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>
35 #define CREATE_TRACE_POINTS
36 #include <trace/events/f2fs.h>
38 static struct kmem_cache *f2fs_inode_cachep;
40 #ifdef CONFIG_F2FS_FAULT_INJECTION
42 const char *f2fs_fault_name[FAULT_MAX] = {
43 [FAULT_KMALLOC] = "kmalloc",
44 [FAULT_KVMALLOC] = "kvmalloc",
45 [FAULT_PAGE_ALLOC] = "page alloc",
46 [FAULT_PAGE_GET] = "page get",
47 [FAULT_ALLOC_BIO] = "alloc bio",
48 [FAULT_ALLOC_NID] = "alloc nid",
49 [FAULT_ORPHAN] = "orphan",
50 [FAULT_BLOCK] = "no more block",
51 [FAULT_DIR_DEPTH] = "too big dir depth",
52 [FAULT_EVICT_INODE] = "evict_inode fail",
53 [FAULT_TRUNCATE] = "truncate fail",
54 [FAULT_READ_IO] = "read IO error",
55 [FAULT_CHECKPOINT] = "checkpoint error",
56 [FAULT_DISCARD] = "discard error",
57 [FAULT_WRITE_IO] = "write IO error",
60 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
63 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
66 atomic_set(&ffi->inject_ops, 0);
67 ffi->inject_rate = rate;
71 ffi->inject_type = type;
74 memset(ffi, 0, sizeof(struct f2fs_fault_info));
78 /* f2fs-wide shrinker description */
79 static struct shrinker f2fs_shrinker_info = {
80 .scan_objects = f2fs_shrink_scan,
81 .count_objects = f2fs_shrink_count,
82 .seeks = DEFAULT_SEEKS,
87 Opt_disable_roll_forward,
98 Opt_disable_ext_identify,
101 Opt_inline_xattr_size,
139 Opt_test_dummy_encryption,
140 Opt_checkpoint_disable,
141 Opt_checkpoint_disable_cap,
142 Opt_checkpoint_disable_cap_perc,
143 Opt_checkpoint_enable,
144 Opt_compress_algorithm,
145 Opt_compress_log_size,
146 Opt_compress_extension,
150 static match_table_t f2fs_tokens = {
151 {Opt_gc_background, "background_gc=%s"},
152 {Opt_disable_roll_forward, "disable_roll_forward"},
153 {Opt_norecovery, "norecovery"},
154 {Opt_discard, "discard"},
155 {Opt_nodiscard, "nodiscard"},
156 {Opt_noheap, "no_heap"},
158 {Opt_user_xattr, "user_xattr"},
159 {Opt_nouser_xattr, "nouser_xattr"},
161 {Opt_noacl, "noacl"},
162 {Opt_active_logs, "active_logs=%u"},
163 {Opt_disable_ext_identify, "disable_ext_identify"},
164 {Opt_inline_xattr, "inline_xattr"},
165 {Opt_noinline_xattr, "noinline_xattr"},
166 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
167 {Opt_inline_data, "inline_data"},
168 {Opt_inline_dentry, "inline_dentry"},
169 {Opt_noinline_dentry, "noinline_dentry"},
170 {Opt_flush_merge, "flush_merge"},
171 {Opt_noflush_merge, "noflush_merge"},
172 {Opt_nobarrier, "nobarrier"},
173 {Opt_fastboot, "fastboot"},
174 {Opt_extent_cache, "extent_cache"},
175 {Opt_noextent_cache, "noextent_cache"},
176 {Opt_noinline_data, "noinline_data"},
177 {Opt_data_flush, "data_flush"},
178 {Opt_reserve_root, "reserve_root=%u"},
179 {Opt_resgid, "resgid=%u"},
180 {Opt_resuid, "resuid=%u"},
181 {Opt_mode, "mode=%s"},
182 {Opt_io_size_bits, "io_bits=%u"},
183 {Opt_fault_injection, "fault_injection=%u"},
184 {Opt_fault_type, "fault_type=%u"},
185 {Opt_lazytime, "lazytime"},
186 {Opt_nolazytime, "nolazytime"},
187 {Opt_quota, "quota"},
188 {Opt_noquota, "noquota"},
189 {Opt_usrquota, "usrquota"},
190 {Opt_grpquota, "grpquota"},
191 {Opt_prjquota, "prjquota"},
192 {Opt_usrjquota, "usrjquota=%s"},
193 {Opt_grpjquota, "grpjquota=%s"},
194 {Opt_prjjquota, "prjjquota=%s"},
195 {Opt_offusrjquota, "usrjquota="},
196 {Opt_offgrpjquota, "grpjquota="},
197 {Opt_offprjjquota, "prjjquota="},
198 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
199 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
200 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
201 {Opt_whint, "whint_mode=%s"},
202 {Opt_alloc, "alloc_mode=%s"},
203 {Opt_fsync, "fsync_mode=%s"},
204 {Opt_test_dummy_encryption, "test_dummy_encryption"},
205 {Opt_checkpoint_disable, "checkpoint=disable"},
206 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
207 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
208 {Opt_checkpoint_enable, "checkpoint=enable"},
209 {Opt_compress_algorithm, "compress_algorithm=%s"},
210 {Opt_compress_log_size, "compress_log_size=%u"},
211 {Opt_compress_extension, "compress_extension=%s"},
215 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
217 struct va_format vaf;
223 level = printk_get_level(fmt);
224 vaf.fmt = printk_skip_level(fmt);
226 printk("%c%cF2FS-fs (%s): %pV\n",
227 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
232 #ifdef CONFIG_UNICODE
233 static const struct f2fs_sb_encodings {
237 } f2fs_sb_encoding_map[] = {
238 {F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"},
241 static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb,
242 const struct f2fs_sb_encodings **encoding,
245 __u16 magic = le16_to_cpu(sb->s_encoding);
248 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
249 if (magic == f2fs_sb_encoding_map[i].magic)
252 if (i >= ARRAY_SIZE(f2fs_sb_encoding_map))
255 *encoding = &f2fs_sb_encoding_map[i];
256 *flags = le16_to_cpu(sb->s_encoding_flags);
262 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
264 block_t limit = min((sbi->user_block_count << 1) / 1000,
265 sbi->user_block_count - sbi->reserved_blocks);
268 if (test_opt(sbi, RESERVE_ROOT) &&
269 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
270 F2FS_OPTION(sbi).root_reserved_blocks = limit;
271 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
272 F2FS_OPTION(sbi).root_reserved_blocks);
274 if (!test_opt(sbi, RESERVE_ROOT) &&
275 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
276 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
277 !gid_eq(F2FS_OPTION(sbi).s_resgid,
278 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
279 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
280 from_kuid_munged(&init_user_ns,
281 F2FS_OPTION(sbi).s_resuid),
282 from_kgid_munged(&init_user_ns,
283 F2FS_OPTION(sbi).s_resgid));
286 static void init_once(void *foo)
288 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
290 inode_init_once(&fi->vfs_inode);
294 static const char * const quotatypes[] = INITQFNAMES;
295 #define QTYPE2NAME(t) (quotatypes[t])
296 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
299 struct f2fs_sb_info *sbi = F2FS_SB(sb);
303 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
304 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
307 if (f2fs_sb_has_quota_ino(sbi)) {
308 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
312 qname = match_strdup(args);
314 f2fs_err(sbi, "Not enough memory for storing quotafile name");
317 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
318 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
321 f2fs_err(sbi, "%s quota file already specified",
325 if (strchr(qname, '/')) {
326 f2fs_err(sbi, "quotafile must be on filesystem root");
329 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
337 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
339 struct f2fs_sb_info *sbi = F2FS_SB(sb);
341 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
342 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
345 kvfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
346 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
350 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
353 * We do the test below only for project quotas. 'usrquota' and
354 * 'grpquota' mount options are allowed even without quota feature
355 * to support legacy quotas in quota files.
357 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
358 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
361 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
362 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
363 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
364 if (test_opt(sbi, USRQUOTA) &&
365 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
366 clear_opt(sbi, USRQUOTA);
368 if (test_opt(sbi, GRPQUOTA) &&
369 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
370 clear_opt(sbi, GRPQUOTA);
372 if (test_opt(sbi, PRJQUOTA) &&
373 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
374 clear_opt(sbi, PRJQUOTA);
376 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
377 test_opt(sbi, PRJQUOTA)) {
378 f2fs_err(sbi, "old and new quota format mixing");
382 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
383 f2fs_err(sbi, "journaled quota format not specified");
388 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
389 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
390 F2FS_OPTION(sbi).s_jquota_fmt = 0;
396 static int parse_options(struct super_block *sb, char *options)
398 struct f2fs_sb_info *sbi = F2FS_SB(sb);
399 substring_t args[MAX_OPT_ARGS];
400 unsigned char (*ext)[F2FS_EXTENSION_LEN];
402 int arg = 0, ext_cnt;
412 while ((p = strsep(&options, ",")) != NULL) {
417 * Initialize args struct so we know whether arg was
418 * found; some options take optional arguments.
420 args[0].to = args[0].from = NULL;
421 token = match_token(p, f2fs_tokens, args);
424 case Opt_gc_background:
425 name = match_strdup(&args[0]);
429 if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
431 clear_opt(sbi, FORCE_FG_GC);
432 } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
433 clear_opt(sbi, BG_GC);
434 clear_opt(sbi, FORCE_FG_GC);
435 } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
437 set_opt(sbi, FORCE_FG_GC);
444 case Opt_disable_roll_forward:
445 set_opt(sbi, DISABLE_ROLL_FORWARD);
448 /* this option mounts f2fs with ro */
449 set_opt(sbi, DISABLE_ROLL_FORWARD);
450 if (!f2fs_readonly(sb))
454 set_opt(sbi, DISCARD);
457 if (f2fs_sb_has_blkzoned(sbi)) {
458 f2fs_warn(sbi, "discard is required for zoned block devices");
461 clear_opt(sbi, DISCARD);
464 set_opt(sbi, NOHEAP);
467 clear_opt(sbi, NOHEAP);
469 #ifdef CONFIG_F2FS_FS_XATTR
471 set_opt(sbi, XATTR_USER);
473 case Opt_nouser_xattr:
474 clear_opt(sbi, XATTR_USER);
476 case Opt_inline_xattr:
477 set_opt(sbi, INLINE_XATTR);
479 case Opt_noinline_xattr:
480 clear_opt(sbi, INLINE_XATTR);
482 case Opt_inline_xattr_size:
483 if (args->from && match_int(args, &arg))
485 set_opt(sbi, INLINE_XATTR_SIZE);
486 F2FS_OPTION(sbi).inline_xattr_size = arg;
490 f2fs_info(sbi, "user_xattr options not supported");
492 case Opt_nouser_xattr:
493 f2fs_info(sbi, "nouser_xattr options not supported");
495 case Opt_inline_xattr:
496 f2fs_info(sbi, "inline_xattr options not supported");
498 case Opt_noinline_xattr:
499 f2fs_info(sbi, "noinline_xattr options not supported");
502 #ifdef CONFIG_F2FS_FS_POSIX_ACL
504 set_opt(sbi, POSIX_ACL);
507 clear_opt(sbi, POSIX_ACL);
511 f2fs_info(sbi, "acl options not supported");
514 f2fs_info(sbi, "noacl options not supported");
517 case Opt_active_logs:
518 if (args->from && match_int(args, &arg))
520 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
522 F2FS_OPTION(sbi).active_logs = arg;
524 case Opt_disable_ext_identify:
525 set_opt(sbi, DISABLE_EXT_IDENTIFY);
527 case Opt_inline_data:
528 set_opt(sbi, INLINE_DATA);
530 case Opt_inline_dentry:
531 set_opt(sbi, INLINE_DENTRY);
533 case Opt_noinline_dentry:
534 clear_opt(sbi, INLINE_DENTRY);
536 case Opt_flush_merge:
537 set_opt(sbi, FLUSH_MERGE);
539 case Opt_noflush_merge:
540 clear_opt(sbi, FLUSH_MERGE);
543 set_opt(sbi, NOBARRIER);
546 set_opt(sbi, FASTBOOT);
548 case Opt_extent_cache:
549 set_opt(sbi, EXTENT_CACHE);
551 case Opt_noextent_cache:
552 clear_opt(sbi, EXTENT_CACHE);
554 case Opt_noinline_data:
555 clear_opt(sbi, INLINE_DATA);
558 set_opt(sbi, DATA_FLUSH);
560 case Opt_reserve_root:
561 if (args->from && match_int(args, &arg))
563 if (test_opt(sbi, RESERVE_ROOT)) {
564 f2fs_info(sbi, "Preserve previous reserve_root=%u",
565 F2FS_OPTION(sbi).root_reserved_blocks);
567 F2FS_OPTION(sbi).root_reserved_blocks = arg;
568 set_opt(sbi, RESERVE_ROOT);
572 if (args->from && match_int(args, &arg))
574 uid = make_kuid(current_user_ns(), arg);
575 if (!uid_valid(uid)) {
576 f2fs_err(sbi, "Invalid uid value %d", arg);
579 F2FS_OPTION(sbi).s_resuid = uid;
582 if (args->from && match_int(args, &arg))
584 gid = make_kgid(current_user_ns(), arg);
585 if (!gid_valid(gid)) {
586 f2fs_err(sbi, "Invalid gid value %d", arg);
589 F2FS_OPTION(sbi).s_resgid = gid;
592 name = match_strdup(&args[0]);
596 if (strlen(name) == 8 &&
597 !strncmp(name, "adaptive", 8)) {
598 if (f2fs_sb_has_blkzoned(sbi)) {
599 f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
603 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
604 } else if (strlen(name) == 3 &&
605 !strncmp(name, "lfs", 3)) {
606 set_opt_mode(sbi, F2FS_MOUNT_LFS);
613 case Opt_io_size_bits:
614 if (args->from && match_int(args, &arg))
616 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_PAGES)) {
617 f2fs_warn(sbi, "Not support %d, larger than %d",
618 1 << arg, BIO_MAX_PAGES);
621 F2FS_OPTION(sbi).write_io_size_bits = arg;
623 #ifdef CONFIG_F2FS_FAULT_INJECTION
624 case Opt_fault_injection:
625 if (args->from && match_int(args, &arg))
627 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
628 set_opt(sbi, FAULT_INJECTION);
632 if (args->from && match_int(args, &arg))
634 f2fs_build_fault_attr(sbi, 0, arg);
635 set_opt(sbi, FAULT_INJECTION);
638 case Opt_fault_injection:
639 f2fs_info(sbi, "fault_injection options not supported");
643 f2fs_info(sbi, "fault_type options not supported");
647 sb->s_flags |= SB_LAZYTIME;
650 sb->s_flags &= ~SB_LAZYTIME;
655 set_opt(sbi, USRQUOTA);
658 set_opt(sbi, GRPQUOTA);
661 set_opt(sbi, PRJQUOTA);
664 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
669 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
674 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
678 case Opt_offusrjquota:
679 ret = f2fs_clear_qf_name(sb, USRQUOTA);
683 case Opt_offgrpjquota:
684 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
688 case Opt_offprjjquota:
689 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
693 case Opt_jqfmt_vfsold:
694 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
696 case Opt_jqfmt_vfsv0:
697 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
699 case Opt_jqfmt_vfsv1:
700 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
703 clear_opt(sbi, QUOTA);
704 clear_opt(sbi, USRQUOTA);
705 clear_opt(sbi, GRPQUOTA);
706 clear_opt(sbi, PRJQUOTA);
716 case Opt_offusrjquota:
717 case Opt_offgrpjquota:
718 case Opt_offprjjquota:
719 case Opt_jqfmt_vfsold:
720 case Opt_jqfmt_vfsv0:
721 case Opt_jqfmt_vfsv1:
723 f2fs_info(sbi, "quota operations not supported");
727 name = match_strdup(&args[0]);
730 if (strlen(name) == 10 &&
731 !strncmp(name, "user-based", 10)) {
732 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
733 } else if (strlen(name) == 3 &&
734 !strncmp(name, "off", 3)) {
735 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
736 } else if (strlen(name) == 8 &&
737 !strncmp(name, "fs-based", 8)) {
738 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
746 name = match_strdup(&args[0]);
750 if (strlen(name) == 7 &&
751 !strncmp(name, "default", 7)) {
752 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
753 } else if (strlen(name) == 5 &&
754 !strncmp(name, "reuse", 5)) {
755 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
763 name = match_strdup(&args[0]);
766 if (strlen(name) == 5 &&
767 !strncmp(name, "posix", 5)) {
768 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
769 } else if (strlen(name) == 6 &&
770 !strncmp(name, "strict", 6)) {
771 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
772 } else if (strlen(name) == 9 &&
773 !strncmp(name, "nobarrier", 9)) {
774 F2FS_OPTION(sbi).fsync_mode =
775 FSYNC_MODE_NOBARRIER;
782 case Opt_test_dummy_encryption:
783 #ifdef CONFIG_FS_ENCRYPTION
784 if (!f2fs_sb_has_encrypt(sbi)) {
785 f2fs_err(sbi, "Encrypt feature is off");
789 F2FS_OPTION(sbi).test_dummy_encryption = true;
790 f2fs_info(sbi, "Test dummy encryption mode enabled");
792 f2fs_info(sbi, "Test dummy encryption mount option ignored");
795 case Opt_checkpoint_disable_cap_perc:
796 if (args->from && match_int(args, &arg))
798 if (arg < 0 || arg > 100)
801 F2FS_OPTION(sbi).unusable_cap =
802 sbi->user_block_count;
804 F2FS_OPTION(sbi).unusable_cap =
805 (sbi->user_block_count / 100) * arg;
806 set_opt(sbi, DISABLE_CHECKPOINT);
808 case Opt_checkpoint_disable_cap:
809 if (args->from && match_int(args, &arg))
811 F2FS_OPTION(sbi).unusable_cap = arg;
812 set_opt(sbi, DISABLE_CHECKPOINT);
814 case Opt_checkpoint_disable:
815 set_opt(sbi, DISABLE_CHECKPOINT);
817 case Opt_checkpoint_enable:
818 clear_opt(sbi, DISABLE_CHECKPOINT);
820 case Opt_compress_algorithm:
821 if (!f2fs_sb_has_compression(sbi)) {
822 f2fs_err(sbi, "Compression feature if off");
825 name = match_strdup(&args[0]);
828 if (strlen(name) == 3 && !strcmp(name, "lzo")) {
829 F2FS_OPTION(sbi).compress_algorithm =
831 } else if (strlen(name) == 3 &&
832 !strcmp(name, "lz4")) {
833 F2FS_OPTION(sbi).compress_algorithm =
841 case Opt_compress_log_size:
842 if (!f2fs_sb_has_compression(sbi)) {
843 f2fs_err(sbi, "Compression feature is off");
846 if (args->from && match_int(args, &arg))
848 if (arg < MIN_COMPRESS_LOG_SIZE ||
849 arg > MAX_COMPRESS_LOG_SIZE) {
851 "Compress cluster log size is out of range");
854 F2FS_OPTION(sbi).compress_log_size = arg;
856 case Opt_compress_extension:
857 if (!f2fs_sb_has_compression(sbi)) {
858 f2fs_err(sbi, "Compression feature is off");
861 name = match_strdup(&args[0]);
865 ext = F2FS_OPTION(sbi).extensions;
866 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
868 if (strlen(name) >= F2FS_EXTENSION_LEN ||
869 ext_cnt >= COMPRESS_EXT_NUM) {
871 "invalid extension length/number");
876 strcpy(ext[ext_cnt], name);
877 F2FS_OPTION(sbi).compress_ext_cnt++;
881 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
887 if (f2fs_check_quota_options(sbi))
890 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
891 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
894 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
895 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
899 #ifndef CONFIG_UNICODE
900 if (f2fs_sb_has_casefold(sbi)) {
902 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
907 if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) {
908 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
909 F2FS_IO_SIZE_KB(sbi));
913 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
914 int min_size, max_size;
916 if (!f2fs_sb_has_extra_attr(sbi) ||
917 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
918 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
921 if (!test_opt(sbi, INLINE_XATTR)) {
922 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
926 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
927 max_size = MAX_INLINE_XATTR_SIZE;
929 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
930 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
931 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
937 if (test_opt(sbi, DISABLE_CHECKPOINT) && test_opt(sbi, LFS)) {
938 f2fs_err(sbi, "LFS not compatible with checkpoint=disable\n");
942 /* Not pass down write hints if the number of active logs is lesser
943 * than NR_CURSEG_TYPE.
945 if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
946 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
950 static struct inode *f2fs_alloc_inode(struct super_block *sb)
952 struct f2fs_inode_info *fi;
954 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
958 init_once((void *) fi);
960 /* Initialize f2fs-specific inode info */
961 atomic_set(&fi->dirty_pages, 0);
962 init_rwsem(&fi->i_sem);
963 INIT_LIST_HEAD(&fi->dirty_list);
964 INIT_LIST_HEAD(&fi->gdirty_list);
965 INIT_LIST_HEAD(&fi->inmem_ilist);
966 INIT_LIST_HEAD(&fi->inmem_pages);
967 mutex_init(&fi->inmem_lock);
968 init_rwsem(&fi->i_gc_rwsem[READ]);
969 init_rwsem(&fi->i_gc_rwsem[WRITE]);
970 init_rwsem(&fi->i_mmap_sem);
971 init_rwsem(&fi->i_xattr_sem);
973 /* Will be used by directory only */
974 fi->i_dir_level = F2FS_SB(sb)->dir_level;
976 return &fi->vfs_inode;
979 static int f2fs_drop_inode(struct inode *inode)
981 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
985 * during filesystem shutdown, if checkpoint is disabled,
986 * drop useless meta/node dirty pages.
988 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
989 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
990 inode->i_ino == F2FS_META_INO(sbi)) {
991 trace_f2fs_drop_inode(inode, 1);
997 * This is to avoid a deadlock condition like below.
998 * writeback_single_inode(inode)
999 * - f2fs_write_data_page
1000 * - f2fs_gc -> iput -> evict
1001 * - inode_wait_for_writeback(inode)
1003 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1004 if (!inode->i_nlink && !is_bad_inode(inode)) {
1005 /* to avoid evict_inode call simultaneously */
1006 atomic_inc(&inode->i_count);
1007 spin_unlock(&inode->i_lock);
1009 /* some remained atomic pages should discarded */
1010 if (f2fs_is_atomic_file(inode))
1011 f2fs_drop_inmem_pages(inode);
1013 /* should remain fi->extent_tree for writepage */
1014 f2fs_destroy_extent_node(inode);
1016 sb_start_intwrite(inode->i_sb);
1017 f2fs_i_size_write(inode, 0);
1019 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1020 inode, NULL, 0, DATA);
1021 truncate_inode_pages_final(inode->i_mapping);
1023 if (F2FS_HAS_BLOCKS(inode))
1024 f2fs_truncate(inode);
1026 sb_end_intwrite(inode->i_sb);
1028 spin_lock(&inode->i_lock);
1029 atomic_dec(&inode->i_count);
1031 trace_f2fs_drop_inode(inode, 0);
1034 ret = generic_drop_inode(inode);
1036 ret = fscrypt_drop_inode(inode);
1037 trace_f2fs_drop_inode(inode, ret);
1041 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1043 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1046 spin_lock(&sbi->inode_lock[DIRTY_META]);
1047 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1050 set_inode_flag(inode, FI_DIRTY_INODE);
1051 stat_inc_dirty_inode(sbi, DIRTY_META);
1053 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1054 list_add_tail(&F2FS_I(inode)->gdirty_list,
1055 &sbi->inode_list[DIRTY_META]);
1056 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1058 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1062 void f2fs_inode_synced(struct inode *inode)
1064 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1066 spin_lock(&sbi->inode_lock[DIRTY_META]);
1067 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1068 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1071 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1072 list_del_init(&F2FS_I(inode)->gdirty_list);
1073 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1075 clear_inode_flag(inode, FI_DIRTY_INODE);
1076 clear_inode_flag(inode, FI_AUTO_RECOVER);
1077 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1078 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1082 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1084 * We should call set_dirty_inode to write the dirty inode through write_inode.
1086 static void f2fs_dirty_inode(struct inode *inode, int flags)
1088 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1090 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1091 inode->i_ino == F2FS_META_INO(sbi))
1094 if (flags == I_DIRTY_TIME)
1097 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1098 clear_inode_flag(inode, FI_AUTO_RECOVER);
1100 f2fs_inode_dirtied(inode, false);
1103 static void f2fs_free_inode(struct inode *inode)
1105 fscrypt_free_inode(inode);
1106 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1109 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1111 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1112 percpu_counter_destroy(&sbi->total_valid_inode_count);
1115 static void destroy_device_list(struct f2fs_sb_info *sbi)
1119 for (i = 0; i < sbi->s_ndevs; i++) {
1120 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1121 #ifdef CONFIG_BLK_DEV_ZONED
1122 kvfree(FDEV(i).blkz_seq);
1128 static void f2fs_put_super(struct super_block *sb)
1130 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1134 f2fs_quota_off_umount(sb);
1136 /* prevent remaining shrinker jobs */
1137 mutex_lock(&sbi->umount_mutex);
1140 * We don't need to do checkpoint when superblock is clean.
1141 * But, the previous checkpoint was not done by umount, it needs to do
1142 * clean checkpoint again.
1144 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1145 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1146 struct cp_control cpc = {
1147 .reason = CP_UMOUNT,
1149 f2fs_write_checkpoint(sbi, &cpc);
1152 /* be sure to wait for any on-going discard commands */
1153 dropped = f2fs_issue_discard_timeout(sbi);
1155 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1156 !sbi->discard_blks && !dropped) {
1157 struct cp_control cpc = {
1158 .reason = CP_UMOUNT | CP_TRIMMED,
1160 f2fs_write_checkpoint(sbi, &cpc);
1164 * normally superblock is clean, so we need to release this.
1165 * In addition, EIO will skip do checkpoint, we need this as well.
1167 f2fs_release_ino_entry(sbi, true);
1169 f2fs_leave_shrinker(sbi);
1170 mutex_unlock(&sbi->umount_mutex);
1172 /* our cp_error case, we can wait for any writeback page */
1173 f2fs_flush_merged_writes(sbi);
1175 f2fs_wait_on_all_pages_writeback(sbi);
1177 f2fs_bug_on(sbi, sbi->fsync_node_num);
1179 iput(sbi->node_inode);
1180 sbi->node_inode = NULL;
1182 iput(sbi->meta_inode);
1183 sbi->meta_inode = NULL;
1186 * iput() can update stat information, if f2fs_write_checkpoint()
1187 * above failed with error.
1189 f2fs_destroy_stats(sbi);
1191 /* destroy f2fs internal modules */
1192 f2fs_destroy_node_manager(sbi);
1193 f2fs_destroy_segment_manager(sbi);
1195 f2fs_destroy_post_read_wq(sbi);
1199 f2fs_unregister_sysfs(sbi);
1201 sb->s_fs_info = NULL;
1202 if (sbi->s_chksum_driver)
1203 crypto_free_shash(sbi->s_chksum_driver);
1204 kvfree(sbi->raw_super);
1206 destroy_device_list(sbi);
1207 mempool_destroy(sbi->write_io_dummy);
1209 for (i = 0; i < MAXQUOTAS; i++)
1210 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1212 destroy_percpu_info(sbi);
1213 for (i = 0; i < NR_PAGE_TYPE; i++)
1214 kvfree(sbi->write_io[i]);
1215 #ifdef CONFIG_UNICODE
1216 utf8_unload(sbi->s_encoding);
1221 int f2fs_sync_fs(struct super_block *sb, int sync)
1223 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1226 if (unlikely(f2fs_cp_error(sbi)))
1228 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1231 trace_f2fs_sync_fs(sb, sync);
1233 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1237 struct cp_control cpc;
1239 cpc.reason = __get_cp_reason(sbi);
1241 mutex_lock(&sbi->gc_mutex);
1242 err = f2fs_write_checkpoint(sbi, &cpc);
1243 mutex_unlock(&sbi->gc_mutex);
1245 f2fs_trace_ios(NULL, 1);
1250 static int f2fs_freeze(struct super_block *sb)
1252 if (f2fs_readonly(sb))
1255 /* IO error happened before */
1256 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1259 /* must be clean, since sync_filesystem() was already called */
1260 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1265 static int f2fs_unfreeze(struct super_block *sb)
1271 static int f2fs_statfs_project(struct super_block *sb,
1272 kprojid_t projid, struct kstatfs *buf)
1275 struct dquot *dquot;
1279 qid = make_kqid_projid(projid);
1280 dquot = dqget(sb, qid);
1282 return PTR_ERR(dquot);
1283 spin_lock(&dquot->dq_dqb_lock);
1286 if (dquot->dq_dqb.dqb_bsoftlimit)
1287 limit = dquot->dq_dqb.dqb_bsoftlimit;
1288 if (dquot->dq_dqb.dqb_bhardlimit &&
1289 (!limit || dquot->dq_dqb.dqb_bhardlimit < limit))
1290 limit = dquot->dq_dqb.dqb_bhardlimit;
1292 limit >>= sb->s_blocksize_bits;
1294 if (limit && buf->f_blocks > limit) {
1295 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1296 buf->f_blocks = limit;
1297 buf->f_bfree = buf->f_bavail =
1298 (buf->f_blocks > curblock) ?
1299 (buf->f_blocks - curblock) : 0;
1303 if (dquot->dq_dqb.dqb_isoftlimit)
1304 limit = dquot->dq_dqb.dqb_isoftlimit;
1305 if (dquot->dq_dqb.dqb_ihardlimit &&
1306 (!limit || dquot->dq_dqb.dqb_ihardlimit < limit))
1307 limit = dquot->dq_dqb.dqb_ihardlimit;
1309 if (limit && buf->f_files > limit) {
1310 buf->f_files = limit;
1312 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1313 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1316 spin_unlock(&dquot->dq_dqb_lock);
1322 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1324 struct super_block *sb = dentry->d_sb;
1325 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1326 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1327 block_t total_count, user_block_count, start_count;
1328 u64 avail_node_count;
1330 total_count = le64_to_cpu(sbi->raw_super->block_count);
1331 user_block_count = sbi->user_block_count;
1332 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1333 buf->f_type = F2FS_SUPER_MAGIC;
1334 buf->f_bsize = sbi->blocksize;
1336 buf->f_blocks = total_count - start_count;
1337 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1338 sbi->current_reserved_blocks;
1340 spin_lock(&sbi->stat_lock);
1341 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1344 buf->f_bfree -= sbi->unusable_block_count;
1345 spin_unlock(&sbi->stat_lock);
1347 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1348 buf->f_bavail = buf->f_bfree -
1349 F2FS_OPTION(sbi).root_reserved_blocks;
1353 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1355 if (avail_node_count > user_block_count) {
1356 buf->f_files = user_block_count;
1357 buf->f_ffree = buf->f_bavail;
1359 buf->f_files = avail_node_count;
1360 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1364 buf->f_namelen = F2FS_NAME_LEN;
1365 buf->f_fsid.val[0] = (u32)id;
1366 buf->f_fsid.val[1] = (u32)(id >> 32);
1369 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1370 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1371 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1377 static inline void f2fs_show_quota_options(struct seq_file *seq,
1378 struct super_block *sb)
1381 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1383 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1386 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1397 seq_printf(seq, ",jqfmt=%s", fmtname);
1400 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1401 seq_show_option(seq, "usrjquota",
1402 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1404 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1405 seq_show_option(seq, "grpjquota",
1406 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1408 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1409 seq_show_option(seq, "prjjquota",
1410 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1414 static inline void f2fs_show_compress_options(struct seq_file *seq,
1415 struct super_block *sb)
1417 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1421 if (!f2fs_sb_has_compression(sbi))
1424 switch (F2FS_OPTION(sbi).compress_algorithm) {
1432 seq_printf(seq, ",compress_algorithm=%s", algtype);
1434 seq_printf(seq, ",compress_log_size=%u",
1435 F2FS_OPTION(sbi).compress_log_size);
1437 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1438 seq_printf(seq, ",compress_extension=%s",
1439 F2FS_OPTION(sbi).extensions[i]);
1443 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1445 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1447 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1448 if (test_opt(sbi, FORCE_FG_GC))
1449 seq_printf(seq, ",background_gc=%s", "sync");
1451 seq_printf(seq, ",background_gc=%s", "on");
1453 seq_printf(seq, ",background_gc=%s", "off");
1455 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1456 seq_puts(seq, ",disable_roll_forward");
1457 if (test_opt(sbi, DISCARD))
1458 seq_puts(seq, ",discard");
1460 seq_puts(seq, ",nodiscard");
1461 if (test_opt(sbi, NOHEAP))
1462 seq_puts(seq, ",no_heap");
1464 seq_puts(seq, ",heap");
1465 #ifdef CONFIG_F2FS_FS_XATTR
1466 if (test_opt(sbi, XATTR_USER))
1467 seq_puts(seq, ",user_xattr");
1469 seq_puts(seq, ",nouser_xattr");
1470 if (test_opt(sbi, INLINE_XATTR))
1471 seq_puts(seq, ",inline_xattr");
1473 seq_puts(seq, ",noinline_xattr");
1474 if (test_opt(sbi, INLINE_XATTR_SIZE))
1475 seq_printf(seq, ",inline_xattr_size=%u",
1476 F2FS_OPTION(sbi).inline_xattr_size);
1478 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1479 if (test_opt(sbi, POSIX_ACL))
1480 seq_puts(seq, ",acl");
1482 seq_puts(seq, ",noacl");
1484 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1485 seq_puts(seq, ",disable_ext_identify");
1486 if (test_opt(sbi, INLINE_DATA))
1487 seq_puts(seq, ",inline_data");
1489 seq_puts(seq, ",noinline_data");
1490 if (test_opt(sbi, INLINE_DENTRY))
1491 seq_puts(seq, ",inline_dentry");
1493 seq_puts(seq, ",noinline_dentry");
1494 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1495 seq_puts(seq, ",flush_merge");
1496 if (test_opt(sbi, NOBARRIER))
1497 seq_puts(seq, ",nobarrier");
1498 if (test_opt(sbi, FASTBOOT))
1499 seq_puts(seq, ",fastboot");
1500 if (test_opt(sbi, EXTENT_CACHE))
1501 seq_puts(seq, ",extent_cache");
1503 seq_puts(seq, ",noextent_cache");
1504 if (test_opt(sbi, DATA_FLUSH))
1505 seq_puts(seq, ",data_flush");
1507 seq_puts(seq, ",mode=");
1508 if (test_opt(sbi, ADAPTIVE))
1509 seq_puts(seq, "adaptive");
1510 else if (test_opt(sbi, LFS))
1511 seq_puts(seq, "lfs");
1512 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1513 if (test_opt(sbi, RESERVE_ROOT))
1514 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1515 F2FS_OPTION(sbi).root_reserved_blocks,
1516 from_kuid_munged(&init_user_ns,
1517 F2FS_OPTION(sbi).s_resuid),
1518 from_kgid_munged(&init_user_ns,
1519 F2FS_OPTION(sbi).s_resgid));
1520 if (F2FS_IO_SIZE_BITS(sbi))
1521 seq_printf(seq, ",io_bits=%u",
1522 F2FS_OPTION(sbi).write_io_size_bits);
1523 #ifdef CONFIG_F2FS_FAULT_INJECTION
1524 if (test_opt(sbi, FAULT_INJECTION)) {
1525 seq_printf(seq, ",fault_injection=%u",
1526 F2FS_OPTION(sbi).fault_info.inject_rate);
1527 seq_printf(seq, ",fault_type=%u",
1528 F2FS_OPTION(sbi).fault_info.inject_type);
1532 if (test_opt(sbi, QUOTA))
1533 seq_puts(seq, ",quota");
1534 if (test_opt(sbi, USRQUOTA))
1535 seq_puts(seq, ",usrquota");
1536 if (test_opt(sbi, GRPQUOTA))
1537 seq_puts(seq, ",grpquota");
1538 if (test_opt(sbi, PRJQUOTA))
1539 seq_puts(seq, ",prjquota");
1541 f2fs_show_quota_options(seq, sbi->sb);
1542 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1543 seq_printf(seq, ",whint_mode=%s", "user-based");
1544 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1545 seq_printf(seq, ",whint_mode=%s", "fs-based");
1546 #ifdef CONFIG_FS_ENCRYPTION
1547 if (F2FS_OPTION(sbi).test_dummy_encryption)
1548 seq_puts(seq, ",test_dummy_encryption");
1551 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1552 seq_printf(seq, ",alloc_mode=%s", "default");
1553 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1554 seq_printf(seq, ",alloc_mode=%s", "reuse");
1556 if (test_opt(sbi, DISABLE_CHECKPOINT))
1557 seq_printf(seq, ",checkpoint=disable:%u",
1558 F2FS_OPTION(sbi).unusable_cap);
1559 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1560 seq_printf(seq, ",fsync_mode=%s", "posix");
1561 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1562 seq_printf(seq, ",fsync_mode=%s", "strict");
1563 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1564 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1566 f2fs_show_compress_options(seq, sbi->sb);
1570 static void default_options(struct f2fs_sb_info *sbi)
1572 /* init some FS parameters */
1573 F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1574 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1575 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1576 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1577 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1578 F2FS_OPTION(sbi).test_dummy_encryption = false;
1579 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1580 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1581 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZO;
1582 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
1583 F2FS_OPTION(sbi).compress_ext_cnt = 0;
1585 set_opt(sbi, BG_GC);
1586 set_opt(sbi, INLINE_XATTR);
1587 set_opt(sbi, INLINE_DATA);
1588 set_opt(sbi, INLINE_DENTRY);
1589 set_opt(sbi, EXTENT_CACHE);
1590 set_opt(sbi, NOHEAP);
1591 clear_opt(sbi, DISABLE_CHECKPOINT);
1592 F2FS_OPTION(sbi).unusable_cap = 0;
1593 sbi->sb->s_flags |= SB_LAZYTIME;
1594 set_opt(sbi, FLUSH_MERGE);
1595 set_opt(sbi, DISCARD);
1596 if (f2fs_sb_has_blkzoned(sbi))
1597 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1599 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1601 #ifdef CONFIG_F2FS_FS_XATTR
1602 set_opt(sbi, XATTR_USER);
1604 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1605 set_opt(sbi, POSIX_ACL);
1608 f2fs_build_fault_attr(sbi, 0, 0);
1612 static int f2fs_enable_quotas(struct super_block *sb);
1615 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1617 unsigned int s_flags = sbi->sb->s_flags;
1618 struct cp_control cpc;
1623 if (s_flags & SB_RDONLY) {
1624 f2fs_err(sbi, "checkpoint=disable on readonly fs");
1627 sbi->sb->s_flags |= SB_ACTIVE;
1629 f2fs_update_time(sbi, DISABLE_TIME);
1631 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1632 mutex_lock(&sbi->gc_mutex);
1633 err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1634 if (err == -ENODATA) {
1638 if (err && err != -EAGAIN)
1642 ret = sync_filesystem(sbi->sb);
1644 err = ret ? ret: err;
1648 unusable = f2fs_get_unusable_blocks(sbi);
1649 if (f2fs_disable_cp_again(sbi, unusable)) {
1654 mutex_lock(&sbi->gc_mutex);
1655 cpc.reason = CP_PAUSE;
1656 set_sbi_flag(sbi, SBI_CP_DISABLED);
1657 err = f2fs_write_checkpoint(sbi, &cpc);
1661 spin_lock(&sbi->stat_lock);
1662 sbi->unusable_block_count = unusable;
1663 spin_unlock(&sbi->stat_lock);
1666 mutex_unlock(&sbi->gc_mutex);
1668 sbi->sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1672 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1674 mutex_lock(&sbi->gc_mutex);
1675 f2fs_dirty_to_prefree(sbi);
1677 clear_sbi_flag(sbi, SBI_CP_DISABLED);
1678 set_sbi_flag(sbi, SBI_IS_DIRTY);
1679 mutex_unlock(&sbi->gc_mutex);
1681 f2fs_sync_fs(sbi->sb, 1);
1684 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1686 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1687 struct f2fs_mount_info org_mount_opt;
1688 unsigned long old_sb_flags;
1690 bool need_restart_gc = false;
1691 bool need_stop_gc = false;
1692 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1693 bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1694 bool no_io_align = !F2FS_IO_ALIGNED(sbi);
1695 bool checkpoint_changed;
1701 * Save the old mount options in case we
1702 * need to restore them.
1704 org_mount_opt = sbi->mount_opt;
1705 old_sb_flags = sb->s_flags;
1708 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1709 for (i = 0; i < MAXQUOTAS; i++) {
1710 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1711 org_mount_opt.s_qf_names[i] =
1712 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1714 if (!org_mount_opt.s_qf_names[i]) {
1715 for (j = 0; j < i; j++)
1716 kvfree(org_mount_opt.s_qf_names[j]);
1720 org_mount_opt.s_qf_names[i] = NULL;
1725 /* recover superblocks we couldn't write due to previous RO mount */
1726 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1727 err = f2fs_commit_super(sbi, false);
1728 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
1731 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1734 default_options(sbi);
1736 /* parse mount options */
1737 err = parse_options(sb, data);
1740 checkpoint_changed =
1741 disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1744 * Previous and new state of filesystem is RO,
1745 * so skip checking GC and FLUSH_MERGE conditions.
1747 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1751 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1752 err = dquot_suspend(sb, -1);
1755 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1756 /* dquot_resume needs RW */
1757 sb->s_flags &= ~SB_RDONLY;
1758 if (sb_any_quota_suspended(sb)) {
1759 dquot_resume(sb, -1);
1760 } else if (f2fs_sb_has_quota_ino(sbi)) {
1761 err = f2fs_enable_quotas(sb);
1767 /* disallow enable/disable extent_cache dynamically */
1768 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1770 f2fs_warn(sbi, "switch extent_cache option is not allowed");
1774 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
1776 f2fs_warn(sbi, "switch io_bits option is not allowed");
1780 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1782 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
1787 * We stop the GC thread if FS is mounted as RO
1788 * or if background_gc = off is passed in mount
1789 * option. Also sync the filesystem.
1791 if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) {
1792 if (sbi->gc_thread) {
1793 f2fs_stop_gc_thread(sbi);
1794 need_restart_gc = true;
1796 } else if (!sbi->gc_thread) {
1797 err = f2fs_start_gc_thread(sbi);
1800 need_stop_gc = true;
1803 if (*flags & SB_RDONLY ||
1804 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1805 writeback_inodes_sb(sb, WB_REASON_SYNC);
1808 set_sbi_flag(sbi, SBI_IS_DIRTY);
1809 set_sbi_flag(sbi, SBI_IS_CLOSE);
1810 f2fs_sync_fs(sb, 1);
1811 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1814 if (checkpoint_changed) {
1815 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1816 err = f2fs_disable_checkpoint(sbi);
1820 f2fs_enable_checkpoint(sbi);
1825 * We stop issue flush thread if FS is mounted as RO
1826 * or if flush_merge is not passed in mount option.
1828 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1829 clear_opt(sbi, FLUSH_MERGE);
1830 f2fs_destroy_flush_cmd_control(sbi, false);
1832 err = f2fs_create_flush_cmd_control(sbi);
1838 /* Release old quota file names */
1839 for (i = 0; i < MAXQUOTAS; i++)
1840 kvfree(org_mount_opt.s_qf_names[i]);
1842 /* Update the POSIXACL Flag */
1843 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1844 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1846 limit_reserve_root(sbi);
1847 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1850 if (need_restart_gc) {
1851 if (f2fs_start_gc_thread(sbi))
1852 f2fs_warn(sbi, "background gc thread has stopped");
1853 } else if (need_stop_gc) {
1854 f2fs_stop_gc_thread(sbi);
1858 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1859 for (i = 0; i < MAXQUOTAS; i++) {
1860 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1861 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1864 sbi->mount_opt = org_mount_opt;
1865 sb->s_flags = old_sb_flags;
1870 /* Read data from quotafile */
1871 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1872 size_t len, loff_t off)
1874 struct inode *inode = sb_dqopt(sb)->files[type];
1875 struct address_space *mapping = inode->i_mapping;
1876 block_t blkidx = F2FS_BYTES_TO_BLK(off);
1877 int offset = off & (sb->s_blocksize - 1);
1880 loff_t i_size = i_size_read(inode);
1887 if (off + len > i_size)
1890 while (toread > 0) {
1891 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1893 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1895 if (PTR_ERR(page) == -ENOMEM) {
1896 congestion_wait(BLK_RW_ASYNC, HZ/50);
1899 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1900 return PTR_ERR(page);
1905 if (unlikely(page->mapping != mapping)) {
1906 f2fs_put_page(page, 1);
1909 if (unlikely(!PageUptodate(page))) {
1910 f2fs_put_page(page, 1);
1911 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1915 kaddr = kmap_atomic(page);
1916 memcpy(data, kaddr + offset, tocopy);
1917 kunmap_atomic(kaddr);
1918 f2fs_put_page(page, 1);
1928 /* Write to quotafile */
1929 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1930 const char *data, size_t len, loff_t off)
1932 struct inode *inode = sb_dqopt(sb)->files[type];
1933 struct address_space *mapping = inode->i_mapping;
1934 const struct address_space_operations *a_ops = mapping->a_ops;
1935 int offset = off & (sb->s_blocksize - 1);
1936 size_t towrite = len;
1942 while (towrite > 0) {
1943 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1946 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1948 if (unlikely(err)) {
1949 if (err == -ENOMEM) {
1950 congestion_wait(BLK_RW_ASYNC, HZ/50);
1953 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1957 kaddr = kmap_atomic(page);
1958 memcpy(kaddr + offset, data, tocopy);
1959 kunmap_atomic(kaddr);
1960 flush_dcache_page(page);
1962 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1973 inode->i_mtime = inode->i_ctime = current_time(inode);
1974 f2fs_mark_inode_dirty_sync(inode, false);
1975 return len - towrite;
1978 static struct dquot **f2fs_get_dquots(struct inode *inode)
1980 return F2FS_I(inode)->i_dquot;
1983 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1985 return &F2FS_I(inode)->i_reserved_quota;
1988 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1990 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
1991 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
1995 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
1996 F2FS_OPTION(sbi).s_jquota_fmt, type);
1999 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2004 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2005 err = f2fs_enable_quotas(sbi->sb);
2007 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2013 for (i = 0; i < MAXQUOTAS; i++) {
2014 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2015 err = f2fs_quota_on_mount(sbi, i);
2020 f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2027 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2030 struct inode *qf_inode;
2031 unsigned long qf_inum;
2034 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2036 qf_inum = f2fs_qf_ino(sb, type);
2040 qf_inode = f2fs_iget(sb, qf_inum);
2041 if (IS_ERR(qf_inode)) {
2042 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2043 return PTR_ERR(qf_inode);
2046 /* Don't account quota for quota files to avoid recursion */
2047 qf_inode->i_flags |= S_NOQUOTA;
2048 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2053 static int f2fs_enable_quotas(struct super_block *sb)
2055 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2057 unsigned long qf_inum;
2058 bool quota_mopt[MAXQUOTAS] = {
2059 test_opt(sbi, USRQUOTA),
2060 test_opt(sbi, GRPQUOTA),
2061 test_opt(sbi, PRJQUOTA),
2064 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2065 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2069 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2071 for (type = 0; type < MAXQUOTAS; type++) {
2072 qf_inum = f2fs_qf_ino(sb, type);
2074 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2075 DQUOT_USAGE_ENABLED |
2076 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2078 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2080 for (type--; type >= 0; type--)
2081 dquot_quota_off(sb, type);
2082 set_sbi_flag(F2FS_SB(sb),
2083 SBI_QUOTA_NEED_REPAIR);
2091 int f2fs_quota_sync(struct super_block *sb, int type)
2093 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2094 struct quota_info *dqopt = sb_dqopt(sb);
2101 * down_read(quota_sem)
2102 * dquot_writeback_dquots()
2105 * down_read(quota_sem)
2109 down_read(&sbi->quota_sem);
2110 ret = dquot_writeback_dquots(sb, type);
2115 * Now when everything is written we can discard the pagecache so
2116 * that userspace sees the changes.
2118 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2119 struct address_space *mapping;
2121 if (type != -1 && cnt != type)
2123 if (!sb_has_quota_active(sb, cnt))
2126 mapping = dqopt->files[cnt]->i_mapping;
2128 ret = filemap_fdatawrite(mapping);
2132 /* if we are using journalled quota */
2133 if (is_journalled_quota(sbi))
2136 ret = filemap_fdatawait(mapping);
2138 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2140 inode_lock(dqopt->files[cnt]);
2141 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
2142 inode_unlock(dqopt->files[cnt]);
2146 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2147 up_read(&sbi->quota_sem);
2148 f2fs_unlock_op(sbi);
2152 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2153 const struct path *path)
2155 struct inode *inode;
2158 /* if quota sysfile exists, deny enabling quota with specific file */
2159 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2160 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2164 err = f2fs_quota_sync(sb, type);
2168 err = dquot_quota_on(sb, type, format_id, path);
2172 inode = d_inode(path->dentry);
2175 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2176 f2fs_set_inode_flags(inode);
2177 inode_unlock(inode);
2178 f2fs_mark_inode_dirty_sync(inode, false);
2183 static int __f2fs_quota_off(struct super_block *sb, int type)
2185 struct inode *inode = sb_dqopt(sb)->files[type];
2188 if (!inode || !igrab(inode))
2189 return dquot_quota_off(sb, type);
2191 err = f2fs_quota_sync(sb, type);
2195 err = dquot_quota_off(sb, type);
2196 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2200 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2201 f2fs_set_inode_flags(inode);
2202 inode_unlock(inode);
2203 f2fs_mark_inode_dirty_sync(inode, false);
2209 static int f2fs_quota_off(struct super_block *sb, int type)
2211 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2214 err = __f2fs_quota_off(sb, type);
2217 * quotactl can shutdown journalled quota, result in inconsistence
2218 * between quota record and fs data by following updates, tag the
2219 * flag to let fsck be aware of it.
2221 if (is_journalled_quota(sbi))
2222 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2226 void f2fs_quota_off_umount(struct super_block *sb)
2231 for (type = 0; type < MAXQUOTAS; type++) {
2232 err = __f2fs_quota_off(sb, type);
2234 int ret = dquot_quota_off(sb, type);
2236 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2238 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2242 * In case of checkpoint=disable, we must flush quota blocks.
2243 * This can cause NULL exception for node_inode in end_io, since
2244 * put_super already dropped it.
2246 sync_filesystem(sb);
2249 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2251 struct quota_info *dqopt = sb_dqopt(sb);
2254 for (type = 0; type < MAXQUOTAS; type++) {
2255 if (!dqopt->files[type])
2257 f2fs_inode_synced(dqopt->files[type]);
2261 static int f2fs_dquot_commit(struct dquot *dquot)
2263 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2266 down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
2267 ret = dquot_commit(dquot);
2269 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2270 up_read(&sbi->quota_sem);
2274 static int f2fs_dquot_acquire(struct dquot *dquot)
2276 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2279 down_read(&sbi->quota_sem);
2280 ret = dquot_acquire(dquot);
2282 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2283 up_read(&sbi->quota_sem);
2287 static int f2fs_dquot_release(struct dquot *dquot)
2289 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2290 int ret = dquot_release(dquot);
2293 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2297 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2299 struct super_block *sb = dquot->dq_sb;
2300 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2301 int ret = dquot_mark_dquot_dirty(dquot);
2303 /* if we are using journalled quota */
2304 if (is_journalled_quota(sbi))
2305 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2310 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2312 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2313 int ret = dquot_commit_info(sb, type);
2316 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2320 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2322 *projid = F2FS_I(inode)->i_projid;
2326 static const struct dquot_operations f2fs_quota_operations = {
2327 .get_reserved_space = f2fs_get_reserved_space,
2328 .write_dquot = f2fs_dquot_commit,
2329 .acquire_dquot = f2fs_dquot_acquire,
2330 .release_dquot = f2fs_dquot_release,
2331 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
2332 .write_info = f2fs_dquot_commit_info,
2333 .alloc_dquot = dquot_alloc,
2334 .destroy_dquot = dquot_destroy,
2335 .get_projid = f2fs_get_projid,
2336 .get_next_id = dquot_get_next_id,
2339 static const struct quotactl_ops f2fs_quotactl_ops = {
2340 .quota_on = f2fs_quota_on,
2341 .quota_off = f2fs_quota_off,
2342 .quota_sync = f2fs_quota_sync,
2343 .get_state = dquot_get_state,
2344 .set_info = dquot_set_dqinfo,
2345 .get_dqblk = dquot_get_dqblk,
2346 .set_dqblk = dquot_set_dqblk,
2347 .get_nextdqblk = dquot_get_next_dqblk,
2350 int f2fs_quota_sync(struct super_block *sb, int type)
2355 void f2fs_quota_off_umount(struct super_block *sb)
2360 static const struct super_operations f2fs_sops = {
2361 .alloc_inode = f2fs_alloc_inode,
2362 .free_inode = f2fs_free_inode,
2363 .drop_inode = f2fs_drop_inode,
2364 .write_inode = f2fs_write_inode,
2365 .dirty_inode = f2fs_dirty_inode,
2366 .show_options = f2fs_show_options,
2368 .quota_read = f2fs_quota_read,
2369 .quota_write = f2fs_quota_write,
2370 .get_dquots = f2fs_get_dquots,
2372 .evict_inode = f2fs_evict_inode,
2373 .put_super = f2fs_put_super,
2374 .sync_fs = f2fs_sync_fs,
2375 .freeze_fs = f2fs_freeze,
2376 .unfreeze_fs = f2fs_unfreeze,
2377 .statfs = f2fs_statfs,
2378 .remount_fs = f2fs_remount,
2381 #ifdef CONFIG_FS_ENCRYPTION
2382 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2384 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2385 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2389 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2392 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2395 * Encrypting the root directory is not allowed because fsck
2396 * expects lost+found directory to exist and remain unencrypted
2397 * if LOST_FOUND feature is enabled.
2400 if (f2fs_sb_has_lost_found(sbi) &&
2401 inode->i_ino == F2FS_ROOT_INO(sbi))
2404 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2405 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2406 ctx, len, fs_data, XATTR_CREATE);
2409 static bool f2fs_dummy_context(struct inode *inode)
2411 return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
2414 static bool f2fs_has_stable_inodes(struct super_block *sb)
2419 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
2420 int *ino_bits_ret, int *lblk_bits_ret)
2422 *ino_bits_ret = 8 * sizeof(nid_t);
2423 *lblk_bits_ret = 8 * sizeof(block_t);
2426 static const struct fscrypt_operations f2fs_cryptops = {
2427 .key_prefix = "f2fs:",
2428 .get_context = f2fs_get_context,
2429 .set_context = f2fs_set_context,
2430 .dummy_context = f2fs_dummy_context,
2431 .empty_dir = f2fs_empty_dir,
2432 .max_namelen = F2FS_NAME_LEN,
2433 .has_stable_inodes = f2fs_has_stable_inodes,
2434 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits,
2438 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2439 u64 ino, u32 generation)
2441 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2442 struct inode *inode;
2444 if (f2fs_check_nid_range(sbi, ino))
2445 return ERR_PTR(-ESTALE);
2448 * f2fs_iget isn't quite right if the inode is currently unallocated!
2449 * However f2fs_iget currently does appropriate checks to handle stale
2450 * inodes so everything is OK.
2452 inode = f2fs_iget(sb, ino);
2454 return ERR_CAST(inode);
2455 if (unlikely(generation && inode->i_generation != generation)) {
2456 /* we didn't find the right inode.. */
2458 return ERR_PTR(-ESTALE);
2463 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2464 int fh_len, int fh_type)
2466 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2467 f2fs_nfs_get_inode);
2470 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2471 int fh_len, int fh_type)
2473 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2474 f2fs_nfs_get_inode);
2477 static const struct export_operations f2fs_export_ops = {
2478 .fh_to_dentry = f2fs_fh_to_dentry,
2479 .fh_to_parent = f2fs_fh_to_parent,
2480 .get_parent = f2fs_get_parent,
2483 static loff_t max_file_blocks(void)
2486 loff_t leaf_count = DEF_ADDRS_PER_BLOCK;
2489 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2490 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2491 * space in inode.i_addr, it will be more safe to reassign
2495 /* two direct node blocks */
2496 result += (leaf_count * 2);
2498 /* two indirect node blocks */
2499 leaf_count *= NIDS_PER_BLOCK;
2500 result += (leaf_count * 2);
2502 /* one double indirect node block */
2503 leaf_count *= NIDS_PER_BLOCK;
2504 result += leaf_count;
2509 static int __f2fs_commit_super(struct buffer_head *bh,
2510 struct f2fs_super_block *super)
2514 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2515 set_buffer_dirty(bh);
2518 /* it's rare case, we can do fua all the time */
2519 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2522 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2523 struct buffer_head *bh)
2525 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2526 (bh->b_data + F2FS_SUPER_OFFSET);
2527 struct super_block *sb = sbi->sb;
2528 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2529 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2530 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2531 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2532 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2533 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2534 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2535 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2536 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2537 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2538 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2539 u32 segment_count = le32_to_cpu(raw_super->segment_count);
2540 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2541 u64 main_end_blkaddr = main_blkaddr +
2542 (segment_count_main << log_blocks_per_seg);
2543 u64 seg_end_blkaddr = segment0_blkaddr +
2544 (segment_count << log_blocks_per_seg);
2546 if (segment0_blkaddr != cp_blkaddr) {
2547 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2548 segment0_blkaddr, cp_blkaddr);
2552 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2554 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2555 cp_blkaddr, sit_blkaddr,
2556 segment_count_ckpt << log_blocks_per_seg);
2560 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2562 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2563 sit_blkaddr, nat_blkaddr,
2564 segment_count_sit << log_blocks_per_seg);
2568 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2570 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2571 nat_blkaddr, ssa_blkaddr,
2572 segment_count_nat << log_blocks_per_seg);
2576 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2578 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2579 ssa_blkaddr, main_blkaddr,
2580 segment_count_ssa << log_blocks_per_seg);
2584 if (main_end_blkaddr > seg_end_blkaddr) {
2585 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2588 (segment_count << log_blocks_per_seg),
2589 segment_count_main << log_blocks_per_seg);
2591 } else if (main_end_blkaddr < seg_end_blkaddr) {
2595 /* fix in-memory information all the time */
2596 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2597 segment0_blkaddr) >> log_blocks_per_seg);
2599 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2600 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2603 err = __f2fs_commit_super(bh, NULL);
2604 res = err ? "failed" : "done";
2606 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%u) block(%u)",
2609 (segment_count << log_blocks_per_seg),
2610 segment_count_main << log_blocks_per_seg);
2617 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2618 struct buffer_head *bh)
2620 block_t segment_count, segs_per_sec, secs_per_zone;
2621 block_t total_sections, blocks_per_seg;
2622 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2623 (bh->b_data + F2FS_SUPER_OFFSET);
2624 unsigned int blocksize;
2625 size_t crc_offset = 0;
2628 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
2629 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
2630 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2634 /* Check checksum_offset and crc in superblock */
2635 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2636 crc_offset = le32_to_cpu(raw_super->checksum_offset);
2638 offsetof(struct f2fs_super_block, crc)) {
2639 f2fs_info(sbi, "Invalid SB checksum offset: %zu",
2641 return -EFSCORRUPTED;
2643 crc = le32_to_cpu(raw_super->crc);
2644 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2645 f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
2646 return -EFSCORRUPTED;
2650 /* Currently, support only 4KB page cache size */
2651 if (F2FS_BLKSIZE != PAGE_SIZE) {
2652 f2fs_info(sbi, "Invalid page_cache_size (%lu), supports only 4KB",
2654 return -EFSCORRUPTED;
2657 /* Currently, support only 4KB block size */
2658 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2659 if (blocksize != F2FS_BLKSIZE) {
2660 f2fs_info(sbi, "Invalid blocksize (%u), supports only 4KB",
2662 return -EFSCORRUPTED;
2665 /* check log blocks per segment */
2666 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2667 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
2668 le32_to_cpu(raw_super->log_blocks_per_seg));
2669 return -EFSCORRUPTED;
2672 /* Currently, support 512/1024/2048/4096 bytes sector size */
2673 if (le32_to_cpu(raw_super->log_sectorsize) >
2674 F2FS_MAX_LOG_SECTOR_SIZE ||
2675 le32_to_cpu(raw_super->log_sectorsize) <
2676 F2FS_MIN_LOG_SECTOR_SIZE) {
2677 f2fs_info(sbi, "Invalid log sectorsize (%u)",
2678 le32_to_cpu(raw_super->log_sectorsize));
2679 return -EFSCORRUPTED;
2681 if (le32_to_cpu(raw_super->log_sectors_per_block) +
2682 le32_to_cpu(raw_super->log_sectorsize) !=
2683 F2FS_MAX_LOG_SECTOR_SIZE) {
2684 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
2685 le32_to_cpu(raw_super->log_sectors_per_block),
2686 le32_to_cpu(raw_super->log_sectorsize));
2687 return -EFSCORRUPTED;
2690 segment_count = le32_to_cpu(raw_super->segment_count);
2691 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2692 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2693 total_sections = le32_to_cpu(raw_super->section_count);
2695 /* blocks_per_seg should be 512, given the above check */
2696 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2698 if (segment_count > F2FS_MAX_SEGMENT ||
2699 segment_count < F2FS_MIN_SEGMENTS) {
2700 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
2701 return -EFSCORRUPTED;
2704 if (total_sections > segment_count ||
2705 total_sections < F2FS_MIN_SEGMENTS ||
2706 segs_per_sec > segment_count || !segs_per_sec) {
2707 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
2708 segment_count, total_sections, segs_per_sec);
2709 return -EFSCORRUPTED;
2712 if ((segment_count / segs_per_sec) < total_sections) {
2713 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
2714 segment_count, segs_per_sec, total_sections);
2715 return -EFSCORRUPTED;
2718 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2719 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
2720 segment_count, le64_to_cpu(raw_super->block_count));
2721 return -EFSCORRUPTED;
2724 if (RDEV(0).path[0]) {
2725 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
2728 while (i < MAX_DEVICES && RDEV(i).path[0]) {
2729 dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
2732 if (segment_count != dev_seg_count) {
2733 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
2734 segment_count, dev_seg_count);
2735 return -EFSCORRUPTED;
2739 if (secs_per_zone > total_sections || !secs_per_zone) {
2740 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
2741 secs_per_zone, total_sections);
2742 return -EFSCORRUPTED;
2744 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2745 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2746 (le32_to_cpu(raw_super->extension_count) +
2747 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2748 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
2749 le32_to_cpu(raw_super->extension_count),
2750 raw_super->hot_ext_count,
2751 F2FS_MAX_EXTENSION);
2752 return -EFSCORRUPTED;
2755 if (le32_to_cpu(raw_super->cp_payload) >
2756 (blocks_per_seg - F2FS_CP_PACKS)) {
2757 f2fs_info(sbi, "Insane cp_payload (%u > %u)",
2758 le32_to_cpu(raw_super->cp_payload),
2759 blocks_per_seg - F2FS_CP_PACKS);
2760 return -EFSCORRUPTED;
2763 /* check reserved ino info */
2764 if (le32_to_cpu(raw_super->node_ino) != 1 ||
2765 le32_to_cpu(raw_super->meta_ino) != 2 ||
2766 le32_to_cpu(raw_super->root_ino) != 3) {
2767 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2768 le32_to_cpu(raw_super->node_ino),
2769 le32_to_cpu(raw_super->meta_ino),
2770 le32_to_cpu(raw_super->root_ino));
2771 return -EFSCORRUPTED;
2774 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2775 if (sanity_check_area_boundary(sbi, bh))
2776 return -EFSCORRUPTED;
2781 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2783 unsigned int total, fsmeta;
2784 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2785 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2786 unsigned int ovp_segments, reserved_segments;
2787 unsigned int main_segs, blocks_per_seg;
2788 unsigned int sit_segs, nat_segs;
2789 unsigned int sit_bitmap_size, nat_bitmap_size;
2790 unsigned int log_blocks_per_seg;
2791 unsigned int segment_count_main;
2792 unsigned int cp_pack_start_sum, cp_payload;
2793 block_t user_block_count, valid_user_blocks;
2794 block_t avail_node_count, valid_node_count;
2797 total = le32_to_cpu(raw_super->segment_count);
2798 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2799 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2801 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2803 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2804 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2806 if (unlikely(fsmeta >= total))
2809 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2810 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2812 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2813 ovp_segments == 0 || reserved_segments == 0)) {
2814 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
2818 user_block_count = le64_to_cpu(ckpt->user_block_count);
2819 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2820 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2821 if (!user_block_count || user_block_count >=
2822 segment_count_main << log_blocks_per_seg) {
2823 f2fs_err(sbi, "Wrong user_block_count: %u",
2828 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
2829 if (valid_user_blocks > user_block_count) {
2830 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
2831 valid_user_blocks, user_block_count);
2835 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
2836 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
2837 if (valid_node_count > avail_node_count) {
2838 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
2839 valid_node_count, avail_node_count);
2843 main_segs = le32_to_cpu(raw_super->segment_count_main);
2844 blocks_per_seg = sbi->blocks_per_seg;
2846 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2847 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2848 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2850 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2851 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2852 le32_to_cpu(ckpt->cur_node_segno[j])) {
2853 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
2855 le32_to_cpu(ckpt->cur_node_segno[i]));
2860 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2861 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2862 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2864 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2865 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2866 le32_to_cpu(ckpt->cur_data_segno[j])) {
2867 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
2869 le32_to_cpu(ckpt->cur_data_segno[i]));
2874 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2875 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
2876 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2877 le32_to_cpu(ckpt->cur_data_segno[j])) {
2878 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
2880 le32_to_cpu(ckpt->cur_node_segno[i]));
2886 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2887 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2889 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2890 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2891 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
2892 sit_bitmap_size, nat_bitmap_size);
2896 cp_pack_start_sum = __start_sum_addr(sbi);
2897 cp_payload = __cp_payload(sbi);
2898 if (cp_pack_start_sum < cp_payload + 1 ||
2899 cp_pack_start_sum > blocks_per_seg - 1 -
2901 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
2906 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
2907 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
2908 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
2909 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
2910 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
2911 le32_to_cpu(ckpt->checksum_offset));
2915 if (unlikely(f2fs_cp_error(sbi))) {
2916 f2fs_err(sbi, "A bug case: need to run fsck");
2922 static void init_sb_info(struct f2fs_sb_info *sbi)
2924 struct f2fs_super_block *raw_super = sbi->raw_super;
2927 sbi->log_sectors_per_block =
2928 le32_to_cpu(raw_super->log_sectors_per_block);
2929 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2930 sbi->blocksize = 1 << sbi->log_blocksize;
2931 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2932 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2933 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2934 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2935 sbi->total_sections = le32_to_cpu(raw_super->section_count);
2936 sbi->total_node_count =
2937 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2938 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2939 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2940 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2941 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2942 sbi->cur_victim_sec = NULL_SECNO;
2943 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
2944 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
2945 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2946 sbi->migration_granularity = sbi->segs_per_sec;
2948 sbi->dir_level = DEF_DIR_LEVEL;
2949 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2950 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2951 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
2952 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
2953 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
2954 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
2955 DEF_UMOUNT_DISCARD_TIMEOUT;
2956 clear_sbi_flag(sbi, SBI_NEED_FSCK);
2958 for (i = 0; i < NR_COUNT_TYPE; i++)
2959 atomic_set(&sbi->nr_pages[i], 0);
2961 for (i = 0; i < META; i++)
2962 atomic_set(&sbi->wb_sync_req[i], 0);
2964 INIT_LIST_HEAD(&sbi->s_list);
2965 mutex_init(&sbi->umount_mutex);
2966 init_rwsem(&sbi->io_order_lock);
2967 spin_lock_init(&sbi->cp_lock);
2969 sbi->dirty_device = 0;
2970 spin_lock_init(&sbi->dev_lock);
2972 init_rwsem(&sbi->sb_lock);
2973 init_rwsem(&sbi->pin_sem);
2976 static int init_percpu_info(struct f2fs_sb_info *sbi)
2980 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2984 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
2987 percpu_counter_destroy(&sbi->alloc_valid_block_count);
2992 #ifdef CONFIG_BLK_DEV_ZONED
2993 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
2996 struct f2fs_dev_info *dev = data;
2998 if (zone->type != BLK_ZONE_TYPE_CONVENTIONAL)
2999 set_bit(idx, dev->blkz_seq);
3003 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3005 struct block_device *bdev = FDEV(devi).bdev;
3006 sector_t nr_sectors = bdev->bd_part->nr_sects;
3009 if (!f2fs_sb_has_blkzoned(sbi))
3012 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3013 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
3015 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
3016 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
3017 __ilog2_u32(sbi->blocks_per_blkz))
3019 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
3020 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
3021 sbi->log_blocks_per_blkz;
3022 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
3023 FDEV(devi).nr_blkz++;
3025 FDEV(devi).blkz_seq = f2fs_kzalloc(sbi,
3026 BITS_TO_LONGS(FDEV(devi).nr_blkz)
3027 * sizeof(unsigned long),
3029 if (!FDEV(devi).blkz_seq)
3032 /* Get block zones type */
3033 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3043 * Read f2fs raw super block.
3044 * Because we have two copies of super block, so read both of them
3045 * to get the first valid one. If any one of them is broken, we pass
3046 * them recovery flag back to the caller.
3048 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3049 struct f2fs_super_block **raw_super,
3050 int *valid_super_block, int *recovery)
3052 struct super_block *sb = sbi->sb;
3054 struct buffer_head *bh;
3055 struct f2fs_super_block *super;
3058 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3062 for (block = 0; block < 2; block++) {
3063 bh = sb_bread(sb, block);
3065 f2fs_err(sbi, "Unable to read %dth superblock",
3072 /* sanity checking of raw super */
3073 err = sanity_check_raw_super(sbi, bh);
3075 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3083 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3085 *valid_super_block = block;
3091 /* No valid superblock */
3100 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3102 struct buffer_head *bh;
3106 if ((recover && f2fs_readonly(sbi->sb)) ||
3107 bdev_read_only(sbi->sb->s_bdev)) {
3108 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3112 /* we should update superblock crc here */
3113 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3114 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3115 offsetof(struct f2fs_super_block, crc));
3116 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3119 /* write back-up superblock first */
3120 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3123 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3126 /* if we are in recovery path, skip writing valid superblock */
3130 /* write current valid superblock */
3131 bh = sb_bread(sbi->sb, sbi->valid_super_block);
3134 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3139 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3141 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3142 unsigned int max_devices = MAX_DEVICES;
3145 /* Initialize single device information */
3146 if (!RDEV(0).path[0]) {
3147 if (!bdev_is_zoned(sbi->sb->s_bdev))
3153 * Initialize multiple devices information, or single
3154 * zoned block device information.
3156 sbi->devs = f2fs_kzalloc(sbi,
3157 array_size(max_devices,
3158 sizeof(struct f2fs_dev_info)),
3163 for (i = 0; i < max_devices; i++) {
3165 if (i > 0 && !RDEV(i).path[0])
3168 if (max_devices == 1) {
3169 /* Single zoned block device mount */
3171 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3172 sbi->sb->s_mode, sbi->sb->s_type);
3174 /* Multi-device mount */
3175 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3176 FDEV(i).total_segments =
3177 le32_to_cpu(RDEV(i).total_segments);
3179 FDEV(i).start_blk = 0;
3180 FDEV(i).end_blk = FDEV(i).start_blk +
3181 (FDEV(i).total_segments <<
3182 sbi->log_blocks_per_seg) - 1 +
3183 le32_to_cpu(raw_super->segment0_blkaddr);
3185 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3186 FDEV(i).end_blk = FDEV(i).start_blk +
3187 (FDEV(i).total_segments <<
3188 sbi->log_blocks_per_seg) - 1;
3190 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3191 sbi->sb->s_mode, sbi->sb->s_type);
3193 if (IS_ERR(FDEV(i).bdev))
3194 return PTR_ERR(FDEV(i).bdev);
3196 /* to release errored devices */
3197 sbi->s_ndevs = i + 1;
3199 #ifdef CONFIG_BLK_DEV_ZONED
3200 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3201 !f2fs_sb_has_blkzoned(sbi)) {
3202 f2fs_err(sbi, "Zoned block device feature not enabled\n");
3205 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3206 if (init_blkz_info(sbi, i)) {
3207 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3210 if (max_devices == 1)
3212 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3214 FDEV(i).total_segments,
3215 FDEV(i).start_blk, FDEV(i).end_blk,
3216 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3217 "Host-aware" : "Host-managed");
3221 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3223 FDEV(i).total_segments,
3224 FDEV(i).start_blk, FDEV(i).end_blk);
3227 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3231 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
3233 #ifdef CONFIG_UNICODE
3234 if (f2fs_sb_has_casefold(sbi) && !sbi->s_encoding) {
3235 const struct f2fs_sb_encodings *encoding_info;
3236 struct unicode_map *encoding;
3237 __u16 encoding_flags;
3239 if (f2fs_sb_has_encrypt(sbi)) {
3241 "Can't mount with encoding and encryption");
3245 if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info,
3248 "Encoding requested by superblock is unknown");
3252 encoding = utf8_load(encoding_info->version);
3253 if (IS_ERR(encoding)) {
3255 "can't mount with superblock charset: %s-%s "
3256 "not supported by the kernel. flags: 0x%x.",
3257 encoding_info->name, encoding_info->version,
3259 return PTR_ERR(encoding);
3261 f2fs_info(sbi, "Using encoding defined by superblock: "
3262 "%s-%s with flags 0x%hx", encoding_info->name,
3263 encoding_info->version?:"\b", encoding_flags);
3265 sbi->s_encoding = encoding;
3266 sbi->s_encoding_flags = encoding_flags;
3267 sbi->sb->s_d_op = &f2fs_dentry_ops;
3270 if (f2fs_sb_has_casefold(sbi)) {
3271 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
3278 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3280 struct f2fs_sm_info *sm_i = SM_I(sbi);
3282 /* adjust parameters according to the volume size */
3283 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3284 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3285 sm_i->dcc_info->discard_granularity = 1;
3286 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3289 sbi->readdir_ra = 1;
3292 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3294 struct f2fs_sb_info *sbi;
3295 struct f2fs_super_block *raw_super;
3298 bool skip_recovery = false, need_fsck = false;
3299 char *options = NULL;
3300 int recovery, i, valid_super_block;
3301 struct curseg_info *seg_i;
3307 valid_super_block = -1;
3310 /* allocate memory for f2fs-specific super block info */
3311 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3317 /* Load the checksum driver */
3318 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3319 if (IS_ERR(sbi->s_chksum_driver)) {
3320 f2fs_err(sbi, "Cannot load crc32 driver.");
3321 err = PTR_ERR(sbi->s_chksum_driver);
3322 sbi->s_chksum_driver = NULL;
3326 /* set a block size */
3327 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3328 f2fs_err(sbi, "unable to set blocksize");
3332 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3337 sb->s_fs_info = sbi;
3338 sbi->raw_super = raw_super;
3340 /* precompute checksum seed for metadata */
3341 if (f2fs_sb_has_inode_chksum(sbi))
3342 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3343 sizeof(raw_super->uuid));
3346 * The BLKZONED feature indicates that the drive was formatted with
3347 * zone alignment optimization. This is optional for host-aware
3348 * devices, but mandatory for host-managed zoned block devices.
3350 #ifndef CONFIG_BLK_DEV_ZONED
3351 if (f2fs_sb_has_blkzoned(sbi)) {
3352 f2fs_err(sbi, "Zoned block device support is not enabled");
3357 default_options(sbi);
3358 /* parse mount options */
3359 options = kstrdup((const char *)data, GFP_KERNEL);
3360 if (data && !options) {
3365 err = parse_options(sb, options);
3369 sbi->max_file_blocks = max_file_blocks();
3370 sb->s_maxbytes = sbi->max_file_blocks <<
3371 le32_to_cpu(raw_super->log_blocksize);
3372 sb->s_max_links = F2FS_LINK_MAX;
3374 err = f2fs_setup_casefold(sbi);
3379 sb->dq_op = &f2fs_quota_operations;
3380 sb->s_qcop = &f2fs_quotactl_ops;
3381 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3383 if (f2fs_sb_has_quota_ino(sbi)) {
3384 for (i = 0; i < MAXQUOTAS; i++) {
3385 if (f2fs_qf_ino(sbi->sb, i))
3386 sbi->nquota_files++;
3391 sb->s_op = &f2fs_sops;
3392 #ifdef CONFIG_FS_ENCRYPTION
3393 sb->s_cop = &f2fs_cryptops;
3395 #ifdef CONFIG_FS_VERITY
3396 sb->s_vop = &f2fs_verityops;
3398 sb->s_xattr = f2fs_xattr_handlers;
3399 sb->s_export_op = &f2fs_export_ops;
3400 sb->s_magic = F2FS_SUPER_MAGIC;
3401 sb->s_time_gran = 1;
3402 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3403 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3404 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3405 sb->s_iflags |= SB_I_CGROUPWB;
3407 /* init f2fs-specific super block info */
3408 sbi->valid_super_block = valid_super_block;
3409 mutex_init(&sbi->gc_mutex);
3410 mutex_init(&sbi->writepages);
3411 mutex_init(&sbi->cp_mutex);
3412 mutex_init(&sbi->resize_mutex);
3413 init_rwsem(&sbi->node_write);
3414 init_rwsem(&sbi->node_change);
3416 /* disallow all the data/node/meta page writes */
3417 set_sbi_flag(sbi, SBI_POR_DOING);
3418 spin_lock_init(&sbi->stat_lock);
3420 /* init iostat info */
3421 spin_lock_init(&sbi->iostat_lock);
3422 sbi->iostat_enable = false;
3424 for (i = 0; i < NR_PAGE_TYPE; i++) {
3425 int n = (i == META) ? 1: NR_TEMP_TYPE;
3431 sizeof(struct f2fs_bio_info)),
3433 if (!sbi->write_io[i]) {
3438 for (j = HOT; j < n; j++) {
3439 init_rwsem(&sbi->write_io[i][j].io_rwsem);
3440 sbi->write_io[i][j].sbi = sbi;
3441 sbi->write_io[i][j].bio = NULL;
3442 spin_lock_init(&sbi->write_io[i][j].io_lock);
3443 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3444 INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list);
3445 init_rwsem(&sbi->write_io[i][j].bio_list_lock);
3449 init_rwsem(&sbi->cp_rwsem);
3450 init_rwsem(&sbi->quota_sem);
3451 init_waitqueue_head(&sbi->cp_wait);
3454 err = init_percpu_info(sbi);
3458 if (F2FS_IO_ALIGNED(sbi)) {
3459 sbi->write_io_dummy =
3460 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3461 if (!sbi->write_io_dummy) {
3467 /* get an inode for meta space */
3468 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3469 if (IS_ERR(sbi->meta_inode)) {
3470 f2fs_err(sbi, "Failed to read F2FS meta data inode");
3471 err = PTR_ERR(sbi->meta_inode);
3475 err = f2fs_get_valid_checkpoint(sbi);
3477 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
3478 goto free_meta_inode;
3481 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3482 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3483 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3484 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3485 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3488 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
3489 set_sbi_flag(sbi, SBI_NEED_FSCK);
3491 /* Initialize device list */
3492 err = f2fs_scan_devices(sbi);
3494 f2fs_err(sbi, "Failed to find devices");
3498 err = f2fs_init_post_read_wq(sbi);
3500 f2fs_err(sbi, "Failed to initialize post read workqueue");
3504 sbi->total_valid_node_count =
3505 le32_to_cpu(sbi->ckpt->valid_node_count);
3506 percpu_counter_set(&sbi->total_valid_inode_count,
3507 le32_to_cpu(sbi->ckpt->valid_inode_count));
3508 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3509 sbi->total_valid_block_count =
3510 le64_to_cpu(sbi->ckpt->valid_block_count);
3511 sbi->last_valid_block_count = sbi->total_valid_block_count;
3512 sbi->reserved_blocks = 0;
3513 sbi->current_reserved_blocks = 0;
3514 limit_reserve_root(sbi);
3516 for (i = 0; i < NR_INODE_TYPE; i++) {
3517 INIT_LIST_HEAD(&sbi->inode_list[i]);
3518 spin_lock_init(&sbi->inode_lock[i]);
3520 mutex_init(&sbi->flush_lock);
3522 f2fs_init_extent_cache_info(sbi);
3524 f2fs_init_ino_entry_info(sbi);
3526 f2fs_init_fsync_node_info(sbi);
3528 /* setup f2fs internal modules */
3529 err = f2fs_build_segment_manager(sbi);
3531 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
3535 err = f2fs_build_node_manager(sbi);
3537 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
3542 /* For write statistics */
3543 if (sb->s_bdev->bd_part)
3544 sbi->sectors_written_start =
3545 (u64)part_stat_read(sb->s_bdev->bd_part,
3546 sectors[STAT_WRITE]);
3548 /* Read accumulated write IO statistics if exists */
3549 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3550 if (__exist_node_summaries(sbi))
3551 sbi->kbytes_written =
3552 le64_to_cpu(seg_i->journal->info.kbytes_written);
3554 f2fs_build_gc_manager(sbi);
3556 err = f2fs_build_stats(sbi);
3560 /* get an inode for node space */
3561 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3562 if (IS_ERR(sbi->node_inode)) {
3563 f2fs_err(sbi, "Failed to read node inode");
3564 err = PTR_ERR(sbi->node_inode);
3568 /* read root inode and dentry */
3569 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3571 f2fs_err(sbi, "Failed to read root inode");
3572 err = PTR_ERR(root);
3573 goto free_node_inode;
3575 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3576 !root->i_size || !root->i_nlink) {
3579 goto free_node_inode;
3582 sb->s_root = d_make_root(root); /* allocate root dentry */
3585 goto free_node_inode;
3588 err = f2fs_register_sysfs(sbi);
3590 goto free_root_inode;
3593 /* Enable quota usage during mount */
3594 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3595 err = f2fs_enable_quotas(sb);
3597 f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
3600 /* if there are nt orphan nodes free them */
3601 err = f2fs_recover_orphan_inodes(sbi);
3605 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3606 goto reset_checkpoint;
3608 /* recover fsynced data */
3609 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
3611 * mount should be failed, when device has readonly mode, and
3612 * previous checkpoint was not done by clean system shutdown.
3614 if (f2fs_hw_is_readonly(sbi)) {
3615 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3617 f2fs_err(sbi, "Need to recover fsync data, but write access unavailable");
3620 f2fs_info(sbi, "write access unavailable, skipping recovery");
3621 goto reset_checkpoint;
3625 set_sbi_flag(sbi, SBI_NEED_FSCK);
3628 goto reset_checkpoint;
3630 err = f2fs_recover_fsync_data(sbi, false);
3633 skip_recovery = true;
3635 f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
3640 err = f2fs_recover_fsync_data(sbi, true);
3642 if (!f2fs_readonly(sb) && err > 0) {
3644 f2fs_err(sbi, "Need to recover fsync data");
3650 * If the f2fs is not readonly and fsync data recovery succeeds,
3651 * check zoned block devices' write pointer consistency.
3653 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
3654 err = f2fs_check_write_pointer(sbi);
3660 /* f2fs_recover_fsync_data() cleared this already */
3661 clear_sbi_flag(sbi, SBI_POR_DOING);
3663 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3664 err = f2fs_disable_checkpoint(sbi);
3666 goto sync_free_meta;
3667 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3668 f2fs_enable_checkpoint(sbi);
3672 * If filesystem is not mounted as read-only then
3673 * do start the gc_thread.
3675 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
3676 /* After POR, we can run background GC thread.*/
3677 err = f2fs_start_gc_thread(sbi);
3679 goto sync_free_meta;
3683 /* recover broken superblock */
3685 err = f2fs_commit_super(sbi, true);
3686 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
3687 sbi->valid_super_block ? 1 : 2, err);
3690 f2fs_join_shrinker(sbi);
3692 f2fs_tuning_parameters(sbi);
3694 f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
3695 cur_cp_version(F2FS_CKPT(sbi)));
3696 f2fs_update_time(sbi, CP_TIME);
3697 f2fs_update_time(sbi, REQ_TIME);
3698 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3702 /* safe to flush all the data */
3703 sync_filesystem(sbi->sb);
3708 f2fs_truncate_quota_inode_pages(sb);
3709 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3710 f2fs_quota_off_umount(sbi->sb);
3713 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3714 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3715 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3716 * falls into an infinite loop in f2fs_sync_meta_pages().
3718 truncate_inode_pages_final(META_MAPPING(sbi));
3719 /* evict some inodes being cached by GC */
3721 f2fs_unregister_sysfs(sbi);
3726 f2fs_release_ino_entry(sbi, true);
3727 truncate_inode_pages_final(NODE_MAPPING(sbi));
3728 iput(sbi->node_inode);
3729 sbi->node_inode = NULL;
3731 f2fs_destroy_stats(sbi);
3733 f2fs_destroy_node_manager(sbi);
3735 f2fs_destroy_segment_manager(sbi);
3736 f2fs_destroy_post_read_wq(sbi);
3738 destroy_device_list(sbi);
3741 make_bad_inode(sbi->meta_inode);
3742 iput(sbi->meta_inode);
3743 sbi->meta_inode = NULL;
3745 mempool_destroy(sbi->write_io_dummy);
3747 destroy_percpu_info(sbi);
3749 for (i = 0; i < NR_PAGE_TYPE; i++)
3750 kvfree(sbi->write_io[i]);
3752 #ifdef CONFIG_UNICODE
3753 utf8_unload(sbi->s_encoding);
3757 for (i = 0; i < MAXQUOTAS; i++)
3758 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
3764 if (sbi->s_chksum_driver)
3765 crypto_free_shash(sbi->s_chksum_driver);
3768 /* give only one another chance */
3769 if (retry_cnt > 0 && skip_recovery) {
3771 shrink_dcache_sb(sb);
3777 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3778 const char *dev_name, void *data)
3780 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3783 static void kill_f2fs_super(struct super_block *sb)
3786 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3788 set_sbi_flag(sbi, SBI_IS_CLOSE);
3789 f2fs_stop_gc_thread(sbi);
3790 f2fs_stop_discard_thread(sbi);
3792 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3793 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3794 struct cp_control cpc = {
3795 .reason = CP_UMOUNT,
3797 f2fs_write_checkpoint(sbi, &cpc);
3800 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3801 sb->s_flags &= ~SB_RDONLY;
3803 kill_block_super(sb);
3806 static struct file_system_type f2fs_fs_type = {
3807 .owner = THIS_MODULE,
3809 .mount = f2fs_mount,
3810 .kill_sb = kill_f2fs_super,
3811 .fs_flags = FS_REQUIRES_DEV,
3813 MODULE_ALIAS_FS("f2fs");
3815 static int __init init_inodecache(void)
3817 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3818 sizeof(struct f2fs_inode_info), 0,
3819 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3820 if (!f2fs_inode_cachep)
3825 static void destroy_inodecache(void)
3828 * Make sure all delayed rcu free inodes are flushed before we
3832 kmem_cache_destroy(f2fs_inode_cachep);
3835 static int __init init_f2fs_fs(void)
3839 if (PAGE_SIZE != F2FS_BLKSIZE) {
3840 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3841 PAGE_SIZE, F2FS_BLKSIZE);
3845 f2fs_build_trace_ios();
3847 err = init_inodecache();
3850 err = f2fs_create_node_manager_caches();
3852 goto free_inodecache;
3853 err = f2fs_create_segment_manager_caches();
3855 goto free_node_manager_caches;
3856 err = f2fs_create_checkpoint_caches();
3858 goto free_segment_manager_caches;
3859 err = f2fs_create_extent_cache();
3861 goto free_checkpoint_caches;
3862 err = f2fs_init_sysfs();
3864 goto free_extent_cache;
3865 err = register_shrinker(&f2fs_shrinker_info);
3868 err = register_filesystem(&f2fs_fs_type);
3871 f2fs_create_root_stats();
3872 err = f2fs_init_post_read_processing();
3874 goto free_root_stats;
3875 err = f2fs_init_bio_entry_cache();
3877 goto free_post_read;
3878 err = f2fs_init_bioset();
3880 goto free_bio_enrty_cache;
3882 free_bio_enrty_cache:
3883 f2fs_destroy_bio_entry_cache();
3885 f2fs_destroy_post_read_processing();
3887 f2fs_destroy_root_stats();
3888 unregister_filesystem(&f2fs_fs_type);
3890 unregister_shrinker(&f2fs_shrinker_info);
3894 f2fs_destroy_extent_cache();
3895 free_checkpoint_caches:
3896 f2fs_destroy_checkpoint_caches();
3897 free_segment_manager_caches:
3898 f2fs_destroy_segment_manager_caches();
3899 free_node_manager_caches:
3900 f2fs_destroy_node_manager_caches();
3902 destroy_inodecache();
3907 static void __exit exit_f2fs_fs(void)
3909 f2fs_destroy_bioset();
3910 f2fs_destroy_bio_entry_cache();
3911 f2fs_destroy_post_read_processing();
3912 f2fs_destroy_root_stats();
3913 unregister_filesystem(&f2fs_fs_type);
3914 unregister_shrinker(&f2fs_shrinker_info);
3916 f2fs_destroy_extent_cache();
3917 f2fs_destroy_checkpoint_caches();
3918 f2fs_destroy_segment_manager_caches();
3919 f2fs_destroy_node_manager_caches();
3920 destroy_inodecache();
3921 f2fs_destroy_trace_ios();
3924 module_init(init_f2fs_fs)
3925 module_exit(exit_f2fs_fs)
3927 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3928 MODULE_DESCRIPTION("Flash Friendly File System");
3929 MODULE_LICENSE("GPL");