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/statfs.h>
12 #include <linux/buffer_head.h>
13 #include <linux/backing-dev.h>
14 #include <linux/kthread.h>
15 #include <linux/parser.h>
16 #include <linux/mount.h>
17 #include <linux/seq_file.h>
18 #include <linux/proc_fs.h>
19 #include <linux/random.h>
20 #include <linux/exportfs.h>
21 #include <linux/blkdev.h>
22 #include <linux/quotaops.h>
23 #include <linux/f2fs_fs.h>
24 #include <linux/sysfs.h>
25 #include <linux/quota.h>
26 #include <linux/unicode.h>
27 #include <linux/part_stat.h>
36 #define CREATE_TRACE_POINTS
37 #include <trace/events/f2fs.h>
39 static struct kmem_cache *f2fs_inode_cachep;
41 #ifdef CONFIG_F2FS_FAULT_INJECTION
43 const char *f2fs_fault_name[FAULT_MAX] = {
44 [FAULT_KMALLOC] = "kmalloc",
45 [FAULT_KVMALLOC] = "kvmalloc",
46 [FAULT_PAGE_ALLOC] = "page alloc",
47 [FAULT_PAGE_GET] = "page get",
48 [FAULT_ALLOC_BIO] = "alloc bio",
49 [FAULT_ALLOC_NID] = "alloc nid",
50 [FAULT_ORPHAN] = "orphan",
51 [FAULT_BLOCK] = "no more block",
52 [FAULT_DIR_DEPTH] = "too big dir depth",
53 [FAULT_EVICT_INODE] = "evict_inode fail",
54 [FAULT_TRUNCATE] = "truncate fail",
55 [FAULT_READ_IO] = "read IO error",
56 [FAULT_CHECKPOINT] = "checkpoint error",
57 [FAULT_DISCARD] = "discard error",
58 [FAULT_WRITE_IO] = "write IO error",
61 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
64 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
67 atomic_set(&ffi->inject_ops, 0);
68 ffi->inject_rate = rate;
72 ffi->inject_type = type;
75 memset(ffi, 0, sizeof(struct f2fs_fault_info));
79 /* f2fs-wide shrinker description */
80 static struct shrinker f2fs_shrinker_info = {
81 .scan_objects = f2fs_shrink_scan,
82 .count_objects = f2fs_shrink_count,
83 .seeks = DEFAULT_SEEKS,
88 Opt_disable_roll_forward,
99 Opt_disable_ext_identify,
102 Opt_inline_xattr_size,
140 Opt_test_dummy_encryption,
142 Opt_checkpoint_disable,
143 Opt_checkpoint_disable_cap,
144 Opt_checkpoint_disable_cap_perc,
145 Opt_checkpoint_enable,
146 Opt_compress_algorithm,
147 Opt_compress_log_size,
148 Opt_compress_extension,
152 static match_table_t f2fs_tokens = {
153 {Opt_gc_background, "background_gc=%s"},
154 {Opt_disable_roll_forward, "disable_roll_forward"},
155 {Opt_norecovery, "norecovery"},
156 {Opt_discard, "discard"},
157 {Opt_nodiscard, "nodiscard"},
158 {Opt_noheap, "no_heap"},
160 {Opt_user_xattr, "user_xattr"},
161 {Opt_nouser_xattr, "nouser_xattr"},
163 {Opt_noacl, "noacl"},
164 {Opt_active_logs, "active_logs=%u"},
165 {Opt_disable_ext_identify, "disable_ext_identify"},
166 {Opt_inline_xattr, "inline_xattr"},
167 {Opt_noinline_xattr, "noinline_xattr"},
168 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
169 {Opt_inline_data, "inline_data"},
170 {Opt_inline_dentry, "inline_dentry"},
171 {Opt_noinline_dentry, "noinline_dentry"},
172 {Opt_flush_merge, "flush_merge"},
173 {Opt_noflush_merge, "noflush_merge"},
174 {Opt_nobarrier, "nobarrier"},
175 {Opt_fastboot, "fastboot"},
176 {Opt_extent_cache, "extent_cache"},
177 {Opt_noextent_cache, "noextent_cache"},
178 {Opt_noinline_data, "noinline_data"},
179 {Opt_data_flush, "data_flush"},
180 {Opt_reserve_root, "reserve_root=%u"},
181 {Opt_resgid, "resgid=%u"},
182 {Opt_resuid, "resuid=%u"},
183 {Opt_mode, "mode=%s"},
184 {Opt_io_size_bits, "io_bits=%u"},
185 {Opt_fault_injection, "fault_injection=%u"},
186 {Opt_fault_type, "fault_type=%u"},
187 {Opt_lazytime, "lazytime"},
188 {Opt_nolazytime, "nolazytime"},
189 {Opt_quota, "quota"},
190 {Opt_noquota, "noquota"},
191 {Opt_usrquota, "usrquota"},
192 {Opt_grpquota, "grpquota"},
193 {Opt_prjquota, "prjquota"},
194 {Opt_usrjquota, "usrjquota=%s"},
195 {Opt_grpjquota, "grpjquota=%s"},
196 {Opt_prjjquota, "prjjquota=%s"},
197 {Opt_offusrjquota, "usrjquota="},
198 {Opt_offgrpjquota, "grpjquota="},
199 {Opt_offprjjquota, "prjjquota="},
200 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
201 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
202 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
203 {Opt_whint, "whint_mode=%s"},
204 {Opt_alloc, "alloc_mode=%s"},
205 {Opt_fsync, "fsync_mode=%s"},
206 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
207 {Opt_test_dummy_encryption, "test_dummy_encryption"},
208 {Opt_inlinecrypt, "inlinecrypt"},
209 {Opt_checkpoint_disable, "checkpoint=disable"},
210 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
211 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
212 {Opt_checkpoint_enable, "checkpoint=enable"},
213 {Opt_compress_algorithm, "compress_algorithm=%s"},
214 {Opt_compress_log_size, "compress_log_size=%u"},
215 {Opt_compress_extension, "compress_extension=%s"},
219 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
221 struct va_format vaf;
227 level = printk_get_level(fmt);
228 vaf.fmt = printk_skip_level(fmt);
230 printk("%c%cF2FS-fs (%s): %pV\n",
231 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
236 #ifdef CONFIG_UNICODE
237 static const struct f2fs_sb_encodings {
241 } f2fs_sb_encoding_map[] = {
242 {F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"},
245 static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb,
246 const struct f2fs_sb_encodings **encoding,
249 __u16 magic = le16_to_cpu(sb->s_encoding);
252 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
253 if (magic == f2fs_sb_encoding_map[i].magic)
256 if (i >= ARRAY_SIZE(f2fs_sb_encoding_map))
259 *encoding = &f2fs_sb_encoding_map[i];
260 *flags = le16_to_cpu(sb->s_encoding_flags);
266 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
268 block_t limit = min((sbi->user_block_count << 1) / 1000,
269 sbi->user_block_count - sbi->reserved_blocks);
272 if (test_opt(sbi, RESERVE_ROOT) &&
273 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
274 F2FS_OPTION(sbi).root_reserved_blocks = limit;
275 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
276 F2FS_OPTION(sbi).root_reserved_blocks);
278 if (!test_opt(sbi, RESERVE_ROOT) &&
279 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
280 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
281 !gid_eq(F2FS_OPTION(sbi).s_resgid,
282 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
283 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
284 from_kuid_munged(&init_user_ns,
285 F2FS_OPTION(sbi).s_resuid),
286 from_kgid_munged(&init_user_ns,
287 F2FS_OPTION(sbi).s_resgid));
290 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
292 if (!F2FS_OPTION(sbi).unusable_cap_perc)
295 if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
296 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
298 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
299 F2FS_OPTION(sbi).unusable_cap_perc;
301 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
302 F2FS_OPTION(sbi).unusable_cap,
303 F2FS_OPTION(sbi).unusable_cap_perc);
306 static void init_once(void *foo)
308 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
310 inode_init_once(&fi->vfs_inode);
314 static const char * const quotatypes[] = INITQFNAMES;
315 #define QTYPE2NAME(t) (quotatypes[t])
316 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
319 struct f2fs_sb_info *sbi = F2FS_SB(sb);
323 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
324 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
327 if (f2fs_sb_has_quota_ino(sbi)) {
328 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
332 qname = match_strdup(args);
334 f2fs_err(sbi, "Not enough memory for storing quotafile name");
337 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
338 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
341 f2fs_err(sbi, "%s quota file already specified",
345 if (strchr(qname, '/')) {
346 f2fs_err(sbi, "quotafile must be on filesystem root");
349 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
357 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
359 struct f2fs_sb_info *sbi = F2FS_SB(sb);
361 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
362 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
365 kvfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
366 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
370 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
373 * We do the test below only for project quotas. 'usrquota' and
374 * 'grpquota' mount options are allowed even without quota feature
375 * to support legacy quotas in quota files.
377 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
378 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
381 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
382 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
383 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
384 if (test_opt(sbi, USRQUOTA) &&
385 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
386 clear_opt(sbi, USRQUOTA);
388 if (test_opt(sbi, GRPQUOTA) &&
389 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
390 clear_opt(sbi, GRPQUOTA);
392 if (test_opt(sbi, PRJQUOTA) &&
393 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
394 clear_opt(sbi, PRJQUOTA);
396 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
397 test_opt(sbi, PRJQUOTA)) {
398 f2fs_err(sbi, "old and new quota format mixing");
402 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
403 f2fs_err(sbi, "journaled quota format not specified");
408 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
409 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
410 F2FS_OPTION(sbi).s_jquota_fmt = 0;
416 static int f2fs_set_test_dummy_encryption(struct super_block *sb,
418 const substring_t *arg,
421 struct f2fs_sb_info *sbi = F2FS_SB(sb);
422 #ifdef CONFIG_FS_ENCRYPTION
425 if (!f2fs_sb_has_encrypt(sbi)) {
426 f2fs_err(sbi, "Encrypt feature is off");
431 * This mount option is just for testing, and it's not worthwhile to
432 * implement the extra complexity (e.g. RCU protection) that would be
433 * needed to allow it to be set or changed during remount. We do allow
434 * it to be specified during remount, but only if there is no change.
436 if (is_remount && !F2FS_OPTION(sbi).dummy_enc_ctx.ctx) {
437 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
440 err = fscrypt_set_test_dummy_encryption(
441 sb, arg, &F2FS_OPTION(sbi).dummy_enc_ctx);
445 "Can't change test_dummy_encryption on remount");
446 else if (err == -EINVAL)
447 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
450 f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
454 f2fs_warn(sbi, "Test dummy encryption mode enabled");
456 f2fs_warn(sbi, "Test dummy encryption mount option ignored");
461 static int parse_options(struct super_block *sb, char *options, bool is_remount)
463 struct f2fs_sb_info *sbi = F2FS_SB(sb);
464 substring_t args[MAX_OPT_ARGS];
465 unsigned char (*ext)[F2FS_EXTENSION_LEN];
467 int arg = 0, ext_cnt;
475 while ((p = strsep(&options, ",")) != NULL) {
480 * Initialize args struct so we know whether arg was
481 * found; some options take optional arguments.
483 args[0].to = args[0].from = NULL;
484 token = match_token(p, f2fs_tokens, args);
487 case Opt_gc_background:
488 name = match_strdup(&args[0]);
492 if (!strcmp(name, "on")) {
493 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
494 } else if (!strcmp(name, "off")) {
495 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
496 } else if (!strcmp(name, "sync")) {
497 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
504 case Opt_disable_roll_forward:
505 set_opt(sbi, DISABLE_ROLL_FORWARD);
508 /* this option mounts f2fs with ro */
509 set_opt(sbi, NORECOVERY);
510 if (!f2fs_readonly(sb))
514 set_opt(sbi, DISCARD);
517 if (f2fs_sb_has_blkzoned(sbi)) {
518 f2fs_warn(sbi, "discard is required for zoned block devices");
521 clear_opt(sbi, DISCARD);
524 set_opt(sbi, NOHEAP);
527 clear_opt(sbi, NOHEAP);
529 #ifdef CONFIG_F2FS_FS_XATTR
531 set_opt(sbi, XATTR_USER);
533 case Opt_nouser_xattr:
534 clear_opt(sbi, XATTR_USER);
536 case Opt_inline_xattr:
537 set_opt(sbi, INLINE_XATTR);
539 case Opt_noinline_xattr:
540 clear_opt(sbi, INLINE_XATTR);
542 case Opt_inline_xattr_size:
543 if (args->from && match_int(args, &arg))
545 set_opt(sbi, INLINE_XATTR_SIZE);
546 F2FS_OPTION(sbi).inline_xattr_size = arg;
550 f2fs_info(sbi, "user_xattr options not supported");
552 case Opt_nouser_xattr:
553 f2fs_info(sbi, "nouser_xattr options not supported");
555 case Opt_inline_xattr:
556 f2fs_info(sbi, "inline_xattr options not supported");
558 case Opt_noinline_xattr:
559 f2fs_info(sbi, "noinline_xattr options not supported");
562 #ifdef CONFIG_F2FS_FS_POSIX_ACL
564 set_opt(sbi, POSIX_ACL);
567 clear_opt(sbi, POSIX_ACL);
571 f2fs_info(sbi, "acl options not supported");
574 f2fs_info(sbi, "noacl options not supported");
577 case Opt_active_logs:
578 if (args->from && match_int(args, &arg))
580 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
582 F2FS_OPTION(sbi).active_logs = arg;
584 case Opt_disable_ext_identify:
585 set_opt(sbi, DISABLE_EXT_IDENTIFY);
587 case Opt_inline_data:
588 set_opt(sbi, INLINE_DATA);
590 case Opt_inline_dentry:
591 set_opt(sbi, INLINE_DENTRY);
593 case Opt_noinline_dentry:
594 clear_opt(sbi, INLINE_DENTRY);
596 case Opt_flush_merge:
597 set_opt(sbi, FLUSH_MERGE);
599 case Opt_noflush_merge:
600 clear_opt(sbi, FLUSH_MERGE);
603 set_opt(sbi, NOBARRIER);
606 set_opt(sbi, FASTBOOT);
608 case Opt_extent_cache:
609 set_opt(sbi, EXTENT_CACHE);
611 case Opt_noextent_cache:
612 clear_opt(sbi, EXTENT_CACHE);
614 case Opt_noinline_data:
615 clear_opt(sbi, INLINE_DATA);
618 set_opt(sbi, DATA_FLUSH);
620 case Opt_reserve_root:
621 if (args->from && match_int(args, &arg))
623 if (test_opt(sbi, RESERVE_ROOT)) {
624 f2fs_info(sbi, "Preserve previous reserve_root=%u",
625 F2FS_OPTION(sbi).root_reserved_blocks);
627 F2FS_OPTION(sbi).root_reserved_blocks = arg;
628 set_opt(sbi, RESERVE_ROOT);
632 if (args->from && match_int(args, &arg))
634 uid = make_kuid(current_user_ns(), arg);
635 if (!uid_valid(uid)) {
636 f2fs_err(sbi, "Invalid uid value %d", arg);
639 F2FS_OPTION(sbi).s_resuid = uid;
642 if (args->from && match_int(args, &arg))
644 gid = make_kgid(current_user_ns(), arg);
645 if (!gid_valid(gid)) {
646 f2fs_err(sbi, "Invalid gid value %d", arg);
649 F2FS_OPTION(sbi).s_resgid = gid;
652 name = match_strdup(&args[0]);
656 if (!strcmp(name, "adaptive")) {
657 if (f2fs_sb_has_blkzoned(sbi)) {
658 f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
662 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
663 } else if (!strcmp(name, "lfs")) {
664 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
671 case Opt_io_size_bits:
672 if (args->from && match_int(args, &arg))
674 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_PAGES)) {
675 f2fs_warn(sbi, "Not support %d, larger than %d",
676 1 << arg, BIO_MAX_PAGES);
679 F2FS_OPTION(sbi).write_io_size_bits = arg;
681 #ifdef CONFIG_F2FS_FAULT_INJECTION
682 case Opt_fault_injection:
683 if (args->from && match_int(args, &arg))
685 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
686 set_opt(sbi, FAULT_INJECTION);
690 if (args->from && match_int(args, &arg))
692 f2fs_build_fault_attr(sbi, 0, arg);
693 set_opt(sbi, FAULT_INJECTION);
696 case Opt_fault_injection:
697 f2fs_info(sbi, "fault_injection options not supported");
701 f2fs_info(sbi, "fault_type options not supported");
705 sb->s_flags |= SB_LAZYTIME;
708 sb->s_flags &= ~SB_LAZYTIME;
713 set_opt(sbi, USRQUOTA);
716 set_opt(sbi, GRPQUOTA);
719 set_opt(sbi, PRJQUOTA);
722 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
727 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
732 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
736 case Opt_offusrjquota:
737 ret = f2fs_clear_qf_name(sb, USRQUOTA);
741 case Opt_offgrpjquota:
742 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
746 case Opt_offprjjquota:
747 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
751 case Opt_jqfmt_vfsold:
752 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
754 case Opt_jqfmt_vfsv0:
755 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
757 case Opt_jqfmt_vfsv1:
758 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
761 clear_opt(sbi, QUOTA);
762 clear_opt(sbi, USRQUOTA);
763 clear_opt(sbi, GRPQUOTA);
764 clear_opt(sbi, PRJQUOTA);
774 case Opt_offusrjquota:
775 case Opt_offgrpjquota:
776 case Opt_offprjjquota:
777 case Opt_jqfmt_vfsold:
778 case Opt_jqfmt_vfsv0:
779 case Opt_jqfmt_vfsv1:
781 f2fs_info(sbi, "quota operations not supported");
785 name = match_strdup(&args[0]);
788 if (!strcmp(name, "user-based")) {
789 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
790 } else if (!strcmp(name, "off")) {
791 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
792 } else if (!strcmp(name, "fs-based")) {
793 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
801 name = match_strdup(&args[0]);
805 if (!strcmp(name, "default")) {
806 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
807 } else if (!strcmp(name, "reuse")) {
808 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
816 name = match_strdup(&args[0]);
819 if (!strcmp(name, "posix")) {
820 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
821 } else if (!strcmp(name, "strict")) {
822 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
823 } else if (!strcmp(name, "nobarrier")) {
824 F2FS_OPTION(sbi).fsync_mode =
825 FSYNC_MODE_NOBARRIER;
832 case Opt_test_dummy_encryption:
833 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
838 case Opt_inlinecrypt:
839 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
840 sb->s_flags |= SB_INLINECRYPT;
842 f2fs_info(sbi, "inline encryption not supported");
845 case Opt_checkpoint_disable_cap_perc:
846 if (args->from && match_int(args, &arg))
848 if (arg < 0 || arg > 100)
850 F2FS_OPTION(sbi).unusable_cap_perc = arg;
851 set_opt(sbi, DISABLE_CHECKPOINT);
853 case Opt_checkpoint_disable_cap:
854 if (args->from && match_int(args, &arg))
856 F2FS_OPTION(sbi).unusable_cap = arg;
857 set_opt(sbi, DISABLE_CHECKPOINT);
859 case Opt_checkpoint_disable:
860 set_opt(sbi, DISABLE_CHECKPOINT);
862 case Opt_checkpoint_enable:
863 clear_opt(sbi, DISABLE_CHECKPOINT);
865 case Opt_compress_algorithm:
866 if (!f2fs_sb_has_compression(sbi)) {
867 f2fs_err(sbi, "Compression feature if off");
870 name = match_strdup(&args[0]);
873 if (!strcmp(name, "lzo")) {
874 F2FS_OPTION(sbi).compress_algorithm =
876 } else if (!strcmp(name, "lz4")) {
877 F2FS_OPTION(sbi).compress_algorithm =
879 } else if (!strcmp(name, "zstd")) {
880 F2FS_OPTION(sbi).compress_algorithm =
882 } else if (!strcmp(name, "lzo-rle")) {
883 F2FS_OPTION(sbi).compress_algorithm =
891 case Opt_compress_log_size:
892 if (!f2fs_sb_has_compression(sbi)) {
893 f2fs_err(sbi, "Compression feature is off");
896 if (args->from && match_int(args, &arg))
898 if (arg < MIN_COMPRESS_LOG_SIZE ||
899 arg > MAX_COMPRESS_LOG_SIZE) {
901 "Compress cluster log size is out of range");
904 F2FS_OPTION(sbi).compress_log_size = arg;
906 case Opt_compress_extension:
907 if (!f2fs_sb_has_compression(sbi)) {
908 f2fs_err(sbi, "Compression feature is off");
911 name = match_strdup(&args[0]);
915 ext = F2FS_OPTION(sbi).extensions;
916 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
918 if (strlen(name) >= F2FS_EXTENSION_LEN ||
919 ext_cnt >= COMPRESS_EXT_NUM) {
921 "invalid extension length/number");
926 strcpy(ext[ext_cnt], name);
927 F2FS_OPTION(sbi).compress_ext_cnt++;
931 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
937 if (f2fs_check_quota_options(sbi))
940 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
941 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
944 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
945 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
949 #ifndef CONFIG_UNICODE
950 if (f2fs_sb_has_casefold(sbi)) {
952 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
957 if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) {
958 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
959 F2FS_IO_SIZE_KB(sbi));
963 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
964 int min_size, max_size;
966 if (!f2fs_sb_has_extra_attr(sbi) ||
967 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
968 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
971 if (!test_opt(sbi, INLINE_XATTR)) {
972 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
976 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
977 max_size = MAX_INLINE_XATTR_SIZE;
979 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
980 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
981 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
987 if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) {
988 f2fs_err(sbi, "LFS not compatible with checkpoint=disable\n");
992 /* Not pass down write hints if the number of active logs is lesser
993 * than NR_CURSEG_TYPE.
995 if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
996 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1000 static struct inode *f2fs_alloc_inode(struct super_block *sb)
1002 struct f2fs_inode_info *fi;
1004 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
1008 init_once((void *) fi);
1010 /* Initialize f2fs-specific inode info */
1011 atomic_set(&fi->dirty_pages, 0);
1012 init_rwsem(&fi->i_sem);
1013 spin_lock_init(&fi->i_size_lock);
1014 INIT_LIST_HEAD(&fi->dirty_list);
1015 INIT_LIST_HEAD(&fi->gdirty_list);
1016 INIT_LIST_HEAD(&fi->inmem_ilist);
1017 INIT_LIST_HEAD(&fi->inmem_pages);
1018 mutex_init(&fi->inmem_lock);
1019 init_rwsem(&fi->i_gc_rwsem[READ]);
1020 init_rwsem(&fi->i_gc_rwsem[WRITE]);
1021 init_rwsem(&fi->i_mmap_sem);
1022 init_rwsem(&fi->i_xattr_sem);
1024 /* Will be used by directory only */
1025 fi->i_dir_level = F2FS_SB(sb)->dir_level;
1027 return &fi->vfs_inode;
1030 static int f2fs_drop_inode(struct inode *inode)
1032 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1036 * during filesystem shutdown, if checkpoint is disabled,
1037 * drop useless meta/node dirty pages.
1039 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1040 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1041 inode->i_ino == F2FS_META_INO(sbi)) {
1042 trace_f2fs_drop_inode(inode, 1);
1048 * This is to avoid a deadlock condition like below.
1049 * writeback_single_inode(inode)
1050 * - f2fs_write_data_page
1051 * - f2fs_gc -> iput -> evict
1052 * - inode_wait_for_writeback(inode)
1054 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1055 if (!inode->i_nlink && !is_bad_inode(inode)) {
1056 /* to avoid evict_inode call simultaneously */
1057 atomic_inc(&inode->i_count);
1058 spin_unlock(&inode->i_lock);
1060 /* some remained atomic pages should discarded */
1061 if (f2fs_is_atomic_file(inode))
1062 f2fs_drop_inmem_pages(inode);
1064 /* should remain fi->extent_tree for writepage */
1065 f2fs_destroy_extent_node(inode);
1067 sb_start_intwrite(inode->i_sb);
1068 f2fs_i_size_write(inode, 0);
1070 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1071 inode, NULL, 0, DATA);
1072 truncate_inode_pages_final(inode->i_mapping);
1074 if (F2FS_HAS_BLOCKS(inode))
1075 f2fs_truncate(inode);
1077 sb_end_intwrite(inode->i_sb);
1079 spin_lock(&inode->i_lock);
1080 atomic_dec(&inode->i_count);
1082 trace_f2fs_drop_inode(inode, 0);
1085 ret = generic_drop_inode(inode);
1087 ret = fscrypt_drop_inode(inode);
1088 trace_f2fs_drop_inode(inode, ret);
1092 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1094 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1097 spin_lock(&sbi->inode_lock[DIRTY_META]);
1098 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1101 set_inode_flag(inode, FI_DIRTY_INODE);
1102 stat_inc_dirty_inode(sbi, DIRTY_META);
1104 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1105 list_add_tail(&F2FS_I(inode)->gdirty_list,
1106 &sbi->inode_list[DIRTY_META]);
1107 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1109 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1113 void f2fs_inode_synced(struct inode *inode)
1115 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1117 spin_lock(&sbi->inode_lock[DIRTY_META]);
1118 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1119 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1122 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1123 list_del_init(&F2FS_I(inode)->gdirty_list);
1124 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1126 clear_inode_flag(inode, FI_DIRTY_INODE);
1127 clear_inode_flag(inode, FI_AUTO_RECOVER);
1128 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1129 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1133 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1135 * We should call set_dirty_inode to write the dirty inode through write_inode.
1137 static void f2fs_dirty_inode(struct inode *inode, int flags)
1139 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1141 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1142 inode->i_ino == F2FS_META_INO(sbi))
1145 if (flags == I_DIRTY_TIME)
1148 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1149 clear_inode_flag(inode, FI_AUTO_RECOVER);
1151 f2fs_inode_dirtied(inode, false);
1154 static void f2fs_free_inode(struct inode *inode)
1156 fscrypt_free_inode(inode);
1157 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1160 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1162 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1163 percpu_counter_destroy(&sbi->total_valid_inode_count);
1166 static void destroy_device_list(struct f2fs_sb_info *sbi)
1170 for (i = 0; i < sbi->s_ndevs; i++) {
1171 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1172 #ifdef CONFIG_BLK_DEV_ZONED
1173 kvfree(FDEV(i).blkz_seq);
1179 static void f2fs_put_super(struct super_block *sb)
1181 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1185 f2fs_quota_off_umount(sb);
1187 /* prevent remaining shrinker jobs */
1188 mutex_lock(&sbi->umount_mutex);
1191 * We don't need to do checkpoint when superblock is clean.
1192 * But, the previous checkpoint was not done by umount, it needs to do
1193 * clean checkpoint again.
1195 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1196 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1197 struct cp_control cpc = {
1198 .reason = CP_UMOUNT,
1200 f2fs_write_checkpoint(sbi, &cpc);
1203 /* be sure to wait for any on-going discard commands */
1204 dropped = f2fs_issue_discard_timeout(sbi);
1206 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1207 !sbi->discard_blks && !dropped) {
1208 struct cp_control cpc = {
1209 .reason = CP_UMOUNT | CP_TRIMMED,
1211 f2fs_write_checkpoint(sbi, &cpc);
1215 * normally superblock is clean, so we need to release this.
1216 * In addition, EIO will skip do checkpoint, we need this as well.
1218 f2fs_release_ino_entry(sbi, true);
1220 f2fs_leave_shrinker(sbi);
1221 mutex_unlock(&sbi->umount_mutex);
1223 /* our cp_error case, we can wait for any writeback page */
1224 f2fs_flush_merged_writes(sbi);
1226 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1228 f2fs_bug_on(sbi, sbi->fsync_node_num);
1230 iput(sbi->node_inode);
1231 sbi->node_inode = NULL;
1233 iput(sbi->meta_inode);
1234 sbi->meta_inode = NULL;
1237 * iput() can update stat information, if f2fs_write_checkpoint()
1238 * above failed with error.
1240 f2fs_destroy_stats(sbi);
1242 /* destroy f2fs internal modules */
1243 f2fs_destroy_node_manager(sbi);
1244 f2fs_destroy_segment_manager(sbi);
1246 f2fs_destroy_post_read_wq(sbi);
1250 f2fs_unregister_sysfs(sbi);
1252 sb->s_fs_info = NULL;
1253 if (sbi->s_chksum_driver)
1254 crypto_free_shash(sbi->s_chksum_driver);
1255 kvfree(sbi->raw_super);
1257 destroy_device_list(sbi);
1258 f2fs_destroy_xattr_caches(sbi);
1259 mempool_destroy(sbi->write_io_dummy);
1261 for (i = 0; i < MAXQUOTAS; i++)
1262 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1264 fscrypt_free_dummy_context(&F2FS_OPTION(sbi).dummy_enc_ctx);
1265 destroy_percpu_info(sbi);
1266 for (i = 0; i < NR_PAGE_TYPE; i++)
1267 kvfree(sbi->write_io[i]);
1268 #ifdef CONFIG_UNICODE
1269 utf8_unload(sbi->s_encoding);
1274 int f2fs_sync_fs(struct super_block *sb, int sync)
1276 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1279 if (unlikely(f2fs_cp_error(sbi)))
1281 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1284 trace_f2fs_sync_fs(sb, sync);
1286 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1290 struct cp_control cpc;
1292 cpc.reason = __get_cp_reason(sbi);
1294 down_write(&sbi->gc_lock);
1295 err = f2fs_write_checkpoint(sbi, &cpc);
1296 up_write(&sbi->gc_lock);
1298 f2fs_trace_ios(NULL, 1);
1303 static int f2fs_freeze(struct super_block *sb)
1305 if (f2fs_readonly(sb))
1308 /* IO error happened before */
1309 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1312 /* must be clean, since sync_filesystem() was already called */
1313 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1318 static int f2fs_unfreeze(struct super_block *sb)
1324 static int f2fs_statfs_project(struct super_block *sb,
1325 kprojid_t projid, struct kstatfs *buf)
1328 struct dquot *dquot;
1332 qid = make_kqid_projid(projid);
1333 dquot = dqget(sb, qid);
1335 return PTR_ERR(dquot);
1336 spin_lock(&dquot->dq_dqb_lock);
1338 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1339 dquot->dq_dqb.dqb_bhardlimit);
1341 limit >>= sb->s_blocksize_bits;
1343 if (limit && buf->f_blocks > limit) {
1344 curblock = (dquot->dq_dqb.dqb_curspace +
1345 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1346 buf->f_blocks = limit;
1347 buf->f_bfree = buf->f_bavail =
1348 (buf->f_blocks > curblock) ?
1349 (buf->f_blocks - curblock) : 0;
1352 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1353 dquot->dq_dqb.dqb_ihardlimit);
1355 if (limit && buf->f_files > limit) {
1356 buf->f_files = limit;
1358 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1359 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1362 spin_unlock(&dquot->dq_dqb_lock);
1368 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1370 struct super_block *sb = dentry->d_sb;
1371 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1372 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1373 block_t total_count, user_block_count, start_count;
1374 u64 avail_node_count;
1376 total_count = le64_to_cpu(sbi->raw_super->block_count);
1377 user_block_count = sbi->user_block_count;
1378 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1379 buf->f_type = F2FS_SUPER_MAGIC;
1380 buf->f_bsize = sbi->blocksize;
1382 buf->f_blocks = total_count - start_count;
1383 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1384 sbi->current_reserved_blocks;
1386 spin_lock(&sbi->stat_lock);
1387 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1390 buf->f_bfree -= sbi->unusable_block_count;
1391 spin_unlock(&sbi->stat_lock);
1393 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1394 buf->f_bavail = buf->f_bfree -
1395 F2FS_OPTION(sbi).root_reserved_blocks;
1399 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1401 if (avail_node_count > user_block_count) {
1402 buf->f_files = user_block_count;
1403 buf->f_ffree = buf->f_bavail;
1405 buf->f_files = avail_node_count;
1406 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1410 buf->f_namelen = F2FS_NAME_LEN;
1411 buf->f_fsid.val[0] = (u32)id;
1412 buf->f_fsid.val[1] = (u32)(id >> 32);
1415 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1416 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1417 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1423 static inline void f2fs_show_quota_options(struct seq_file *seq,
1424 struct super_block *sb)
1427 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1429 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1432 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1443 seq_printf(seq, ",jqfmt=%s", fmtname);
1446 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1447 seq_show_option(seq, "usrjquota",
1448 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1450 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1451 seq_show_option(seq, "grpjquota",
1452 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1454 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1455 seq_show_option(seq, "prjjquota",
1456 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1460 static inline void f2fs_show_compress_options(struct seq_file *seq,
1461 struct super_block *sb)
1463 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1467 if (!f2fs_sb_has_compression(sbi))
1470 switch (F2FS_OPTION(sbi).compress_algorithm) {
1480 case COMPRESS_LZORLE:
1481 algtype = "lzo-rle";
1484 seq_printf(seq, ",compress_algorithm=%s", algtype);
1486 seq_printf(seq, ",compress_log_size=%u",
1487 F2FS_OPTION(sbi).compress_log_size);
1489 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1490 seq_printf(seq, ",compress_extension=%s",
1491 F2FS_OPTION(sbi).extensions[i]);
1495 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1497 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1499 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1500 seq_printf(seq, ",background_gc=%s", "sync");
1501 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1502 seq_printf(seq, ",background_gc=%s", "on");
1503 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1504 seq_printf(seq, ",background_gc=%s", "off");
1506 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1507 seq_puts(seq, ",disable_roll_forward");
1508 if (test_opt(sbi, NORECOVERY))
1509 seq_puts(seq, ",norecovery");
1510 if (test_opt(sbi, DISCARD))
1511 seq_puts(seq, ",discard");
1513 seq_puts(seq, ",nodiscard");
1514 if (test_opt(sbi, NOHEAP))
1515 seq_puts(seq, ",no_heap");
1517 seq_puts(seq, ",heap");
1518 #ifdef CONFIG_F2FS_FS_XATTR
1519 if (test_opt(sbi, XATTR_USER))
1520 seq_puts(seq, ",user_xattr");
1522 seq_puts(seq, ",nouser_xattr");
1523 if (test_opt(sbi, INLINE_XATTR))
1524 seq_puts(seq, ",inline_xattr");
1526 seq_puts(seq, ",noinline_xattr");
1527 if (test_opt(sbi, INLINE_XATTR_SIZE))
1528 seq_printf(seq, ",inline_xattr_size=%u",
1529 F2FS_OPTION(sbi).inline_xattr_size);
1531 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1532 if (test_opt(sbi, POSIX_ACL))
1533 seq_puts(seq, ",acl");
1535 seq_puts(seq, ",noacl");
1537 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1538 seq_puts(seq, ",disable_ext_identify");
1539 if (test_opt(sbi, INLINE_DATA))
1540 seq_puts(seq, ",inline_data");
1542 seq_puts(seq, ",noinline_data");
1543 if (test_opt(sbi, INLINE_DENTRY))
1544 seq_puts(seq, ",inline_dentry");
1546 seq_puts(seq, ",noinline_dentry");
1547 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1548 seq_puts(seq, ",flush_merge");
1549 if (test_opt(sbi, NOBARRIER))
1550 seq_puts(seq, ",nobarrier");
1551 if (test_opt(sbi, FASTBOOT))
1552 seq_puts(seq, ",fastboot");
1553 if (test_opt(sbi, EXTENT_CACHE))
1554 seq_puts(seq, ",extent_cache");
1556 seq_puts(seq, ",noextent_cache");
1557 if (test_opt(sbi, DATA_FLUSH))
1558 seq_puts(seq, ",data_flush");
1560 seq_puts(seq, ",mode=");
1561 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
1562 seq_puts(seq, "adaptive");
1563 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
1564 seq_puts(seq, "lfs");
1565 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1566 if (test_opt(sbi, RESERVE_ROOT))
1567 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1568 F2FS_OPTION(sbi).root_reserved_blocks,
1569 from_kuid_munged(&init_user_ns,
1570 F2FS_OPTION(sbi).s_resuid),
1571 from_kgid_munged(&init_user_ns,
1572 F2FS_OPTION(sbi).s_resgid));
1573 if (F2FS_IO_SIZE_BITS(sbi))
1574 seq_printf(seq, ",io_bits=%u",
1575 F2FS_OPTION(sbi).write_io_size_bits);
1576 #ifdef CONFIG_F2FS_FAULT_INJECTION
1577 if (test_opt(sbi, FAULT_INJECTION)) {
1578 seq_printf(seq, ",fault_injection=%u",
1579 F2FS_OPTION(sbi).fault_info.inject_rate);
1580 seq_printf(seq, ",fault_type=%u",
1581 F2FS_OPTION(sbi).fault_info.inject_type);
1585 if (test_opt(sbi, QUOTA))
1586 seq_puts(seq, ",quota");
1587 if (test_opt(sbi, USRQUOTA))
1588 seq_puts(seq, ",usrquota");
1589 if (test_opt(sbi, GRPQUOTA))
1590 seq_puts(seq, ",grpquota");
1591 if (test_opt(sbi, PRJQUOTA))
1592 seq_puts(seq, ",prjquota");
1594 f2fs_show_quota_options(seq, sbi->sb);
1595 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1596 seq_printf(seq, ",whint_mode=%s", "user-based");
1597 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1598 seq_printf(seq, ",whint_mode=%s", "fs-based");
1600 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
1602 if (sbi->sb->s_flags & SB_INLINECRYPT)
1603 seq_puts(seq, ",inlinecrypt");
1605 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1606 seq_printf(seq, ",alloc_mode=%s", "default");
1607 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1608 seq_printf(seq, ",alloc_mode=%s", "reuse");
1610 if (test_opt(sbi, DISABLE_CHECKPOINT))
1611 seq_printf(seq, ",checkpoint=disable:%u",
1612 F2FS_OPTION(sbi).unusable_cap);
1613 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1614 seq_printf(seq, ",fsync_mode=%s", "posix");
1615 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1616 seq_printf(seq, ",fsync_mode=%s", "strict");
1617 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1618 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1620 f2fs_show_compress_options(seq, sbi->sb);
1624 static void default_options(struct f2fs_sb_info *sbi)
1626 /* init some FS parameters */
1627 F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1628 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1629 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1630 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1631 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1632 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1633 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1634 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
1635 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
1636 F2FS_OPTION(sbi).compress_ext_cnt = 0;
1637 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
1639 sbi->sb->s_flags &= ~SB_INLINECRYPT;
1641 set_opt(sbi, INLINE_XATTR);
1642 set_opt(sbi, INLINE_DATA);
1643 set_opt(sbi, INLINE_DENTRY);
1644 set_opt(sbi, EXTENT_CACHE);
1645 set_opt(sbi, NOHEAP);
1646 clear_opt(sbi, DISABLE_CHECKPOINT);
1647 F2FS_OPTION(sbi).unusable_cap = 0;
1648 sbi->sb->s_flags |= SB_LAZYTIME;
1649 set_opt(sbi, FLUSH_MERGE);
1650 set_opt(sbi, DISCARD);
1651 if (f2fs_sb_has_blkzoned(sbi))
1652 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
1654 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
1656 #ifdef CONFIG_F2FS_FS_XATTR
1657 set_opt(sbi, XATTR_USER);
1659 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1660 set_opt(sbi, POSIX_ACL);
1663 f2fs_build_fault_attr(sbi, 0, 0);
1667 static int f2fs_enable_quotas(struct super_block *sb);
1670 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1672 unsigned int s_flags = sbi->sb->s_flags;
1673 struct cp_control cpc;
1678 if (s_flags & SB_RDONLY) {
1679 f2fs_err(sbi, "checkpoint=disable on readonly fs");
1682 sbi->sb->s_flags |= SB_ACTIVE;
1684 f2fs_update_time(sbi, DISABLE_TIME);
1686 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1687 down_write(&sbi->gc_lock);
1688 err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1689 if (err == -ENODATA) {
1693 if (err && err != -EAGAIN)
1697 ret = sync_filesystem(sbi->sb);
1699 err = ret ? ret: err;
1703 unusable = f2fs_get_unusable_blocks(sbi);
1704 if (f2fs_disable_cp_again(sbi, unusable)) {
1709 down_write(&sbi->gc_lock);
1710 cpc.reason = CP_PAUSE;
1711 set_sbi_flag(sbi, SBI_CP_DISABLED);
1712 err = f2fs_write_checkpoint(sbi, &cpc);
1716 spin_lock(&sbi->stat_lock);
1717 sbi->unusable_block_count = unusable;
1718 spin_unlock(&sbi->stat_lock);
1721 up_write(&sbi->gc_lock);
1723 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */
1727 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1729 down_write(&sbi->gc_lock);
1730 f2fs_dirty_to_prefree(sbi);
1732 clear_sbi_flag(sbi, SBI_CP_DISABLED);
1733 set_sbi_flag(sbi, SBI_IS_DIRTY);
1734 up_write(&sbi->gc_lock);
1736 f2fs_sync_fs(sbi->sb, 1);
1739 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1741 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1742 struct f2fs_mount_info org_mount_opt;
1743 unsigned long old_sb_flags;
1745 bool need_restart_gc = false;
1746 bool need_stop_gc = false;
1747 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1748 bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1749 bool no_io_align = !F2FS_IO_ALIGNED(sbi);
1750 bool checkpoint_changed;
1756 * Save the old mount options in case we
1757 * need to restore them.
1759 org_mount_opt = sbi->mount_opt;
1760 old_sb_flags = sb->s_flags;
1763 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1764 for (i = 0; i < MAXQUOTAS; i++) {
1765 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1766 org_mount_opt.s_qf_names[i] =
1767 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1769 if (!org_mount_opt.s_qf_names[i]) {
1770 for (j = 0; j < i; j++)
1771 kvfree(org_mount_opt.s_qf_names[j]);
1775 org_mount_opt.s_qf_names[i] = NULL;
1780 /* recover superblocks we couldn't write due to previous RO mount */
1781 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1782 err = f2fs_commit_super(sbi, false);
1783 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
1786 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1789 default_options(sbi);
1791 /* parse mount options */
1792 err = parse_options(sb, data, true);
1795 checkpoint_changed =
1796 disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1799 * Previous and new state of filesystem is RO,
1800 * so skip checking GC and FLUSH_MERGE conditions.
1802 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1806 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1807 err = dquot_suspend(sb, -1);
1810 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1811 /* dquot_resume needs RW */
1812 sb->s_flags &= ~SB_RDONLY;
1813 if (sb_any_quota_suspended(sb)) {
1814 dquot_resume(sb, -1);
1815 } else if (f2fs_sb_has_quota_ino(sbi)) {
1816 err = f2fs_enable_quotas(sb);
1822 /* disallow enable/disable extent_cache dynamically */
1823 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1825 f2fs_warn(sbi, "switch extent_cache option is not allowed");
1829 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
1831 f2fs_warn(sbi, "switch io_bits option is not allowed");
1835 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1837 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
1842 * We stop the GC thread if FS is mounted as RO
1843 * or if background_gc = off is passed in mount
1844 * option. Also sync the filesystem.
1846 if ((*flags & SB_RDONLY) ||
1847 F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) {
1848 if (sbi->gc_thread) {
1849 f2fs_stop_gc_thread(sbi);
1850 need_restart_gc = true;
1852 } else if (!sbi->gc_thread) {
1853 err = f2fs_start_gc_thread(sbi);
1856 need_stop_gc = true;
1859 if (*flags & SB_RDONLY ||
1860 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1861 writeback_inodes_sb(sb, WB_REASON_SYNC);
1864 set_sbi_flag(sbi, SBI_IS_DIRTY);
1865 set_sbi_flag(sbi, SBI_IS_CLOSE);
1866 f2fs_sync_fs(sb, 1);
1867 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1870 if (checkpoint_changed) {
1871 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1872 err = f2fs_disable_checkpoint(sbi);
1876 f2fs_enable_checkpoint(sbi);
1881 * We stop issue flush thread if FS is mounted as RO
1882 * or if flush_merge is not passed in mount option.
1884 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1885 clear_opt(sbi, FLUSH_MERGE);
1886 f2fs_destroy_flush_cmd_control(sbi, false);
1888 err = f2fs_create_flush_cmd_control(sbi);
1894 /* Release old quota file names */
1895 for (i = 0; i < MAXQUOTAS; i++)
1896 kvfree(org_mount_opt.s_qf_names[i]);
1898 /* Update the POSIXACL Flag */
1899 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1900 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1902 limit_reserve_root(sbi);
1903 adjust_unusable_cap_perc(sbi);
1904 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1907 if (need_restart_gc) {
1908 if (f2fs_start_gc_thread(sbi))
1909 f2fs_warn(sbi, "background gc thread has stopped");
1910 } else if (need_stop_gc) {
1911 f2fs_stop_gc_thread(sbi);
1915 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1916 for (i = 0; i < MAXQUOTAS; i++) {
1917 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1918 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1921 sbi->mount_opt = org_mount_opt;
1922 sb->s_flags = old_sb_flags;
1927 /* Read data from quotafile */
1928 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1929 size_t len, loff_t off)
1931 struct inode *inode = sb_dqopt(sb)->files[type];
1932 struct address_space *mapping = inode->i_mapping;
1933 block_t blkidx = F2FS_BYTES_TO_BLK(off);
1934 int offset = off & (sb->s_blocksize - 1);
1937 loff_t i_size = i_size_read(inode);
1944 if (off + len > i_size)
1947 while (toread > 0) {
1948 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1950 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1952 if (PTR_ERR(page) == -ENOMEM) {
1953 congestion_wait(BLK_RW_ASYNC,
1954 DEFAULT_IO_TIMEOUT);
1957 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1958 return PTR_ERR(page);
1963 if (unlikely(page->mapping != mapping)) {
1964 f2fs_put_page(page, 1);
1967 if (unlikely(!PageUptodate(page))) {
1968 f2fs_put_page(page, 1);
1969 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1973 kaddr = kmap_atomic(page);
1974 memcpy(data, kaddr + offset, tocopy);
1975 kunmap_atomic(kaddr);
1976 f2fs_put_page(page, 1);
1986 /* Write to quotafile */
1987 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1988 const char *data, size_t len, loff_t off)
1990 struct inode *inode = sb_dqopt(sb)->files[type];
1991 struct address_space *mapping = inode->i_mapping;
1992 const struct address_space_operations *a_ops = mapping->a_ops;
1993 int offset = off & (sb->s_blocksize - 1);
1994 size_t towrite = len;
1996 void *fsdata = NULL;
2001 while (towrite > 0) {
2002 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2005 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
2007 if (unlikely(err)) {
2008 if (err == -ENOMEM) {
2009 congestion_wait(BLK_RW_ASYNC,
2010 DEFAULT_IO_TIMEOUT);
2013 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2017 kaddr = kmap_atomic(page);
2018 memcpy(kaddr + offset, data, tocopy);
2019 kunmap_atomic(kaddr);
2020 flush_dcache_page(page);
2022 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2033 inode->i_mtime = inode->i_ctime = current_time(inode);
2034 f2fs_mark_inode_dirty_sync(inode, false);
2035 return len - towrite;
2038 static struct dquot **f2fs_get_dquots(struct inode *inode)
2040 return F2FS_I(inode)->i_dquot;
2043 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2045 return &F2FS_I(inode)->i_reserved_quota;
2048 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2050 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2051 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2055 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2056 F2FS_OPTION(sbi).s_jquota_fmt, type);
2059 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2064 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2065 err = f2fs_enable_quotas(sbi->sb);
2067 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2073 for (i = 0; i < MAXQUOTAS; i++) {
2074 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2075 err = f2fs_quota_on_mount(sbi, i);
2080 f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2087 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2090 struct inode *qf_inode;
2091 unsigned long qf_inum;
2094 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2096 qf_inum = f2fs_qf_ino(sb, type);
2100 qf_inode = f2fs_iget(sb, qf_inum);
2101 if (IS_ERR(qf_inode)) {
2102 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2103 return PTR_ERR(qf_inode);
2106 /* Don't account quota for quota files to avoid recursion */
2107 qf_inode->i_flags |= S_NOQUOTA;
2108 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2113 static int f2fs_enable_quotas(struct super_block *sb)
2115 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2117 unsigned long qf_inum;
2118 bool quota_mopt[MAXQUOTAS] = {
2119 test_opt(sbi, USRQUOTA),
2120 test_opt(sbi, GRPQUOTA),
2121 test_opt(sbi, PRJQUOTA),
2124 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2125 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2129 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2131 for (type = 0; type < MAXQUOTAS; type++) {
2132 qf_inum = f2fs_qf_ino(sb, type);
2134 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2135 DQUOT_USAGE_ENABLED |
2136 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2138 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2140 for (type--; type >= 0; type--)
2141 dquot_quota_off(sb, type);
2142 set_sbi_flag(F2FS_SB(sb),
2143 SBI_QUOTA_NEED_REPAIR);
2151 int f2fs_quota_sync(struct super_block *sb, int type)
2153 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2154 struct quota_info *dqopt = sb_dqopt(sb);
2161 * down_read(quota_sem)
2162 * dquot_writeback_dquots()
2165 * down_read(quota_sem)
2169 down_read(&sbi->quota_sem);
2170 ret = dquot_writeback_dquots(sb, type);
2175 * Now when everything is written we can discard the pagecache so
2176 * that userspace sees the changes.
2178 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2179 struct address_space *mapping;
2181 if (type != -1 && cnt != type)
2183 if (!sb_has_quota_active(sb, cnt))
2186 mapping = dqopt->files[cnt]->i_mapping;
2188 ret = filemap_fdatawrite(mapping);
2192 /* if we are using journalled quota */
2193 if (is_journalled_quota(sbi))
2196 ret = filemap_fdatawait(mapping);
2198 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2200 inode_lock(dqopt->files[cnt]);
2201 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
2202 inode_unlock(dqopt->files[cnt]);
2206 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2207 up_read(&sbi->quota_sem);
2208 f2fs_unlock_op(sbi);
2212 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2213 const struct path *path)
2215 struct inode *inode;
2218 /* if quota sysfile exists, deny enabling quota with specific file */
2219 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2220 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2224 err = f2fs_quota_sync(sb, type);
2228 err = dquot_quota_on(sb, type, format_id, path);
2232 inode = d_inode(path->dentry);
2235 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2236 f2fs_set_inode_flags(inode);
2237 inode_unlock(inode);
2238 f2fs_mark_inode_dirty_sync(inode, false);
2243 static int __f2fs_quota_off(struct super_block *sb, int type)
2245 struct inode *inode = sb_dqopt(sb)->files[type];
2248 if (!inode || !igrab(inode))
2249 return dquot_quota_off(sb, type);
2251 err = f2fs_quota_sync(sb, type);
2255 err = dquot_quota_off(sb, type);
2256 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2260 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2261 f2fs_set_inode_flags(inode);
2262 inode_unlock(inode);
2263 f2fs_mark_inode_dirty_sync(inode, false);
2269 static int f2fs_quota_off(struct super_block *sb, int type)
2271 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2274 err = __f2fs_quota_off(sb, type);
2277 * quotactl can shutdown journalled quota, result in inconsistence
2278 * between quota record and fs data by following updates, tag the
2279 * flag to let fsck be aware of it.
2281 if (is_journalled_quota(sbi))
2282 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2286 void f2fs_quota_off_umount(struct super_block *sb)
2291 for (type = 0; type < MAXQUOTAS; type++) {
2292 err = __f2fs_quota_off(sb, type);
2294 int ret = dquot_quota_off(sb, type);
2296 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2298 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2302 * In case of checkpoint=disable, we must flush quota blocks.
2303 * This can cause NULL exception for node_inode in end_io, since
2304 * put_super already dropped it.
2306 sync_filesystem(sb);
2309 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2311 struct quota_info *dqopt = sb_dqopt(sb);
2314 for (type = 0; type < MAXQUOTAS; type++) {
2315 if (!dqopt->files[type])
2317 f2fs_inode_synced(dqopt->files[type]);
2321 static int f2fs_dquot_commit(struct dquot *dquot)
2323 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2326 down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
2327 ret = dquot_commit(dquot);
2329 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2330 up_read(&sbi->quota_sem);
2334 static int f2fs_dquot_acquire(struct dquot *dquot)
2336 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2339 down_read(&sbi->quota_sem);
2340 ret = dquot_acquire(dquot);
2342 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2343 up_read(&sbi->quota_sem);
2347 static int f2fs_dquot_release(struct dquot *dquot)
2349 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2350 int ret = dquot_release(dquot);
2353 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2357 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2359 struct super_block *sb = dquot->dq_sb;
2360 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2361 int ret = dquot_mark_dquot_dirty(dquot);
2363 /* if we are using journalled quota */
2364 if (is_journalled_quota(sbi))
2365 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2370 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2372 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2373 int ret = dquot_commit_info(sb, type);
2376 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2380 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2382 *projid = F2FS_I(inode)->i_projid;
2386 static const struct dquot_operations f2fs_quota_operations = {
2387 .get_reserved_space = f2fs_get_reserved_space,
2388 .write_dquot = f2fs_dquot_commit,
2389 .acquire_dquot = f2fs_dquot_acquire,
2390 .release_dquot = f2fs_dquot_release,
2391 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
2392 .write_info = f2fs_dquot_commit_info,
2393 .alloc_dquot = dquot_alloc,
2394 .destroy_dquot = dquot_destroy,
2395 .get_projid = f2fs_get_projid,
2396 .get_next_id = dquot_get_next_id,
2399 static const struct quotactl_ops f2fs_quotactl_ops = {
2400 .quota_on = f2fs_quota_on,
2401 .quota_off = f2fs_quota_off,
2402 .quota_sync = f2fs_quota_sync,
2403 .get_state = dquot_get_state,
2404 .set_info = dquot_set_dqinfo,
2405 .get_dqblk = dquot_get_dqblk,
2406 .set_dqblk = dquot_set_dqblk,
2407 .get_nextdqblk = dquot_get_next_dqblk,
2410 int f2fs_quota_sync(struct super_block *sb, int type)
2415 void f2fs_quota_off_umount(struct super_block *sb)
2420 static const struct super_operations f2fs_sops = {
2421 .alloc_inode = f2fs_alloc_inode,
2422 .free_inode = f2fs_free_inode,
2423 .drop_inode = f2fs_drop_inode,
2424 .write_inode = f2fs_write_inode,
2425 .dirty_inode = f2fs_dirty_inode,
2426 .show_options = f2fs_show_options,
2428 .quota_read = f2fs_quota_read,
2429 .quota_write = f2fs_quota_write,
2430 .get_dquots = f2fs_get_dquots,
2432 .evict_inode = f2fs_evict_inode,
2433 .put_super = f2fs_put_super,
2434 .sync_fs = f2fs_sync_fs,
2435 .freeze_fs = f2fs_freeze,
2436 .unfreeze_fs = f2fs_unfreeze,
2437 .statfs = f2fs_statfs,
2438 .remount_fs = f2fs_remount,
2441 #ifdef CONFIG_FS_ENCRYPTION
2442 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2444 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2445 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2449 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2452 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2455 * Encrypting the root directory is not allowed because fsck
2456 * expects lost+found directory to exist and remain unencrypted
2457 * if LOST_FOUND feature is enabled.
2460 if (f2fs_sb_has_lost_found(sbi) &&
2461 inode->i_ino == F2FS_ROOT_INO(sbi))
2464 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2465 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2466 ctx, len, fs_data, XATTR_CREATE);
2469 static const union fscrypt_context *
2470 f2fs_get_dummy_context(struct super_block *sb)
2472 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_ctx.ctx;
2475 static bool f2fs_has_stable_inodes(struct super_block *sb)
2480 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
2481 int *ino_bits_ret, int *lblk_bits_ret)
2483 *ino_bits_ret = 8 * sizeof(nid_t);
2484 *lblk_bits_ret = 8 * sizeof(block_t);
2487 static int f2fs_get_num_devices(struct super_block *sb)
2489 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2491 if (f2fs_is_multi_device(sbi))
2492 return sbi->s_ndevs;
2496 static void f2fs_get_devices(struct super_block *sb,
2497 struct request_queue **devs)
2499 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2502 for (i = 0; i < sbi->s_ndevs; i++)
2503 devs[i] = bdev_get_queue(FDEV(i).bdev);
2506 static const struct fscrypt_operations f2fs_cryptops = {
2507 .key_prefix = "f2fs:",
2508 .get_context = f2fs_get_context,
2509 .set_context = f2fs_set_context,
2510 .get_dummy_context = f2fs_get_dummy_context,
2511 .empty_dir = f2fs_empty_dir,
2512 .max_namelen = F2FS_NAME_LEN,
2513 .has_stable_inodes = f2fs_has_stable_inodes,
2514 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits,
2515 .get_num_devices = f2fs_get_num_devices,
2516 .get_devices = f2fs_get_devices,
2520 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2521 u64 ino, u32 generation)
2523 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2524 struct inode *inode;
2526 if (f2fs_check_nid_range(sbi, ino))
2527 return ERR_PTR(-ESTALE);
2530 * f2fs_iget isn't quite right if the inode is currently unallocated!
2531 * However f2fs_iget currently does appropriate checks to handle stale
2532 * inodes so everything is OK.
2534 inode = f2fs_iget(sb, ino);
2536 return ERR_CAST(inode);
2537 if (unlikely(generation && inode->i_generation != generation)) {
2538 /* we didn't find the right inode.. */
2540 return ERR_PTR(-ESTALE);
2545 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2546 int fh_len, int fh_type)
2548 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2549 f2fs_nfs_get_inode);
2552 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2553 int fh_len, int fh_type)
2555 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2556 f2fs_nfs_get_inode);
2559 static const struct export_operations f2fs_export_ops = {
2560 .fh_to_dentry = f2fs_fh_to_dentry,
2561 .fh_to_parent = f2fs_fh_to_parent,
2562 .get_parent = f2fs_get_parent,
2565 static loff_t max_file_blocks(void)
2568 loff_t leaf_count = DEF_ADDRS_PER_BLOCK;
2571 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2572 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2573 * space in inode.i_addr, it will be more safe to reassign
2577 /* two direct node blocks */
2578 result += (leaf_count * 2);
2580 /* two indirect node blocks */
2581 leaf_count *= NIDS_PER_BLOCK;
2582 result += (leaf_count * 2);
2584 /* one double indirect node block */
2585 leaf_count *= NIDS_PER_BLOCK;
2586 result += leaf_count;
2591 static int __f2fs_commit_super(struct buffer_head *bh,
2592 struct f2fs_super_block *super)
2596 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2597 set_buffer_dirty(bh);
2600 /* it's rare case, we can do fua all the time */
2601 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2604 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2605 struct buffer_head *bh)
2607 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2608 (bh->b_data + F2FS_SUPER_OFFSET);
2609 struct super_block *sb = sbi->sb;
2610 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2611 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2612 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2613 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2614 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2615 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2616 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2617 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2618 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2619 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2620 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2621 u32 segment_count = le32_to_cpu(raw_super->segment_count);
2622 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2623 u64 main_end_blkaddr = main_blkaddr +
2624 (segment_count_main << log_blocks_per_seg);
2625 u64 seg_end_blkaddr = segment0_blkaddr +
2626 (segment_count << log_blocks_per_seg);
2628 if (segment0_blkaddr != cp_blkaddr) {
2629 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2630 segment0_blkaddr, cp_blkaddr);
2634 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2636 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2637 cp_blkaddr, sit_blkaddr,
2638 segment_count_ckpt << log_blocks_per_seg);
2642 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2644 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2645 sit_blkaddr, nat_blkaddr,
2646 segment_count_sit << log_blocks_per_seg);
2650 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2652 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2653 nat_blkaddr, ssa_blkaddr,
2654 segment_count_nat << log_blocks_per_seg);
2658 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2660 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2661 ssa_blkaddr, main_blkaddr,
2662 segment_count_ssa << log_blocks_per_seg);
2666 if (main_end_blkaddr > seg_end_blkaddr) {
2667 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2670 (segment_count << log_blocks_per_seg),
2671 segment_count_main << log_blocks_per_seg);
2673 } else if (main_end_blkaddr < seg_end_blkaddr) {
2677 /* fix in-memory information all the time */
2678 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2679 segment0_blkaddr) >> log_blocks_per_seg);
2681 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2682 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2685 err = __f2fs_commit_super(bh, NULL);
2686 res = err ? "failed" : "done";
2688 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%u) block(%u)",
2691 (segment_count << log_blocks_per_seg),
2692 segment_count_main << log_blocks_per_seg);
2699 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2700 struct buffer_head *bh)
2702 block_t segment_count, segs_per_sec, secs_per_zone;
2703 block_t total_sections, blocks_per_seg;
2704 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2705 (bh->b_data + F2FS_SUPER_OFFSET);
2706 unsigned int blocksize;
2707 size_t crc_offset = 0;
2710 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
2711 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
2712 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2716 /* Check checksum_offset and crc in superblock */
2717 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2718 crc_offset = le32_to_cpu(raw_super->checksum_offset);
2720 offsetof(struct f2fs_super_block, crc)) {
2721 f2fs_info(sbi, "Invalid SB checksum offset: %zu",
2723 return -EFSCORRUPTED;
2725 crc = le32_to_cpu(raw_super->crc);
2726 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2727 f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
2728 return -EFSCORRUPTED;
2732 /* Currently, support only 4KB page cache size */
2733 if (F2FS_BLKSIZE != PAGE_SIZE) {
2734 f2fs_info(sbi, "Invalid page_cache_size (%lu), supports only 4KB",
2736 return -EFSCORRUPTED;
2739 /* Currently, support only 4KB block size */
2740 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2741 if (blocksize != F2FS_BLKSIZE) {
2742 f2fs_info(sbi, "Invalid blocksize (%u), supports only 4KB",
2744 return -EFSCORRUPTED;
2747 /* check log blocks per segment */
2748 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2749 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
2750 le32_to_cpu(raw_super->log_blocks_per_seg));
2751 return -EFSCORRUPTED;
2754 /* Currently, support 512/1024/2048/4096 bytes sector size */
2755 if (le32_to_cpu(raw_super->log_sectorsize) >
2756 F2FS_MAX_LOG_SECTOR_SIZE ||
2757 le32_to_cpu(raw_super->log_sectorsize) <
2758 F2FS_MIN_LOG_SECTOR_SIZE) {
2759 f2fs_info(sbi, "Invalid log sectorsize (%u)",
2760 le32_to_cpu(raw_super->log_sectorsize));
2761 return -EFSCORRUPTED;
2763 if (le32_to_cpu(raw_super->log_sectors_per_block) +
2764 le32_to_cpu(raw_super->log_sectorsize) !=
2765 F2FS_MAX_LOG_SECTOR_SIZE) {
2766 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
2767 le32_to_cpu(raw_super->log_sectors_per_block),
2768 le32_to_cpu(raw_super->log_sectorsize));
2769 return -EFSCORRUPTED;
2772 segment_count = le32_to_cpu(raw_super->segment_count);
2773 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2774 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2775 total_sections = le32_to_cpu(raw_super->section_count);
2777 /* blocks_per_seg should be 512, given the above check */
2778 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2780 if (segment_count > F2FS_MAX_SEGMENT ||
2781 segment_count < F2FS_MIN_SEGMENTS) {
2782 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
2783 return -EFSCORRUPTED;
2786 if (total_sections > segment_count ||
2787 total_sections < F2FS_MIN_SEGMENTS ||
2788 segs_per_sec > segment_count || !segs_per_sec) {
2789 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
2790 segment_count, total_sections, segs_per_sec);
2791 return -EFSCORRUPTED;
2794 if ((segment_count / segs_per_sec) < total_sections) {
2795 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
2796 segment_count, segs_per_sec, total_sections);
2797 return -EFSCORRUPTED;
2800 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2801 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
2802 segment_count, le64_to_cpu(raw_super->block_count));
2803 return -EFSCORRUPTED;
2806 if (RDEV(0).path[0]) {
2807 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
2810 while (i < MAX_DEVICES && RDEV(i).path[0]) {
2811 dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
2814 if (segment_count != dev_seg_count) {
2815 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
2816 segment_count, dev_seg_count);
2817 return -EFSCORRUPTED;
2821 if (secs_per_zone > total_sections || !secs_per_zone) {
2822 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
2823 secs_per_zone, total_sections);
2824 return -EFSCORRUPTED;
2826 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2827 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2828 (le32_to_cpu(raw_super->extension_count) +
2829 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2830 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
2831 le32_to_cpu(raw_super->extension_count),
2832 raw_super->hot_ext_count,
2833 F2FS_MAX_EXTENSION);
2834 return -EFSCORRUPTED;
2837 if (le32_to_cpu(raw_super->cp_payload) >
2838 (blocks_per_seg - F2FS_CP_PACKS)) {
2839 f2fs_info(sbi, "Insane cp_payload (%u > %u)",
2840 le32_to_cpu(raw_super->cp_payload),
2841 blocks_per_seg - F2FS_CP_PACKS);
2842 return -EFSCORRUPTED;
2845 /* check reserved ino info */
2846 if (le32_to_cpu(raw_super->node_ino) != 1 ||
2847 le32_to_cpu(raw_super->meta_ino) != 2 ||
2848 le32_to_cpu(raw_super->root_ino) != 3) {
2849 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2850 le32_to_cpu(raw_super->node_ino),
2851 le32_to_cpu(raw_super->meta_ino),
2852 le32_to_cpu(raw_super->root_ino));
2853 return -EFSCORRUPTED;
2856 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2857 if (sanity_check_area_boundary(sbi, bh))
2858 return -EFSCORRUPTED;
2863 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2865 unsigned int total, fsmeta;
2866 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2867 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2868 unsigned int ovp_segments, reserved_segments;
2869 unsigned int main_segs, blocks_per_seg;
2870 unsigned int sit_segs, nat_segs;
2871 unsigned int sit_bitmap_size, nat_bitmap_size;
2872 unsigned int log_blocks_per_seg;
2873 unsigned int segment_count_main;
2874 unsigned int cp_pack_start_sum, cp_payload;
2875 block_t user_block_count, valid_user_blocks;
2876 block_t avail_node_count, valid_node_count;
2879 total = le32_to_cpu(raw_super->segment_count);
2880 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2881 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2883 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2885 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2886 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2888 if (unlikely(fsmeta >= total))
2891 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2892 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2894 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2895 ovp_segments == 0 || reserved_segments == 0)) {
2896 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
2900 user_block_count = le64_to_cpu(ckpt->user_block_count);
2901 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2902 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2903 if (!user_block_count || user_block_count >=
2904 segment_count_main << log_blocks_per_seg) {
2905 f2fs_err(sbi, "Wrong user_block_count: %u",
2910 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
2911 if (valid_user_blocks > user_block_count) {
2912 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
2913 valid_user_blocks, user_block_count);
2917 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
2918 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
2919 if (valid_node_count > avail_node_count) {
2920 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
2921 valid_node_count, avail_node_count);
2925 main_segs = le32_to_cpu(raw_super->segment_count_main);
2926 blocks_per_seg = sbi->blocks_per_seg;
2928 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2929 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2930 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2932 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2933 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2934 le32_to_cpu(ckpt->cur_node_segno[j])) {
2935 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
2937 le32_to_cpu(ckpt->cur_node_segno[i]));
2942 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2943 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2944 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2946 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2947 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2948 le32_to_cpu(ckpt->cur_data_segno[j])) {
2949 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
2951 le32_to_cpu(ckpt->cur_data_segno[i]));
2956 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2957 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
2958 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2959 le32_to_cpu(ckpt->cur_data_segno[j])) {
2960 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
2962 le32_to_cpu(ckpt->cur_node_segno[i]));
2968 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2969 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2971 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2972 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2973 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
2974 sit_bitmap_size, nat_bitmap_size);
2978 cp_pack_start_sum = __start_sum_addr(sbi);
2979 cp_payload = __cp_payload(sbi);
2980 if (cp_pack_start_sum < cp_payload + 1 ||
2981 cp_pack_start_sum > blocks_per_seg - 1 -
2983 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
2988 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
2989 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
2990 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
2991 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
2992 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
2993 le32_to_cpu(ckpt->checksum_offset));
2997 if (unlikely(f2fs_cp_error(sbi))) {
2998 f2fs_err(sbi, "A bug case: need to run fsck");
3004 static void init_sb_info(struct f2fs_sb_info *sbi)
3006 struct f2fs_super_block *raw_super = sbi->raw_super;
3009 sbi->log_sectors_per_block =
3010 le32_to_cpu(raw_super->log_sectors_per_block);
3011 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3012 sbi->blocksize = 1 << sbi->log_blocksize;
3013 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3014 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
3015 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3016 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3017 sbi->total_sections = le32_to_cpu(raw_super->section_count);
3018 sbi->total_node_count =
3019 (le32_to_cpu(raw_super->segment_count_nat) / 2)
3020 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
3021 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
3022 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
3023 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
3024 sbi->cur_victim_sec = NULL_SECNO;
3025 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3026 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3027 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3028 sbi->migration_granularity = sbi->segs_per_sec;
3030 sbi->dir_level = DEF_DIR_LEVEL;
3031 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3032 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3033 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3034 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3035 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3036 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3037 DEF_UMOUNT_DISCARD_TIMEOUT;
3038 clear_sbi_flag(sbi, SBI_NEED_FSCK);
3040 for (i = 0; i < NR_COUNT_TYPE; i++)
3041 atomic_set(&sbi->nr_pages[i], 0);
3043 for (i = 0; i < META; i++)
3044 atomic_set(&sbi->wb_sync_req[i], 0);
3046 INIT_LIST_HEAD(&sbi->s_list);
3047 mutex_init(&sbi->umount_mutex);
3048 init_rwsem(&sbi->io_order_lock);
3049 spin_lock_init(&sbi->cp_lock);
3051 sbi->dirty_device = 0;
3052 spin_lock_init(&sbi->dev_lock);
3054 init_rwsem(&sbi->sb_lock);
3055 init_rwsem(&sbi->pin_sem);
3058 static int init_percpu_info(struct f2fs_sb_info *sbi)
3062 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3066 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3069 percpu_counter_destroy(&sbi->alloc_valid_block_count);
3074 #ifdef CONFIG_BLK_DEV_ZONED
3075 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3078 struct f2fs_dev_info *dev = data;
3080 if (zone->type != BLK_ZONE_TYPE_CONVENTIONAL)
3081 set_bit(idx, dev->blkz_seq);
3085 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3087 struct block_device *bdev = FDEV(devi).bdev;
3088 sector_t nr_sectors = bdev->bd_part->nr_sects;
3091 if (!f2fs_sb_has_blkzoned(sbi))
3094 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3095 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
3097 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
3098 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
3099 __ilog2_u32(sbi->blocks_per_blkz))
3101 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
3102 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
3103 sbi->log_blocks_per_blkz;
3104 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
3105 FDEV(devi).nr_blkz++;
3107 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3108 BITS_TO_LONGS(FDEV(devi).nr_blkz)
3109 * sizeof(unsigned long),
3111 if (!FDEV(devi).blkz_seq)
3114 /* Get block zones type */
3115 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3125 * Read f2fs raw super block.
3126 * Because we have two copies of super block, so read both of them
3127 * to get the first valid one. If any one of them is broken, we pass
3128 * them recovery flag back to the caller.
3130 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3131 struct f2fs_super_block **raw_super,
3132 int *valid_super_block, int *recovery)
3134 struct super_block *sb = sbi->sb;
3136 struct buffer_head *bh;
3137 struct f2fs_super_block *super;
3140 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3144 for (block = 0; block < 2; block++) {
3145 bh = sb_bread(sb, block);
3147 f2fs_err(sbi, "Unable to read %dth superblock",
3154 /* sanity checking of raw super */
3155 err = sanity_check_raw_super(sbi, bh);
3157 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3165 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3167 *valid_super_block = block;
3173 /* No valid superblock */
3182 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3184 struct buffer_head *bh;
3188 if ((recover && f2fs_readonly(sbi->sb)) ||
3189 bdev_read_only(sbi->sb->s_bdev)) {
3190 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3194 /* we should update superblock crc here */
3195 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3196 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3197 offsetof(struct f2fs_super_block, crc));
3198 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3201 /* write back-up superblock first */
3202 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3205 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3208 /* if we are in recovery path, skip writing valid superblock */
3212 /* write current valid superblock */
3213 bh = sb_bread(sbi->sb, sbi->valid_super_block);
3216 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3221 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3223 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3224 unsigned int max_devices = MAX_DEVICES;
3227 /* Initialize single device information */
3228 if (!RDEV(0).path[0]) {
3229 if (!bdev_is_zoned(sbi->sb->s_bdev))
3235 * Initialize multiple devices information, or single
3236 * zoned block device information.
3238 sbi->devs = f2fs_kzalloc(sbi,
3239 array_size(max_devices,
3240 sizeof(struct f2fs_dev_info)),
3245 for (i = 0; i < max_devices; i++) {
3247 if (i > 0 && !RDEV(i).path[0])
3250 if (max_devices == 1) {
3251 /* Single zoned block device mount */
3253 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3254 sbi->sb->s_mode, sbi->sb->s_type);
3256 /* Multi-device mount */
3257 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3258 FDEV(i).total_segments =
3259 le32_to_cpu(RDEV(i).total_segments);
3261 FDEV(i).start_blk = 0;
3262 FDEV(i).end_blk = FDEV(i).start_blk +
3263 (FDEV(i).total_segments <<
3264 sbi->log_blocks_per_seg) - 1 +
3265 le32_to_cpu(raw_super->segment0_blkaddr);
3267 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3268 FDEV(i).end_blk = FDEV(i).start_blk +
3269 (FDEV(i).total_segments <<
3270 sbi->log_blocks_per_seg) - 1;
3272 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3273 sbi->sb->s_mode, sbi->sb->s_type);
3275 if (IS_ERR(FDEV(i).bdev))
3276 return PTR_ERR(FDEV(i).bdev);
3278 /* to release errored devices */
3279 sbi->s_ndevs = i + 1;
3281 #ifdef CONFIG_BLK_DEV_ZONED
3282 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3283 !f2fs_sb_has_blkzoned(sbi)) {
3284 f2fs_err(sbi, "Zoned block device feature not enabled\n");
3287 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3288 if (init_blkz_info(sbi, i)) {
3289 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3292 if (max_devices == 1)
3294 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3296 FDEV(i).total_segments,
3297 FDEV(i).start_blk, FDEV(i).end_blk,
3298 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3299 "Host-aware" : "Host-managed");
3303 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3305 FDEV(i).total_segments,
3306 FDEV(i).start_blk, FDEV(i).end_blk);
3309 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3313 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
3315 #ifdef CONFIG_UNICODE
3316 if (f2fs_sb_has_casefold(sbi) && !sbi->s_encoding) {
3317 const struct f2fs_sb_encodings *encoding_info;
3318 struct unicode_map *encoding;
3319 __u16 encoding_flags;
3321 if (f2fs_sb_has_encrypt(sbi)) {
3323 "Can't mount with encoding and encryption");
3327 if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info,
3330 "Encoding requested by superblock is unknown");
3334 encoding = utf8_load(encoding_info->version);
3335 if (IS_ERR(encoding)) {
3337 "can't mount with superblock charset: %s-%s "
3338 "not supported by the kernel. flags: 0x%x.",
3339 encoding_info->name, encoding_info->version,
3341 return PTR_ERR(encoding);
3343 f2fs_info(sbi, "Using encoding defined by superblock: "
3344 "%s-%s with flags 0x%hx", encoding_info->name,
3345 encoding_info->version?:"\b", encoding_flags);
3347 sbi->s_encoding = encoding;
3348 sbi->s_encoding_flags = encoding_flags;
3349 sbi->sb->s_d_op = &f2fs_dentry_ops;
3352 if (f2fs_sb_has_casefold(sbi)) {
3353 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
3360 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3362 struct f2fs_sm_info *sm_i = SM_I(sbi);
3364 /* adjust parameters according to the volume size */
3365 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3366 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3367 sm_i->dcc_info->discard_granularity = 1;
3368 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3371 sbi->readdir_ra = 1;
3374 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3376 struct f2fs_sb_info *sbi;
3377 struct f2fs_super_block *raw_super;
3380 bool skip_recovery = false, need_fsck = false;
3381 char *options = NULL;
3382 int recovery, i, valid_super_block;
3383 struct curseg_info *seg_i;
3389 valid_super_block = -1;
3392 /* allocate memory for f2fs-specific super block info */
3393 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3399 /* Load the checksum driver */
3400 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3401 if (IS_ERR(sbi->s_chksum_driver)) {
3402 f2fs_err(sbi, "Cannot load crc32 driver.");
3403 err = PTR_ERR(sbi->s_chksum_driver);
3404 sbi->s_chksum_driver = NULL;
3408 /* set a block size */
3409 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3410 f2fs_err(sbi, "unable to set blocksize");
3414 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3419 sb->s_fs_info = sbi;
3420 sbi->raw_super = raw_super;
3422 /* precompute checksum seed for metadata */
3423 if (f2fs_sb_has_inode_chksum(sbi))
3424 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3425 sizeof(raw_super->uuid));
3428 * The BLKZONED feature indicates that the drive was formatted with
3429 * zone alignment optimization. This is optional for host-aware
3430 * devices, but mandatory for host-managed zoned block devices.
3432 #ifndef CONFIG_BLK_DEV_ZONED
3433 if (f2fs_sb_has_blkzoned(sbi)) {
3434 f2fs_err(sbi, "Zoned block device support is not enabled");
3439 default_options(sbi);
3440 /* parse mount options */
3441 options = kstrdup((const char *)data, GFP_KERNEL);
3442 if (data && !options) {
3447 err = parse_options(sb, options, false);
3451 sbi->max_file_blocks = max_file_blocks();
3452 sb->s_maxbytes = sbi->max_file_blocks <<
3453 le32_to_cpu(raw_super->log_blocksize);
3454 sb->s_max_links = F2FS_LINK_MAX;
3456 err = f2fs_setup_casefold(sbi);
3461 sb->dq_op = &f2fs_quota_operations;
3462 sb->s_qcop = &f2fs_quotactl_ops;
3463 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3465 if (f2fs_sb_has_quota_ino(sbi)) {
3466 for (i = 0; i < MAXQUOTAS; i++) {
3467 if (f2fs_qf_ino(sbi->sb, i))
3468 sbi->nquota_files++;
3473 sb->s_op = &f2fs_sops;
3474 #ifdef CONFIG_FS_ENCRYPTION
3475 sb->s_cop = &f2fs_cryptops;
3477 #ifdef CONFIG_FS_VERITY
3478 sb->s_vop = &f2fs_verityops;
3480 sb->s_xattr = f2fs_xattr_handlers;
3481 sb->s_export_op = &f2fs_export_ops;
3482 sb->s_magic = F2FS_SUPER_MAGIC;
3483 sb->s_time_gran = 1;
3484 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3485 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3486 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3487 sb->s_iflags |= SB_I_CGROUPWB;
3489 /* init f2fs-specific super block info */
3490 sbi->valid_super_block = valid_super_block;
3491 init_rwsem(&sbi->gc_lock);
3492 mutex_init(&sbi->writepages);
3493 mutex_init(&sbi->cp_mutex);
3494 init_rwsem(&sbi->node_write);
3495 init_rwsem(&sbi->node_change);
3497 /* disallow all the data/node/meta page writes */
3498 set_sbi_flag(sbi, SBI_POR_DOING);
3499 spin_lock_init(&sbi->stat_lock);
3501 /* init iostat info */
3502 spin_lock_init(&sbi->iostat_lock);
3503 sbi->iostat_enable = false;
3504 sbi->iostat_period_ms = DEFAULT_IOSTAT_PERIOD_MS;
3506 for (i = 0; i < NR_PAGE_TYPE; i++) {
3507 int n = (i == META) ? 1: NR_TEMP_TYPE;
3513 sizeof(struct f2fs_bio_info)),
3515 if (!sbi->write_io[i]) {
3520 for (j = HOT; j < n; j++) {
3521 init_rwsem(&sbi->write_io[i][j].io_rwsem);
3522 sbi->write_io[i][j].sbi = sbi;
3523 sbi->write_io[i][j].bio = NULL;
3524 spin_lock_init(&sbi->write_io[i][j].io_lock);
3525 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3526 INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list);
3527 init_rwsem(&sbi->write_io[i][j].bio_list_lock);
3531 init_rwsem(&sbi->cp_rwsem);
3532 init_rwsem(&sbi->quota_sem);
3533 init_waitqueue_head(&sbi->cp_wait);
3536 err = init_percpu_info(sbi);
3540 if (F2FS_IO_ALIGNED(sbi)) {
3541 sbi->write_io_dummy =
3542 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3543 if (!sbi->write_io_dummy) {
3549 /* init per sbi slab cache */
3550 err = f2fs_init_xattr_caches(sbi);
3554 /* get an inode for meta space */
3555 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3556 if (IS_ERR(sbi->meta_inode)) {
3557 f2fs_err(sbi, "Failed to read F2FS meta data inode");
3558 err = PTR_ERR(sbi->meta_inode);
3559 goto free_xattr_cache;
3562 err = f2fs_get_valid_checkpoint(sbi);
3564 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
3565 goto free_meta_inode;
3568 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3569 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3570 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3571 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3572 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3575 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
3576 set_sbi_flag(sbi, SBI_NEED_FSCK);
3578 /* Initialize device list */
3579 err = f2fs_scan_devices(sbi);
3581 f2fs_err(sbi, "Failed to find devices");
3585 err = f2fs_init_post_read_wq(sbi);
3587 f2fs_err(sbi, "Failed to initialize post read workqueue");
3591 sbi->total_valid_node_count =
3592 le32_to_cpu(sbi->ckpt->valid_node_count);
3593 percpu_counter_set(&sbi->total_valid_inode_count,
3594 le32_to_cpu(sbi->ckpt->valid_inode_count));
3595 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3596 sbi->total_valid_block_count =
3597 le64_to_cpu(sbi->ckpt->valid_block_count);
3598 sbi->last_valid_block_count = sbi->total_valid_block_count;
3599 sbi->reserved_blocks = 0;
3600 sbi->current_reserved_blocks = 0;
3601 limit_reserve_root(sbi);
3602 adjust_unusable_cap_perc(sbi);
3604 for (i = 0; i < NR_INODE_TYPE; i++) {
3605 INIT_LIST_HEAD(&sbi->inode_list[i]);
3606 spin_lock_init(&sbi->inode_lock[i]);
3608 mutex_init(&sbi->flush_lock);
3610 f2fs_init_extent_cache_info(sbi);
3612 f2fs_init_ino_entry_info(sbi);
3614 f2fs_init_fsync_node_info(sbi);
3616 /* setup f2fs internal modules */
3617 err = f2fs_build_segment_manager(sbi);
3619 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
3623 err = f2fs_build_node_manager(sbi);
3625 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
3630 /* For write statistics */
3631 if (sb->s_bdev->bd_part)
3632 sbi->sectors_written_start =
3633 (u64)part_stat_read(sb->s_bdev->bd_part,
3634 sectors[STAT_WRITE]);
3636 /* Read accumulated write IO statistics if exists */
3637 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3638 if (__exist_node_summaries(sbi))
3639 sbi->kbytes_written =
3640 le64_to_cpu(seg_i->journal->info.kbytes_written);
3642 f2fs_build_gc_manager(sbi);
3644 err = f2fs_build_stats(sbi);
3648 /* get an inode for node space */
3649 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3650 if (IS_ERR(sbi->node_inode)) {
3651 f2fs_err(sbi, "Failed to read node inode");
3652 err = PTR_ERR(sbi->node_inode);
3656 /* read root inode and dentry */
3657 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3659 f2fs_err(sbi, "Failed to read root inode");
3660 err = PTR_ERR(root);
3661 goto free_node_inode;
3663 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3664 !root->i_size || !root->i_nlink) {
3667 goto free_node_inode;
3670 sb->s_root = d_make_root(root); /* allocate root dentry */
3673 goto free_node_inode;
3676 err = f2fs_register_sysfs(sbi);
3678 goto free_root_inode;
3681 /* Enable quota usage during mount */
3682 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3683 err = f2fs_enable_quotas(sb);
3685 f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
3688 /* if there are any orphan inodes, free them */
3689 err = f2fs_recover_orphan_inodes(sbi);
3693 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3694 goto reset_checkpoint;
3696 /* recover fsynced data */
3697 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
3698 !test_opt(sbi, NORECOVERY)) {
3700 * mount should be failed, when device has readonly mode, and
3701 * previous checkpoint was not done by clean system shutdown.
3703 if (f2fs_hw_is_readonly(sbi)) {
3704 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3706 f2fs_err(sbi, "Need to recover fsync data, but write access unavailable");
3709 f2fs_info(sbi, "write access unavailable, skipping recovery");
3710 goto reset_checkpoint;
3714 set_sbi_flag(sbi, SBI_NEED_FSCK);
3717 goto reset_checkpoint;
3719 err = f2fs_recover_fsync_data(sbi, false);
3722 skip_recovery = true;
3724 f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
3729 err = f2fs_recover_fsync_data(sbi, true);
3731 if (!f2fs_readonly(sb) && err > 0) {
3733 f2fs_err(sbi, "Need to recover fsync data");
3739 * If the f2fs is not readonly and fsync data recovery succeeds,
3740 * check zoned block devices' write pointer consistency.
3742 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
3743 err = f2fs_check_write_pointer(sbi);
3749 /* f2fs_recover_fsync_data() cleared this already */
3750 clear_sbi_flag(sbi, SBI_POR_DOING);
3752 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3753 err = f2fs_disable_checkpoint(sbi);
3755 goto sync_free_meta;
3756 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3757 f2fs_enable_checkpoint(sbi);
3761 * If filesystem is not mounted as read-only then
3762 * do start the gc_thread.
3764 if (F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF && !f2fs_readonly(sb)) {
3765 /* After POR, we can run background GC thread.*/
3766 err = f2fs_start_gc_thread(sbi);
3768 goto sync_free_meta;
3772 /* recover broken superblock */
3774 err = f2fs_commit_super(sbi, true);
3775 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
3776 sbi->valid_super_block ? 1 : 2, err);
3779 f2fs_join_shrinker(sbi);
3781 f2fs_tuning_parameters(sbi);
3783 f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
3784 cur_cp_version(F2FS_CKPT(sbi)));
3785 f2fs_update_time(sbi, CP_TIME);
3786 f2fs_update_time(sbi, REQ_TIME);
3787 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3791 /* safe to flush all the data */
3792 sync_filesystem(sbi->sb);
3797 f2fs_truncate_quota_inode_pages(sb);
3798 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3799 f2fs_quota_off_umount(sbi->sb);
3802 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3803 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3804 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3805 * falls into an infinite loop in f2fs_sync_meta_pages().
3807 truncate_inode_pages_final(META_MAPPING(sbi));
3808 /* evict some inodes being cached by GC */
3810 f2fs_unregister_sysfs(sbi);
3815 f2fs_release_ino_entry(sbi, true);
3816 truncate_inode_pages_final(NODE_MAPPING(sbi));
3817 iput(sbi->node_inode);
3818 sbi->node_inode = NULL;
3820 f2fs_destroy_stats(sbi);
3822 f2fs_destroy_node_manager(sbi);
3824 f2fs_destroy_segment_manager(sbi);
3825 f2fs_destroy_post_read_wq(sbi);
3827 destroy_device_list(sbi);
3830 make_bad_inode(sbi->meta_inode);
3831 iput(sbi->meta_inode);
3832 sbi->meta_inode = NULL;
3834 f2fs_destroy_xattr_caches(sbi);
3836 mempool_destroy(sbi->write_io_dummy);
3838 destroy_percpu_info(sbi);
3840 for (i = 0; i < NR_PAGE_TYPE; i++)
3841 kvfree(sbi->write_io[i]);
3843 #ifdef CONFIG_UNICODE
3844 utf8_unload(sbi->s_encoding);
3848 for (i = 0; i < MAXQUOTAS; i++)
3849 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
3851 fscrypt_free_dummy_context(&F2FS_OPTION(sbi).dummy_enc_ctx);
3856 if (sbi->s_chksum_driver)
3857 crypto_free_shash(sbi->s_chksum_driver);
3860 /* give only one another chance */
3861 if (retry_cnt > 0 && skip_recovery) {
3863 shrink_dcache_sb(sb);
3869 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3870 const char *dev_name, void *data)
3872 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3875 static void kill_f2fs_super(struct super_block *sb)
3878 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3880 set_sbi_flag(sbi, SBI_IS_CLOSE);
3881 f2fs_stop_gc_thread(sbi);
3882 f2fs_stop_discard_thread(sbi);
3884 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3885 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3886 struct cp_control cpc = {
3887 .reason = CP_UMOUNT,
3889 f2fs_write_checkpoint(sbi, &cpc);
3892 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3893 sb->s_flags &= ~SB_RDONLY;
3895 kill_block_super(sb);
3898 static struct file_system_type f2fs_fs_type = {
3899 .owner = THIS_MODULE,
3901 .mount = f2fs_mount,
3902 .kill_sb = kill_f2fs_super,
3903 .fs_flags = FS_REQUIRES_DEV,
3905 MODULE_ALIAS_FS("f2fs");
3907 static int __init init_inodecache(void)
3909 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3910 sizeof(struct f2fs_inode_info), 0,
3911 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3912 if (!f2fs_inode_cachep)
3917 static void destroy_inodecache(void)
3920 * Make sure all delayed rcu free inodes are flushed before we
3924 kmem_cache_destroy(f2fs_inode_cachep);
3927 static int __init init_f2fs_fs(void)
3931 if (PAGE_SIZE != F2FS_BLKSIZE) {
3932 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3933 PAGE_SIZE, F2FS_BLKSIZE);
3937 f2fs_build_trace_ios();
3939 err = init_inodecache();
3942 err = f2fs_create_node_manager_caches();
3944 goto free_inodecache;
3945 err = f2fs_create_segment_manager_caches();
3947 goto free_node_manager_caches;
3948 err = f2fs_create_checkpoint_caches();
3950 goto free_segment_manager_caches;
3951 err = f2fs_create_extent_cache();
3953 goto free_checkpoint_caches;
3954 err = f2fs_init_sysfs();
3956 goto free_extent_cache;
3957 err = register_shrinker(&f2fs_shrinker_info);
3960 err = register_filesystem(&f2fs_fs_type);
3963 f2fs_create_root_stats();
3964 err = f2fs_init_post_read_processing();
3966 goto free_root_stats;
3967 err = f2fs_init_bio_entry_cache();
3969 goto free_post_read;
3970 err = f2fs_init_bioset();
3972 goto free_bio_enrty_cache;
3973 err = f2fs_init_compress_mempool();
3978 f2fs_destroy_bioset();
3979 free_bio_enrty_cache:
3980 f2fs_destroy_bio_entry_cache();
3982 f2fs_destroy_post_read_processing();
3984 f2fs_destroy_root_stats();
3985 unregister_filesystem(&f2fs_fs_type);
3987 unregister_shrinker(&f2fs_shrinker_info);
3991 f2fs_destroy_extent_cache();
3992 free_checkpoint_caches:
3993 f2fs_destroy_checkpoint_caches();
3994 free_segment_manager_caches:
3995 f2fs_destroy_segment_manager_caches();
3996 free_node_manager_caches:
3997 f2fs_destroy_node_manager_caches();
3999 destroy_inodecache();
4004 static void __exit exit_f2fs_fs(void)
4006 f2fs_destroy_compress_mempool();
4007 f2fs_destroy_bioset();
4008 f2fs_destroy_bio_entry_cache();
4009 f2fs_destroy_post_read_processing();
4010 f2fs_destroy_root_stats();
4011 unregister_filesystem(&f2fs_fs_type);
4012 unregister_shrinker(&f2fs_shrinker_info);
4014 f2fs_destroy_extent_cache();
4015 f2fs_destroy_checkpoint_caches();
4016 f2fs_destroy_segment_manager_caches();
4017 f2fs_destroy_node_manager_caches();
4018 destroy_inodecache();
4019 f2fs_destroy_trace_ios();
4022 module_init(init_f2fs_fs)
4023 module_exit(exit_f2fs_fs)
4025 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
4026 MODULE_DESCRIPTION("Flash Friendly File System");
4027 MODULE_LICENSE("GPL");