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 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1008 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1010 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1011 percpu_counter_destroy(&sbi->total_valid_inode_count);
1014 static void destroy_device_list(struct f2fs_sb_info *sbi)
1018 for (i = 0; i < sbi->s_ndevs; i++) {
1019 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1020 #ifdef CONFIG_BLK_DEV_ZONED
1021 kvfree(FDEV(i).blkz_type);
1027 static void f2fs_put_super(struct super_block *sb)
1029 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1033 f2fs_quota_off_umount(sb);
1035 /* prevent remaining shrinker jobs */
1036 mutex_lock(&sbi->umount_mutex);
1039 * We don't need to do checkpoint when superblock is clean.
1040 * But, the previous checkpoint was not done by umount, it needs to do
1041 * clean checkpoint again.
1043 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1044 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1045 struct cp_control cpc = {
1046 .reason = CP_UMOUNT,
1048 f2fs_write_checkpoint(sbi, &cpc);
1051 /* be sure to wait for any on-going discard commands */
1052 dropped = f2fs_issue_discard_timeout(sbi);
1054 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1055 !sbi->discard_blks && !dropped) {
1056 struct cp_control cpc = {
1057 .reason = CP_UMOUNT | CP_TRIMMED,
1059 f2fs_write_checkpoint(sbi, &cpc);
1063 * normally superblock is clean, so we need to release this.
1064 * In addition, EIO will skip do checkpoint, we need this as well.
1066 f2fs_release_ino_entry(sbi, true);
1068 f2fs_leave_shrinker(sbi);
1069 mutex_unlock(&sbi->umount_mutex);
1071 /* our cp_error case, we can wait for any writeback page */
1072 f2fs_flush_merged_writes(sbi);
1074 f2fs_wait_on_all_pages_writeback(sbi);
1076 f2fs_bug_on(sbi, sbi->fsync_node_num);
1078 iput(sbi->node_inode);
1079 sbi->node_inode = NULL;
1081 iput(sbi->meta_inode);
1082 sbi->meta_inode = NULL;
1085 * iput() can update stat information, if f2fs_write_checkpoint()
1086 * above failed with error.
1088 f2fs_destroy_stats(sbi);
1090 /* destroy f2fs internal modules */
1091 f2fs_destroy_node_manager(sbi);
1092 f2fs_destroy_segment_manager(sbi);
1096 f2fs_unregister_sysfs(sbi);
1098 sb->s_fs_info = NULL;
1099 if (sbi->s_chksum_driver)
1100 crypto_free_shash(sbi->s_chksum_driver);
1101 kvfree(sbi->raw_super);
1103 destroy_device_list(sbi);
1104 mempool_destroy(sbi->write_io_dummy);
1106 for (i = 0; i < MAXQUOTAS; i++)
1107 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1109 destroy_percpu_info(sbi);
1110 for (i = 0; i < NR_PAGE_TYPE; i++)
1111 kvfree(sbi->write_io[i]);
1115 int f2fs_sync_fs(struct super_block *sb, int sync)
1117 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1120 if (unlikely(f2fs_cp_error(sbi)))
1122 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1125 trace_f2fs_sync_fs(sb, sync);
1127 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1131 struct cp_control cpc;
1133 cpc.reason = __get_cp_reason(sbi);
1135 mutex_lock(&sbi->gc_mutex);
1136 err = f2fs_write_checkpoint(sbi, &cpc);
1137 mutex_unlock(&sbi->gc_mutex);
1139 f2fs_trace_ios(NULL, 1);
1144 static int f2fs_freeze(struct super_block *sb)
1146 if (f2fs_readonly(sb))
1149 /* IO error happened before */
1150 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1153 /* must be clean, since sync_filesystem() was already called */
1154 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1159 static int f2fs_unfreeze(struct super_block *sb)
1165 static int f2fs_statfs_project(struct super_block *sb,
1166 kprojid_t projid, struct kstatfs *buf)
1169 struct dquot *dquot;
1173 qid = make_kqid_projid(projid);
1174 dquot = dqget(sb, qid);
1176 return PTR_ERR(dquot);
1177 spin_lock(&dquot->dq_dqb_lock);
1179 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
1180 dquot->dq_dqb.dqb_bsoftlimit :
1181 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
1182 if (limit && buf->f_blocks > limit) {
1183 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1184 buf->f_blocks = limit;
1185 buf->f_bfree = buf->f_bavail =
1186 (buf->f_blocks > curblock) ?
1187 (buf->f_blocks - curblock) : 0;
1190 limit = dquot->dq_dqb.dqb_isoftlimit ?
1191 dquot->dq_dqb.dqb_isoftlimit :
1192 dquot->dq_dqb.dqb_ihardlimit;
1193 if (limit && buf->f_files > limit) {
1194 buf->f_files = limit;
1196 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1197 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1200 spin_unlock(&dquot->dq_dqb_lock);
1206 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1208 struct super_block *sb = dentry->d_sb;
1209 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1210 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1211 block_t total_count, user_block_count, start_count;
1212 u64 avail_node_count;
1214 total_count = le64_to_cpu(sbi->raw_super->block_count);
1215 user_block_count = sbi->user_block_count;
1216 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1217 buf->f_type = F2FS_SUPER_MAGIC;
1218 buf->f_bsize = sbi->blocksize;
1220 buf->f_blocks = total_count - start_count;
1221 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1222 sbi->current_reserved_blocks;
1223 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1226 buf->f_bfree -= sbi->unusable_block_count;
1228 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1229 buf->f_bavail = buf->f_bfree -
1230 F2FS_OPTION(sbi).root_reserved_blocks;
1234 avail_node_count = sbi->total_node_count - sbi->nquota_files -
1235 F2FS_RESERVED_NODE_NUM;
1237 if (avail_node_count > user_block_count) {
1238 buf->f_files = user_block_count;
1239 buf->f_ffree = buf->f_bavail;
1241 buf->f_files = avail_node_count;
1242 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1246 buf->f_namelen = F2FS_NAME_LEN;
1247 buf->f_fsid.val[0] = (u32)id;
1248 buf->f_fsid.val[1] = (u32)(id >> 32);
1251 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1252 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1253 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1259 static inline void f2fs_show_quota_options(struct seq_file *seq,
1260 struct super_block *sb)
1263 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1265 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1268 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1279 seq_printf(seq, ",jqfmt=%s", fmtname);
1282 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1283 seq_show_option(seq, "usrjquota",
1284 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1286 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1287 seq_show_option(seq, "grpjquota",
1288 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1290 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1291 seq_show_option(seq, "prjjquota",
1292 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1296 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1298 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1300 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1301 if (test_opt(sbi, FORCE_FG_GC))
1302 seq_printf(seq, ",background_gc=%s", "sync");
1304 seq_printf(seq, ",background_gc=%s", "on");
1306 seq_printf(seq, ",background_gc=%s", "off");
1308 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1309 seq_puts(seq, ",disable_roll_forward");
1310 if (test_opt(sbi, DISCARD))
1311 seq_puts(seq, ",discard");
1312 if (test_opt(sbi, NOHEAP))
1313 seq_puts(seq, ",no_heap");
1315 seq_puts(seq, ",heap");
1316 #ifdef CONFIG_F2FS_FS_XATTR
1317 if (test_opt(sbi, XATTR_USER))
1318 seq_puts(seq, ",user_xattr");
1320 seq_puts(seq, ",nouser_xattr");
1321 if (test_opt(sbi, INLINE_XATTR))
1322 seq_puts(seq, ",inline_xattr");
1324 seq_puts(seq, ",noinline_xattr");
1325 if (test_opt(sbi, INLINE_XATTR_SIZE))
1326 seq_printf(seq, ",inline_xattr_size=%u",
1327 F2FS_OPTION(sbi).inline_xattr_size);
1329 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1330 if (test_opt(sbi, POSIX_ACL))
1331 seq_puts(seq, ",acl");
1333 seq_puts(seq, ",noacl");
1335 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1336 seq_puts(seq, ",disable_ext_identify");
1337 if (test_opt(sbi, INLINE_DATA))
1338 seq_puts(seq, ",inline_data");
1340 seq_puts(seq, ",noinline_data");
1341 if (test_opt(sbi, INLINE_DENTRY))
1342 seq_puts(seq, ",inline_dentry");
1344 seq_puts(seq, ",noinline_dentry");
1345 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1346 seq_puts(seq, ",flush_merge");
1347 if (test_opt(sbi, NOBARRIER))
1348 seq_puts(seq, ",nobarrier");
1349 if (test_opt(sbi, FASTBOOT))
1350 seq_puts(seq, ",fastboot");
1351 if (test_opt(sbi, EXTENT_CACHE))
1352 seq_puts(seq, ",extent_cache");
1354 seq_puts(seq, ",noextent_cache");
1355 if (test_opt(sbi, DATA_FLUSH))
1356 seq_puts(seq, ",data_flush");
1358 seq_puts(seq, ",mode=");
1359 if (test_opt(sbi, ADAPTIVE))
1360 seq_puts(seq, "adaptive");
1361 else if (test_opt(sbi, LFS))
1362 seq_puts(seq, "lfs");
1363 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1364 if (test_opt(sbi, RESERVE_ROOT))
1365 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1366 F2FS_OPTION(sbi).root_reserved_blocks,
1367 from_kuid_munged(&init_user_ns,
1368 F2FS_OPTION(sbi).s_resuid),
1369 from_kgid_munged(&init_user_ns,
1370 F2FS_OPTION(sbi).s_resgid));
1371 if (F2FS_IO_SIZE_BITS(sbi))
1372 seq_printf(seq, ",io_bits=%u",
1373 F2FS_OPTION(sbi).write_io_size_bits);
1374 #ifdef CONFIG_F2FS_FAULT_INJECTION
1375 if (test_opt(sbi, FAULT_INJECTION)) {
1376 seq_printf(seq, ",fault_injection=%u",
1377 F2FS_OPTION(sbi).fault_info.inject_rate);
1378 seq_printf(seq, ",fault_type=%u",
1379 F2FS_OPTION(sbi).fault_info.inject_type);
1383 if (test_opt(sbi, QUOTA))
1384 seq_puts(seq, ",quota");
1385 if (test_opt(sbi, USRQUOTA))
1386 seq_puts(seq, ",usrquota");
1387 if (test_opt(sbi, GRPQUOTA))
1388 seq_puts(seq, ",grpquota");
1389 if (test_opt(sbi, PRJQUOTA))
1390 seq_puts(seq, ",prjquota");
1392 f2fs_show_quota_options(seq, sbi->sb);
1393 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1394 seq_printf(seq, ",whint_mode=%s", "user-based");
1395 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1396 seq_printf(seq, ",whint_mode=%s", "fs-based");
1397 #ifdef CONFIG_FS_ENCRYPTION
1398 if (F2FS_OPTION(sbi).test_dummy_encryption)
1399 seq_puts(seq, ",test_dummy_encryption");
1402 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1403 seq_printf(seq, ",alloc_mode=%s", "default");
1404 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1405 seq_printf(seq, ",alloc_mode=%s", "reuse");
1407 if (test_opt(sbi, DISABLE_CHECKPOINT))
1408 seq_puts(seq, ",checkpoint=disable");
1410 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1411 seq_printf(seq, ",fsync_mode=%s", "posix");
1412 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1413 seq_printf(seq, ",fsync_mode=%s", "strict");
1414 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1415 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1419 static void default_options(struct f2fs_sb_info *sbi)
1421 /* init some FS parameters */
1422 F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1423 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1424 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1425 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1426 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1427 F2FS_OPTION(sbi).test_dummy_encryption = false;
1428 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1429 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1431 set_opt(sbi, BG_GC);
1432 set_opt(sbi, INLINE_XATTR);
1433 set_opt(sbi, INLINE_DATA);
1434 set_opt(sbi, INLINE_DENTRY);
1435 set_opt(sbi, EXTENT_CACHE);
1436 set_opt(sbi, NOHEAP);
1437 clear_opt(sbi, DISABLE_CHECKPOINT);
1438 sbi->sb->s_flags |= SB_LAZYTIME;
1439 set_opt(sbi, FLUSH_MERGE);
1440 set_opt(sbi, DISCARD);
1441 if (f2fs_sb_has_blkzoned(sbi))
1442 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1444 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1446 #ifdef CONFIG_F2FS_FS_XATTR
1447 set_opt(sbi, XATTR_USER);
1449 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1450 set_opt(sbi, POSIX_ACL);
1453 f2fs_build_fault_attr(sbi, 0, 0);
1457 static int f2fs_enable_quotas(struct super_block *sb);
1460 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1462 unsigned int s_flags = sbi->sb->s_flags;
1463 struct cp_control cpc;
1467 if (s_flags & SB_RDONLY) {
1468 f2fs_msg(sbi->sb, KERN_ERR,
1469 "checkpoint=disable on readonly fs");
1472 sbi->sb->s_flags |= SB_ACTIVE;
1474 f2fs_update_time(sbi, DISABLE_TIME);
1476 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1477 mutex_lock(&sbi->gc_mutex);
1478 err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1479 if (err == -ENODATA) {
1483 if (err && err != -EAGAIN)
1487 ret = sync_filesystem(sbi->sb);
1489 err = ret ? ret: err;
1493 if (f2fs_disable_cp_again(sbi)) {
1498 mutex_lock(&sbi->gc_mutex);
1499 cpc.reason = CP_PAUSE;
1500 set_sbi_flag(sbi, SBI_CP_DISABLED);
1501 f2fs_write_checkpoint(sbi, &cpc);
1503 sbi->unusable_block_count = 0;
1504 mutex_unlock(&sbi->gc_mutex);
1506 sbi->sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1510 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1512 mutex_lock(&sbi->gc_mutex);
1513 f2fs_dirty_to_prefree(sbi);
1515 clear_sbi_flag(sbi, SBI_CP_DISABLED);
1516 set_sbi_flag(sbi, SBI_IS_DIRTY);
1517 mutex_unlock(&sbi->gc_mutex);
1519 f2fs_sync_fs(sbi->sb, 1);
1522 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1524 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1525 struct f2fs_mount_info org_mount_opt;
1526 unsigned long old_sb_flags;
1528 bool need_restart_gc = false;
1529 bool need_stop_gc = false;
1530 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1531 bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1532 bool checkpoint_changed;
1538 * Save the old mount options in case we
1539 * need to restore them.
1541 org_mount_opt = sbi->mount_opt;
1542 old_sb_flags = sb->s_flags;
1545 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1546 for (i = 0; i < MAXQUOTAS; i++) {
1547 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1548 org_mount_opt.s_qf_names[i] =
1549 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1551 if (!org_mount_opt.s_qf_names[i]) {
1552 for (j = 0; j < i; j++)
1553 kvfree(org_mount_opt.s_qf_names[j]);
1557 org_mount_opt.s_qf_names[i] = NULL;
1562 /* recover superblocks we couldn't write due to previous RO mount */
1563 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1564 err = f2fs_commit_super(sbi, false);
1565 f2fs_msg(sb, KERN_INFO,
1566 "Try to recover all the superblocks, ret: %d", err);
1568 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1571 default_options(sbi);
1573 /* parse mount options */
1574 err = parse_options(sb, data);
1577 checkpoint_changed =
1578 disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1581 * Previous and new state of filesystem is RO,
1582 * so skip checking GC and FLUSH_MERGE conditions.
1584 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1588 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1589 err = dquot_suspend(sb, -1);
1592 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1593 /* dquot_resume needs RW */
1594 sb->s_flags &= ~SB_RDONLY;
1595 if (sb_any_quota_suspended(sb)) {
1596 dquot_resume(sb, -1);
1597 } else if (f2fs_sb_has_quota_ino(sbi)) {
1598 err = f2fs_enable_quotas(sb);
1604 /* disallow enable/disable extent_cache dynamically */
1605 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1607 f2fs_msg(sbi->sb, KERN_WARNING,
1608 "switch extent_cache option is not allowed");
1612 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1614 f2fs_msg(sbi->sb, KERN_WARNING,
1615 "disabling checkpoint not compatible with read-only");
1620 * We stop the GC thread if FS is mounted as RO
1621 * or if background_gc = off is passed in mount
1622 * option. Also sync the filesystem.
1624 if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) {
1625 if (sbi->gc_thread) {
1626 f2fs_stop_gc_thread(sbi);
1627 need_restart_gc = true;
1629 } else if (!sbi->gc_thread) {
1630 err = f2fs_start_gc_thread(sbi);
1633 need_stop_gc = true;
1636 if (*flags & SB_RDONLY ||
1637 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1638 writeback_inodes_sb(sb, WB_REASON_SYNC);
1641 set_sbi_flag(sbi, SBI_IS_DIRTY);
1642 set_sbi_flag(sbi, SBI_IS_CLOSE);
1643 f2fs_sync_fs(sb, 1);
1644 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1647 if (checkpoint_changed) {
1648 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1649 err = f2fs_disable_checkpoint(sbi);
1653 f2fs_enable_checkpoint(sbi);
1658 * We stop issue flush thread if FS is mounted as RO
1659 * or if flush_merge is not passed in mount option.
1661 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1662 clear_opt(sbi, FLUSH_MERGE);
1663 f2fs_destroy_flush_cmd_control(sbi, false);
1665 err = f2fs_create_flush_cmd_control(sbi);
1671 /* Release old quota file names */
1672 for (i = 0; i < MAXQUOTAS; i++)
1673 kvfree(org_mount_opt.s_qf_names[i]);
1675 /* Update the POSIXACL Flag */
1676 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1677 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1679 limit_reserve_root(sbi);
1680 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1683 if (need_restart_gc) {
1684 if (f2fs_start_gc_thread(sbi))
1685 f2fs_msg(sbi->sb, KERN_WARNING,
1686 "background gc thread has stopped");
1687 } else if (need_stop_gc) {
1688 f2fs_stop_gc_thread(sbi);
1692 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1693 for (i = 0; i < MAXQUOTAS; i++) {
1694 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1695 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1698 sbi->mount_opt = org_mount_opt;
1699 sb->s_flags = old_sb_flags;
1704 /* Read data from quotafile */
1705 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1706 size_t len, loff_t off)
1708 struct inode *inode = sb_dqopt(sb)->files[type];
1709 struct address_space *mapping = inode->i_mapping;
1710 block_t blkidx = F2FS_BYTES_TO_BLK(off);
1711 int offset = off & (sb->s_blocksize - 1);
1714 loff_t i_size = i_size_read(inode);
1721 if (off + len > i_size)
1724 while (toread > 0) {
1725 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1727 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1729 if (PTR_ERR(page) == -ENOMEM) {
1730 congestion_wait(BLK_RW_ASYNC, HZ/50);
1733 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1734 return PTR_ERR(page);
1739 if (unlikely(page->mapping != mapping)) {
1740 f2fs_put_page(page, 1);
1743 if (unlikely(!PageUptodate(page))) {
1744 f2fs_put_page(page, 1);
1745 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1749 kaddr = kmap_atomic(page);
1750 memcpy(data, kaddr + offset, tocopy);
1751 kunmap_atomic(kaddr);
1752 f2fs_put_page(page, 1);
1762 /* Write to quotafile */
1763 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1764 const char *data, size_t len, loff_t off)
1766 struct inode *inode = sb_dqopt(sb)->files[type];
1767 struct address_space *mapping = inode->i_mapping;
1768 const struct address_space_operations *a_ops = mapping->a_ops;
1769 int offset = off & (sb->s_blocksize - 1);
1770 size_t towrite = len;
1776 while (towrite > 0) {
1777 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1780 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1782 if (unlikely(err)) {
1783 if (err == -ENOMEM) {
1784 congestion_wait(BLK_RW_ASYNC, HZ/50);
1787 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1791 kaddr = kmap_atomic(page);
1792 memcpy(kaddr + offset, data, tocopy);
1793 kunmap_atomic(kaddr);
1794 flush_dcache_page(page);
1796 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1807 inode->i_mtime = inode->i_ctime = current_time(inode);
1808 f2fs_mark_inode_dirty_sync(inode, false);
1809 return len - towrite;
1812 static struct dquot **f2fs_get_dquots(struct inode *inode)
1814 return F2FS_I(inode)->i_dquot;
1817 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1819 return &F2FS_I(inode)->i_reserved_quota;
1822 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1824 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
1825 f2fs_msg(sbi->sb, KERN_ERR,
1826 "quota sysfile may be corrupted, skip loading it");
1830 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
1831 F2FS_OPTION(sbi).s_jquota_fmt, type);
1834 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1839 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
1840 err = f2fs_enable_quotas(sbi->sb);
1842 f2fs_msg(sbi->sb, KERN_ERR,
1843 "Cannot turn on quota_ino: %d", err);
1849 for (i = 0; i < MAXQUOTAS; i++) {
1850 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1851 err = f2fs_quota_on_mount(sbi, i);
1856 f2fs_msg(sbi->sb, KERN_ERR,
1857 "Cannot turn on quotas: %d on %d", err, i);
1863 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
1866 struct inode *qf_inode;
1867 unsigned long qf_inum;
1870 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
1872 qf_inum = f2fs_qf_ino(sb, type);
1876 qf_inode = f2fs_iget(sb, qf_inum);
1877 if (IS_ERR(qf_inode)) {
1878 f2fs_msg(sb, KERN_ERR,
1879 "Bad quota inode %u:%lu", type, qf_inum);
1880 return PTR_ERR(qf_inode);
1883 /* Don't account quota for quota files to avoid recursion */
1884 qf_inode->i_flags |= S_NOQUOTA;
1885 err = dquot_enable(qf_inode, type, format_id, flags);
1890 static int f2fs_enable_quotas(struct super_block *sb)
1893 unsigned long qf_inum;
1894 bool quota_mopt[MAXQUOTAS] = {
1895 test_opt(F2FS_SB(sb), USRQUOTA),
1896 test_opt(F2FS_SB(sb), GRPQUOTA),
1897 test_opt(F2FS_SB(sb), PRJQUOTA),
1900 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
1901 f2fs_msg(sb, KERN_ERR,
1902 "quota file may be corrupted, skip loading it");
1906 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
1908 for (type = 0; type < MAXQUOTAS; type++) {
1909 qf_inum = f2fs_qf_ino(sb, type);
1911 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
1912 DQUOT_USAGE_ENABLED |
1913 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
1915 f2fs_msg(sb, KERN_ERR,
1916 "Failed to enable quota tracking "
1917 "(type=%d, err=%d). Please run "
1918 "fsck to fix.", type, err);
1919 for (type--; type >= 0; type--)
1920 dquot_quota_off(sb, type);
1921 set_sbi_flag(F2FS_SB(sb),
1922 SBI_QUOTA_NEED_REPAIR);
1930 int f2fs_quota_sync(struct super_block *sb, int type)
1932 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1933 struct quota_info *dqopt = sb_dqopt(sb);
1937 ret = dquot_writeback_dquots(sb, type);
1942 * Now when everything is written we can discard the pagecache so
1943 * that userspace sees the changes.
1945 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
1946 struct address_space *mapping;
1948 if (type != -1 && cnt != type)
1950 if (!sb_has_quota_active(sb, cnt))
1953 mapping = dqopt->files[cnt]->i_mapping;
1955 ret = filemap_fdatawrite(mapping);
1959 /* if we are using journalled quota */
1960 if (is_journalled_quota(sbi))
1963 ret = filemap_fdatawait(mapping);
1965 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1967 inode_lock(dqopt->files[cnt]);
1968 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
1969 inode_unlock(dqopt->files[cnt]);
1973 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1977 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
1978 const struct path *path)
1980 struct inode *inode;
1983 err = f2fs_quota_sync(sb, type);
1987 err = dquot_quota_on(sb, type, format_id, path);
1991 inode = d_inode(path->dentry);
1994 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
1995 f2fs_set_inode_flags(inode);
1996 inode_unlock(inode);
1997 f2fs_mark_inode_dirty_sync(inode, false);
2002 static int f2fs_quota_off(struct super_block *sb, int type)
2004 struct inode *inode = sb_dqopt(sb)->files[type];
2007 if (!inode || !igrab(inode))
2008 return dquot_quota_off(sb, type);
2010 err = f2fs_quota_sync(sb, type);
2014 err = dquot_quota_off(sb, type);
2015 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2019 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2020 f2fs_set_inode_flags(inode);
2021 inode_unlock(inode);
2022 f2fs_mark_inode_dirty_sync(inode, false);
2028 void f2fs_quota_off_umount(struct super_block *sb)
2033 for (type = 0; type < MAXQUOTAS; type++) {
2034 err = f2fs_quota_off(sb, type);
2036 int ret = dquot_quota_off(sb, type);
2038 f2fs_msg(sb, KERN_ERR,
2039 "Fail to turn off disk quota "
2040 "(type: %d, err: %d, ret:%d), Please "
2041 "run fsck to fix it.", type, err, ret);
2042 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2046 * In case of checkpoint=disable, we must flush quota blocks.
2047 * This can cause NULL exception for node_inode in end_io, since
2048 * put_super already dropped it.
2050 sync_filesystem(sb);
2053 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2055 struct quota_info *dqopt = sb_dqopt(sb);
2058 for (type = 0; type < MAXQUOTAS; type++) {
2059 if (!dqopt->files[type])
2061 f2fs_inode_synced(dqopt->files[type]);
2065 static int f2fs_dquot_commit(struct dquot *dquot)
2069 ret = dquot_commit(dquot);
2071 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2075 static int f2fs_dquot_acquire(struct dquot *dquot)
2079 ret = dquot_acquire(dquot);
2081 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2086 static int f2fs_dquot_release(struct dquot *dquot)
2090 ret = dquot_release(dquot);
2092 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2096 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2098 struct super_block *sb = dquot->dq_sb;
2099 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2102 ret = dquot_mark_dquot_dirty(dquot);
2104 /* if we are using journalled quota */
2105 if (is_journalled_quota(sbi))
2106 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2111 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2115 ret = dquot_commit_info(sb, type);
2117 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2121 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2123 *projid = F2FS_I(inode)->i_projid;
2127 static const struct dquot_operations f2fs_quota_operations = {
2128 .get_reserved_space = f2fs_get_reserved_space,
2129 .write_dquot = f2fs_dquot_commit,
2130 .acquire_dquot = f2fs_dquot_acquire,
2131 .release_dquot = f2fs_dquot_release,
2132 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
2133 .write_info = f2fs_dquot_commit_info,
2134 .alloc_dquot = dquot_alloc,
2135 .destroy_dquot = dquot_destroy,
2136 .get_projid = f2fs_get_projid,
2137 .get_next_id = dquot_get_next_id,
2140 static const struct quotactl_ops f2fs_quotactl_ops = {
2141 .quota_on = f2fs_quota_on,
2142 .quota_off = f2fs_quota_off,
2143 .quota_sync = f2fs_quota_sync,
2144 .get_state = dquot_get_state,
2145 .set_info = dquot_set_dqinfo,
2146 .get_dqblk = dquot_get_dqblk,
2147 .set_dqblk = dquot_set_dqblk,
2148 .get_nextdqblk = dquot_get_next_dqblk,
2151 int f2fs_quota_sync(struct super_block *sb, int type)
2156 void f2fs_quota_off_umount(struct super_block *sb)
2161 static const struct super_operations f2fs_sops = {
2162 .alloc_inode = f2fs_alloc_inode,
2163 .free_inode = f2fs_free_inode,
2164 .drop_inode = f2fs_drop_inode,
2165 .write_inode = f2fs_write_inode,
2166 .dirty_inode = f2fs_dirty_inode,
2167 .show_options = f2fs_show_options,
2169 .quota_read = f2fs_quota_read,
2170 .quota_write = f2fs_quota_write,
2171 .get_dquots = f2fs_get_dquots,
2173 .evict_inode = f2fs_evict_inode,
2174 .put_super = f2fs_put_super,
2175 .sync_fs = f2fs_sync_fs,
2176 .freeze_fs = f2fs_freeze,
2177 .unfreeze_fs = f2fs_unfreeze,
2178 .statfs = f2fs_statfs,
2179 .remount_fs = f2fs_remount,
2182 #ifdef CONFIG_FS_ENCRYPTION
2183 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2185 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2186 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2190 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2193 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2196 * Encrypting the root directory is not allowed because fsck
2197 * expects lost+found directory to exist and remain unencrypted
2198 * if LOST_FOUND feature is enabled.
2201 if (f2fs_sb_has_lost_found(sbi) &&
2202 inode->i_ino == F2FS_ROOT_INO(sbi))
2205 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2206 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2207 ctx, len, fs_data, XATTR_CREATE);
2210 static bool f2fs_dummy_context(struct inode *inode)
2212 return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
2215 static const struct fscrypt_operations f2fs_cryptops = {
2216 .key_prefix = "f2fs:",
2217 .get_context = f2fs_get_context,
2218 .set_context = f2fs_set_context,
2219 .dummy_context = f2fs_dummy_context,
2220 .empty_dir = f2fs_empty_dir,
2221 .max_namelen = F2FS_NAME_LEN,
2225 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2226 u64 ino, u32 generation)
2228 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2229 struct inode *inode;
2231 if (f2fs_check_nid_range(sbi, ino))
2232 return ERR_PTR(-ESTALE);
2235 * f2fs_iget isn't quite right if the inode is currently unallocated!
2236 * However f2fs_iget currently does appropriate checks to handle stale
2237 * inodes so everything is OK.
2239 inode = f2fs_iget(sb, ino);
2241 return ERR_CAST(inode);
2242 if (unlikely(generation && inode->i_generation != generation)) {
2243 /* we didn't find the right inode.. */
2245 return ERR_PTR(-ESTALE);
2250 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2251 int fh_len, int fh_type)
2253 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2254 f2fs_nfs_get_inode);
2257 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2258 int fh_len, int fh_type)
2260 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2261 f2fs_nfs_get_inode);
2264 static const struct export_operations f2fs_export_ops = {
2265 .fh_to_dentry = f2fs_fh_to_dentry,
2266 .fh_to_parent = f2fs_fh_to_parent,
2267 .get_parent = f2fs_get_parent,
2270 static loff_t max_file_blocks(void)
2273 loff_t leaf_count = ADDRS_PER_BLOCK;
2276 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2277 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2278 * space in inode.i_addr, it will be more safe to reassign
2282 /* two direct node blocks */
2283 result += (leaf_count * 2);
2285 /* two indirect node blocks */
2286 leaf_count *= NIDS_PER_BLOCK;
2287 result += (leaf_count * 2);
2289 /* one double indirect node block */
2290 leaf_count *= NIDS_PER_BLOCK;
2291 result += leaf_count;
2296 static int __f2fs_commit_super(struct buffer_head *bh,
2297 struct f2fs_super_block *super)
2301 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2302 set_buffer_dirty(bh);
2305 /* it's rare case, we can do fua all the time */
2306 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2309 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2310 struct buffer_head *bh)
2312 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2313 (bh->b_data + F2FS_SUPER_OFFSET);
2314 struct super_block *sb = sbi->sb;
2315 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2316 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2317 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2318 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2319 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2320 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2321 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2322 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2323 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2324 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2325 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2326 u32 segment_count = le32_to_cpu(raw_super->segment_count);
2327 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2328 u64 main_end_blkaddr = main_blkaddr +
2329 (segment_count_main << log_blocks_per_seg);
2330 u64 seg_end_blkaddr = segment0_blkaddr +
2331 (segment_count << log_blocks_per_seg);
2333 if (segment0_blkaddr != cp_blkaddr) {
2334 f2fs_msg(sb, KERN_INFO,
2335 "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2336 segment0_blkaddr, cp_blkaddr);
2340 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2342 f2fs_msg(sb, KERN_INFO,
2343 "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2344 cp_blkaddr, sit_blkaddr,
2345 segment_count_ckpt << log_blocks_per_seg);
2349 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2351 f2fs_msg(sb, KERN_INFO,
2352 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2353 sit_blkaddr, nat_blkaddr,
2354 segment_count_sit << log_blocks_per_seg);
2358 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2360 f2fs_msg(sb, KERN_INFO,
2361 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2362 nat_blkaddr, ssa_blkaddr,
2363 segment_count_nat << log_blocks_per_seg);
2367 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2369 f2fs_msg(sb, KERN_INFO,
2370 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2371 ssa_blkaddr, main_blkaddr,
2372 segment_count_ssa << log_blocks_per_seg);
2376 if (main_end_blkaddr > seg_end_blkaddr) {
2377 f2fs_msg(sb, KERN_INFO,
2378 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2381 (segment_count << log_blocks_per_seg),
2382 segment_count_main << log_blocks_per_seg);
2384 } else if (main_end_blkaddr < seg_end_blkaddr) {
2388 /* fix in-memory information all the time */
2389 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2390 segment0_blkaddr) >> log_blocks_per_seg);
2392 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2393 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2396 err = __f2fs_commit_super(bh, NULL);
2397 res = err ? "failed" : "done";
2399 f2fs_msg(sb, KERN_INFO,
2400 "Fix alignment : %s, start(%u) end(%u) block(%u)",
2403 (segment_count << log_blocks_per_seg),
2404 segment_count_main << log_blocks_per_seg);
2411 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2412 struct buffer_head *bh)
2414 block_t segment_count, segs_per_sec, secs_per_zone;
2415 block_t total_sections, blocks_per_seg;
2416 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2417 (bh->b_data + F2FS_SUPER_OFFSET);
2418 struct super_block *sb = sbi->sb;
2419 unsigned int blocksize;
2420 size_t crc_offset = 0;
2423 /* Check checksum_offset and crc in superblock */
2424 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2425 crc_offset = le32_to_cpu(raw_super->checksum_offset);
2427 offsetof(struct f2fs_super_block, crc)) {
2428 f2fs_msg(sb, KERN_INFO,
2429 "Invalid SB checksum offset: %zu",
2433 crc = le32_to_cpu(raw_super->crc);
2434 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2435 f2fs_msg(sb, KERN_INFO,
2436 "Invalid SB checksum value: %u", crc);
2441 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
2442 f2fs_msg(sb, KERN_INFO,
2443 "Magic Mismatch, valid(0x%x) - read(0x%x)",
2444 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2448 /* Currently, support only 4KB page cache size */
2449 if (F2FS_BLKSIZE != PAGE_SIZE) {
2450 f2fs_msg(sb, KERN_INFO,
2451 "Invalid page_cache_size (%lu), supports only 4KB\n",
2456 /* Currently, support only 4KB block size */
2457 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2458 if (blocksize != F2FS_BLKSIZE) {
2459 f2fs_msg(sb, KERN_INFO,
2460 "Invalid blocksize (%u), supports only 4KB\n",
2465 /* check log blocks per segment */
2466 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2467 f2fs_msg(sb, KERN_INFO,
2468 "Invalid log blocks per segment (%u)\n",
2469 le32_to_cpu(raw_super->log_blocks_per_seg));
2473 /* Currently, support 512/1024/2048/4096 bytes sector size */
2474 if (le32_to_cpu(raw_super->log_sectorsize) >
2475 F2FS_MAX_LOG_SECTOR_SIZE ||
2476 le32_to_cpu(raw_super->log_sectorsize) <
2477 F2FS_MIN_LOG_SECTOR_SIZE) {
2478 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
2479 le32_to_cpu(raw_super->log_sectorsize));
2482 if (le32_to_cpu(raw_super->log_sectors_per_block) +
2483 le32_to_cpu(raw_super->log_sectorsize) !=
2484 F2FS_MAX_LOG_SECTOR_SIZE) {
2485 f2fs_msg(sb, KERN_INFO,
2486 "Invalid log sectors per block(%u) log sectorsize(%u)",
2487 le32_to_cpu(raw_super->log_sectors_per_block),
2488 le32_to_cpu(raw_super->log_sectorsize));
2492 segment_count = le32_to_cpu(raw_super->segment_count);
2493 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2494 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2495 total_sections = le32_to_cpu(raw_super->section_count);
2497 /* blocks_per_seg should be 512, given the above check */
2498 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2500 if (segment_count > F2FS_MAX_SEGMENT ||
2501 segment_count < F2FS_MIN_SEGMENTS) {
2502 f2fs_msg(sb, KERN_INFO,
2503 "Invalid segment count (%u)",
2508 if (total_sections > segment_count ||
2509 total_sections < F2FS_MIN_SEGMENTS ||
2510 segs_per_sec > segment_count || !segs_per_sec) {
2511 f2fs_msg(sb, KERN_INFO,
2512 "Invalid segment/section count (%u, %u x %u)",
2513 segment_count, total_sections, segs_per_sec);
2517 if ((segment_count / segs_per_sec) < total_sections) {
2518 f2fs_msg(sb, KERN_INFO,
2519 "Small segment_count (%u < %u * %u)",
2520 segment_count, segs_per_sec, total_sections);
2524 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2525 f2fs_msg(sb, KERN_INFO,
2526 "Wrong segment_count / block_count (%u > %llu)",
2527 segment_count, le64_to_cpu(raw_super->block_count));
2531 if (secs_per_zone > total_sections || !secs_per_zone) {
2532 f2fs_msg(sb, KERN_INFO,
2533 "Wrong secs_per_zone / total_sections (%u, %u)",
2534 secs_per_zone, total_sections);
2537 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2538 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2539 (le32_to_cpu(raw_super->extension_count) +
2540 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2541 f2fs_msg(sb, KERN_INFO,
2542 "Corrupted extension count (%u + %u > %u)",
2543 le32_to_cpu(raw_super->extension_count),
2544 raw_super->hot_ext_count,
2545 F2FS_MAX_EXTENSION);
2549 if (le32_to_cpu(raw_super->cp_payload) >
2550 (blocks_per_seg - F2FS_CP_PACKS)) {
2551 f2fs_msg(sb, KERN_INFO,
2552 "Insane cp_payload (%u > %u)",
2553 le32_to_cpu(raw_super->cp_payload),
2554 blocks_per_seg - F2FS_CP_PACKS);
2558 /* check reserved ino info */
2559 if (le32_to_cpu(raw_super->node_ino) != 1 ||
2560 le32_to_cpu(raw_super->meta_ino) != 2 ||
2561 le32_to_cpu(raw_super->root_ino) != 3) {
2562 f2fs_msg(sb, KERN_INFO,
2563 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2564 le32_to_cpu(raw_super->node_ino),
2565 le32_to_cpu(raw_super->meta_ino),
2566 le32_to_cpu(raw_super->root_ino));
2570 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2571 if (sanity_check_area_boundary(sbi, bh))
2577 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2579 unsigned int total, fsmeta;
2580 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2581 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2582 unsigned int ovp_segments, reserved_segments;
2583 unsigned int main_segs, blocks_per_seg;
2584 unsigned int sit_segs, nat_segs;
2585 unsigned int sit_bitmap_size, nat_bitmap_size;
2586 unsigned int log_blocks_per_seg;
2587 unsigned int segment_count_main;
2588 unsigned int cp_pack_start_sum, cp_payload;
2589 block_t user_block_count;
2592 total = le32_to_cpu(raw_super->segment_count);
2593 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2594 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2596 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2598 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2599 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2601 if (unlikely(fsmeta >= total))
2604 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2605 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2607 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2608 ovp_segments == 0 || reserved_segments == 0)) {
2609 f2fs_msg(sbi->sb, KERN_ERR,
2610 "Wrong layout: check mkfs.f2fs version");
2614 user_block_count = le64_to_cpu(ckpt->user_block_count);
2615 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2616 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2617 if (!user_block_count || user_block_count >=
2618 segment_count_main << log_blocks_per_seg) {
2619 f2fs_msg(sbi->sb, KERN_ERR,
2620 "Wrong user_block_count: %u", user_block_count);
2624 main_segs = le32_to_cpu(raw_super->segment_count_main);
2625 blocks_per_seg = sbi->blocks_per_seg;
2627 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2628 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2629 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2631 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2632 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2633 le32_to_cpu(ckpt->cur_node_segno[j])) {
2634 f2fs_msg(sbi->sb, KERN_ERR,
2635 "Node segment (%u, %u) has the same "
2637 le32_to_cpu(ckpt->cur_node_segno[i]));
2642 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2643 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2644 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2646 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2647 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2648 le32_to_cpu(ckpt->cur_data_segno[j])) {
2649 f2fs_msg(sbi->sb, KERN_ERR,
2650 "Data segment (%u, %u) has the same "
2652 le32_to_cpu(ckpt->cur_data_segno[i]));
2657 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2658 for (j = i; j < NR_CURSEG_DATA_TYPE; j++) {
2659 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2660 le32_to_cpu(ckpt->cur_data_segno[j])) {
2661 f2fs_msg(sbi->sb, KERN_ERR,
2662 "Data segment (%u) and Data segment (%u)"
2663 " has the same segno: %u", i, j,
2664 le32_to_cpu(ckpt->cur_node_segno[i]));
2670 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2671 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2673 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2674 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2675 f2fs_msg(sbi->sb, KERN_ERR,
2676 "Wrong bitmap size: sit: %u, nat:%u",
2677 sit_bitmap_size, nat_bitmap_size);
2681 cp_pack_start_sum = __start_sum_addr(sbi);
2682 cp_payload = __cp_payload(sbi);
2683 if (cp_pack_start_sum < cp_payload + 1 ||
2684 cp_pack_start_sum > blocks_per_seg - 1 -
2686 f2fs_msg(sbi->sb, KERN_ERR,
2687 "Wrong cp_pack_start_sum: %u",
2692 if (unlikely(f2fs_cp_error(sbi))) {
2693 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
2699 static void init_sb_info(struct f2fs_sb_info *sbi)
2701 struct f2fs_super_block *raw_super = sbi->raw_super;
2704 sbi->log_sectors_per_block =
2705 le32_to_cpu(raw_super->log_sectors_per_block);
2706 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2707 sbi->blocksize = 1 << sbi->log_blocksize;
2708 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2709 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2710 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2711 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2712 sbi->total_sections = le32_to_cpu(raw_super->section_count);
2713 sbi->total_node_count =
2714 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2715 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2716 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2717 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2718 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2719 sbi->cur_victim_sec = NULL_SECNO;
2720 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
2721 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
2722 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2723 sbi->migration_granularity = sbi->segs_per_sec;
2725 sbi->dir_level = DEF_DIR_LEVEL;
2726 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2727 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2728 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
2729 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
2730 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
2731 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
2732 DEF_UMOUNT_DISCARD_TIMEOUT;
2733 clear_sbi_flag(sbi, SBI_NEED_FSCK);
2735 for (i = 0; i < NR_COUNT_TYPE; i++)
2736 atomic_set(&sbi->nr_pages[i], 0);
2738 for (i = 0; i < META; i++)
2739 atomic_set(&sbi->wb_sync_req[i], 0);
2741 INIT_LIST_HEAD(&sbi->s_list);
2742 mutex_init(&sbi->umount_mutex);
2743 init_rwsem(&sbi->io_order_lock);
2744 spin_lock_init(&sbi->cp_lock);
2746 sbi->dirty_device = 0;
2747 spin_lock_init(&sbi->dev_lock);
2749 init_rwsem(&sbi->sb_lock);
2752 static int init_percpu_info(struct f2fs_sb_info *sbi)
2756 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2760 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
2763 percpu_counter_destroy(&sbi->alloc_valid_block_count);
2768 #ifdef CONFIG_BLK_DEV_ZONED
2769 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2771 struct block_device *bdev = FDEV(devi).bdev;
2772 sector_t nr_sectors = bdev->bd_part->nr_sects;
2773 sector_t sector = 0;
2774 struct blk_zone *zones;
2775 unsigned int i, nr_zones;
2779 if (!f2fs_sb_has_blkzoned(sbi))
2782 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2783 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2785 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2786 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2787 __ilog2_u32(sbi->blocks_per_blkz))
2789 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2790 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2791 sbi->log_blocks_per_blkz;
2792 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2793 FDEV(devi).nr_blkz++;
2795 FDEV(devi).blkz_type = f2fs_kmalloc(sbi, FDEV(devi).nr_blkz,
2797 if (!FDEV(devi).blkz_type)
2800 #define F2FS_REPORT_NR_ZONES 4096
2802 zones = f2fs_kzalloc(sbi,
2803 array_size(F2FS_REPORT_NR_ZONES,
2804 sizeof(struct blk_zone)),
2809 /* Get block zones type */
2810 while (zones && sector < nr_sectors) {
2812 nr_zones = F2FS_REPORT_NR_ZONES;
2813 err = blkdev_report_zones(bdev, sector,
2823 for (i = 0; i < nr_zones; i++) {
2824 FDEV(devi).blkz_type[n] = zones[i].type;
2825 sector += zones[i].len;
2837 * Read f2fs raw super block.
2838 * Because we have two copies of super block, so read both of them
2839 * to get the first valid one. If any one of them is broken, we pass
2840 * them recovery flag back to the caller.
2842 static int read_raw_super_block(struct f2fs_sb_info *sbi,
2843 struct f2fs_super_block **raw_super,
2844 int *valid_super_block, int *recovery)
2846 struct super_block *sb = sbi->sb;
2848 struct buffer_head *bh;
2849 struct f2fs_super_block *super;
2852 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2856 for (block = 0; block < 2; block++) {
2857 bh = sb_bread(sb, block);
2859 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
2865 /* sanity checking of raw super */
2866 if (sanity_check_raw_super(sbi, bh)) {
2867 f2fs_msg(sb, KERN_ERR,
2868 "Can't find valid F2FS filesystem in %dth superblock",
2876 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2878 *valid_super_block = block;
2884 /* Fail to read any one of the superblocks*/
2888 /* No valid superblock */
2897 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
2899 struct buffer_head *bh;
2903 if ((recover && f2fs_readonly(sbi->sb)) ||
2904 bdev_read_only(sbi->sb->s_bdev)) {
2905 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2909 /* we should update superblock crc here */
2910 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
2911 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
2912 offsetof(struct f2fs_super_block, crc));
2913 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
2916 /* write back-up superblock first */
2917 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
2920 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2923 /* if we are in recovery path, skip writing valid superblock */
2927 /* write current valid superblock */
2928 bh = sb_bread(sbi->sb, sbi->valid_super_block);
2931 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2936 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
2938 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2939 unsigned int max_devices = MAX_DEVICES;
2942 /* Initialize single device information */
2943 if (!RDEV(0).path[0]) {
2944 if (!bdev_is_zoned(sbi->sb->s_bdev))
2950 * Initialize multiple devices information, or single
2951 * zoned block device information.
2953 sbi->devs = f2fs_kzalloc(sbi,
2954 array_size(max_devices,
2955 sizeof(struct f2fs_dev_info)),
2960 for (i = 0; i < max_devices; i++) {
2962 if (i > 0 && !RDEV(i).path[0])
2965 if (max_devices == 1) {
2966 /* Single zoned block device mount */
2968 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
2969 sbi->sb->s_mode, sbi->sb->s_type);
2971 /* Multi-device mount */
2972 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
2973 FDEV(i).total_segments =
2974 le32_to_cpu(RDEV(i).total_segments);
2976 FDEV(i).start_blk = 0;
2977 FDEV(i).end_blk = FDEV(i).start_blk +
2978 (FDEV(i).total_segments <<
2979 sbi->log_blocks_per_seg) - 1 +
2980 le32_to_cpu(raw_super->segment0_blkaddr);
2982 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
2983 FDEV(i).end_blk = FDEV(i).start_blk +
2984 (FDEV(i).total_segments <<
2985 sbi->log_blocks_per_seg) - 1;
2987 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
2988 sbi->sb->s_mode, sbi->sb->s_type);
2990 if (IS_ERR(FDEV(i).bdev))
2991 return PTR_ERR(FDEV(i).bdev);
2993 /* to release errored devices */
2994 sbi->s_ndevs = i + 1;
2996 #ifdef CONFIG_BLK_DEV_ZONED
2997 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
2998 !f2fs_sb_has_blkzoned(sbi)) {
2999 f2fs_msg(sbi->sb, KERN_ERR,
3000 "Zoned block device feature not enabled\n");
3003 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3004 if (init_blkz_info(sbi, i)) {
3005 f2fs_msg(sbi->sb, KERN_ERR,
3006 "Failed to initialize F2FS blkzone information");
3009 if (max_devices == 1)
3011 f2fs_msg(sbi->sb, KERN_INFO,
3012 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3014 FDEV(i).total_segments,
3015 FDEV(i).start_blk, FDEV(i).end_blk,
3016 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3017 "Host-aware" : "Host-managed");
3021 f2fs_msg(sbi->sb, KERN_INFO,
3022 "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3024 FDEV(i).total_segments,
3025 FDEV(i).start_blk, FDEV(i).end_blk);
3027 f2fs_msg(sbi->sb, KERN_INFO,
3028 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3032 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3034 struct f2fs_sm_info *sm_i = SM_I(sbi);
3036 /* adjust parameters according to the volume size */
3037 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3038 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3039 sm_i->dcc_info->discard_granularity = 1;
3040 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3043 sbi->readdir_ra = 1;
3046 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3048 struct f2fs_sb_info *sbi;
3049 struct f2fs_super_block *raw_super;
3052 bool skip_recovery = false, need_fsck = false;
3053 char *options = NULL;
3054 int recovery, i, valid_super_block;
3055 struct curseg_info *seg_i;
3061 valid_super_block = -1;
3064 /* allocate memory for f2fs-specific super block info */
3065 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3071 /* Load the checksum driver */
3072 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3073 if (IS_ERR(sbi->s_chksum_driver)) {
3074 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
3075 err = PTR_ERR(sbi->s_chksum_driver);
3076 sbi->s_chksum_driver = NULL;
3080 /* set a block size */
3081 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3082 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
3086 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3091 sb->s_fs_info = sbi;
3092 sbi->raw_super = raw_super;
3094 /* precompute checksum seed for metadata */
3095 if (f2fs_sb_has_inode_chksum(sbi))
3096 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3097 sizeof(raw_super->uuid));
3100 * The BLKZONED feature indicates that the drive was formatted with
3101 * zone alignment optimization. This is optional for host-aware
3102 * devices, but mandatory for host-managed zoned block devices.
3104 #ifndef CONFIG_BLK_DEV_ZONED
3105 if (f2fs_sb_has_blkzoned(sbi)) {
3106 f2fs_msg(sb, KERN_ERR,
3107 "Zoned block device support is not enabled\n");
3112 default_options(sbi);
3113 /* parse mount options */
3114 options = kstrdup((const char *)data, GFP_KERNEL);
3115 if (data && !options) {
3120 err = parse_options(sb, options);
3124 sbi->max_file_blocks = max_file_blocks();
3125 sb->s_maxbytes = sbi->max_file_blocks <<
3126 le32_to_cpu(raw_super->log_blocksize);
3127 sb->s_max_links = F2FS_LINK_MAX;
3130 sb->dq_op = &f2fs_quota_operations;
3131 if (f2fs_sb_has_quota_ino(sbi))
3132 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3134 sb->s_qcop = &f2fs_quotactl_ops;
3135 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3137 if (f2fs_sb_has_quota_ino(sbi)) {
3138 for (i = 0; i < MAXQUOTAS; i++) {
3139 if (f2fs_qf_ino(sbi->sb, i))
3140 sbi->nquota_files++;
3145 sb->s_op = &f2fs_sops;
3146 #ifdef CONFIG_FS_ENCRYPTION
3147 sb->s_cop = &f2fs_cryptops;
3149 sb->s_xattr = f2fs_xattr_handlers;
3150 sb->s_export_op = &f2fs_export_ops;
3151 sb->s_magic = F2FS_SUPER_MAGIC;
3152 sb->s_time_gran = 1;
3153 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3154 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3155 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3156 sb->s_iflags |= SB_I_CGROUPWB;
3158 /* init f2fs-specific super block info */
3159 sbi->valid_super_block = valid_super_block;
3160 mutex_init(&sbi->gc_mutex);
3161 mutex_init(&sbi->writepages);
3162 mutex_init(&sbi->cp_mutex);
3163 init_rwsem(&sbi->node_write);
3164 init_rwsem(&sbi->node_change);
3166 /* disallow all the data/node/meta page writes */
3167 set_sbi_flag(sbi, SBI_POR_DOING);
3168 spin_lock_init(&sbi->stat_lock);
3170 /* init iostat info */
3171 spin_lock_init(&sbi->iostat_lock);
3172 sbi->iostat_enable = false;
3174 for (i = 0; i < NR_PAGE_TYPE; i++) {
3175 int n = (i == META) ? 1: NR_TEMP_TYPE;
3181 sizeof(struct f2fs_bio_info)),
3183 if (!sbi->write_io[i]) {
3188 for (j = HOT; j < n; j++) {
3189 init_rwsem(&sbi->write_io[i][j].io_rwsem);
3190 sbi->write_io[i][j].sbi = sbi;
3191 sbi->write_io[i][j].bio = NULL;
3192 spin_lock_init(&sbi->write_io[i][j].io_lock);
3193 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3197 init_rwsem(&sbi->cp_rwsem);
3198 init_waitqueue_head(&sbi->cp_wait);
3201 err = init_percpu_info(sbi);
3205 if (F2FS_IO_SIZE(sbi) > 1) {
3206 sbi->write_io_dummy =
3207 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3208 if (!sbi->write_io_dummy) {
3214 /* get an inode for meta space */
3215 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3216 if (IS_ERR(sbi->meta_inode)) {
3217 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
3218 err = PTR_ERR(sbi->meta_inode);
3222 err = f2fs_get_valid_checkpoint(sbi);
3224 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
3225 goto free_meta_inode;
3228 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3229 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3230 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3231 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3232 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3235 /* Initialize device list */
3236 err = f2fs_scan_devices(sbi);
3238 f2fs_msg(sb, KERN_ERR, "Failed to find devices");
3242 sbi->total_valid_node_count =
3243 le32_to_cpu(sbi->ckpt->valid_node_count);
3244 percpu_counter_set(&sbi->total_valid_inode_count,
3245 le32_to_cpu(sbi->ckpt->valid_inode_count));
3246 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3247 sbi->total_valid_block_count =
3248 le64_to_cpu(sbi->ckpt->valid_block_count);
3249 sbi->last_valid_block_count = sbi->total_valid_block_count;
3250 sbi->reserved_blocks = 0;
3251 sbi->current_reserved_blocks = 0;
3252 limit_reserve_root(sbi);
3254 for (i = 0; i < NR_INODE_TYPE; i++) {
3255 INIT_LIST_HEAD(&sbi->inode_list[i]);
3256 spin_lock_init(&sbi->inode_lock[i]);
3259 f2fs_init_extent_cache_info(sbi);
3261 f2fs_init_ino_entry_info(sbi);
3263 f2fs_init_fsync_node_info(sbi);
3265 /* setup f2fs internal modules */
3266 err = f2fs_build_segment_manager(sbi);
3268 f2fs_msg(sb, KERN_ERR,
3269 "Failed to initialize F2FS segment manager");
3272 err = f2fs_build_node_manager(sbi);
3274 f2fs_msg(sb, KERN_ERR,
3275 "Failed to initialize F2FS node manager");
3279 /* For write statistics */
3280 if (sb->s_bdev->bd_part)
3281 sbi->sectors_written_start =
3282 (u64)part_stat_read(sb->s_bdev->bd_part,
3283 sectors[STAT_WRITE]);
3285 /* Read accumulated write IO statistics if exists */
3286 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3287 if (__exist_node_summaries(sbi))
3288 sbi->kbytes_written =
3289 le64_to_cpu(seg_i->journal->info.kbytes_written);
3291 f2fs_build_gc_manager(sbi);
3293 err = f2fs_build_stats(sbi);
3297 /* get an inode for node space */
3298 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3299 if (IS_ERR(sbi->node_inode)) {
3300 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
3301 err = PTR_ERR(sbi->node_inode);
3305 /* read root inode and dentry */
3306 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3308 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
3309 err = PTR_ERR(root);
3310 goto free_node_inode;
3312 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3313 !root->i_size || !root->i_nlink) {
3316 goto free_node_inode;
3319 sb->s_root = d_make_root(root); /* allocate root dentry */
3322 goto free_node_inode;