCK;
2521 result += leaf_count;
2523 return result;
2526 static int __f2fs_commit_super(struct buffer_head *bh,
2527 struct f2fs_super_block *super)
2529 lock_buffer(bh);
2530 if (super)
2531 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2532 set_buffer_dirty(bh);
2533 unlock_buffer(bh);
2535 /* it's rare case, we can do fua all the time */
2536 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2539 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2540 struct buffer_head *bh)
2542 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2543 (bh->b_data + F2FS_SUPER_OFFSET);
2544 struct super_block *sb = sbi->sb;
2545 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2546 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2547 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2548 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2549 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2550 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2551 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2552 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2553 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2554 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2555 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2556 u32 segment_count = le32_to_cpu(raw_super->segment_count);
2557 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2558 u64 main_end_blkaddr = main_blkaddr +
2559 (segment_count_main << log_blocks_per_seg);
2560 u64 seg_end_blkaddr = segment0_blkaddr +
2561 (segment_count << log_blocks_per_seg);
2563 if (segment0_blkaddr != cp_blkaddr) {
2564 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2565 segment0_blkaddr, cp_blkaddr);
2566 return true;
2569 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2570 sit_blkaddr) {
2571 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2572 cp_blkaddr, sit_blkaddr,
2573 segment_count_ckpt << log_blocks_per_seg);
2574 return true;
2577 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2578 nat_blkaddr) {
2579 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2580 sit_blkaddr, nat_blkaddr,
2581 segment_count_sit << log_blocks_per_seg);
2582 return true;
2585 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2586 ssa_blkaddr) {
2587 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2588 nat_blkaddr, ssa_blkaddr,
2589 segment_count_nat << log_blocks_per_seg);
2590 return true;
2593 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2594 main_blkaddr) {
2595 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2596 ssa_blkaddr, main_blkaddr,
2597 segment_count_ssa << log_blocks_per_seg);
2598 return true;
2601 if (main_end_blkaddr > seg_end_blkaddr) {
2602 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2603 main_blkaddr,
2604 segment0_blkaddr +
2605 (segment_count << log_blocks_per_seg),
2606 segment_count_main << log_blocks_per_seg);
2607 return true;
2608 } else if (main_end_blkaddr < seg_end_blkaddr) {
2609 int err = 0;
2610 char *res;
2612 /* fix in-memory information all the time */
2613 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2614 segment0_blkaddr) >> log_blocks_per_seg);
2616 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2617 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2618 res = "internally";
2619 } else {
2620 err = __f2fs_commit_super(bh, NULL);
2621 res = err ? "failed" : "done";
2623 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%u) block(%u)",
2624 res, main_blkaddr,
2625 segment0_blkaddr +
2626 (segment_count << log_blocks_per_seg),
2627 segment_count_main << log_blocks_per_seg);
2628 if (err)
2629 return true;
2631 return false;
2634 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2635 struct buffer_head *bh)
2637 block_t segment_count, segs_per_sec, secs_per_zone;
2638 block_t total_sections, blocks_per_seg;
2639 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2640 (bh->b_data + F2FS_SUPER_OFFSET);
2641 unsigned int blocksize;
2642 size_t crc_offset = 0;
2643 __u32 crc = 0;
2645 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
2646 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
2647 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2648 return -EINVAL;
2651 /* Check checksum_offset and crc in superblock */
2652 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2653 crc_offset = le32_to_cpu(raw_super->checksum_offset);
2654 if (crc_offset !=
2655 offsetof(struct f2fs_super_block, crc)) {
2656 f2fs_info(sbi, "Invalid SB checksum offset: %zu",
2657 crc_offset);
2658 return -EFSCORRUPTED;
2660 crc = le32_to_cpu(raw_super->crc);
2661 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2662 f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
2663 return -EFSCORRUPTED;
2667 /* Currently, support only 4KB page cache size */
2668 if (F2FS_BLKSIZE != PAGE_SIZE) {
2669 f2fs_info(sbi, "Invalid page_cache_size (%lu), supports only 4KB",
2670 PAGE_SIZE);
2671 return -EFSCORRUPTED;
2674 /* Currently, support only 4KB block size */
2675 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2676 if (blocksize != F2FS_BLKSIZE) {
2677 f2fs_info(sbi, "Invalid blocksize (%u), supports only 4KB",
2678 blocksize);
2679 return -EFSCORRUPTED;
2682 /* check log blocks per segment */
2683 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2684 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
2685 le32_to_cpu(raw_super->log_blocks_per_seg));
2686 return -EFSCORRUPTED;
2689 /* Currently, support 512/1024/2048/4096 bytes sector size */
2690 if (le32_to_cpu(raw_super->log_sectorsize) >
2691 F2FS_MAX_LOG_SECTOR_SIZE ||
2692 le32_to_cpu(raw_super->log_sectorsize) <
2693 F2FS_MIN_LOG_SECTOR_SIZE) {
2694 f2fs_info(sbi, "Invalid log sectorsize (%u)",
2695 le32_to_cpu(raw_super->log_sectorsize));
2696 return -EFSCORRUPTED;
2698 if (le32_to_cpu(raw_super->log_sectors_per_block) +
2699 le32_to_cpu(raw_super->log_sectorsize) !=
2700 F2FS_MAX_LOG_SECTOR_SIZE) {
2701 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
2702 le32_to_cpu(raw_super->log_sectors_per_block),
2703 le32_to_cpu(raw_super->log_sectorsize));
2704 return -EFSCORRUPTED;
2707 segment_count = le32_to_cpu(raw_super->segment_count);
2708 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2709 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2710 total_sections = le32_to_cpu(raw_super->section_count);
2712 /* blocks_per_seg should be 512, given the above check */
2713 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2715 if (segment_count > F2FS_MAX_SEGMENT ||
2716 segment_count < F2FS_MIN_SEGMENTS) {
2717 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
2718 return -EFSCORRUPTED;
2721 if (total_sections > segment_count ||
2722 total_sections < F2FS_MIN_SEGMENTS ||
2723 segs_per_sec > segment_count || !segs_per_sec) {
2724 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
2725 segment_count, total_sections, segs_per_sec);
2726 return -EFSCORRUPTED;
2729 if ((segment_count / segs_per_sec) < total_sections) {
2730 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
2731 segment_count, segs_per_sec, total_sections);
2732 return -EFSCORRUPTED;
2735 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2736 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
2737 segment_count, le64_to_cpu(raw_super->block_count));
2738 return -EFSCORRUPTED;
2741 if (RDEV(0).path[0]) {
2742 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
2743 int i = 1;
2745 while (i < MAX_DEVICES && RDEV(i).path[0]) {
2746 dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
2747 i++;
2749 if (segment_count != dev_seg_count) {
2750 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
2751 segment_count, dev_seg_count);
2752 return -EFSCORRUPTED;
2756 if (secs_per_zone > total_sections || !secs_per_zone) {
2757 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
2758 secs_per_zone, total_sections);
2759 return -EFSCORRUPTED;
2761 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2762 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2763 (le32_to_cpu(raw_super->extension_count) +
2764 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2765 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
2766 le32_to_cpu(raw_super->extension_count),
2767 raw_super->hot_ext_count,
2768 F2FS_MAX_EXTENSION);
2769 return -EFSCORRUPTED;
2772 if (le32_to_cpu(raw_super->cp_payload) >
2773 (blocks_per_seg - F2FS_CP_PACKS)) {
2774 f2fs_info(sbi, "Insane cp_payload (%u > %u)",
2775 le32_to_cpu(raw_super->cp_payload),
2776 blocks_per_seg - F2FS_CP_PACKS);
2777 return -EFSCORRUPTED;
2780 /* check reserved ino info */
2781 if (le32_to_cpu(raw_super->node_ino) != 1 ||
2782 le32_to_cpu(raw_super->meta_ino) != 2 ||
2783 le32_to_cpu(raw_super->root_ino) != 3) {
2784 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2785 le32_to_cpu(raw_super->node_ino),
2786 le32_to_cpu(raw_super->meta_ino),
2787 le32_to_cpu(raw_super->root_ino));
2788 return -EFSCORRUPTED;
2791 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2792 if (sanity_check_area_boundary(sbi, bh))
2793 return -EFSCORRUPTED;
2795 return 0;
2798 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2800 unsigned int total, fsmeta;
2801 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2802 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2803 unsigned int ovp_segments, reserved_segments;
2804 unsigned int main_segs, blocks_per_seg;
2805 unsigned int sit_segs, nat_segs;
2806 unsigned int sit_bitmap_size, nat_bitmap_size;
2807 unsigned int log_blocks_per_seg;
2808 unsigned int segment_count_main;
2809 unsigned int cp_pack_start_sum, cp_payload;
2810 block_t user_block_count, valid_user_blocks;
2811 block_t avail_node_count, valid_node_count;
2812 int i, j;
2814 total = le32_to_cpu(raw_super->segment_count);
2815 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2816 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2817 fsmeta += sit_segs;
2818 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2819 fsmeta += nat_segs;
2820 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2821 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2823 if (unlikely(fsmeta >= total))
2824 return 1;
2826 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2827 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2829 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2830 ovp_segments == 0 || reserved_segments == 0)) {
2831 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
2832 return 1;
2835 user_block_count = le64_to_cpu(ckpt->user_block_count);
2836 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2837 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2838 if (!user_block_count || user_block_count >=
2839 segment_count_main << log_blocks_per_seg) {
2840 f2fs_err(sbi, "Wrong user_block_count: %u",
2841 user_block_count);
2842 return 1;
2845 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
2846 if (valid_user_blocks > user_block_count) {
2847 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
2848 valid_user_blocks, user_block_count);
2849 return 1;
2852 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
2853 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
2854 if (valid_node_count > avail_node_count) {
2855 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
2856 valid_node_count, avail_node_count);
2857 return 1;
2860 main_segs = le32_to_cpu(raw_super->segment_count_main);
2861 blocks_per_seg = sbi->blocks_per_seg;
2863 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2864 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2865 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2866 return 1;
2867 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2868 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2869 le32_to_cpu(ckpt->cur_node_segno[j])) {
2870 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
2871 i, j,
2872 le32_to_cpu(ckpt->cur_node_segno[i]));
2873 return 1;
2877 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2878 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2879 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2880 return 1;
2881 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2882 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2883 le32_to_cpu(ckpt->cur_data_segno[j])) {
2884 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
2885 i, j,
2886 le32_to_cpu(ckpt->cur_data_segno[i]));
2887 return 1;
2891 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2892 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
2893 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2894 le32_to_cpu(ckpt->cur_data_segno[j])) {
2895 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
2896 i, j,
2897 le32_to_cpu(ckpt->cur_node_segno[i]));
2898 return 1;
2903 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2904 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2906 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2907 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2908 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
2909 sit_bitmap_size, nat_bitmap_size);
2910 return 1;
2913 cp_pack_start_sum = __start_sum_addr(sbi);
2914 cp_payload = __cp_payload(sbi);
2915 if (cp_pack_start_sum < cp_payload + 1 ||
2916 cp_pack_start_sum > blocks_per_seg - 1 -
2917 NR_CURSEG_TYPE) {
2918 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
2919 cp_pack_start_sum);
2920 return 1;
2923 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
2924 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
2925 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
2926 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
2927 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
2928 le32_to_cpu(ckpt->checksum_offset));
2929 return 1;
2932 if (unlikely(f2fs_cp_error(sbi))) {
2933 f2fs_err(sbi, "A bug case: need to run fsck");
2934 return 1;
2936 return 0;
2939 static void init_sb_info(struct f2fs_sb_info *sbi)
2941 struct f2fs_super_block *raw_super = sbi->raw_super;
2942 int i;
2944 sbi->log_sectors_per_block =
2945 le32_to_cpu(raw_super->log_sectors_per_block);
2946 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2947 sbi->blocksize = 1 << sbi->log_blocksize;
2948 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2949 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2950 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2951 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2952 sbi->total_sections = le32_to_cpu(raw_super->section_count);
2953 sbi->total_node_count =
2954 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2955 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2956 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2957 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2958 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2959 sbi->cur_victim_sec = NULL_SECNO;
2960 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
2961 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
2962 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2963 sbi->migration_granularity = sbi->segs_per_sec;
2965 sbi->dir_level = DEF_DIR_LEVEL;
2966 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2967 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2968 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
2969 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
2970 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
2971 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
2972 DEF_UMOUNT_DISCARD_TIMEOUT;
2973 clear_sbi_flag(sbi, SBI_NEED_FSCK);
2975 for (i = 0; i < NR_COUNT_TYPE; i++)
2976 atomic_set(&sbi->nr_pages[i], 0);
2978 for (i = 0; i < META; i++)
2979 atomic_set(&sbi->wb_sync_req[i], 0);
2981 INIT_LIST_HEAD(&sbi->s_list);
2982 mutex_init(&sbi->umount_mutex);
2983 init_rwsem(&sbi->io_order_lock);
2984 spin_lock_init(&sbi->cp_lock);
2986 sbi->dirty_device = 0;
2987 spin_lock_init(&sbi->dev_lock);
2989 init_rwsem(&sbi->sb_lock);
2990 init_rwsem(&sbi->pin_sem);
2993 static int init_percpu_info(struct f2fs_sb_info *sbi)
2995 int err;
2997 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2998 if (err)
2999 return err;
3001 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3002 GFP_KERNEL);
3003 if (err)
3004 percpu_counter_destroy(&sbi->alloc_valid_block_count);
3006 return err;
3009 #ifdef CONFIG_BLK_DEV_ZONED
3010 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3011 void *data)
3013 struct f2fs_dev_info *dev = data;
3015 if (zone->type != BLK_ZONE_TYPE_CONVENTIONAL)
3016 set_bit(idx, dev->blkz_seq);
3017 return 0;
3020 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3022 struct block_device *bdev = FDEV(devi).bdev;
3023 sector_t nr_sectors = bdev->bd_part->nr_sects;
3024 int ret;
3026 if (!f2fs_sb_has_blkzoned(sbi))
3027 return 0;
3029 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3030 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
3031 return -EINVAL;
3032 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
3033 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
3034 __ilog2_u32(sbi->blocks_per_blkz))
3035 return -EINVAL;
3036 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
3037 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
3038 sbi->log_blocks_per_blkz;
3039 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
3040 FDEV(devi).nr_blkz++;
3042 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3043 BITS_TO_LONGS(FDEV(devi).nr_blkz)
3044 * sizeof(unsigned long),
3045 GFP_KERNEL);
3046 if (!FDEV(devi).blkz_seq)
3047 return -ENOMEM;
3049 /* Get block zones type */
3050 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3051 &FDEV(devi));
3052 if (ret < 0)
3053 return ret;
3055 return 0;
3057 #endif
3060 * Read f2fs raw super block.
3061 * Because we have two copies of super block, so read both of them
3062 * to get the first valid one. If any one of them is broken, we pass
3063 * them recovery flag back to the caller.
3065 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3066 struct f2fs_super_block **raw_super,
3067 int *valid_super_block, int *recovery)
3069 struct super_block *sb = sbi->sb;
3070 int block;
3071 struct buffer_head *bh;
3072 struct f2fs_super_block *super;
3073 int err = 0;
3075 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3076 if (!super)
3077 return -ENOMEM;
3079 for (block = 0; block < 2; block++) {
3080 bh = sb_bread(sb, block);
3081 if (!bh) {
3082 f2fs_err(sbi, "Unable to read %dth superblock",
3083 block + 1);
3084 err = -EIO;
3085 *recovery = 1;
3086 continue;
3089 /* sanity checking of raw super */
3090 err = sanity_check_raw_super(sbi, bh);
3091 if (err) {
3092 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3093 block + 1);
3094 brelse(bh);
3095 *recovery = 1;
3096 continue;
3099 if (!*raw_super) {
3100 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3101 sizeof(*super));
3102 *valid_super_block = block;
3103 *raw_super = super;
3105 brelse(bh);
3108 /* No valid superblock */
3109 if (!*raw_super)
3110 kvfree(super);
3111 else
3112 err = 0;
3114 return err;
3117 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3119 struct buffer_head *bh;
3120 __u32 crc = 0;
3121 int err;
3123 if ((recover && f2fs_readonly(sbi->sb)) ||
3124 bdev_read_only(sbi->sb->s_bdev)) {
3125 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3126 return -EROFS;
3129 /* we should update superblock crc here */
3130 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3131 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3132 offsetof(struct f2fs_super_block, crc));
3133 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3136 /* write back-up superblock first */
3137 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3138 if (!bh)
3139 return -EIO;
3140 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3141 brelse(bh);
3143 /* if we are in recovery path, skip writing valid superblock */
3144 if (recover || err)
3145 return err;
3147 /* write current valid superblock */
3148 bh = sb_bread(sbi->sb, sbi->valid_super_block);
3149 if (!bh)
3150 return -EIO;
3151 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3152 brelse(bh);
3153 return err;
3156 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3158 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3159 unsigned int max_devices = MAX_DEVICES;
3160 int i;
3162 /* Initialize single device information */
3163 if (!RDEV(0).path[0]) {
3164 if (!bdev_is_zoned(sbi->sb->s_bdev))
3165 return 0;
3166 max_devices = 1;
3170 * Initialize multiple devices information, or single
3171 * zoned block device information.
3173 sbi->devs = f2fs_kzalloc(sbi,
3174 array_size(max_devices,
3175 sizeof(struct f2fs_dev_info)),
3176 GFP_KERNEL);
3177 if (!sbi->devs)
3178 return -ENOMEM;
3180 for (i = 0; i < max_devices; i++) {
3182 if (i > 0 && !RDEV(i).path[0])
3183 break;
3185 if (max_devices == 1) {
3186 /* Single zoned block device mount */
3187 FDEV(0).bdev =
3188 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3189 sbi->sb->s_mode, sbi->sb->s_type);
3190 } else {
3191 /* Multi-device mount */
3192 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3193 FDEV(i).total_segments =
3194 le32_to_cpu(RDEV(i).total_segments);
3195 if (i == 0) {
3196 FDEV(i).start_blk = 0;
3197 FDEV(i).end_blk = FDEV(i).start_blk +
3198 (FDEV(i).total_segments <<
3199 sbi->log_blocks_per_seg) - 1 +
3200 le32_to_cpu(raw_super->segment0_blkaddr);
3201 } else {
3202 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3203 FDEV(i).end_blk = FDEV(i).start_blk +
3204 (FDEV(i).total_segments <<
3205 sbi->log_blocks_per_seg) - 1;
3207 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3208 sbi->sb->s_mode, sbi->sb->s_type);
3210 if (IS_ERR(FDEV(i).bdev))
3211 return PTR_ERR(FDEV(i).bdev);
3213 /* to release errored devices */
3214 sbi->s_ndevs = i + 1;
3216 #ifdef CONFIG_BLK_DEV_ZONED
3217 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3218 !f2fs_sb_has_blkzoned(sbi)) {
3219 f2fs_err(sbi, "Zoned block device feature not enabled\n");
3220 return -EINVAL;
3222 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3223 if (init_blkz_info(sbi, i)) {
3224 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3225 return -EINVAL;
3227 if (max_devices == 1)
3228 break;
3229 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3230 i, FDEV(i).path,
3231 FDEV(i).total_segments,
3232 FDEV(i).start_blk, FDEV(i).end_blk,
3233 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3234 "Host-aware" : "Host-managed");
3235 continue;
3237 #endif
3238 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3239 i, FDEV(i).path,
3240 FDEV(i).total_segments,
3241 FDEV(i).start_blk, FDEV(i).end_blk);
3243 f2fs_info(sbi,
3244 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3245 return 0;
3248 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
3250 #ifdef CONFIG_UNICODE
3251 if (f2fs_sb_has_casefold(sbi) && !sbi->s_encoding) {
3252 const struct f2fs_sb_encodings *encoding_info;
3253 struct unicode_map *encoding;
3254 __u16 encoding_flags;
3256 if (f2fs_sb_has_encrypt(sbi)) {
3257 f2fs_err(sbi,
3258 "Can't mount with encoding and encryption");
3259 return -EINVAL;
3262 if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info,
3263 &encoding_flags)) {
3264 f2fs_err(sbi,
3265 "Encoding requested by superblock is unknown");
3266 return -EINVAL;
3269 encoding = utf8_load(encoding_info->version);
3270 if (IS_ERR(encoding)) {
3271 f2fs_err(sbi,
3272 "can't mount with superblock charset: %s-%s "
3273 "not supported by the kernel. flags: 0x%x.",
3274 encoding_info->name, encoding_info->version,
3275 encoding_flags);
3276 return PTR_ERR(encoding);
3278 f2fs_info(sbi, "Using encoding defined by superblock: "
3279 "%s-%s with flags 0x%hx", encoding_info->name,
3280 encoding_info->version?:"\b", encoding_flags);
3282 sbi->s_encoding = encoding;
3283 sbi->s_encoding_flags = encoding_flags;
3284 sbi->sb->s_d_op = &f2fs_dentry_ops;
3286 #else
3287 if (f2fs_sb_has_casefold(sbi)) {
3288 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
3289 return -EINVAL;
3291 #endif
3292 return 0;
3295 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3297 struct f2fs_sm_info *sm_i = SM_I(sbi);
3299 /* adjust parameters according to the volume size */
3300 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3301 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3302 sm_i->dcc_info->discard_granularity = 1;
3303 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3306 sbi->readdir_ra = 1;
3309 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3311 struct f2fs_sb_info *sbi;
3312 struct f2fs_super_block *raw_super;
3313 struct inode *root;
3314 int err;
3315 bool skip_recovery = false, need_fsck = false;
3316 char *options = NULL;
3317 int recovery, i, valid_super_block;
3318 struct curseg_info *seg_i;
3319 int retry_cnt = 1;
3321 try_onemore:
3322 err = -EINVAL;
3323 raw_super = NULL;
3324 valid_super_block = -1;
3325 recovery = 0;
3327 /* allocate memory for f2fs-specific super block info */
3328 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3329 if (!sbi)
3330 return -ENOMEM;
3332 sbi->sb = sb;
3334 /* Load the checksum driver */
3335 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3336 if (IS_ERR(sbi->s_chksum_driver)) {
3337 f2fs_err(sbi, "Cannot load crc32 driver.");
3338 err = PTR_ERR(sbi->s_chksum_driver);
3339 sbi->s_chksum_driver = NULL;
3340 goto free_sbi;
3343 /* set a block size */
3344 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3345 f2fs_err(sbi, "unable to set blocksize");
3346 goto free_sbi;
3349 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3350 &recovery);
3351 if (err)
3352 goto free_sbi;
3354 sb->s_fs_info = sbi;
3355 sbi->raw_super = raw_super;
3357 /* precompute checksum seed for metadata */
3358 if (f2fs_sb_has_inode_chksum(sbi))
3359 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3360 sizeof(raw_super->uuid));
3363 * The BLKZONED feature indicates that the drive was formatted with
3364 * zone alignment optimization. This is optional for host-aware
3365 * devices, but mandatory for host-managed zoned block devices.
3367 #ifndef CONFIG_BLK_DEV_ZONED
3368 if (f2fs_sb_has_blkzoned(sbi)) {
3369 f2fs_err(sbi, "Zoned block device support is not enabled");
3370 err = -EOPNOTSUPP;
3371 goto free_sb_buf;
3373 #endif
3374 default_options(sbi);
3375 /* parse mount options */
3376 options = kstrdup((const char *)data, GFP_KERNEL);
3377 if (data && !options) {
3378 err = -ENOMEM;
3379 goto free_sb_buf;
3382 err = parse_options(sb, options);
3383 if (err)
3384 goto free_options;
3386 sbi->max_file_blocks = max_file_blocks();
3387 sb->s_maxbytes = sbi->max_file_blocks <<
3388 le32_to_cpu(raw_super->log_blocksize);
3389 sb->s_max_links = F2FS_LINK_MAX;
3391 err = f2fs_setup_casefold(sbi);
3392 if (err)
3393 goto free_options;
3395 #ifdef CONFIG_QUOTA
3396 sb->dq_op = &f2fs_quota_operations;
3397 sb->s_qcop = &f2fs_quotactl_ops;
3398 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3400 if (f2fs_sb_has_quota_ino(sbi)) {
3401 for (i = 0; i < MAXQUOTAS; i++) {
3402 if (f2fs_qf_ino(sbi->sb, i))
3403 sbi->nquota_files++;
3406 #endif
3408 sb->s_op = &f2fs_sops;
3409 #ifdef CONFIG_FS_ENCRYPTION
3410 sb->s_cop = &f2fs_cryptops;
3411 #endif
3412 #ifdef CONFIG_FS_VERITY
3413 sb->s_vop = &f2fs_verityops;
3414 #endif
3415 sb->s_xattr = f2fs_xattr_handlers;
3416 sb->s_export_op = &f2fs_export_ops;
3417 sb->s_magic = F2FS_SUPER_MAGIC;
3418 sb->s_time_gran = 1;
3419 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3420 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3421 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3422 sb->s_iflags |= SB_I_CGROUPWB;
3424 /* init f2fs-specific super block info */
3425 sbi->valid_super_block = valid_super_block;
3426 init_rwsem(&sbi->gc_lock);
3427 mutex_init(&sbi->writepages);
3428 mutex_init(&sbi->cp_mutex);
3429 mutex_init(&sbi->resize_mutex);
3430 init_rwsem(&sbi->node_write);
3431 init_rwsem(&sbi->node_change);
3433 /* disallow all the data/node/meta page writes */
3434 set_sbi_flag(sbi, SBI_POR_DOING);
3435 spin_lock_init(&sbi->stat_lock);
3437 /* init iostat info */
3438 spin_lock_init(&sbi->iostat_lock);
3439 sbi->iostat_enable = false;
3441 for (i = 0; i < NR_PAGE_TYPE; i++) {
3442 int n = (i == META) ? 1: NR_TEMP_TYPE;
3443 int j;
3445 sbi->write_io[i] =
3446 f2fs_kmalloc(sbi,
3447 array_size(n,
3448 sizeof(struct f2fs_bio_info)),
3449 GFP_KERNEL);
3450 if (!sbi->write_io[i]) {
3451 err = -ENOMEM;
3452 goto free_bio_info;
3455 for (j = HOT; j < n; j++) {
3456 init_rwsem(&sbi->write_io[i][j].io_rwsem);
3457 sbi->write_io[i][j].sbi = sbi;
3458 sbi->write_io[i][j].bio = NULL;
3459 spin_lock_init(&sbi->write_io[i][j].io_lock);
3460 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3461 INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list);
3462 init_rwsem(&sbi->write_io[i][j].bio_list_lock);
3466 init_rwsem(&sbi->cp_rwsem);
3467 init_rwsem(&sbi->quota_sem);
3468 init_waitqueue_head(&sbi->cp_wait);
3469 init_sb_info(sbi);
3471 err = init_percpu_info(sbi);
3472 if (err)
3473 goto free_bio_info;
3475 if (F2FS_IO_ALIGNED(sbi)) {
3476 sbi->write_io_dummy =
3477 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3478 if (!sbi->write_io_dummy) {
3479 err = -ENOMEM;
3480 goto free_percpu;
3484 /* init per sbi slab cache */
3485 err = f2fs_init_xattr_caches(sbi);
3486 if (err)
3487 goto free_io_dummy;
3489 /* get an inode for meta space */
3490 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3491 if (IS_ERR(sbi->meta_inode)) {
3492 f2fs_err(sbi, "Failed to read F2FS meta data inode");
3493 err = PTR_ERR(sbi->meta_inode);
3494 goto free_xattr_cache;
3497 err = f2fs_get_valid_checkpoint(sbi);
3498 if (err) {
3499 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
3500 goto free_meta_inode;
3503 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3504 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3505 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3506 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3507 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3510 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
3511 set_sbi_flag(sbi, SBI_NEED_FSCK);
3513 /* Initialize device list */
3514 err = f2fs_scan_devices(sbi);
3515 if (err) {
3516 f2fs_err(sbi, "Failed to find devices");
3517 goto free_devices;
3520 err = f2fs_init_post_read_wq(sbi);
3521 if (err) {
3522 f2fs_err(sbi, "Failed to initialize post read workqueue");
3523 goto free_devices;
3526 sbi->total_valid_node_count =
3527 le32_to_cpu(sbi->ckpt->valid_node_count);
3528 percpu_counter_set(&sbi->total_valid_inode_count,
3529 le32_to_cpu(sbi->ckpt->valid_inode_count));
3530 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3531 sbi->total_valid_block_count =
3532 le64_to_cpu(sbi->ckpt->valid_block_count);
3533 sbi->last_valid_block_count = sbi->total_valid_block_count;
3534 sbi->reserved_blocks = 0;
3535 sbi->current_reserved_blocks = 0;
3536 limit_reserve_root(sbi);
3537 adjust_unusable_cap_perc(sbi);
3539 for (i = 0; i < NR_INODE_TYPE; i++) {
3540 INIT_LIST_HEAD(&sbi->inode_list[i]);
3541 spin_lock_init(&sbi->inode_lock[i]);
3543 mutex_init(&sbi->flush_lock);
3545 f2fs_init_extent_cache_info(sbi);
3547 f2fs_init_ino_entry_info(sbi);
3549 f2fs_init_fsync_node_info(sbi);
3551 /* setup f2fs internal modules */
3552 err = f2fs_build_segment_manager(sbi);
3553 if (err) {
3554 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
3555 err);
3556 goto free_sm;
3558 err = f2fs_build_node_manager(sbi);
3559 if (err) {
3560 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
3561 err);
3562 goto free_nm;
3565 /* For write statistics */
3566 if (sb->s_bdev->bd_part)
3567 sbi->sectors_written_start =
3568 (u64)part_stat_read(sb->s_bdev->bd_part,
3569 sectors[STAT_WRITE]);
3571 /* Read accumulated write IO statistics if exists */
3572 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3573 if (__exist_node_summaries(sbi))
3574 sbi->kbytes_written =
3575 le64_to_cpu(seg_i->journal->info.kbytes_written);
3577 f2fs_build_gc_manager(sbi);
3579 err = f2fs_build_stats(sbi);
3580 if (err)
3581 goto free_nm;
3583 /* get an inode for node space */
3584 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3585 if (IS_ERR(sbi->node_inode)) {
3586 f2fs_err(sbi, "Failed to read node inode");
3587 err = PTR_ERR(sbi->node_inode);
3588 goto free_stats;
3591 /* read root inode and dentry */
3592 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3593 if (IS_ERR(root)) {
3594 f2fs_err(sbi, "Failed to read root inode");
3595 err = PTR_ERR(root);
3596 goto free_node_inode;
3598 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3599 !root->i_size || !root->i_nlink) {
3600 iput(root);
3601 err = -EINVAL;
3602 goto free_node_inode;
3605 sb->s_root = d_make_root(root); /* allocate root dentry */
3606 if (!sb->s_root) {
3607 err = -ENOMEM;
3608 goto free_node_inode;
3611 err = f2fs_register_sysfs(sbi);
3612 if (err)
3613 goto free_root_inode;
3615 #ifdef CONFIG_QUOTA
3616 /* Enable quota usage during mount */
3617 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3618 err = f2fs_enable_quotas(sb);
3619 if (err)
3620 f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
3622 #endif
3623 /* if there are any orphan inodes, free them */
3624 err = f2fs_recover_orphan_inodes(sbi);
3625 if (err)
3626 goto free_meta;
3628 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3629 goto reset_checkpoint;
3631 /* recover fsynced data */
3632 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
3633 !test_opt(sbi, NORECOVERY)) {
3635 * mount should be failed, when device has readonly mode, and
3636 * previous checkpoint was not done by clean system shutdown.
3638 if (f2fs_hw_is_readonly(sbi)) {
3639 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3640 err = -EROFS;
3641 f2fs_err(sbi, "Need to recover fsync data, but write access unavailable");
3642 goto free_meta;
3644 f2fs_info(sbi, "write access unavailable, skipping recovery");
3645 goto reset_checkpoint;
3648 if (need_fsck)
3649 set_sbi_flag(sbi, SBI_NEED_FSCK);
3651 if (skip_recovery)
3652 goto reset_checkpoint;
3654 err = f2fs_recover_fsync_data(sbi, false);
3655 if (err < 0) {
3656 if (err != -ENOMEM)
3657 skip_recovery = true;
3658 need_fsck = true;
3659 f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
3660 err);
3661 goto free_meta;
3663 } else {
3664 err = f2fs_recover_fsync_data(sbi, true);
3666 if (!f2fs_readonly(sb) && err > 0) {
3667 err = -EINVAL;
3668 f2fs_err(sbi, "Need to recover fsync data");
3669 goto free_meta;
3674 * If the f2fs is not readonly and fsync data recovery succeeds,
3675 * check zoned block devices' write pointer consistency.
3677 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
3678 err = f2fs_check_write_pointer(sbi);
3679 if (err)
3680 goto free_meta;
3683 reset_checkpoint:
3684 /* f2fs_recover_fsync_data() cleared this already */
3685 clear_sbi_flag(sbi, SBI_POR_DOING);
3687 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3688 err = f2fs_disable_checkpoint(sbi);
3689 if (err)
3690 goto sync_free_meta;
3691 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3692 f2fs_enable_checkpoint(sbi);
3696 * If filesystem is not mounted as read-only then
3697 * do start the gc_thread.
3699 if (F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF && !f2fs_readonly(sb)) {
3700 /* After POR, we can run background GC thread.*/
3701 err = f2fs_start_gc_thread(sbi);
3702 if (err)
3703 goto sync_free_meta;
3705 kvfree(options);
3707 /* recover broken superblock */
3708 if (recovery) {
3709 err = f2fs_commit_super(sbi, true);
3710 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
3711 sbi->valid_super_block ? 1 : 2, err);
3714 f2fs_join_shrinker(sbi);
3716 f2fs_tuning_parameters(sbi);
3718 f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
3719 cur_cp_version(F2FS_CKPT(sbi)));
3720 f2fs_update_time(sbi, CP_TIME);
3721 f2fs_update_time(sbi, REQ_TIME);
3722 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3723 return 0;
3725 sync_free_meta:
3726 /* safe to flush all the data */
3727 sync_filesystem(sbi->sb);
3728 retry_cnt = 0;
3730 free_meta:
3731 #ifdef CONFIG_QUOTA
3732 f2fs_truncate_quota_inode_pages(sb);
3733 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3734 f2fs_quota_off_umount(sbi->sb);
3735 #endif
3737 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3738 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3739 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3740 * falls into an infinite loop in f2fs_sync_meta_pages().
3742 truncate_inode_pages_final(META_MAPPING(sbi));
3743 /* evict some inodes being cached by GC */
3744 evict_inodes(sb);
3745 f2fs_unregister_sysfs(sbi);
3746 free_root_inode:
3747 dput(sb->s_root);
3748 sb->s_root = NULL;
3749 free_node_inode:
3750 f2fs_release_ino_entry(sbi, true);
3751 truncate_inode_pages_final(NODE_MAPPING(sbi));
3752 iput(sbi->node_inode);
3753 sbi->node_inode = NULL;
3754 free_stats:
3755 f2fs_destroy_stats(sbi);
3756 free_nm:
3757 f2fs_destroy_node_manager(sbi);
3758 free_sm:
3759 f2fs_destroy_segment_manager(sbi);
3760 f2fs_destroy_post_read_wq(sbi);
3761 free_devices:
3762 destroy_device_list(sbi);
3763 kvfree(sbi->ckpt);
3764 free_meta_inode:
3765 make_bad_inode(sbi->meta_inode);
3766 iput(sbi->meta_inode);
3767 sbi->meta_inode = NULL;
3768 free_xattr_cache:
3769 f2fs_destroy_xattr_caches(sbi);
3770 free_io_dummy:
3771 mempool_destroy(sbi->write_io_dummy);
3772 free_percpu:
3773 destroy_percpu_info(sbi);
3774 free_bio_info:
3775 for (i = 0; i < NR_PAGE_TYPE; i++)
3776 kvfree(sbi->write_io[i]);
3778 #ifdef CONFIG_UNICODE
3779 utf8_unload(sbi->s_encoding);
3780 #endif
3781 free_options:
3782 #ifdef CONFIG_QUOTA
3783 for (i = 0; i < MAXQUOTAS; i++)
3784 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
3785 #endif
3786 kvfree(options);
3787 free_sb_buf:
3788 kvfree(raw_super);
3789 free_sbi:
3790 if (sbi->s_chksum_driver)
3791 crypto_free_shash(sbi->s_chksum_driver);
3792 kvfree(sbi);
3794 /* give only one another chance */
3795 if (retry_cnt > 0 && skip_recovery) {
3796 retry_cnt--;
3797 shrink_dcache_sb(sb);
3798 goto try_onemore;
3800 return err;
3803 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3804 const char *dev_name, void *data)
3806 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3809 static void kill_f2fs_super(struct super_block *sb)
3811 if (sb->s_root) {
3812 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3814 set_sbi_flag(sbi, SBI_IS_CLOSE);
3815 f2fs_stop_gc_thread(sbi);
3816 f2fs_stop_discard_thread(sbi);
3818 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3819 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3820 struct cp_control cpc = {
3821 .reason = CP_UMOUNT,
3823 f2fs_write_checkpoint(sbi, &cpc);
3826 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3827 sb->s_flags &= ~SB_RDONLY;
3829 kill_block_super(sb);
3832 static struct file_system_type f2fs_fs_type = {
3833 .owner = THIS_MODULE,
3834 .name = "f2fs",
3835 .mount = f2fs_mount,
3836 .kill_sb = kill_f2fs_super,
3837 .fs_flags = FS_REQUIRES_DEV,
3839 MODULE_ALIAS_FS("f2fs");
3841 static int __init init_inodecache(void)
3843 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3844 sizeof(struct f2fs_inode_info), 0,
3845 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3846 if (!f2fs_inode_cachep)
3847 return -ENOMEM;
3848 return 0;
3851 static void destroy_inodecache(void)
3854 * Make sure all delayed rcu free inodes are flushed before we
3855 * destroy cache.
3857 rcu_barrier();
3858 kmem_cache_destroy(f2fs_inode_cachep);
3861 static int __init init_f2fs_fs(void)
3863 int err;
3865 if (PAGE_SIZE != F2FS_BLKSIZE) {
3866 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3867 PAGE_SIZE, F2FS_BLKSIZE);
3868 return -EINVAL;
3871 f2fs_build_trace_ios();
3873 err = init_inodecache();
3874 if (err)
3875 goto fail;
3876 err = f2fs_create_node_manager_caches();
3877 if (err)
3878 goto free_inodecache;
3879 err = f2fs_create_segment_manager_caches();
3880 if (err)
3881 goto free_node_manager_caches;
3882 err = f2fs_create_checkpoint_caches();
3883 if (err)
3884 goto free_segment_manager_caches;
3885 err = f2fs_create_extent_cache();
3886 if (err)
3887 goto free_checkpoint_caches;
3888 err = f2fs_init_sysfs();
3889 if (err)
3890 goto free_extent_cache;
3891 err = register_shrinker(&f2fs_shrinker_info);
3892 if (err)
3893 goto free_sysfs;
3894 err = register_filesystem(&f2fs_fs_type);
3895 if (err)
3896 goto free_shrinker;
3897 f2fs_create_root_stats();
3898 err = f2fs_init_post_read_processing();
3899 if (err)
3900 goto free_root_stats;
3901 err = f2fs_init_bio_entry_cache();
3902 if (err)
3903 goto free_post_read;
3904 err = f2fs_init_bioset();
3905 if (err)
3906 goto free_bio_enrty_cache;
3907 return 0;
3908 free_bio_enrty_cache:
3909 f2fs_destroy_bio_entry_cache();
3910 free_post_read:
3911 f2fs_destroy_post_read_processing();
3912 free_root_stats:
3913 f2fs_destroy_root_stats();
3914 unregister_filesystem(&f2fs_fs_type);
3915 free_shrinker:
3916 unregister_shrinker(&f2fs_shrinker_info);
3917 free_sysfs:
3918 f2fs_exit_sysfs();
3919 free_extent_cache:
3920 f2fs_destroy_extent_cache();
3921 free_checkpoint_caches:
3922 f2fs_destroy_checkpoint_caches();
3923 free_segment_manager_caches:
3924 f2fs_destroy_segment_manager_caches();
3925 free_node_manager_caches:
3926 f2fs_destroy_node_manager_caches();
3927 free_inodecache:
3928 destroy_inodecache();
3929 fail:
3930 return err;
3933 static void __exit exit_f2fs_fs(void)
3935 f2fs_destroy_bioset();
3936 f2fs_destroy_bio_entry_cache();
3937 f2fs_destroy_post_read_processing();
3938 f2fs_destroy_root_stats();
3939 unregister_filesystem(&f2fs_fs_type);
3940 unregister_shrinker(&f2fs_shrinker_info);
3941 f2fs_exit_sysfs();
3942 f2fs_destroy_extent_cache();
3943 f2fs_destroy_checkpoint_caches();
3944 f2fs_destroy_segment_manager_caches();
3945 f2fs_destroy_node_manager_caches();
3946 destroy_inodecache();
3947 f2fs_destroy_trace_ios();
3950 module_init(init_f2fs_fs)
3951 module_exit(exit_f2fs_fs)
3953 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3954 MODULE_DESCRIPTION("Flash Friendly File System");
3955 MODULE_LICENSE("GPL");