1 // SPDX-License-Identifier: GPL-2.0
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
8 #include <linux/module.h>
9 #include <linux/init.h>
11 #include <linux/fs_context.h>
12 #include <linux/sched/mm.h>
13 #include <linux/statfs.h>
14 #include <linux/buffer_head.h>
15 #include <linux/kthread.h>
16 #include <linux/parser.h>
17 #include <linux/mount.h>
18 #include <linux/seq_file.h>
19 #include <linux/proc_fs.h>
20 #include <linux/random.h>
21 #include <linux/exportfs.h>
22 #include <linux/blkdev.h>
23 #include <linux/quotaops.h>
24 #include <linux/f2fs_fs.h>
25 #include <linux/sysfs.h>
26 #include <linux/quota.h>
27 #include <linux/unicode.h>
28 #include <linux/part_stat.h>
29 #include <linux/zstd.h>
30 #include <linux/lz4.h>
39 #define CREATE_TRACE_POINTS
40 #include <trace/events/f2fs.h>
42 static struct kmem_cache *f2fs_inode_cachep;
44 #ifdef CONFIG_F2FS_FAULT_INJECTION
46 const char *f2fs_fault_name[FAULT_MAX] = {
47 [FAULT_KMALLOC] = "kmalloc",
48 [FAULT_KVMALLOC] = "kvmalloc",
49 [FAULT_PAGE_ALLOC] = "page alloc",
50 [FAULT_PAGE_GET] = "page get",
51 [FAULT_ALLOC_NID] = "alloc nid",
52 [FAULT_ORPHAN] = "orphan",
53 [FAULT_BLOCK] = "no more block",
54 [FAULT_DIR_DEPTH] = "too big dir depth",
55 [FAULT_EVICT_INODE] = "evict_inode fail",
56 [FAULT_TRUNCATE] = "truncate fail",
57 [FAULT_READ_IO] = "read IO error",
58 [FAULT_CHECKPOINT] = "checkpoint error",
59 [FAULT_DISCARD] = "discard error",
60 [FAULT_WRITE_IO] = "write IO error",
61 [FAULT_SLAB_ALLOC] = "slab alloc",
62 [FAULT_DQUOT_INIT] = "dquot initialize",
63 [FAULT_LOCK_OP] = "lock_op",
64 [FAULT_BLKADDR_VALIDITY] = "invalid blkaddr",
65 [FAULT_BLKADDR_CONSISTENCE] = "inconsistent blkaddr",
66 [FAULT_NO_SEGMENT] = "no free segment",
69 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
72 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
75 atomic_set(&ffi->inject_ops, 0);
76 ffi->inject_rate = rate;
80 ffi->inject_type = type;
83 memset(ffi, 0, sizeof(struct f2fs_fault_info));
87 /* f2fs-wide shrinker description */
88 static struct shrinker *f2fs_shrinker_info;
90 static int __init f2fs_init_shrinker(void)
92 f2fs_shrinker_info = shrinker_alloc(0, "f2fs-shrinker");
93 if (!f2fs_shrinker_info)
96 f2fs_shrinker_info->count_objects = f2fs_shrink_count;
97 f2fs_shrinker_info->scan_objects = f2fs_shrink_scan;
99 shrinker_register(f2fs_shrinker_info);
104 static void f2fs_exit_shrinker(void)
106 shrinker_free(f2fs_shrinker_info);
111 Opt_disable_roll_forward,
122 Opt_disable_ext_identify,
125 Opt_inline_xattr_size,
162 Opt_test_dummy_encryption,
164 Opt_checkpoint_disable,
165 Opt_checkpoint_disable_cap,
166 Opt_checkpoint_disable_cap_perc,
167 Opt_checkpoint_enable,
168 Opt_checkpoint_merge,
169 Opt_nocheckpoint_merge,
170 Opt_compress_algorithm,
171 Opt_compress_log_size,
172 Opt_compress_extension,
173 Opt_nocompress_extension,
182 Opt_age_extent_cache,
187 static match_table_t f2fs_tokens = {
188 {Opt_gc_background, "background_gc=%s"},
189 {Opt_disable_roll_forward, "disable_roll_forward"},
190 {Opt_norecovery, "norecovery"},
191 {Opt_discard, "discard"},
192 {Opt_nodiscard, "nodiscard"},
193 {Opt_noheap, "no_heap"},
195 {Opt_user_xattr, "user_xattr"},
196 {Opt_nouser_xattr, "nouser_xattr"},
198 {Opt_noacl, "noacl"},
199 {Opt_active_logs, "active_logs=%u"},
200 {Opt_disable_ext_identify, "disable_ext_identify"},
201 {Opt_inline_xattr, "inline_xattr"},
202 {Opt_noinline_xattr, "noinline_xattr"},
203 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
204 {Opt_inline_data, "inline_data"},
205 {Opt_inline_dentry, "inline_dentry"},
206 {Opt_noinline_dentry, "noinline_dentry"},
207 {Opt_flush_merge, "flush_merge"},
208 {Opt_noflush_merge, "noflush_merge"},
209 {Opt_barrier, "barrier"},
210 {Opt_nobarrier, "nobarrier"},
211 {Opt_fastboot, "fastboot"},
212 {Opt_extent_cache, "extent_cache"},
213 {Opt_noextent_cache, "noextent_cache"},
214 {Opt_noinline_data, "noinline_data"},
215 {Opt_data_flush, "data_flush"},
216 {Opt_reserve_root, "reserve_root=%u"},
217 {Opt_resgid, "resgid=%u"},
218 {Opt_resuid, "resuid=%u"},
219 {Opt_mode, "mode=%s"},
220 {Opt_fault_injection, "fault_injection=%u"},
221 {Opt_fault_type, "fault_type=%u"},
222 {Opt_lazytime, "lazytime"},
223 {Opt_nolazytime, "nolazytime"},
224 {Opt_quota, "quota"},
225 {Opt_noquota, "noquota"},
226 {Opt_usrquota, "usrquota"},
227 {Opt_grpquota, "grpquota"},
228 {Opt_prjquota, "prjquota"},
229 {Opt_usrjquota, "usrjquota=%s"},
230 {Opt_grpjquota, "grpjquota=%s"},
231 {Opt_prjjquota, "prjjquota=%s"},
232 {Opt_offusrjquota, "usrjquota="},
233 {Opt_offgrpjquota, "grpjquota="},
234 {Opt_offprjjquota, "prjjquota="},
235 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
236 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
237 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
238 {Opt_alloc, "alloc_mode=%s"},
239 {Opt_fsync, "fsync_mode=%s"},
240 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
241 {Opt_test_dummy_encryption, "test_dummy_encryption"},
242 {Opt_inlinecrypt, "inlinecrypt"},
243 {Opt_checkpoint_disable, "checkpoint=disable"},
244 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
245 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
246 {Opt_checkpoint_enable, "checkpoint=enable"},
247 {Opt_checkpoint_merge, "checkpoint_merge"},
248 {Opt_nocheckpoint_merge, "nocheckpoint_merge"},
249 {Opt_compress_algorithm, "compress_algorithm=%s"},
250 {Opt_compress_log_size, "compress_log_size=%u"},
251 {Opt_compress_extension, "compress_extension=%s"},
252 {Opt_nocompress_extension, "nocompress_extension=%s"},
253 {Opt_compress_chksum, "compress_chksum"},
254 {Opt_compress_mode, "compress_mode=%s"},
255 {Opt_compress_cache, "compress_cache"},
257 {Opt_gc_merge, "gc_merge"},
258 {Opt_nogc_merge, "nogc_merge"},
259 {Opt_discard_unit, "discard_unit=%s"},
260 {Opt_memory_mode, "memory=%s"},
261 {Opt_age_extent_cache, "age_extent_cache"},
262 {Opt_errors, "errors=%s"},
266 void f2fs_printk(struct f2fs_sb_info *sbi, bool limit_rate,
267 const char *fmt, ...)
269 struct va_format vaf;
275 level = printk_get_level(fmt);
276 vaf.fmt = printk_skip_level(fmt);
279 printk_ratelimited("%c%cF2FS-fs (%s): %pV\n",
280 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
282 printk("%c%cF2FS-fs (%s): %pV\n",
283 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
288 #if IS_ENABLED(CONFIG_UNICODE)
289 static const struct f2fs_sb_encodings {
292 unsigned int version;
293 } f2fs_sb_encoding_map[] = {
294 {F2FS_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
297 static const struct f2fs_sb_encodings *
298 f2fs_sb_read_encoding(const struct f2fs_super_block *sb)
300 __u16 magic = le16_to_cpu(sb->s_encoding);
303 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
304 if (magic == f2fs_sb_encoding_map[i].magic)
305 return &f2fs_sb_encoding_map[i];
310 struct kmem_cache *f2fs_cf_name_slab;
311 static int __init f2fs_create_casefold_cache(void)
313 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name",
315 return f2fs_cf_name_slab ? 0 : -ENOMEM;
318 static void f2fs_destroy_casefold_cache(void)
320 kmem_cache_destroy(f2fs_cf_name_slab);
323 static int __init f2fs_create_casefold_cache(void) { return 0; }
324 static void f2fs_destroy_casefold_cache(void) { }
327 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
329 block_t limit = min((sbi->user_block_count >> 3),
330 sbi->user_block_count - sbi->reserved_blocks);
333 if (test_opt(sbi, RESERVE_ROOT) &&
334 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
335 F2FS_OPTION(sbi).root_reserved_blocks = limit;
336 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
337 F2FS_OPTION(sbi).root_reserved_blocks);
339 if (!test_opt(sbi, RESERVE_ROOT) &&
340 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
341 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
342 !gid_eq(F2FS_OPTION(sbi).s_resgid,
343 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
344 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
345 from_kuid_munged(&init_user_ns,
346 F2FS_OPTION(sbi).s_resuid),
347 from_kgid_munged(&init_user_ns,
348 F2FS_OPTION(sbi).s_resgid));
351 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
353 if (!F2FS_OPTION(sbi).unusable_cap_perc)
356 if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
357 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
359 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
360 F2FS_OPTION(sbi).unusable_cap_perc;
362 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
363 F2FS_OPTION(sbi).unusable_cap,
364 F2FS_OPTION(sbi).unusable_cap_perc);
367 static void init_once(void *foo)
369 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
371 inode_init_once(&fi->vfs_inode);
375 static const char * const quotatypes[] = INITQFNAMES;
376 #define QTYPE2NAME(t) (quotatypes[t])
377 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
380 struct f2fs_sb_info *sbi = F2FS_SB(sb);
384 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
385 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
388 if (f2fs_sb_has_quota_ino(sbi)) {
389 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
393 qname = match_strdup(args);
395 f2fs_err(sbi, "Not enough memory for storing quotafile name");
398 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
399 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
402 f2fs_err(sbi, "%s quota file already specified",
406 if (strchr(qname, '/')) {
407 f2fs_err(sbi, "quotafile must be on filesystem root");
410 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
418 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
420 struct f2fs_sb_info *sbi = F2FS_SB(sb);
422 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
423 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
426 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
427 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
431 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
434 * We do the test below only for project quotas. 'usrquota' and
435 * 'grpquota' mount options are allowed even without quota feature
436 * to support legacy quotas in quota files.
438 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
439 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
442 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
443 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
444 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
445 if (test_opt(sbi, USRQUOTA) &&
446 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
447 clear_opt(sbi, USRQUOTA);
449 if (test_opt(sbi, GRPQUOTA) &&
450 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
451 clear_opt(sbi, GRPQUOTA);
453 if (test_opt(sbi, PRJQUOTA) &&
454 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
455 clear_opt(sbi, PRJQUOTA);
457 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
458 test_opt(sbi, PRJQUOTA)) {
459 f2fs_err(sbi, "old and new quota format mixing");
463 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
464 f2fs_err(sbi, "journaled quota format not specified");
469 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
470 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
471 F2FS_OPTION(sbi).s_jquota_fmt = 0;
477 static int f2fs_set_test_dummy_encryption(struct super_block *sb,
479 const substring_t *arg,
482 struct f2fs_sb_info *sbi = F2FS_SB(sb);
483 struct fs_parameter param = {
484 .type = fs_value_is_string,
485 .string = arg->from ? arg->from : "",
487 struct fscrypt_dummy_policy *policy =
488 &F2FS_OPTION(sbi).dummy_enc_policy;
491 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
492 f2fs_warn(sbi, "test_dummy_encryption option not supported");
496 if (!f2fs_sb_has_encrypt(sbi)) {
497 f2fs_err(sbi, "Encrypt feature is off");
502 * This mount option is just for testing, and it's not worthwhile to
503 * implement the extra complexity (e.g. RCU protection) that would be
504 * needed to allow it to be set or changed during remount. We do allow
505 * it to be specified during remount, but only if there is no change.
507 if (is_remount && !fscrypt_is_dummy_policy_set(policy)) {
508 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
512 err = fscrypt_parse_test_dummy_encryption(¶m, policy);
516 "Can't change test_dummy_encryption on remount");
517 else if (err == -EINVAL)
518 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
521 f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
525 f2fs_warn(sbi, "Test dummy encryption mode enabled");
529 #ifdef CONFIG_F2FS_FS_COMPRESSION
530 static bool is_compress_extension_exist(struct f2fs_sb_info *sbi,
531 const char *new_ext, bool is_ext)
533 unsigned char (*ext)[F2FS_EXTENSION_LEN];
538 ext = F2FS_OPTION(sbi).extensions;
539 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
541 ext = F2FS_OPTION(sbi).noextensions;
542 ext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
545 for (i = 0; i < ext_cnt; i++) {
546 if (!strcasecmp(new_ext, ext[i]))
554 * 1. The same extension name cannot not appear in both compress and non-compress extension
556 * 2. If the compress extension specifies all files, the types specified by the non-compress
557 * extension will be treated as special cases and will not be compressed.
558 * 3. Don't allow the non-compress extension specifies all files.
560 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi)
562 unsigned char (*ext)[F2FS_EXTENSION_LEN];
563 unsigned char (*noext)[F2FS_EXTENSION_LEN];
564 int ext_cnt, noext_cnt, index = 0, no_index = 0;
566 ext = F2FS_OPTION(sbi).extensions;
567 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
568 noext = F2FS_OPTION(sbi).noextensions;
569 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
574 for (no_index = 0; no_index < noext_cnt; no_index++) {
575 if (!strcasecmp("*", noext[no_index])) {
576 f2fs_info(sbi, "Don't allow the nocompress extension specifies all files");
579 for (index = 0; index < ext_cnt; index++) {
580 if (!strcasecmp(ext[index], noext[no_index])) {
581 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension",
590 #ifdef CONFIG_F2FS_FS_LZ4
591 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str)
593 #ifdef CONFIG_F2FS_FS_LZ4HC
596 if (strlen(str) == 3) {
597 F2FS_OPTION(sbi).compress_level = 0;
604 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
607 if (kstrtouint(str + 1, 10, &level))
610 if (!f2fs_is_compress_level_valid(COMPRESS_LZ4, level)) {
611 f2fs_info(sbi, "invalid lz4hc compress level: %d", level);
615 F2FS_OPTION(sbi).compress_level = level;
618 if (strlen(str) == 3) {
619 F2FS_OPTION(sbi).compress_level = 0;
622 f2fs_info(sbi, "kernel doesn't support lz4hc compression");
628 #ifdef CONFIG_F2FS_FS_ZSTD
629 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str)
634 if (strlen(str) == len) {
635 F2FS_OPTION(sbi).compress_level = F2FS_ZSTD_DEFAULT_CLEVEL;
642 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
645 if (kstrtoint(str + 1, 10, &level))
648 /* f2fs does not support negative compress level now */
650 f2fs_info(sbi, "do not support negative compress level: %d", level);
654 if (!f2fs_is_compress_level_valid(COMPRESS_ZSTD, level)) {
655 f2fs_info(sbi, "invalid zstd compress level: %d", level);
659 F2FS_OPTION(sbi).compress_level = level;
665 static int parse_options(struct super_block *sb, char *options, bool is_remount)
667 struct f2fs_sb_info *sbi = F2FS_SB(sb);
668 substring_t args[MAX_OPT_ARGS];
669 #ifdef CONFIG_F2FS_FS_COMPRESSION
670 unsigned char (*ext)[F2FS_EXTENSION_LEN];
671 unsigned char (*noext)[F2FS_EXTENSION_LEN];
672 int ext_cnt, noext_cnt;
683 while ((p = strsep(&options, ",")) != NULL) {
689 * Initialize args struct so we know whether arg was
690 * found; some options take optional arguments.
692 args[0].to = args[0].from = NULL;
693 token = match_token(p, f2fs_tokens, args);
696 case Opt_gc_background:
697 name = match_strdup(&args[0]);
701 if (!strcmp(name, "on")) {
702 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
703 } else if (!strcmp(name, "off")) {
704 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
705 } else if (!strcmp(name, "sync")) {
706 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
713 case Opt_disable_roll_forward:
714 set_opt(sbi, DISABLE_ROLL_FORWARD);
717 /* this option mounts f2fs with ro */
718 set_opt(sbi, NORECOVERY);
719 if (!f2fs_readonly(sb))
723 if (!f2fs_hw_support_discard(sbi)) {
724 f2fs_warn(sbi, "device does not support discard");
727 set_opt(sbi, DISCARD);
730 if (f2fs_hw_should_discard(sbi)) {
731 f2fs_warn(sbi, "discard is required for zoned block devices");
734 clear_opt(sbi, DISCARD);
738 f2fs_warn(sbi, "heap/no_heap options were deprecated");
740 #ifdef CONFIG_F2FS_FS_XATTR
742 set_opt(sbi, XATTR_USER);
744 case Opt_nouser_xattr:
745 clear_opt(sbi, XATTR_USER);
747 case Opt_inline_xattr:
748 set_opt(sbi, INLINE_XATTR);
750 case Opt_noinline_xattr:
751 clear_opt(sbi, INLINE_XATTR);
753 case Opt_inline_xattr_size:
754 if (args->from && match_int(args, &arg))
756 set_opt(sbi, INLINE_XATTR_SIZE);
757 F2FS_OPTION(sbi).inline_xattr_size = arg;
761 f2fs_info(sbi, "user_xattr options not supported");
763 case Opt_nouser_xattr:
764 f2fs_info(sbi, "nouser_xattr options not supported");
766 case Opt_inline_xattr:
767 f2fs_info(sbi, "inline_xattr options not supported");
769 case Opt_noinline_xattr:
770 f2fs_info(sbi, "noinline_xattr options not supported");
773 #ifdef CONFIG_F2FS_FS_POSIX_ACL
775 set_opt(sbi, POSIX_ACL);
778 clear_opt(sbi, POSIX_ACL);
782 f2fs_info(sbi, "acl options not supported");
785 f2fs_info(sbi, "noacl options not supported");
788 case Opt_active_logs:
789 if (args->from && match_int(args, &arg))
791 if (arg != 2 && arg != 4 &&
792 arg != NR_CURSEG_PERSIST_TYPE)
794 F2FS_OPTION(sbi).active_logs = arg;
796 case Opt_disable_ext_identify:
797 set_opt(sbi, DISABLE_EXT_IDENTIFY);
799 case Opt_inline_data:
800 set_opt(sbi, INLINE_DATA);
802 case Opt_inline_dentry:
803 set_opt(sbi, INLINE_DENTRY);
805 case Opt_noinline_dentry:
806 clear_opt(sbi, INLINE_DENTRY);
808 case Opt_flush_merge:
809 set_opt(sbi, FLUSH_MERGE);
811 case Opt_noflush_merge:
812 clear_opt(sbi, FLUSH_MERGE);
815 set_opt(sbi, NOBARRIER);
818 clear_opt(sbi, NOBARRIER);
821 set_opt(sbi, FASTBOOT);
823 case Opt_extent_cache:
824 set_opt(sbi, READ_EXTENT_CACHE);
826 case Opt_noextent_cache:
827 clear_opt(sbi, READ_EXTENT_CACHE);
829 case Opt_noinline_data:
830 clear_opt(sbi, INLINE_DATA);
833 set_opt(sbi, DATA_FLUSH);
835 case Opt_reserve_root:
836 if (args->from && match_int(args, &arg))
838 if (test_opt(sbi, RESERVE_ROOT)) {
839 f2fs_info(sbi, "Preserve previous reserve_root=%u",
840 F2FS_OPTION(sbi).root_reserved_blocks);
842 F2FS_OPTION(sbi).root_reserved_blocks = arg;
843 set_opt(sbi, RESERVE_ROOT);
847 if (args->from && match_int(args, &arg))
849 uid = make_kuid(current_user_ns(), arg);
850 if (!uid_valid(uid)) {
851 f2fs_err(sbi, "Invalid uid value %d", arg);
854 F2FS_OPTION(sbi).s_resuid = uid;
857 if (args->from && match_int(args, &arg))
859 gid = make_kgid(current_user_ns(), arg);
860 if (!gid_valid(gid)) {
861 f2fs_err(sbi, "Invalid gid value %d", arg);
864 F2FS_OPTION(sbi).s_resgid = gid;
867 name = match_strdup(&args[0]);
871 if (!strcmp(name, "adaptive")) {
872 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
873 } else if (!strcmp(name, "lfs")) {
874 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
875 } else if (!strcmp(name, "fragment:segment")) {
876 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG;
877 } else if (!strcmp(name, "fragment:block")) {
878 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK;
885 #ifdef CONFIG_F2FS_FAULT_INJECTION
886 case Opt_fault_injection:
887 if (args->from && match_int(args, &arg))
889 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
890 set_opt(sbi, FAULT_INJECTION);
894 if (args->from && match_int(args, &arg))
896 f2fs_build_fault_attr(sbi, 0, arg);
897 set_opt(sbi, FAULT_INJECTION);
900 case Opt_fault_injection:
901 f2fs_info(sbi, "fault_injection options not supported");
905 f2fs_info(sbi, "fault_type options not supported");
909 sb->s_flags |= SB_LAZYTIME;
912 sb->s_flags &= ~SB_LAZYTIME;
917 set_opt(sbi, USRQUOTA);
920 set_opt(sbi, GRPQUOTA);
923 set_opt(sbi, PRJQUOTA);
926 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
931 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
936 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
940 case Opt_offusrjquota:
941 ret = f2fs_clear_qf_name(sb, USRQUOTA);
945 case Opt_offgrpjquota:
946 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
950 case Opt_offprjjquota:
951 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
955 case Opt_jqfmt_vfsold:
956 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
958 case Opt_jqfmt_vfsv0:
959 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
961 case Opt_jqfmt_vfsv1:
962 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
965 clear_opt(sbi, QUOTA);
966 clear_opt(sbi, USRQUOTA);
967 clear_opt(sbi, GRPQUOTA);
968 clear_opt(sbi, PRJQUOTA);
978 case Opt_offusrjquota:
979 case Opt_offgrpjquota:
980 case Opt_offprjjquota:
981 case Opt_jqfmt_vfsold:
982 case Opt_jqfmt_vfsv0:
983 case Opt_jqfmt_vfsv1:
985 f2fs_info(sbi, "quota operations not supported");
989 name = match_strdup(&args[0]);
993 if (!strcmp(name, "default")) {
994 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
995 } else if (!strcmp(name, "reuse")) {
996 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
1004 name = match_strdup(&args[0]);
1007 if (!strcmp(name, "posix")) {
1008 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1009 } else if (!strcmp(name, "strict")) {
1010 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
1011 } else if (!strcmp(name, "nobarrier")) {
1012 F2FS_OPTION(sbi).fsync_mode =
1013 FSYNC_MODE_NOBARRIER;
1020 case Opt_test_dummy_encryption:
1021 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
1026 case Opt_inlinecrypt:
1027 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
1028 sb->s_flags |= SB_INLINECRYPT;
1030 f2fs_info(sbi, "inline encryption not supported");
1033 case Opt_checkpoint_disable_cap_perc:
1034 if (args->from && match_int(args, &arg))
1036 if (arg < 0 || arg > 100)
1038 F2FS_OPTION(sbi).unusable_cap_perc = arg;
1039 set_opt(sbi, DISABLE_CHECKPOINT);
1041 case Opt_checkpoint_disable_cap:
1042 if (args->from && match_int(args, &arg))
1044 F2FS_OPTION(sbi).unusable_cap = arg;
1045 set_opt(sbi, DISABLE_CHECKPOINT);
1047 case Opt_checkpoint_disable:
1048 set_opt(sbi, DISABLE_CHECKPOINT);
1050 case Opt_checkpoint_enable:
1051 clear_opt(sbi, DISABLE_CHECKPOINT);
1053 case Opt_checkpoint_merge:
1054 set_opt(sbi, MERGE_CHECKPOINT);
1056 case Opt_nocheckpoint_merge:
1057 clear_opt(sbi, MERGE_CHECKPOINT);
1059 #ifdef CONFIG_F2FS_FS_COMPRESSION
1060 case Opt_compress_algorithm:
1061 if (!f2fs_sb_has_compression(sbi)) {
1062 f2fs_info(sbi, "Image doesn't support compression");
1065 name = match_strdup(&args[0]);
1068 if (!strcmp(name, "lzo")) {
1069 #ifdef CONFIG_F2FS_FS_LZO
1070 F2FS_OPTION(sbi).compress_level = 0;
1071 F2FS_OPTION(sbi).compress_algorithm =
1074 f2fs_info(sbi, "kernel doesn't support lzo compression");
1076 } else if (!strncmp(name, "lz4", 3)) {
1077 #ifdef CONFIG_F2FS_FS_LZ4
1078 ret = f2fs_set_lz4hc_level(sbi, name);
1083 F2FS_OPTION(sbi).compress_algorithm =
1086 f2fs_info(sbi, "kernel doesn't support lz4 compression");
1088 } else if (!strncmp(name, "zstd", 4)) {
1089 #ifdef CONFIG_F2FS_FS_ZSTD
1090 ret = f2fs_set_zstd_level(sbi, name);
1095 F2FS_OPTION(sbi).compress_algorithm =
1098 f2fs_info(sbi, "kernel doesn't support zstd compression");
1100 } else if (!strcmp(name, "lzo-rle")) {
1101 #ifdef CONFIG_F2FS_FS_LZORLE
1102 F2FS_OPTION(sbi).compress_level = 0;
1103 F2FS_OPTION(sbi).compress_algorithm =
1106 f2fs_info(sbi, "kernel doesn't support lzorle compression");
1114 case Opt_compress_log_size:
1115 if (!f2fs_sb_has_compression(sbi)) {
1116 f2fs_info(sbi, "Image doesn't support compression");
1119 if (args->from && match_int(args, &arg))
1121 if (arg < MIN_COMPRESS_LOG_SIZE ||
1122 arg > MAX_COMPRESS_LOG_SIZE) {
1124 "Compress cluster log size is out of range");
1127 F2FS_OPTION(sbi).compress_log_size = arg;
1129 case Opt_compress_extension:
1130 if (!f2fs_sb_has_compression(sbi)) {
1131 f2fs_info(sbi, "Image doesn't support compression");
1134 name = match_strdup(&args[0]);
1138 ext = F2FS_OPTION(sbi).extensions;
1139 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
1141 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1142 ext_cnt >= COMPRESS_EXT_NUM) {
1144 "invalid extension length/number");
1149 if (is_compress_extension_exist(sbi, name, true)) {
1154 strcpy(ext[ext_cnt], name);
1155 F2FS_OPTION(sbi).compress_ext_cnt++;
1158 case Opt_nocompress_extension:
1159 if (!f2fs_sb_has_compression(sbi)) {
1160 f2fs_info(sbi, "Image doesn't support compression");
1163 name = match_strdup(&args[0]);
1167 noext = F2FS_OPTION(sbi).noextensions;
1168 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
1170 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1171 noext_cnt >= COMPRESS_EXT_NUM) {
1173 "invalid extension length/number");
1178 if (is_compress_extension_exist(sbi, name, false)) {
1183 strcpy(noext[noext_cnt], name);
1184 F2FS_OPTION(sbi).nocompress_ext_cnt++;
1187 case Opt_compress_chksum:
1188 if (!f2fs_sb_has_compression(sbi)) {
1189 f2fs_info(sbi, "Image doesn't support compression");
1192 F2FS_OPTION(sbi).compress_chksum = true;
1194 case Opt_compress_mode:
1195 if (!f2fs_sb_has_compression(sbi)) {
1196 f2fs_info(sbi, "Image doesn't support compression");
1199 name = match_strdup(&args[0]);
1202 if (!strcmp(name, "fs")) {
1203 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
1204 } else if (!strcmp(name, "user")) {
1205 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER;
1212 case Opt_compress_cache:
1213 if (!f2fs_sb_has_compression(sbi)) {
1214 f2fs_info(sbi, "Image doesn't support compression");
1217 set_opt(sbi, COMPRESS_CACHE);
1220 case Opt_compress_algorithm:
1221 case Opt_compress_log_size:
1222 case Opt_compress_extension:
1223 case Opt_nocompress_extension:
1224 case Opt_compress_chksum:
1225 case Opt_compress_mode:
1226 case Opt_compress_cache:
1227 f2fs_info(sbi, "compression options not supported");
1234 set_opt(sbi, GC_MERGE);
1236 case Opt_nogc_merge:
1237 clear_opt(sbi, GC_MERGE);
1239 case Opt_discard_unit:
1240 name = match_strdup(&args[0]);
1243 if (!strcmp(name, "block")) {
1244 F2FS_OPTION(sbi).discard_unit =
1246 } else if (!strcmp(name, "segment")) {
1247 F2FS_OPTION(sbi).discard_unit =
1248 DISCARD_UNIT_SEGMENT;
1249 } else if (!strcmp(name, "section")) {
1250 F2FS_OPTION(sbi).discard_unit =
1251 DISCARD_UNIT_SECTION;
1258 case Opt_memory_mode:
1259 name = match_strdup(&args[0]);
1262 if (!strcmp(name, "normal")) {
1263 F2FS_OPTION(sbi).memory_mode =
1265 } else if (!strcmp(name, "low")) {
1266 F2FS_OPTION(sbi).memory_mode =
1274 case Opt_age_extent_cache:
1275 set_opt(sbi, AGE_EXTENT_CACHE);
1278 name = match_strdup(&args[0]);
1281 if (!strcmp(name, "remount-ro")) {
1282 F2FS_OPTION(sbi).errors =
1283 MOUNT_ERRORS_READONLY;
1284 } else if (!strcmp(name, "continue")) {
1285 F2FS_OPTION(sbi).errors =
1286 MOUNT_ERRORS_CONTINUE;
1287 } else if (!strcmp(name, "panic")) {
1288 F2FS_OPTION(sbi).errors =
1297 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
1304 if (f2fs_check_quota_options(sbi))
1307 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
1308 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1311 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
1312 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1316 #if !IS_ENABLED(CONFIG_UNICODE)
1317 if (f2fs_sb_has_casefold(sbi)) {
1319 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
1324 * The BLKZONED feature indicates that the drive was formatted with
1325 * zone alignment optimization. This is optional for host-aware
1326 * devices, but mandatory for host-managed zoned block devices.
1328 if (f2fs_sb_has_blkzoned(sbi)) {
1329 #ifdef CONFIG_BLK_DEV_ZONED
1330 if (F2FS_OPTION(sbi).discard_unit !=
1331 DISCARD_UNIT_SECTION) {
1332 f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default");
1333 F2FS_OPTION(sbi).discard_unit =
1334 DISCARD_UNIT_SECTION;
1337 if (F2FS_OPTION(sbi).fs_mode != FS_MODE_LFS) {
1338 f2fs_info(sbi, "Only lfs mode is allowed with zoned block device feature");
1342 f2fs_err(sbi, "Zoned block device support is not enabled");
1347 #ifdef CONFIG_F2FS_FS_COMPRESSION
1348 if (f2fs_test_compress_extension(sbi)) {
1349 f2fs_err(sbi, "invalid compress or nocompress extension");
1354 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
1355 int min_size, max_size;
1357 if (!f2fs_sb_has_extra_attr(sbi) ||
1358 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
1359 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
1362 if (!test_opt(sbi, INLINE_XATTR)) {
1363 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
1367 min_size = MIN_INLINE_XATTR_SIZE;
1368 max_size = MAX_INLINE_XATTR_SIZE;
1370 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
1371 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
1372 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
1373 min_size, max_size);
1378 if (test_opt(sbi, ATGC) && f2fs_lfs_mode(sbi)) {
1379 f2fs_err(sbi, "LFS is not compatible with ATGC");
1383 if (f2fs_is_readonly(sbi) && test_opt(sbi, FLUSH_MERGE)) {
1384 f2fs_err(sbi, "FLUSH_MERGE not compatible with readonly mode");
1388 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) {
1389 f2fs_err(sbi, "Allow to mount readonly mode only");
1395 static struct inode *f2fs_alloc_inode(struct super_block *sb)
1397 struct f2fs_inode_info *fi;
1399 if (time_to_inject(F2FS_SB(sb), FAULT_SLAB_ALLOC))
1402 fi = alloc_inode_sb(sb, f2fs_inode_cachep, GFP_F2FS_ZERO);
1406 init_once((void *) fi);
1408 /* Initialize f2fs-specific inode info */
1409 atomic_set(&fi->dirty_pages, 0);
1410 atomic_set(&fi->i_compr_blocks, 0);
1411 init_f2fs_rwsem(&fi->i_sem);
1412 spin_lock_init(&fi->i_size_lock);
1413 INIT_LIST_HEAD(&fi->dirty_list);
1414 INIT_LIST_HEAD(&fi->gdirty_list);
1415 init_f2fs_rwsem(&fi->i_gc_rwsem[READ]);
1416 init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]);
1417 init_f2fs_rwsem(&fi->i_xattr_sem);
1419 /* Will be used by directory only */
1420 fi->i_dir_level = F2FS_SB(sb)->dir_level;
1422 return &fi->vfs_inode;
1425 static int f2fs_drop_inode(struct inode *inode)
1427 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1431 * during filesystem shutdown, if checkpoint is disabled,
1432 * drop useless meta/node dirty pages.
1434 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1435 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1436 inode->i_ino == F2FS_META_INO(sbi)) {
1437 trace_f2fs_drop_inode(inode, 1);
1443 * This is to avoid a deadlock condition like below.
1444 * writeback_single_inode(inode)
1445 * - f2fs_write_data_page
1446 * - f2fs_gc -> iput -> evict
1447 * - inode_wait_for_writeback(inode)
1449 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1450 if (!inode->i_nlink && !is_bad_inode(inode)) {
1451 /* to avoid evict_inode call simultaneously */
1452 atomic_inc(&inode->i_count);
1453 spin_unlock(&inode->i_lock);
1455 /* should remain fi->extent_tree for writepage */
1456 f2fs_destroy_extent_node(inode);
1458 sb_start_intwrite(inode->i_sb);
1459 f2fs_i_size_write(inode, 0);
1461 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1462 inode, NULL, 0, DATA);
1463 truncate_inode_pages_final(inode->i_mapping);
1465 if (F2FS_HAS_BLOCKS(inode))
1466 f2fs_truncate(inode);
1468 sb_end_intwrite(inode->i_sb);
1470 spin_lock(&inode->i_lock);
1471 atomic_dec(&inode->i_count);
1473 trace_f2fs_drop_inode(inode, 0);
1476 ret = generic_drop_inode(inode);
1478 ret = fscrypt_drop_inode(inode);
1479 trace_f2fs_drop_inode(inode, ret);
1483 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1485 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1488 spin_lock(&sbi->inode_lock[DIRTY_META]);
1489 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1492 set_inode_flag(inode, FI_DIRTY_INODE);
1493 stat_inc_dirty_inode(sbi, DIRTY_META);
1495 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1496 list_add_tail(&F2FS_I(inode)->gdirty_list,
1497 &sbi->inode_list[DIRTY_META]);
1498 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1500 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1504 void f2fs_inode_synced(struct inode *inode)
1506 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1508 spin_lock(&sbi->inode_lock[DIRTY_META]);
1509 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1510 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1513 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1514 list_del_init(&F2FS_I(inode)->gdirty_list);
1515 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1517 clear_inode_flag(inode, FI_DIRTY_INODE);
1518 clear_inode_flag(inode, FI_AUTO_RECOVER);
1519 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1520 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1524 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1526 * We should call set_dirty_inode to write the dirty inode through write_inode.
1528 static void f2fs_dirty_inode(struct inode *inode, int flags)
1530 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1532 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1533 inode->i_ino == F2FS_META_INO(sbi))
1536 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1537 clear_inode_flag(inode, FI_AUTO_RECOVER);
1539 f2fs_inode_dirtied(inode, false);
1542 static void f2fs_free_inode(struct inode *inode)
1544 fscrypt_free_inode(inode);
1545 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1548 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1550 percpu_counter_destroy(&sbi->total_valid_inode_count);
1551 percpu_counter_destroy(&sbi->rf_node_block_count);
1552 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1555 static void destroy_device_list(struct f2fs_sb_info *sbi)
1559 for (i = 0; i < sbi->s_ndevs; i++) {
1561 bdev_release(FDEV(i).bdev_handle);
1562 #ifdef CONFIG_BLK_DEV_ZONED
1563 kvfree(FDEV(i).blkz_seq);
1569 static void f2fs_put_super(struct super_block *sb)
1571 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1576 /* unregister procfs/sysfs entries in advance to avoid race case */
1577 f2fs_unregister_sysfs(sbi);
1579 f2fs_quota_off_umount(sb);
1581 /* prevent remaining shrinker jobs */
1582 mutex_lock(&sbi->umount_mutex);
1585 * flush all issued checkpoints and stop checkpoint issue thread.
1586 * after then, all checkpoints should be done by each process context.
1588 f2fs_stop_ckpt_thread(sbi);
1591 * We don't need to do checkpoint when superblock is clean.
1592 * But, the previous checkpoint was not done by umount, it needs to do
1593 * clean checkpoint again.
1595 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1596 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1597 struct cp_control cpc = {
1598 .reason = CP_UMOUNT,
1600 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1601 err = f2fs_write_checkpoint(sbi, &cpc);
1604 /* be sure to wait for any on-going discard commands */
1605 done = f2fs_issue_discard_timeout(sbi);
1606 if (f2fs_realtime_discard_enable(sbi) && !sbi->discard_blks && done) {
1607 struct cp_control cpc = {
1608 .reason = CP_UMOUNT | CP_TRIMMED,
1610 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1611 err = f2fs_write_checkpoint(sbi, &cpc);
1615 * normally superblock is clean, so we need to release this.
1616 * In addition, EIO will skip do checkpoint, we need this as well.
1618 f2fs_release_ino_entry(sbi, true);
1620 f2fs_leave_shrinker(sbi);
1621 mutex_unlock(&sbi->umount_mutex);
1623 /* our cp_error case, we can wait for any writeback page */
1624 f2fs_flush_merged_writes(sbi);
1626 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1628 if (err || f2fs_cp_error(sbi)) {
1629 truncate_inode_pages_final(NODE_MAPPING(sbi));
1630 truncate_inode_pages_final(META_MAPPING(sbi));
1633 for (i = 0; i < NR_COUNT_TYPE; i++) {
1634 if (!get_pages(sbi, i))
1636 f2fs_err(sbi, "detect filesystem reference count leak during "
1637 "umount, type: %d, count: %lld", i, get_pages(sbi, i));
1638 f2fs_bug_on(sbi, 1);
1641 f2fs_bug_on(sbi, sbi->fsync_node_num);
1643 f2fs_destroy_compress_inode(sbi);
1645 iput(sbi->node_inode);
1646 sbi->node_inode = NULL;
1648 iput(sbi->meta_inode);
1649 sbi->meta_inode = NULL;
1652 * iput() can update stat information, if f2fs_write_checkpoint()
1653 * above failed with error.
1655 f2fs_destroy_stats(sbi);
1657 /* destroy f2fs internal modules */
1658 f2fs_destroy_node_manager(sbi);
1659 f2fs_destroy_segment_manager(sbi);
1661 /* flush s_error_work before sbi destroy */
1662 flush_work(&sbi->s_error_work);
1664 f2fs_destroy_post_read_wq(sbi);
1668 if (sbi->s_chksum_driver)
1669 crypto_free_shash(sbi->s_chksum_driver);
1670 kfree(sbi->raw_super);
1672 f2fs_destroy_page_array_cache(sbi);
1673 f2fs_destroy_xattr_caches(sbi);
1675 for (i = 0; i < MAXQUOTAS; i++)
1676 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1678 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
1679 destroy_percpu_info(sbi);
1680 f2fs_destroy_iostat(sbi);
1681 for (i = 0; i < NR_PAGE_TYPE; i++)
1682 kvfree(sbi->write_io[i]);
1683 #if IS_ENABLED(CONFIG_UNICODE)
1684 utf8_unload(sb->s_encoding);
1688 int f2fs_sync_fs(struct super_block *sb, int sync)
1690 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1693 if (unlikely(f2fs_cp_error(sbi)))
1695 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1698 trace_f2fs_sync_fs(sb, sync);
1700 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1704 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1705 err = f2fs_issue_checkpoint(sbi);
1711 static int f2fs_freeze(struct super_block *sb)
1713 if (f2fs_readonly(sb))
1716 /* IO error happened before */
1717 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1720 /* must be clean, since sync_filesystem() was already called */
1721 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1724 /* Let's flush checkpoints and stop the thread. */
1725 f2fs_flush_ckpt_thread(F2FS_SB(sb));
1727 /* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */
1728 set_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1732 static int f2fs_unfreeze(struct super_block *sb)
1734 clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1739 static int f2fs_statfs_project(struct super_block *sb,
1740 kprojid_t projid, struct kstatfs *buf)
1743 struct dquot *dquot;
1747 qid = make_kqid_projid(projid);
1748 dquot = dqget(sb, qid);
1750 return PTR_ERR(dquot);
1751 spin_lock(&dquot->dq_dqb_lock);
1753 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1754 dquot->dq_dqb.dqb_bhardlimit);
1756 limit >>= sb->s_blocksize_bits;
1758 if (limit && buf->f_blocks > limit) {
1759 curblock = (dquot->dq_dqb.dqb_curspace +
1760 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1761 buf->f_blocks = limit;
1762 buf->f_bfree = buf->f_bavail =
1763 (buf->f_blocks > curblock) ?
1764 (buf->f_blocks - curblock) : 0;
1767 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1768 dquot->dq_dqb.dqb_ihardlimit);
1770 if (limit && buf->f_files > limit) {
1771 buf->f_files = limit;
1773 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1774 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1777 spin_unlock(&dquot->dq_dqb_lock);
1783 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1785 struct super_block *sb = dentry->d_sb;
1786 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1787 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1788 block_t total_count, user_block_count, start_count;
1789 u64 avail_node_count;
1790 unsigned int total_valid_node_count;
1792 total_count = le64_to_cpu(sbi->raw_super->block_count);
1793 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1794 buf->f_type = F2FS_SUPER_MAGIC;
1795 buf->f_bsize = sbi->blocksize;
1797 buf->f_blocks = total_count - start_count;
1799 spin_lock(&sbi->stat_lock);
1801 user_block_count = sbi->user_block_count;
1802 total_valid_node_count = valid_node_count(sbi);
1803 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1804 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1805 sbi->current_reserved_blocks;
1807 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1810 buf->f_bfree -= sbi->unusable_block_count;
1811 spin_unlock(&sbi->stat_lock);
1813 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1814 buf->f_bavail = buf->f_bfree -
1815 F2FS_OPTION(sbi).root_reserved_blocks;
1819 if (avail_node_count > user_block_count) {
1820 buf->f_files = user_block_count;
1821 buf->f_ffree = buf->f_bavail;
1823 buf->f_files = avail_node_count;
1824 buf->f_ffree = min(avail_node_count - total_valid_node_count,
1828 buf->f_namelen = F2FS_NAME_LEN;
1829 buf->f_fsid = u64_to_fsid(id);
1832 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1833 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1834 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1840 static inline void f2fs_show_quota_options(struct seq_file *seq,
1841 struct super_block *sb)
1844 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1846 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1849 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1860 seq_printf(seq, ",jqfmt=%s", fmtname);
1863 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1864 seq_show_option(seq, "usrjquota",
1865 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1867 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1868 seq_show_option(seq, "grpjquota",
1869 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1871 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1872 seq_show_option(seq, "prjjquota",
1873 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1877 #ifdef CONFIG_F2FS_FS_COMPRESSION
1878 static inline void f2fs_show_compress_options(struct seq_file *seq,
1879 struct super_block *sb)
1881 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1885 if (!f2fs_sb_has_compression(sbi))
1888 switch (F2FS_OPTION(sbi).compress_algorithm) {
1898 case COMPRESS_LZORLE:
1899 algtype = "lzo-rle";
1902 seq_printf(seq, ",compress_algorithm=%s", algtype);
1904 if (F2FS_OPTION(sbi).compress_level)
1905 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level);
1907 seq_printf(seq, ",compress_log_size=%u",
1908 F2FS_OPTION(sbi).compress_log_size);
1910 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1911 seq_printf(seq, ",compress_extension=%s",
1912 F2FS_OPTION(sbi).extensions[i]);
1915 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) {
1916 seq_printf(seq, ",nocompress_extension=%s",
1917 F2FS_OPTION(sbi).noextensions[i]);
1920 if (F2FS_OPTION(sbi).compress_chksum)
1921 seq_puts(seq, ",compress_chksum");
1923 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS)
1924 seq_printf(seq, ",compress_mode=%s", "fs");
1925 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER)
1926 seq_printf(seq, ",compress_mode=%s", "user");
1928 if (test_opt(sbi, COMPRESS_CACHE))
1929 seq_puts(seq, ",compress_cache");
1933 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1935 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1937 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1938 seq_printf(seq, ",background_gc=%s", "sync");
1939 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1940 seq_printf(seq, ",background_gc=%s", "on");
1941 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1942 seq_printf(seq, ",background_gc=%s", "off");
1944 if (test_opt(sbi, GC_MERGE))
1945 seq_puts(seq, ",gc_merge");
1947 seq_puts(seq, ",nogc_merge");
1949 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1950 seq_puts(seq, ",disable_roll_forward");
1951 if (test_opt(sbi, NORECOVERY))
1952 seq_puts(seq, ",norecovery");
1953 if (test_opt(sbi, DISCARD)) {
1954 seq_puts(seq, ",discard");
1955 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK)
1956 seq_printf(seq, ",discard_unit=%s", "block");
1957 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT)
1958 seq_printf(seq, ",discard_unit=%s", "segment");
1959 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION)
1960 seq_printf(seq, ",discard_unit=%s", "section");
1962 seq_puts(seq, ",nodiscard");
1964 #ifdef CONFIG_F2FS_FS_XATTR
1965 if (test_opt(sbi, XATTR_USER))
1966 seq_puts(seq, ",user_xattr");
1968 seq_puts(seq, ",nouser_xattr");
1969 if (test_opt(sbi, INLINE_XATTR))
1970 seq_puts(seq, ",inline_xattr");
1972 seq_puts(seq, ",noinline_xattr");
1973 if (test_opt(sbi, INLINE_XATTR_SIZE))
1974 seq_printf(seq, ",inline_xattr_size=%u",
1975 F2FS_OPTION(sbi).inline_xattr_size);
1977 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1978 if (test_opt(sbi, POSIX_ACL))
1979 seq_puts(seq, ",acl");
1981 seq_puts(seq, ",noacl");
1983 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1984 seq_puts(seq, ",disable_ext_identify");
1985 if (test_opt(sbi, INLINE_DATA))
1986 seq_puts(seq, ",inline_data");
1988 seq_puts(seq, ",noinline_data");
1989 if (test_opt(sbi, INLINE_DENTRY))
1990 seq_puts(seq, ",inline_dentry");
1992 seq_puts(seq, ",noinline_dentry");
1993 if (test_opt(sbi, FLUSH_MERGE))
1994 seq_puts(seq, ",flush_merge");
1996 seq_puts(seq, ",noflush_merge");
1997 if (test_opt(sbi, NOBARRIER))
1998 seq_puts(seq, ",nobarrier");
2000 seq_puts(seq, ",barrier");
2001 if (test_opt(sbi, FASTBOOT))
2002 seq_puts(seq, ",fastboot");
2003 if (test_opt(sbi, READ_EXTENT_CACHE))
2004 seq_puts(seq, ",extent_cache");
2006 seq_puts(seq, ",noextent_cache");
2007 if (test_opt(sbi, AGE_EXTENT_CACHE))
2008 seq_puts(seq, ",age_extent_cache");
2009 if (test_opt(sbi, DATA_FLUSH))
2010 seq_puts(seq, ",data_flush");
2012 seq_puts(seq, ",mode=");
2013 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
2014 seq_puts(seq, "adaptive");
2015 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
2016 seq_puts(seq, "lfs");
2017 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG)
2018 seq_puts(seq, "fragment:segment");
2019 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK)
2020 seq_puts(seq, "fragment:block");
2021 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
2022 if (test_opt(sbi, RESERVE_ROOT))
2023 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
2024 F2FS_OPTION(sbi).root_reserved_blocks,
2025 from_kuid_munged(&init_user_ns,
2026 F2FS_OPTION(sbi).s_resuid),
2027 from_kgid_munged(&init_user_ns,
2028 F2FS_OPTION(sbi).s_resgid));
2029 #ifdef CONFIG_F2FS_FAULT_INJECTION
2030 if (test_opt(sbi, FAULT_INJECTION)) {
2031 seq_printf(seq, ",fault_injection=%u",
2032 F2FS_OPTION(sbi).fault_info.inject_rate);
2033 seq_printf(seq, ",fault_type=%u",
2034 F2FS_OPTION(sbi).fault_info.inject_type);
2038 if (test_opt(sbi, QUOTA))
2039 seq_puts(seq, ",quota");
2040 if (test_opt(sbi, USRQUOTA))
2041 seq_puts(seq, ",usrquota");
2042 if (test_opt(sbi, GRPQUOTA))
2043 seq_puts(seq, ",grpquota");
2044 if (test_opt(sbi, PRJQUOTA))
2045 seq_puts(seq, ",prjquota");
2047 f2fs_show_quota_options(seq, sbi->sb);
2049 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
2051 if (sbi->sb->s_flags & SB_INLINECRYPT)
2052 seq_puts(seq, ",inlinecrypt");
2054 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
2055 seq_printf(seq, ",alloc_mode=%s", "default");
2056 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
2057 seq_printf(seq, ",alloc_mode=%s", "reuse");
2059 if (test_opt(sbi, DISABLE_CHECKPOINT))
2060 seq_printf(seq, ",checkpoint=disable:%u",
2061 F2FS_OPTION(sbi).unusable_cap);
2062 if (test_opt(sbi, MERGE_CHECKPOINT))
2063 seq_puts(seq, ",checkpoint_merge");
2065 seq_puts(seq, ",nocheckpoint_merge");
2066 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
2067 seq_printf(seq, ",fsync_mode=%s", "posix");
2068 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
2069 seq_printf(seq, ",fsync_mode=%s", "strict");
2070 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
2071 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
2073 #ifdef CONFIG_F2FS_FS_COMPRESSION
2074 f2fs_show_compress_options(seq, sbi->sb);
2077 if (test_opt(sbi, ATGC))
2078 seq_puts(seq, ",atgc");
2080 if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_NORMAL)
2081 seq_printf(seq, ",memory=%s", "normal");
2082 else if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_LOW)
2083 seq_printf(seq, ",memory=%s", "low");
2085 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_READONLY)
2086 seq_printf(seq, ",errors=%s", "remount-ro");
2087 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE)
2088 seq_printf(seq, ",errors=%s", "continue");
2089 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC)
2090 seq_printf(seq, ",errors=%s", "panic");
2095 static void default_options(struct f2fs_sb_info *sbi, bool remount)
2097 /* init some FS parameters */
2099 set_opt(sbi, READ_EXTENT_CACHE);
2100 clear_opt(sbi, DISABLE_CHECKPOINT);
2102 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi))
2103 set_opt(sbi, DISCARD);
2105 if (f2fs_sb_has_blkzoned(sbi))
2106 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION;
2108 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK;
2111 if (f2fs_sb_has_readonly(sbi))
2112 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE;
2114 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
2116 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
2117 if (le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count_main) <=
2118 SMALL_VOLUME_SEGMENTS)
2119 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
2121 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
2122 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
2123 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
2124 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
2125 if (f2fs_sb_has_compression(sbi)) {
2126 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
2127 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
2128 F2FS_OPTION(sbi).compress_ext_cnt = 0;
2129 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
2131 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
2132 F2FS_OPTION(sbi).memory_mode = MEMORY_MODE_NORMAL;
2133 F2FS_OPTION(sbi).errors = MOUNT_ERRORS_CONTINUE;
2135 sbi->sb->s_flags &= ~SB_INLINECRYPT;
2137 set_opt(sbi, INLINE_XATTR);
2138 set_opt(sbi, INLINE_DATA);
2139 set_opt(sbi, INLINE_DENTRY);
2140 set_opt(sbi, MERGE_CHECKPOINT);
2141 F2FS_OPTION(sbi).unusable_cap = 0;
2142 sbi->sb->s_flags |= SB_LAZYTIME;
2143 if (!f2fs_is_readonly(sbi))
2144 set_opt(sbi, FLUSH_MERGE);
2145 if (f2fs_sb_has_blkzoned(sbi))
2146 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
2148 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
2150 #ifdef CONFIG_F2FS_FS_XATTR
2151 set_opt(sbi, XATTR_USER);
2153 #ifdef CONFIG_F2FS_FS_POSIX_ACL
2154 set_opt(sbi, POSIX_ACL);
2157 f2fs_build_fault_attr(sbi, 0, 0);
2161 static int f2fs_enable_quotas(struct super_block *sb);
2164 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
2166 unsigned int s_flags = sbi->sb->s_flags;
2167 struct cp_control cpc;
2168 unsigned int gc_mode = sbi->gc_mode;
2173 if (s_flags & SB_RDONLY) {
2174 f2fs_err(sbi, "checkpoint=disable on readonly fs");
2177 sbi->sb->s_flags |= SB_ACTIVE;
2179 /* check if we need more GC first */
2180 unusable = f2fs_get_unusable_blocks(sbi);
2181 if (!f2fs_disable_cp_again(sbi, unusable))
2184 f2fs_update_time(sbi, DISABLE_TIME);
2186 sbi->gc_mode = GC_URGENT_HIGH;
2188 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
2189 struct f2fs_gc_control gc_control = {
2190 .victim_segno = NULL_SEGNO,
2191 .init_gc_type = FG_GC,
2192 .should_migrate_blocks = false,
2193 .err_gc_skipped = true,
2195 .nr_free_secs = 1 };
2197 f2fs_down_write(&sbi->gc_lock);
2198 stat_inc_gc_call_count(sbi, FOREGROUND);
2199 err = f2fs_gc(sbi, &gc_control);
2200 if (err == -ENODATA) {
2204 if (err && err != -EAGAIN)
2208 ret = sync_filesystem(sbi->sb);
2210 err = ret ? ret : err;
2214 unusable = f2fs_get_unusable_blocks(sbi);
2215 if (f2fs_disable_cp_again(sbi, unusable)) {
2221 f2fs_down_write(&sbi->gc_lock);
2222 cpc.reason = CP_PAUSE;
2223 set_sbi_flag(sbi, SBI_CP_DISABLED);
2224 stat_inc_cp_call_count(sbi, TOTAL_CALL);
2225 err = f2fs_write_checkpoint(sbi, &cpc);
2229 spin_lock(&sbi->stat_lock);
2230 sbi->unusable_block_count = unusable;
2231 spin_unlock(&sbi->stat_lock);
2234 f2fs_up_write(&sbi->gc_lock);
2236 sbi->gc_mode = gc_mode;
2237 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2241 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
2243 int retry = DEFAULT_RETRY_IO_COUNT;
2245 /* we should flush all the data to keep data consistency */
2247 sync_inodes_sb(sbi->sb);
2248 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2249 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--);
2251 if (unlikely(retry < 0))
2252 f2fs_warn(sbi, "checkpoint=enable has some unwritten data.");
2254 f2fs_down_write(&sbi->gc_lock);
2255 f2fs_dirty_to_prefree(sbi);
2257 clear_sbi_flag(sbi, SBI_CP_DISABLED);
2258 set_sbi_flag(sbi, SBI_IS_DIRTY);
2259 f2fs_up_write(&sbi->gc_lock);
2261 f2fs_sync_fs(sbi->sb, 1);
2263 /* Let's ensure there's no pending checkpoint anymore */
2264 f2fs_flush_ckpt_thread(sbi);
2267 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
2269 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2270 struct f2fs_mount_info org_mount_opt;
2271 unsigned long old_sb_flags;
2273 bool need_restart_gc = false, need_stop_gc = false;
2274 bool need_restart_flush = false, need_stop_flush = false;
2275 bool need_restart_discard = false, need_stop_discard = false;
2276 bool need_enable_checkpoint = false, need_disable_checkpoint = false;
2277 bool no_read_extent_cache = !test_opt(sbi, READ_EXTENT_CACHE);
2278 bool no_age_extent_cache = !test_opt(sbi, AGE_EXTENT_CACHE);
2279 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT);
2280 bool no_atgc = !test_opt(sbi, ATGC);
2281 bool no_discard = !test_opt(sbi, DISCARD);
2282 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE);
2283 bool block_unit_discard = f2fs_block_unit_discard(sbi);
2289 * Save the old mount options in case we
2290 * need to restore them.
2292 org_mount_opt = sbi->mount_opt;
2293 old_sb_flags = sb->s_flags;
2296 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
2297 for (i = 0; i < MAXQUOTAS; i++) {
2298 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2299 org_mount_opt.s_qf_names[i] =
2300 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
2302 if (!org_mount_opt.s_qf_names[i]) {
2303 for (j = 0; j < i; j++)
2304 kfree(org_mount_opt.s_qf_names[j]);
2308 org_mount_opt.s_qf_names[i] = NULL;
2313 /* recover superblocks we couldn't write due to previous RO mount */
2314 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
2315 err = f2fs_commit_super(sbi, false);
2316 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
2319 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2322 default_options(sbi, true);
2324 /* parse mount options */
2325 err = parse_options(sb, data, true);
2329 /* flush outstanding errors before changing fs state */
2330 flush_work(&sbi->s_error_work);
2333 * Previous and new state of filesystem is RO,
2334 * so skip checking GC and FLUSH_MERGE conditions.
2336 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
2339 if (f2fs_dev_is_readonly(sbi) && !(*flags & SB_RDONLY)) {
2345 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
2346 err = dquot_suspend(sb, -1);
2349 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
2350 /* dquot_resume needs RW */
2351 sb->s_flags &= ~SB_RDONLY;
2352 if (sb_any_quota_suspended(sb)) {
2353 dquot_resume(sb, -1);
2354 } else if (f2fs_sb_has_quota_ino(sbi)) {
2355 err = f2fs_enable_quotas(sb);
2361 if (f2fs_lfs_mode(sbi) && !IS_F2FS_IPU_DISABLE(sbi)) {
2363 f2fs_warn(sbi, "LFS is not compatible with IPU");
2367 /* disallow enable atgc dynamically */
2368 if (no_atgc == !!test_opt(sbi, ATGC)) {
2370 f2fs_warn(sbi, "switch atgc option is not allowed");
2374 /* disallow enable/disable extent_cache dynamically */
2375 if (no_read_extent_cache == !!test_opt(sbi, READ_EXTENT_CACHE)) {
2377 f2fs_warn(sbi, "switch extent_cache option is not allowed");
2380 /* disallow enable/disable age extent_cache dynamically */
2381 if (no_age_extent_cache == !!test_opt(sbi, AGE_EXTENT_CACHE)) {
2383 f2fs_warn(sbi, "switch age_extent_cache option is not allowed");
2387 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) {
2389 f2fs_warn(sbi, "switch compress_cache option is not allowed");
2393 if (block_unit_discard != f2fs_block_unit_discard(sbi)) {
2395 f2fs_warn(sbi, "switch discard_unit option is not allowed");
2399 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
2401 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
2406 * We stop the GC thread if FS is mounted as RO
2407 * or if background_gc = off is passed in mount
2408 * option. Also sync the filesystem.
2410 if ((*flags & SB_RDONLY) ||
2411 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF &&
2412 !test_opt(sbi, GC_MERGE))) {
2413 if (sbi->gc_thread) {
2414 f2fs_stop_gc_thread(sbi);
2415 need_restart_gc = true;
2417 } else if (!sbi->gc_thread) {
2418 err = f2fs_start_gc_thread(sbi);
2421 need_stop_gc = true;
2424 if (*flags & SB_RDONLY) {
2427 set_sbi_flag(sbi, SBI_IS_DIRTY);
2428 set_sbi_flag(sbi, SBI_IS_CLOSE);
2429 f2fs_sync_fs(sb, 1);
2430 clear_sbi_flag(sbi, SBI_IS_CLOSE);
2434 * We stop issue flush thread if FS is mounted as RO
2435 * or if flush_merge is not passed in mount option.
2437 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2438 clear_opt(sbi, FLUSH_MERGE);
2439 f2fs_destroy_flush_cmd_control(sbi, false);
2440 need_restart_flush = true;
2442 err = f2fs_create_flush_cmd_control(sbi);
2445 need_stop_flush = true;
2448 if (no_discard == !!test_opt(sbi, DISCARD)) {
2449 if (test_opt(sbi, DISCARD)) {
2450 err = f2fs_start_discard_thread(sbi);
2453 need_stop_discard = true;
2455 f2fs_stop_discard_thread(sbi);
2456 f2fs_issue_discard_timeout(sbi);
2457 need_restart_discard = true;
2461 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) {
2462 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
2463 err = f2fs_disable_checkpoint(sbi);
2465 goto restore_discard;
2466 need_enable_checkpoint = true;
2468 f2fs_enable_checkpoint(sbi);
2469 need_disable_checkpoint = true;
2474 * Place this routine at the end, since a new checkpoint would be
2475 * triggered while remount and we need to take care of it before
2476 * returning from remount.
2478 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) ||
2479 !test_opt(sbi, MERGE_CHECKPOINT)) {
2480 f2fs_stop_ckpt_thread(sbi);
2482 /* Flush if the prevous checkpoint, if exists. */
2483 f2fs_flush_ckpt_thread(sbi);
2485 err = f2fs_start_ckpt_thread(sbi);
2488 "Failed to start F2FS issue_checkpoint_thread (%d)",
2490 goto restore_checkpoint;
2496 /* Release old quota file names */
2497 for (i = 0; i < MAXQUOTAS; i++)
2498 kfree(org_mount_opt.s_qf_names[i]);
2500 /* Update the POSIXACL Flag */
2501 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2502 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2504 limit_reserve_root(sbi);
2505 adjust_unusable_cap_perc(sbi);
2506 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
2509 if (need_enable_checkpoint) {
2510 f2fs_enable_checkpoint(sbi);
2511 } else if (need_disable_checkpoint) {
2512 if (f2fs_disable_checkpoint(sbi))
2513 f2fs_warn(sbi, "checkpoint has not been disabled");
2516 if (need_restart_discard) {
2517 if (f2fs_start_discard_thread(sbi))
2518 f2fs_warn(sbi, "discard has been stopped");
2519 } else if (need_stop_discard) {
2520 f2fs_stop_discard_thread(sbi);
2523 if (need_restart_flush) {
2524 if (f2fs_create_flush_cmd_control(sbi))
2525 f2fs_warn(sbi, "background flush thread has stopped");
2526 } else if (need_stop_flush) {
2527 clear_opt(sbi, FLUSH_MERGE);
2528 f2fs_destroy_flush_cmd_control(sbi, false);
2531 if (need_restart_gc) {
2532 if (f2fs_start_gc_thread(sbi))
2533 f2fs_warn(sbi, "background gc thread has stopped");
2534 } else if (need_stop_gc) {
2535 f2fs_stop_gc_thread(sbi);
2539 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
2540 for (i = 0; i < MAXQUOTAS; i++) {
2541 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2542 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
2545 sbi->mount_opt = org_mount_opt;
2546 sb->s_flags = old_sb_flags;
2551 static bool f2fs_need_recovery(struct f2fs_sb_info *sbi)
2553 /* need to recovery orphan */
2554 if (is_set_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG))
2556 /* need to recovery data */
2557 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
2559 if (test_opt(sbi, NORECOVERY))
2561 return !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG);
2564 static bool f2fs_recover_quota_begin(struct f2fs_sb_info *sbi)
2566 bool readonly = f2fs_readonly(sbi->sb);
2568 if (!f2fs_need_recovery(sbi))
2571 /* it doesn't need to check f2fs_sb_has_readonly() */
2572 if (f2fs_hw_is_readonly(sbi))
2576 sbi->sb->s_flags &= ~SB_RDONLY;
2577 set_sbi_flag(sbi, SBI_IS_WRITABLE);
2581 * Turn on quotas which were not enabled for read-only mounts if
2582 * filesystem has quota feature, so that they are updated correctly.
2584 return f2fs_enable_quota_files(sbi, readonly);
2587 static void f2fs_recover_quota_end(struct f2fs_sb_info *sbi,
2591 f2fs_quota_off_umount(sbi->sb);
2593 if (is_sbi_flag_set(sbi, SBI_IS_WRITABLE)) {
2594 clear_sbi_flag(sbi, SBI_IS_WRITABLE);
2595 sbi->sb->s_flags |= SB_RDONLY;
2599 /* Read data from quotafile */
2600 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
2601 size_t len, loff_t off)
2603 struct inode *inode = sb_dqopt(sb)->files[type];
2604 struct address_space *mapping = inode->i_mapping;
2605 block_t blkidx = F2FS_BYTES_TO_BLK(off);
2606 int offset = off & (sb->s_blocksize - 1);
2609 loff_t i_size = i_size_read(inode);
2615 if (off + len > i_size)
2618 while (toread > 0) {
2619 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
2621 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
2623 if (PTR_ERR(page) == -ENOMEM) {
2624 memalloc_retry_wait(GFP_NOFS);
2627 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2628 return PTR_ERR(page);
2633 if (unlikely(page->mapping != mapping)) {
2634 f2fs_put_page(page, 1);
2637 if (unlikely(!PageUptodate(page))) {
2638 f2fs_put_page(page, 1);
2639 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2643 memcpy_from_page(data, page, offset, tocopy);
2644 f2fs_put_page(page, 1);
2654 /* Write to quotafile */
2655 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2656 const char *data, size_t len, loff_t off)
2658 struct inode *inode = sb_dqopt(sb)->files[type];
2659 struct address_space *mapping = inode->i_mapping;
2660 const struct address_space_operations *a_ops = mapping->a_ops;
2661 int offset = off & (sb->s_blocksize - 1);
2662 size_t towrite = len;
2664 void *fsdata = NULL;
2668 while (towrite > 0) {
2669 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2672 err = a_ops->write_begin(NULL, mapping, off, tocopy,
2674 if (unlikely(err)) {
2675 if (err == -ENOMEM) {
2676 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2679 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2683 memcpy_to_page(page, offset, data, tocopy);
2685 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2696 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
2697 f2fs_mark_inode_dirty_sync(inode, false);
2698 return len - towrite;
2701 int f2fs_dquot_initialize(struct inode *inode)
2703 if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT))
2706 return dquot_initialize(inode);
2709 static struct dquot **f2fs_get_dquots(struct inode *inode)
2711 return F2FS_I(inode)->i_dquot;
2714 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2716 return &F2FS_I(inode)->i_reserved_quota;
2719 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2721 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2722 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2726 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2727 F2FS_OPTION(sbi).s_jquota_fmt, type);
2730 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2735 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2736 err = f2fs_enable_quotas(sbi->sb);
2738 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2744 for (i = 0; i < MAXQUOTAS; i++) {
2745 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2746 err = f2fs_quota_on_mount(sbi, i);
2751 f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2758 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2761 struct inode *qf_inode;
2762 unsigned long qf_inum;
2763 unsigned long qf_flag = F2FS_QUOTA_DEFAULT_FL;
2766 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2768 qf_inum = f2fs_qf_ino(sb, type);
2772 qf_inode = f2fs_iget(sb, qf_inum);
2773 if (IS_ERR(qf_inode)) {
2774 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2775 return PTR_ERR(qf_inode);
2778 /* Don't account quota for quota files to avoid recursion */
2779 inode_lock(qf_inode);
2780 qf_inode->i_flags |= S_NOQUOTA;
2782 if ((F2FS_I(qf_inode)->i_flags & qf_flag) != qf_flag) {
2783 F2FS_I(qf_inode)->i_flags |= qf_flag;
2784 f2fs_set_inode_flags(qf_inode);
2786 inode_unlock(qf_inode);
2788 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2793 static int f2fs_enable_quotas(struct super_block *sb)
2795 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2797 unsigned long qf_inum;
2798 bool quota_mopt[MAXQUOTAS] = {
2799 test_opt(sbi, USRQUOTA),
2800 test_opt(sbi, GRPQUOTA),
2801 test_opt(sbi, PRJQUOTA),
2804 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2805 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2809 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2811 for (type = 0; type < MAXQUOTAS; type++) {
2812 qf_inum = f2fs_qf_ino(sb, type);
2814 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2815 DQUOT_USAGE_ENABLED |
2816 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2818 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2820 for (type--; type >= 0; type--)
2821 dquot_quota_off(sb, type);
2822 set_sbi_flag(F2FS_SB(sb),
2823 SBI_QUOTA_NEED_REPAIR);
2831 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type)
2833 struct quota_info *dqopt = sb_dqopt(sbi->sb);
2834 struct address_space *mapping = dqopt->files[type]->i_mapping;
2837 ret = dquot_writeback_dquots(sbi->sb, type);
2841 ret = filemap_fdatawrite(mapping);
2845 /* if we are using journalled quota */
2846 if (is_journalled_quota(sbi))
2849 ret = filemap_fdatawait(mapping);
2851 truncate_inode_pages(&dqopt->files[type]->i_data, 0);
2854 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2858 int f2fs_quota_sync(struct super_block *sb, int type)
2860 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2861 struct quota_info *dqopt = sb_dqopt(sb);
2866 * Now when everything is written we can discard the pagecache so
2867 * that userspace sees the changes.
2869 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2871 if (type != -1 && cnt != type)
2874 if (!sb_has_quota_active(sb, cnt))
2877 if (!f2fs_sb_has_quota_ino(sbi))
2878 inode_lock(dqopt->files[cnt]);
2883 * f2fs_down_read(quota_sem)
2884 * dquot_writeback_dquots()
2887 * f2fs_down_read(quota_sem)
2890 f2fs_down_read(&sbi->quota_sem);
2892 ret = f2fs_quota_sync_file(sbi, cnt);
2894 f2fs_up_read(&sbi->quota_sem);
2895 f2fs_unlock_op(sbi);
2897 if (!f2fs_sb_has_quota_ino(sbi))
2898 inode_unlock(dqopt->files[cnt]);
2906 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2907 const struct path *path)
2909 struct inode *inode;
2912 /* if quota sysfile exists, deny enabling quota with specific file */
2913 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2914 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2918 if (path->dentry->d_sb != sb)
2921 err = f2fs_quota_sync(sb, type);
2925 inode = d_inode(path->dentry);
2927 err = filemap_fdatawrite(inode->i_mapping);
2931 err = filemap_fdatawait(inode->i_mapping);
2935 err = dquot_quota_on(sb, type, format_id, path);
2940 F2FS_I(inode)->i_flags |= F2FS_QUOTA_DEFAULT_FL;
2941 f2fs_set_inode_flags(inode);
2942 inode_unlock(inode);
2943 f2fs_mark_inode_dirty_sync(inode, false);
2948 static int __f2fs_quota_off(struct super_block *sb, int type)
2950 struct inode *inode = sb_dqopt(sb)->files[type];
2953 if (!inode || !igrab(inode))
2954 return dquot_quota_off(sb, type);
2956 err = f2fs_quota_sync(sb, type);
2960 err = dquot_quota_off(sb, type);
2961 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2965 F2FS_I(inode)->i_flags &= ~F2FS_QUOTA_DEFAULT_FL;
2966 f2fs_set_inode_flags(inode);
2967 inode_unlock(inode);
2968 f2fs_mark_inode_dirty_sync(inode, false);
2974 static int f2fs_quota_off(struct super_block *sb, int type)
2976 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2979 err = __f2fs_quota_off(sb, type);
2982 * quotactl can shutdown journalled quota, result in inconsistence
2983 * between quota record and fs data by following updates, tag the
2984 * flag to let fsck be aware of it.
2986 if (is_journalled_quota(sbi))
2987 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2991 void f2fs_quota_off_umount(struct super_block *sb)
2996 for (type = 0; type < MAXQUOTAS; type++) {
2997 err = __f2fs_quota_off(sb, type);
2999 int ret = dquot_quota_off(sb, type);
3001 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
3003 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
3007 * In case of checkpoint=disable, we must flush quota blocks.
3008 * This can cause NULL exception for node_inode in end_io, since
3009 * put_super already dropped it.
3011 sync_filesystem(sb);
3014 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
3016 struct quota_info *dqopt = sb_dqopt(sb);
3019 for (type = 0; type < MAXQUOTAS; type++) {
3020 if (!dqopt->files[type])
3022 f2fs_inode_synced(dqopt->files[type]);
3026 static int f2fs_dquot_commit(struct dquot *dquot)
3028 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3031 f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
3032 ret = dquot_commit(dquot);
3034 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3035 f2fs_up_read(&sbi->quota_sem);
3039 static int f2fs_dquot_acquire(struct dquot *dquot)
3041 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3044 f2fs_down_read(&sbi->quota_sem);
3045 ret = dquot_acquire(dquot);
3047 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3048 f2fs_up_read(&sbi->quota_sem);
3052 static int f2fs_dquot_release(struct dquot *dquot)
3054 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3055 int ret = dquot_release(dquot);
3058 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3062 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
3064 struct super_block *sb = dquot->dq_sb;
3065 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3066 int ret = dquot_mark_dquot_dirty(dquot);
3068 /* if we are using journalled quota */
3069 if (is_journalled_quota(sbi))
3070 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
3075 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
3077 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3078 int ret = dquot_commit_info(sb, type);
3081 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3085 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
3087 *projid = F2FS_I(inode)->i_projid;
3091 static const struct dquot_operations f2fs_quota_operations = {
3092 .get_reserved_space = f2fs_get_reserved_space,
3093 .write_dquot = f2fs_dquot_commit,
3094 .acquire_dquot = f2fs_dquot_acquire,
3095 .release_dquot = f2fs_dquot_release,
3096 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
3097 .write_info = f2fs_dquot_commit_info,
3098 .alloc_dquot = dquot_alloc,
3099 .destroy_dquot = dquot_destroy,
3100 .get_projid = f2fs_get_projid,
3101 .get_next_id = dquot_get_next_id,
3104 static const struct quotactl_ops f2fs_quotactl_ops = {
3105 .quota_on = f2fs_quota_on,
3106 .quota_off = f2fs_quota_off,
3107 .quota_sync = f2fs_quota_sync,
3108 .get_state = dquot_get_state,
3109 .set_info = dquot_set_dqinfo,
3110 .get_dqblk = dquot_get_dqblk,
3111 .set_dqblk = dquot_set_dqblk,
3112 .get_nextdqblk = dquot_get_next_dqblk,
3115 int f2fs_dquot_initialize(struct inode *inode)
3120 int f2fs_quota_sync(struct super_block *sb, int type)
3125 void f2fs_quota_off_umount(struct super_block *sb)
3130 static const struct super_operations f2fs_sops = {
3131 .alloc_inode = f2fs_alloc_inode,
3132 .free_inode = f2fs_free_inode,
3133 .drop_inode = f2fs_drop_inode,
3134 .write_inode = f2fs_write_inode,
3135 .dirty_inode = f2fs_dirty_inode,
3136 .show_options = f2fs_show_options,
3138 .quota_read = f2fs_quota_read,
3139 .quota_write = f2fs_quota_write,
3140 .get_dquots = f2fs_get_dquots,
3142 .evict_inode = f2fs_evict_inode,
3143 .put_super = f2fs_put_super,
3144 .sync_fs = f2fs_sync_fs,
3145 .freeze_fs = f2fs_freeze,
3146 .unfreeze_fs = f2fs_unfreeze,
3147 .statfs = f2fs_statfs,
3148 .remount_fs = f2fs_remount,
3151 #ifdef CONFIG_FS_ENCRYPTION
3152 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
3154 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3155 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3159 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
3162 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3165 * Encrypting the root directory is not allowed because fsck
3166 * expects lost+found directory to exist and remain unencrypted
3167 * if LOST_FOUND feature is enabled.
3170 if (f2fs_sb_has_lost_found(sbi) &&
3171 inode->i_ino == F2FS_ROOT_INO(sbi))
3174 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3175 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3176 ctx, len, fs_data, XATTR_CREATE);
3179 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb)
3181 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy;
3184 static bool f2fs_has_stable_inodes(struct super_block *sb)
3189 static struct block_device **f2fs_get_devices(struct super_block *sb,
3190 unsigned int *num_devs)
3192 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3193 struct block_device **devs;
3196 if (!f2fs_is_multi_device(sbi))
3199 devs = kmalloc_array(sbi->s_ndevs, sizeof(*devs), GFP_KERNEL);
3201 return ERR_PTR(-ENOMEM);
3203 for (i = 0; i < sbi->s_ndevs; i++)
3204 devs[i] = FDEV(i).bdev;
3205 *num_devs = sbi->s_ndevs;
3209 static const struct fscrypt_operations f2fs_cryptops = {
3210 .needs_bounce_pages = 1,
3211 .has_32bit_inodes = 1,
3212 .supports_subblock_data_units = 1,
3213 .legacy_key_prefix = "f2fs:",
3214 .get_context = f2fs_get_context,
3215 .set_context = f2fs_set_context,
3216 .get_dummy_policy = f2fs_get_dummy_policy,
3217 .empty_dir = f2fs_empty_dir,
3218 .has_stable_inodes = f2fs_has_stable_inodes,
3219 .get_devices = f2fs_get_devices,
3223 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
3224 u64 ino, u32 generation)
3226 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3227 struct inode *inode;
3229 if (f2fs_check_nid_range(sbi, ino))
3230 return ERR_PTR(-ESTALE);
3233 * f2fs_iget isn't quite right if the inode is currently unallocated!
3234 * However f2fs_iget currently does appropriate checks to handle stale
3235 * inodes so everything is OK.
3237 inode = f2fs_iget(sb, ino);
3239 return ERR_CAST(inode);
3240 if (unlikely(generation && inode->i_generation != generation)) {
3241 /* we didn't find the right inode.. */
3243 return ERR_PTR(-ESTALE);
3248 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
3249 int fh_len, int fh_type)
3251 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
3252 f2fs_nfs_get_inode);
3255 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
3256 int fh_len, int fh_type)
3258 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
3259 f2fs_nfs_get_inode);
3262 static const struct export_operations f2fs_export_ops = {
3263 .encode_fh = generic_encode_ino32_fh,
3264 .fh_to_dentry = f2fs_fh_to_dentry,
3265 .fh_to_parent = f2fs_fh_to_parent,
3266 .get_parent = f2fs_get_parent,
3269 loff_t max_file_blocks(struct inode *inode)
3275 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
3276 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
3277 * space in inode.i_addr, it will be more safe to reassign
3281 if (inode && f2fs_compressed_file(inode))
3282 leaf_count = ADDRS_PER_BLOCK(inode);
3284 leaf_count = DEF_ADDRS_PER_BLOCK;
3286 /* two direct node blocks */
3287 result += (leaf_count * 2);
3289 /* two indirect node blocks */
3290 leaf_count *= NIDS_PER_BLOCK;
3291 result += (leaf_count * 2);
3293 /* one double indirect node block */
3294 leaf_count *= NIDS_PER_BLOCK;
3295 result += leaf_count;
3298 * For compatibility with FSCRYPT_POLICY_FLAG_IV_INO_LBLK_{64,32} with
3299 * a 4K crypto data unit, we must restrict the max filesize to what can
3300 * fit within U32_MAX + 1 data units.
3303 result = min(result, (((loff_t)U32_MAX + 1) * 4096) >> F2FS_BLKSIZE_BITS);
3308 static int __f2fs_commit_super(struct buffer_head *bh,
3309 struct f2fs_super_block *super)
3313 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
3314 set_buffer_dirty(bh);
3317 /* it's rare case, we can do fua all the time */
3318 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
3321 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
3322 struct buffer_head *bh)
3324 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3325 (bh->b_data + F2FS_SUPER_OFFSET);
3326 struct super_block *sb = sbi->sb;
3327 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3328 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
3329 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
3330 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
3331 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
3332 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3333 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
3334 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
3335 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
3336 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
3337 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3338 u32 segment_count = le32_to_cpu(raw_super->segment_count);
3339 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3340 u64 main_end_blkaddr = main_blkaddr +
3341 (segment_count_main << log_blocks_per_seg);
3342 u64 seg_end_blkaddr = segment0_blkaddr +
3343 (segment_count << log_blocks_per_seg);
3345 if (segment0_blkaddr != cp_blkaddr) {
3346 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
3347 segment0_blkaddr, cp_blkaddr);
3351 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
3353 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
3354 cp_blkaddr, sit_blkaddr,
3355 segment_count_ckpt << log_blocks_per_seg);
3359 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
3361 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
3362 sit_blkaddr, nat_blkaddr,
3363 segment_count_sit << log_blocks_per_seg);
3367 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
3369 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
3370 nat_blkaddr, ssa_blkaddr,
3371 segment_count_nat << log_blocks_per_seg);
3375 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
3377 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
3378 ssa_blkaddr, main_blkaddr,
3379 segment_count_ssa << log_blocks_per_seg);
3383 if (main_end_blkaddr > seg_end_blkaddr) {
3384 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)",
3385 main_blkaddr, seg_end_blkaddr,
3386 segment_count_main << log_blocks_per_seg);
3388 } else if (main_end_blkaddr < seg_end_blkaddr) {
3392 /* fix in-memory information all the time */
3393 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
3394 segment0_blkaddr) >> log_blocks_per_seg);
3396 if (f2fs_readonly(sb) || f2fs_hw_is_readonly(sbi)) {
3397 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3400 err = __f2fs_commit_super(bh, NULL);
3401 res = err ? "failed" : "done";
3403 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)",
3404 res, main_blkaddr, seg_end_blkaddr,
3405 segment_count_main << log_blocks_per_seg);
3412 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
3413 struct buffer_head *bh)
3415 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main;
3416 block_t total_sections, blocks_per_seg;
3417 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3418 (bh->b_data + F2FS_SUPER_OFFSET);
3419 size_t crc_offset = 0;
3422 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
3423 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
3424 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
3428 /* Check checksum_offset and crc in superblock */
3429 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
3430 crc_offset = le32_to_cpu(raw_super->checksum_offset);
3432 offsetof(struct f2fs_super_block, crc)) {
3433 f2fs_info(sbi, "Invalid SB checksum offset: %zu",
3435 return -EFSCORRUPTED;
3437 crc = le32_to_cpu(raw_super->crc);
3438 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
3439 f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
3440 return -EFSCORRUPTED;
3444 /* Currently, support only 4KB block size */
3445 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) {
3446 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u",
3447 le32_to_cpu(raw_super->log_blocksize),
3449 return -EFSCORRUPTED;
3452 /* check log blocks per segment */
3453 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
3454 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
3455 le32_to_cpu(raw_super->log_blocks_per_seg));
3456 return -EFSCORRUPTED;
3459 /* Currently, support 512/1024/2048/4096/16K bytes sector size */
3460 if (le32_to_cpu(raw_super->log_sectorsize) >
3461 F2FS_MAX_LOG_SECTOR_SIZE ||
3462 le32_to_cpu(raw_super->log_sectorsize) <
3463 F2FS_MIN_LOG_SECTOR_SIZE) {
3464 f2fs_info(sbi, "Invalid log sectorsize (%u)",
3465 le32_to_cpu(raw_super->log_sectorsize));
3466 return -EFSCORRUPTED;
3468 if (le32_to_cpu(raw_super->log_sectors_per_block) +
3469 le32_to_cpu(raw_super->log_sectorsize) !=
3470 F2FS_MAX_LOG_SECTOR_SIZE) {
3471 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
3472 le32_to_cpu(raw_super->log_sectors_per_block),
3473 le32_to_cpu(raw_super->log_sectorsize));
3474 return -EFSCORRUPTED;
3477 segment_count = le32_to_cpu(raw_super->segment_count);
3478 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3479 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3480 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3481 total_sections = le32_to_cpu(raw_super->section_count);
3483 /* blocks_per_seg should be 512, given the above check */
3484 blocks_per_seg = BIT(le32_to_cpu(raw_super->log_blocks_per_seg));
3486 if (segment_count > F2FS_MAX_SEGMENT ||
3487 segment_count < F2FS_MIN_SEGMENTS) {
3488 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
3489 return -EFSCORRUPTED;
3492 if (total_sections > segment_count_main || total_sections < 1 ||
3493 segs_per_sec > segment_count || !segs_per_sec) {
3494 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
3495 segment_count, total_sections, segs_per_sec);
3496 return -EFSCORRUPTED;
3499 if (segment_count_main != total_sections * segs_per_sec) {
3500 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)",
3501 segment_count_main, total_sections, segs_per_sec);
3502 return -EFSCORRUPTED;
3505 if ((segment_count / segs_per_sec) < total_sections) {
3506 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
3507 segment_count, segs_per_sec, total_sections);
3508 return -EFSCORRUPTED;
3511 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
3512 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
3513 segment_count, le64_to_cpu(raw_super->block_count));
3514 return -EFSCORRUPTED;
3517 if (RDEV(0).path[0]) {
3518 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
3521 while (i < MAX_DEVICES && RDEV(i).path[0]) {
3522 dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
3525 if (segment_count != dev_seg_count) {
3526 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
3527 segment_count, dev_seg_count);
3528 return -EFSCORRUPTED;
3531 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
3532 !bdev_is_zoned(sbi->sb->s_bdev)) {
3533 f2fs_info(sbi, "Zoned block device path is missing");
3534 return -EFSCORRUPTED;
3538 if (secs_per_zone > total_sections || !secs_per_zone) {
3539 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
3540 secs_per_zone, total_sections);
3541 return -EFSCORRUPTED;
3543 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
3544 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
3545 (le32_to_cpu(raw_super->extension_count) +
3546 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
3547 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
3548 le32_to_cpu(raw_super->extension_count),
3549 raw_super->hot_ext_count,
3550 F2FS_MAX_EXTENSION);
3551 return -EFSCORRUPTED;
3554 if (le32_to_cpu(raw_super->cp_payload) >=
3555 (blocks_per_seg - F2FS_CP_PACKS -
3556 NR_CURSEG_PERSIST_TYPE)) {
3557 f2fs_info(sbi, "Insane cp_payload (%u >= %u)",
3558 le32_to_cpu(raw_super->cp_payload),
3559 blocks_per_seg - F2FS_CP_PACKS -
3560 NR_CURSEG_PERSIST_TYPE);
3561 return -EFSCORRUPTED;
3564 /* check reserved ino info */
3565 if (le32_to_cpu(raw_super->node_ino) != 1 ||
3566 le32_to_cpu(raw_super->meta_ino) != 2 ||
3567 le32_to_cpu(raw_super->root_ino) != 3) {
3568 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
3569 le32_to_cpu(raw_super->node_ino),
3570 le32_to_cpu(raw_super->meta_ino),
3571 le32_to_cpu(raw_super->root_ino));
3572 return -EFSCORRUPTED;
3575 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
3576 if (sanity_check_area_boundary(sbi, bh))
3577 return -EFSCORRUPTED;
3582 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
3584 unsigned int total, fsmeta;
3585 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3586 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3587 unsigned int ovp_segments, reserved_segments;
3588 unsigned int main_segs, blocks_per_seg;
3589 unsigned int sit_segs, nat_segs;
3590 unsigned int sit_bitmap_size, nat_bitmap_size;
3591 unsigned int log_blocks_per_seg;
3592 unsigned int segment_count_main;
3593 unsigned int cp_pack_start_sum, cp_payload;
3594 block_t user_block_count, valid_user_blocks;
3595 block_t avail_node_count, valid_node_count;
3596 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks;
3599 total = le32_to_cpu(raw_super->segment_count);
3600 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
3601 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
3603 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
3605 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
3606 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
3608 if (unlikely(fsmeta >= total))
3611 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3612 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3614 if (!f2fs_sb_has_readonly(sbi) &&
3615 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS ||
3616 ovp_segments == 0 || reserved_segments == 0)) {
3617 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
3620 user_block_count = le64_to_cpu(ckpt->user_block_count);
3621 segment_count_main = le32_to_cpu(raw_super->segment_count_main) +
3622 (f2fs_sb_has_readonly(sbi) ? 1 : 0);
3623 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3624 if (!user_block_count || user_block_count >=
3625 segment_count_main << log_blocks_per_seg) {
3626 f2fs_err(sbi, "Wrong user_block_count: %u",
3631 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
3632 if (valid_user_blocks > user_block_count) {
3633 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
3634 valid_user_blocks, user_block_count);
3638 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
3639 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
3640 if (valid_node_count > avail_node_count) {
3641 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
3642 valid_node_count, avail_node_count);
3646 main_segs = le32_to_cpu(raw_super->segment_count_main);
3647 blocks_per_seg = BLKS_PER_SEG(sbi);
3649 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3650 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
3651 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
3654 if (f2fs_sb_has_readonly(sbi))
3657 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
3658 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3659 le32_to_cpu(ckpt->cur_node_segno[j])) {
3660 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
3662 le32_to_cpu(ckpt->cur_node_segno[i]));
3668 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
3669 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
3670 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
3673 if (f2fs_sb_has_readonly(sbi))
3676 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
3677 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
3678 le32_to_cpu(ckpt->cur_data_segno[j])) {
3679 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
3681 le32_to_cpu(ckpt->cur_data_segno[i]));
3686 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3687 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
3688 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3689 le32_to_cpu(ckpt->cur_data_segno[j])) {
3690 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3692 le32_to_cpu(ckpt->cur_node_segno[i]));
3698 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3699 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3701 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3702 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3703 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3704 sit_bitmap_size, nat_bitmap_size);
3708 cp_pack_start_sum = __start_sum_addr(sbi);
3709 cp_payload = __cp_payload(sbi);
3710 if (cp_pack_start_sum < cp_payload + 1 ||
3711 cp_pack_start_sum > blocks_per_seg - 1 -
3712 NR_CURSEG_PERSIST_TYPE) {
3713 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3718 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3719 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3720 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3721 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3722 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3723 le32_to_cpu(ckpt->checksum_offset));
3727 nat_blocks = nat_segs << log_blocks_per_seg;
3728 nat_bits_bytes = nat_blocks / BITS_PER_BYTE;
3729 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
3730 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) &&
3731 (cp_payload + F2FS_CP_PACKS +
3732 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) {
3733 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)",
3734 cp_payload, nat_bits_blocks);
3738 if (unlikely(f2fs_cp_error(sbi))) {
3739 f2fs_err(sbi, "A bug case: need to run fsck");
3745 static void init_sb_info(struct f2fs_sb_info *sbi)
3747 struct f2fs_super_block *raw_super = sbi->raw_super;
3750 sbi->log_sectors_per_block =
3751 le32_to_cpu(raw_super->log_sectors_per_block);
3752 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3753 sbi->blocksize = BIT(sbi->log_blocksize);
3754 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3755 sbi->blocks_per_seg = BIT(sbi->log_blocks_per_seg);
3756 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3757 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3758 sbi->total_sections = le32_to_cpu(raw_super->section_count);
3759 sbi->total_node_count = SEGS_TO_BLKS(sbi,
3760 ((le32_to_cpu(raw_super->segment_count_nat) / 2) *
3761 NAT_ENTRY_PER_BLOCK));
3762 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino);
3763 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino);
3764 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino);
3765 sbi->cur_victim_sec = NULL_SECNO;
3766 sbi->gc_mode = GC_NORMAL;
3767 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3768 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3769 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3770 sbi->migration_granularity = SEGS_PER_SEC(sbi);
3771 sbi->seq_file_ra_mul = MIN_RA_MUL;
3772 sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE;
3773 sbi->max_fragment_hole = DEF_FRAGMENT_SIZE;
3774 spin_lock_init(&sbi->gc_remaining_trials_lock);
3775 atomic64_set(&sbi->current_atomic_write, 0);
3777 sbi->dir_level = DEF_DIR_LEVEL;
3778 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3779 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3780 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3781 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3782 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3783 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3784 DEF_UMOUNT_DISCARD_TIMEOUT;
3785 clear_sbi_flag(sbi, SBI_NEED_FSCK);
3787 for (i = 0; i < NR_COUNT_TYPE; i++)
3788 atomic_set(&sbi->nr_pages[i], 0);
3790 for (i = 0; i < META; i++)
3791 atomic_set(&sbi->wb_sync_req[i], 0);
3793 INIT_LIST_HEAD(&sbi->s_list);
3794 mutex_init(&sbi->umount_mutex);
3795 init_f2fs_rwsem(&sbi->io_order_lock);
3796 spin_lock_init(&sbi->cp_lock);
3798 sbi->dirty_device = 0;
3799 spin_lock_init(&sbi->dev_lock);
3801 init_f2fs_rwsem(&sbi->sb_lock);
3802 init_f2fs_rwsem(&sbi->pin_sem);
3805 static int init_percpu_info(struct f2fs_sb_info *sbi)
3809 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3813 err = percpu_counter_init(&sbi->rf_node_block_count, 0, GFP_KERNEL);
3815 goto err_valid_block;
3817 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3820 goto err_node_block;
3824 percpu_counter_destroy(&sbi->rf_node_block_count);
3826 percpu_counter_destroy(&sbi->alloc_valid_block_count);
3830 #ifdef CONFIG_BLK_DEV_ZONED
3832 struct f2fs_report_zones_args {
3833 struct f2fs_sb_info *sbi;
3834 struct f2fs_dev_info *dev;
3837 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3840 struct f2fs_report_zones_args *rz_args = data;
3841 block_t unusable_blocks = (zone->len - zone->capacity) >>
3842 F2FS_LOG_SECTORS_PER_BLOCK;
3844 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3847 set_bit(idx, rz_args->dev->blkz_seq);
3848 if (!rz_args->sbi->unusable_blocks_per_sec) {
3849 rz_args->sbi->unusable_blocks_per_sec = unusable_blocks;
3852 if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) {
3853 f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n");
3859 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3861 struct block_device *bdev = FDEV(devi).bdev;
3862 sector_t nr_sectors = bdev_nr_sectors(bdev);
3863 struct f2fs_report_zones_args rep_zone_arg;
3867 if (!f2fs_sb_has_blkzoned(sbi))
3870 zone_sectors = bdev_zone_sectors(bdev);
3871 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3872 SECTOR_TO_BLOCK(zone_sectors))
3874 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(zone_sectors);
3875 FDEV(devi).nr_blkz = div_u64(SECTOR_TO_BLOCK(nr_sectors),
3876 sbi->blocks_per_blkz);
3877 if (nr_sectors & (zone_sectors - 1))
3878 FDEV(devi).nr_blkz++;
3880 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3881 BITS_TO_LONGS(FDEV(devi).nr_blkz)
3882 * sizeof(unsigned long),
3884 if (!FDEV(devi).blkz_seq)
3887 rep_zone_arg.sbi = sbi;
3888 rep_zone_arg.dev = &FDEV(devi);
3890 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3899 * Read f2fs raw super block.
3900 * Because we have two copies of super block, so read both of them
3901 * to get the first valid one. If any one of them is broken, we pass
3902 * them recovery flag back to the caller.
3904 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3905 struct f2fs_super_block **raw_super,
3906 int *valid_super_block, int *recovery)
3908 struct super_block *sb = sbi->sb;
3910 struct buffer_head *bh;
3911 struct f2fs_super_block *super;
3914 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3918 for (block = 0; block < 2; block++) {
3919 bh = sb_bread(sb, block);
3921 f2fs_err(sbi, "Unable to read %dth superblock",
3928 /* sanity checking of raw super */
3929 err = sanity_check_raw_super(sbi, bh);
3931 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3939 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3941 *valid_super_block = block;
3947 /* No valid superblock */
3956 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3958 struct buffer_head *bh;
3962 if ((recover && f2fs_readonly(sbi->sb)) ||
3963 f2fs_hw_is_readonly(sbi)) {
3964 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3968 /* we should update superblock crc here */
3969 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3970 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3971 offsetof(struct f2fs_super_block, crc));
3972 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3975 /* write back-up superblock first */
3976 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3979 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3982 /* if we are in recovery path, skip writing valid superblock */
3986 /* write current valid superblock */
3987 bh = sb_bread(sbi->sb, sbi->valid_super_block);
3990 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3995 static void save_stop_reason(struct f2fs_sb_info *sbi, unsigned char reason)
3997 unsigned long flags;
3999 spin_lock_irqsave(&sbi->error_lock, flags);
4000 if (sbi->stop_reason[reason] < GENMASK(BITS_PER_BYTE - 1, 0))
4001 sbi->stop_reason[reason]++;
4002 spin_unlock_irqrestore(&sbi->error_lock, flags);
4005 static void f2fs_record_stop_reason(struct f2fs_sb_info *sbi)
4007 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
4008 unsigned long flags;
4011 f2fs_down_write(&sbi->sb_lock);
4013 spin_lock_irqsave(&sbi->error_lock, flags);
4014 if (sbi->error_dirty) {
4015 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
4017 sbi->error_dirty = false;
4019 memcpy(raw_super->s_stop_reason, sbi->stop_reason, MAX_STOP_REASON);
4020 spin_unlock_irqrestore(&sbi->error_lock, flags);
4022 err = f2fs_commit_super(sbi, false);
4024 f2fs_up_write(&sbi->sb_lock);
4026 f2fs_err_ratelimited(sbi,
4027 "f2fs_commit_super fails to record stop_reason, err:%d",
4031 void f2fs_save_errors(struct f2fs_sb_info *sbi, unsigned char flag)
4033 unsigned long flags;
4035 spin_lock_irqsave(&sbi->error_lock, flags);
4036 if (!test_bit(flag, (unsigned long *)sbi->errors)) {
4037 set_bit(flag, (unsigned long *)sbi->errors);
4038 sbi->error_dirty = true;
4040 spin_unlock_irqrestore(&sbi->error_lock, flags);
4043 static bool f2fs_update_errors(struct f2fs_sb_info *sbi)
4045 unsigned long flags;
4046 bool need_update = false;
4048 spin_lock_irqsave(&sbi->error_lock, flags);
4049 if (sbi->error_dirty) {
4050 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
4052 sbi->error_dirty = false;
4055 spin_unlock_irqrestore(&sbi->error_lock, flags);
4060 static void f2fs_record_errors(struct f2fs_sb_info *sbi, unsigned char error)
4064 f2fs_down_write(&sbi->sb_lock);
4066 if (!f2fs_update_errors(sbi))
4069 err = f2fs_commit_super(sbi, false);
4071 f2fs_err_ratelimited(sbi,
4072 "f2fs_commit_super fails to record errors:%u, err:%d",
4075 f2fs_up_write(&sbi->sb_lock);
4078 void f2fs_handle_error(struct f2fs_sb_info *sbi, unsigned char error)
4080 f2fs_save_errors(sbi, error);
4081 f2fs_record_errors(sbi, error);
4084 void f2fs_handle_error_async(struct f2fs_sb_info *sbi, unsigned char error)
4086 f2fs_save_errors(sbi, error);
4088 if (!sbi->error_dirty)
4090 if (!test_bit(error, (unsigned long *)sbi->errors))
4092 schedule_work(&sbi->s_error_work);
4095 static bool system_going_down(void)
4097 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
4098 || system_state == SYSTEM_RESTART;
4101 void f2fs_handle_critical_error(struct f2fs_sb_info *sbi, unsigned char reason,
4104 struct super_block *sb = sbi->sb;
4105 bool shutdown = reason == STOP_CP_REASON_SHUTDOWN;
4106 bool continue_fs = !shutdown &&
4107 F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE;
4109 set_ckpt_flags(sbi, CP_ERROR_FLAG);
4111 if (!f2fs_hw_is_readonly(sbi)) {
4112 save_stop_reason(sbi, reason);
4114 if (irq_context && !shutdown)
4115 schedule_work(&sbi->s_error_work);
4117 f2fs_record_stop_reason(sbi);
4121 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
4122 * could panic during 'reboot -f' as the underlying device got already
4125 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC &&
4126 !shutdown && !system_going_down() &&
4127 !is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN))
4128 panic("F2FS-fs (device %s): panic forced after error\n",
4132 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
4134 /* continue filesystem operators if errors=continue */
4135 if (continue_fs || f2fs_readonly(sb))
4138 f2fs_warn(sbi, "Remounting filesystem read-only");
4140 * Make sure updated value of ->s_mount_flags will be visible before
4144 sb->s_flags |= SB_RDONLY;
4147 static void f2fs_record_error_work(struct work_struct *work)
4149 struct f2fs_sb_info *sbi = container_of(work,
4150 struct f2fs_sb_info, s_error_work);
4152 f2fs_record_stop_reason(sbi);
4155 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
4157 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
4158 unsigned int max_devices = MAX_DEVICES;
4159 unsigned int logical_blksize;
4160 blk_mode_t mode = sb_open_mode(sbi->sb->s_flags);
4163 /* Initialize single device information */
4164 if (!RDEV(0).path[0]) {
4165 if (!bdev_is_zoned(sbi->sb->s_bdev))
4171 * Initialize multiple devices information, or single
4172 * zoned block device information.
4174 sbi->devs = f2fs_kzalloc(sbi,
4175 array_size(max_devices,
4176 sizeof(struct f2fs_dev_info)),
4181 logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev);
4182 sbi->aligned_blksize = true;
4184 for (i = 0; i < max_devices; i++) {
4186 FDEV(0).bdev_handle = sbi->sb->s_bdev_handle;
4187 else if (!RDEV(i).path[0])
4190 if (max_devices > 1) {
4191 /* Multi-device mount */
4192 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
4193 FDEV(i).total_segments =
4194 le32_to_cpu(RDEV(i).total_segments);
4196 FDEV(i).start_blk = 0;
4197 FDEV(i).end_blk = FDEV(i).start_blk +
4199 FDEV(i).total_segments) - 1 +
4200 le32_to_cpu(raw_super->segment0_blkaddr);
4202 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
4203 FDEV(i).end_blk = FDEV(i).start_blk +
4205 FDEV(i).total_segments) - 1;
4206 FDEV(i).bdev_handle = bdev_open_by_path(
4207 FDEV(i).path, mode, sbi->sb, NULL);
4210 if (IS_ERR(FDEV(i).bdev_handle))
4211 return PTR_ERR(FDEV(i).bdev_handle);
4213 FDEV(i).bdev = FDEV(i).bdev_handle->bdev;
4214 /* to release errored devices */
4215 sbi->s_ndevs = i + 1;
4217 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev))
4218 sbi->aligned_blksize = false;
4220 #ifdef CONFIG_BLK_DEV_ZONED
4221 if (bdev_is_zoned(FDEV(i).bdev)) {
4222 if (!f2fs_sb_has_blkzoned(sbi)) {
4223 f2fs_err(sbi, "Zoned block device feature not enabled");
4226 if (init_blkz_info(sbi, i)) {
4227 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
4230 if (max_devices == 1)
4232 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: Host-managed)",
4234 FDEV(i).total_segments,
4235 FDEV(i).start_blk, FDEV(i).end_blk);
4239 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
4241 FDEV(i).total_segments,
4242 FDEV(i).start_blk, FDEV(i).end_blk);
4247 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
4249 #if IS_ENABLED(CONFIG_UNICODE)
4250 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
4251 const struct f2fs_sb_encodings *encoding_info;
4252 struct unicode_map *encoding;
4253 __u16 encoding_flags;
4255 encoding_info = f2fs_sb_read_encoding(sbi->raw_super);
4256 if (!encoding_info) {
4258 "Encoding requested by superblock is unknown");
4262 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags);
4263 encoding = utf8_load(encoding_info->version);
4264 if (IS_ERR(encoding)) {
4266 "can't mount with superblock charset: %s-%u.%u.%u "
4267 "not supported by the kernel. flags: 0x%x.",
4268 encoding_info->name,
4269 unicode_major(encoding_info->version),
4270 unicode_minor(encoding_info->version),
4271 unicode_rev(encoding_info->version),
4273 return PTR_ERR(encoding);
4275 f2fs_info(sbi, "Using encoding defined by superblock: "
4276 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4277 unicode_major(encoding_info->version),
4278 unicode_minor(encoding_info->version),
4279 unicode_rev(encoding_info->version),
4282 sbi->sb->s_encoding = encoding;
4283 sbi->sb->s_encoding_flags = encoding_flags;
4286 if (f2fs_sb_has_casefold(sbi)) {
4287 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
4294 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
4296 /* adjust parameters according to the volume size */
4297 if (MAIN_SEGS(sbi) <= SMALL_VOLUME_SEGMENTS) {
4298 if (f2fs_block_unit_discard(sbi))
4299 SM_I(sbi)->dcc_info->discard_granularity =
4300 MIN_DISCARD_GRANULARITY;
4301 if (!f2fs_lfs_mode(sbi))
4302 SM_I(sbi)->ipu_policy = BIT(F2FS_IPU_FORCE) |
4303 BIT(F2FS_IPU_HONOR_OPU_WRITE);
4306 sbi->readdir_ra = true;
4309 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
4311 struct f2fs_sb_info *sbi;
4312 struct f2fs_super_block *raw_super;
4315 bool skip_recovery = false, need_fsck = false;
4316 char *options = NULL;
4317 int recovery, i, valid_super_block;
4318 struct curseg_info *seg_i;
4321 bool quota_enabled = false;
4327 valid_super_block = -1;
4330 /* allocate memory for f2fs-specific super block info */
4331 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
4337 /* initialize locks within allocated memory */
4338 init_f2fs_rwsem(&sbi->gc_lock);
4339 mutex_init(&sbi->writepages);
4340 init_f2fs_rwsem(&sbi->cp_global_sem);
4341 init_f2fs_rwsem(&sbi->node_write);
4342 init_f2fs_rwsem(&sbi->node_change);
4343 spin_lock_init(&sbi->stat_lock);
4344 init_f2fs_rwsem(&sbi->cp_rwsem);
4345 init_f2fs_rwsem(&sbi->quota_sem);
4346 init_waitqueue_head(&sbi->cp_wait);
4347 spin_lock_init(&sbi->error_lock);
4349 for (i = 0; i < NR_INODE_TYPE; i++) {
4350 INIT_LIST_HEAD(&sbi->inode_list[i]);
4351 spin_lock_init(&sbi->inode_lock[i]);
4353 mutex_init(&sbi->flush_lock);
4355 /* Load the checksum driver */
4356 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
4357 if (IS_ERR(sbi->s_chksum_driver)) {
4358 f2fs_err(sbi, "Cannot load crc32 driver.");
4359 err = PTR_ERR(sbi->s_chksum_driver);
4360 sbi->s_chksum_driver = NULL;
4364 /* set a block size */
4365 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
4366 f2fs_err(sbi, "unable to set blocksize");
4370 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
4375 sb->s_fs_info = sbi;
4376 sbi->raw_super = raw_super;
4378 INIT_WORK(&sbi->s_error_work, f2fs_record_error_work);
4379 memcpy(sbi->errors, raw_super->s_errors, MAX_F2FS_ERRORS);
4380 memcpy(sbi->stop_reason, raw_super->s_stop_reason, MAX_STOP_REASON);
4382 /* precompute checksum seed for metadata */
4383 if (f2fs_sb_has_inode_chksum(sbi))
4384 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
4385 sizeof(raw_super->uuid));
4387 default_options(sbi, false);
4388 /* parse mount options */
4389 options = kstrdup((const char *)data, GFP_KERNEL);
4390 if (data && !options) {
4395 err = parse_options(sb, options, false);
4399 sb->s_maxbytes = max_file_blocks(NULL) <<
4400 le32_to_cpu(raw_super->log_blocksize);
4401 sb->s_max_links = F2FS_LINK_MAX;
4403 err = f2fs_setup_casefold(sbi);
4408 sb->dq_op = &f2fs_quota_operations;
4409 sb->s_qcop = &f2fs_quotactl_ops;
4410 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4412 if (f2fs_sb_has_quota_ino(sbi)) {
4413 for (i = 0; i < MAXQUOTAS; i++) {
4414 if (f2fs_qf_ino(sbi->sb, i))
4415 sbi->nquota_files++;
4420 sb->s_op = &f2fs_sops;
4421 #ifdef CONFIG_FS_ENCRYPTION
4422 sb->s_cop = &f2fs_cryptops;
4424 #ifdef CONFIG_FS_VERITY
4425 sb->s_vop = &f2fs_verityops;
4427 sb->s_xattr = f2fs_xattr_handlers;
4428 sb->s_export_op = &f2fs_export_ops;
4429 sb->s_magic = F2FS_SUPER_MAGIC;
4430 sb->s_time_gran = 1;
4431 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4432 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
4433 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
4434 sb->s_iflags |= SB_I_CGROUPWB;
4436 /* init f2fs-specific super block info */
4437 sbi->valid_super_block = valid_super_block;
4439 /* disallow all the data/node/meta page writes */
4440 set_sbi_flag(sbi, SBI_POR_DOING);
4442 err = f2fs_init_write_merge_io(sbi);
4448 err = f2fs_init_iostat(sbi);
4452 err = init_percpu_info(sbi);
4456 /* init per sbi slab cache */
4457 err = f2fs_init_xattr_caches(sbi);
4460 err = f2fs_init_page_array_cache(sbi);
4462 goto free_xattr_cache;
4464 /* get an inode for meta space */
4465 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
4466 if (IS_ERR(sbi->meta_inode)) {
4467 f2fs_err(sbi, "Failed to read F2FS meta data inode");
4468 err = PTR_ERR(sbi->meta_inode);
4469 goto free_page_array_cache;
4472 err = f2fs_get_valid_checkpoint(sbi);
4474 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
4475 goto free_meta_inode;
4478 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
4479 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
4480 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
4481 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4482 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
4485 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
4486 set_sbi_flag(sbi, SBI_NEED_FSCK);
4488 /* Initialize device list */
4489 err = f2fs_scan_devices(sbi);
4491 f2fs_err(sbi, "Failed to find devices");
4495 err = f2fs_init_post_read_wq(sbi);
4497 f2fs_err(sbi, "Failed to initialize post read workqueue");
4501 sbi->total_valid_node_count =
4502 le32_to_cpu(sbi->ckpt->valid_node_count);
4503 percpu_counter_set(&sbi->total_valid_inode_count,
4504 le32_to_cpu(sbi->ckpt->valid_inode_count));
4505 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
4506 sbi->total_valid_block_count =
4507 le64_to_cpu(sbi->ckpt->valid_block_count);
4508 sbi->last_valid_block_count = sbi->total_valid_block_count;
4509 sbi->reserved_blocks = 0;
4510 sbi->current_reserved_blocks = 0;
4511 limit_reserve_root(sbi);
4512 adjust_unusable_cap_perc(sbi);
4514 f2fs_init_extent_cache_info(sbi);
4516 f2fs_init_ino_entry_info(sbi);
4518 f2fs_init_fsync_node_info(sbi);
4520 /* setup checkpoint request control and start checkpoint issue thread */
4521 f2fs_init_ckpt_req_control(sbi);
4522 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) &&
4523 test_opt(sbi, MERGE_CHECKPOINT)) {
4524 err = f2fs_start_ckpt_thread(sbi);
4527 "Failed to start F2FS issue_checkpoint_thread (%d)",
4529 goto stop_ckpt_thread;
4533 /* setup f2fs internal modules */
4534 err = f2fs_build_segment_manager(sbi);
4536 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
4540 err = f2fs_build_node_manager(sbi);
4542 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
4547 /* For write statistics */
4548 sbi->sectors_written_start = f2fs_get_sectors_written(sbi);
4550 /* Read accumulated write IO statistics if exists */
4551 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
4552 if (__exist_node_summaries(sbi))
4553 sbi->kbytes_written =
4554 le64_to_cpu(seg_i->journal->info.kbytes_written);
4556 f2fs_build_gc_manager(sbi);
4558 err = f2fs_build_stats(sbi);
4562 /* get an inode for node space */
4563 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
4564 if (IS_ERR(sbi->node_inode)) {
4565 f2fs_err(sbi, "Failed to read node inode");
4566 err = PTR_ERR(sbi->node_inode);
4570 /* read root inode and dentry */
4571 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
4573 f2fs_err(sbi, "Failed to read root inode");
4574 err = PTR_ERR(root);
4575 goto free_node_inode;
4577 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
4578 !root->i_size || !root->i_nlink) {
4581 goto free_node_inode;
4584 sb->s_root = d_make_root(root); /* allocate root dentry */
4587 goto free_node_inode;
4590 err = f2fs_init_compress_inode(sbi);
4592 goto free_root_inode;
4594 err = f2fs_register_sysfs(sbi);
4596 goto free_compress_inode;
4599 /* Enable quota usage during mount */
4600 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
4601 err = f2fs_enable_quotas(sb);
4603 f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
4606 quota_enabled = f2fs_recover_quota_begin(sbi);
4608 /* if there are any orphan inodes, free them */
4609 err = f2fs_recover_orphan_inodes(sbi);
4613 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
4614 goto reset_checkpoint;
4616 /* recover fsynced data */
4617 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
4618 !test_opt(sbi, NORECOVERY)) {
4620 * mount should be failed, when device has readonly mode, and
4621 * previous checkpoint was not done by clean system shutdown.
4623 if (f2fs_hw_is_readonly(sbi)) {
4624 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4625 err = f2fs_recover_fsync_data(sbi, true);
4628 f2fs_err(sbi, "Need to recover fsync data, but "
4629 "write access unavailable, please try "
4630 "mount w/ disable_roll_forward or norecovery");
4635 f2fs_info(sbi, "write access unavailable, skipping recovery");
4636 goto reset_checkpoint;
4640 set_sbi_flag(sbi, SBI_NEED_FSCK);
4643 goto reset_checkpoint;
4645 err = f2fs_recover_fsync_data(sbi, false);
4648 skip_recovery = true;
4650 f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
4655 err = f2fs_recover_fsync_data(sbi, true);
4657 if (!f2fs_readonly(sb) && err > 0) {
4659 f2fs_err(sbi, "Need to recover fsync data");
4665 f2fs_recover_quota_end(sbi, quota_enabled);
4669 * If the f2fs is not readonly and fsync data recovery succeeds,
4670 * check zoned block devices' write pointer consistency.
4672 if (f2fs_sb_has_blkzoned(sbi) && !f2fs_readonly(sb)) {
4675 f2fs_notice(sbi, "Checking entire write pointers");
4676 err2 = f2fs_check_write_pointer(sbi);
4683 f2fs_init_inmem_curseg(sbi);
4685 /* f2fs_recover_fsync_data() cleared this already */
4686 clear_sbi_flag(sbi, SBI_POR_DOING);
4688 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
4689 err = f2fs_disable_checkpoint(sbi);
4691 goto sync_free_meta;
4692 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
4693 f2fs_enable_checkpoint(sbi);
4697 * If filesystem is not mounted as read-only then
4698 * do start the gc_thread.
4700 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF ||
4701 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) {
4702 /* After POR, we can run background GC thread.*/
4703 err = f2fs_start_gc_thread(sbi);
4705 goto sync_free_meta;
4709 /* recover broken superblock */
4711 err = f2fs_commit_super(sbi, true);
4712 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
4713 sbi->valid_super_block ? 1 : 2, err);
4716 f2fs_join_shrinker(sbi);
4718 f2fs_tuning_parameters(sbi);
4720 f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
4721 cur_cp_version(F2FS_CKPT(sbi)));
4722 f2fs_update_time(sbi, CP_TIME);
4723 f2fs_update_time(sbi, REQ_TIME);
4724 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4728 /* safe to flush all the data */
4729 sync_filesystem(sbi->sb);
4734 f2fs_truncate_quota_inode_pages(sb);
4735 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
4736 f2fs_quota_off_umount(sbi->sb);
4739 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
4740 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
4741 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
4742 * falls into an infinite loop in f2fs_sync_meta_pages().
4744 truncate_inode_pages_final(META_MAPPING(sbi));
4745 /* evict some inodes being cached by GC */
4747 f2fs_unregister_sysfs(sbi);
4748 free_compress_inode:
4749 f2fs_destroy_compress_inode(sbi);
4754 f2fs_release_ino_entry(sbi, true);
4755 truncate_inode_pages_final(NODE_MAPPING(sbi));
4756 iput(sbi->node_inode);
4757 sbi->node_inode = NULL;
4759 f2fs_destroy_stats(sbi);
4761 /* stop discard thread before destroying node manager */
4762 f2fs_stop_discard_thread(sbi);
4763 f2fs_destroy_node_manager(sbi);
4765 f2fs_destroy_segment_manager(sbi);
4767 f2fs_stop_ckpt_thread(sbi);
4768 /* flush s_error_work before sbi destroy */
4769 flush_work(&sbi->s_error_work);
4770 f2fs_destroy_post_read_wq(sbi);
4772 destroy_device_list(sbi);
4775 make_bad_inode(sbi->meta_inode);
4776 iput(sbi->meta_inode);
4777 sbi->meta_inode = NULL;
4778 free_page_array_cache:
4779 f2fs_destroy_page_array_cache(sbi);
4781 f2fs_destroy_xattr_caches(sbi);
4783 destroy_percpu_info(sbi);
4785 f2fs_destroy_iostat(sbi);
4787 for (i = 0; i < NR_PAGE_TYPE; i++)
4788 kvfree(sbi->write_io[i]);
4790 #if IS_ENABLED(CONFIG_UNICODE)
4791 utf8_unload(sb->s_encoding);
4792 sb->s_encoding = NULL;
4796 for (i = 0; i < MAXQUOTAS; i++)
4797 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
4799 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
4804 if (sbi->s_chksum_driver)
4805 crypto_free_shash(sbi->s_chksum_driver);
4807 sb->s_fs_info = NULL;
4809 /* give only one another chance */
4810 if (retry_cnt > 0 && skip_recovery) {
4812 shrink_dcache_sb(sb);
4818 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
4819 const char *dev_name, void *data)
4821 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
4824 static void kill_f2fs_super(struct super_block *sb)
4826 struct f2fs_sb_info *sbi = F2FS_SB(sb);
4829 set_sbi_flag(sbi, SBI_IS_CLOSE);
4830 f2fs_stop_gc_thread(sbi);
4831 f2fs_stop_discard_thread(sbi);
4833 #ifdef CONFIG_F2FS_FS_COMPRESSION
4835 * latter evict_inode() can bypass checking and invalidating
4836 * compress inode cache.
4838 if (test_opt(sbi, COMPRESS_CACHE))
4839 truncate_inode_pages_final(COMPRESS_MAPPING(sbi));
4842 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
4843 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4844 struct cp_control cpc = {
4845 .reason = CP_UMOUNT,
4847 stat_inc_cp_call_count(sbi, TOTAL_CALL);
4848 f2fs_write_checkpoint(sbi, &cpc);
4851 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
4852 sb->s_flags &= ~SB_RDONLY;
4854 kill_block_super(sb);
4855 /* Release block devices last, after fscrypt_destroy_keyring(). */
4857 destroy_device_list(sbi);
4859 sb->s_fs_info = NULL;
4863 static struct file_system_type f2fs_fs_type = {
4864 .owner = THIS_MODULE,
4866 .mount = f2fs_mount,
4867 .kill_sb = kill_f2fs_super,
4868 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
4870 MODULE_ALIAS_FS("f2fs");
4872 static int __init init_inodecache(void)
4874 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
4875 sizeof(struct f2fs_inode_info), 0,
4876 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
4877 return f2fs_inode_cachep ? 0 : -ENOMEM;
4880 static void destroy_inodecache(void)
4883 * Make sure all delayed rcu free inodes are flushed before we
4887 kmem_cache_destroy(f2fs_inode_cachep);
4890 static int __init init_f2fs_fs(void)
4894 if (PAGE_SIZE != F2FS_BLKSIZE) {
4895 printk("F2FS not supported on PAGE_SIZE(%lu) != BLOCK_SIZE(%lu)\n",
4896 PAGE_SIZE, F2FS_BLKSIZE);
4900 err = init_inodecache();
4903 err = f2fs_create_node_manager_caches();
4905 goto free_inodecache;
4906 err = f2fs_create_segment_manager_caches();
4908 goto free_node_manager_caches;
4909 err = f2fs_create_checkpoint_caches();
4911 goto free_segment_manager_caches;
4912 err = f2fs_create_recovery_cache();
4914 goto free_checkpoint_caches;
4915 err = f2fs_create_extent_cache();
4917 goto free_recovery_cache;
4918 err = f2fs_create_garbage_collection_cache();
4920 goto free_extent_cache;
4921 err = f2fs_init_sysfs();
4923 goto free_garbage_collection_cache;
4924 err = f2fs_init_shrinker();
4927 err = register_filesystem(&f2fs_fs_type);
4930 f2fs_create_root_stats();
4931 err = f2fs_init_post_read_processing();
4933 goto free_root_stats;
4934 err = f2fs_init_iostat_processing();
4936 goto free_post_read;
4937 err = f2fs_init_bio_entry_cache();
4940 err = f2fs_init_bioset();
4942 goto free_bio_entry_cache;
4943 err = f2fs_init_compress_mempool();
4946 err = f2fs_init_compress_cache();
4948 goto free_compress_mempool;
4949 err = f2fs_create_casefold_cache();
4951 goto free_compress_cache;
4953 free_compress_cache:
4954 f2fs_destroy_compress_cache();
4955 free_compress_mempool:
4956 f2fs_destroy_compress_mempool();
4958 f2fs_destroy_bioset();
4959 free_bio_entry_cache:
4960 f2fs_destroy_bio_entry_cache();
4962 f2fs_destroy_iostat_processing();
4964 f2fs_destroy_post_read_processing();
4966 f2fs_destroy_root_stats();
4967 unregister_filesystem(&f2fs_fs_type);
4969 f2fs_exit_shrinker();
4972 free_garbage_collection_cache:
4973 f2fs_destroy_garbage_collection_cache();
4975 f2fs_destroy_extent_cache();
4976 free_recovery_cache:
4977 f2fs_destroy_recovery_cache();
4978 free_checkpoint_caches:
4979 f2fs_destroy_checkpoint_caches();
4980 free_segment_manager_caches:
4981 f2fs_destroy_segment_manager_caches();
4982 free_node_manager_caches:
4983 f2fs_destroy_node_manager_caches();
4985 destroy_inodecache();
4990 static void __exit exit_f2fs_fs(void)
4992 f2fs_destroy_casefold_cache();
4993 f2fs_destroy_compress_cache();
4994 f2fs_destroy_compress_mempool();
4995 f2fs_destroy_bioset();
4996 f2fs_destroy_bio_entry_cache();
4997 f2fs_destroy_iostat_processing();
4998 f2fs_destroy_post_read_processing();
4999 f2fs_destroy_root_stats();
5000 unregister_filesystem(&f2fs_fs_type);
5001 f2fs_exit_shrinker();
5003 f2fs_destroy_garbage_collection_cache();
5004 f2fs_destroy_extent_cache();
5005 f2fs_destroy_recovery_cache();
5006 f2fs_destroy_checkpoint_caches();
5007 f2fs_destroy_segment_manager_caches();
5008 f2fs_destroy_node_manager_caches();
5009 destroy_inodecache();
5012 module_init(init_f2fs_fs)
5013 module_exit(exit_f2fs_fs)
5015 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
5016 MODULE_DESCRIPTION("Flash Friendly File System");
5017 MODULE_LICENSE("GPL");
5018 MODULE_SOFTDEP("pre: crc32");