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 > __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(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 if (!f2fs_sb_has_extra_attr(sbi) ||
825 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
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 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
919 inode, NULL, 0, DATA);
920 truncate_inode_pages_final(inode->i_mapping);
922 if (F2FS_HAS_BLOCKS(inode))
923 f2fs_truncate(inode);
925 sb_end_intwrite(inode->i_sb);
927 spin_lock(&inode->i_lock);
928 atomic_dec(&inode->i_count);
930 trace_f2fs_drop_inode(inode, 0);
933 ret = generic_drop_inode(inode);
934 trace_f2fs_drop_inode(inode, ret);
938 int f2fs_inode_dirtied(struct inode *inode, bool sync)
940 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
943 spin_lock(&sbi->inode_lock[DIRTY_META]);
944 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
947 set_inode_flag(inode, FI_DIRTY_INODE);
948 stat_inc_dirty_inode(sbi, DIRTY_META);
950 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
951 list_add_tail(&F2FS_I(inode)->gdirty_list,
952 &sbi->inode_list[DIRTY_META]);
953 inc_page_count(sbi, F2FS_DIRTY_IMETA);
955 spin_unlock(&sbi->inode_lock[DIRTY_META]);
959 void f2fs_inode_synced(struct inode *inode)
961 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
963 spin_lock(&sbi->inode_lock[DIRTY_META]);
964 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
965 spin_unlock(&sbi->inode_lock[DIRTY_META]);
968 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
969 list_del_init(&F2FS_I(inode)->gdirty_list);
970 dec_page_count(sbi, F2FS_DIRTY_IMETA);
972 clear_inode_flag(inode, FI_DIRTY_INODE);
973 clear_inode_flag(inode, FI_AUTO_RECOVER);
974 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
975 spin_unlock(&sbi->inode_lock[DIRTY_META]);
979 * f2fs_dirty_inode() is called from __mark_inode_dirty()
981 * We should call set_dirty_inode to write the dirty inode through write_inode.
983 static void f2fs_dirty_inode(struct inode *inode, int flags)
985 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
987 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
988 inode->i_ino == F2FS_META_INO(sbi))
991 if (flags == I_DIRTY_TIME)
994 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
995 clear_inode_flag(inode, FI_AUTO_RECOVER);
997 f2fs_inode_dirtied(inode, false);
1000 static void f2fs_i_callback(struct rcu_head *head)
1002 struct inode *inode = container_of(head, struct inode, i_rcu);
1003 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1006 static void f2fs_destroy_inode(struct inode *inode)
1008 call_rcu(&inode->i_rcu, f2fs_i_callback);
1011 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1013 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1014 percpu_counter_destroy(&sbi->total_valid_inode_count);
1017 static void destroy_device_list(struct f2fs_sb_info *sbi)
1021 for (i = 0; i < sbi->s_ndevs; i++) {
1022 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1023 #ifdef CONFIG_BLK_DEV_ZONED
1024 kvfree(FDEV(i).blkz_type);
1030 static void f2fs_put_super(struct super_block *sb)
1032 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1036 f2fs_quota_off_umount(sb);
1038 /* prevent remaining shrinker jobs */
1039 mutex_lock(&sbi->umount_mutex);
1042 * We don't need to do checkpoint when superblock is clean.
1043 * But, the previous checkpoint was not done by umount, it needs to do
1044 * clean checkpoint again.
1046 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1047 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1048 struct cp_control cpc = {
1049 .reason = CP_UMOUNT,
1051 f2fs_write_checkpoint(sbi, &cpc);
1054 /* be sure to wait for any on-going discard commands */
1055 dropped = f2fs_issue_discard_timeout(sbi);
1057 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1058 !sbi->discard_blks && !dropped) {
1059 struct cp_control cpc = {
1060 .reason = CP_UMOUNT | CP_TRIMMED,
1062 f2fs_write_checkpoint(sbi, &cpc);
1066 * normally superblock is clean, so we need to release this.
1067 * In addition, EIO will skip do checkpoint, we need this as well.
1069 f2fs_release_ino_entry(sbi, true);
1071 f2fs_leave_shrinker(sbi);
1072 mutex_unlock(&sbi->umount_mutex);
1074 /* our cp_error case, we can wait for any writeback page */
1075 f2fs_flush_merged_writes(sbi);
1077 f2fs_wait_on_all_pages_writeback(sbi);
1079 f2fs_bug_on(sbi, sbi->fsync_node_num);
1081 iput(sbi->node_inode);
1082 sbi->node_inode = NULL;
1084 iput(sbi->meta_inode);
1085 sbi->meta_inode = NULL;
1088 * iput() can update stat information, if f2fs_write_checkpoint()
1089 * above failed with error.
1091 f2fs_destroy_stats(sbi);
1093 /* destroy f2fs internal modules */
1094 f2fs_destroy_node_manager(sbi);
1095 f2fs_destroy_segment_manager(sbi);
1099 f2fs_unregister_sysfs(sbi);
1101 sb->s_fs_info = NULL;
1102 if (sbi->s_chksum_driver)
1103 crypto_free_shash(sbi->s_chksum_driver);
1104 kvfree(sbi->raw_super);
1106 destroy_device_list(sbi);
1107 mempool_destroy(sbi->write_io_dummy);
1109 for (i = 0; i < MAXQUOTAS; i++)
1110 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1112 destroy_percpu_info(sbi);
1113 for (i = 0; i < NR_PAGE_TYPE; i++)
1114 kvfree(sbi->write_io[i]);
1118 int f2fs_sync_fs(struct super_block *sb, int sync)
1120 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1123 if (unlikely(f2fs_cp_error(sbi)))
1125 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1128 trace_f2fs_sync_fs(sb, sync);
1130 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1134 struct cp_control cpc;
1136 cpc.reason = __get_cp_reason(sbi);
1138 mutex_lock(&sbi->gc_mutex);
1139 err = f2fs_write_checkpoint(sbi, &cpc);
1140 mutex_unlock(&sbi->gc_mutex);
1142 f2fs_trace_ios(NULL, 1);
1147 static int f2fs_freeze(struct super_block *sb)
1149 if (f2fs_readonly(sb))
1152 /* IO error happened before */
1153 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1156 /* must be clean, since sync_filesystem() was already called */
1157 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1162 static int f2fs_unfreeze(struct super_block *sb)
1168 static int f2fs_statfs_project(struct super_block *sb,
1169 kprojid_t projid, struct kstatfs *buf)
1172 struct dquot *dquot;
1176 qid = make_kqid_projid(projid);
1177 dquot = dqget(sb, qid);
1179 return PTR_ERR(dquot);
1180 spin_lock(&dquot->dq_dqb_lock);
1182 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
1183 dquot->dq_dqb.dqb_bsoftlimit :
1184 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
1185 if (limit && buf->f_blocks > limit) {
1186 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1187 buf->f_blocks = limit;
1188 buf->f_bfree = buf->f_bavail =
1189 (buf->f_blocks > curblock) ?
1190 (buf->f_blocks - curblock) : 0;
1193 limit = dquot->dq_dqb.dqb_isoftlimit ?
1194 dquot->dq_dqb.dqb_isoftlimit :
1195 dquot->dq_dqb.dqb_ihardlimit;
1196 if (limit && buf->f_files > limit) {
1197 buf->f_files = limit;
1199 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1200 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1203 spin_unlock(&dquot->dq_dqb_lock);
1209 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1211 struct super_block *sb = dentry->d_sb;
1212 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1213 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1214 block_t total_count, user_block_count, start_count;
1215 u64 avail_node_count;
1217 total_count = le64_to_cpu(sbi->raw_super->block_count);
1218 user_block_count = sbi->user_block_count;
1219 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1220 buf->f_type = F2FS_SUPER_MAGIC;
1221 buf->f_bsize = sbi->blocksize;
1223 buf->f_blocks = total_count - start_count;
1224 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1225 sbi->current_reserved_blocks;
1226 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1229 buf->f_bfree -= sbi->unusable_block_count;
1231 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1232 buf->f_bavail = buf->f_bfree -
1233 F2FS_OPTION(sbi).root_reserved_blocks;
1237 avail_node_count = sbi->total_node_count - sbi->nquota_files -
1238 F2FS_RESERVED_NODE_NUM;
1240 if (avail_node_count > user_block_count) {
1241 buf->f_files = user_block_count;
1242 buf->f_ffree = buf->f_bavail;
1244 buf->f_files = avail_node_count;
1245 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1249 buf->f_namelen = F2FS_NAME_LEN;
1250 buf->f_fsid.val[0] = (u32)id;
1251 buf->f_fsid.val[1] = (u32)(id >> 32);
1254 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1255 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1256 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1262 static inline void f2fs_show_quota_options(struct seq_file *seq,
1263 struct super_block *sb)
1266 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1268 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1271 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1282 seq_printf(seq, ",jqfmt=%s", fmtname);
1285 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1286 seq_show_option(seq, "usrjquota",
1287 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1289 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1290 seq_show_option(seq, "grpjquota",
1291 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1293 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1294 seq_show_option(seq, "prjjquota",
1295 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1299 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1301 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1303 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1304 if (test_opt(sbi, FORCE_FG_GC))
1305 seq_printf(seq, ",background_gc=%s", "sync");
1307 seq_printf(seq, ",background_gc=%s", "on");
1309 seq_printf(seq, ",background_gc=%s", "off");
1311 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1312 seq_puts(seq, ",disable_roll_forward");
1313 if (test_opt(sbi, DISCARD))
1314 seq_puts(seq, ",discard");
1315 if (test_opt(sbi, NOHEAP))
1316 seq_puts(seq, ",no_heap");
1318 seq_puts(seq, ",heap");
1319 #ifdef CONFIG_F2FS_FS_XATTR
1320 if (test_opt(sbi, XATTR_USER))
1321 seq_puts(seq, ",user_xattr");
1323 seq_puts(seq, ",nouser_xattr");
1324 if (test_opt(sbi, INLINE_XATTR))
1325 seq_puts(seq, ",inline_xattr");
1327 seq_puts(seq, ",noinline_xattr");
1328 if (test_opt(sbi, INLINE_XATTR_SIZE))
1329 seq_printf(seq, ",inline_xattr_size=%u",
1330 F2FS_OPTION(sbi).inline_xattr_size);
1332 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1333 if (test_opt(sbi, POSIX_ACL))
1334 seq_puts(seq, ",acl");
1336 seq_puts(seq, ",noacl");
1338 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1339 seq_puts(seq, ",disable_ext_identify");
1340 if (test_opt(sbi, INLINE_DATA))
1341 seq_puts(seq, ",inline_data");
1343 seq_puts(seq, ",noinline_data");
1344 if (test_opt(sbi, INLINE_DENTRY))
1345 seq_puts(seq, ",inline_dentry");
1347 seq_puts(seq, ",noinline_dentry");
1348 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1349 seq_puts(seq, ",flush_merge");
1350 if (test_opt(sbi, NOBARRIER))
1351 seq_puts(seq, ",nobarrier");
1352 if (test_opt(sbi, FASTBOOT))
1353 seq_puts(seq, ",fastboot");
1354 if (test_opt(sbi, EXTENT_CACHE))
1355 seq_puts(seq, ",extent_cache");
1357 seq_puts(seq, ",noextent_cache");
1358 if (test_opt(sbi, DATA_FLUSH))
1359 seq_puts(seq, ",data_flush");
1361 seq_puts(seq, ",mode=");
1362 if (test_opt(sbi, ADAPTIVE))
1363 seq_puts(seq, "adaptive");
1364 else if (test_opt(sbi, LFS))
1365 seq_puts(seq, "lfs");
1366 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1367 if (test_opt(sbi, RESERVE_ROOT))
1368 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1369 F2FS_OPTION(sbi).root_reserved_blocks,
1370 from_kuid_munged(&init_user_ns,
1371 F2FS_OPTION(sbi).s_resuid),
1372 from_kgid_munged(&init_user_ns,
1373 F2FS_OPTION(sbi).s_resgid));
1374 if (F2FS_IO_SIZE_BITS(sbi))
1375 seq_printf(seq, ",io_bits=%u",
1376 F2FS_OPTION(sbi).write_io_size_bits);
1377 #ifdef CONFIG_F2FS_FAULT_INJECTION
1378 if (test_opt(sbi, FAULT_INJECTION)) {
1379 seq_printf(seq, ",fault_injection=%u",
1380 F2FS_OPTION(sbi).fault_info.inject_rate);
1381 seq_printf(seq, ",fault_type=%u",
1382 F2FS_OPTION(sbi).fault_info.inject_type);
1386 if (test_opt(sbi, QUOTA))
1387 seq_puts(seq, ",quota");
1388 if (test_opt(sbi, USRQUOTA))
1389 seq_puts(seq, ",usrquota");
1390 if (test_opt(sbi, GRPQUOTA))
1391 seq_puts(seq, ",grpquota");
1392 if (test_opt(sbi, PRJQUOTA))
1393 seq_puts(seq, ",prjquota");
1395 f2fs_show_quota_options(seq, sbi->sb);
1396 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1397 seq_printf(seq, ",whint_mode=%s", "user-based");
1398 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1399 seq_printf(seq, ",whint_mode=%s", "fs-based");
1400 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1401 if (F2FS_OPTION(sbi).test_dummy_encryption)
1402 seq_puts(seq, ",test_dummy_encryption");
1405 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1406 seq_printf(seq, ",alloc_mode=%s", "default");
1407 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1408 seq_printf(seq, ",alloc_mode=%s", "reuse");
1410 if (test_opt(sbi, DISABLE_CHECKPOINT))
1411 seq_puts(seq, ",checkpoint=disable");
1413 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1414 seq_printf(seq, ",fsync_mode=%s", "posix");
1415 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1416 seq_printf(seq, ",fsync_mode=%s", "strict");
1417 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1418 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1422 static void default_options(struct f2fs_sb_info *sbi)
1424 /* init some FS parameters */
1425 F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1426 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1427 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1428 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1429 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1430 F2FS_OPTION(sbi).test_dummy_encryption = false;
1431 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1432 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1434 set_opt(sbi, BG_GC);
1435 set_opt(sbi, INLINE_XATTR);
1436 set_opt(sbi, INLINE_DATA);
1437 set_opt(sbi, INLINE_DENTRY);
1438 set_opt(sbi, EXTENT_CACHE);
1439 set_opt(sbi, NOHEAP);
1440 clear_opt(sbi, DISABLE_CHECKPOINT);
1441 sbi->sb->s_flags |= SB_LAZYTIME;
1442 set_opt(sbi, FLUSH_MERGE);
1443 set_opt(sbi, DISCARD);
1444 if (f2fs_sb_has_blkzoned(sbi))
1445 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1447 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1449 #ifdef CONFIG_F2FS_FS_XATTR
1450 set_opt(sbi, XATTR_USER);
1452 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1453 set_opt(sbi, POSIX_ACL);
1456 f2fs_build_fault_attr(sbi, 0, 0);
1460 static int f2fs_enable_quotas(struct super_block *sb);
1463 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1465 unsigned int s_flags = sbi->sb->s_flags;
1466 struct cp_control cpc;
1470 if (s_flags & SB_RDONLY) {
1471 f2fs_msg(sbi->sb, KERN_ERR,
1472 "checkpoint=disable on readonly fs");
1475 sbi->sb->s_flags |= SB_ACTIVE;
1477 f2fs_update_time(sbi, DISABLE_TIME);
1479 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1480 mutex_lock(&sbi->gc_mutex);
1481 err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1482 if (err == -ENODATA) {
1486 if (err && err != -EAGAIN)
1490 ret = sync_filesystem(sbi->sb);
1492 err = ret ? ret: err;
1496 if (f2fs_disable_cp_again(sbi)) {
1501 mutex_lock(&sbi->gc_mutex);
1502 cpc.reason = CP_PAUSE;
1503 set_sbi_flag(sbi, SBI_CP_DISABLED);
1504 f2fs_write_checkpoint(sbi, &cpc);
1506 sbi->unusable_block_count = 0;
1507 mutex_unlock(&sbi->gc_mutex);
1509 sbi->sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1513 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1515 mutex_lock(&sbi->gc_mutex);
1516 f2fs_dirty_to_prefree(sbi);
1518 clear_sbi_flag(sbi, SBI_CP_DISABLED);
1519 set_sbi_flag(sbi, SBI_IS_DIRTY);
1520 mutex_unlock(&sbi->gc_mutex);
1522 f2fs_sync_fs(sbi->sb, 1);
1525 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1527 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1528 struct f2fs_mount_info org_mount_opt;
1529 unsigned long old_sb_flags;
1531 bool need_restart_gc = false;
1532 bool need_stop_gc = false;
1533 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1534 bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1535 bool checkpoint_changed;
1541 * Save the old mount options in case we
1542 * need to restore them.
1544 org_mount_opt = sbi->mount_opt;
1545 old_sb_flags = sb->s_flags;
1548 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1549 for (i = 0; i < MAXQUOTAS; i++) {
1550 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1551 org_mount_opt.s_qf_names[i] =
1552 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1554 if (!org_mount_opt.s_qf_names[i]) {
1555 for (j = 0; j < i; j++)
1556 kvfree(org_mount_opt.s_qf_names[j]);
1560 org_mount_opt.s_qf_names[i] = NULL;
1565 /* recover superblocks we couldn't write due to previous RO mount */
1566 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1567 err = f2fs_commit_super(sbi, false);
1568 f2fs_msg(sb, KERN_INFO,
1569 "Try to recover all the superblocks, ret: %d", err);
1571 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1574 default_options(sbi);
1576 /* parse mount options */
1577 err = parse_options(sb, data);
1580 checkpoint_changed =
1581 disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1584 * Previous and new state of filesystem is RO,
1585 * so skip checking GC and FLUSH_MERGE conditions.
1587 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1591 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1592 err = dquot_suspend(sb, -1);
1595 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1596 /* dquot_resume needs RW */
1597 sb->s_flags &= ~SB_RDONLY;
1598 if (sb_any_quota_suspended(sb)) {
1599 dquot_resume(sb, -1);
1600 } else if (f2fs_sb_has_quota_ino(sbi)) {
1601 err = f2fs_enable_quotas(sb);
1607 /* disallow enable/disable extent_cache dynamically */
1608 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1610 f2fs_msg(sbi->sb, KERN_WARNING,
1611 "switch extent_cache option is not allowed");
1615 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1617 f2fs_msg(sbi->sb, KERN_WARNING,
1618 "disabling checkpoint not compatible with read-only");
1623 * We stop the GC thread if FS is mounted as RO
1624 * or if background_gc = off is passed in mount
1625 * option. Also sync the filesystem.
1627 if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) {
1628 if (sbi->gc_thread) {
1629 f2fs_stop_gc_thread(sbi);
1630 need_restart_gc = true;
1632 } else if (!sbi->gc_thread) {
1633 err = f2fs_start_gc_thread(sbi);
1636 need_stop_gc = true;
1639 if (*flags & SB_RDONLY ||
1640 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1641 writeback_inodes_sb(sb, WB_REASON_SYNC);
1644 set_sbi_flag(sbi, SBI_IS_DIRTY);
1645 set_sbi_flag(sbi, SBI_IS_CLOSE);
1646 f2fs_sync_fs(sb, 1);
1647 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1650 if (checkpoint_changed) {
1651 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1652 err = f2fs_disable_checkpoint(sbi);
1656 f2fs_enable_checkpoint(sbi);
1661 * We stop issue flush thread if FS is mounted as RO
1662 * or if flush_merge is not passed in mount option.
1664 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1665 clear_opt(sbi, FLUSH_MERGE);
1666 f2fs_destroy_flush_cmd_control(sbi, false);
1668 err = f2fs_create_flush_cmd_control(sbi);
1674 /* Release old quota file names */
1675 for (i = 0; i < MAXQUOTAS; i++)
1676 kvfree(org_mount_opt.s_qf_names[i]);
1678 /* Update the POSIXACL Flag */
1679 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1680 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1682 limit_reserve_root(sbi);
1683 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1686 if (need_restart_gc) {
1687 if (f2fs_start_gc_thread(sbi))
1688 f2fs_msg(sbi->sb, KERN_WARNING,
1689 "background gc thread has stopped");
1690 } else if (need_stop_gc) {
1691 f2fs_stop_gc_thread(sbi);
1695 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1696 for (i = 0; i < MAXQUOTAS; i++) {
1697 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1698 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1701 sbi->mount_opt = org_mount_opt;
1702 sb->s_flags = old_sb_flags;
1707 /* Read data from quotafile */
1708 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1709 size_t len, loff_t off)
1711 struct inode *inode = sb_dqopt(sb)->files[type];
1712 struct address_space *mapping = inode->i_mapping;
1713 block_t blkidx = F2FS_BYTES_TO_BLK(off);
1714 int offset = off & (sb->s_blocksize - 1);
1717 loff_t i_size = i_size_read(inode);
1724 if (off + len > i_size)
1727 while (toread > 0) {
1728 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1730 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1732 if (PTR_ERR(page) == -ENOMEM) {
1733 congestion_wait(BLK_RW_ASYNC, HZ/50);
1736 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1737 return PTR_ERR(page);
1742 if (unlikely(page->mapping != mapping)) {
1743 f2fs_put_page(page, 1);
1746 if (unlikely(!PageUptodate(page))) {
1747 f2fs_put_page(page, 1);
1748 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1752 kaddr = kmap_atomic(page);
1753 memcpy(data, kaddr + offset, tocopy);
1754 kunmap_atomic(kaddr);
1755 f2fs_put_page(page, 1);
1765 /* Write to quotafile */
1766 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1767 const char *data, size_t len, loff_t off)
1769 struct inode *inode = sb_dqopt(sb)->files[type];
1770 struct address_space *mapping = inode->i_mapping;
1771 const struct address_space_operations *a_ops = mapping->a_ops;
1772 int offset = off & (sb->s_blocksize - 1);
1773 size_t towrite = len;
1779 while (towrite > 0) {
1780 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1783 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1785 if (unlikely(err)) {
1786 if (err == -ENOMEM) {
1787 congestion_wait(BLK_RW_ASYNC, HZ/50);
1790 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1794 kaddr = kmap_atomic(page);
1795 memcpy(kaddr + offset, data, tocopy);
1796 kunmap_atomic(kaddr);
1797 flush_dcache_page(page);
1799 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1810 inode->i_mtime = inode->i_ctime = current_time(inode);
1811 f2fs_mark_inode_dirty_sync(inode, false);
1812 return len - towrite;
1815 static struct dquot **f2fs_get_dquots(struct inode *inode)
1817 return F2FS_I(inode)->i_dquot;
1820 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1822 return &F2FS_I(inode)->i_reserved_quota;
1825 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1827 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
1828 f2fs_msg(sbi->sb, KERN_ERR,
1829 "quota sysfile may be corrupted, skip loading it");
1833 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
1834 F2FS_OPTION(sbi).s_jquota_fmt, type);
1837 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1842 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
1843 err = f2fs_enable_quotas(sbi->sb);
1845 f2fs_msg(sbi->sb, KERN_ERR,
1846 "Cannot turn on quota_ino: %d", err);
1852 for (i = 0; i < MAXQUOTAS; i++) {
1853 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1854 err = f2fs_quota_on_mount(sbi, i);
1859 f2fs_msg(sbi->sb, KERN_ERR,
1860 "Cannot turn on quotas: %d on %d", err, i);
1866 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
1869 struct inode *qf_inode;
1870 unsigned long qf_inum;
1873 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
1875 qf_inum = f2fs_qf_ino(sb, type);
1879 qf_inode = f2fs_iget(sb, qf_inum);
1880 if (IS_ERR(qf_inode)) {
1881 f2fs_msg(sb, KERN_ERR,
1882 "Bad quota inode %u:%lu", type, qf_inum);
1883 return PTR_ERR(qf_inode);
1886 /* Don't account quota for quota files to avoid recursion */
1887 qf_inode->i_flags |= S_NOQUOTA;
1888 err = dquot_enable(qf_inode, type, format_id, flags);
1893 static int f2fs_enable_quotas(struct super_block *sb)
1896 unsigned long qf_inum;
1897 bool quota_mopt[MAXQUOTAS] = {
1898 test_opt(F2FS_SB(sb), USRQUOTA),
1899 test_opt(F2FS_SB(sb), GRPQUOTA),
1900 test_opt(F2FS_SB(sb), PRJQUOTA),
1903 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
1904 f2fs_msg(sb, KERN_ERR,
1905 "quota file may be corrupted, skip loading it");
1909 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
1911 for (type = 0; type < MAXQUOTAS; type++) {
1912 qf_inum = f2fs_qf_ino(sb, type);
1914 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
1915 DQUOT_USAGE_ENABLED |
1916 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
1918 f2fs_msg(sb, KERN_ERR,
1919 "Failed to enable quota tracking "
1920 "(type=%d, err=%d). Please run "
1921 "fsck to fix.", type, err);
1922 for (type--; type >= 0; type--)
1923 dquot_quota_off(sb, type);
1924 set_sbi_flag(F2FS_SB(sb),
1925 SBI_QUOTA_NEED_REPAIR);
1933 int f2fs_quota_sync(struct super_block *sb, int type)
1935 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1936 struct quota_info *dqopt = sb_dqopt(sb);
1940 ret = dquot_writeback_dquots(sb, type);
1945 * Now when everything is written we can discard the pagecache so
1946 * that userspace sees the changes.
1948 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
1949 struct address_space *mapping;
1951 if (type != -1 && cnt != type)
1953 if (!sb_has_quota_active(sb, cnt))
1956 mapping = dqopt->files[cnt]->i_mapping;
1958 ret = filemap_fdatawrite(mapping);
1962 /* if we are using journalled quota */
1963 if (is_journalled_quota(sbi))
1966 ret = filemap_fdatawait(mapping);
1968 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1970 inode_lock(dqopt->files[cnt]);
1971 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
1972 inode_unlock(dqopt->files[cnt]);
1976 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1980 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
1981 const struct path *path)
1983 struct inode *inode;
1986 err = f2fs_quota_sync(sb, type);
1990 err = dquot_quota_on(sb, type, format_id, path);
1994 inode = d_inode(path->dentry);
1997 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
1998 f2fs_set_inode_flags(inode);
1999 inode_unlock(inode);
2000 f2fs_mark_inode_dirty_sync(inode, false);
2005 static int f2fs_quota_off(struct super_block *sb, int type)
2007 struct inode *inode = sb_dqopt(sb)->files[type];
2010 if (!inode || !igrab(inode))
2011 return dquot_quota_off(sb, type);
2013 err = f2fs_quota_sync(sb, type);
2017 err = dquot_quota_off(sb, type);
2018 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2022 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2023 f2fs_set_inode_flags(inode);
2024 inode_unlock(inode);
2025 f2fs_mark_inode_dirty_sync(inode, false);
2031 void f2fs_quota_off_umount(struct super_block *sb)
2036 for (type = 0; type < MAXQUOTAS; type++) {
2037 err = f2fs_quota_off(sb, type);
2039 int ret = dquot_quota_off(sb, type);
2041 f2fs_msg(sb, KERN_ERR,
2042 "Fail to turn off disk quota "
2043 "(type: %d, err: %d, ret:%d), Please "
2044 "run fsck to fix it.", type, err, ret);
2045 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2049 * In case of checkpoint=disable, we must flush quota blocks.
2050 * This can cause NULL exception for node_inode in end_io, since
2051 * put_super already dropped it.
2053 sync_filesystem(sb);
2056 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2058 struct quota_info *dqopt = sb_dqopt(sb);
2061 for (type = 0; type < MAXQUOTAS; type++) {
2062 if (!dqopt->files[type])
2064 f2fs_inode_synced(dqopt->files[type]);
2068 static int f2fs_dquot_commit(struct dquot *dquot)
2072 ret = dquot_commit(dquot);
2074 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2078 static int f2fs_dquot_acquire(struct dquot *dquot)
2082 ret = dquot_acquire(dquot);
2084 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2089 static int f2fs_dquot_release(struct dquot *dquot)
2093 ret = dquot_release(dquot);
2095 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2099 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2101 struct super_block *sb = dquot->dq_sb;
2102 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2105 ret = dquot_mark_dquot_dirty(dquot);
2107 /* if we are using journalled quota */
2108 if (is_journalled_quota(sbi))
2109 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2114 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2118 ret = dquot_commit_info(sb, type);
2120 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2124 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2126 *projid = F2FS_I(inode)->i_projid;
2130 static const struct dquot_operations f2fs_quota_operations = {
2131 .get_reserved_space = f2fs_get_reserved_space,
2132 .write_dquot = f2fs_dquot_commit,
2133 .acquire_dquot = f2fs_dquot_acquire,
2134 .release_dquot = f2fs_dquot_release,
2135 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
2136 .write_info = f2fs_dquot_commit_info,
2137 .alloc_dquot = dquot_alloc,
2138 .destroy_dquot = dquot_destroy,
2139 .get_projid = f2fs_get_projid,
2140 .get_next_id = dquot_get_next_id,
2143 static const struct quotactl_ops f2fs_quotactl_ops = {
2144 .quota_on = f2fs_quota_on,
2145 .quota_off = f2fs_quota_off,
2146 .quota_sync = f2fs_quota_sync,
2147 .get_state = dquot_get_state,
2148 .set_info = dquot_set_dqinfo,
2149 .get_dqblk = dquot_get_dqblk,
2150 .set_dqblk = dquot_set_dqblk,
2151 .get_nextdqblk = dquot_get_next_dqblk,
2154 int f2fs_quota_sync(struct super_block *sb, int type)
2159 void f2fs_quota_off_umount(struct super_block *sb)
2164 static const struct super_operations f2fs_sops = {
2165 .alloc_inode = f2fs_alloc_inode,
2166 .drop_inode = f2fs_drop_inode,
2167 .destroy_inode = f2fs_destroy_inode,
2168 .write_inode = f2fs_write_inode,
2169 .dirty_inode = f2fs_dirty_inode,
2170 .show_options = f2fs_show_options,
2172 .quota_read = f2fs_quota_read,
2173 .quota_write = f2fs_quota_write,
2174 .get_dquots = f2fs_get_dquots,
2176 .evict_inode = f2fs_evict_inode,
2177 .put_super = f2fs_put_super,
2178 .sync_fs = f2fs_sync_fs,
2179 .freeze_fs = f2fs_freeze,
2180 .unfreeze_fs = f2fs_unfreeze,
2181 .statfs = f2fs_statfs,
2182 .remount_fs = f2fs_remount,
2185 #ifdef CONFIG_F2FS_FS_ENCRYPTION
2186 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2188 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2189 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2193 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2196 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2199 * Encrypting the root directory is not allowed because fsck
2200 * expects lost+found directory to exist and remain unencrypted
2201 * if LOST_FOUND feature is enabled.
2204 if (f2fs_sb_has_lost_found(sbi) &&
2205 inode->i_ino == F2FS_ROOT_INO(sbi))
2208 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2209 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2210 ctx, len, fs_data, XATTR_CREATE);
2213 static bool f2fs_dummy_context(struct inode *inode)
2215 return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
2218 static const struct fscrypt_operations f2fs_cryptops = {
2219 .key_prefix = "f2fs:",
2220 .get_context = f2fs_get_context,
2221 .set_context = f2fs_set_context,
2222 .dummy_context = f2fs_dummy_context,
2223 .empty_dir = f2fs_empty_dir,
2224 .max_namelen = F2FS_NAME_LEN,
2228 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2229 u64 ino, u32 generation)
2231 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2232 struct inode *inode;
2234 if (f2fs_check_nid_range(sbi, ino))
2235 return ERR_PTR(-ESTALE);
2238 * f2fs_iget isn't quite right if the inode is currently unallocated!
2239 * However f2fs_iget currently does appropriate checks to handle stale
2240 * inodes so everything is OK.
2242 inode = f2fs_iget(sb, ino);
2244 return ERR_CAST(inode);
2245 if (unlikely(generation && inode->i_generation != generation)) {
2246 /* we didn't find the right inode.. */
2248 return ERR_PTR(-ESTALE);
2253 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2254 int fh_len, int fh_type)
2256 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2257 f2fs_nfs_get_inode);
2260 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2261 int fh_len, int fh_type)
2263 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2264 f2fs_nfs_get_inode);
2267 static const struct export_operations f2fs_export_ops = {
2268 .fh_to_dentry = f2fs_fh_to_dentry,
2269 .fh_to_parent = f2fs_fh_to_parent,
2270 .get_parent = f2fs_get_parent,
2273 static loff_t max_file_blocks(void)
2276 loff_t leaf_count = ADDRS_PER_BLOCK;
2279 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2280 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2281 * space in inode.i_addr, it will be more safe to reassign
2285 /* two direct node blocks */
2286 result += (leaf_count * 2);
2288 /* two indirect node blocks */
2289 leaf_count *= NIDS_PER_BLOCK;
2290 result += (leaf_count * 2);
2292 /* one double indirect node block */
2293 leaf_count *= NIDS_PER_BLOCK;
2294 result += leaf_count;
2299 static int __f2fs_commit_super(struct buffer_head *bh,
2300 struct f2fs_super_block *super)
2304 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2305 set_buffer_dirty(bh);
2308 /* it's rare case, we can do fua all the time */
2309 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2312 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2313 struct buffer_head *bh)
2315 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2316 (bh->b_data + F2FS_SUPER_OFFSET);
2317 struct super_block *sb = sbi->sb;
2318 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2319 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2320 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2321 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2322 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2323 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2324 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2325 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2326 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2327 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2328 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2329 u32 segment_count = le32_to_cpu(raw_super->segment_count);
2330 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2331 u64 main_end_blkaddr = main_blkaddr +
2332 (segment_count_main << log_blocks_per_seg);
2333 u64 seg_end_blkaddr = segment0_blkaddr +
2334 (segment_count << log_blocks_per_seg);
2336 if (segment0_blkaddr != cp_blkaddr) {
2337 f2fs_msg(sb, KERN_INFO,
2338 "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2339 segment0_blkaddr, cp_blkaddr);
2343 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2345 f2fs_msg(sb, KERN_INFO,
2346 "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2347 cp_blkaddr, sit_blkaddr,
2348 segment_count_ckpt << log_blocks_per_seg);
2352 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2354 f2fs_msg(sb, KERN_INFO,
2355 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2356 sit_blkaddr, nat_blkaddr,
2357 segment_count_sit << log_blocks_per_seg);
2361 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2363 f2fs_msg(sb, KERN_INFO,
2364 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2365 nat_blkaddr, ssa_blkaddr,
2366 segment_count_nat << log_blocks_per_seg);
2370 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2372 f2fs_msg(sb, KERN_INFO,
2373 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2374 ssa_blkaddr, main_blkaddr,
2375 segment_count_ssa << log_blocks_per_seg);
2379 if (main_end_blkaddr > seg_end_blkaddr) {
2380 f2fs_msg(sb, KERN_INFO,
2381 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2384 (segment_count << log_blocks_per_seg),
2385 segment_count_main << log_blocks_per_seg);
2387 } else if (main_end_blkaddr < seg_end_blkaddr) {
2391 /* fix in-memory information all the time */
2392 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2393 segment0_blkaddr) >> log_blocks_per_seg);
2395 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2396 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2399 err = __f2fs_commit_super(bh, NULL);
2400 res = err ? "failed" : "done";
2402 f2fs_msg(sb, KERN_INFO,
2403 "Fix alignment : %s, start(%u) end(%u) block(%u)",
2406 (segment_count << log_blocks_per_seg),
2407 segment_count_main << log_blocks_per_seg);
2414 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2415 struct buffer_head *bh)
2417 block_t segment_count, segs_per_sec, secs_per_zone;
2418 block_t total_sections, blocks_per_seg;
2419 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2420 (bh->b_data + F2FS_SUPER_OFFSET);
2421 struct super_block *sb = sbi->sb;
2422 unsigned int blocksize;
2423 size_t crc_offset = 0;
2426 /* Check checksum_offset and crc in superblock */
2427 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2428 crc_offset = le32_to_cpu(raw_super->checksum_offset);
2430 offsetof(struct f2fs_super_block, crc)) {
2431 f2fs_msg(sb, KERN_INFO,
2432 "Invalid SB checksum offset: %zu",
2436 crc = le32_to_cpu(raw_super->crc);
2437 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2438 f2fs_msg(sb, KERN_INFO,
2439 "Invalid SB checksum value: %u", crc);
2444 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
2445 f2fs_msg(sb, KERN_INFO,
2446 "Magic Mismatch, valid(0x%x) - read(0x%x)",
2447 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2451 /* Currently, support only 4KB page cache size */
2452 if (F2FS_BLKSIZE != PAGE_SIZE) {
2453 f2fs_msg(sb, KERN_INFO,
2454 "Invalid page_cache_size (%lu), supports only 4KB\n",
2459 /* Currently, support only 4KB block size */
2460 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2461 if (blocksize != F2FS_BLKSIZE) {
2462 f2fs_msg(sb, KERN_INFO,
2463 "Invalid blocksize (%u), supports only 4KB\n",
2468 /* check log blocks per segment */
2469 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2470 f2fs_msg(sb, KERN_INFO,
2471 "Invalid log blocks per segment (%u)\n",
2472 le32_to_cpu(raw_super->log_blocks_per_seg));
2476 /* Currently, support 512/1024/2048/4096 bytes sector size */
2477 if (le32_to_cpu(raw_super->log_sectorsize) >
2478 F2FS_MAX_LOG_SECTOR_SIZE ||
2479 le32_to_cpu(raw_super->log_sectorsize) <
2480 F2FS_MIN_LOG_SECTOR_SIZE) {
2481 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
2482 le32_to_cpu(raw_super->log_sectorsize));
2485 if (le32_to_cpu(raw_super->log_sectors_per_block) +
2486 le32_to_cpu(raw_super->log_sectorsize) !=
2487 F2FS_MAX_LOG_SECTOR_SIZE) {
2488 f2fs_msg(sb, KERN_INFO,
2489 "Invalid log sectors per block(%u) log sectorsize(%u)",
2490 le32_to_cpu(raw_super->log_sectors_per_block),
2491 le32_to_cpu(raw_super->log_sectorsize));
2495 segment_count = le32_to_cpu(raw_super->segment_count);
2496 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2497 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2498 total_sections = le32_to_cpu(raw_super->section_count);
2500 /* blocks_per_seg should be 512, given the above check */
2501 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2503 if (segment_count > F2FS_MAX_SEGMENT ||
2504 segment_count < F2FS_MIN_SEGMENTS) {
2505 f2fs_msg(sb, KERN_INFO,
2506 "Invalid segment count (%u)",
2511 if (total_sections > segment_count ||
2512 total_sections < F2FS_MIN_SEGMENTS ||
2513 segs_per_sec > segment_count || !segs_per_sec) {
2514 f2fs_msg(sb, KERN_INFO,
2515 "Invalid segment/section count (%u, %u x %u)",
2516 segment_count, total_sections, segs_per_sec);
2520 if ((segment_count / segs_per_sec) < total_sections) {
2521 f2fs_msg(sb, KERN_INFO,
2522 "Small segment_count (%u < %u * %u)",
2523 segment_count, segs_per_sec, total_sections);
2527 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2528 f2fs_msg(sb, KERN_INFO,
2529 "Wrong segment_count / block_count (%u > %llu)",
2530 segment_count, le64_to_cpu(raw_super->block_count));
2534 if (secs_per_zone > total_sections || !secs_per_zone) {
2535 f2fs_msg(sb, KERN_INFO,
2536 "Wrong secs_per_zone / total_sections (%u, %u)",
2537 secs_per_zone, total_sections);
2540 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2541 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2542 (le32_to_cpu(raw_super->extension_count) +
2543 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2544 f2fs_msg(sb, KERN_INFO,
2545 "Corrupted extension count (%u + %u > %u)",
2546 le32_to_cpu(raw_super->extension_count),
2547 raw_super->hot_ext_count,
2548 F2FS_MAX_EXTENSION);
2552 if (le32_to_cpu(raw_super->cp_payload) >
2553 (blocks_per_seg - F2FS_CP_PACKS)) {
2554 f2fs_msg(sb, KERN_INFO,
2555 "Insane cp_payload (%u > %u)",
2556 le32_to_cpu(raw_super->cp_payload),
2557 blocks_per_seg - F2FS_CP_PACKS);
2561 /* check reserved ino info */
2562 if (le32_to_cpu(raw_super->node_ino) != 1 ||
2563 le32_to_cpu(raw_super->meta_ino) != 2 ||
2564 le32_to_cpu(raw_super->root_ino) != 3) {
2565 f2fs_msg(sb, KERN_INFO,
2566 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2567 le32_to_cpu(raw_super->node_ino),
2568 le32_to_cpu(raw_super->meta_ino),
2569 le32_to_cpu(raw_super->root_ino));
2573 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2574 if (sanity_check_area_boundary(sbi, bh))
2580 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2582 unsigned int total, fsmeta;
2583 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2584 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2585 unsigned int ovp_segments, reserved_segments;
2586 unsigned int main_segs, blocks_per_seg;
2587 unsigned int sit_segs, nat_segs;
2588 unsigned int sit_bitmap_size, nat_bitmap_size;
2589 unsigned int log_blocks_per_seg;
2590 unsigned int segment_count_main;
2591 unsigned int cp_pack_start_sum, cp_payload;
2592 block_t user_block_count;
2595 total = le32_to_cpu(raw_super->segment_count);
2596 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2597 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2599 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2601 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2602 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2604 if (unlikely(fsmeta >= total))
2607 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2608 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2610 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2611 ovp_segments == 0 || reserved_segments == 0)) {
2612 f2fs_msg(sbi->sb, KERN_ERR,
2613 "Wrong layout: check mkfs.f2fs version");
2617 user_block_count = le64_to_cpu(ckpt->user_block_count);
2618 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2619 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2620 if (!user_block_count || user_block_count >=
2621 segment_count_main << log_blocks_per_seg) {
2622 f2fs_msg(sbi->sb, KERN_ERR,
2623 "Wrong user_block_count: %u", user_block_count);
2627 main_segs = le32_to_cpu(raw_super->segment_count_main);
2628 blocks_per_seg = sbi->blocks_per_seg;
2630 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2631 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2632 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2634 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2635 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2636 le32_to_cpu(ckpt->cur_node_segno[j])) {
2637 f2fs_msg(sbi->sb, KERN_ERR,
2638 "Node segment (%u, %u) has the same "
2640 le32_to_cpu(ckpt->cur_node_segno[i]));
2645 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2646 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2647 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2649 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2650 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2651 le32_to_cpu(ckpt->cur_data_segno[j])) {
2652 f2fs_msg(sbi->sb, KERN_ERR,
2653 "Data segment (%u, %u) has the same "
2655 le32_to_cpu(ckpt->cur_data_segno[i]));
2660 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2661 for (j = i; j < NR_CURSEG_DATA_TYPE; j++) {
2662 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2663 le32_to_cpu(ckpt->cur_data_segno[j])) {
2664 f2fs_msg(sbi->sb, KERN_ERR,
2665 "Data segment (%u) and Data segment (%u)"
2666 " has the same segno: %u", i, j,
2667 le32_to_cpu(ckpt->cur_node_segno[i]));
2673 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2674 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2676 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2677 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2678 f2fs_msg(sbi->sb, KERN_ERR,
2679 "Wrong bitmap size: sit: %u, nat:%u",
2680 sit_bitmap_size, nat_bitmap_size);
2684 cp_pack_start_sum = __start_sum_addr(sbi);
2685 cp_payload = __cp_payload(sbi);
2686 if (cp_pack_start_sum < cp_payload + 1 ||
2687 cp_pack_start_sum > blocks_per_seg - 1 -
2689 f2fs_msg(sbi->sb, KERN_ERR,
2690 "Wrong cp_pack_start_sum: %u",
2695 if (unlikely(f2fs_cp_error(sbi))) {
2696 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
2702 static void init_sb_info(struct f2fs_sb_info *sbi)
2704 struct f2fs_super_block *raw_super = sbi->raw_super;
2707 sbi->log_sectors_per_block =
2708 le32_to_cpu(raw_super->log_sectors_per_block);
2709 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2710 sbi->blocksize = 1 << sbi->log_blocksize;
2711 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2712 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2713 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2714 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2715 sbi->total_sections = le32_to_cpu(raw_super->section_count);
2716 sbi->total_node_count =
2717 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2718 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2719 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2720 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2721 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2722 sbi->cur_victim_sec = NULL_SECNO;
2723 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
2724 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
2725 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2726 sbi->migration_granularity = sbi->segs_per_sec;
2728 sbi->dir_level = DEF_DIR_LEVEL;
2729 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2730 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2731 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
2732 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
2733 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
2734 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
2735 DEF_UMOUNT_DISCARD_TIMEOUT;
2736 clear_sbi_flag(sbi, SBI_NEED_FSCK);
2738 for (i = 0; i < NR_COUNT_TYPE; i++)
2739 atomic_set(&sbi->nr_pages[i], 0);
2741 for (i = 0; i < META; i++)
2742 atomic_set(&sbi->wb_sync_req[i], 0);
2744 INIT_LIST_HEAD(&sbi->s_list);
2745 mutex_init(&sbi->umount_mutex);
2746 init_rwsem(&sbi->io_order_lock);
2747 spin_lock_init(&sbi->cp_lock);
2749 sbi->dirty_device = 0;
2750 spin_lock_init(&sbi->dev_lock);
2752 init_rwsem(&sbi->sb_lock);
2755 static int init_percpu_info(struct f2fs_sb_info *sbi)
2759 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2763 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
2766 percpu_counter_destroy(&sbi->alloc_valid_block_count);
2771 #ifdef CONFIG_BLK_DEV_ZONED
2772 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2774 struct block_device *bdev = FDEV(devi).bdev;
2775 sector_t nr_sectors = bdev->bd_part->nr_sects;
2776 sector_t sector = 0;
2777 struct blk_zone *zones;
2778 unsigned int i, nr_zones;
2782 if (!f2fs_sb_has_blkzoned(sbi))
2785 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2786 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2788 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2789 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2790 __ilog2_u32(sbi->blocks_per_blkz))
2792 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2793 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2794 sbi->log_blocks_per_blkz;
2795 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2796 FDEV(devi).nr_blkz++;
2798 FDEV(devi).blkz_type = f2fs_kmalloc(sbi, FDEV(devi).nr_blkz,
2800 if (!FDEV(devi).blkz_type)
2803 #define F2FS_REPORT_NR_ZONES 4096
2805 zones = f2fs_kzalloc(sbi,
2806 array_size(F2FS_REPORT_NR_ZONES,
2807 sizeof(struct blk_zone)),
2812 /* Get block zones type */
2813 while (zones && sector < nr_sectors) {
2815 nr_zones = F2FS_REPORT_NR_ZONES;
2816 err = blkdev_report_zones(bdev, sector,
2826 for (i = 0; i < nr_zones; i++) {
2827 FDEV(devi).blkz_type[n] = zones[i].type;
2828 sector += zones[i].len;
2840 * Read f2fs raw super block.
2841 * Because we have two copies of super block, so read both of them
2842 * to get the first valid one. If any one of them is broken, we pass
2843 * them recovery flag back to the caller.
2845 static int read_raw_super_block(struct f2fs_sb_info *sbi,
2846 struct f2fs_super_block **raw_super,
2847 int *valid_super_block, int *recovery)
2849 struct super_block *sb = sbi->sb;
2851 struct buffer_head *bh;
2852 struct f2fs_super_block *super;
2855 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2859 for (block = 0; block < 2; block++) {
2860 bh = sb_bread(sb, block);
2862 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
2868 /* sanity checking of raw super */
2869 if (sanity_check_raw_super(sbi, bh)) {
2870 f2fs_msg(sb, KERN_ERR,
2871 "Can't find valid F2FS filesystem in %dth superblock",
2879 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2881 *valid_super_block = block;
2887 /* Fail to read any one of the superblocks*/
2891 /* No valid superblock */
2900 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
2902 struct buffer_head *bh;
2906 if ((recover && f2fs_readonly(sbi->sb)) ||
2907 bdev_read_only(sbi->sb->s_bdev)) {
2908 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2912 /* we should update superblock crc here */
2913 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
2914 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
2915 offsetof(struct f2fs_super_block, crc));
2916 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
2919 /* write back-up superblock first */
2920 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
2923 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2926 /* if we are in recovery path, skip writing valid superblock */
2930 /* write current valid superblock */
2931 bh = sb_bread(sbi->sb, sbi->valid_super_block);
2934 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2939 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
2941 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2942 unsigned int max_devices = MAX_DEVICES;
2945 /* Initialize single device information */
2946 if (!RDEV(0).path[0]) {
2947 if (!bdev_is_zoned(sbi->sb->s_bdev))
2953 * Initialize multiple devices information, or single
2954 * zoned block device information.
2956 sbi->devs = f2fs_kzalloc(sbi,
2957 array_size(max_devices,
2958 sizeof(struct f2fs_dev_info)),
2963 for (i = 0; i < max_devices; i++) {
2965 if (i > 0 && !RDEV(i).path[0])
2968 if (max_devices == 1) {
2969 /* Single zoned block device mount */
2971 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
2972 sbi->sb->s_mode, sbi->sb->s_type);
2974 /* Multi-device mount */
2975 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
2976 FDEV(i).total_segments =
2977 le32_to_cpu(RDEV(i).total_segments);
2979 FDEV(i).start_blk = 0;
2980 FDEV(i).end_blk = FDEV(i).start_blk +
2981 (FDEV(i).total_segments <<
2982 sbi->log_blocks_per_seg) - 1 +
2983 le32_to_cpu(raw_super->segment0_blkaddr);
2985 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
2986 FDEV(i).end_blk = FDEV(i).start_blk +
2987 (FDEV(i).total_segments <<
2988 sbi->log_blocks_per_seg) - 1;
2990 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
2991 sbi->sb->s_mode, sbi->sb->s_type);
2993 if (IS_ERR(FDEV(i).bdev))
2994 return PTR_ERR(FDEV(i).bdev);
2996 /* to release errored devices */
2997 sbi->s_ndevs = i + 1;
2999 #ifdef CONFIG_BLK_DEV_ZONED
3000 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3001 !f2fs_sb_has_blkzoned(sbi)) {
3002 f2fs_msg(sbi->sb, KERN_ERR,
3003 "Zoned block device feature not enabled\n");
3006 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3007 if (init_blkz_info(sbi, i)) {
3008 f2fs_msg(sbi->sb, KERN_ERR,
3009 "Failed to initialize F2FS blkzone information");
3012 if (max_devices == 1)
3014 f2fs_msg(sbi->sb, KERN_INFO,
3015 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3017 FDEV(i).total_segments,
3018 FDEV(i).start_blk, FDEV(i).end_blk,
3019 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3020 "Host-aware" : "Host-managed");
3024 f2fs_msg(sbi->sb, KERN_INFO,
3025 "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3027 FDEV(i).total_segments,
3028 FDEV(i).start_blk, FDEV(i).end_blk);
3030 f2fs_msg(sbi->sb, KERN_INFO,
3031 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3035 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3037 struct f2fs_sm_info *sm_i = SM_I(sbi);
3039 /* adjust parameters according to the volume size */
3040 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3041 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3042 sm_i->dcc_info->discard_granularity = 1;
3043 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3046 sbi->readdir_ra = 1;
3049 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3051 struct f2fs_sb_info *sbi;
3052 struct f2fs_super_block *raw_super;
3055 bool retry = true, need_fsck = false;
3056 char *options = NULL;
3057 int recovery, i, valid_super_block;
3058 struct curseg_info *seg_i;
3063 valid_super_block = -1;
3066 /* allocate memory for f2fs-specific super block info */
3067 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3073 /* Load the checksum driver */
3074 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3075 if (IS_ERR(sbi->s_chksum_driver)) {
3076 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
3077 err = PTR_ERR(sbi->s_chksum_driver);
3078 sbi->s_chksum_driver = NULL;
3082 /* set a block size */
3083 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3084 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
3088 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3093 sb->s_fs_info = sbi;
3094 sbi->raw_super = raw_super;
3096 /* precompute checksum seed for metadata */
3097 if (f2fs_sb_has_inode_chksum(sbi))
3098 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3099 sizeof(raw_super->uuid));
3102 * The BLKZONED feature indicates that the drive was formatted with
3103 * zone alignment optimization. This is optional for host-aware
3104 * devices, but mandatory for host-managed zoned block devices.
3106 #ifndef CONFIG_BLK_DEV_ZONED
3107 if (f2fs_sb_has_blkzoned(sbi)) {
3108 f2fs_msg(sb, KERN_ERR,
3109 "Zoned block device support is not enabled\n");
3114 default_options(sbi);
3115 /* parse mount options */
3116 options = kstrdup((const char *)data, GFP_KERNEL);
3117 if (data && !options) {
3122 err = parse_options(sb, options);
3126 sbi->max_file_blocks = max_file_blocks();
3127 sb->s_maxbytes = sbi->max_file_blocks <<
3128 le32_to_cpu(raw_super->log_blocksize);
3129 sb->s_max_links = F2FS_LINK_MAX;
3130 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3133 sb->dq_op = &f2fs_quota_operations;
3134 if (f2fs_sb_has_quota_ino(sbi))
3135 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3137 sb->s_qcop = &f2fs_quotactl_ops;
3138 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3140 if (f2fs_sb_has_quota_ino(sbi)) {
3141 for (i = 0; i < MAXQUOTAS; i++) {
3142 if (f2fs_qf_ino(sbi->sb, i))
3143 sbi->nquota_files++;
3148 sb->s_op = &f2fs_sops;
3149 #ifdef CONFIG_F2FS_FS_ENCRYPTION
3150 sb->s_cop = &f2fs_cryptops;
3152 sb->s_xattr = f2fs_xattr_handlers;
3153 sb->s_export_op = &f2fs_export_ops;
3154 sb->s_magic = F2FS_SUPER_MAGIC;
3155 sb->s_time_gran = 1;
3156 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3157 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3158 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3159 sb->s_iflags |= SB_I_CGROUPWB;
3161 /* init f2fs-specific super block info */
3162 sbi->valid_super_block = valid_super_block;
3163 mutex_init(&sbi->gc_mutex);
3164 mutex_init(&sbi->writepages);
3165 mutex_init(&sbi->cp_mutex);
3166 init_rwsem(&sbi->node_write);
3167 init_rwsem(&sbi->node_change);
3169 /* disallow all the data/node/meta page writes */
3170 set_sbi_flag(sbi, SBI_POR_DOING);
3171 spin_lock_init(&sbi->stat_lock);
3173 /* init iostat info */
3174 spin_lock_init(&sbi->iostat_lock);
3175 sbi->iostat_enable = false;
3177 for (i = 0; i < NR_PAGE_TYPE; i++) {
3178 int n = (i == META) ? 1: NR_TEMP_TYPE;
3184 sizeof(struct f2fs_bio_info)),
3186 if (!sbi->write_io[i]) {
3191 for (j = HOT; j < n; j++) {
3192 init_rwsem(&sbi->write_io[i][j].io_rwsem);
3193 sbi->write_io[i][j].sbi = sbi;
3194 sbi->write_io[i][j].bio = NULL;
3195 spin_lock_init(&sbi->write_io[i][j].io_lock);
3196 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3200 init_rwsem(&sbi->cp_rwsem);
3201 init_waitqueue_head(&sbi->cp_wait);
3204 err = init_percpu_info(sbi);
3208 if (F2FS_IO_SIZE(sbi) > 1) {
3209 sbi->write_io_dummy =
3210 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3211 if (!sbi->write_io_dummy) {
3217 /* get an inode for meta space */
3218 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3219 if (IS_ERR(sbi->meta_inode)) {
3220 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
3221 err = PTR_ERR(sbi->meta_inode);
3225 err = f2fs_get_valid_checkpoint(sbi);
3227 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
3228 goto free_meta_inode;
3231 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3232 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3233 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3234 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3235 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3238 /* Initialize device list */
3239 err = f2fs_scan_devices(sbi);
3241 f2fs_msg(sb, KERN_ERR, "Failed to find devices");
3245 sbi->total_valid_node_count =
3246 le32_to_cpu(sbi->ckpt->valid_node_count);
3247 percpu_counter_set(&sbi->total_valid_inode_count,
3248 le32_to_cpu(sbi->ckpt->valid_inode_count));
3249 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3250 sbi->total_valid_block_count =
3251 le64_to_cpu(sbi->ckpt->valid_block_count);
3252 sbi->last_valid_block_count = sbi->total_valid_block_count;
3253 sbi->reserved_blocks = 0;
3254 sbi->current_reserved_blocks = 0;
3255 limit_reserve_root(sbi);
3257 for (i = 0; i < NR_INODE_TYPE; i++) {
3258 INIT_LIST_HEAD(&sbi->inode_list[i]);
3259 spin_lock_init(&sbi->inode_lock[i]);
3262 f2fs_init_extent_cache_info(sbi);
3264 f2fs_init_ino_entry_info(sbi);
3266 f2fs_init_fsync_node_info(sbi);
3268 /* setup f2fs internal modules */
3269 err = f2fs_build_segment_manager(sbi);
3271 f2fs_msg(sb, KERN_ERR,
3272 "Failed to initialize F2FS segment manager");
3275 err = f2fs_build_node_manager(sbi);
3277 f2fs_msg(sb, KERN_ERR,
3278 "Failed to initialize F2FS node manager");
3282 /* For write statistics */
3283 if (sb->s_bdev->bd_part)
3284 sbi->sectors_written_start =
3285 (u64)part_stat_read(sb->s_bdev->bd_part,
3286 sectors[STAT_WRITE]);
3288 /* Read accumulated write IO statistics if exists */
3289 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3290 if (__exist_node_summaries(sbi))
3291 sbi->kbytes_written =
3292 le64_to_cpu(seg_i->journal->info.kbytes_written);
3294 f2fs_build_gc_manager(sbi);
3296 err = f2fs_build_stats(sbi);
3300 /* get an inode for node space */
3301 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3302 if (IS_ERR(sbi->node_inode)) {
3303 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
3304 err = PTR_ERR(sbi->node_inode);
3308 /* read root inode and dentry */
3309 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3311 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
3312 err = PTR_ERR(root);
3313 goto free_node_inode;
3315 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3316 !root->i_size || !root->i_nlink) {
3319 goto free_node_inode;
3322 sb->s_root = d_make_root(root); /* allocate root dentry */
3325 goto free_root_inode;
3328 err = f2fs_register_sysfs(sbi);
3330 goto free_root_inode;
3333 /* Enable quota usage during mount */
3334 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3335 err = f2fs_enable_quotas(sb);
3337 f2fs_msg(sb, KERN_ERR,
3338 "Cannot turn on quotas: error %d", err);
3341 /* if there are nt orphan nodes free them */
3342 err = f2fs_recover_orphan_inodes(sbi);
3346 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3349 /* recover fsynced data */
3350 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
3352 * mount should be failed, when device has readonly mode, and
3353 * previous checkpoint was not done by clean system shutdown.
3355 if (bdev_read_only(sb->s_bdev) &&
3356 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3362 set_sbi_flag(sbi, SBI_NEED_FSCK);
3367 err = f2fs_recover_fsync_data(sbi, false);
3370 f2fs_msg(sb, KERN_ERR,
3371 "Cannot recover all fsync data errno=%d", err);
3375 err = f2fs_recover_fsync_data(sbi, true);
3377 if (!f2fs_readonly(sb) && err > 0) {
3379 f2fs_msg(sb, KERN_ERR,
3380 "Need to recover fsync data");
3385 /* f2fs_recover_fsync_data() cleared this already */
3386 clear_sbi_flag(sbi, SBI_POR_DOING);
3388 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3389 err = f2fs_disable_checkpoint(sbi);
3391 goto sync_free_meta;
3392 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3393 f2fs_enable_checkpoint(sbi);
3397 * If filesystem is not mounted as read-only then
3398 * do start the gc_thread.
3400 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
3401 /* After POR, we can run background GC thread.*/
3402 err = f2fs_start_gc_thread(sbi);
3404 goto sync_free_meta;
3408 /* recover broken superblock */
3410 err = f2fs_commit_super(sbi, true);
3411 f2fs_msg(sb, KERN_INFO,
3412 "Try to recover %dth superblock, ret: %d",
3413 sbi->valid_super_block ? 1 : 2, err);
3416 f2fs_join_shrinker(sbi);
3418 f2fs_tuning_parameters(sbi);
3420 f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx",
3421 cur_cp_version(F2FS_CKPT(sbi)));
3422 f2fs_update_time(sbi, CP_TIME);
3423 f2fs_update_time(sbi, REQ_TIME);
3424 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3428 /* safe to flush all the data */
3429 sync_filesystem(sbi->sb);
3434 f2fs_truncate_quota_inode_pages(sb);
3435 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3436 f2fs_quota_off_umount(sbi->sb);
3439 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3440 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3441 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3442 * falls into an infinite loop in f2fs_sync_meta_pages().
3444 truncate_inode_pages_final(META_MAPPING(sbi));
3445 /* evict some inodes being cached by GC */
3447 f2fs_unregister_sysfs(sbi);
3452 f2fs_release_ino_entry(sbi, true);
3453 truncate_inode_pages_final(NODE_MAPPING(sbi));
3454 iput(sbi->node_inode);
3455 sbi->node_inode = NULL;
3457 f2fs_destroy_stats(sbi);
3459 f2fs_destroy_node_manager(sbi);
3461 f2fs_destroy_segment_manager(sbi);
3463 destroy_device_list(sbi);
3466 make_bad_inode(sbi->meta_inode);
3467 iput(sbi->meta_inode);
3468 sbi->meta_inode = NULL;
3470 mempool_destroy(sbi->write_io_dummy);
3472 destroy_percpu_info(sbi);
3474 for (i = 0; i < NR_PAGE_TYPE; i++)
3475 kvfree(sbi->write_io[i]);
3478 for (i = 0; i < MAXQUOTAS; i++)
3479 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
3485 if (sbi->s_chksum_driver)
3486 crypto_free_shash(sbi->s_chksum_driver);
3489 /* give only one another chance */
3492 shrink_dcache_sb(sb);
3498 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3499 const char *dev_name, void *data)
3501 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3504 static void kill_f2fs_super(struct super_block *sb)
3507 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3509 set_sbi_flag(sbi, SBI_IS_CLOSE);
3510 f2fs_stop_gc_thread(sbi);
3511 f2fs_stop_discard_thread(sbi);
3513 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3514 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3515 struct cp_control cpc = {
3516 .reason = CP_UMOUNT,
3518 f2fs_write_checkpoint(sbi, &cpc);
3521 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3522 sb->s_flags &= ~SB_RDONLY;
3524 kill_block_super(sb);
3527 static struct file_system_type f2fs_fs_type = {
3528 .owner = THIS_MODULE,
3530 .mount = f2fs_mount,
3531 .kill_sb = kill_f2fs_super,
3532 .fs_flags = FS_REQUIRES_DEV,
3534 MODULE_ALIAS_FS("f2fs");
3536 static int __init init_inodecache(void)
3538 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3539 sizeof(struct f2fs_inode_info), 0,
3540 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3541 if (!f2fs_inode_cachep)
3546 static void destroy_inodecache(void)
3549 * Make sure all delayed rcu free inodes are flushed before we
3553 kmem_cache_destroy(f2fs_inode_cachep);
3556 static int __init init_f2fs_fs(void)
3560 if (PAGE_SIZE != F2FS_BLKSIZE) {
3561 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3562 PAGE_SIZE, F2FS_BLKSIZE);
3566 f2fs_build_trace_ios();
3568 err = init_inodecache();
3571 err = f2fs_create_node_manager_caches();
3573 goto free_inodecache;
3574 err = f2fs_create_segment_manager_caches();
3576 goto free_node_manager_caches;
3577 err = f2fs_create_checkpoint_caches();
3579 goto free_segment_manager_caches;
3580 err = f2fs_create_extent_cache();
3582 goto free_checkpoint_caches;
3583 err = f2fs_init_sysfs();
3585 goto free_extent_cache;
3586 err = register_shrinker(&f2fs_shrinker_info);
3589 err = register_filesystem(&f2fs_fs_type);
3592 f2fs_create_root_stats();
3593 err = f2fs_init_post_read_processing();
3595 goto free_root_stats;
3599 f2fs_destroy_root_stats();
3600 unregister_filesystem(&f2fs_fs_type);
3602 unregister_shrinker(&f2fs_shrinker_info);
3606 f2fs_destroy_extent_cache();
3607 free_checkpoint_caches:
3608 f2fs_destroy_checkpoint_caches();
3609 free_segment_manager_caches:
3610 f2fs_destroy_segment_manager_caches();
3611 free_node_manager_caches:
3612 f2fs_destroy_node_manager_caches();
3614 destroy_inodecache();
3619 static void __exit exit_f2fs_fs(void)
3621 f2fs_destroy_post_read_processing();
3622 f2fs_destroy_root_stats();
3623 unregister_filesystem(&f2fs_fs_type);
3624 unregister_shrinker(&f2fs_shrinker_info);
3626 f2fs_destroy_extent_cache();
3627 f2fs_destroy_checkpoint_caches();
3628 f2fs_destroy_segment_manager_caches();
3629 f2fs_destroy_node_manager_caches();
3630 destroy_inodecache();
3631 f2fs_destroy_trace_ios();
3634 module_init(init_f2fs_fs)
3635 module_exit(exit_f2fs_fs)
3637 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3638 MODULE_DESCRIPTION("Flash Friendly File System");
3639 MODULE_LICENSE("GPL");