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 const 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(sbi)) {
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 kvfree(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)) {
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) && 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(sbi)) {
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(sbi)) {
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 <= 0 || 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_FS_ENCRYPTION
761 if (!f2fs_sb_has_encrypt(sbi)) {
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) && !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) && !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 int min_size, max_size;
826 if (!f2fs_sb_has_extra_attr(sbi) ||
827 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
828 f2fs_msg(sb, KERN_ERR,
829 "extra_attr or flexible_inline_xattr "
833 if (!test_opt(sbi, INLINE_XATTR)) {
834 f2fs_msg(sb, KERN_ERR,
835 "inline_xattr_size option should be "
836 "set with inline_xattr option");
840 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
841 max_size = MAX_INLINE_XATTR_SIZE;
843 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
844 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
845 f2fs_msg(sb, KERN_ERR,
846 "inline xattr size is out of range: %d ~ %d",
852 if (test_opt(sbi, DISABLE_CHECKPOINT) && test_opt(sbi, LFS)) {
853 f2fs_msg(sb, KERN_ERR,
854 "LFS not compatible with checkpoint=disable\n");
858 /* Not pass down write hints if the number of active logs is lesser
859 * than NR_CURSEG_TYPE.
861 if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
862 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
866 static struct inode *f2fs_alloc_inode(struct super_block *sb)
868 struct f2fs_inode_info *fi;
870 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
874 init_once((void *) fi);
876 /* Initialize f2fs-specific inode info */
877 atomic_set(&fi->dirty_pages, 0);
878 init_rwsem(&fi->i_sem);
879 INIT_LIST_HEAD(&fi->dirty_list);
880 INIT_LIST_HEAD(&fi->gdirty_list);
881 INIT_LIST_HEAD(&fi->inmem_ilist);
882 INIT_LIST_HEAD(&fi->inmem_pages);
883 mutex_init(&fi->inmem_lock);
884 init_rwsem(&fi->i_gc_rwsem[READ]);
885 init_rwsem(&fi->i_gc_rwsem[WRITE]);
886 init_rwsem(&fi->i_mmap_sem);
887 init_rwsem(&fi->i_xattr_sem);
889 /* Will be used by directory only */
890 fi->i_dir_level = F2FS_SB(sb)->dir_level;
892 return &fi->vfs_inode;
895 static int f2fs_drop_inode(struct inode *inode)
899 * This is to avoid a deadlock condition like below.
900 * writeback_single_inode(inode)
901 * - f2fs_write_data_page
902 * - f2fs_gc -> iput -> evict
903 * - inode_wait_for_writeback(inode)
905 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
906 if (!inode->i_nlink && !is_bad_inode(inode)) {
907 /* to avoid evict_inode call simultaneously */
908 atomic_inc(&inode->i_count);
909 spin_unlock(&inode->i_lock);
911 /* some remained atomic pages should discarded */
912 if (f2fs_is_atomic_file(inode))
913 f2fs_drop_inmem_pages(inode);
915 /* should remain fi->extent_tree for writepage */
916 f2fs_destroy_extent_node(inode);
918 sb_start_intwrite(inode->i_sb);
919 f2fs_i_size_write(inode, 0);
921 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
922 inode, NULL, 0, DATA);
923 truncate_inode_pages_final(inode->i_mapping);
925 if (F2FS_HAS_BLOCKS(inode))
926 f2fs_truncate(inode);
928 sb_end_intwrite(inode->i_sb);
930 spin_lock(&inode->i_lock);
931 atomic_dec(&inode->i_count);
933 trace_f2fs_drop_inode(inode, 0);
936 ret = generic_drop_inode(inode);
937 trace_f2fs_drop_inode(inode, ret);
941 int f2fs_inode_dirtied(struct inode *inode, bool sync)
943 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
946 spin_lock(&sbi->inode_lock[DIRTY_META]);
947 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
950 set_inode_flag(inode, FI_DIRTY_INODE);
951 stat_inc_dirty_inode(sbi, DIRTY_META);
953 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
954 list_add_tail(&F2FS_I(inode)->gdirty_list,
955 &sbi->inode_list[DIRTY_META]);
956 inc_page_count(sbi, F2FS_DIRTY_IMETA);
958 spin_unlock(&sbi->inode_lock[DIRTY_META]);
962 void f2fs_inode_synced(struct inode *inode)
964 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
966 spin_lock(&sbi->inode_lock[DIRTY_META]);
967 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
968 spin_unlock(&sbi->inode_lock[DIRTY_META]);
971 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
972 list_del_init(&F2FS_I(inode)->gdirty_list);
973 dec_page_count(sbi, F2FS_DIRTY_IMETA);
975 clear_inode_flag(inode, FI_DIRTY_INODE);
976 clear_inode_flag(inode, FI_AUTO_RECOVER);
977 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
978 spin_unlock(&sbi->inode_lock[DIRTY_META]);
982 * f2fs_dirty_inode() is called from __mark_inode_dirty()
984 * We should call set_dirty_inode to write the dirty inode through write_inode.
986 static void f2fs_dirty_inode(struct inode *inode, int flags)
988 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
990 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
991 inode->i_ino == F2FS_META_INO(sbi))
994 if (flags == I_DIRTY_TIME)
997 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
998 clear_inode_flag(inode, FI_AUTO_RECOVER);
1000 f2fs_inode_dirtied(inode, false);
1003 static void f2fs_free_inode(struct inode *inode)
1005 fscrypt_free_inode(inode);
1006 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1009 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1011 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1012 percpu_counter_destroy(&sbi->total_valid_inode_count);
1015 static void destroy_device_list(struct f2fs_sb_info *sbi)
1019 for (i = 0; i < sbi->s_ndevs; i++) {
1020 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1021 #ifdef CONFIG_BLK_DEV_ZONED
1022 kvfree(FDEV(i).blkz_seq);
1028 static void f2fs_put_super(struct super_block *sb)
1030 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1034 f2fs_quota_off_umount(sb);
1036 /* prevent remaining shrinker jobs */
1037 mutex_lock(&sbi->umount_mutex);
1040 * We don't need to do checkpoint when superblock is clean.
1041 * But, the previous checkpoint was not done by umount, it needs to do
1042 * clean checkpoint again.
1044 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1045 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1046 struct cp_control cpc = {
1047 .reason = CP_UMOUNT,
1049 f2fs_write_checkpoint(sbi, &cpc);
1052 /* be sure to wait for any on-going discard commands */
1053 dropped = f2fs_issue_discard_timeout(sbi);
1055 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1056 !sbi->discard_blks && !dropped) {
1057 struct cp_control cpc = {
1058 .reason = CP_UMOUNT | CP_TRIMMED,
1060 f2fs_write_checkpoint(sbi, &cpc);
1064 * normally superblock is clean, so we need to release this.
1065 * In addition, EIO will skip do checkpoint, we need this as well.
1067 f2fs_release_ino_entry(sbi, true);
1069 f2fs_leave_shrinker(sbi);
1070 mutex_unlock(&sbi->umount_mutex);
1072 /* our cp_error case, we can wait for any writeback page */
1073 f2fs_flush_merged_writes(sbi);
1075 f2fs_wait_on_all_pages_writeback(sbi);
1077 f2fs_bug_on(sbi, sbi->fsync_node_num);
1079 iput(sbi->node_inode);
1080 sbi->node_inode = NULL;
1082 iput(sbi->meta_inode);
1083 sbi->meta_inode = NULL;
1086 * iput() can update stat information, if f2fs_write_checkpoint()
1087 * above failed with error.
1089 f2fs_destroy_stats(sbi);
1091 /* destroy f2fs internal modules */
1092 f2fs_destroy_node_manager(sbi);
1093 f2fs_destroy_segment_manager(sbi);
1097 f2fs_unregister_sysfs(sbi);
1099 sb->s_fs_info = NULL;
1100 if (sbi->s_chksum_driver)
1101 crypto_free_shash(sbi->s_chksum_driver);
1102 kvfree(sbi->raw_super);
1104 destroy_device_list(sbi);
1105 mempool_destroy(sbi->write_io_dummy);
1107 for (i = 0; i < MAXQUOTAS; i++)
1108 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1110 destroy_percpu_info(sbi);
1111 for (i = 0; i < NR_PAGE_TYPE; i++)
1112 kvfree(sbi->write_io[i]);
1116 int f2fs_sync_fs(struct super_block *sb, int sync)
1118 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1121 if (unlikely(f2fs_cp_error(sbi)))
1123 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1126 trace_f2fs_sync_fs(sb, sync);
1128 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1132 struct cp_control cpc;
1134 cpc.reason = __get_cp_reason(sbi);
1136 mutex_lock(&sbi->gc_mutex);
1137 err = f2fs_write_checkpoint(sbi, &cpc);
1138 mutex_unlock(&sbi->gc_mutex);
1140 f2fs_trace_ios(NULL, 1);
1145 static int f2fs_freeze(struct super_block *sb)
1147 if (f2fs_readonly(sb))
1150 /* IO error happened before */
1151 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1154 /* must be clean, since sync_filesystem() was already called */
1155 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1160 static int f2fs_unfreeze(struct super_block *sb)
1166 static int f2fs_statfs_project(struct super_block *sb,
1167 kprojid_t projid, struct kstatfs *buf)
1170 struct dquot *dquot;
1174 qid = make_kqid_projid(projid);
1175 dquot = dqget(sb, qid);
1177 return PTR_ERR(dquot);
1178 spin_lock(&dquot->dq_dqb_lock);
1180 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
1181 dquot->dq_dqb.dqb_bsoftlimit :
1182 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
1183 if (limit && buf->f_blocks > limit) {
1184 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1185 buf->f_blocks = limit;
1186 buf->f_bfree = buf->f_bavail =
1187 (buf->f_blocks > curblock) ?
1188 (buf->f_blocks - curblock) : 0;
1191 limit = dquot->dq_dqb.dqb_isoftlimit ?
1192 dquot->dq_dqb.dqb_isoftlimit :
1193 dquot->dq_dqb.dqb_ihardlimit;
1194 if (limit && buf->f_files > limit) {
1195 buf->f_files = limit;
1197 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1198 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1201 spin_unlock(&dquot->dq_dqb_lock);
1207 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1209 struct super_block *sb = dentry->d_sb;
1210 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1211 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1212 block_t total_count, user_block_count, start_count;
1213 u64 avail_node_count;
1215 total_count = le64_to_cpu(sbi->raw_super->block_count);
1216 user_block_count = sbi->user_block_count;
1217 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1218 buf->f_type = F2FS_SUPER_MAGIC;
1219 buf->f_bsize = sbi->blocksize;
1221 buf->f_blocks = total_count - start_count;
1222 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1223 sbi->current_reserved_blocks;
1225 spin_lock(&sbi->stat_lock);
1226 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1229 buf->f_bfree -= sbi->unusable_block_count;
1230 spin_unlock(&sbi->stat_lock);
1232 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1233 buf->f_bavail = buf->f_bfree -
1234 F2FS_OPTION(sbi).root_reserved_blocks;
1238 avail_node_count = sbi->total_node_count - sbi->nquota_files -
1239 F2FS_RESERVED_NODE_NUM;
1241 if (avail_node_count > user_block_count) {
1242 buf->f_files = user_block_count;
1243 buf->f_ffree = buf->f_bavail;
1245 buf->f_files = avail_node_count;
1246 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1250 buf->f_namelen = F2FS_NAME_LEN;
1251 buf->f_fsid.val[0] = (u32)id;
1252 buf->f_fsid.val[1] = (u32)(id >> 32);
1255 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1256 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1257 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1263 static inline void f2fs_show_quota_options(struct seq_file *seq,
1264 struct super_block *sb)
1267 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1269 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1272 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1283 seq_printf(seq, ",jqfmt=%s", fmtname);
1286 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1287 seq_show_option(seq, "usrjquota",
1288 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1290 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1291 seq_show_option(seq, "grpjquota",
1292 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1294 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1295 seq_show_option(seq, "prjjquota",
1296 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1300 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1302 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1304 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1305 if (test_opt(sbi, FORCE_FG_GC))
1306 seq_printf(seq, ",background_gc=%s", "sync");
1308 seq_printf(seq, ",background_gc=%s", "on");
1310 seq_printf(seq, ",background_gc=%s", "off");
1312 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1313 seq_puts(seq, ",disable_roll_forward");
1314 if (test_opt(sbi, DISCARD))
1315 seq_puts(seq, ",discard");
1316 if (test_opt(sbi, NOHEAP))
1317 seq_puts(seq, ",no_heap");
1319 seq_puts(seq, ",heap");
1320 #ifdef CONFIG_F2FS_FS_XATTR
1321 if (test_opt(sbi, XATTR_USER))
1322 seq_puts(seq, ",user_xattr");
1324 seq_puts(seq, ",nouser_xattr");
1325 if (test_opt(sbi, INLINE_XATTR))
1326 seq_puts(seq, ",inline_xattr");
1328 seq_puts(seq, ",noinline_xattr");
1329 if (test_opt(sbi, INLINE_XATTR_SIZE))
1330 seq_printf(seq, ",inline_xattr_size=%u",
1331 F2FS_OPTION(sbi).inline_xattr_size);
1333 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1334 if (test_opt(sbi, POSIX_ACL))
1335 seq_puts(seq, ",acl");
1337 seq_puts(seq, ",noacl");
1339 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1340 seq_puts(seq, ",disable_ext_identify");
1341 if (test_opt(sbi, INLINE_DATA))
1342 seq_puts(seq, ",inline_data");
1344 seq_puts(seq, ",noinline_data");
1345 if (test_opt(sbi, INLINE_DENTRY))
1346 seq_puts(seq, ",inline_dentry");
1348 seq_puts(seq, ",noinline_dentry");
1349 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1350 seq_puts(seq, ",flush_merge");
1351 if (test_opt(sbi, NOBARRIER))
1352 seq_puts(seq, ",nobarrier");
1353 if (test_opt(sbi, FASTBOOT))
1354 seq_puts(seq, ",fastboot");
1355 if (test_opt(sbi, EXTENT_CACHE))
1356 seq_puts(seq, ",extent_cache");
1358 seq_puts(seq, ",noextent_cache");
1359 if (test_opt(sbi, DATA_FLUSH))
1360 seq_puts(seq, ",data_flush");
1362 seq_puts(seq, ",mode=");
1363 if (test_opt(sbi, ADAPTIVE))
1364 seq_puts(seq, "adaptive");
1365 else if (test_opt(sbi, LFS))
1366 seq_puts(seq, "lfs");
1367 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1368 if (test_opt(sbi, RESERVE_ROOT))
1369 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1370 F2FS_OPTION(sbi).root_reserved_blocks,
1371 from_kuid_munged(&init_user_ns,
1372 F2FS_OPTION(sbi).s_resuid),
1373 from_kgid_munged(&init_user_ns,
1374 F2FS_OPTION(sbi).s_resgid));
1375 if (F2FS_IO_SIZE_BITS(sbi))
1376 seq_printf(seq, ",io_bits=%u",
1377 F2FS_OPTION(sbi).write_io_size_bits);
1378 #ifdef CONFIG_F2FS_FAULT_INJECTION
1379 if (test_opt(sbi, FAULT_INJECTION)) {
1380 seq_printf(seq, ",fault_injection=%u",
1381 F2FS_OPTION(sbi).fault_info.inject_rate);
1382 seq_printf(seq, ",fault_type=%u",
1383 F2FS_OPTION(sbi).fault_info.inject_type);
1387 if (test_opt(sbi, QUOTA))
1388 seq_puts(seq, ",quota");
1389 if (test_opt(sbi, USRQUOTA))
1390 seq_puts(seq, ",usrquota");
1391 if (test_opt(sbi, GRPQUOTA))
1392 seq_puts(seq, ",grpquota");
1393 if (test_opt(sbi, PRJQUOTA))
1394 seq_puts(seq, ",prjquota");
1396 f2fs_show_quota_options(seq, sbi->sb);
1397 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1398 seq_printf(seq, ",whint_mode=%s", "user-based");
1399 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1400 seq_printf(seq, ",whint_mode=%s", "fs-based");
1401 #ifdef CONFIG_FS_ENCRYPTION
1402 if (F2FS_OPTION(sbi).test_dummy_encryption)
1403 seq_puts(seq, ",test_dummy_encryption");
1406 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1407 seq_printf(seq, ",alloc_mode=%s", "default");
1408 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1409 seq_printf(seq, ",alloc_mode=%s", "reuse");
1411 if (test_opt(sbi, DISABLE_CHECKPOINT))
1412 seq_puts(seq, ",checkpoint=disable");
1414 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1415 seq_printf(seq, ",fsync_mode=%s", "posix");
1416 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1417 seq_printf(seq, ",fsync_mode=%s", "strict");
1418 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1419 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1423 static void default_options(struct f2fs_sb_info *sbi)
1425 /* init some FS parameters */
1426 F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1427 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1428 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1429 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1430 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1431 F2FS_OPTION(sbi).test_dummy_encryption = false;
1432 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1433 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1435 set_opt(sbi, BG_GC);
1436 set_opt(sbi, INLINE_XATTR);
1437 set_opt(sbi, INLINE_DATA);
1438 set_opt(sbi, INLINE_DENTRY);
1439 set_opt(sbi, EXTENT_CACHE);
1440 set_opt(sbi, NOHEAP);
1441 clear_opt(sbi, DISABLE_CHECKPOINT);
1442 sbi->sb->s_flags |= SB_LAZYTIME;
1443 set_opt(sbi, FLUSH_MERGE);
1444 set_opt(sbi, DISCARD);
1445 if (f2fs_sb_has_blkzoned(sbi))
1446 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1448 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1450 #ifdef CONFIG_F2FS_FS_XATTR
1451 set_opt(sbi, XATTR_USER);
1453 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1454 set_opt(sbi, POSIX_ACL);
1457 f2fs_build_fault_attr(sbi, 0, 0);
1461 static int f2fs_enable_quotas(struct super_block *sb);
1464 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1466 unsigned int s_flags = sbi->sb->s_flags;
1467 struct cp_control cpc;
1471 if (s_flags & SB_RDONLY) {
1472 f2fs_msg(sbi->sb, KERN_ERR,
1473 "checkpoint=disable on readonly fs");
1476 sbi->sb->s_flags |= SB_ACTIVE;
1478 f2fs_update_time(sbi, DISABLE_TIME);
1480 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1481 mutex_lock(&sbi->gc_mutex);
1482 err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1483 if (err == -ENODATA) {
1487 if (err && err != -EAGAIN)
1491 ret = sync_filesystem(sbi->sb);
1493 err = ret ? ret: err;
1497 if (f2fs_disable_cp_again(sbi)) {
1502 mutex_lock(&sbi->gc_mutex);
1503 cpc.reason = CP_PAUSE;
1504 set_sbi_flag(sbi, SBI_CP_DISABLED);
1505 err = f2fs_write_checkpoint(sbi, &cpc);
1509 spin_lock(&sbi->stat_lock);
1510 sbi->unusable_block_count = 0;
1511 spin_unlock(&sbi->stat_lock);
1514 mutex_unlock(&sbi->gc_mutex);
1516 sbi->sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1520 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1522 mutex_lock(&sbi->gc_mutex);
1523 f2fs_dirty_to_prefree(sbi);
1525 clear_sbi_flag(sbi, SBI_CP_DISABLED);
1526 set_sbi_flag(sbi, SBI_IS_DIRTY);
1527 mutex_unlock(&sbi->gc_mutex);
1529 f2fs_sync_fs(sbi->sb, 1);
1532 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1534 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1535 struct f2fs_mount_info org_mount_opt;
1536 unsigned long old_sb_flags;
1538 bool need_restart_gc = false;
1539 bool need_stop_gc = false;
1540 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1541 bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1542 bool checkpoint_changed;
1548 * Save the old mount options in case we
1549 * need to restore them.
1551 org_mount_opt = sbi->mount_opt;
1552 old_sb_flags = sb->s_flags;
1555 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1556 for (i = 0; i < MAXQUOTAS; i++) {
1557 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1558 org_mount_opt.s_qf_names[i] =
1559 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1561 if (!org_mount_opt.s_qf_names[i]) {
1562 for (j = 0; j < i; j++)
1563 kvfree(org_mount_opt.s_qf_names[j]);
1567 org_mount_opt.s_qf_names[i] = NULL;
1572 /* recover superblocks we couldn't write due to previous RO mount */
1573 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1574 err = f2fs_commit_super(sbi, false);
1575 f2fs_msg(sb, KERN_INFO,
1576 "Try to recover all the superblocks, ret: %d", err);
1578 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1581 default_options(sbi);
1583 /* parse mount options */
1584 err = parse_options(sb, data);
1587 checkpoint_changed =
1588 disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1591 * Previous and new state of filesystem is RO,
1592 * so skip checking GC and FLUSH_MERGE conditions.
1594 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1598 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1599 err = dquot_suspend(sb, -1);
1602 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1603 /* dquot_resume needs RW */
1604 sb->s_flags &= ~SB_RDONLY;
1605 if (sb_any_quota_suspended(sb)) {
1606 dquot_resume(sb, -1);
1607 } else if (f2fs_sb_has_quota_ino(sbi)) {
1608 err = f2fs_enable_quotas(sb);
1614 /* disallow enable/disable extent_cache dynamically */
1615 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1617 f2fs_msg(sbi->sb, KERN_WARNING,
1618 "switch extent_cache option is not allowed");
1622 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1624 f2fs_msg(sbi->sb, KERN_WARNING,
1625 "disabling checkpoint not compatible with read-only");
1630 * We stop the GC thread if FS is mounted as RO
1631 * or if background_gc = off is passed in mount
1632 * option. Also sync the filesystem.
1634 if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) {
1635 if (sbi->gc_thread) {
1636 f2fs_stop_gc_thread(sbi);
1637 need_restart_gc = true;
1639 } else if (!sbi->gc_thread) {
1640 err = f2fs_start_gc_thread(sbi);
1643 need_stop_gc = true;
1646 if (*flags & SB_RDONLY ||
1647 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1648 writeback_inodes_sb(sb, WB_REASON_SYNC);
1651 set_sbi_flag(sbi, SBI_IS_DIRTY);
1652 set_sbi_flag(sbi, SBI_IS_CLOSE);
1653 f2fs_sync_fs(sb, 1);
1654 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1657 if (checkpoint_changed) {
1658 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1659 err = f2fs_disable_checkpoint(sbi);
1663 f2fs_enable_checkpoint(sbi);
1668 * We stop issue flush thread if FS is mounted as RO
1669 * or if flush_merge is not passed in mount option.
1671 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1672 clear_opt(sbi, FLUSH_MERGE);
1673 f2fs_destroy_flush_cmd_control(sbi, false);
1675 err = f2fs_create_flush_cmd_control(sbi);
1681 /* Release old quota file names */
1682 for (i = 0; i < MAXQUOTAS; i++)
1683 kvfree(org_mount_opt.s_qf_names[i]);
1685 /* Update the POSIXACL Flag */
1686 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1687 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1689 limit_reserve_root(sbi);
1690 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1693 if (need_restart_gc) {
1694 if (f2fs_start_gc_thread(sbi))
1695 f2fs_msg(sbi->sb, KERN_WARNING,
1696 "background gc thread has stopped");
1697 } else if (need_stop_gc) {
1698 f2fs_stop_gc_thread(sbi);
1702 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1703 for (i = 0; i < MAXQUOTAS; i++) {
1704 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1705 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1708 sbi->mount_opt = org_mount_opt;
1709 sb->s_flags = old_sb_flags;
1714 /* Read data from quotafile */
1715 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1716 size_t len, loff_t off)
1718 struct inode *inode = sb_dqopt(sb)->files[type];
1719 struct address_space *mapping = inode->i_mapping;
1720 block_t blkidx = F2FS_BYTES_TO_BLK(off);
1721 int offset = off & (sb->s_blocksize - 1);
1724 loff_t i_size = i_size_read(inode);
1731 if (off + len > i_size)
1734 while (toread > 0) {
1735 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1737 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1739 if (PTR_ERR(page) == -ENOMEM) {
1740 congestion_wait(BLK_RW_ASYNC, HZ/50);
1743 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1744 return PTR_ERR(page);
1749 if (unlikely(page->mapping != mapping)) {
1750 f2fs_put_page(page, 1);
1753 if (unlikely(!PageUptodate(page))) {
1754 f2fs_put_page(page, 1);
1755 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1759 kaddr = kmap_atomic(page);
1760 memcpy(data, kaddr + offset, tocopy);
1761 kunmap_atomic(kaddr);
1762 f2fs_put_page(page, 1);
1772 /* Write to quotafile */
1773 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1774 const char *data, size_t len, loff_t off)
1776 struct inode *inode = sb_dqopt(sb)->files[type];
1777 struct address_space *mapping = inode->i_mapping;
1778 const struct address_space_operations *a_ops = mapping->a_ops;
1779 int offset = off & (sb->s_blocksize - 1);
1780 size_t towrite = len;
1786 while (towrite > 0) {
1787 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1790 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1792 if (unlikely(err)) {
1793 if (err == -ENOMEM) {
1794 congestion_wait(BLK_RW_ASYNC, HZ/50);
1797 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1801 kaddr = kmap_atomic(page);
1802 memcpy(kaddr + offset, data, tocopy);
1803 kunmap_atomic(kaddr);
1804 flush_dcache_page(page);
1806 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1817 inode->i_mtime = inode->i_ctime = current_time(inode);
1818 f2fs_mark_inode_dirty_sync(inode, false);
1819 return len - towrite;
1822 static struct dquot **f2fs_get_dquots(struct inode *inode)
1824 return F2FS_I(inode)->i_dquot;
1827 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1829 return &F2FS_I(inode)->i_reserved_quota;
1832 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1834 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
1835 f2fs_msg(sbi->sb, KERN_ERR,
1836 "quota sysfile may be corrupted, skip loading it");
1840 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
1841 F2FS_OPTION(sbi).s_jquota_fmt, type);
1844 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1849 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
1850 err = f2fs_enable_quotas(sbi->sb);
1852 f2fs_msg(sbi->sb, KERN_ERR,
1853 "Cannot turn on quota_ino: %d", err);
1859 for (i = 0; i < MAXQUOTAS; i++) {
1860 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1861 err = f2fs_quota_on_mount(sbi, i);
1866 f2fs_msg(sbi->sb, KERN_ERR,
1867 "Cannot turn on quotas: %d on %d", err, i);
1873 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
1876 struct inode *qf_inode;
1877 unsigned long qf_inum;
1880 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
1882 qf_inum = f2fs_qf_ino(sb, type);
1886 qf_inode = f2fs_iget(sb, qf_inum);
1887 if (IS_ERR(qf_inode)) {
1888 f2fs_msg(sb, KERN_ERR,
1889 "Bad quota inode %u:%lu", type, qf_inum);
1890 return PTR_ERR(qf_inode);
1893 /* Don't account quota for quota files to avoid recursion */
1894 qf_inode->i_flags |= S_NOQUOTA;
1895 err = dquot_enable(qf_inode, type, format_id, flags);
1900 static int f2fs_enable_quotas(struct super_block *sb)
1903 unsigned long qf_inum;
1904 bool quota_mopt[MAXQUOTAS] = {
1905 test_opt(F2FS_SB(sb), USRQUOTA),
1906 test_opt(F2FS_SB(sb), GRPQUOTA),
1907 test_opt(F2FS_SB(sb), PRJQUOTA),
1910 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
1911 f2fs_msg(sb, KERN_ERR,
1912 "quota file may be corrupted, skip loading it");
1916 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
1918 for (type = 0; type < MAXQUOTAS; type++) {
1919 qf_inum = f2fs_qf_ino(sb, type);
1921 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
1922 DQUOT_USAGE_ENABLED |
1923 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
1925 f2fs_msg(sb, KERN_ERR,
1926 "Failed to enable quota tracking "
1927 "(type=%d, err=%d). Please run "
1928 "fsck to fix.", type, err);
1929 for (type--; type >= 0; type--)
1930 dquot_quota_off(sb, type);
1931 set_sbi_flag(F2FS_SB(sb),
1932 SBI_QUOTA_NEED_REPAIR);
1940 int f2fs_quota_sync(struct super_block *sb, int type)
1942 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1943 struct quota_info *dqopt = sb_dqopt(sb);
1947 ret = dquot_writeback_dquots(sb, type);
1952 * Now when everything is written we can discard the pagecache so
1953 * that userspace sees the changes.
1955 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
1956 struct address_space *mapping;
1958 if (type != -1 && cnt != type)
1960 if (!sb_has_quota_active(sb, cnt))
1963 mapping = dqopt->files[cnt]->i_mapping;
1965 ret = filemap_fdatawrite(mapping);
1969 /* if we are using journalled quota */
1970 if (is_journalled_quota(sbi))
1973 ret = filemap_fdatawait(mapping);
1975 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1977 inode_lock(dqopt->files[cnt]);
1978 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
1979 inode_unlock(dqopt->files[cnt]);
1983 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1987 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
1988 const struct path *path)
1990 struct inode *inode;
1993 err = f2fs_quota_sync(sb, type);
1997 err = dquot_quota_on(sb, type, format_id, path);
2001 inode = d_inode(path->dentry);
2004 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2005 f2fs_set_inode_flags(inode);
2006 inode_unlock(inode);
2007 f2fs_mark_inode_dirty_sync(inode, false);
2012 static int f2fs_quota_off(struct super_block *sb, int type)
2014 struct inode *inode = sb_dqopt(sb)->files[type];
2017 if (!inode || !igrab(inode))
2018 return dquot_quota_off(sb, type);
2020 err = f2fs_quota_sync(sb, type);
2024 err = dquot_quota_off(sb, type);
2025 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2029 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2030 f2fs_set_inode_flags(inode);
2031 inode_unlock(inode);
2032 f2fs_mark_inode_dirty_sync(inode, false);
2038 void f2fs_quota_off_umount(struct super_block *sb)
2043 for (type = 0; type < MAXQUOTAS; type++) {
2044 err = f2fs_quota_off(sb, type);
2046 int ret = dquot_quota_off(sb, type);
2048 f2fs_msg(sb, KERN_ERR,
2049 "Fail to turn off disk quota "
2050 "(type: %d, err: %d, ret:%d), Please "
2051 "run fsck to fix it.", type, err, ret);
2052 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2056 * In case of checkpoint=disable, we must flush quota blocks.
2057 * This can cause NULL exception for node_inode in end_io, since
2058 * put_super already dropped it.
2060 sync_filesystem(sb);
2063 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2065 struct quota_info *dqopt = sb_dqopt(sb);
2068 for (type = 0; type < MAXQUOTAS; type++) {
2069 if (!dqopt->files[type])
2071 f2fs_inode_synced(dqopt->files[type]);
2075 static int f2fs_dquot_commit(struct dquot *dquot)
2079 ret = dquot_commit(dquot);
2081 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2085 static int f2fs_dquot_acquire(struct dquot *dquot)
2089 ret = dquot_acquire(dquot);
2091 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2096 static int f2fs_dquot_release(struct dquot *dquot)
2100 ret = dquot_release(dquot);
2102 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2106 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2108 struct super_block *sb = dquot->dq_sb;
2109 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2112 ret = dquot_mark_dquot_dirty(dquot);
2114 /* if we are using journalled quota */
2115 if (is_journalled_quota(sbi))
2116 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2121 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2125 ret = dquot_commit_info(sb, type);
2127 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2131 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2133 *projid = F2FS_I(inode)->i_projid;
2137 static const struct dquot_operations f2fs_quota_operations = {
2138 .get_reserved_space = f2fs_get_reserved_space,
2139 .write_dquot = f2fs_dquot_commit,
2140 .acquire_dquot = f2fs_dquot_acquire,
2141 .release_dquot = f2fs_dquot_release,
2142 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
2143 .write_info = f2fs_dquot_commit_info,
2144 .alloc_dquot = dquot_alloc,
2145 .destroy_dquot = dquot_destroy,
2146 .get_projid = f2fs_get_projid,
2147 .get_next_id = dquot_get_next_id,
2150 static const struct quotactl_ops f2fs_quotactl_ops = {
2151 .quota_on = f2fs_quota_on,
2152 .quota_off = f2fs_quota_off,
2153 .quota_sync = f2fs_quota_sync,
2154 .get_state = dquot_get_state,
2155 .set_info = dquot_set_dqinfo,
2156 .get_dqblk = dquot_get_dqblk,
2157 .set_dqblk = dquot_set_dqblk,
2158 .get_nextdqblk = dquot_get_next_dqblk,
2161 int f2fs_quota_sync(struct super_block *sb, int type)
2166 void f2fs_quota_off_umount(struct super_block *sb)
2171 static const struct super_operations f2fs_sops = {
2172 .alloc_inode = f2fs_alloc_inode,
2173 .free_inode = f2fs_free_inode,
2174 .drop_inode = f2fs_drop_inode,
2175 .write_inode = f2fs_write_inode,
2176 .dirty_inode = f2fs_dirty_inode,
2177 .show_options = f2fs_show_options,
2179 .quota_read = f2fs_quota_read,
2180 .quota_write = f2fs_quota_write,
2181 .get_dquots = f2fs_get_dquots,
2183 .evict_inode = f2fs_evict_inode,
2184 .put_super = f2fs_put_super,
2185 .sync_fs = f2fs_sync_fs,
2186 .freeze_fs = f2fs_freeze,
2187 .unfreeze_fs = f2fs_unfreeze,
2188 .statfs = f2fs_statfs,
2189 .remount_fs = f2fs_remount,
2192 #ifdef CONFIG_FS_ENCRYPTION
2193 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2195 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2196 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2200 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2203 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2206 * Encrypting the root directory is not allowed because fsck
2207 * expects lost+found directory to exist and remain unencrypted
2208 * if LOST_FOUND feature is enabled.
2211 if (f2fs_sb_has_lost_found(sbi) &&
2212 inode->i_ino == F2FS_ROOT_INO(sbi))
2215 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2216 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2217 ctx, len, fs_data, XATTR_CREATE);
2220 static bool f2fs_dummy_context(struct inode *inode)
2222 return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
2225 static const struct fscrypt_operations f2fs_cryptops = {
2226 .key_prefix = "f2fs:",
2227 .get_context = f2fs_get_context,
2228 .set_context = f2fs_set_context,
2229 .dummy_context = f2fs_dummy_context,
2230 .empty_dir = f2fs_empty_dir,
2231 .max_namelen = F2FS_NAME_LEN,
2235 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2236 u64 ino, u32 generation)
2238 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2239 struct inode *inode;
2241 if (f2fs_check_nid_range(sbi, ino))
2242 return ERR_PTR(-ESTALE);
2245 * f2fs_iget isn't quite right if the inode is currently unallocated!
2246 * However f2fs_iget currently does appropriate checks to handle stale
2247 * inodes so everything is OK.
2249 inode = f2fs_iget(sb, ino);
2251 return ERR_CAST(inode);
2252 if (unlikely(generation && inode->i_generation != generation)) {
2253 /* we didn't find the right inode.. */
2255 return ERR_PTR(-ESTALE);
2260 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2261 int fh_len, int fh_type)
2263 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2264 f2fs_nfs_get_inode);
2267 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2268 int fh_len, int fh_type)
2270 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2271 f2fs_nfs_get_inode);
2274 static const struct export_operations f2fs_export_ops = {
2275 .fh_to_dentry = f2fs_fh_to_dentry,
2276 .fh_to_parent = f2fs_fh_to_parent,
2277 .get_parent = f2fs_get_parent,
2280 static loff_t max_file_blocks(void)
2283 loff_t leaf_count = DEF_ADDRS_PER_BLOCK;
2286 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2287 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2288 * space in inode.i_addr, it will be more safe to reassign
2292 /* two direct node blocks */
2293 result += (leaf_count * 2);
2295 /* two indirect node blocks */
2296 leaf_count *= NIDS_PER_BLOCK;
2297 result += (leaf_count * 2);
2299 /* one double indirect node block */
2300 leaf_count *= NIDS_PER_BLOCK;
2301 result += leaf_count;
2306 static int __f2fs_commit_super(struct buffer_head *bh,
2307 struct f2fs_super_block *super)
2311 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2312 set_buffer_dirty(bh);
2315 /* it's rare case, we can do fua all the time */
2316 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2319 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2320 struct buffer_head *bh)
2322 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2323 (bh->b_data + F2FS_SUPER_OFFSET);
2324 struct super_block *sb = sbi->sb;
2325 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2326 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2327 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2328 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2329 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2330 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2331 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2332 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2333 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2334 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2335 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2336 u32 segment_count = le32_to_cpu(raw_super->segment_count);
2337 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2338 u64 main_end_blkaddr = main_blkaddr +
2339 (segment_count_main << log_blocks_per_seg);
2340 u64 seg_end_blkaddr = segment0_blkaddr +
2341 (segment_count << log_blocks_per_seg);
2343 if (segment0_blkaddr != cp_blkaddr) {
2344 f2fs_msg(sb, KERN_INFO,
2345 "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2346 segment0_blkaddr, cp_blkaddr);
2350 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2352 f2fs_msg(sb, KERN_INFO,
2353 "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2354 cp_blkaddr, sit_blkaddr,
2355 segment_count_ckpt << log_blocks_per_seg);
2359 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2361 f2fs_msg(sb, KERN_INFO,
2362 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2363 sit_blkaddr, nat_blkaddr,
2364 segment_count_sit << log_blocks_per_seg);
2368 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2370 f2fs_msg(sb, KERN_INFO,
2371 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2372 nat_blkaddr, ssa_blkaddr,
2373 segment_count_nat << log_blocks_per_seg);
2377 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2379 f2fs_msg(sb, KERN_INFO,
2380 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2381 ssa_blkaddr, main_blkaddr,
2382 segment_count_ssa << log_blocks_per_seg);
2386 if (main_end_blkaddr > seg_end_blkaddr) {
2387 f2fs_msg(sb, KERN_INFO,
2388 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2391 (segment_count << log_blocks_per_seg),
2392 segment_count_main << log_blocks_per_seg);
2394 } else if (main_end_blkaddr < seg_end_blkaddr) {
2398 /* fix in-memory information all the time */
2399 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2400 segment0_blkaddr) >> log_blocks_per_seg);
2402 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2403 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2406 err = __f2fs_commit_super(bh, NULL);
2407 res = err ? "failed" : "done";
2409 f2fs_msg(sb, KERN_INFO,
2410 "Fix alignment : %s, start(%u) end(%u) block(%u)",
2413 (segment_count << log_blocks_per_seg),
2414 segment_count_main << log_blocks_per_seg);
2421 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2422 struct buffer_head *bh)
2424 block_t segment_count, segs_per_sec, secs_per_zone;
2425 block_t total_sections, blocks_per_seg;
2426 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2427 (bh->b_data + F2FS_SUPER_OFFSET);
2428 struct super_block *sb = sbi->sb;
2429 unsigned int blocksize;
2430 size_t crc_offset = 0;
2433 /* Check checksum_offset and crc in superblock */
2434 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2435 crc_offset = le32_to_cpu(raw_super->checksum_offset);
2437 offsetof(struct f2fs_super_block, crc)) {
2438 f2fs_msg(sb, KERN_INFO,
2439 "Invalid SB checksum offset: %zu",
2443 crc = le32_to_cpu(raw_super->crc);
2444 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2445 f2fs_msg(sb, KERN_INFO,
2446 "Invalid SB checksum value: %u", crc);
2451 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
2452 f2fs_msg(sb, KERN_INFO,
2453 "Magic Mismatch, valid(0x%x) - read(0x%x)",
2454 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2458 /* Currently, support only 4KB page cache size */
2459 if (F2FS_BLKSIZE != PAGE_SIZE) {
2460 f2fs_msg(sb, KERN_INFO,
2461 "Invalid page_cache_size (%lu), supports only 4KB",
2466 /* Currently, support only 4KB block size */
2467 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2468 if (blocksize != F2FS_BLKSIZE) {
2469 f2fs_msg(sb, KERN_INFO,
2470 "Invalid blocksize (%u), supports only 4KB",
2475 /* check log blocks per segment */
2476 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2477 f2fs_msg(sb, KERN_INFO,
2478 "Invalid log blocks per segment (%u)",
2479 le32_to_cpu(raw_super->log_blocks_per_seg));
2483 /* Currently, support 512/1024/2048/4096 bytes sector size */
2484 if (le32_to_cpu(raw_super->log_sectorsize) >
2485 F2FS_MAX_LOG_SECTOR_SIZE ||
2486 le32_to_cpu(raw_super->log_sectorsize) <
2487 F2FS_MIN_LOG_SECTOR_SIZE) {
2488 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
2489 le32_to_cpu(raw_super->log_sectorsize));
2492 if (le32_to_cpu(raw_super->log_sectors_per_block) +
2493 le32_to_cpu(raw_super->log_sectorsize) !=
2494 F2FS_MAX_LOG_SECTOR_SIZE) {
2495 f2fs_msg(sb, KERN_INFO,
2496 "Invalid log sectors per block(%u) log sectorsize(%u)",
2497 le32_to_cpu(raw_super->log_sectors_per_block),
2498 le32_to_cpu(raw_super->log_sectorsize));
2502 segment_count = le32_to_cpu(raw_super->segment_count);
2503 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2504 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2505 total_sections = le32_to_cpu(raw_super->section_count);
2507 /* blocks_per_seg should be 512, given the above check */
2508 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2510 if (segment_count > F2FS_MAX_SEGMENT ||
2511 segment_count < F2FS_MIN_SEGMENTS) {
2512 f2fs_msg(sb, KERN_INFO,
2513 "Invalid segment count (%u)",
2518 if (total_sections > segment_count ||
2519 total_sections < F2FS_MIN_SEGMENTS ||
2520 segs_per_sec > segment_count || !segs_per_sec) {
2521 f2fs_msg(sb, KERN_INFO,
2522 "Invalid segment/section count (%u, %u x %u)",
2523 segment_count, total_sections, segs_per_sec);
2527 if ((segment_count / segs_per_sec) < total_sections) {
2528 f2fs_msg(sb, KERN_INFO,
2529 "Small segment_count (%u < %u * %u)",
2530 segment_count, segs_per_sec, total_sections);
2534 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2535 f2fs_msg(sb, KERN_INFO,
2536 "Wrong segment_count / block_count (%u > %llu)",
2537 segment_count, le64_to_cpu(raw_super->block_count));
2541 if (secs_per_zone > total_sections || !secs_per_zone) {
2542 f2fs_msg(sb, KERN_INFO,
2543 "Wrong secs_per_zone / total_sections (%u, %u)",
2544 secs_per_zone, total_sections);
2547 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2548 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2549 (le32_to_cpu(raw_super->extension_count) +
2550 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2551 f2fs_msg(sb, KERN_INFO,
2552 "Corrupted extension count (%u + %u > %u)",
2553 le32_to_cpu(raw_super->extension_count),
2554 raw_super->hot_ext_count,
2555 F2FS_MAX_EXTENSION);
2559 if (le32_to_cpu(raw_super->cp_payload) >
2560 (blocks_per_seg - F2FS_CP_PACKS)) {
2561 f2fs_msg(sb, KERN_INFO,
2562 "Insane cp_payload (%u > %u)",
2563 le32_to_cpu(raw_super->cp_payload),
2564 blocks_per_seg - F2FS_CP_PACKS);
2568 /* check reserved ino info */
2569 if (le32_to_cpu(raw_super->node_ino) != 1 ||
2570 le32_to_cpu(raw_super->meta_ino) != 2 ||
2571 le32_to_cpu(raw_super->root_ino) != 3) {
2572 f2fs_msg(sb, KERN_INFO,
2573 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2574 le32_to_cpu(raw_super->node_ino),
2575 le32_to_cpu(raw_super->meta_ino),
2576 le32_to_cpu(raw_super->root_ino));
2580 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2581 if (sanity_check_area_boundary(sbi, bh))
2587 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2589 unsigned int total, fsmeta;
2590 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2591 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2592 unsigned int ovp_segments, reserved_segments;
2593 unsigned int main_segs, blocks_per_seg;
2594 unsigned int sit_segs, nat_segs;
2595 unsigned int sit_bitmap_size, nat_bitmap_size;
2596 unsigned int log_blocks_per_seg;
2597 unsigned int segment_count_main;
2598 unsigned int cp_pack_start_sum, cp_payload;
2599 block_t user_block_count, valid_user_blocks;
2600 block_t avail_node_count, valid_node_count;
2603 total = le32_to_cpu(raw_super->segment_count);
2604 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2605 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2607 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2609 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2610 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2612 if (unlikely(fsmeta >= total))
2615 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2616 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2618 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2619 ovp_segments == 0 || reserved_segments == 0)) {
2620 f2fs_msg(sbi->sb, KERN_ERR,
2621 "Wrong layout: check mkfs.f2fs version");
2625 user_block_count = le64_to_cpu(ckpt->user_block_count);
2626 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2627 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2628 if (!user_block_count || user_block_count >=
2629 segment_count_main << log_blocks_per_seg) {
2630 f2fs_msg(sbi->sb, KERN_ERR,
2631 "Wrong user_block_count: %u", user_block_count);
2635 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
2636 if (valid_user_blocks > user_block_count) {
2637 f2fs_msg(sbi->sb, KERN_ERR,
2638 "Wrong valid_user_blocks: %u, user_block_count: %u",
2639 valid_user_blocks, user_block_count);
2643 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
2644 avail_node_count = sbi->total_node_count - sbi->nquota_files -
2645 F2FS_RESERVED_NODE_NUM;
2646 if (valid_node_count > avail_node_count) {
2647 f2fs_msg(sbi->sb, KERN_ERR,
2648 "Wrong valid_node_count: %u, avail_node_count: %u",
2649 valid_node_count, avail_node_count);
2653 main_segs = le32_to_cpu(raw_super->segment_count_main);
2654 blocks_per_seg = sbi->blocks_per_seg;
2656 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2657 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2658 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2660 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2661 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2662 le32_to_cpu(ckpt->cur_node_segno[j])) {
2663 f2fs_msg(sbi->sb, KERN_ERR,
2664 "Node segment (%u, %u) has the same "
2666 le32_to_cpu(ckpt->cur_node_segno[i]));
2671 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2672 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2673 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2675 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2676 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2677 le32_to_cpu(ckpt->cur_data_segno[j])) {
2678 f2fs_msg(sbi->sb, KERN_ERR,
2679 "Data segment (%u, %u) has the same "
2681 le32_to_cpu(ckpt->cur_data_segno[i]));
2686 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2687 for (j = i; j < NR_CURSEG_DATA_TYPE; j++) {
2688 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2689 le32_to_cpu(ckpt->cur_data_segno[j])) {
2690 f2fs_msg(sbi->sb, KERN_ERR,
2691 "Data segment (%u) and Data segment (%u)"
2692 " has the same segno: %u", i, j,
2693 le32_to_cpu(ckpt->cur_node_segno[i]));
2699 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2700 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2702 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2703 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2704 f2fs_msg(sbi->sb, KERN_ERR,
2705 "Wrong bitmap size: sit: %u, nat:%u",
2706 sit_bitmap_size, nat_bitmap_size);
2710 cp_pack_start_sum = __start_sum_addr(sbi);
2711 cp_payload = __cp_payload(sbi);
2712 if (cp_pack_start_sum < cp_payload + 1 ||
2713 cp_pack_start_sum > blocks_per_seg - 1 -
2715 f2fs_msg(sbi->sb, KERN_ERR,
2716 "Wrong cp_pack_start_sum: %u",
2721 if (unlikely(f2fs_cp_error(sbi))) {
2722 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
2728 static void init_sb_info(struct f2fs_sb_info *sbi)
2730 struct f2fs_super_block *raw_super = sbi->raw_super;
2733 sbi->log_sectors_per_block =
2734 le32_to_cpu(raw_super->log_sectors_per_block);
2735 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2736 sbi->blocksize = 1 << sbi->log_blocksize;
2737 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2738 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2739 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2740 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2741 sbi->total_sections = le32_to_cpu(raw_super->section_count);
2742 sbi->total_node_count =
2743 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2744 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2745 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2746 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2747 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2748 sbi->cur_victim_sec = NULL_SECNO;
2749 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
2750 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
2751 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2752 sbi->migration_granularity = sbi->segs_per_sec;
2754 sbi->dir_level = DEF_DIR_LEVEL;
2755 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2756 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2757 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
2758 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
2759 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
2760 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
2761 DEF_UMOUNT_DISCARD_TIMEOUT;
2762 clear_sbi_flag(sbi, SBI_NEED_FSCK);
2764 for (i = 0; i < NR_COUNT_TYPE; i++)
2765 atomic_set(&sbi->nr_pages[i], 0);
2767 for (i = 0; i < META; i++)
2768 atomic_set(&sbi->wb_sync_req[i], 0);
2770 INIT_LIST_HEAD(&sbi->s_list);
2771 mutex_init(&sbi->umount_mutex);
2772 init_rwsem(&sbi->io_order_lock);
2773 spin_lock_init(&sbi->cp_lock);
2775 sbi->dirty_device = 0;
2776 spin_lock_init(&sbi->dev_lock);
2778 init_rwsem(&sbi->sb_lock);
2781 static int init_percpu_info(struct f2fs_sb_info *sbi)
2785 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2789 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
2792 percpu_counter_destroy(&sbi->alloc_valid_block_count);
2797 #ifdef CONFIG_BLK_DEV_ZONED
2798 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2800 struct block_device *bdev = FDEV(devi).bdev;
2801 sector_t nr_sectors = bdev->bd_part->nr_sects;
2802 sector_t sector = 0;
2803 struct blk_zone *zones;
2804 unsigned int i, nr_zones;
2808 if (!f2fs_sb_has_blkzoned(sbi))
2811 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2812 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2814 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2815 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2816 __ilog2_u32(sbi->blocks_per_blkz))
2818 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2819 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2820 sbi->log_blocks_per_blkz;
2821 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2822 FDEV(devi).nr_blkz++;
2824 FDEV(devi).blkz_seq = f2fs_kzalloc(sbi,
2825 BITS_TO_LONGS(FDEV(devi).nr_blkz)
2826 * sizeof(unsigned long),
2828 if (!FDEV(devi).blkz_seq)
2831 #define F2FS_REPORT_NR_ZONES 4096
2833 zones = f2fs_kzalloc(sbi,
2834 array_size(F2FS_REPORT_NR_ZONES,
2835 sizeof(struct blk_zone)),
2840 /* Get block zones type */
2841 while (zones && sector < nr_sectors) {
2843 nr_zones = F2FS_REPORT_NR_ZONES;
2844 err = blkdev_report_zones(bdev, sector,
2854 for (i = 0; i < nr_zones; i++) {
2855 if (zones[i].type != BLK_ZONE_TYPE_CONVENTIONAL)
2856 set_bit(n, FDEV(devi).blkz_seq);
2857 sector += zones[i].len;
2869 * Read f2fs raw super block.
2870 * Because we have two copies of super block, so read both of them
2871 * to get the first valid one. If any one of them is broken, we pass
2872 * them recovery flag back to the caller.
2874 static int read_raw_super_block(struct f2fs_sb_info *sbi,
2875 struct f2fs_super_block **raw_super,
2876 int *valid_super_block, int *recovery)
2878 struct super_block *sb = sbi->sb;
2880 struct buffer_head *bh;
2881 struct f2fs_super_block *super;
2884 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2888 for (block = 0; block < 2; block++) {
2889 bh = sb_bread(sb, block);
2891 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
2897 /* sanity checking of raw super */
2898 if (sanity_check_raw_super(sbi, bh)) {
2899 f2fs_msg(sb, KERN_ERR,
2900 "Can't find valid F2FS filesystem in %dth superblock",
2908 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2910 *valid_super_block = block;
2916 /* Fail to read any one of the superblocks*/
2920 /* No valid superblock */
2929 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
2931 struct buffer_head *bh;
2935 if ((recover && f2fs_readonly(sbi->sb)) ||
2936 bdev_read_only(sbi->sb->s_bdev)) {
2937 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2941 /* we should update superblock crc here */
2942 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
2943 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
2944 offsetof(struct f2fs_super_block, crc));
2945 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
2948 /* write back-up superblock first */
2949 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
2952 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2955 /* if we are in recovery path, skip writing valid superblock */
2959 /* write current valid superblock */
2960 bh = sb_bread(sbi->sb, sbi->valid_super_block);
2963 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2968 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
2970 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2971 unsigned int max_devices = MAX_DEVICES;
2974 /* Initialize single device information */
2975 if (!RDEV(0).path[0]) {
2976 if (!bdev_is_zoned(sbi->sb->s_bdev))
2982 * Initialize multiple devices information, or single
2983 * zoned block device information.
2985 sbi->devs = f2fs_kzalloc(sbi,
2986 array_size(max_devices,
2987 sizeof(struct f2fs_dev_info)),
2992 for (i = 0; i < max_devices; i++) {
2994 if (i > 0 && !RDEV(i).path[0])
2997 if (max_devices == 1) {
2998 /* Single zoned block device mount */
3000 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3001 sbi->sb->s_mode, sbi->sb->s_type);
3003 /* Multi-device mount */
3004 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3005 FDEV(i).total_segments =
3006 le32_to_cpu(RDEV(i).total_segments);
3008 FDEV(i).start_blk = 0;
3009 FDEV(i).end_blk = FDEV(i).start_blk +
3010 (FDEV(i).total_segments <<
3011 sbi->log_blocks_per_seg) - 1 +
3012 le32_to_cpu(raw_super->segment0_blkaddr);
3014 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3015 FDEV(i).end_blk = FDEV(i).start_blk +
3016 (FDEV(i).total_segments <<
3017 sbi->log_blocks_per_seg) - 1;
3019 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3020 sbi->sb->s_mode, sbi->sb->s_type);
3022 if (IS_ERR(FDEV(i).bdev))
3023 return PTR_ERR(FDEV(i).bdev);
3025 /* to release errored devices */
3026 sbi->s_ndevs = i + 1;
3028 #ifdef CONFIG_BLK_DEV_ZONED
3029 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3030 !f2fs_sb_has_blkzoned(sbi)) {
3031 f2fs_msg(sbi->sb, KERN_ERR,
3032 "Zoned block device feature not enabled\n");
3035 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3036 if (init_blkz_info(sbi, i)) {
3037 f2fs_msg(sbi->sb, KERN_ERR,
3038 "Failed to initialize F2FS blkzone information");
3041 if (max_devices == 1)
3043 f2fs_msg(sbi->sb, KERN_INFO,
3044 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3046 FDEV(i).total_segments,
3047 FDEV(i).start_blk, FDEV(i).end_blk,
3048 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3049 "Host-aware" : "Host-managed");
3053 f2fs_msg(sbi->sb, KERN_INFO,
3054 "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3056 FDEV(i).total_segments,
3057 FDEV(i).start_blk, FDEV(i).end_blk);
3059 f2fs_msg(sbi->sb, KERN_INFO,
3060 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3064 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3066 struct f2fs_sm_info *sm_i = SM_I(sbi);
3068 /* adjust parameters according to the volume size */
3069 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3070 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3071 sm_i->dcc_info->discard_granularity = 1;
3072 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3075 sbi->readdir_ra = 1;
3078 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3080 struct f2fs_sb_info *sbi;
3081 struct f2fs_super_block *raw_super;
3084 bool skip_recovery = false, need_fsck = false;
3085 char *options = NULL;
3086 int recovery, i, valid_super_block;
3087 struct curseg_info *seg_i;
3093 valid_super_block = -1;
3096 /* allocate memory for f2fs-specific super block info */
3097 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3103 /* Load the checksum driver */
3104 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3105 if (IS_ERR(sbi->s_chksum_driver)) {
3106 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
3107 err = PTR_ERR(sbi->s_chksum_driver);
3108 sbi->s_chksum_driver = NULL;
3112 /* set a block size */
3113 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3114 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
3118 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3123 sb->s_fs_info = sbi;
3124 sbi->raw_super = raw_super;
3126 /* precompute checksum seed for metadata */
3127 if (f2fs_sb_has_inode_chksum(sbi))
3128 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3129 sizeof(raw_super->uuid));
3132 * The BLKZONED feature indicates that the drive was formatted with
3133 * zone alignment optimization. This is optional for host-aware
3134 * devices, but mandatory for host-managed zoned block devices.
3136 #ifndef CONFIG_BLK_DEV_ZONED
3137 if (f2fs_sb_has_blkzoned(sbi)) {
3138 f2fs_msg(sb, KERN_ERR,
3139 "Zoned block device support is not enabled");
3144 default_options(sbi);
3145 /* parse mount options */
3146 options = kstrdup((const char *)data, GFP_KERNEL);
3147 if (data && !options) {
3152 err = parse_options(sb, options);
3156 sbi->max_file_blocks = max_file_blocks();
3157 sb->s_maxbytes = sbi->max_file_blocks <<
3158 le32_to_cpu(raw_super->log_blocksize);
3159 sb->s_max_links = F2FS_LINK_MAX;
3162 sb->dq_op = &f2fs_quota_operations;
3163 if (f2fs_sb_has_quota_ino(sbi))
3164 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3166 sb->s_qcop = &f2fs_quotactl_ops;
3167 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3169 if (f2fs_sb_has_quota_ino(sbi)) {
3170 for (i = 0; i < MAXQUOTAS; i++) {
3171 if (f2fs_qf_ino(sbi->sb, i))
3172 sbi->nquota_files++;
3177 sb->s_op = &f2fs_sops;
3178 #ifdef CONFIG_FS_ENCRYPTION
3179 sb->s_cop = &f2fs_cryptops;
3181 sb->s_xattr = f2fs_xattr_handlers;
3182 sb->s_export_op = &f2fs_export_ops;
3183 sb->s_magic = F2FS_SUPER_MAGIC;
3184 sb->s_time_gran = 1;
3185 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3186 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3187 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3188 sb->s_iflags |= SB_I_CGROUPWB;
3190 /* init f2fs-specific super block info */
3191 sbi->valid_super_block = valid_super_block;
3192 mutex_init(&sbi->gc_mutex);
3193 mutex_init(&sbi->writepages);
3194 mutex_init(&sbi->cp_mutex);
3195 init_rwsem(&sbi->node_write);
3196 init_rwsem(&sbi->node_change);
3198 /* disallow all the data/node/meta page writes */
3199 set_sbi_flag(sbi, SBI_POR_DOING);
3200 spin_lock_init(&sbi->stat_lock);
3202 /* init iostat info */
3203 spin_lock_init(&sbi->iostat_lock);
3204 sbi->iostat_enable = false;
3206 for (i = 0; i < NR_PAGE_TYPE; i++) {
3207 int n = (i == META) ? 1: NR_TEMP_TYPE;
3213 sizeof(struct f2fs_bio_info)),
3215 if (!sbi->write_io[i]) {
3220 for (j = HOT; j < n; j++) {
3221 init_rwsem(&sbi->write_io[i][j].io_rwsem);
3222 sbi->write_io[i][j].sbi = sbi;
3223 sbi->write_io[i][j].bio = NULL;
3224 spin_lock_init(&sbi->write_io[i][j].io_lock);
3225 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3229 init_rwsem(&sbi->cp_rwsem);
3230 init_waitqueue_head(&sbi->cp_wait);
3233 err = init_percpu_info(sbi);
3237 if (F2FS_IO_SIZE(sbi) > 1) {
3238 sbi->write_io_dummy =
3239 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3240 if (!sbi->write_io_dummy) {
3246 /* get an inode for meta space */
3247 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3248 if (IS_ERR(sbi->meta_inode)) {
3249 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
3250 err = PTR_ERR(sbi->meta_inode);
3254 err = f2fs_get_valid_checkpoint(sbi);
3256 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
3257 goto free_meta_inode;
3260 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3261 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3262 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3263 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3264 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3267 /* Initialize device list */
3268 err = f2fs_scan_devices(sbi);
3270 f2fs_msg(sb, KERN_ERR, "Failed to find devices");
3274 sbi->total_valid_node_count =
3275 le32_to_cpu(sbi->ckpt->valid_node_count);
3276 percpu_counter_set(&sbi->total_valid_inode_count,
3277 le32_to_cpu(sbi->ckpt->valid_inode_count));
3278 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3279 sbi->total_valid_block_count =
3280 le64_to_cpu(sbi->ckpt->valid_block_count);
3281 sbi->last_valid_block_count = sbi->total_valid_block_count;
3282 sbi->reserved_blocks = 0;
3283 sbi->current_reserved_blocks = 0;
3284 limit_reserve_root(sbi);
3286 for (i = 0; i < NR_INODE_TYPE; i++) {
3287 INIT_LIST_HEAD(&sbi->inode_list[i]);
3288 spin_lock_init(&sbi->inode_lock[i]);
3291 f2fs_init_extent_cache_info(sbi);
3293 f2fs_init_ino_entry_info(sbi);
3295 f2fs_init_fsync_node_info(sbi);
3297 /* setup f2fs internal modules */
3298 err = f2fs_build_segment_manager(sbi);
3300 f2fs_msg(sb, KERN_ERR,
3301 "Failed to initialize F2FS segment manager");
3304 err = f2fs_build_node_manager(sbi);
3306 f2fs_msg(sb, KERN_ERR,
3307 "Failed to initialize F2FS node manager");
3311 /* For write statistics */
3312 if (sb->s_bdev->bd_part)
3313 sbi->sectors_written_start =
3314 (u64)part_stat_read(sb->s_bdev->bd_part,
3315 sectors[STAT_WRITE]);
3317 /* Read accumulated write IO statistics if exists */
3318 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3319 if (__exist_node_summaries(sbi))
3320 sbi->kbytes_written =
3321 le64_to_cpu(seg_i->journal->info.kbytes_written);
3323 f2fs_build_gc_manager(sbi);
3325 err = f2fs_build_stats(sbi);
3329 /* get an inode for node space */
3330 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3331 if (IS_ERR(sbi->node_inode)) {
3332 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
3333 err = PTR_ERR(sbi->node_inode);
3337 /* read root inode and dentry */
3338 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3340 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
3341 err = PTR_ERR(root);
3342 goto free_node_inode;
3344 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3345 !root->i_size || !root->i_nlink) {
3348 goto free_node_inode;
3351 sb->s_root = d_make_root(root); /* allocate root dentry */
3354 goto free_node_inode;
3357 err = f2fs_register_sysfs(sbi);
3359 goto free_root_inode;
3362 /* Enable quota usage during mount */
3363 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3364 err = f2fs_enable_quotas(sb);
3366 f2fs_msg(sb, KERN_ERR,
3367 "Cannot turn on quotas: error %d", err);
3370 /* if there are nt orphan nodes free them */
3371 err = f2fs_recover_orphan_inodes(sbi);
3375 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3376 goto reset_checkpoint;
3378 /* recover fsynced data */
3379 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
3381 * mount should be failed, when device has readonly mode, and
3382 * previous checkpoint was not done by clean system shutdown.
3384 if (f2fs_hw_is_readonly(sbi)) {
3385 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3387 f2fs_msg(sb, KERN_ERR,
3388 "Need to recover fsync data, but "
3389 "write access unavailable");
3392 f2fs_msg(sbi->sb, KERN_INFO, "write access "
3393 "unavailable, skipping recovery");
3394 goto reset_checkpoint;
3398 set_sbi_flag(sbi, SBI_NEED_FSCK);
3401 goto reset_checkpoint;
3403 err = f2fs_recover_fsync_data(sbi, false);
3406 skip_recovery = true;
3408 f2fs_msg(sb, KERN_ERR,
3409 "Cannot recover all fsync data errno=%d", err);
3413 err = f2fs_recover_fsync_data(sbi, true);
3415 if (!f2fs_readonly(sb) && err > 0) {
3417 f2fs_msg(sb, KERN_ERR,
3418 "Need to recover fsync data");
3423 /* f2fs_recover_fsync_data() cleared this already */
3424 clear_sbi_flag(sbi, SBI_POR_DOING);
3426 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3427 err = f2fs_disable_checkpoint(sbi);
3429 goto sync_free_meta;
3430 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3431 f2fs_enable_checkpoint(sbi);
3435 * If filesystem is not mounted as read-only then
3436 * do start the gc_thread.
3438 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
3439 /* After POR, we can run background GC thread.*/
3440 err = f2fs_start_gc_thread(sbi);
3442 goto sync_free_meta;
3446 /* recover broken superblock */
3448 err = f2fs_commit_super(sbi, true);
3449 f2fs_msg(sb, KERN_INFO,
3450 "Try to recover %dth superblock, ret: %d",
3451 sbi->valid_super_block ? 1 : 2, err);
3454 f2fs_join_shrinker(sbi);
3456 f2fs_tuning_parameters(sbi);
3458 f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx",
3459 cur_cp_version(F2FS_CKPT(sbi)));
3460 f2fs_update_time(sbi, CP_TIME);
3461 f2fs_update_time(sbi, REQ_TIME);
3462 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3466 /* safe to flush all the data */
3467 sync_filesystem(sbi->sb);
3472 f2fs_truncate_quota_inode_pages(sb);
3473 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3474 f2fs_quota_off_umount(sbi->sb);
3477 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3478 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3479 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3480 * falls into an infinite loop in f2fs_sync_meta_pages().
3482 truncate_inode_pages_final(META_MAPPING(sbi));
3483 /* evict some inodes being cached by GC */
3485 f2fs_unregister_sysfs(sbi);
3490 f2fs_release_ino_entry(sbi, true);
3491 truncate_inode_pages_final(NODE_MAPPING(sbi));
3492 iput(sbi->node_inode);
3493 sbi->node_inode = NULL;
3495 f2fs_destroy_stats(sbi);
3497 f2fs_destroy_node_manager(sbi);
3499 f2fs_destroy_segment_manager(sbi);
3501 destroy_device_list(sbi);
3504 make_bad_inode(sbi->meta_inode);
3505 iput(sbi->meta_inode);
3506 sbi->meta_inode = NULL;
3508 mempool_destroy(sbi->write_io_dummy);
3510 destroy_percpu_info(sbi);
3512 for (i = 0; i < NR_PAGE_TYPE; i++)
3513 kvfree(sbi->write_io[i]);
3516 for (i = 0; i < MAXQUOTAS; i++)
3517 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
3523 if (sbi->s_chksum_driver)
3524 crypto_free_shash(sbi->s_chksum_driver);
3527 /* give only one another chance */
3528 if (retry_cnt > 0 && skip_recovery) {
3530 shrink_dcache_sb(sb);
3536 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3537 const char *dev_name, void *data)
3539 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3542 static void kill_f2fs_super(struct super_block *sb)
3545 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3547 set_sbi_flag(sbi, SBI_IS_CLOSE);
3548 f2fs_stop_gc_thread(sbi);
3549 f2fs_stop_discard_thread(sbi);
3551 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3552 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3553 struct cp_control cpc = {
3554 .reason = CP_UMOUNT,
3556 f2fs_write_checkpoint(sbi, &cpc);
3559 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3560 sb->s_flags &= ~SB_RDONLY;
3562 kill_block_super(sb);
3565 static struct file_system_type f2fs_fs_type = {
3566 .owner = THIS_MODULE,
3568 .mount = f2fs_mount,
3569 .kill_sb = kill_f2fs_super,
3570 .fs_flags = FS_REQUIRES_DEV,
3572 MODULE_ALIAS_FS("f2fs");
3574 static int __init init_inodecache(void)
3576 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3577 sizeof(struct f2fs_inode_info), 0,
3578 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3579 if (!f2fs_inode_cachep)
3584 static void destroy_inodecache(void)
3587 * Make sure all delayed rcu free inodes are flushed before we
3591 kmem_cache_destroy(f2fs_inode_cachep);
3594 static int __init init_f2fs_fs(void)
3598 if (PAGE_SIZE != F2FS_BLKSIZE) {
3599 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3600 PAGE_SIZE, F2FS_BLKSIZE);
3604 f2fs_build_trace_ios();
3606 err = init_inodecache();
3609 err = f2fs_create_node_manager_caches();
3611 goto free_inodecache;
3612 err = f2fs_create_segment_manager_caches();
3614 goto free_node_manager_caches;
3615 err = f2fs_create_checkpoint_caches();
3617 goto free_segment_manager_caches;
3618 err = f2fs_create_extent_cache();
3620 goto free_checkpoint_caches;
3621 err = f2fs_init_sysfs();
3623 goto free_extent_cache;
3624 err = register_shrinker(&f2fs_shrinker_info);
3627 err = register_filesystem(&f2fs_fs_type);
3630 f2fs_create_root_stats();
3631 err = f2fs_init_post_read_processing();
3633 goto free_root_stats;
3637 f2fs_destroy_root_stats();
3638 unregister_filesystem(&f2fs_fs_type);
3640 unregister_shrinker(&f2fs_shrinker_info);
3644 f2fs_destroy_extent_cache();
3645 free_checkpoint_caches:
3646 f2fs_destroy_checkpoint_caches();
3647 free_segment_manager_caches:
3648 f2fs_destroy_segment_manager_caches();
3649 free_node_manager_caches:
3650 f2fs_destroy_node_manager_caches();
3652 destroy_inodecache();
3657 static void __exit exit_f2fs_fs(void)
3659 f2fs_destroy_post_read_processing();
3660 f2fs_destroy_root_stats();
3661 unregister_filesystem(&f2fs_fs_type);
3662 unregister_shrinker(&f2fs_shrinker_info);
3664 f2fs_destroy_extent_cache();
3665 f2fs_destroy_checkpoint_caches();
3666 f2fs_destroy_segment_manager_caches();
3667 f2fs_destroy_node_manager_caches();
3668 destroy_inodecache();
3669 f2fs_destroy_trace_ios();
3672 module_init(init_f2fs_fs)
3673 module_exit(exit_f2fs_fs)
3675 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3676 MODULE_DESCRIPTION("Flash Friendly File System");
3677 MODULE_LICENSE("GPL");