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,
141 Opt_checkpoint_disable,
142 Opt_checkpoint_disable_cap,
143 Opt_checkpoint_disable_cap_perc,
144 Opt_checkpoint_enable,
145 Opt_compress_algorithm,
146 Opt_compress_log_size,
147 Opt_compress_extension,
151 static match_table_t f2fs_tokens = {
152 {Opt_gc_background, "background_gc=%s"},
153 {Opt_disable_roll_forward, "disable_roll_forward"},
154 {Opt_norecovery, "norecovery"},
155 {Opt_discard, "discard"},
156 {Opt_nodiscard, "nodiscard"},
157 {Opt_noheap, "no_heap"},
159 {Opt_user_xattr, "user_xattr"},
160 {Opt_nouser_xattr, "nouser_xattr"},
162 {Opt_noacl, "noacl"},
163 {Opt_active_logs, "active_logs=%u"},
164 {Opt_disable_ext_identify, "disable_ext_identify"},
165 {Opt_inline_xattr, "inline_xattr"},
166 {Opt_noinline_xattr, "noinline_xattr"},
167 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
168 {Opt_inline_data, "inline_data"},
169 {Opt_inline_dentry, "inline_dentry"},
170 {Opt_noinline_dentry, "noinline_dentry"},
171 {Opt_flush_merge, "flush_merge"},
172 {Opt_noflush_merge, "noflush_merge"},
173 {Opt_nobarrier, "nobarrier"},
174 {Opt_fastboot, "fastboot"},
175 {Opt_extent_cache, "extent_cache"},
176 {Opt_noextent_cache, "noextent_cache"},
177 {Opt_noinline_data, "noinline_data"},
178 {Opt_data_flush, "data_flush"},
179 {Opt_reserve_root, "reserve_root=%u"},
180 {Opt_resgid, "resgid=%u"},
181 {Opt_resuid, "resuid=%u"},
182 {Opt_mode, "mode=%s"},
183 {Opt_io_size_bits, "io_bits=%u"},
184 {Opt_fault_injection, "fault_injection=%u"},
185 {Opt_fault_type, "fault_type=%u"},
186 {Opt_lazytime, "lazytime"},
187 {Opt_nolazytime, "nolazytime"},
188 {Opt_quota, "quota"},
189 {Opt_noquota, "noquota"},
190 {Opt_usrquota, "usrquota"},
191 {Opt_grpquota, "grpquota"},
192 {Opt_prjquota, "prjquota"},
193 {Opt_usrjquota, "usrjquota=%s"},
194 {Opt_grpjquota, "grpjquota=%s"},
195 {Opt_prjjquota, "prjjquota=%s"},
196 {Opt_offusrjquota, "usrjquota="},
197 {Opt_offgrpjquota, "grpjquota="},
198 {Opt_offprjjquota, "prjjquota="},
199 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
200 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
201 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
202 {Opt_whint, "whint_mode=%s"},
203 {Opt_alloc, "alloc_mode=%s"},
204 {Opt_fsync, "fsync_mode=%s"},
205 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
206 {Opt_test_dummy_encryption, "test_dummy_encryption"},
207 {Opt_checkpoint_disable, "checkpoint=disable"},
208 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
209 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
210 {Opt_checkpoint_enable, "checkpoint=enable"},
211 {Opt_compress_algorithm, "compress_algorithm=%s"},
212 {Opt_compress_log_size, "compress_log_size=%u"},
213 {Opt_compress_extension, "compress_extension=%s"},
217 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
219 struct va_format vaf;
225 level = printk_get_level(fmt);
226 vaf.fmt = printk_skip_level(fmt);
228 printk("%c%cF2FS-fs (%s): %pV\n",
229 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
234 #ifdef CONFIG_UNICODE
235 static const struct f2fs_sb_encodings {
239 } f2fs_sb_encoding_map[] = {
240 {F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"},
243 static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb,
244 const struct f2fs_sb_encodings **encoding,
247 __u16 magic = le16_to_cpu(sb->s_encoding);
250 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
251 if (magic == f2fs_sb_encoding_map[i].magic)
254 if (i >= ARRAY_SIZE(f2fs_sb_encoding_map))
257 *encoding = &f2fs_sb_encoding_map[i];
258 *flags = le16_to_cpu(sb->s_encoding_flags);
264 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
266 block_t limit = min((sbi->user_block_count << 1) / 1000,
267 sbi->user_block_count - sbi->reserved_blocks);
270 if (test_opt(sbi, RESERVE_ROOT) &&
271 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
272 F2FS_OPTION(sbi).root_reserved_blocks = limit;
273 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
274 F2FS_OPTION(sbi).root_reserved_blocks);
276 if (!test_opt(sbi, RESERVE_ROOT) &&
277 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
278 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
279 !gid_eq(F2FS_OPTION(sbi).s_resgid,
280 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
281 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
282 from_kuid_munged(&init_user_ns,
283 F2FS_OPTION(sbi).s_resuid),
284 from_kgid_munged(&init_user_ns,
285 F2FS_OPTION(sbi).s_resgid));
288 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
290 if (!F2FS_OPTION(sbi).unusable_cap_perc)
293 if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
294 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
296 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
297 F2FS_OPTION(sbi).unusable_cap_perc;
299 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
300 F2FS_OPTION(sbi).unusable_cap,
301 F2FS_OPTION(sbi).unusable_cap_perc);
304 static void init_once(void *foo)
306 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
308 inode_init_once(&fi->vfs_inode);
312 static const char * const quotatypes[] = INITQFNAMES;
313 #define QTYPE2NAME(t) (quotatypes[t])
314 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
317 struct f2fs_sb_info *sbi = F2FS_SB(sb);
321 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
322 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
325 if (f2fs_sb_has_quota_ino(sbi)) {
326 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
330 qname = match_strdup(args);
332 f2fs_err(sbi, "Not enough memory for storing quotafile name");
335 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
336 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
339 f2fs_err(sbi, "%s quota file already specified",
343 if (strchr(qname, '/')) {
344 f2fs_err(sbi, "quotafile must be on filesystem root");
347 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
355 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
357 struct f2fs_sb_info *sbi = F2FS_SB(sb);
359 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
360 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
363 kvfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
364 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
368 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
371 * We do the test below only for project quotas. 'usrquota' and
372 * 'grpquota' mount options are allowed even without quota feature
373 * to support legacy quotas in quota files.
375 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
376 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
379 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
380 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
381 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
382 if (test_opt(sbi, USRQUOTA) &&
383 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
384 clear_opt(sbi, USRQUOTA);
386 if (test_opt(sbi, GRPQUOTA) &&
387 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
388 clear_opt(sbi, GRPQUOTA);
390 if (test_opt(sbi, PRJQUOTA) &&
391 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
392 clear_opt(sbi, PRJQUOTA);
394 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
395 test_opt(sbi, PRJQUOTA)) {
396 f2fs_err(sbi, "old and new quota format mixing");
400 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
401 f2fs_err(sbi, "journaled quota format not specified");
406 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
407 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
408 F2FS_OPTION(sbi).s_jquota_fmt = 0;
414 static int f2fs_set_test_dummy_encryption(struct super_block *sb,
416 const substring_t *arg,
419 struct f2fs_sb_info *sbi = F2FS_SB(sb);
420 #ifdef CONFIG_FS_ENCRYPTION
423 if (!f2fs_sb_has_encrypt(sbi)) {
424 f2fs_err(sbi, "Encrypt feature is off");
429 * This mount option is just for testing, and it's not worthwhile to
430 * implement the extra complexity (e.g. RCU protection) that would be
431 * needed to allow it to be set or changed during remount. We do allow
432 * it to be specified during remount, but only if there is no change.
434 if (is_remount && !F2FS_OPTION(sbi).dummy_enc_ctx.ctx) {
435 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
438 err = fscrypt_set_test_dummy_encryption(
439 sb, arg, &F2FS_OPTION(sbi).dummy_enc_ctx);
443 "Can't change test_dummy_encryption on remount");
444 else if (err == -EINVAL)
445 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
448 f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
452 f2fs_warn(sbi, "Test dummy encryption mode enabled");
454 f2fs_warn(sbi, "Test dummy encryption mount option ignored");
459 static int parse_options(struct super_block *sb, char *options, bool is_remount)
461 struct f2fs_sb_info *sbi = F2FS_SB(sb);
462 substring_t args[MAX_OPT_ARGS];
463 unsigned char (*ext)[F2FS_EXTENSION_LEN];
465 int arg = 0, ext_cnt;
473 while ((p = strsep(&options, ",")) != NULL) {
478 * Initialize args struct so we know whether arg was
479 * found; some options take optional arguments.
481 args[0].to = args[0].from = NULL;
482 token = match_token(p, f2fs_tokens, args);
485 case Opt_gc_background:
486 name = match_strdup(&args[0]);
490 if (!strcmp(name, "on")) {
491 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
492 } else if (!strcmp(name, "off")) {
493 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
494 } else if (!strcmp(name, "sync")) {
495 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
502 case Opt_disable_roll_forward:
503 set_opt(sbi, DISABLE_ROLL_FORWARD);
506 /* this option mounts f2fs with ro */
507 set_opt(sbi, NORECOVERY);
508 if (!f2fs_readonly(sb))
512 set_opt(sbi, DISCARD);
515 if (f2fs_sb_has_blkzoned(sbi)) {
516 f2fs_warn(sbi, "discard is required for zoned block devices");
519 clear_opt(sbi, DISCARD);
522 set_opt(sbi, NOHEAP);
525 clear_opt(sbi, NOHEAP);
527 #ifdef CONFIG_F2FS_FS_XATTR
529 set_opt(sbi, XATTR_USER);
531 case Opt_nouser_xattr:
532 clear_opt(sbi, XATTR_USER);
534 case Opt_inline_xattr:
535 set_opt(sbi, INLINE_XATTR);
537 case Opt_noinline_xattr:
538 clear_opt(sbi, INLINE_XATTR);
540 case Opt_inline_xattr_size:
541 if (args->from && match_int(args, &arg))
543 set_opt(sbi, INLINE_XATTR_SIZE);
544 F2FS_OPTION(sbi).inline_xattr_size = arg;
548 f2fs_info(sbi, "user_xattr options not supported");
550 case Opt_nouser_xattr:
551 f2fs_info(sbi, "nouser_xattr options not supported");
553 case Opt_inline_xattr:
554 f2fs_info(sbi, "inline_xattr options not supported");
556 case Opt_noinline_xattr:
557 f2fs_info(sbi, "noinline_xattr options not supported");
560 #ifdef CONFIG_F2FS_FS_POSIX_ACL
562 set_opt(sbi, POSIX_ACL);
565 clear_opt(sbi, POSIX_ACL);
569 f2fs_info(sbi, "acl options not supported");
572 f2fs_info(sbi, "noacl options not supported");
575 case Opt_active_logs:
576 if (args->from && match_int(args, &arg))
578 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
580 F2FS_OPTION(sbi).active_logs = arg;
582 case Opt_disable_ext_identify:
583 set_opt(sbi, DISABLE_EXT_IDENTIFY);
585 case Opt_inline_data:
586 set_opt(sbi, INLINE_DATA);
588 case Opt_inline_dentry:
589 set_opt(sbi, INLINE_DENTRY);
591 case Opt_noinline_dentry:
592 clear_opt(sbi, INLINE_DENTRY);
594 case Opt_flush_merge:
595 set_opt(sbi, FLUSH_MERGE);
597 case Opt_noflush_merge:
598 clear_opt(sbi, FLUSH_MERGE);
601 set_opt(sbi, NOBARRIER);
604 set_opt(sbi, FASTBOOT);
606 case Opt_extent_cache:
607 set_opt(sbi, EXTENT_CACHE);
609 case Opt_noextent_cache:
610 clear_opt(sbi, EXTENT_CACHE);
612 case Opt_noinline_data:
613 clear_opt(sbi, INLINE_DATA);
616 set_opt(sbi, DATA_FLUSH);
618 case Opt_reserve_root:
619 if (args->from && match_int(args, &arg))
621 if (test_opt(sbi, RESERVE_ROOT)) {
622 f2fs_info(sbi, "Preserve previous reserve_root=%u",
623 F2FS_OPTION(sbi).root_reserved_blocks);
625 F2FS_OPTION(sbi).root_reserved_blocks = arg;
626 set_opt(sbi, RESERVE_ROOT);
630 if (args->from && match_int(args, &arg))
632 uid = make_kuid(current_user_ns(), arg);
633 if (!uid_valid(uid)) {
634 f2fs_err(sbi, "Invalid uid value %d", arg);
637 F2FS_OPTION(sbi).s_resuid = uid;
640 if (args->from && match_int(args, &arg))
642 gid = make_kgid(current_user_ns(), arg);
643 if (!gid_valid(gid)) {
644 f2fs_err(sbi, "Invalid gid value %d", arg);
647 F2FS_OPTION(sbi).s_resgid = gid;
650 name = match_strdup(&args[0]);
654 if (!strcmp(name, "adaptive")) {
655 if (f2fs_sb_has_blkzoned(sbi)) {
656 f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
660 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
661 } else if (!strcmp(name, "lfs")) {
662 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
669 case Opt_io_size_bits:
670 if (args->from && match_int(args, &arg))
672 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_PAGES)) {
673 f2fs_warn(sbi, "Not support %d, larger than %d",
674 1 << arg, BIO_MAX_PAGES);
677 F2FS_OPTION(sbi).write_io_size_bits = arg;
679 #ifdef CONFIG_F2FS_FAULT_INJECTION
680 case Opt_fault_injection:
681 if (args->from && match_int(args, &arg))
683 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
684 set_opt(sbi, FAULT_INJECTION);
688 if (args->from && match_int(args, &arg))
690 f2fs_build_fault_attr(sbi, 0, arg);
691 set_opt(sbi, FAULT_INJECTION);
694 case Opt_fault_injection:
695 f2fs_info(sbi, "fault_injection options not supported");
699 f2fs_info(sbi, "fault_type options not supported");
703 sb->s_flags |= SB_LAZYTIME;
706 sb->s_flags &= ~SB_LAZYTIME;
711 set_opt(sbi, USRQUOTA);
714 set_opt(sbi, GRPQUOTA);
717 set_opt(sbi, PRJQUOTA);
720 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
725 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
730 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
734 case Opt_offusrjquota:
735 ret = f2fs_clear_qf_name(sb, USRQUOTA);
739 case Opt_offgrpjquota:
740 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
744 case Opt_offprjjquota:
745 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
749 case Opt_jqfmt_vfsold:
750 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
752 case Opt_jqfmt_vfsv0:
753 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
755 case Opt_jqfmt_vfsv1:
756 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
759 clear_opt(sbi, QUOTA);
760 clear_opt(sbi, USRQUOTA);
761 clear_opt(sbi, GRPQUOTA);
762 clear_opt(sbi, PRJQUOTA);
772 case Opt_offusrjquota:
773 case Opt_offgrpjquota:
774 case Opt_offprjjquota:
775 case Opt_jqfmt_vfsold:
776 case Opt_jqfmt_vfsv0:
777 case Opt_jqfmt_vfsv1:
779 f2fs_info(sbi, "quota operations not supported");
783 name = match_strdup(&args[0]);
786 if (!strcmp(name, "user-based")) {
787 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
788 } else if (!strcmp(name, "off")) {
789 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
790 } else if (!strcmp(name, "fs-based")) {
791 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
799 name = match_strdup(&args[0]);
803 if (!strcmp(name, "default")) {
804 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
805 } else if (!strcmp(name, "reuse")) {
806 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
814 name = match_strdup(&args[0]);
817 if (!strcmp(name, "posix")) {
818 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
819 } else if (!strcmp(name, "strict")) {
820 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
821 } else if (!strcmp(name, "nobarrier")) {
822 F2FS_OPTION(sbi).fsync_mode =
823 FSYNC_MODE_NOBARRIER;
830 case Opt_test_dummy_encryption:
831 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
836 case Opt_checkpoint_disable_cap_perc:
837 if (args->from && match_int(args, &arg))
839 if (arg < 0 || arg > 100)
841 F2FS_OPTION(sbi).unusable_cap_perc = arg;
842 set_opt(sbi, DISABLE_CHECKPOINT);
844 case Opt_checkpoint_disable_cap:
845 if (args->from && match_int(args, &arg))
847 F2FS_OPTION(sbi).unusable_cap = arg;
848 set_opt(sbi, DISABLE_CHECKPOINT);
850 case Opt_checkpoint_disable:
851 set_opt(sbi, DISABLE_CHECKPOINT);
853 case Opt_checkpoint_enable:
854 clear_opt(sbi, DISABLE_CHECKPOINT);
856 case Opt_compress_algorithm:
857 if (!f2fs_sb_has_compression(sbi)) {
858 f2fs_err(sbi, "Compression feature if off");
861 name = match_strdup(&args[0]);
864 if (!strcmp(name, "lzo")) {
865 F2FS_OPTION(sbi).compress_algorithm =
867 } else if (!strcmp(name, "lz4")) {
868 F2FS_OPTION(sbi).compress_algorithm =
870 } else if (!strcmp(name, "zstd")) {
871 F2FS_OPTION(sbi).compress_algorithm =
873 } else if (!strcmp(name, "lzo-rle")) {
874 F2FS_OPTION(sbi).compress_algorithm =
882 case Opt_compress_log_size:
883 if (!f2fs_sb_has_compression(sbi)) {
884 f2fs_err(sbi, "Compression feature is off");
887 if (args->from && match_int(args, &arg))
889 if (arg < MIN_COMPRESS_LOG_SIZE ||
890 arg > MAX_COMPRESS_LOG_SIZE) {
892 "Compress cluster log size is out of range");
895 F2FS_OPTION(sbi).compress_log_size = arg;
897 case Opt_compress_extension:
898 if (!f2fs_sb_has_compression(sbi)) {
899 f2fs_err(sbi, "Compression feature is off");
902 name = match_strdup(&args[0]);
906 ext = F2FS_OPTION(sbi).extensions;
907 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
909 if (strlen(name) >= F2FS_EXTENSION_LEN ||
910 ext_cnt >= COMPRESS_EXT_NUM) {
912 "invalid extension length/number");
917 strcpy(ext[ext_cnt], name);
918 F2FS_OPTION(sbi).compress_ext_cnt++;
922 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
928 if (f2fs_check_quota_options(sbi))
931 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
932 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
935 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
936 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
940 #ifndef CONFIG_UNICODE
941 if (f2fs_sb_has_casefold(sbi)) {
943 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
948 if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) {
949 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
950 F2FS_IO_SIZE_KB(sbi));
954 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
955 int min_size, max_size;
957 if (!f2fs_sb_has_extra_attr(sbi) ||
958 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
959 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
962 if (!test_opt(sbi, INLINE_XATTR)) {
963 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
967 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
968 max_size = MAX_INLINE_XATTR_SIZE;
970 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
971 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
972 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
978 if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) {
979 f2fs_err(sbi, "LFS not compatible with checkpoint=disable\n");
983 /* Not pass down write hints if the number of active logs is lesser
984 * than NR_CURSEG_TYPE.
986 if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
987 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
991 static struct inode *f2fs_alloc_inode(struct super_block *sb)
993 struct f2fs_inode_info *fi;
995 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
999 init_once((void *) fi);
1001 /* Initialize f2fs-specific inode info */
1002 atomic_set(&fi->dirty_pages, 0);
1003 init_rwsem(&fi->i_sem);
1004 spin_lock_init(&fi->i_size_lock);
1005 INIT_LIST_HEAD(&fi->dirty_list);
1006 INIT_LIST_HEAD(&fi->gdirty_list);
1007 INIT_LIST_HEAD(&fi->inmem_ilist);
1008 INIT_LIST_HEAD(&fi->inmem_pages);
1009 mutex_init(&fi->inmem_lock);
1010 init_rwsem(&fi->i_gc_rwsem[READ]);
1011 init_rwsem(&fi->i_gc_rwsem[WRITE]);
1012 init_rwsem(&fi->i_mmap_sem);
1013 init_rwsem(&fi->i_xattr_sem);
1015 /* Will be used by directory only */
1016 fi->i_dir_level = F2FS_SB(sb)->dir_level;
1018 return &fi->vfs_inode;
1021 static int f2fs_drop_inode(struct inode *inode)
1023 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1027 * during filesystem shutdown, if checkpoint is disabled,
1028 * drop useless meta/node dirty pages.
1030 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1031 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1032 inode->i_ino == F2FS_META_INO(sbi)) {
1033 trace_f2fs_drop_inode(inode, 1);
1039 * This is to avoid a deadlock condition like below.
1040 * writeback_single_inode(inode)
1041 * - f2fs_write_data_page
1042 * - f2fs_gc -> iput -> evict
1043 * - inode_wait_for_writeback(inode)
1045 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1046 if (!inode->i_nlink && !is_bad_inode(inode)) {
1047 /* to avoid evict_inode call simultaneously */
1048 atomic_inc(&inode->i_count);
1049 spin_unlock(&inode->i_lock);
1051 /* some remained atomic pages should discarded */
1052 if (f2fs_is_atomic_file(inode))
1053 f2fs_drop_inmem_pages(inode);
1055 /* should remain fi->extent_tree for writepage */
1056 f2fs_destroy_extent_node(inode);
1058 sb_start_intwrite(inode->i_sb);
1059 f2fs_i_size_write(inode, 0);
1061 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1062 inode, NULL, 0, DATA);
1063 truncate_inode_pages_final(inode->i_mapping);
1065 if (F2FS_HAS_BLOCKS(inode))
1066 f2fs_truncate(inode);
1068 sb_end_intwrite(inode->i_sb);
1070 spin_lock(&inode->i_lock);
1071 atomic_dec(&inode->i_count);
1073 trace_f2fs_drop_inode(inode, 0);
1076 ret = generic_drop_inode(inode);
1078 ret = fscrypt_drop_inode(inode);
1079 trace_f2fs_drop_inode(inode, ret);
1083 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1085 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1088 spin_lock(&sbi->inode_lock[DIRTY_META]);
1089 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1092 set_inode_flag(inode, FI_DIRTY_INODE);
1093 stat_inc_dirty_inode(sbi, DIRTY_META);
1095 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1096 list_add_tail(&F2FS_I(inode)->gdirty_list,
1097 &sbi->inode_list[DIRTY_META]);
1098 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1100 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1104 void f2fs_inode_synced(struct inode *inode)
1106 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1108 spin_lock(&sbi->inode_lock[DIRTY_META]);
1109 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1110 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1113 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1114 list_del_init(&F2FS_I(inode)->gdirty_list);
1115 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1117 clear_inode_flag(inode, FI_DIRTY_INODE);
1118 clear_inode_flag(inode, FI_AUTO_RECOVER);
1119 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1120 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1124 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1126 * We should call set_dirty_inode to write the dirty inode through write_inode.
1128 static void f2fs_dirty_inode(struct inode *inode, int flags)
1130 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1132 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1133 inode->i_ino == F2FS_META_INO(sbi))
1136 if (flags == I_DIRTY_TIME)
1139 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1140 clear_inode_flag(inode, FI_AUTO_RECOVER);
1142 f2fs_inode_dirtied(inode, false);
1145 static void f2fs_free_inode(struct inode *inode)
1147 fscrypt_free_inode(inode);
1148 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1151 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1153 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1154 percpu_counter_destroy(&sbi->total_valid_inode_count);
1157 static void destroy_device_list(struct f2fs_sb_info *sbi)
1161 for (i = 0; i < sbi->s_ndevs; i++) {
1162 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1163 #ifdef CONFIG_BLK_DEV_ZONED
1164 kvfree(FDEV(i).blkz_seq);
1170 static void f2fs_put_super(struct super_block *sb)
1172 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1176 f2fs_quota_off_umount(sb);
1178 /* prevent remaining shrinker jobs */
1179 mutex_lock(&sbi->umount_mutex);
1182 * We don't need to do checkpoint when superblock is clean.
1183 * But, the previous checkpoint was not done by umount, it needs to do
1184 * clean checkpoint again.
1186 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1187 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1188 struct cp_control cpc = {
1189 .reason = CP_UMOUNT,
1191 f2fs_write_checkpoint(sbi, &cpc);
1194 /* be sure to wait for any on-going discard commands */
1195 dropped = f2fs_issue_discard_timeout(sbi);
1197 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1198 !sbi->discard_blks && !dropped) {
1199 struct cp_control cpc = {
1200 .reason = CP_UMOUNT | CP_TRIMMED,
1202 f2fs_write_checkpoint(sbi, &cpc);
1206 * normally superblock is clean, so we need to release this.
1207 * In addition, EIO will skip do checkpoint, we need this as well.
1209 f2fs_release_ino_entry(sbi, true);
1211 f2fs_leave_shrinker(sbi);
1212 mutex_unlock(&sbi->umount_mutex);
1214 /* our cp_error case, we can wait for any writeback page */
1215 f2fs_flush_merged_writes(sbi);
1217 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1219 f2fs_bug_on(sbi, sbi->fsync_node_num);
1221 iput(sbi->node_inode);
1222 sbi->node_inode = NULL;
1224 iput(sbi->meta_inode);
1225 sbi->meta_inode = NULL;
1228 * iput() can update stat information, if f2fs_write_checkpoint()
1229 * above failed with error.
1231 f2fs_destroy_stats(sbi);
1233 /* destroy f2fs internal modules */
1234 f2fs_destroy_node_manager(sbi);
1235 f2fs_destroy_segment_manager(sbi);
1237 f2fs_destroy_post_read_wq(sbi);
1241 f2fs_unregister_sysfs(sbi);
1243 sb->s_fs_info = NULL;
1244 if (sbi->s_chksum_driver)
1245 crypto_free_shash(sbi->s_chksum_driver);
1246 kvfree(sbi->raw_super);
1248 destroy_device_list(sbi);
1249 f2fs_destroy_xattr_caches(sbi);
1250 mempool_destroy(sbi->write_io_dummy);
1252 for (i = 0; i < MAXQUOTAS; i++)
1253 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1255 fscrypt_free_dummy_context(&F2FS_OPTION(sbi).dummy_enc_ctx);
1256 destroy_percpu_info(sbi);
1257 for (i = 0; i < NR_PAGE_TYPE; i++)
1258 kvfree(sbi->write_io[i]);
1259 #ifdef CONFIG_UNICODE
1260 utf8_unload(sbi->s_encoding);
1265 int f2fs_sync_fs(struct super_block *sb, int sync)
1267 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1270 if (unlikely(f2fs_cp_error(sbi)))
1272 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1275 trace_f2fs_sync_fs(sb, sync);
1277 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1281 struct cp_control cpc;
1283 cpc.reason = __get_cp_reason(sbi);
1285 down_write(&sbi->gc_lock);
1286 err = f2fs_write_checkpoint(sbi, &cpc);
1287 up_write(&sbi->gc_lock);
1289 f2fs_trace_ios(NULL, 1);
1294 static int f2fs_freeze(struct super_block *sb)
1296 if (f2fs_readonly(sb))
1299 /* IO error happened before */
1300 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1303 /* must be clean, since sync_filesystem() was already called */
1304 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1309 static int f2fs_unfreeze(struct super_block *sb)
1315 static int f2fs_statfs_project(struct super_block *sb,
1316 kprojid_t projid, struct kstatfs *buf)
1319 struct dquot *dquot;
1323 qid = make_kqid_projid(projid);
1324 dquot = dqget(sb, qid);
1326 return PTR_ERR(dquot);
1327 spin_lock(&dquot->dq_dqb_lock);
1329 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1330 dquot->dq_dqb.dqb_bhardlimit);
1332 limit >>= sb->s_blocksize_bits;
1334 if (limit && buf->f_blocks > limit) {
1335 curblock = (dquot->dq_dqb.dqb_curspace +
1336 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1337 buf->f_blocks = limit;
1338 buf->f_bfree = buf->f_bavail =
1339 (buf->f_blocks > curblock) ?
1340 (buf->f_blocks - curblock) : 0;
1343 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1344 dquot->dq_dqb.dqb_ihardlimit);
1346 if (limit && buf->f_files > limit) {
1347 buf->f_files = limit;
1349 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1350 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1353 spin_unlock(&dquot->dq_dqb_lock);
1359 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1361 struct super_block *sb = dentry->d_sb;
1362 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1363 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1364 block_t total_count, user_block_count, start_count;
1365 u64 avail_node_count;
1367 total_count = le64_to_cpu(sbi->raw_super->block_count);
1368 user_block_count = sbi->user_block_count;
1369 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1370 buf->f_type = F2FS_SUPER_MAGIC;
1371 buf->f_bsize = sbi->blocksize;
1373 buf->f_blocks = total_count - start_count;
1374 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1375 sbi->current_reserved_blocks;
1377 spin_lock(&sbi->stat_lock);
1378 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1381 buf->f_bfree -= sbi->unusable_block_count;
1382 spin_unlock(&sbi->stat_lock);
1384 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1385 buf->f_bavail = buf->f_bfree -
1386 F2FS_OPTION(sbi).root_reserved_blocks;
1390 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1392 if (avail_node_count > user_block_count) {
1393 buf->f_files = user_block_count;
1394 buf->f_ffree = buf->f_bavail;
1396 buf->f_files = avail_node_count;
1397 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1401 buf->f_namelen = F2FS_NAME_LEN;
1402 buf->f_fsid.val[0] = (u32)id;
1403 buf->f_fsid.val[1] = (u32)(id >> 32);
1406 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1407 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1408 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1414 static inline void f2fs_show_quota_options(struct seq_file *seq,
1415 struct super_block *sb)
1418 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1420 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1423 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1434 seq_printf(seq, ",jqfmt=%s", fmtname);
1437 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1438 seq_show_option(seq, "usrjquota",
1439 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1441 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1442 seq_show_option(seq, "grpjquota",
1443 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1445 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1446 seq_show_option(seq, "prjjquota",
1447 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1451 static inline void f2fs_show_compress_options(struct seq_file *seq,
1452 struct super_block *sb)
1454 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1458 if (!f2fs_sb_has_compression(sbi))
1461 switch (F2FS_OPTION(sbi).compress_algorithm) {
1471 case COMPRESS_LZORLE:
1472 algtype = "lzo-rle";
1475 seq_printf(seq, ",compress_algorithm=%s", algtype);
1477 seq_printf(seq, ",compress_log_size=%u",
1478 F2FS_OPTION(sbi).compress_log_size);
1480 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1481 seq_printf(seq, ",compress_extension=%s",
1482 F2FS_OPTION(sbi).extensions[i]);
1486 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1488 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1490 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1491 seq_printf(seq, ",background_gc=%s", "sync");
1492 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1493 seq_printf(seq, ",background_gc=%s", "on");
1494 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1495 seq_printf(seq, ",background_gc=%s", "off");
1497 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1498 seq_puts(seq, ",disable_roll_forward");
1499 if (test_opt(sbi, NORECOVERY))
1500 seq_puts(seq, ",norecovery");
1501 if (test_opt(sbi, DISCARD))
1502 seq_puts(seq, ",discard");
1504 seq_puts(seq, ",nodiscard");
1505 if (test_opt(sbi, NOHEAP))
1506 seq_puts(seq, ",no_heap");
1508 seq_puts(seq, ",heap");
1509 #ifdef CONFIG_F2FS_FS_XATTR
1510 if (test_opt(sbi, XATTR_USER))
1511 seq_puts(seq, ",user_xattr");
1513 seq_puts(seq, ",nouser_xattr");
1514 if (test_opt(sbi, INLINE_XATTR))
1515 seq_puts(seq, ",inline_xattr");
1517 seq_puts(seq, ",noinline_xattr");
1518 if (test_opt(sbi, INLINE_XATTR_SIZE))
1519 seq_printf(seq, ",inline_xattr_size=%u",
1520 F2FS_OPTION(sbi).inline_xattr_size);
1522 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1523 if (test_opt(sbi, POSIX_ACL))
1524 seq_puts(seq, ",acl");
1526 seq_puts(seq, ",noacl");
1528 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1529 seq_puts(seq, ",disable_ext_identify");
1530 if (test_opt(sbi, INLINE_DATA))
1531 seq_puts(seq, ",inline_data");
1533 seq_puts(seq, ",noinline_data");
1534 if (test_opt(sbi, INLINE_DENTRY))
1535 seq_puts(seq, ",inline_dentry");
1537 seq_puts(seq, ",noinline_dentry");
1538 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1539 seq_puts(seq, ",flush_merge");
1540 if (test_opt(sbi, NOBARRIER))
1541 seq_puts(seq, ",nobarrier");
1542 if (test_opt(sbi, FASTBOOT))
1543 seq_puts(seq, ",fastboot");
1544 if (test_opt(sbi, EXTENT_CACHE))
1545 seq_puts(seq, ",extent_cache");
1547 seq_puts(seq, ",noextent_cache");
1548 if (test_opt(sbi, DATA_FLUSH))
1549 seq_puts(seq, ",data_flush");
1551 seq_puts(seq, ",mode=");
1552 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
1553 seq_puts(seq, "adaptive");
1554 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
1555 seq_puts(seq, "lfs");
1556 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1557 if (test_opt(sbi, RESERVE_ROOT))
1558 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1559 F2FS_OPTION(sbi).root_reserved_blocks,
1560 from_kuid_munged(&init_user_ns,
1561 F2FS_OPTION(sbi).s_resuid),
1562 from_kgid_munged(&init_user_ns,
1563 F2FS_OPTION(sbi).s_resgid));
1564 if (F2FS_IO_SIZE_BITS(sbi))
1565 seq_printf(seq, ",io_bits=%u",
1566 F2FS_OPTION(sbi).write_io_size_bits);
1567 #ifdef CONFIG_F2FS_FAULT_INJECTION
1568 if (test_opt(sbi, FAULT_INJECTION)) {
1569 seq_printf(seq, ",fault_injection=%u",
1570 F2FS_OPTION(sbi).fault_info.inject_rate);
1571 seq_printf(seq, ",fault_type=%u",
1572 F2FS_OPTION(sbi).fault_info.inject_type);
1576 if (test_opt(sbi, QUOTA))
1577 seq_puts(seq, ",quota");
1578 if (test_opt(sbi, USRQUOTA))
1579 seq_puts(seq, ",usrquota");
1580 if (test_opt(sbi, GRPQUOTA))
1581 seq_puts(seq, ",grpquota");
1582 if (test_opt(sbi, PRJQUOTA))
1583 seq_puts(seq, ",prjquota");
1585 f2fs_show_quota_options(seq, sbi->sb);
1586 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1587 seq_printf(seq, ",whint_mode=%s", "user-based");
1588 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1589 seq_printf(seq, ",whint_mode=%s", "fs-based");
1591 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
1593 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1594 seq_printf(seq, ",alloc_mode=%s", "default");
1595 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1596 seq_printf(seq, ",alloc_mode=%s", "reuse");
1598 if (test_opt(sbi, DISABLE_CHECKPOINT))
1599 seq_printf(seq, ",checkpoint=disable:%u",
1600 F2FS_OPTION(sbi).unusable_cap);
1601 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1602 seq_printf(seq, ",fsync_mode=%s", "posix");
1603 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1604 seq_printf(seq, ",fsync_mode=%s", "strict");
1605 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1606 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1608 f2fs_show_compress_options(seq, sbi->sb);
1612 static void default_options(struct f2fs_sb_info *sbi)
1614 /* init some FS parameters */
1615 F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1616 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1617 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1618 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1619 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1620 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1621 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1622 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
1623 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
1624 F2FS_OPTION(sbi).compress_ext_cnt = 0;
1625 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
1627 set_opt(sbi, INLINE_XATTR);
1628 set_opt(sbi, INLINE_DATA);
1629 set_opt(sbi, INLINE_DENTRY);
1630 set_opt(sbi, EXTENT_CACHE);
1631 set_opt(sbi, NOHEAP);
1632 clear_opt(sbi, DISABLE_CHECKPOINT);
1633 F2FS_OPTION(sbi).unusable_cap = 0;
1634 sbi->sb->s_flags |= SB_LAZYTIME;
1635 set_opt(sbi, FLUSH_MERGE);
1636 set_opt(sbi, DISCARD);
1637 if (f2fs_sb_has_blkzoned(sbi))
1638 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
1640 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
1642 #ifdef CONFIG_F2FS_FS_XATTR
1643 set_opt(sbi, XATTR_USER);
1645 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1646 set_opt(sbi, POSIX_ACL);
1649 f2fs_build_fault_attr(sbi, 0, 0);
1653 static int f2fs_enable_quotas(struct super_block *sb);
1656 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1658 unsigned int s_flags = sbi->sb->s_flags;
1659 struct cp_control cpc;
1664 if (s_flags & SB_RDONLY) {
1665 f2fs_err(sbi, "checkpoint=disable on readonly fs");
1668 sbi->sb->s_flags |= SB_ACTIVE;
1670 f2fs_update_time(sbi, DISABLE_TIME);
1672 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1673 down_write(&sbi->gc_lock);
1674 err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1675 if (err == -ENODATA) {
1679 if (err && err != -EAGAIN)
1683 ret = sync_filesystem(sbi->sb);
1685 err = ret ? ret: err;
1689 unusable = f2fs_get_unusable_blocks(sbi);
1690 if (f2fs_disable_cp_again(sbi, unusable)) {
1695 down_write(&sbi->gc_lock);
1696 cpc.reason = CP_PAUSE;
1697 set_sbi_flag(sbi, SBI_CP_DISABLED);
1698 err = f2fs_write_checkpoint(sbi, &cpc);
1702 spin_lock(&sbi->stat_lock);
1703 sbi->unusable_block_count = unusable;
1704 spin_unlock(&sbi->stat_lock);
1707 up_write(&sbi->gc_lock);
1709 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */
1713 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1715 down_write(&sbi->gc_lock);
1716 f2fs_dirty_to_prefree(sbi);
1718 clear_sbi_flag(sbi, SBI_CP_DISABLED);
1719 set_sbi_flag(sbi, SBI_IS_DIRTY);
1720 up_write(&sbi->gc_lock);
1722 f2fs_sync_fs(sbi->sb, 1);
1725 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1727 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1728 struct f2fs_mount_info org_mount_opt;
1729 unsigned long old_sb_flags;
1731 bool need_restart_gc = false;
1732 bool need_stop_gc = false;
1733 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1734 bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1735 bool no_io_align = !F2FS_IO_ALIGNED(sbi);
1736 bool checkpoint_changed;
1742 * Save the old mount options in case we
1743 * need to restore them.
1745 org_mount_opt = sbi->mount_opt;
1746 old_sb_flags = sb->s_flags;
1749 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1750 for (i = 0; i < MAXQUOTAS; i++) {
1751 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1752 org_mount_opt.s_qf_names[i] =
1753 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1755 if (!org_mount_opt.s_qf_names[i]) {
1756 for (j = 0; j < i; j++)
1757 kvfree(org_mount_opt.s_qf_names[j]);
1761 org_mount_opt.s_qf_names[i] = NULL;
1766 /* recover superblocks we couldn't write due to previous RO mount */
1767 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1768 err = f2fs_commit_super(sbi, false);
1769 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
1772 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1775 default_options(sbi);
1777 /* parse mount options */
1778 err = parse_options(sb, data, true);
1781 checkpoint_changed =
1782 disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1785 * Previous and new state of filesystem is RO,
1786 * so skip checking GC and FLUSH_MERGE conditions.
1788 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1792 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1793 err = dquot_suspend(sb, -1);
1796 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1797 /* dquot_resume needs RW */
1798 sb->s_flags &= ~SB_RDONLY;
1799 if (sb_any_quota_suspended(sb)) {
1800 dquot_resume(sb, -1);
1801 } else if (f2fs_sb_has_quota_ino(sbi)) {
1802 err = f2fs_enable_quotas(sb);
1808 /* disallow enable/disable extent_cache dynamically */
1809 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1811 f2fs_warn(sbi, "switch extent_cache option is not allowed");
1815 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
1817 f2fs_warn(sbi, "switch io_bits option is not allowed");
1821 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1823 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
1828 * We stop the GC thread if FS is mounted as RO
1829 * or if background_gc = off is passed in mount
1830 * option. Also sync the filesystem.
1832 if ((*flags & SB_RDONLY) ||
1833 F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) {
1834 if (sbi->gc_thread) {
1835 f2fs_stop_gc_thread(sbi);
1836 need_restart_gc = true;
1838 } else if (!sbi->gc_thread) {
1839 err = f2fs_start_gc_thread(sbi);
1842 need_stop_gc = true;
1845 if (*flags & SB_RDONLY ||
1846 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1847 writeback_inodes_sb(sb, WB_REASON_SYNC);
1850 set_sbi_flag(sbi, SBI_IS_DIRTY);
1851 set_sbi_flag(sbi, SBI_IS_CLOSE);
1852 f2fs_sync_fs(sb, 1);
1853 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1856 if (checkpoint_changed) {
1857 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1858 err = f2fs_disable_checkpoint(sbi);
1862 f2fs_enable_checkpoint(sbi);
1867 * We stop issue flush thread if FS is mounted as RO
1868 * or if flush_merge is not passed in mount option.
1870 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1871 clear_opt(sbi, FLUSH_MERGE);
1872 f2fs_destroy_flush_cmd_control(sbi, false);
1874 err = f2fs_create_flush_cmd_control(sbi);
1880 /* Release old quota file names */
1881 for (i = 0; i < MAXQUOTAS; i++)
1882 kvfree(org_mount_opt.s_qf_names[i]);
1884 /* Update the POSIXACL Flag */
1885 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1886 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1888 limit_reserve_root(sbi);
1889 adjust_unusable_cap_perc(sbi);
1890 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1893 if (need_restart_gc) {
1894 if (f2fs_start_gc_thread(sbi))
1895 f2fs_warn(sbi, "background gc thread has stopped");
1896 } else if (need_stop_gc) {
1897 f2fs_stop_gc_thread(sbi);
1901 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1902 for (i = 0; i < MAXQUOTAS; i++) {
1903 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1904 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1907 sbi->mount_opt = org_mount_opt;
1908 sb->s_flags = old_sb_flags;
1913 /* Read data from quotafile */
1914 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1915 size_t len, loff_t off)
1917 struct inode *inode = sb_dqopt(sb)->files[type];
1918 struct address_space *mapping = inode->i_mapping;
1919 block_t blkidx = F2FS_BYTES_TO_BLK(off);
1920 int offset = off & (sb->s_blocksize - 1);
1923 loff_t i_size = i_size_read(inode);
1930 if (off + len > i_size)
1933 while (toread > 0) {
1934 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1936 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1938 if (PTR_ERR(page) == -ENOMEM) {
1939 congestion_wait(BLK_RW_ASYNC,
1940 DEFAULT_IO_TIMEOUT);
1943 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1944 return PTR_ERR(page);
1949 if (unlikely(page->mapping != mapping)) {
1950 f2fs_put_page(page, 1);
1953 if (unlikely(!PageUptodate(page))) {
1954 f2fs_put_page(page, 1);
1955 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1959 kaddr = kmap_atomic(page);
1960 memcpy(data, kaddr + offset, tocopy);
1961 kunmap_atomic(kaddr);
1962 f2fs_put_page(page, 1);
1972 /* Write to quotafile */
1973 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1974 const char *data, size_t len, loff_t off)
1976 struct inode *inode = sb_dqopt(sb)->files[type];
1977 struct address_space *mapping = inode->i_mapping;
1978 const struct address_space_operations *a_ops = mapping->a_ops;
1979 int offset = off & (sb->s_blocksize - 1);
1980 size_t towrite = len;
1982 void *fsdata = NULL;
1987 while (towrite > 0) {
1988 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1991 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1993 if (unlikely(err)) {
1994 if (err == -ENOMEM) {
1995 congestion_wait(BLK_RW_ASYNC,
1996 DEFAULT_IO_TIMEOUT);
1999 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2003 kaddr = kmap_atomic(page);
2004 memcpy(kaddr + offset, data, tocopy);
2005 kunmap_atomic(kaddr);
2006 flush_dcache_page(page);
2008 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2019 inode->i_mtime = inode->i_ctime = current_time(inode);
2020 f2fs_mark_inode_dirty_sync(inode, false);
2021 return len - towrite;
2024 static struct dquot **f2fs_get_dquots(struct inode *inode)
2026 return F2FS_I(inode)->i_dquot;
2029 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2031 return &F2FS_I(inode)->i_reserved_quota;
2034 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2036 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2037 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2041 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2042 F2FS_OPTION(sbi).s_jquota_fmt, type);
2045 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2050 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2051 err = f2fs_enable_quotas(sbi->sb);
2053 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2059 for (i = 0; i < MAXQUOTAS; i++) {
2060 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2061 err = f2fs_quota_on_mount(sbi, i);
2066 f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2073 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2076 struct inode *qf_inode;
2077 unsigned long qf_inum;
2080 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2082 qf_inum = f2fs_qf_ino(sb, type);
2086 qf_inode = f2fs_iget(sb, qf_inum);
2087 if (IS_ERR(qf_inode)) {
2088 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2089 return PTR_ERR(qf_inode);
2092 /* Don't account quota for quota files to avoid recursion */
2093 qf_inode->i_flags |= S_NOQUOTA;
2094 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2099 static int f2fs_enable_quotas(struct super_block *sb)
2101 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2103 unsigned long qf_inum;
2104 bool quota_mopt[MAXQUOTAS] = {
2105 test_opt(sbi, USRQUOTA),
2106 test_opt(sbi, GRPQUOTA),
2107 test_opt(sbi, PRJQUOTA),
2110 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2111 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2115 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2117 for (type = 0; type < MAXQUOTAS; type++) {
2118 qf_inum = f2fs_qf_ino(sb, type);
2120 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2121 DQUOT_USAGE_ENABLED |
2122 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2124 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2126 for (type--; type >= 0; type--)
2127 dquot_quota_off(sb, type);
2128 set_sbi_flag(F2FS_SB(sb),
2129 SBI_QUOTA_NEED_REPAIR);
2137 int f2fs_quota_sync(struct super_block *sb, int type)
2139 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2140 struct quota_info *dqopt = sb_dqopt(sb);
2147 * down_read(quota_sem)
2148 * dquot_writeback_dquots()
2151 * down_read(quota_sem)
2155 down_read(&sbi->quota_sem);
2156 ret = dquot_writeback_dquots(sb, type);
2161 * Now when everything is written we can discard the pagecache so
2162 * that userspace sees the changes.
2164 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2165 struct address_space *mapping;
2167 if (type != -1 && cnt != type)
2169 if (!sb_has_quota_active(sb, cnt))
2172 mapping = dqopt->files[cnt]->i_mapping;
2174 ret = filemap_fdatawrite(mapping);
2178 /* if we are using journalled quota */
2179 if (is_journalled_quota(sbi))
2182 ret = filemap_fdatawait(mapping);
2184 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2186 inode_lock(dqopt->files[cnt]);
2187 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
2188 inode_unlock(dqopt->files[cnt]);
2192 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2193 up_read(&sbi->quota_sem);
2194 f2fs_unlock_op(sbi);
2198 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2199 const struct path *path)
2201 struct inode *inode;
2204 /* if quota sysfile exists, deny enabling quota with specific file */
2205 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2206 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2210 err = f2fs_quota_sync(sb, type);
2214 err = dquot_quota_on(sb, type, format_id, path);
2218 inode = d_inode(path->dentry);
2221 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2222 f2fs_set_inode_flags(inode);
2223 inode_unlock(inode);
2224 f2fs_mark_inode_dirty_sync(inode, false);
2229 static int __f2fs_quota_off(struct super_block *sb, int type)
2231 struct inode *inode = sb_dqopt(sb)->files[type];
2234 if (!inode || !igrab(inode))
2235 return dquot_quota_off(sb, type);
2237 err = f2fs_quota_sync(sb, type);
2241 err = dquot_quota_off(sb, type);
2242 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2246 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2247 f2fs_set_inode_flags(inode);
2248 inode_unlock(inode);
2249 f2fs_mark_inode_dirty_sync(inode, false);
2255 static int f2fs_quota_off(struct super_block *sb, int type)
2257 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2260 err = __f2fs_quota_off(sb, type);
2263 * quotactl can shutdown journalled quota, result in inconsistence
2264 * between quota record and fs data by following updates, tag the
2265 * flag to let fsck be aware of it.
2267 if (is_journalled_quota(sbi))
2268 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2272 void f2fs_quota_off_umount(struct super_block *sb)
2277 for (type = 0; type < MAXQUOTAS; type++) {
2278 err = __f2fs_quota_off(sb, type);
2280 int ret = dquot_quota_off(sb, type);
2282 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2284 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2288 * In case of checkpoint=disable, we must flush quota blocks.
2289 * This can cause NULL exception for node_inode in end_io, since
2290 * put_super already dropped it.
2292 sync_filesystem(sb);
2295 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2297 struct quota_info *dqopt = sb_dqopt(sb);
2300 for (type = 0; type < MAXQUOTAS; type++) {
2301 if (!dqopt->files[type])
2303 f2fs_inode_synced(dqopt->files[type]);
2307 static int f2fs_dquot_commit(struct dquot *dquot)
2309 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2312 down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
2313 ret = dquot_commit(dquot);
2315 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2316 up_read(&sbi->quota_sem);
2320 static int f2fs_dquot_acquire(struct dquot *dquot)
2322 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2325 down_read(&sbi->quota_sem);
2326 ret = dquot_acquire(dquot);
2328 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2329 up_read(&sbi->quota_sem);
2333 static int f2fs_dquot_release(struct dquot *dquot)
2335 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2336 int ret = dquot_release(dquot);
2339 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2343 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2345 struct super_block *sb = dquot->dq_sb;
2346 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2347 int ret = dquot_mark_dquot_dirty(dquot);
2349 /* if we are using journalled quota */
2350 if (is_journalled_quota(sbi))
2351 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2356 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2358 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2359 int ret = dquot_commit_info(sb, type);
2362 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2366 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2368 *projid = F2FS_I(inode)->i_projid;
2372 static const struct dquot_operations f2fs_quota_operations = {
2373 .get_reserved_space = f2fs_get_reserved_space,
2374 .write_dquot = f2fs_dquot_commit,
2375 .acquire_dquot = f2fs_dquot_acquire,
2376 .release_dquot = f2fs_dquot_release,
2377 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
2378 .write_info = f2fs_dquot_commit_info,
2379 .alloc_dquot = dquot_alloc,
2380 .destroy_dquot = dquot_destroy,
2381 .get_projid = f2fs_get_projid,
2382 .get_next_id = dquot_get_next_id,
2385 static const struct quotactl_ops f2fs_quotactl_ops = {
2386 .quota_on = f2fs_quota_on,
2387 .quota_off = f2fs_quota_off,
2388 .quota_sync = f2fs_quota_sync,
2389 .get_state = dquot_get_state,
2390 .set_info = dquot_set_dqinfo,
2391 .get_dqblk = dquot_get_dqblk,
2392 .set_dqblk = dquot_set_dqblk,
2393 .get_nextdqblk = dquot_get_next_dqblk,
2396 int f2fs_quota_sync(struct super_block *sb, int type)
2401 void f2fs_quota_off_umount(struct super_block *sb)
2406 static const struct super_operations f2fs_sops = {
2407 .alloc_inode = f2fs_alloc_inode,
2408 .free_inode = f2fs_free_inode,
2409 .drop_inode = f2fs_drop_inode,
2410 .write_inode = f2fs_write_inode,
2411 .dirty_inode = f2fs_dirty_inode,
2412 .show_options = f2fs_show_options,
2414 .quota_read = f2fs_quota_read,
2415 .quota_write = f2fs_quota_write,
2416 .get_dquots = f2fs_get_dquots,
2418 .evict_inode = f2fs_evict_inode,
2419 .put_super = f2fs_put_super,
2420 .sync_fs = f2fs_sync_fs,
2421 .freeze_fs = f2fs_freeze,
2422 .unfreeze_fs = f2fs_unfreeze,
2423 .statfs = f2fs_statfs,
2424 .remount_fs = f2fs_remount,
2427 #ifdef CONFIG_FS_ENCRYPTION
2428 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2430 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2431 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2435 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2438 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2441 * Encrypting the root directory is not allowed because fsck
2442 * expects lost+found directory to exist and remain unencrypted
2443 * if LOST_FOUND feature is enabled.
2446 if (f2fs_sb_has_lost_found(sbi) &&
2447 inode->i_ino == F2FS_ROOT_INO(sbi))
2450 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2451 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2452 ctx, len, fs_data, XATTR_CREATE);
2455 static const union fscrypt_context *
2456 f2fs_get_dummy_context(struct super_block *sb)
2458 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_ctx.ctx;
2461 static bool f2fs_has_stable_inodes(struct super_block *sb)
2466 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
2467 int *ino_bits_ret, int *lblk_bits_ret)
2469 *ino_bits_ret = 8 * sizeof(nid_t);
2470 *lblk_bits_ret = 8 * sizeof(block_t);
2473 static const struct fscrypt_operations f2fs_cryptops = {
2474 .key_prefix = "f2fs:",
2475 .get_context = f2fs_get_context,
2476 .set_context = f2fs_set_context,
2477 .get_dummy_context = f2fs_get_dummy_context,
2478 .empty_dir = f2fs_empty_dir,
2479 .max_namelen = F2FS_NAME_LEN,
2480 .has_stable_inodes = f2fs_has_stable_inodes,
2481 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits,
2485 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2486 u64 ino, u32 generation)
2488 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2489 struct inode *inode;
2491 if (f2fs_check_nid_range(sbi, ino))
2492 return ERR_PTR(-ESTALE);
2495 * f2fs_iget isn't quite right if the inode is currently unallocated!
2496 * However f2fs_iget currently does appropriate checks to handle stale
2497 * inodes so everything is OK.
2499 inode = f2fs_iget(sb, ino);
2501 return ERR_CAST(inode);
2502 if (unlikely(generation && inode->i_generation != generation)) {
2503 /* we didn't find the right inode.. */
2505 return ERR_PTR(-ESTALE);
2510 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2511 int fh_len, int fh_type)
2513 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2514 f2fs_nfs_get_inode);
2517 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2518 int fh_len, int fh_type)
2520 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2521 f2fs_nfs_get_inode);
2524 static const struct export_operations f2fs_export_ops = {
2525 .fh_to_dentry = f2fs_fh_to_dentry,
2526 .fh_to_parent = f2fs_fh_to_parent,
2527 .get_parent = f2fs_get_parent,
2530 static loff_t max_file_blocks(void)
2533 loff_t leaf_count = DEF_ADDRS_PER_BLOCK;
2536 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2537 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2538 * space in inode.i_addr, it will be more safe to reassign
2542 /* two direct node blocks */
2543 result += (leaf_count * 2);
2545 /* two indirect node blocks */
2546 leaf_count *= NIDS_PER_BLOCK;
2547 result += (leaf_count * 2);
2549 /* one double indirect node block */
2550 leaf_count *= NIDS_PER_BLOCK;
2551 result += leaf_count;
2556 static int __f2fs_commit_super(struct buffer_head *bh,
2557 struct f2fs_super_block *super)
2561 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2562 set_buffer_dirty(bh);
2565 /* it's rare case, we can do fua all the time */
2566 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2569 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2570 struct buffer_head *bh)
2572 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2573 (bh->b_data + F2FS_SUPER_OFFSET);
2574 struct super_block *sb = sbi->sb;
2575 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2576 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2577 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2578 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2579 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2580 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2581 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2582 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2583 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2584 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2585 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2586 u32 segment_count = le32_to_cpu(raw_super->segment_count);
2587 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2588 u64 main_end_blkaddr = main_blkaddr +
2589 (segment_count_main << log_blocks_per_seg);
2590 u64 seg_end_blkaddr = segment0_blkaddr +
2591 (segment_count << log_blocks_per_seg);
2593 if (segment0_blkaddr != cp_blkaddr) {
2594 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2595 segment0_blkaddr, cp_blkaddr);
2599 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2601 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2602 cp_blkaddr, sit_blkaddr,
2603 segment_count_ckpt << log_blocks_per_seg);
2607 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2609 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2610 sit_blkaddr, nat_blkaddr,
2611 segment_count_sit << log_blocks_per_seg);
2615 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2617 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2618 nat_blkaddr, ssa_blkaddr,
2619 segment_count_nat << log_blocks_per_seg);
2623 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2625 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2626 ssa_blkaddr, main_blkaddr,
2627 segment_count_ssa << log_blocks_per_seg);
2631 if (main_end_blkaddr > seg_end_blkaddr) {
2632 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2635 (segment_count << log_blocks_per_seg),
2636 segment_count_main << log_blocks_per_seg);
2638 } else if (main_end_blkaddr < seg_end_blkaddr) {
2642 /* fix in-memory information all the time */
2643 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2644 segment0_blkaddr) >> log_blocks_per_seg);
2646 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2647 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2650 err = __f2fs_commit_super(bh, NULL);
2651 res = err ? "failed" : "done";
2653 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%u) block(%u)",
2656 (segment_count << log_blocks_per_seg),
2657 segment_count_main << log_blocks_per_seg);
2664 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2665 struct buffer_head *bh)
2667 block_t segment_count, segs_per_sec, secs_per_zone;
2668 block_t total_sections, blocks_per_seg;
2669 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2670 (bh->b_data + F2FS_SUPER_OFFSET);
2671 unsigned int blocksize;
2672 size_t crc_offset = 0;
2675 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
2676 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
2677 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2681 /* Check checksum_offset and crc in superblock */
2682 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2683 crc_offset = le32_to_cpu(raw_super->checksum_offset);
2685 offsetof(struct f2fs_super_block, crc)) {
2686 f2fs_info(sbi, "Invalid SB checksum offset: %zu",
2688 return -EFSCORRUPTED;
2690 crc = le32_to_cpu(raw_super->crc);
2691 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2692 f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
2693 return -EFSCORRUPTED;
2697 /* Currently, support only 4KB page cache size */
2698 if (F2FS_BLKSIZE != PAGE_SIZE) {
2699 f2fs_info(sbi, "Invalid page_cache_size (%lu), supports only 4KB",
2701 return -EFSCORRUPTED;
2704 /* Currently, support only 4KB block size */
2705 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2706 if (blocksize != F2FS_BLKSIZE) {
2707 f2fs_info(sbi, "Invalid blocksize (%u), supports only 4KB",
2709 return -EFSCORRUPTED;
2712 /* check log blocks per segment */
2713 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2714 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
2715 le32_to_cpu(raw_super->log_blocks_per_seg));
2716 return -EFSCORRUPTED;
2719 /* Currently, support 512/1024/2048/4096 bytes sector size */
2720 if (le32_to_cpu(raw_super->log_sectorsize) >
2721 F2FS_MAX_LOG_SECTOR_SIZE ||
2722 le32_to_cpu(raw_super->log_sectorsize) <
2723 F2FS_MIN_LOG_SECTOR_SIZE) {
2724 f2fs_info(sbi, "Invalid log sectorsize (%u)",
2725 le32_to_cpu(raw_super->log_sectorsize));
2726 return -EFSCORRUPTED;
2728 if (le32_to_cpu(raw_super->log_sectors_per_block) +
2729 le32_to_cpu(raw_super->log_sectorsize) !=
2730 F2FS_MAX_LOG_SECTOR_SIZE) {
2731 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
2732 le32_to_cpu(raw_super->log_sectors_per_block),
2733 le32_to_cpu(raw_super->log_sectorsize));
2734 return -EFSCORRUPTED;
2737 segment_count = le32_to_cpu(raw_super->segment_count);
2738 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2739 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2740 total_sections = le32_to_cpu(raw_super->section_count);
2742 /* blocks_per_seg should be 512, given the above check */
2743 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2745 if (segment_count > F2FS_MAX_SEGMENT ||
2746 segment_count < F2FS_MIN_SEGMENTS) {
2747 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
2748 return -EFSCORRUPTED;
2751 if (total_sections > segment_count ||
2752 total_sections < F2FS_MIN_SEGMENTS ||
2753 segs_per_sec > segment_count || !segs_per_sec) {
2754 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
2755 segment_count, total_sections, segs_per_sec);
2756 return -EFSCORRUPTED;
2759 if ((segment_count / segs_per_sec) < total_sections) {
2760 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
2761 segment_count, segs_per_sec, total_sections);
2762 return -EFSCORRUPTED;
2765 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2766 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
2767 segment_count, le64_to_cpu(raw_super->block_count));
2768 return -EFSCORRUPTED;
2771 if (RDEV(0).path[0]) {
2772 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
2775 while (i < MAX_DEVICES && RDEV(i).path[0]) {
2776 dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
2779 if (segment_count != dev_seg_count) {
2780 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
2781 segment_count, dev_seg_count);
2782 return -EFSCORRUPTED;
2786 if (secs_per_zone > total_sections || !secs_per_zone) {
2787 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
2788 secs_per_zone, total_sections);
2789 return -EFSCORRUPTED;
2791 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2792 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2793 (le32_to_cpu(raw_super->extension_count) +
2794 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2795 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
2796 le32_to_cpu(raw_super->extension_count),
2797 raw_super->hot_ext_count,
2798 F2FS_MAX_EXTENSION);
2799 return -EFSCORRUPTED;
2802 if (le32_to_cpu(raw_super->cp_payload) >
2803 (blocks_per_seg - F2FS_CP_PACKS)) {
2804 f2fs_info(sbi, "Insane cp_payload (%u > %u)",
2805 le32_to_cpu(raw_super->cp_payload),
2806 blocks_per_seg - F2FS_CP_PACKS);
2807 return -EFSCORRUPTED;
2810 /* check reserved ino info */
2811 if (le32_to_cpu(raw_super->node_ino) != 1 ||
2812 le32_to_cpu(raw_super->meta_ino) != 2 ||
2813 le32_to_cpu(raw_super->root_ino) != 3) {
2814 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2815 le32_to_cpu(raw_super->node_ino),
2816 le32_to_cpu(raw_super->meta_ino),
2817 le32_to_cpu(raw_super->root_ino));
2818 return -EFSCORRUPTED;
2821 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2822 if (sanity_check_area_boundary(sbi, bh))
2823 return -EFSCORRUPTED;
2828 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2830 unsigned int total, fsmeta;
2831 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2832 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2833 unsigned int ovp_segments, reserved_segments;
2834 unsigned int main_segs, blocks_per_seg;
2835 unsigned int sit_segs, nat_segs;
2836 unsigned int sit_bitmap_size, nat_bitmap_size;
2837 unsigned int log_blocks_per_seg;
2838 unsigned int segment_count_main;
2839 unsigned int cp_pack_start_sum, cp_payload;
2840 block_t user_block_count, valid_user_blocks;
2841 block_t avail_node_count, valid_node_count;
2844 total = le32_to_cpu(raw_super->segment_count);
2845 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2846 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2848 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2850 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2851 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2853 if (unlikely(fsmeta >= total))
2856 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2857 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2859 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2860 ovp_segments == 0 || reserved_segments == 0)) {
2861 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
2865 user_block_count = le64_to_cpu(ckpt->user_block_count);
2866 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2867 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2868 if (!user_block_count || user_block_count >=
2869 segment_count_main << log_blocks_per_seg) {
2870 f2fs_err(sbi, "Wrong user_block_count: %u",
2875 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
2876 if (valid_user_blocks > user_block_count) {
2877 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
2878 valid_user_blocks, user_block_count);
2882 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
2883 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
2884 if (valid_node_count > avail_node_count) {
2885 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
2886 valid_node_count, avail_node_count);
2890 main_segs = le32_to_cpu(raw_super->segment_count_main);
2891 blocks_per_seg = sbi->blocks_per_seg;
2893 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2894 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2895 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2897 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2898 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2899 le32_to_cpu(ckpt->cur_node_segno[j])) {
2900 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
2902 le32_to_cpu(ckpt->cur_node_segno[i]));
2907 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2908 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2909 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2911 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2912 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2913 le32_to_cpu(ckpt->cur_data_segno[j])) {
2914 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
2916 le32_to_cpu(ckpt->cur_data_segno[i]));
2921 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2922 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
2923 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2924 le32_to_cpu(ckpt->cur_data_segno[j])) {
2925 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
2927 le32_to_cpu(ckpt->cur_node_segno[i]));
2933 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2934 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2936 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2937 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2938 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
2939 sit_bitmap_size, nat_bitmap_size);
2943 cp_pack_start_sum = __start_sum_addr(sbi);
2944 cp_payload = __cp_payload(sbi);
2945 if (cp_pack_start_sum < cp_payload + 1 ||
2946 cp_pack_start_sum > blocks_per_seg - 1 -
2948 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
2953 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
2954 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
2955 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
2956 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
2957 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
2958 le32_to_cpu(ckpt->checksum_offset));
2962 if (unlikely(f2fs_cp_error(sbi))) {
2963 f2fs_err(sbi, "A bug case: need to run fsck");
2969 static void init_sb_info(struct f2fs_sb_info *sbi)
2971 struct f2fs_super_block *raw_super = sbi->raw_super;
2974 sbi->log_sectors_per_block =
2975 le32_to_cpu(raw_super->log_sectors_per_block);
2976 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2977 sbi->blocksize = 1 << sbi->log_blocksize;
2978 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2979 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2980 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2981 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2982 sbi->total_sections = le32_to_cpu(raw_super->section_count);
2983 sbi->total_node_count =
2984 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2985 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2986 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2987 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2988 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2989 sbi->cur_victim_sec = NULL_SECNO;
2990 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
2991 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
2992 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2993 sbi->migration_granularity = sbi->segs_per_sec;
2995 sbi->dir_level = DEF_DIR_LEVEL;
2996 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2997 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2998 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
2999 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3000 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3001 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3002 DEF_UMOUNT_DISCARD_TIMEOUT;
3003 clear_sbi_flag(sbi, SBI_NEED_FSCK);
3005 for (i = 0; i < NR_COUNT_TYPE; i++)
3006 atomic_set(&sbi->nr_pages[i], 0);
3008 for (i = 0; i < META; i++)
3009 atomic_set(&sbi->wb_sync_req[i], 0);
3011 INIT_LIST_HEAD(&sbi->s_list);
3012 mutex_init(&sbi->umount_mutex);
3013 init_rwsem(&sbi->io_order_lock);
3014 spin_lock_init(&sbi->cp_lock);
3016 sbi->dirty_device = 0;
3017 spin_lock_init(&sbi->dev_lock);
3019 init_rwsem(&sbi->sb_lock);
3020 init_rwsem(&sbi->pin_sem);
3023 static int init_percpu_info(struct f2fs_sb_info *sbi)
3027 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3031 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3034 percpu_counter_destroy(&sbi->alloc_valid_block_count);
3039 #ifdef CONFIG_BLK_DEV_ZONED
3040 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3043 struct f2fs_dev_info *dev = data;
3045 if (zone->type != BLK_ZONE_TYPE_CONVENTIONAL)
3046 set_bit(idx, dev->blkz_seq);
3050 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3052 struct block_device *bdev = FDEV(devi).bdev;
3053 sector_t nr_sectors = bdev->bd_part->nr_sects;
3056 if (!f2fs_sb_has_blkzoned(sbi))
3059 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3060 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
3062 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
3063 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
3064 __ilog2_u32(sbi->blocks_per_blkz))
3066 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
3067 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
3068 sbi->log_blocks_per_blkz;
3069 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
3070 FDEV(devi).nr_blkz++;
3072 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3073 BITS_TO_LONGS(FDEV(devi).nr_blkz)
3074 * sizeof(unsigned long),
3076 if (!FDEV(devi).blkz_seq)
3079 /* Get block zones type */
3080 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3090 * Read f2fs raw super block.
3091 * Because we have two copies of super block, so read both of them
3092 * to get the first valid one. If any one of them is broken, we pass
3093 * them recovery flag back to the caller.
3095 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3096 struct f2fs_super_block **raw_super,
3097 int *valid_super_block, int *recovery)
3099 struct super_block *sb = sbi->sb;
3101 struct buffer_head *bh;
3102 struct f2fs_super_block *super;
3105 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3109 for (block = 0; block < 2; block++) {
3110 bh = sb_bread(sb, block);
3112 f2fs_err(sbi, "Unable to read %dth superblock",
3119 /* sanity checking of raw super */
3120 err = sanity_check_raw_super(sbi, bh);
3122 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3130 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3132 *valid_super_block = block;
3138 /* No valid superblock */
3147 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3149 struct buffer_head *bh;
3153 if ((recover && f2fs_readonly(sbi->sb)) ||
3154 bdev_read_only(sbi->sb->s_bdev)) {
3155 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3159 /* we should update superblock crc here */
3160 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3161 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3162 offsetof(struct f2fs_super_block, crc));
3163 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3166 /* write back-up superblock first */
3167 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3170 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3173 /* if we are in recovery path, skip writing valid superblock */
3177 /* write current valid superblock */
3178 bh = sb_bread(sbi->sb, sbi->valid_super_block);
3181 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3186 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3188 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3189 unsigned int max_devices = MAX_DEVICES;
3192 /* Initialize single device information */
3193 if (!RDEV(0).path[0]) {
3194 if (!bdev_is_zoned(sbi->sb->s_bdev))
3200 * Initialize multiple devices information, or single
3201 * zoned block device information.
3203 sbi->devs = f2fs_kzalloc(sbi,
3204 array_size(max_devices,
3205 sizeof(struct f2fs_dev_info)),
3210 for (i = 0; i < max_devices; i++) {
3212 if (i > 0 && !RDEV(i).path[0])
3215 if (max_devices == 1) {
3216 /* Single zoned block device mount */
3218 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3219 sbi->sb->s_mode, sbi->sb->s_type);
3221 /* Multi-device mount */
3222 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3223 FDEV(i).total_segments =
3224 le32_to_cpu(RDEV(i).total_segments);
3226 FDEV(i).start_blk = 0;
3227 FDEV(i).end_blk = FDEV(i).start_blk +
3228 (FDEV(i).total_segments <<
3229 sbi->log_blocks_per_seg) - 1 +
3230 le32_to_cpu(raw_super->segment0_blkaddr);
3232 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3233 FDEV(i).end_blk = FDEV(i).start_blk +
3234 (FDEV(i).total_segments <<
3235 sbi->log_blocks_per_seg) - 1;
3237 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3238 sbi->sb->s_mode, sbi->sb->s_type);
3240 if (IS_ERR(FDEV(i).bdev))
3241 return PTR_ERR(FDEV(i).bdev);
3243 /* to release errored devices */
3244 sbi->s_ndevs = i + 1;
3246 #ifdef CONFIG_BLK_DEV_ZONED
3247 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3248 !f2fs_sb_has_blkzoned(sbi)) {
3249 f2fs_err(sbi, "Zoned block device feature not enabled\n");
3252 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3253 if (init_blkz_info(sbi, i)) {
3254 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3257 if (max_devices == 1)
3259 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3261 FDEV(i).total_segments,
3262 FDEV(i).start_blk, FDEV(i).end_blk,
3263 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3264 "Host-aware" : "Host-managed");
3268 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3270 FDEV(i).total_segments,
3271 FDEV(i).start_blk, FDEV(i).end_blk);
3274 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3278 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
3280 #ifdef CONFIG_UNICODE
3281 if (f2fs_sb_has_casefold(sbi) && !sbi->s_encoding) {
3282 const struct f2fs_sb_encodings *encoding_info;
3283 struct unicode_map *encoding;
3284 __u16 encoding_flags;
3286 if (f2fs_sb_has_encrypt(sbi)) {
3288 "Can't mount with encoding and encryption");
3292 if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info,
3295 "Encoding requested by superblock is unknown");
3299 encoding = utf8_load(encoding_info->version);
3300 if (IS_ERR(encoding)) {
3302 "can't mount with superblock charset: %s-%s "
3303 "not supported by the kernel. flags: 0x%x.",
3304 encoding_info->name, encoding_info->version,
3306 return PTR_ERR(encoding);
3308 f2fs_info(sbi, "Using encoding defined by superblock: "
3309 "%s-%s with flags 0x%hx", encoding_info->name,
3310 encoding_info->version?:"\b", encoding_flags);
3312 sbi->s_encoding = encoding;
3313 sbi->s_encoding_flags = encoding_flags;
3314 sbi->sb->s_d_op = &f2fs_dentry_ops;
3317 if (f2fs_sb_has_casefold(sbi)) {
3318 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
3325 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3327 struct f2fs_sm_info *sm_i = SM_I(sbi);
3329 /* adjust parameters according to the volume size */
3330 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3331 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3332 sm_i->dcc_info->discard_granularity = 1;
3333 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3336 sbi->readdir_ra = 1;
3339 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3341 struct f2fs_sb_info *sbi;
3342 struct f2fs_super_block *raw_super;
3345 bool skip_recovery = false, need_fsck = false;
3346 char *options = NULL;
3347 int recovery, i, valid_super_block;
3348 struct curseg_info *seg_i;
3354 valid_super_block = -1;
3357 /* allocate memory for f2fs-specific super block info */
3358 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3364 /* Load the checksum driver */
3365 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3366 if (IS_ERR(sbi->s_chksum_driver)) {
3367 f2fs_err(sbi, "Cannot load crc32 driver.");
3368 err = PTR_ERR(sbi->s_chksum_driver);
3369 sbi->s_chksum_driver = NULL;
3373 /* set a block size */
3374 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3375 f2fs_err(sbi, "unable to set blocksize");
3379 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3384 sb->s_fs_info = sbi;
3385 sbi->raw_super = raw_super;
3387 /* precompute checksum seed for metadata */
3388 if (f2fs_sb_has_inode_chksum(sbi))
3389 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3390 sizeof(raw_super->uuid));
3393 * The BLKZONED feature indicates that the drive was formatted with
3394 * zone alignment optimization. This is optional for host-aware
3395 * devices, but mandatory for host-managed zoned block devices.
3397 #ifndef CONFIG_BLK_DEV_ZONED
3398 if (f2fs_sb_has_blkzoned(sbi)) {
3399 f2fs_err(sbi, "Zoned block device support is not enabled");
3404 default_options(sbi);
3405 /* parse mount options */
3406 options = kstrdup((const char *)data, GFP_KERNEL);
3407 if (data && !options) {
3412 err = parse_options(sb, options, false);
3416 sbi->max_file_blocks = max_file_blocks();
3417 sb->s_maxbytes = sbi->max_file_blocks <<
3418 le32_to_cpu(raw_super->log_blocksize);
3419 sb->s_max_links = F2FS_LINK_MAX;
3421 err = f2fs_setup_casefold(sbi);
3426 sb->dq_op = &f2fs_quota_operations;
3427 sb->s_qcop = &f2fs_quotactl_ops;
3428 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3430 if (f2fs_sb_has_quota_ino(sbi)) {
3431 for (i = 0; i < MAXQUOTAS; i++) {
3432 if (f2fs_qf_ino(sbi->sb, i))
3433 sbi->nquota_files++;
3438 sb->s_op = &f2fs_sops;
3439 #ifdef CONFIG_FS_ENCRYPTION
3440 sb->s_cop = &f2fs_cryptops;
3442 #ifdef CONFIG_FS_VERITY
3443 sb->s_vop = &f2fs_verityops;
3445 sb->s_xattr = f2fs_xattr_handlers;
3446 sb->s_export_op = &f2fs_export_ops;
3447 sb->s_magic = F2FS_SUPER_MAGIC;
3448 sb->s_time_gran = 1;
3449 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3450 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3451 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3452 sb->s_iflags |= SB_I_CGROUPWB;
3454 /* init f2fs-specific super block info */
3455 sbi->valid_super_block = valid_super_block;
3456 init_rwsem(&sbi->gc_lock);
3457 mutex_init(&sbi->writepages);
3458 mutex_init(&sbi->cp_mutex);
3459 init_rwsem(&sbi->node_write);
3460 init_rwsem(&sbi->node_change);
3462 /* disallow all the data/node/meta page writes */
3463 set_sbi_flag(sbi, SBI_POR_DOING);
3464 spin_lock_init(&sbi->stat_lock);
3466 /* init iostat info */
3467 spin_lock_init(&sbi->iostat_lock);
3468 sbi->iostat_enable = false;
3469 sbi->iostat_period_ms = DEFAULT_IOSTAT_PERIOD_MS;
3471 for (i = 0; i < NR_PAGE_TYPE; i++) {
3472 int n = (i == META) ? 1: NR_TEMP_TYPE;
3478 sizeof(struct f2fs_bio_info)),
3480 if (!sbi->write_io[i]) {
3485 for (j = HOT; j < n; j++) {
3486 init_rwsem(&sbi->write_io[i][j].io_rwsem);
3487 sbi->write_io[i][j].sbi = sbi;
3488 sbi->write_io[i][j].bio = NULL;
3489 spin_lock_init(&sbi->write_io[i][j].io_lock);
3490 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3491 INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list);
3492 init_rwsem(&sbi->write_io[i][j].bio_list_lock);
3496 init_rwsem(&sbi->cp_rwsem);
3497 init_rwsem(&sbi->quota_sem);
3498 init_waitqueue_head(&sbi->cp_wait);
3501 err = init_percpu_info(sbi);
3505 if (F2FS_IO_ALIGNED(sbi)) {
3506 sbi->write_io_dummy =
3507 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3508 if (!sbi->write_io_dummy) {
3514 /* init per sbi slab cache */
3515 err = f2fs_init_xattr_caches(sbi);
3519 /* get an inode for meta space */
3520 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3521 if (IS_ERR(sbi->meta_inode)) {
3522 f2fs_err(sbi, "Failed to read F2FS meta data inode");
3523 err = PTR_ERR(sbi->meta_inode);
3524 goto free_xattr_cache;
3527 err = f2fs_get_valid_checkpoint(sbi);
3529 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
3530 goto free_meta_inode;
3533 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3534 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3535 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3536 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3537 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3540 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
3541 set_sbi_flag(sbi, SBI_NEED_FSCK);
3543 /* Initialize device list */
3544 err = f2fs_scan_devices(sbi);
3546 f2fs_err(sbi, "Failed to find devices");
3550 err = f2fs_init_post_read_wq(sbi);
3552 f2fs_err(sbi, "Failed to initialize post read workqueue");
3556 sbi->total_valid_node_count =
3557 le32_to_cpu(sbi->ckpt->valid_node_count);
3558 percpu_counter_set(&sbi->total_valid_inode_count,
3559 le32_to_cpu(sbi->ckpt->valid_inode_count));
3560 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3561 sbi->total_valid_block_count =
3562 le64_to_cpu(sbi->ckpt->valid_block_count);
3563 sbi->last_valid_block_count = sbi->total_valid_block_count;
3564 sbi->reserved_blocks = 0;
3565 sbi->current_reserved_blocks = 0;
3566 limit_reserve_root(sbi);
3567 adjust_unusable_cap_perc(sbi);
3569 for (i = 0; i < NR_INODE_TYPE; i++) {
3570 INIT_LIST_HEAD(&sbi->inode_list[i]);
3571 spin_lock_init(&sbi->inode_lock[i]);
3573 mutex_init(&sbi->flush_lock);
3575 f2fs_init_extent_cache_info(sbi);
3577 f2fs_init_ino_entry_info(sbi);
3579 f2fs_init_fsync_node_info(sbi);
3581 /* setup f2fs internal modules */
3582 err = f2fs_build_segment_manager(sbi);
3584 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
3588 err = f2fs_build_node_manager(sbi);
3590 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
3595 /* For write statistics */
3596 if (sb->s_bdev->bd_part)
3597 sbi->sectors_written_start =
3598 (u64)part_stat_read(sb->s_bdev->bd_part,
3599 sectors[STAT_WRITE]);
3601 /* Read accumulated write IO statistics if exists */
3602 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3603 if (__exist_node_summaries(sbi))
3604 sbi->kbytes_written =
3605 le64_to_cpu(seg_i->journal->info.kbytes_written);
3607 f2fs_build_gc_manager(sbi);
3609 err = f2fs_build_stats(sbi);
3613 /* get an inode for node space */
3614 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3615 if (IS_ERR(sbi->node_inode)) {
3616 f2fs_err(sbi, "Failed to read node inode");
3617 err = PTR_ERR(sbi->node_inode);
3621 /* read root inode and dentry */
3622 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3624 f2fs_err(sbi, "Failed to read root inode");
3625 err = PTR_ERR(root);
3626 goto free_node_inode;
3628 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3629 !root->i_size || !root->i_nlink) {
3632 goto free_node_inode;
3635 sb->s_root = d_make_root(root); /* allocate root dentry */
3638 goto free_node_inode;
3641 err = f2fs_register_sysfs(sbi);
3643 goto free_root_inode;
3646 /* Enable quota usage during mount */
3647 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3648 err = f2fs_enable_quotas(sb);
3650 f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
3653 /* if there are any orphan inodes, free them */
3654 err = f2fs_recover_orphan_inodes(sbi);
3658 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3659 goto reset_checkpoint;
3661 /* recover fsynced data */
3662 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
3663 !test_opt(sbi, NORECOVERY)) {
3665 * mount should be failed, when device has readonly mode, and
3666 * previous checkpoint was not done by clean system shutdown.
3668 if (f2fs_hw_is_readonly(sbi)) {
3669 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3671 f2fs_err(sbi, "Need to recover fsync data, but write access unavailable");
3674 f2fs_info(sbi, "write access unavailable, skipping recovery");
3675 goto reset_checkpoint;
3679 set_sbi_flag(sbi, SBI_NEED_FSCK);
3682 goto reset_checkpoint;
3684 err = f2fs_recover_fsync_data(sbi, false);
3687 skip_recovery = true;
3689 f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
3694 err = f2fs_recover_fsync_data(sbi, true);
3696 if (!f2fs_readonly(sb) && err > 0) {
3698 f2fs_err(sbi, "Need to recover fsync data");
3704 * If the f2fs is not readonly and fsync data recovery succeeds,
3705 * check zoned block devices' write pointer consistency.
3707 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
3708 err = f2fs_check_write_pointer(sbi);
3714 /* f2fs_recover_fsync_data() cleared this already */
3715 clear_sbi_flag(sbi, SBI_POR_DOING);
3717 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3718 err = f2fs_disable_checkpoint(sbi);
3720 goto sync_free_meta;
3721 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3722 f2fs_enable_checkpoint(sbi);
3726 * If filesystem is not mounted as read-only then
3727 * do start the gc_thread.
3729 if (F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF && !f2fs_readonly(sb)) {
3730 /* After POR, we can run background GC thread.*/
3731 err = f2fs_start_gc_thread(sbi);
3733 goto sync_free_meta;
3737 /* recover broken superblock */
3739 err = f2fs_commit_super(sbi, true);
3740 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
3741 sbi->valid_super_block ? 1 : 2, err);
3744 f2fs_join_shrinker(sbi);
3746 f2fs_tuning_parameters(sbi);
3748 f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
3749 cur_cp_version(F2FS_CKPT(sbi)));
3750 f2fs_update_time(sbi, CP_TIME);
3751 f2fs_update_time(sbi, REQ_TIME);
3752 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3756 /* safe to flush all the data */
3757 sync_filesystem(sbi->sb);
3762 f2fs_truncate_quota_inode_pages(sb);
3763 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3764 f2fs_quota_off_umount(sbi->sb);
3767 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3768 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3769 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3770 * falls into an infinite loop in f2fs_sync_meta_pages().
3772 truncate_inode_pages_final(META_MAPPING(sbi));
3773 /* evict some inodes being cached by GC */
3775 f2fs_unregister_sysfs(sbi);
3780 f2fs_release_ino_entry(sbi, true);
3781 truncate_inode_pages_final(NODE_MAPPING(sbi));
3782 iput(sbi->node_inode);
3783 sbi->node_inode = NULL;
3785 f2fs_destroy_stats(sbi);
3787 f2fs_destroy_node_manager(sbi);
3789 f2fs_destroy_segment_manager(sbi);
3790 f2fs_destroy_post_read_wq(sbi);
3792 destroy_device_list(sbi);
3795 make_bad_inode(sbi->meta_inode);
3796 iput(sbi->meta_inode);
3797 sbi->meta_inode = NULL;
3799 f2fs_destroy_xattr_caches(sbi);
3801 mempool_destroy(sbi->write_io_dummy);
3803 destroy_percpu_info(sbi);
3805 for (i = 0; i < NR_PAGE_TYPE; i++)
3806 kvfree(sbi->write_io[i]);
3808 #ifdef CONFIG_UNICODE
3809 utf8_unload(sbi->s_encoding);
3813 for (i = 0; i < MAXQUOTAS; i++)
3814 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
3816 fscrypt_free_dummy_context(&F2FS_OPTION(sbi).dummy_enc_ctx);
3821 if (sbi->s_chksum_driver)
3822 crypto_free_shash(sbi->s_chksum_driver);
3825 /* give only one another chance */
3826 if (retry_cnt > 0 && skip_recovery) {
3828 shrink_dcache_sb(sb);
3834 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3835 const char *dev_name, void *data)
3837 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3840 static void kill_f2fs_super(struct super_block *sb)
3843 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3845 set_sbi_flag(sbi, SBI_IS_CLOSE);
3846 f2fs_stop_gc_thread(sbi);
3847 f2fs_stop_discard_thread(sbi);
3849 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3850 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3851 struct cp_control cpc = {
3852 .reason = CP_UMOUNT,
3854 f2fs_write_checkpoint(sbi, &cpc);
3857 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3858 sb->s_flags &= ~SB_RDONLY;
3860 kill_block_super(sb);
3863 static struct file_system_type f2fs_fs_type = {
3864 .owner = THIS_MODULE,
3866 .mount = f2fs_mount,
3867 .kill_sb = kill_f2fs_super,
3868 .fs_flags = FS_REQUIRES_DEV,
3870 MODULE_ALIAS_FS("f2fs");
3872 static int __init init_inodecache(void)
3874 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3875 sizeof(struct f2fs_inode_info), 0,
3876 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3877 if (!f2fs_inode_cachep)
3882 static void destroy_inodecache(void)
3885 * Make sure all delayed rcu free inodes are flushed before we
3889 kmem_cache_destroy(f2fs_inode_cachep);
3892 static int __init init_f2fs_fs(void)
3896 if (PAGE_SIZE != F2FS_BLKSIZE) {
3897 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3898 PAGE_SIZE, F2FS_BLKSIZE);
3902 f2fs_build_trace_ios();
3904 err = init_inodecache();
3907 err = f2fs_create_node_manager_caches();
3909 goto free_inodecache;
3910 err = f2fs_create_segment_manager_caches();
3912 goto free_node_manager_caches;
3913 err = f2fs_create_checkpoint_caches();
3915 goto free_segment_manager_caches;
3916 err = f2fs_create_extent_cache();
3918 goto free_checkpoint_caches;
3919 err = f2fs_init_sysfs();
3921 goto free_extent_cache;
3922 err = register_shrinker(&f2fs_shrinker_info);
3925 err = register_filesystem(&f2fs_fs_type);
3928 f2fs_create_root_stats();
3929 err = f2fs_init_post_read_processing();
3931 goto free_root_stats;
3932 err = f2fs_init_bio_entry_cache();
3934 goto free_post_read;
3935 err = f2fs_init_bioset();
3937 goto free_bio_enrty_cache;
3938 err = f2fs_init_compress_mempool();
3943 f2fs_destroy_bioset();
3944 free_bio_enrty_cache:
3945 f2fs_destroy_bio_entry_cache();
3947 f2fs_destroy_post_read_processing();
3949 f2fs_destroy_root_stats();
3950 unregister_filesystem(&f2fs_fs_type);
3952 unregister_shrinker(&f2fs_shrinker_info);
3956 f2fs_destroy_extent_cache();
3957 free_checkpoint_caches:
3958 f2fs_destroy_checkpoint_caches();
3959 free_segment_manager_caches:
3960 f2fs_destroy_segment_manager_caches();
3961 free_node_manager_caches:
3962 f2fs_destroy_node_manager_caches();
3964 destroy_inodecache();
3969 static void __exit exit_f2fs_fs(void)
3971 f2fs_destroy_compress_mempool();
3972 f2fs_destroy_bioset();
3973 f2fs_destroy_bio_entry_cache();
3974 f2fs_destroy_post_read_processing();
3975 f2fs_destroy_root_stats();
3976 unregister_filesystem(&f2fs_fs_type);
3977 unregister_shrinker(&f2fs_shrinker_info);
3979 f2fs_destroy_extent_cache();
3980 f2fs_destroy_checkpoint_caches();
3981 f2fs_destroy_segment_manager_caches();
3982 f2fs_destroy_node_manager_caches();
3983 destroy_inodecache();
3984 f2fs_destroy_trace_ios();
3987 module_init(init_f2fs_fs)
3988 module_exit(exit_f2fs_fs)
3990 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3991 MODULE_DESCRIPTION("Flash Friendly File System");
3992 MODULE_LICENSE("GPL");