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>
34 #define CREATE_TRACE_POINTS
35 #include <trace/events/f2fs.h>
37 static struct kmem_cache *f2fs_inode_cachep;
39 #ifdef CONFIG_F2FS_FAULT_INJECTION
41 char *f2fs_fault_name[FAULT_MAX] = {
42 [FAULT_KMALLOC] = "kmalloc",
43 [FAULT_KVMALLOC] = "kvmalloc",
44 [FAULT_PAGE_ALLOC] = "page alloc",
45 [FAULT_PAGE_GET] = "page get",
46 [FAULT_ALLOC_BIO] = "alloc bio",
47 [FAULT_ALLOC_NID] = "alloc nid",
48 [FAULT_ORPHAN] = "orphan",
49 [FAULT_BLOCK] = "no more block",
50 [FAULT_DIR_DEPTH] = "too big dir depth",
51 [FAULT_EVICT_INODE] = "evict_inode fail",
52 [FAULT_TRUNCATE] = "truncate fail",
53 [FAULT_READ_IO] = "read IO error",
54 [FAULT_CHECKPOINT] = "checkpoint error",
55 [FAULT_DISCARD] = "discard error",
56 [FAULT_WRITE_IO] = "write IO error",
59 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
62 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
65 atomic_set(&ffi->inject_ops, 0);
66 ffi->inject_rate = rate;
70 ffi->inject_type = type;
73 memset(ffi, 0, sizeof(struct f2fs_fault_info));
77 /* f2fs-wide shrinker description */
78 static struct shrinker f2fs_shrinker_info = {
79 .scan_objects = f2fs_shrink_scan,
80 .count_objects = f2fs_shrink_count,
81 .seeks = DEFAULT_SEEKS,
86 Opt_disable_roll_forward,
97 Opt_disable_ext_identify,
100 Opt_inline_xattr_size,
138 Opt_test_dummy_encryption,
143 static match_table_t f2fs_tokens = {
144 {Opt_gc_background, "background_gc=%s"},
145 {Opt_disable_roll_forward, "disable_roll_forward"},
146 {Opt_norecovery, "norecovery"},
147 {Opt_discard, "discard"},
148 {Opt_nodiscard, "nodiscard"},
149 {Opt_noheap, "no_heap"},
151 {Opt_user_xattr, "user_xattr"},
152 {Opt_nouser_xattr, "nouser_xattr"},
154 {Opt_noacl, "noacl"},
155 {Opt_active_logs, "active_logs=%u"},
156 {Opt_disable_ext_identify, "disable_ext_identify"},
157 {Opt_inline_xattr, "inline_xattr"},
158 {Opt_noinline_xattr, "noinline_xattr"},
159 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
160 {Opt_inline_data, "inline_data"},
161 {Opt_inline_dentry, "inline_dentry"},
162 {Opt_noinline_dentry, "noinline_dentry"},
163 {Opt_flush_merge, "flush_merge"},
164 {Opt_noflush_merge, "noflush_merge"},
165 {Opt_nobarrier, "nobarrier"},
166 {Opt_fastboot, "fastboot"},
167 {Opt_extent_cache, "extent_cache"},
168 {Opt_noextent_cache, "noextent_cache"},
169 {Opt_noinline_data, "noinline_data"},
170 {Opt_data_flush, "data_flush"},
171 {Opt_reserve_root, "reserve_root=%u"},
172 {Opt_resgid, "resgid=%u"},
173 {Opt_resuid, "resuid=%u"},
174 {Opt_mode, "mode=%s"},
175 {Opt_io_size_bits, "io_bits=%u"},
176 {Opt_fault_injection, "fault_injection=%u"},
177 {Opt_fault_type, "fault_type=%u"},
178 {Opt_lazytime, "lazytime"},
179 {Opt_nolazytime, "nolazytime"},
180 {Opt_quota, "quota"},
181 {Opt_noquota, "noquota"},
182 {Opt_usrquota, "usrquota"},
183 {Opt_grpquota, "grpquota"},
184 {Opt_prjquota, "prjquota"},
185 {Opt_usrjquota, "usrjquota=%s"},
186 {Opt_grpjquota, "grpjquota=%s"},
187 {Opt_prjjquota, "prjjquota=%s"},
188 {Opt_offusrjquota, "usrjquota="},
189 {Opt_offgrpjquota, "grpjquota="},
190 {Opt_offprjjquota, "prjjquota="},
191 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
192 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
193 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
194 {Opt_whint, "whint_mode=%s"},
195 {Opt_alloc, "alloc_mode=%s"},
196 {Opt_fsync, "fsync_mode=%s"},
197 {Opt_test_dummy_encryption, "test_dummy_encryption"},
198 {Opt_checkpoint, "checkpoint=%s"},
202 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
204 struct va_format vaf;
210 printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
214 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
216 block_t limit = (sbi->user_block_count << 1) / 1000;
219 if (test_opt(sbi, RESERVE_ROOT) &&
220 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
221 F2FS_OPTION(sbi).root_reserved_blocks = limit;
222 f2fs_msg(sbi->sb, KERN_INFO,
223 "Reduce reserved blocks for root = %u",
224 F2FS_OPTION(sbi).root_reserved_blocks);
226 if (!test_opt(sbi, RESERVE_ROOT) &&
227 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
228 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
229 !gid_eq(F2FS_OPTION(sbi).s_resgid,
230 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
231 f2fs_msg(sbi->sb, KERN_INFO,
232 "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
233 from_kuid_munged(&init_user_ns,
234 F2FS_OPTION(sbi).s_resuid),
235 from_kgid_munged(&init_user_ns,
236 F2FS_OPTION(sbi).s_resgid));
239 static void init_once(void *foo)
241 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
243 inode_init_once(&fi->vfs_inode);
247 static const char * const quotatypes[] = INITQFNAMES;
248 #define QTYPE2NAME(t) (quotatypes[t])
249 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
252 struct f2fs_sb_info *sbi = F2FS_SB(sb);
256 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
257 f2fs_msg(sb, KERN_ERR,
258 "Cannot change journaled "
259 "quota options when quota turned on");
262 if (f2fs_sb_has_quota_ino(sb)) {
263 f2fs_msg(sb, KERN_INFO,
264 "QUOTA feature is enabled, so ignore qf_name");
268 qname = match_strdup(args);
270 f2fs_msg(sb, KERN_ERR,
271 "Not enough memory for storing quotafile name");
274 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
275 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
278 f2fs_msg(sb, KERN_ERR,
279 "%s quota file already specified",
283 if (strchr(qname, '/')) {
284 f2fs_msg(sb, KERN_ERR,
285 "quotafile must be on filesystem root");
288 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
296 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
298 struct f2fs_sb_info *sbi = F2FS_SB(sb);
300 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
301 f2fs_msg(sb, KERN_ERR, "Cannot change journaled quota options"
302 " when quota turned on");
305 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
306 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
310 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
313 * We do the test below only for project quotas. 'usrquota' and
314 * 'grpquota' mount options are allowed even without quota feature
315 * to support legacy quotas in quota files.
317 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi->sb)) {
318 f2fs_msg(sbi->sb, KERN_ERR, "Project quota feature not enabled. "
319 "Cannot enable project quota enforcement.");
322 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
323 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
324 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
325 if (test_opt(sbi, USRQUOTA) &&
326 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
327 clear_opt(sbi, USRQUOTA);
329 if (test_opt(sbi, GRPQUOTA) &&
330 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
331 clear_opt(sbi, GRPQUOTA);
333 if (test_opt(sbi, PRJQUOTA) &&
334 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
335 clear_opt(sbi, PRJQUOTA);
337 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
338 test_opt(sbi, PRJQUOTA)) {
339 f2fs_msg(sbi->sb, KERN_ERR, "old and new quota "
344 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
345 f2fs_msg(sbi->sb, KERN_ERR, "journaled quota format "
351 if (f2fs_sb_has_quota_ino(sbi->sb) && F2FS_OPTION(sbi).s_jquota_fmt) {
352 f2fs_msg(sbi->sb, KERN_INFO,
353 "QUOTA feature is enabled, so ignore jquota_fmt");
354 F2FS_OPTION(sbi).s_jquota_fmt = 0;
360 static int parse_options(struct super_block *sb, char *options)
362 struct f2fs_sb_info *sbi = F2FS_SB(sb);
363 substring_t args[MAX_OPT_ARGS];
375 while ((p = strsep(&options, ",")) != NULL) {
380 * Initialize args struct so we know whether arg was
381 * found; some options take optional arguments.
383 args[0].to = args[0].from = NULL;
384 token = match_token(p, f2fs_tokens, args);
387 case Opt_gc_background:
388 name = match_strdup(&args[0]);
392 if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
394 clear_opt(sbi, FORCE_FG_GC);
395 } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
396 clear_opt(sbi, BG_GC);
397 clear_opt(sbi, FORCE_FG_GC);
398 } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
400 set_opt(sbi, FORCE_FG_GC);
407 case Opt_disable_roll_forward:
408 set_opt(sbi, DISABLE_ROLL_FORWARD);
411 /* this option mounts f2fs with ro */
412 set_opt(sbi, DISABLE_ROLL_FORWARD);
413 if (!f2fs_readonly(sb))
417 set_opt(sbi, DISCARD);
420 if (f2fs_sb_has_blkzoned(sb)) {
421 f2fs_msg(sb, KERN_WARNING,
422 "discard is required for zoned block devices");
425 clear_opt(sbi, DISCARD);
428 set_opt(sbi, NOHEAP);
431 clear_opt(sbi, NOHEAP);
433 #ifdef CONFIG_F2FS_FS_XATTR
435 set_opt(sbi, XATTR_USER);
437 case Opt_nouser_xattr:
438 clear_opt(sbi, XATTR_USER);
440 case Opt_inline_xattr:
441 set_opt(sbi, INLINE_XATTR);
443 case Opt_noinline_xattr:
444 clear_opt(sbi, INLINE_XATTR);
446 case Opt_inline_xattr_size:
447 if (args->from && match_int(args, &arg))
449 set_opt(sbi, INLINE_XATTR_SIZE);
450 F2FS_OPTION(sbi).inline_xattr_size = arg;
454 f2fs_msg(sb, KERN_INFO,
455 "user_xattr options not supported");
457 case Opt_nouser_xattr:
458 f2fs_msg(sb, KERN_INFO,
459 "nouser_xattr options not supported");
461 case Opt_inline_xattr:
462 f2fs_msg(sb, KERN_INFO,
463 "inline_xattr options not supported");
465 case Opt_noinline_xattr:
466 f2fs_msg(sb, KERN_INFO,
467 "noinline_xattr options not supported");
470 #ifdef CONFIG_F2FS_FS_POSIX_ACL
472 set_opt(sbi, POSIX_ACL);
475 clear_opt(sbi, POSIX_ACL);
479 f2fs_msg(sb, KERN_INFO, "acl options not supported");
482 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
485 case Opt_active_logs:
486 if (args->from && match_int(args, &arg))
488 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
490 F2FS_OPTION(sbi).active_logs = arg;
492 case Opt_disable_ext_identify:
493 set_opt(sbi, DISABLE_EXT_IDENTIFY);
495 case Opt_inline_data:
496 set_opt(sbi, INLINE_DATA);
498 case Opt_inline_dentry:
499 set_opt(sbi, INLINE_DENTRY);
501 case Opt_noinline_dentry:
502 clear_opt(sbi, INLINE_DENTRY);
504 case Opt_flush_merge:
505 set_opt(sbi, FLUSH_MERGE);
507 case Opt_noflush_merge:
508 clear_opt(sbi, FLUSH_MERGE);
511 set_opt(sbi, NOBARRIER);
514 set_opt(sbi, FASTBOOT);
516 case Opt_extent_cache:
517 set_opt(sbi, EXTENT_CACHE);
519 case Opt_noextent_cache:
520 clear_opt(sbi, EXTENT_CACHE);
522 case Opt_noinline_data:
523 clear_opt(sbi, INLINE_DATA);
526 set_opt(sbi, DATA_FLUSH);
528 case Opt_reserve_root:
529 if (args->from && match_int(args, &arg))
531 if (test_opt(sbi, RESERVE_ROOT)) {
532 f2fs_msg(sb, KERN_INFO,
533 "Preserve previous reserve_root=%u",
534 F2FS_OPTION(sbi).root_reserved_blocks);
536 F2FS_OPTION(sbi).root_reserved_blocks = arg;
537 set_opt(sbi, RESERVE_ROOT);
541 if (args->from && match_int(args, &arg))
543 uid = make_kuid(current_user_ns(), arg);
544 if (!uid_valid(uid)) {
545 f2fs_msg(sb, KERN_ERR,
546 "Invalid uid value %d", arg);
549 F2FS_OPTION(sbi).s_resuid = uid;
552 if (args->from && match_int(args, &arg))
554 gid = make_kgid(current_user_ns(), arg);
555 if (!gid_valid(gid)) {
556 f2fs_msg(sb, KERN_ERR,
557 "Invalid gid value %d", arg);
560 F2FS_OPTION(sbi).s_resgid = gid;
563 name = match_strdup(&args[0]);
567 if (strlen(name) == 8 &&
568 !strncmp(name, "adaptive", 8)) {
569 if (f2fs_sb_has_blkzoned(sb)) {
570 f2fs_msg(sb, KERN_WARNING,
571 "adaptive mode is not allowed with "
572 "zoned block device feature");
576 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
577 } else if (strlen(name) == 3 &&
578 !strncmp(name, "lfs", 3)) {
579 set_opt_mode(sbi, F2FS_MOUNT_LFS);
586 case Opt_io_size_bits:
587 if (args->from && match_int(args, &arg))
589 if (arg > __ilog2_u32(BIO_MAX_PAGES)) {
590 f2fs_msg(sb, KERN_WARNING,
591 "Not support %d, larger than %d",
592 1 << arg, BIO_MAX_PAGES);
595 F2FS_OPTION(sbi).write_io_size_bits = arg;
597 #ifdef CONFIG_F2FS_FAULT_INJECTION
598 case Opt_fault_injection:
599 if (args->from && match_int(args, &arg))
601 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
602 set_opt(sbi, FAULT_INJECTION);
606 if (args->from && match_int(args, &arg))
608 f2fs_build_fault_attr(sbi, 0, arg);
609 set_opt(sbi, FAULT_INJECTION);
612 case Opt_fault_injection:
613 f2fs_msg(sb, KERN_INFO,
614 "fault_injection options not supported");
618 f2fs_msg(sb, KERN_INFO,
619 "fault_type options not supported");
623 sb->s_flags |= SB_LAZYTIME;
626 sb->s_flags &= ~SB_LAZYTIME;
631 set_opt(sbi, USRQUOTA);
634 set_opt(sbi, GRPQUOTA);
637 set_opt(sbi, PRJQUOTA);
640 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
645 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
650 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
654 case Opt_offusrjquota:
655 ret = f2fs_clear_qf_name(sb, USRQUOTA);
659 case Opt_offgrpjquota:
660 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
664 case Opt_offprjjquota:
665 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
669 case Opt_jqfmt_vfsold:
670 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
672 case Opt_jqfmt_vfsv0:
673 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
675 case Opt_jqfmt_vfsv1:
676 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
679 clear_opt(sbi, QUOTA);
680 clear_opt(sbi, USRQUOTA);
681 clear_opt(sbi, GRPQUOTA);
682 clear_opt(sbi, PRJQUOTA);
692 case Opt_offusrjquota:
693 case Opt_offgrpjquota:
694 case Opt_offprjjquota:
695 case Opt_jqfmt_vfsold:
696 case Opt_jqfmt_vfsv0:
697 case Opt_jqfmt_vfsv1:
699 f2fs_msg(sb, KERN_INFO,
700 "quota operations not supported");
704 name = match_strdup(&args[0]);
707 if (strlen(name) == 10 &&
708 !strncmp(name, "user-based", 10)) {
709 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
710 } else if (strlen(name) == 3 &&
711 !strncmp(name, "off", 3)) {
712 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
713 } else if (strlen(name) == 8 &&
714 !strncmp(name, "fs-based", 8)) {
715 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
723 name = match_strdup(&args[0]);
727 if (strlen(name) == 7 &&
728 !strncmp(name, "default", 7)) {
729 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
730 } else if (strlen(name) == 5 &&
731 !strncmp(name, "reuse", 5)) {
732 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
740 name = match_strdup(&args[0]);
743 if (strlen(name) == 5 &&
744 !strncmp(name, "posix", 5)) {
745 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
746 } else if (strlen(name) == 6 &&
747 !strncmp(name, "strict", 6)) {
748 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
749 } else if (strlen(name) == 9 &&
750 !strncmp(name, "nobarrier", 9)) {
751 F2FS_OPTION(sbi).fsync_mode =
752 FSYNC_MODE_NOBARRIER;
759 case Opt_test_dummy_encryption:
760 #ifdef CONFIG_F2FS_FS_ENCRYPTION
761 if (!f2fs_sb_has_encrypt(sb)) {
762 f2fs_msg(sb, KERN_ERR, "Encrypt feature is off");
766 F2FS_OPTION(sbi).test_dummy_encryption = true;
767 f2fs_msg(sb, KERN_INFO,
768 "Test dummy encryption mode enabled");
770 f2fs_msg(sb, KERN_INFO,
771 "Test dummy encryption mount option ignored");
775 name = match_strdup(&args[0]);
779 if (strlen(name) == 6 &&
780 !strncmp(name, "enable", 6)) {
781 clear_opt(sbi, DISABLE_CHECKPOINT);
782 } else if (strlen(name) == 7 &&
783 !strncmp(name, "disable", 7)) {
784 set_opt(sbi, DISABLE_CHECKPOINT);
792 f2fs_msg(sb, KERN_ERR,
793 "Unrecognized mount option \"%s\" or missing value",
799 if (f2fs_check_quota_options(sbi))
802 if (f2fs_sb_has_quota_ino(sbi->sb) && !f2fs_readonly(sbi->sb)) {
803 f2fs_msg(sbi->sb, KERN_INFO,
804 "Filesystem with quota feature cannot be mounted RDWR "
805 "without CONFIG_QUOTA");
808 if (f2fs_sb_has_project_quota(sbi->sb) && !f2fs_readonly(sbi->sb)) {
809 f2fs_msg(sb, KERN_ERR,
810 "Filesystem with project quota feature cannot be "
811 "mounted RDWR without CONFIG_QUOTA");
816 if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) {
817 f2fs_msg(sb, KERN_ERR,
818 "Should set mode=lfs with %uKB-sized IO",
819 F2FS_IO_SIZE_KB(sbi));
823 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
824 if (!f2fs_sb_has_extra_attr(sb) ||
825 !f2fs_sb_has_flexible_inline_xattr(sb)) {
826 f2fs_msg(sb, KERN_ERR,
827 "extra_attr or flexible_inline_xattr "
831 if (!test_opt(sbi, INLINE_XATTR)) {
832 f2fs_msg(sb, KERN_ERR,
833 "inline_xattr_size option should be "
834 "set with inline_xattr option");
837 if (!F2FS_OPTION(sbi).inline_xattr_size ||
838 F2FS_OPTION(sbi).inline_xattr_size >=
839 DEF_ADDRS_PER_INODE -
840 F2FS_TOTAL_EXTRA_ATTR_SIZE -
841 DEF_INLINE_RESERVED_SIZE -
842 DEF_MIN_INLINE_SIZE) {
843 f2fs_msg(sb, KERN_ERR,
844 "inline xattr size is out of range");
849 if (test_opt(sbi, DISABLE_CHECKPOINT) && test_opt(sbi, LFS)) {
850 f2fs_msg(sb, KERN_ERR,
851 "LFS not compatible with checkpoint=disable\n");
855 /* Not pass down write hints if the number of active logs is lesser
856 * than NR_CURSEG_TYPE.
858 if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
859 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
863 static struct inode *f2fs_alloc_inode(struct super_block *sb)
865 struct f2fs_inode_info *fi;
867 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
871 init_once((void *) fi);
873 /* Initialize f2fs-specific inode info */
874 atomic_set(&fi->dirty_pages, 0);
875 init_rwsem(&fi->i_sem);
876 INIT_LIST_HEAD(&fi->dirty_list);
877 INIT_LIST_HEAD(&fi->gdirty_list);
878 INIT_LIST_HEAD(&fi->inmem_ilist);
879 INIT_LIST_HEAD(&fi->inmem_pages);
880 mutex_init(&fi->inmem_lock);
881 init_rwsem(&fi->i_gc_rwsem[READ]);
882 init_rwsem(&fi->i_gc_rwsem[WRITE]);
883 init_rwsem(&fi->i_mmap_sem);
884 init_rwsem(&fi->i_xattr_sem);
886 /* Will be used by directory only */
887 fi->i_dir_level = F2FS_SB(sb)->dir_level;
889 return &fi->vfs_inode;
892 static int f2fs_drop_inode(struct inode *inode)
896 * This is to avoid a deadlock condition like below.
897 * writeback_single_inode(inode)
898 * - f2fs_write_data_page
899 * - f2fs_gc -> iput -> evict
900 * - inode_wait_for_writeback(inode)
902 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
903 if (!inode->i_nlink && !is_bad_inode(inode)) {
904 /* to avoid evict_inode call simultaneously */
905 atomic_inc(&inode->i_count);
906 spin_unlock(&inode->i_lock);
908 /* some remained atomic pages should discarded */
909 if (f2fs_is_atomic_file(inode))
910 f2fs_drop_inmem_pages(inode);
912 /* should remain fi->extent_tree for writepage */
913 f2fs_destroy_extent_node(inode);
915 sb_start_intwrite(inode->i_sb);
916 f2fs_i_size_write(inode, 0);
918 if (F2FS_HAS_BLOCKS(inode))
919 f2fs_truncate(inode);
921 sb_end_intwrite(inode->i_sb);
923 spin_lock(&inode->i_lock);
924 atomic_dec(&inode->i_count);
926 trace_f2fs_drop_inode(inode, 0);
929 ret = generic_drop_inode(inode);
930 trace_f2fs_drop_inode(inode, ret);
934 int f2fs_inode_dirtied(struct inode *inode, bool sync)
936 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
939 spin_lock(&sbi->inode_lock[DIRTY_META]);
940 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
943 set_inode_flag(inode, FI_DIRTY_INODE);
944 stat_inc_dirty_inode(sbi, DIRTY_META);
946 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
947 list_add_tail(&F2FS_I(inode)->gdirty_list,
948 &sbi->inode_list[DIRTY_META]);
949 inc_page_count(sbi, F2FS_DIRTY_IMETA);
951 spin_unlock(&sbi->inode_lock[DIRTY_META]);
955 void f2fs_inode_synced(struct inode *inode)
957 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
959 spin_lock(&sbi->inode_lock[DIRTY_META]);
960 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
961 spin_unlock(&sbi->inode_lock[DIRTY_META]);
964 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
965 list_del_init(&F2FS_I(inode)->gdirty_list);
966 dec_page_count(sbi, F2FS_DIRTY_IMETA);
968 clear_inode_flag(inode, FI_DIRTY_INODE);
969 clear_inode_flag(inode, FI_AUTO_RECOVER);
970 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
971 spin_unlock(&sbi->inode_lock[DIRTY_META]);
975 * f2fs_dirty_inode() is called from __mark_inode_dirty()
977 * We should call set_dirty_inode to write the dirty inode through write_inode.
979 static void f2fs_dirty_inode(struct inode *inode, int flags)
981 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
983 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
984 inode->i_ino == F2FS_META_INO(sbi))
987 if (flags == I_DIRTY_TIME)
990 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
991 clear_inode_flag(inode, FI_AUTO_RECOVER);
993 f2fs_inode_dirtied(inode, false);
996 static void f2fs_i_callback(struct rcu_head *head)
998 struct inode *inode = container_of(head, struct inode, i_rcu);
999 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1002 static void f2fs_destroy_inode(struct inode *inode)
1004 call_rcu(&inode->i_rcu, f2fs_i_callback);
1007 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1009 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1010 percpu_counter_destroy(&sbi->total_valid_inode_count);
1013 static void destroy_device_list(struct f2fs_sb_info *sbi)
1017 for (i = 0; i < sbi->s_ndevs; i++) {
1018 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1019 #ifdef CONFIG_BLK_DEV_ZONED
1020 kfree(FDEV(i).blkz_type);
1026 static void f2fs_put_super(struct super_block *sb)
1028 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1032 f2fs_quota_off_umount(sb);
1034 /* prevent remaining shrinker jobs */
1035 mutex_lock(&sbi->umount_mutex);
1038 * We don't need to do checkpoint when superblock is clean.
1039 * But, the previous checkpoint was not done by umount, it needs to do
1040 * clean checkpoint again.
1042 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1043 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1044 struct cp_control cpc = {
1045 .reason = CP_UMOUNT,
1047 f2fs_write_checkpoint(sbi, &cpc);
1050 /* be sure to wait for any on-going discard commands */
1051 dropped = f2fs_wait_discard_bios(sbi);
1053 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1054 !sbi->discard_blks && !dropped) {
1055 struct cp_control cpc = {
1056 .reason = CP_UMOUNT | CP_TRIMMED,
1058 f2fs_write_checkpoint(sbi, &cpc);
1061 /* f2fs_write_checkpoint can update stat informaion */
1062 f2fs_destroy_stats(sbi);
1065 * normally superblock is clean, so we need to release this.
1066 * In addition, EIO will skip do checkpoint, we need this as well.
1068 f2fs_release_ino_entry(sbi, true);
1070 f2fs_leave_shrinker(sbi);
1071 mutex_unlock(&sbi->umount_mutex);
1073 /* our cp_error case, we can wait for any writeback page */
1074 f2fs_flush_merged_writes(sbi);
1076 f2fs_wait_on_all_pages_writeback(sbi);
1078 f2fs_bug_on(sbi, sbi->fsync_node_num);
1080 iput(sbi->node_inode);
1081 iput(sbi->meta_inode);
1083 /* destroy f2fs internal modules */
1084 f2fs_destroy_node_manager(sbi);
1085 f2fs_destroy_segment_manager(sbi);
1089 f2fs_unregister_sysfs(sbi);
1091 sb->s_fs_info = NULL;
1092 if (sbi->s_chksum_driver)
1093 crypto_free_shash(sbi->s_chksum_driver);
1094 kfree(sbi->raw_super);
1096 destroy_device_list(sbi);
1097 mempool_destroy(sbi->write_io_dummy);
1099 for (i = 0; i < MAXQUOTAS; i++)
1100 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1102 destroy_percpu_info(sbi);
1103 for (i = 0; i < NR_PAGE_TYPE; i++)
1104 kfree(sbi->write_io[i]);
1108 int f2fs_sync_fs(struct super_block *sb, int sync)
1110 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1113 if (unlikely(f2fs_cp_error(sbi)))
1115 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1118 trace_f2fs_sync_fs(sb, sync);
1120 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1124 struct cp_control cpc;
1126 cpc.reason = __get_cp_reason(sbi);
1128 mutex_lock(&sbi->gc_mutex);
1129 err = f2fs_write_checkpoint(sbi, &cpc);
1130 mutex_unlock(&sbi->gc_mutex);
1132 f2fs_trace_ios(NULL, 1);
1137 static int f2fs_freeze(struct super_block *sb)
1139 if (f2fs_readonly(sb))
1142 /* IO error happened before */
1143 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1146 /* must be clean, since sync_filesystem() was already called */
1147 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1152 static int f2fs_unfreeze(struct super_block *sb)
1158 static int f2fs_statfs_project(struct super_block *sb,
1159 kprojid_t projid, struct kstatfs *buf)
1162 struct dquot *dquot;
1166 qid = make_kqid_projid(projid);
1167 dquot = dqget(sb, qid);
1169 return PTR_ERR(dquot);
1170 spin_lock(&dquot->dq_dqb_lock);
1172 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
1173 dquot->dq_dqb.dqb_bsoftlimit :
1174 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
1175 if (limit && buf->f_blocks > limit) {
1176 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1177 buf->f_blocks = limit;
1178 buf->f_bfree = buf->f_bavail =
1179 (buf->f_blocks > curblock) ?
1180 (buf->f_blocks - curblock) : 0;
1183 limit = dquot->dq_dqb.dqb_isoftlimit ?
1184 dquot->dq_dqb.dqb_isoftlimit :
1185 dquot->dq_dqb.dqb_ihardlimit;
1186 if (limit && buf->f_files > limit) {
1187 buf->f_files = limit;
1189 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1190 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1193 spin_unlock(&dquot->dq_dqb_lock);
1199 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1201 struct super_block *sb = dentry->d_sb;
1202 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1203 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1204 block_t total_count, user_block_count, start_count;
1205 u64 avail_node_count;
1207 total_count = le64_to_cpu(sbi->raw_super->block_count);
1208 user_block_count = sbi->user_block_count;
1209 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1210 buf->f_type = F2FS_SUPER_MAGIC;
1211 buf->f_bsize = sbi->blocksize;
1213 buf->f_blocks = total_count - start_count;
1214 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1215 sbi->current_reserved_blocks;
1216 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1219 buf->f_bfree -= sbi->unusable_block_count;
1221 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1222 buf->f_bavail = buf->f_bfree -
1223 F2FS_OPTION(sbi).root_reserved_blocks;
1227 avail_node_count = sbi->total_node_count - sbi->nquota_files -
1228 F2FS_RESERVED_NODE_NUM;
1230 if (avail_node_count > user_block_count) {
1231 buf->f_files = user_block_count;
1232 buf->f_ffree = buf->f_bavail;
1234 buf->f_files = avail_node_count;
1235 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1239 buf->f_namelen = F2FS_NAME_LEN;
1240 buf->f_fsid.val[0] = (u32)id;
1241 buf->f_fsid.val[1] = (u32)(id >> 32);
1244 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1245 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1246 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1252 static inline void f2fs_show_quota_options(struct seq_file *seq,
1253 struct super_block *sb)
1256 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1258 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1261 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1272 seq_printf(seq, ",jqfmt=%s", fmtname);
1275 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1276 seq_show_option(seq, "usrjquota",
1277 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1279 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1280 seq_show_option(seq, "grpjquota",
1281 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1283 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1284 seq_show_option(seq, "prjjquota",
1285 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1289 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1291 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1293 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1294 if (test_opt(sbi, FORCE_FG_GC))
1295 seq_printf(seq, ",background_gc=%s", "sync");
1297 seq_printf(seq, ",background_gc=%s", "on");
1299 seq_printf(seq, ",background_gc=%s", "off");
1301 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1302 seq_puts(seq, ",disable_roll_forward");
1303 if (test_opt(sbi, DISCARD))
1304 seq_puts(seq, ",discard");
1305 if (test_opt(sbi, NOHEAP))
1306 seq_puts(seq, ",no_heap");
1308 seq_puts(seq, ",heap");
1309 #ifdef CONFIG_F2FS_FS_XATTR
1310 if (test_opt(sbi, XATTR_USER))
1311 seq_puts(seq, ",user_xattr");
1313 seq_puts(seq, ",nouser_xattr");
1314 if (test_opt(sbi, INLINE_XATTR))
1315 seq_puts(seq, ",inline_xattr");
1317 seq_puts(seq, ",noinline_xattr");
1318 if (test_opt(sbi, INLINE_XATTR_SIZE))
1319 seq_printf(seq, ",inline_xattr_size=%u",
1320 F2FS_OPTION(sbi).inline_xattr_size);
1322 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1323 if (test_opt(sbi, POSIX_ACL))
1324 seq_puts(seq, ",acl");
1326 seq_puts(seq, ",noacl");
1328 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1329 seq_puts(seq, ",disable_ext_identify");
1330 if (test_opt(sbi, INLINE_DATA))
1331 seq_puts(seq, ",inline_data");
1333 seq_puts(seq, ",noinline_data");
1334 if (test_opt(sbi, INLINE_DENTRY))
1335 seq_puts(seq, ",inline_dentry");
1337 seq_puts(seq, ",noinline_dentry");
1338 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1339 seq_puts(seq, ",flush_merge");
1340 if (test_opt(sbi, NOBARRIER))
1341 seq_puts(seq, ",nobarrier");
1342 if (test_opt(sbi, FASTBOOT))
1343 seq_puts(seq, ",fastboot");
1344 if (test_opt(sbi, EXTENT_CACHE))
1345 seq_puts(seq, ",extent_cache");
1347 seq_puts(seq, ",noextent_cache");
1348 if (test_opt(sbi, DATA_FLUSH))
1349 seq_puts(seq, ",data_flush");
1351 seq_puts(seq, ",mode=");
1352 if (test_opt(sbi, ADAPTIVE))
1353 seq_puts(seq, "adaptive");
1354 else if (test_opt(sbi, LFS))
1355 seq_puts(seq, "lfs");
1356 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1357 if (test_opt(sbi, RESERVE_ROOT))
1358 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1359 F2FS_OPTION(sbi).root_reserved_blocks,
1360 from_kuid_munged(&init_user_ns,
1361 F2FS_OPTION(sbi).s_resuid),
1362 from_kgid_munged(&init_user_ns,
1363 F2FS_OPTION(sbi).s_resgid));
1364 if (F2FS_IO_SIZE_BITS(sbi))
1365 seq_printf(seq, ",io_bits=%u",
1366 F2FS_OPTION(sbi).write_io_size_bits);
1367 #ifdef CONFIG_F2FS_FAULT_INJECTION
1368 if (test_opt(sbi, FAULT_INJECTION)) {
1369 seq_printf(seq, ",fault_injection=%u",
1370 F2FS_OPTION(sbi).fault_info.inject_rate);
1371 seq_printf(seq, ",fault_type=%u",
1372 F2FS_OPTION(sbi).fault_info.inject_type);
1376 if (test_opt(sbi, QUOTA))
1377 seq_puts(seq, ",quota");
1378 if (test_opt(sbi, USRQUOTA))
1379 seq_puts(seq, ",usrquota");
1380 if (test_opt(sbi, GRPQUOTA))
1381 seq_puts(seq, ",grpquota");
1382 if (test_opt(sbi, PRJQUOTA))
1383 seq_puts(seq, ",prjquota");
1385 f2fs_show_quota_options(seq, sbi->sb);
1386 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1387 seq_printf(seq, ",whint_mode=%s", "user-based");
1388 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1389 seq_printf(seq, ",whint_mode=%s", "fs-based");
1390 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1391 if (F2FS_OPTION(sbi).test_dummy_encryption)
1392 seq_puts(seq, ",test_dummy_encryption");
1395 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1396 seq_printf(seq, ",alloc_mode=%s", "default");
1397 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1398 seq_printf(seq, ",alloc_mode=%s", "reuse");
1400 if (test_opt(sbi, DISABLE_CHECKPOINT))
1401 seq_puts(seq, ",checkpoint=disable");
1403 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1404 seq_printf(seq, ",fsync_mode=%s", "posix");
1405 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1406 seq_printf(seq, ",fsync_mode=%s", "strict");
1407 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1408 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1412 static void default_options(struct f2fs_sb_info *sbi)
1414 /* init some FS parameters */
1415 F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1416 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1417 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1418 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1419 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1420 F2FS_OPTION(sbi).test_dummy_encryption = false;
1421 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1422 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1424 set_opt(sbi, BG_GC);
1425 set_opt(sbi, INLINE_XATTR);
1426 set_opt(sbi, INLINE_DATA);
1427 set_opt(sbi, INLINE_DENTRY);
1428 set_opt(sbi, EXTENT_CACHE);
1429 set_opt(sbi, NOHEAP);
1430 clear_opt(sbi, DISABLE_CHECKPOINT);
1431 sbi->sb->s_flags |= SB_LAZYTIME;
1432 set_opt(sbi, FLUSH_MERGE);
1433 set_opt(sbi, DISCARD);
1434 if (f2fs_sb_has_blkzoned(sbi->sb))
1435 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1437 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1439 #ifdef CONFIG_F2FS_FS_XATTR
1440 set_opt(sbi, XATTR_USER);
1442 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1443 set_opt(sbi, POSIX_ACL);
1446 f2fs_build_fault_attr(sbi, 0, 0);
1450 static int f2fs_enable_quotas(struct super_block *sb);
1453 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1455 struct cp_control cpc;
1458 sbi->sb->s_flags |= SB_ACTIVE;
1460 mutex_lock(&sbi->gc_mutex);
1461 f2fs_update_time(sbi, DISABLE_TIME);
1463 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1464 err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1465 if (err == -ENODATA)
1467 if (err && err != -EAGAIN) {
1468 mutex_unlock(&sbi->gc_mutex);
1472 mutex_unlock(&sbi->gc_mutex);
1474 err = sync_filesystem(sbi->sb);
1478 if (f2fs_disable_cp_again(sbi))
1481 mutex_lock(&sbi->gc_mutex);
1482 cpc.reason = CP_PAUSE;
1483 set_sbi_flag(sbi, SBI_CP_DISABLED);
1484 f2fs_write_checkpoint(sbi, &cpc);
1486 sbi->unusable_block_count = 0;
1487 mutex_unlock(&sbi->gc_mutex);
1491 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1493 mutex_lock(&sbi->gc_mutex);
1494 f2fs_dirty_to_prefree(sbi);
1496 clear_sbi_flag(sbi, SBI_CP_DISABLED);
1497 set_sbi_flag(sbi, SBI_IS_DIRTY);
1498 mutex_unlock(&sbi->gc_mutex);
1500 f2fs_sync_fs(sbi->sb, 1);
1503 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1505 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1506 struct f2fs_mount_info org_mount_opt;
1507 unsigned long old_sb_flags;
1509 bool need_restart_gc = false;
1510 bool need_stop_gc = false;
1511 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1512 bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1513 bool checkpoint_changed;
1519 * Save the old mount options in case we
1520 * need to restore them.
1522 org_mount_opt = sbi->mount_opt;
1523 old_sb_flags = sb->s_flags;
1526 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1527 for (i = 0; i < MAXQUOTAS; i++) {
1528 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1529 org_mount_opt.s_qf_names[i] =
1530 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1532 if (!org_mount_opt.s_qf_names[i]) {
1533 for (j = 0; j < i; j++)
1534 kfree(org_mount_opt.s_qf_names[j]);
1538 org_mount_opt.s_qf_names[i] = NULL;
1543 /* recover superblocks we couldn't write due to previous RO mount */
1544 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1545 err = f2fs_commit_super(sbi, false);
1546 f2fs_msg(sb, KERN_INFO,
1547 "Try to recover all the superblocks, ret: %d", err);
1549 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1552 default_options(sbi);
1554 /* parse mount options */
1555 err = parse_options(sb, data);
1558 checkpoint_changed =
1559 disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1562 * Previous and new state of filesystem is RO,
1563 * so skip checking GC and FLUSH_MERGE conditions.
1565 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1569 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1570 err = dquot_suspend(sb, -1);
1573 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1574 /* dquot_resume needs RW */
1575 sb->s_flags &= ~SB_RDONLY;
1576 if (sb_any_quota_suspended(sb)) {
1577 dquot_resume(sb, -1);
1578 } else if (f2fs_sb_has_quota_ino(sb)) {
1579 err = f2fs_enable_quotas(sb);
1585 /* disallow enable/disable extent_cache dynamically */
1586 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1588 f2fs_msg(sbi->sb, KERN_WARNING,
1589 "switch extent_cache option is not allowed");
1593 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1595 f2fs_msg(sbi->sb, KERN_WARNING,
1596 "disabling checkpoint not compatible with read-only");
1601 * We stop the GC thread if FS is mounted as RO
1602 * or if background_gc = off is passed in mount
1603 * option. Also sync the filesystem.
1605 if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) {
1606 if (sbi->gc_thread) {
1607 f2fs_stop_gc_thread(sbi);
1608 need_restart_gc = true;
1610 } else if (!sbi->gc_thread) {
1611 err = f2fs_start_gc_thread(sbi);
1614 need_stop_gc = true;
1617 if (*flags & SB_RDONLY ||
1618 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1619 writeback_inodes_sb(sb, WB_REASON_SYNC);
1622 set_sbi_flag(sbi, SBI_IS_DIRTY);
1623 set_sbi_flag(sbi, SBI_IS_CLOSE);
1624 f2fs_sync_fs(sb, 1);
1625 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1628 if (checkpoint_changed) {
1629 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1630 err = f2fs_disable_checkpoint(sbi);
1634 f2fs_enable_checkpoint(sbi);
1639 * We stop issue flush thread if FS is mounted as RO
1640 * or if flush_merge is not passed in mount option.
1642 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1643 clear_opt(sbi, FLUSH_MERGE);
1644 f2fs_destroy_flush_cmd_control(sbi, false);
1646 err = f2fs_create_flush_cmd_control(sbi);
1652 /* Release old quota file names */
1653 for (i = 0; i < MAXQUOTAS; i++)
1654 kfree(org_mount_opt.s_qf_names[i]);
1656 /* Update the POSIXACL Flag */
1657 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1658 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1660 limit_reserve_root(sbi);
1661 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1664 if (need_restart_gc) {
1665 if (f2fs_start_gc_thread(sbi))
1666 f2fs_msg(sbi->sb, KERN_WARNING,
1667 "background gc thread has stopped");
1668 } else if (need_stop_gc) {
1669 f2fs_stop_gc_thread(sbi);
1673 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1674 for (i = 0; i < MAXQUOTAS; i++) {
1675 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1676 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1679 sbi->mount_opt = org_mount_opt;
1680 sb->s_flags = old_sb_flags;
1685 /* Read data from quotafile */
1686 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1687 size_t len, loff_t off)
1689 struct inode *inode = sb_dqopt(sb)->files[type];
1690 struct address_space *mapping = inode->i_mapping;
1691 block_t blkidx = F2FS_BYTES_TO_BLK(off);
1692 int offset = off & (sb->s_blocksize - 1);
1695 loff_t i_size = i_size_read(inode);
1702 if (off + len > i_size)
1705 while (toread > 0) {
1706 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1708 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1710 if (PTR_ERR(page) == -ENOMEM) {
1711 congestion_wait(BLK_RW_ASYNC, HZ/50);
1714 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1715 return PTR_ERR(page);
1720 if (unlikely(page->mapping != mapping)) {
1721 f2fs_put_page(page, 1);
1724 if (unlikely(!PageUptodate(page))) {
1725 f2fs_put_page(page, 1);
1726 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1730 kaddr = kmap_atomic(page);
1731 memcpy(data, kaddr + offset, tocopy);
1732 kunmap_atomic(kaddr);
1733 f2fs_put_page(page, 1);
1743 /* Write to quotafile */
1744 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1745 const char *data, size_t len, loff_t off)
1747 struct inode *inode = sb_dqopt(sb)->files[type];
1748 struct address_space *mapping = inode->i_mapping;
1749 const struct address_space_operations *a_ops = mapping->a_ops;
1750 int offset = off & (sb->s_blocksize - 1);
1751 size_t towrite = len;
1757 while (towrite > 0) {
1758 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1761 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1763 if (unlikely(err)) {
1764 if (err == -ENOMEM) {
1765 congestion_wait(BLK_RW_ASYNC, HZ/50);
1768 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1772 kaddr = kmap_atomic(page);
1773 memcpy(kaddr + offset, data, tocopy);
1774 kunmap_atomic(kaddr);
1775 flush_dcache_page(page);
1777 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1788 inode->i_mtime = inode->i_ctime = current_time(inode);
1789 f2fs_mark_inode_dirty_sync(inode, false);
1790 return len - towrite;
1793 static struct dquot **f2fs_get_dquots(struct inode *inode)
1795 return F2FS_I(inode)->i_dquot;
1798 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1800 return &F2FS_I(inode)->i_reserved_quota;
1803 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1805 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
1806 f2fs_msg(sbi->sb, KERN_ERR,
1807 "quota sysfile may be corrupted, skip loading it");
1811 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
1812 F2FS_OPTION(sbi).s_jquota_fmt, type);
1815 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1820 if (f2fs_sb_has_quota_ino(sbi->sb) && rdonly) {
1821 err = f2fs_enable_quotas(sbi->sb);
1823 f2fs_msg(sbi->sb, KERN_ERR,
1824 "Cannot turn on quota_ino: %d", err);
1830 for (i = 0; i < MAXQUOTAS; i++) {
1831 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1832 err = f2fs_quota_on_mount(sbi, i);
1837 f2fs_msg(sbi->sb, KERN_ERR,
1838 "Cannot turn on quotas: %d on %d", err, i);
1844 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
1847 struct inode *qf_inode;
1848 unsigned long qf_inum;
1851 BUG_ON(!f2fs_sb_has_quota_ino(sb));
1853 qf_inum = f2fs_qf_ino(sb, type);
1857 qf_inode = f2fs_iget(sb, qf_inum);
1858 if (IS_ERR(qf_inode)) {
1859 f2fs_msg(sb, KERN_ERR,
1860 "Bad quota inode %u:%lu", type, qf_inum);
1861 return PTR_ERR(qf_inode);
1864 /* Don't account quota for quota files to avoid recursion */
1865 qf_inode->i_flags |= S_NOQUOTA;
1866 err = dquot_enable(qf_inode, type, format_id, flags);
1871 static int f2fs_enable_quotas(struct super_block *sb)
1874 unsigned long qf_inum;
1875 bool quota_mopt[MAXQUOTAS] = {
1876 test_opt(F2FS_SB(sb), USRQUOTA),
1877 test_opt(F2FS_SB(sb), GRPQUOTA),
1878 test_opt(F2FS_SB(sb), PRJQUOTA),
1881 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
1882 f2fs_msg(sb, KERN_ERR,
1883 "quota file may be corrupted, skip loading it");
1887 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
1889 for (type = 0; type < MAXQUOTAS; type++) {
1890 qf_inum = f2fs_qf_ino(sb, type);
1892 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
1893 DQUOT_USAGE_ENABLED |
1894 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
1896 f2fs_msg(sb, KERN_ERR,
1897 "Failed to enable quota tracking "
1898 "(type=%d, err=%d). Please run "
1899 "fsck to fix.", type, err);
1900 for (type--; type >= 0; type--)
1901 dquot_quota_off(sb, type);
1902 set_sbi_flag(F2FS_SB(sb),
1903 SBI_QUOTA_NEED_REPAIR);
1911 int f2fs_quota_sync(struct super_block *sb, int type)
1913 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1914 struct quota_info *dqopt = sb_dqopt(sb);
1918 ret = dquot_writeback_dquots(sb, type);
1923 * Now when everything is written we can discard the pagecache so
1924 * that userspace sees the changes.
1926 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
1927 struct address_space *mapping;
1929 if (type != -1 && cnt != type)
1931 if (!sb_has_quota_active(sb, cnt))
1934 mapping = dqopt->files[cnt]->i_mapping;
1936 ret = filemap_fdatawrite(mapping);
1940 /* if we are using journalled quota */
1941 if (is_journalled_quota(sbi))
1944 ret = filemap_fdatawait(mapping);
1946 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1948 inode_lock(dqopt->files[cnt]);
1949 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
1950 inode_unlock(dqopt->files[cnt]);
1954 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1958 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
1959 const struct path *path)
1961 struct inode *inode;
1964 err = f2fs_quota_sync(sb, type);
1968 err = dquot_quota_on(sb, type, format_id, path);
1972 inode = d_inode(path->dentry);
1975 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
1976 f2fs_set_inode_flags(inode);
1977 inode_unlock(inode);
1978 f2fs_mark_inode_dirty_sync(inode, false);
1983 static int f2fs_quota_off(struct super_block *sb, int type)
1985 struct inode *inode = sb_dqopt(sb)->files[type];
1988 if (!inode || !igrab(inode))
1989 return dquot_quota_off(sb, type);
1991 err = f2fs_quota_sync(sb, type);
1995 err = dquot_quota_off(sb, type);
1996 if (err || f2fs_sb_has_quota_ino(sb))
2000 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2001 f2fs_set_inode_flags(inode);
2002 inode_unlock(inode);
2003 f2fs_mark_inode_dirty_sync(inode, false);
2009 void f2fs_quota_off_umount(struct super_block *sb)
2014 for (type = 0; type < MAXQUOTAS; type++) {
2015 err = f2fs_quota_off(sb, type);
2017 int ret = dquot_quota_off(sb, type);
2019 f2fs_msg(sb, KERN_ERR,
2020 "Fail to turn off disk quota "
2021 "(type: %d, err: %d, ret:%d), Please "
2022 "run fsck to fix it.", type, err, ret);
2023 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2028 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2030 struct quota_info *dqopt = sb_dqopt(sb);
2033 for (type = 0; type < MAXQUOTAS; type++) {
2034 if (!dqopt->files[type])
2036 f2fs_inode_synced(dqopt->files[type]);
2040 static int f2fs_dquot_commit(struct dquot *dquot)
2044 ret = dquot_commit(dquot);
2046 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2050 static int f2fs_dquot_acquire(struct dquot *dquot)
2054 ret = dquot_acquire(dquot);
2056 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2061 static int f2fs_dquot_release(struct dquot *dquot)
2065 ret = dquot_release(dquot);
2067 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2071 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2073 struct super_block *sb = dquot->dq_sb;
2074 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2077 ret = dquot_mark_dquot_dirty(dquot);
2079 /* if we are using journalled quota */
2080 if (is_journalled_quota(sbi))
2081 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2086 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2090 ret = dquot_commit_info(sb, type);
2092 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2096 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2098 *projid = F2FS_I(inode)->i_projid;
2102 static const struct dquot_operations f2fs_quota_operations = {
2103 .get_reserved_space = f2fs_get_reserved_space,
2104 .write_dquot = f2fs_dquot_commit,
2105 .acquire_dquot = f2fs_dquot_acquire,
2106 .release_dquot = f2fs_dquot_release,
2107 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
2108 .write_info = f2fs_dquot_commit_info,
2109 .alloc_dquot = dquot_alloc,
2110 .destroy_dquot = dquot_destroy,
2111 .get_projid = f2fs_get_projid,
2112 .get_next_id = dquot_get_next_id,
2115 static const struct quotactl_ops f2fs_quotactl_ops = {
2116 .quota_on = f2fs_quota_on,
2117 .quota_off = f2fs_quota_off,
2118 .quota_sync = f2fs_quota_sync,
2119 .get_state = dquot_get_state,
2120 .set_info = dquot_set_dqinfo,
2121 .get_dqblk = dquot_get_dqblk,
2122 .set_dqblk = dquot_set_dqblk,
2123 .get_nextdqblk = dquot_get_next_dqblk,
2126 int f2fs_quota_sync(struct super_block *sb, int type)
2131 void f2fs_quota_off_umount(struct super_block *sb)
2136 static const struct super_operations f2fs_sops = {
2137 .alloc_inode = f2fs_alloc_inode,
2138 .drop_inode = f2fs_drop_inode,
2139 .destroy_inode = f2fs_destroy_inode,
2140 .write_inode = f2fs_write_inode,
2141 .dirty_inode = f2fs_dirty_inode,
2142 .show_options = f2fs_show_options,
2144 .quota_read = f2fs_quota_read,
2145 .quota_write = f2fs_quota_write,
2146 .get_dquots = f2fs_get_dquots,
2148 .evict_inode = f2fs_evict_inode,
2149 .put_super = f2fs_put_super,
2150 .sync_fs = f2fs_sync_fs,
2151 .freeze_fs = f2fs_freeze,
2152 .unfreeze_fs = f2fs_unfreeze,
2153 .statfs = f2fs_statfs,
2154 .remount_fs = f2fs_remount,
2157 #ifdef CONFIG_F2FS_FS_ENCRYPTION
2158 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2160 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2161 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2165 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2168 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2171 * Encrypting the root directory is not allowed because fsck
2172 * expects lost+found directory to exist and remain unencrypted
2173 * if LOST_FOUND feature is enabled.
2176 if (f2fs_sb_has_lost_found(sbi->sb) &&
2177 inode->i_ino == F2FS_ROOT_INO(sbi))
2180 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2181 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2182 ctx, len, fs_data, XATTR_CREATE);
2185 static bool f2fs_dummy_context(struct inode *inode)
2187 return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
2190 static const struct fscrypt_operations f2fs_cryptops = {
2191 .key_prefix = "f2fs:",
2192 .get_context = f2fs_get_context,
2193 .set_context = f2fs_set_context,
2194 .dummy_context = f2fs_dummy_context,
2195 .empty_dir = f2fs_empty_dir,
2196 .max_namelen = F2FS_NAME_LEN,
2200 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2201 u64 ino, u32 generation)
2203 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2204 struct inode *inode;
2206 if (f2fs_check_nid_range(sbi, ino))
2207 return ERR_PTR(-ESTALE);
2210 * f2fs_iget isn't quite right if the inode is currently unallocated!
2211 * However f2fs_iget currently does appropriate checks to handle stale
2212 * inodes so everything is OK.
2214 inode = f2fs_iget(sb, ino);
2216 return ERR_CAST(inode);
2217 if (unlikely(generation && inode->i_generation != generation)) {
2218 /* we didn't find the right inode.. */
2220 return ERR_PTR(-ESTALE);
2225 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2226 int fh_len, int fh_type)
2228 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2229 f2fs_nfs_get_inode);
2232 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2233 int fh_len, int fh_type)
2235 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2236 f2fs_nfs_get_inode);
2239 static const struct export_operations f2fs_export_ops = {
2240 .fh_to_dentry = f2fs_fh_to_dentry,
2241 .fh_to_parent = f2fs_fh_to_parent,
2242 .get_parent = f2fs_get_parent,
2245 static loff_t max_file_blocks(void)
2248 loff_t leaf_count = ADDRS_PER_BLOCK;
2251 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2252 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2253 * space in inode.i_addr, it will be more safe to reassign
2257 /* two direct node blocks */
2258 result += (leaf_count * 2);
2260 /* two indirect node blocks */
2261 leaf_count *= NIDS_PER_BLOCK;
2262 result += (leaf_count * 2);
2264 /* one double indirect node block */
2265 leaf_count *= NIDS_PER_BLOCK;
2266 result += leaf_count;
2271 static int __f2fs_commit_super(struct buffer_head *bh,
2272 struct f2fs_super_block *super)
2276 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2277 set_buffer_dirty(bh);
2280 /* it's rare case, we can do fua all the time */
2281 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2284 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2285 struct buffer_head *bh)
2287 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2288 (bh->b_data + F2FS_SUPER_OFFSET);
2289 struct super_block *sb = sbi->sb;
2290 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2291 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2292 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2293 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2294 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2295 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2296 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2297 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2298 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2299 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2300 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2301 u32 segment_count = le32_to_cpu(raw_super->segment_count);
2302 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2303 u64 main_end_blkaddr = main_blkaddr +
2304 (segment_count_main << log_blocks_per_seg);
2305 u64 seg_end_blkaddr = segment0_blkaddr +
2306 (segment_count << log_blocks_per_seg);
2308 if (segment0_blkaddr != cp_blkaddr) {
2309 f2fs_msg(sb, KERN_INFO,
2310 "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2311 segment0_blkaddr, cp_blkaddr);
2315 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2317 f2fs_msg(sb, KERN_INFO,
2318 "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2319 cp_blkaddr, sit_blkaddr,
2320 segment_count_ckpt << log_blocks_per_seg);
2324 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2326 f2fs_msg(sb, KERN_INFO,
2327 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2328 sit_blkaddr, nat_blkaddr,
2329 segment_count_sit << log_blocks_per_seg);
2333 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2335 f2fs_msg(sb, KERN_INFO,
2336 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2337 nat_blkaddr, ssa_blkaddr,
2338 segment_count_nat << log_blocks_per_seg);
2342 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2344 f2fs_msg(sb, KERN_INFO,
2345 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2346 ssa_blkaddr, main_blkaddr,
2347 segment_count_ssa << log_blocks_per_seg);
2351 if (main_end_blkaddr > seg_end_blkaddr) {
2352 f2fs_msg(sb, KERN_INFO,
2353 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2356 (segment_count << log_blocks_per_seg),
2357 segment_count_main << log_blocks_per_seg);
2359 } else if (main_end_blkaddr < seg_end_blkaddr) {
2363 /* fix in-memory information all the time */
2364 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2365 segment0_blkaddr) >> log_blocks_per_seg);
2367 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2368 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2371 err = __f2fs_commit_super(bh, NULL);
2372 res = err ? "failed" : "done";
2374 f2fs_msg(sb, KERN_INFO,
2375 "Fix alignment : %s, start(%u) end(%u) block(%u)",
2378 (segment_count << log_blocks_per_seg),
2379 segment_count_main << log_blocks_per_seg);
2386 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2387 struct buffer_head *bh)
2389 block_t segment_count, segs_per_sec, secs_per_zone;
2390 block_t total_sections, blocks_per_seg;
2391 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2392 (bh->b_data + F2FS_SUPER_OFFSET);
2393 struct super_block *sb = sbi->sb;
2394 unsigned int blocksize;
2395 size_t crc_offset = 0;
2398 /* Check checksum_offset and crc in superblock */
2399 if (le32_to_cpu(raw_super->feature) & F2FS_FEATURE_SB_CHKSUM) {
2400 crc_offset = le32_to_cpu(raw_super->checksum_offset);
2402 offsetof(struct f2fs_super_block, crc)) {
2403 f2fs_msg(sb, KERN_INFO,
2404 "Invalid SB checksum offset: %zu",
2408 crc = le32_to_cpu(raw_super->crc);
2409 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2410 f2fs_msg(sb, KERN_INFO,
2411 "Invalid SB checksum value: %u", crc);
2416 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
2417 f2fs_msg(sb, KERN_INFO,
2418 "Magic Mismatch, valid(0x%x) - read(0x%x)",
2419 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2423 /* Currently, support only 4KB page cache size */
2424 if (F2FS_BLKSIZE != PAGE_SIZE) {
2425 f2fs_msg(sb, KERN_INFO,
2426 "Invalid page_cache_size (%lu), supports only 4KB\n",
2431 /* Currently, support only 4KB block size */
2432 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2433 if (blocksize != F2FS_BLKSIZE) {
2434 f2fs_msg(sb, KERN_INFO,
2435 "Invalid blocksize (%u), supports only 4KB\n",
2440 /* check log blocks per segment */
2441 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2442 f2fs_msg(sb, KERN_INFO,
2443 "Invalid log blocks per segment (%u)\n",
2444 le32_to_cpu(raw_super->log_blocks_per_seg));
2448 /* Currently, support 512/1024/2048/4096 bytes sector size */
2449 if (le32_to_cpu(raw_super->log_sectorsize) >
2450 F2FS_MAX_LOG_SECTOR_SIZE ||
2451 le32_to_cpu(raw_super->log_sectorsize) <
2452 F2FS_MIN_LOG_SECTOR_SIZE) {
2453 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
2454 le32_to_cpu(raw_super->log_sectorsize));
2457 if (le32_to_cpu(raw_super->log_sectors_per_block) +
2458 le32_to_cpu(raw_super->log_sectorsize) !=
2459 F2FS_MAX_LOG_SECTOR_SIZE) {
2460 f2fs_msg(sb, KERN_INFO,
2461 "Invalid log sectors per block(%u) log sectorsize(%u)",
2462 le32_to_cpu(raw_super->log_sectors_per_block),
2463 le32_to_cpu(raw_super->log_sectorsize));
2467 segment_count = le32_to_cpu(raw_super->segment_count);
2468 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2469 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2470 total_sections = le32_to_cpu(raw_super->section_count);
2472 /* blocks_per_seg should be 512, given the above check */
2473 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2475 if (segment_count > F2FS_MAX_SEGMENT ||
2476 segment_count < F2FS_MIN_SEGMENTS) {
2477 f2fs_msg(sb, KERN_INFO,
2478 "Invalid segment count (%u)",
2483 if (total_sections > segment_count ||
2484 total_sections < F2FS_MIN_SEGMENTS ||
2485 segs_per_sec > segment_count || !segs_per_sec) {
2486 f2fs_msg(sb, KERN_INFO,
2487 "Invalid segment/section count (%u, %u x %u)",
2488 segment_count, total_sections, segs_per_sec);
2492 if ((segment_count / segs_per_sec) < total_sections) {
2493 f2fs_msg(sb, KERN_INFO,
2494 "Small segment_count (%u < %u * %u)",
2495 segment_count, segs_per_sec, total_sections);
2499 if (segment_count > (le32_to_cpu(raw_super->block_count) >> 9)) {
2500 f2fs_msg(sb, KERN_INFO,
2501 "Wrong segment_count / block_count (%u > %u)",
2502 segment_count, le32_to_cpu(raw_super->block_count));
2506 if (secs_per_zone > total_sections || !secs_per_zone) {
2507 f2fs_msg(sb, KERN_INFO,
2508 "Wrong secs_per_zone / total_sections (%u, %u)",
2509 secs_per_zone, total_sections);
2512 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2513 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2514 (le32_to_cpu(raw_super->extension_count) +
2515 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2516 f2fs_msg(sb, KERN_INFO,
2517 "Corrupted extension count (%u + %u > %u)",
2518 le32_to_cpu(raw_super->extension_count),
2519 raw_super->hot_ext_count,
2520 F2FS_MAX_EXTENSION);
2524 if (le32_to_cpu(raw_super->cp_payload) >
2525 (blocks_per_seg - F2FS_CP_PACKS)) {
2526 f2fs_msg(sb, KERN_INFO,
2527 "Insane cp_payload (%u > %u)",
2528 le32_to_cpu(raw_super->cp_payload),
2529 blocks_per_seg - F2FS_CP_PACKS);
2533 /* check reserved ino info */
2534 if (le32_to_cpu(raw_super->node_ino) != 1 ||
2535 le32_to_cpu(raw_super->meta_ino) != 2 ||
2536 le32_to_cpu(raw_super->root_ino) != 3) {
2537 f2fs_msg(sb, KERN_INFO,
2538 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2539 le32_to_cpu(raw_super->node_ino),
2540 le32_to_cpu(raw_super->meta_ino),
2541 le32_to_cpu(raw_super->root_ino));
2545 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2546 if (sanity_check_area_boundary(sbi, bh))
2552 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2554 unsigned int total, fsmeta;
2555 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2556 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2557 unsigned int ovp_segments, reserved_segments;
2558 unsigned int main_segs, blocks_per_seg;
2559 unsigned int sit_segs, nat_segs;
2560 unsigned int sit_bitmap_size, nat_bitmap_size;
2561 unsigned int log_blocks_per_seg;
2562 unsigned int segment_count_main;
2563 unsigned int cp_pack_start_sum, cp_payload;
2564 block_t user_block_count;
2567 total = le32_to_cpu(raw_super->segment_count);
2568 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2569 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2571 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2573 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2574 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2576 if (unlikely(fsmeta >= total))
2579 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2580 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2582 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2583 ovp_segments == 0 || reserved_segments == 0)) {
2584 f2fs_msg(sbi->sb, KERN_ERR,
2585 "Wrong layout: check mkfs.f2fs version");
2589 user_block_count = le64_to_cpu(ckpt->user_block_count);
2590 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2591 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2592 if (!user_block_count || user_block_count >=
2593 segment_count_main << log_blocks_per_seg) {
2594 f2fs_msg(sbi->sb, KERN_ERR,
2595 "Wrong user_block_count: %u", user_block_count);
2599 main_segs = le32_to_cpu(raw_super->segment_count_main);
2600 blocks_per_seg = sbi->blocks_per_seg;
2602 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2603 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2604 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2606 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2607 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2608 le32_to_cpu(ckpt->cur_node_segno[j])) {
2609 f2fs_msg(sbi->sb, KERN_ERR,
2610 "Node segment (%u, %u) has the same "
2612 le32_to_cpu(ckpt->cur_node_segno[i]));
2617 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2618 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2619 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2621 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2622 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2623 le32_to_cpu(ckpt->cur_data_segno[j])) {
2624 f2fs_msg(sbi->sb, KERN_ERR,
2625 "Data segment (%u, %u) has the same "
2627 le32_to_cpu(ckpt->cur_data_segno[i]));
2632 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2633 for (j = i; j < NR_CURSEG_DATA_TYPE; j++) {
2634 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2635 le32_to_cpu(ckpt->cur_data_segno[j])) {
2636 f2fs_msg(sbi->sb, KERN_ERR,
2637 "Data segment (%u) and Data segment (%u)"
2638 " has the same segno: %u", i, j,
2639 le32_to_cpu(ckpt->cur_node_segno[i]));
2645 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2646 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2648 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2649 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2650 f2fs_msg(sbi->sb, KERN_ERR,
2651 "Wrong bitmap size: sit: %u, nat:%u",
2652 sit_bitmap_size, nat_bitmap_size);
2656 cp_pack_start_sum = __start_sum_addr(sbi);
2657 cp_payload = __cp_payload(sbi);
2658 if (cp_pack_start_sum < cp_payload + 1 ||
2659 cp_pack_start_sum > blocks_per_seg - 1 -
2661 f2fs_msg(sbi->sb, KERN_ERR,
2662 "Wrong cp_pack_start_sum: %u",
2667 if (unlikely(f2fs_cp_error(sbi))) {
2668 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
2674 static void init_sb_info(struct f2fs_sb_info *sbi)
2676 struct f2fs_super_block *raw_super = sbi->raw_super;
2679 sbi->log_sectors_per_block =
2680 le32_to_cpu(raw_super->log_sectors_per_block);
2681 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2682 sbi->blocksize = 1 << sbi->log_blocksize;
2683 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2684 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2685 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2686 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2687 sbi->total_sections = le32_to_cpu(raw_super->section_count);
2688 sbi->total_node_count =
2689 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2690 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2691 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2692 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2693 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2694 sbi->cur_victim_sec = NULL_SECNO;
2695 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2697 sbi->dir_level = DEF_DIR_LEVEL;
2698 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2699 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2700 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
2701 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
2702 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
2703 clear_sbi_flag(sbi, SBI_NEED_FSCK);
2705 for (i = 0; i < NR_COUNT_TYPE; i++)
2706 atomic_set(&sbi->nr_pages[i], 0);
2708 for (i = 0; i < META; i++)
2709 atomic_set(&sbi->wb_sync_req[i], 0);
2711 INIT_LIST_HEAD(&sbi->s_list);
2712 mutex_init(&sbi->umount_mutex);
2713 for (i = 0; i < NR_PAGE_TYPE - 1; i++)
2714 for (j = HOT; j < NR_TEMP_TYPE; j++)
2715 mutex_init(&sbi->wio_mutex[i][j]);
2716 init_rwsem(&sbi->io_order_lock);
2717 spin_lock_init(&sbi->cp_lock);
2719 sbi->dirty_device = 0;
2720 spin_lock_init(&sbi->dev_lock);
2722 init_rwsem(&sbi->sb_lock);
2725 static int init_percpu_info(struct f2fs_sb_info *sbi)
2729 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2733 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
2736 percpu_counter_destroy(&sbi->alloc_valid_block_count);
2741 #ifdef CONFIG_BLK_DEV_ZONED
2742 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2744 struct block_device *bdev = FDEV(devi).bdev;
2745 sector_t nr_sectors = bdev->bd_part->nr_sects;
2746 sector_t sector = 0;
2747 struct blk_zone *zones;
2748 unsigned int i, nr_zones;
2752 if (!f2fs_sb_has_blkzoned(sbi->sb))
2755 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2756 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2758 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2759 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2760 __ilog2_u32(sbi->blocks_per_blkz))
2762 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2763 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2764 sbi->log_blocks_per_blkz;
2765 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2766 FDEV(devi).nr_blkz++;
2768 FDEV(devi).blkz_type = f2fs_kmalloc(sbi, FDEV(devi).nr_blkz,
2770 if (!FDEV(devi).blkz_type)
2773 #define F2FS_REPORT_NR_ZONES 4096
2775 zones = f2fs_kzalloc(sbi,
2776 array_size(F2FS_REPORT_NR_ZONES,
2777 sizeof(struct blk_zone)),
2782 /* Get block zones type */
2783 while (zones && sector < nr_sectors) {
2785 nr_zones = F2FS_REPORT_NR_ZONES;
2786 err = blkdev_report_zones(bdev, sector,
2796 for (i = 0; i < nr_zones; i++) {
2797 FDEV(devi).blkz_type[n] = zones[i].type;
2798 sector += zones[i].len;
2810 * Read f2fs raw super block.
2811 * Because we have two copies of super block, so read both of them
2812 * to get the first valid one. If any one of them is broken, we pass
2813 * them recovery flag back to the caller.
2815 static int read_raw_super_block(struct f2fs_sb_info *sbi,
2816 struct f2fs_super_block **raw_super,
2817 int *valid_super_block, int *recovery)
2819 struct super_block *sb = sbi->sb;
2821 struct buffer_head *bh;
2822 struct f2fs_super_block *super;
2825 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2829 for (block = 0; block < 2; block++) {
2830 bh = sb_bread(sb, block);
2832 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
2838 /* sanity checking of raw super */
2839 if (sanity_check_raw_super(sbi, bh)) {
2840 f2fs_msg(sb, KERN_ERR,
2841 "Can't find valid F2FS filesystem in %dth superblock",
2849 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2851 *valid_super_block = block;
2857 /* Fail to read any one of the superblocks*/
2861 /* No valid superblock */
2870 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
2872 struct buffer_head *bh;
2876 if ((recover && f2fs_readonly(sbi->sb)) ||
2877 bdev_read_only(sbi->sb->s_bdev)) {
2878 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2882 /* we should update superblock crc here */
2883 if (!recover && f2fs_sb_has_sb_chksum(sbi->sb)) {
2884 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
2885 offsetof(struct f2fs_super_block, crc));
2886 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
2889 /* write back-up superblock first */
2890 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
2893 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2896 /* if we are in recovery path, skip writing valid superblock */
2900 /* write current valid superblock */
2901 bh = sb_bread(sbi->sb, sbi->valid_super_block);
2904 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2909 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
2911 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2912 unsigned int max_devices = MAX_DEVICES;
2915 /* Initialize single device information */
2916 if (!RDEV(0).path[0]) {
2917 if (!bdev_is_zoned(sbi->sb->s_bdev))
2923 * Initialize multiple devices information, or single
2924 * zoned block device information.
2926 sbi->devs = f2fs_kzalloc(sbi,
2927 array_size(max_devices,
2928 sizeof(struct f2fs_dev_info)),
2933 for (i = 0; i < max_devices; i++) {
2935 if (i > 0 && !RDEV(i).path[0])
2938 if (max_devices == 1) {
2939 /* Single zoned block device mount */
2941 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
2942 sbi->sb->s_mode, sbi->sb->s_type);
2944 /* Multi-device mount */
2945 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
2946 FDEV(i).total_segments =
2947 le32_to_cpu(RDEV(i).total_segments);
2949 FDEV(i).start_blk = 0;
2950 FDEV(i).end_blk = FDEV(i).start_blk +
2951 (FDEV(i).total_segments <<
2952 sbi->log_blocks_per_seg) - 1 +
2953 le32_to_cpu(raw_super->segment0_blkaddr);
2955 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
2956 FDEV(i).end_blk = FDEV(i).start_blk +
2957 (FDEV(i).total_segments <<
2958 sbi->log_blocks_per_seg) - 1;
2960 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
2961 sbi->sb->s_mode, sbi->sb->s_type);
2963 if (IS_ERR(FDEV(i).bdev))
2964 return PTR_ERR(FDEV(i).bdev);
2966 /* to release errored devices */
2967 sbi->s_ndevs = i + 1;
2969 #ifdef CONFIG_BLK_DEV_ZONED
2970 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
2971 !f2fs_sb_has_blkzoned(sbi->sb)) {
2972 f2fs_msg(sbi->sb, KERN_ERR,
2973 "Zoned block device feature not enabled\n");
2976 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
2977 if (init_blkz_info(sbi, i)) {
2978 f2fs_msg(sbi->sb, KERN_ERR,
2979 "Failed to initialize F2FS blkzone information");
2982 if (max_devices == 1)
2984 f2fs_msg(sbi->sb, KERN_INFO,
2985 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
2987 FDEV(i).total_segments,
2988 FDEV(i).start_blk, FDEV(i).end_blk,
2989 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
2990 "Host-aware" : "Host-managed");
2994 f2fs_msg(sbi->sb, KERN_INFO,
2995 "Mount Device [%2d]: %20s, %8u, %8x - %8x",
2997 FDEV(i).total_segments,
2998 FDEV(i).start_blk, FDEV(i).end_blk);
3000 f2fs_msg(sbi->sb, KERN_INFO,
3001 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3005 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3007 struct f2fs_sm_info *sm_i = SM_I(sbi);
3009 /* adjust parameters according to the volume size */
3010 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3011 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3012 sm_i->dcc_info->discard_granularity = 1;
3013 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3016 sbi->readdir_ra = 1;
3019 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3021 struct f2fs_sb_info *sbi;
3022 struct f2fs_super_block *raw_super;
3025 bool retry = true, need_fsck = false;
3026 char *options = NULL;
3027 int recovery, i, valid_super_block;
3028 struct curseg_info *seg_i;
3033 valid_super_block = -1;
3036 /* allocate memory for f2fs-specific super block info */
3037 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3043 /* Load the checksum driver */
3044 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3045 if (IS_ERR(sbi->s_chksum_driver)) {
3046 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
3047 err = PTR_ERR(sbi->s_chksum_driver);
3048 sbi->s_chksum_driver = NULL;
3052 /* set a block size */
3053 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3054 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
3058 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3063 sb->s_fs_info = sbi;
3064 sbi->raw_super = raw_super;
3066 /* precompute checksum seed for metadata */
3067 if (f2fs_sb_has_inode_chksum(sb))
3068 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3069 sizeof(raw_super->uuid));
3072 * The BLKZONED feature indicates that the drive was formatted with
3073 * zone alignment optimization. This is optional for host-aware
3074 * devices, but mandatory for host-managed zoned block devices.
3076 #ifndef CONFIG_BLK_DEV_ZONED
3077 if (f2fs_sb_has_blkzoned(sb)) {
3078 f2fs_msg(sb, KERN_ERR,
3079 "Zoned block device support is not enabled\n");
3084 default_options(sbi);
3085 /* parse mount options */
3086 options = kstrdup((const char *)data, GFP_KERNEL);
3087 if (data && !options) {
3092 err = parse_options(sb, options);
3096 sbi->max_file_blocks = max_file_blocks();
3097 sb->s_maxbytes = sbi->max_file_blocks <<
3098 le32_to_cpu(raw_super->log_blocksize);
3099 sb->s_max_links = F2FS_LINK_MAX;
3100 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3103 sb->dq_op = &f2fs_quota_operations;
3104 if (f2fs_sb_has_quota_ino(sb))
3105 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3107 sb->s_qcop = &f2fs_quotactl_ops;
3108 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3110 if (f2fs_sb_has_quota_ino(sbi->sb)) {
3111 for (i = 0; i < MAXQUOTAS; i++) {
3112 if (f2fs_qf_ino(sbi->sb, i))
3113 sbi->nquota_files++;
3118 sb->s_op = &f2fs_sops;
3119 #ifdef CONFIG_F2FS_FS_ENCRYPTION
3120 sb->s_cop = &f2fs_cryptops;
3122 sb->s_xattr = f2fs_xattr_handlers;
3123 sb->s_export_op = &f2fs_export_ops;
3124 sb->s_magic = F2FS_SUPER_MAGIC;
3125 sb->s_time_gran = 1;
3126 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3127 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3128 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3129 sb->s_iflags |= SB_I_CGROUPWB;
3131 /* init f2fs-specific super block info */
3132 sbi->valid_super_block = valid_super_block;
3133 mutex_init(&sbi->gc_mutex);
3134 mutex_init(&sbi->writepages);
3135 mutex_init(&sbi->cp_mutex);
3136 init_rwsem(&sbi->node_write);
3137 init_rwsem(&sbi->node_change);
3139 /* disallow all the data/node/meta page writes */
3140 set_sbi_flag(sbi, SBI_POR_DOING);
3141 spin_lock_init(&sbi->stat_lock);
3143 /* init iostat info */
3144 spin_lock_init(&sbi->iostat_lock);
3145 sbi->iostat_enable = false;
3147 for (i = 0; i < NR_PAGE_TYPE; i++) {
3148 int n = (i == META) ? 1: NR_TEMP_TYPE;
3154 sizeof(struct f2fs_bio_info)),
3156 if (!sbi->write_io[i]) {
3161 for (j = HOT; j < n; j++) {
3162 init_rwsem(&sbi->write_io[i][j].io_rwsem);
3163 sbi->write_io[i][j].sbi = sbi;
3164 sbi->write_io[i][j].bio = NULL;
3165 spin_lock_init(&sbi->write_io[i][j].io_lock);
3166 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3170 init_rwsem(&sbi->cp_rwsem);
3171 init_waitqueue_head(&sbi->cp_wait);
3174 err = init_percpu_info(sbi);
3178 if (F2FS_IO_SIZE(sbi) > 1) {
3179 sbi->write_io_dummy =
3180 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3181 if (!sbi->write_io_dummy) {
3187 /* get an inode for meta space */
3188 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3189 if (IS_ERR(sbi->meta_inode)) {
3190 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
3191 err = PTR_ERR(sbi->meta_inode);
3195 err = f2fs_get_valid_checkpoint(sbi);
3197 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
3198 goto free_meta_inode;
3201 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3202 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3204 /* Initialize device list */
3205 err = f2fs_scan_devices(sbi);
3207 f2fs_msg(sb, KERN_ERR, "Failed to find devices");
3211 sbi->total_valid_node_count =
3212 le32_to_cpu(sbi->ckpt->valid_node_count);
3213 percpu_counter_set(&sbi->total_valid_inode_count,
3214 le32_to_cpu(sbi->ckpt->valid_inode_count));
3215 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3216 sbi->total_valid_block_count =
3217 le64_to_cpu(sbi->ckpt->valid_block_count);
3218 sbi->last_valid_block_count = sbi->total_valid_block_count;
3219 sbi->reserved_blocks = 0;
3220 sbi->current_reserved_blocks = 0;
3221 limit_reserve_root(sbi);
3223 for (i = 0; i < NR_INODE_TYPE; i++) {
3224 INIT_LIST_HEAD(&sbi->inode_list[i]);
3225 spin_lock_init(&sbi->inode_lock[i]);
3228 f2fs_init_extent_cache_info(sbi);
3230 f2fs_init_ino_entry_info(sbi);
3232 f2fs_init_fsync_node_info(sbi);
3234 /* setup f2fs internal modules */
3235 err = f2fs_build_segment_manager(sbi);
3237 f2fs_msg(sb, KERN_ERR,
3238 "Failed to initialize F2FS segment manager");
3241 err = f2fs_build_node_manager(sbi);
3243 f2fs_msg(sb, KERN_ERR,
3244 "Failed to initialize F2FS node manager");
3248 /* For write statistics */
3249 if (sb->s_bdev->bd_part)
3250 sbi->sectors_written_start =
3251 (u64)part_stat_read(sb->s_bdev->bd_part,
3252 sectors[STAT_WRITE]);
3254 /* Read accumulated write IO statistics if exists */
3255 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3256 if (__exist_node_summaries(sbi))
3257 sbi->kbytes_written =
3258 le64_to_cpu(seg_i->journal->info.kbytes_written);
3260 f2fs_build_gc_manager(sbi);
3262 /* get an inode for node space */
3263 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3264 if (IS_ERR(sbi->node_inode)) {
3265 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
3266 err = PTR_ERR(sbi->node_inode);
3270 err = f2fs_build_stats(sbi);
3272 goto free_node_inode;
3274 /* read root inode and dentry */
3275 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3277 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
3278 err = PTR_ERR(root);
3281 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3282 !root->i_size || !root->i_nlink) {
3288 sb->s_root = d_make_root(root); /* allocate root dentry */
3291 goto free_root_inode;
3294 err = f2fs_register_sysfs(sbi);
3296 goto free_root_inode;
3299 /* Enable quota usage during mount */
3300 if (f2fs_sb_has_quota_ino(sb) && !f2fs_readonly(sb)) {
3301 err = f2fs_enable_quotas(sb);
3303 f2fs_msg(sb, KERN_ERR,
3304 "Cannot turn on quotas: error %d", err);
3307 /* if there are nt orphan nodes free them */
3308 err = f2fs_recover_orphan_inodes(sbi);
3312 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3315 /* recover fsynced data */
3316 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
3318 * mount should be failed, when device has readonly mode, and
3319 * previous checkpoint was not done by clean system shutdown.
3321 if (bdev_read_only(sb->s_bdev) &&
3322 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3328 set_sbi_flag(sbi, SBI_NEED_FSCK);
3333 err = f2fs_recover_fsync_data(sbi, false);
3336 f2fs_msg(sb, KERN_ERR,
3337 "Cannot recover all fsync data errno=%d", err);
3341 err = f2fs_recover_fsync_data(sbi, true);
3343 if (!f2fs_readonly(sb) && err > 0) {
3345 f2fs_msg(sb, KERN_ERR,
3346 "Need to recover fsync data");
3351 /* f2fs_recover_fsync_data() cleared this already */
3352 clear_sbi_flag(sbi, SBI_POR_DOING);
3354 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3355 err = f2fs_disable_checkpoint(sbi);
3358 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3359 f2fs_enable_checkpoint(sbi);
3363 * If filesystem is not mounted as read-only then
3364 * do start the gc_thread.
3366 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
3367 /* After POR, we can run background GC thread.*/
3368 err = f2fs_start_gc_thread(sbi);
3374 /* recover broken superblock */
3376 err = f2fs_commit_super(sbi, true);
3377 f2fs_msg(sb, KERN_INFO,
3378 "Try to recover %dth superblock, ret: %d",
3379 sbi->valid_super_block ? 1 : 2, err);
3382 f2fs_join_shrinker(sbi);
3384 f2fs_tuning_parameters(sbi);
3386 f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx",
3387 cur_cp_version(F2FS_CKPT(sbi)));
3388 f2fs_update_time(sbi, CP_TIME);
3389 f2fs_update_time(sbi, REQ_TIME);
3394 f2fs_truncate_quota_inode_pages(sb);
3395 if (f2fs_sb_has_quota_ino(sb) && !f2fs_readonly(sb))
3396 f2fs_quota_off_umount(sbi->sb);
3399 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3400 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3401 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3402 * falls into an infinite loop in f2fs_sync_meta_pages().
3404 truncate_inode_pages_final(META_MAPPING(sbi));
3405 f2fs_unregister_sysfs(sbi);
3410 f2fs_destroy_stats(sbi);
3412 f2fs_release_ino_entry(sbi, true);
3413 truncate_inode_pages_final(NODE_MAPPING(sbi));
3414 iput(sbi->node_inode);
3416 f2fs_destroy_node_manager(sbi);
3418 f2fs_destroy_segment_manager(sbi);
3420 destroy_device_list(sbi);
3423 make_bad_inode(sbi->meta_inode);
3424 iput(sbi->meta_inode);
3426 mempool_destroy(sbi->write_io_dummy);
3428 destroy_percpu_info(sbi);
3430 for (i = 0; i < NR_PAGE_TYPE; i++)
3431 kfree(sbi->write_io[i]);
3434 for (i = 0; i < MAXQUOTAS; i++)
3435 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
3441 if (sbi->s_chksum_driver)
3442 crypto_free_shash(sbi->s_chksum_driver);
3445 /* give only one another chance */
3448 shrink_dcache_sb(sb);
3454 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3455 const char *dev_name, void *data)
3457 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3460 static void kill_f2fs_super(struct super_block *sb)
3463 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3465 set_sbi_flag(sbi, SBI_IS_CLOSE);
3466 f2fs_stop_gc_thread(sbi);
3467 f2fs_stop_discard_thread(sbi);
3469 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3470 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3471 struct cp_control cpc = {
3472 .reason = CP_UMOUNT,
3474 f2fs_write_checkpoint(sbi, &cpc);
3477 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3478 sb->s_flags &= ~SB_RDONLY;
3480 kill_block_super(sb);
3483 static struct file_system_type f2fs_fs_type = {
3484 .owner = THIS_MODULE,
3486 .mount = f2fs_mount,
3487 .kill_sb = kill_f2fs_super,
3488 .fs_flags = FS_REQUIRES_DEV,
3490 MODULE_ALIAS_FS("f2fs");
3492 static int __init init_inodecache(void)
3494 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3495 sizeof(struct f2fs_inode_info), 0,
3496 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3497 if (!f2fs_inode_cachep)
3502 static void destroy_inodecache(void)
3505 * Make sure all delayed rcu free inodes are flushed before we
3509 kmem_cache_destroy(f2fs_inode_cachep);
3512 static int __init init_f2fs_fs(void)
3516 if (PAGE_SIZE != F2FS_BLKSIZE) {
3517 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3518 PAGE_SIZE, F2FS_BLKSIZE);
3522 f2fs_build_trace_ios();
3524 err = init_inodecache();
3527 err = f2fs_create_node_manager_caches();
3529 goto free_inodecache;
3530 err = f2fs_create_segment_manager_caches();
3532 goto free_node_manager_caches;
3533 err = f2fs_create_checkpoint_caches();
3535 goto free_segment_manager_caches;
3536 err = f2fs_create_extent_cache();
3538 goto free_checkpoint_caches;
3539 err = f2fs_init_sysfs();
3541 goto free_extent_cache;
3542 err = register_shrinker(&f2fs_shrinker_info);
3545 err = register_filesystem(&f2fs_fs_type);
3548 err = f2fs_create_root_stats();
3550 goto free_filesystem;
3551 err = f2fs_init_post_read_processing();
3553 goto free_root_stats;
3557 f2fs_destroy_root_stats();
3559 unregister_filesystem(&f2fs_fs_type);
3561 unregister_shrinker(&f2fs_shrinker_info);
3565 f2fs_destroy_extent_cache();
3566 free_checkpoint_caches:
3567 f2fs_destroy_checkpoint_caches();
3568 free_segment_manager_caches:
3569 f2fs_destroy_segment_manager_caches();
3570 free_node_manager_caches:
3571 f2fs_destroy_node_manager_caches();
3573 destroy_inodecache();
3578 static void __exit exit_f2fs_fs(void)
3580 f2fs_destroy_post_read_processing();
3581 f2fs_destroy_root_stats();
3582 unregister_filesystem(&f2fs_fs_type);
3583 unregister_shrinker(&f2fs_shrinker_info);
3585 f2fs_destroy_extent_cache();
3586 f2fs_destroy_checkpoint_caches();
3587 f2fs_destroy_segment_manager_caches();
3588 f2fs_destroy_node_manager_caches();
3589 destroy_inodecache();
3590 f2fs_destroy_trace_ios();
3593 module_init(init_f2fs_fs)
3594 module_exit(exit_f2fs_fs)
3596 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3597 MODULE_DESCRIPTION("Flash Friendly File System");
3598 MODULE_LICENSE("GPL");