4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
11 #include <linux/module.h>
12 #include <linux/init.h>
14 #include <linux/statfs.h>
15 #include <linux/buffer_head.h>
16 #include <linux/backing-dev.h>
17 #include <linux/kthread.h>
18 #include <linux/parser.h>
19 #include <linux/mount.h>
20 #include <linux/seq_file.h>
21 #include <linux/proc_fs.h>
22 #include <linux/random.h>
23 #include <linux/exportfs.h>
24 #include <linux/blkdev.h>
25 #include <linux/quotaops.h>
26 #include <linux/f2fs_fs.h>
27 #include <linux/sysfs.h>
28 #include <linux/quota.h>
37 #define CREATE_TRACE_POINTS
38 #include <trace/events/f2fs.h>
40 static struct kmem_cache *f2fs_inode_cachep;
42 #ifdef CONFIG_F2FS_FAULT_INJECTION
44 char *fault_name[FAULT_MAX] = {
45 [FAULT_KMALLOC] = "kmalloc",
46 [FAULT_KVMALLOC] = "kvmalloc",
47 [FAULT_PAGE_ALLOC] = "page alloc",
48 [FAULT_PAGE_GET] = "page get",
49 [FAULT_ALLOC_BIO] = "alloc bio",
50 [FAULT_ALLOC_NID] = "alloc nid",
51 [FAULT_ORPHAN] = "orphan",
52 [FAULT_BLOCK] = "no more block",
53 [FAULT_DIR_DEPTH] = "too big dir depth",
54 [FAULT_EVICT_INODE] = "evict_inode fail",
55 [FAULT_TRUNCATE] = "truncate fail",
56 [FAULT_IO] = "IO error",
57 [FAULT_CHECKPOINT] = "checkpoint error",
60 static void f2fs_build_fault_attr(struct f2fs_sb_info *sbi,
63 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
66 atomic_set(&ffi->inject_ops, 0);
67 ffi->inject_rate = rate;
68 ffi->inject_type = (1 << FAULT_MAX) - 1;
70 memset(ffi, 0, sizeof(struct f2fs_fault_info));
75 /* f2fs-wide shrinker description */
76 static struct shrinker f2fs_shrinker_info = {
77 .scan_objects = f2fs_shrink_scan,
78 .count_objects = f2fs_shrink_count,
79 .seeks = DEFAULT_SEEKS,
84 Opt_disable_roll_forward,
95 Opt_disable_ext_identify,
98 Opt_inline_xattr_size,
135 Opt_test_dummy_encryption,
139 static match_table_t f2fs_tokens = {
140 {Opt_gc_background, "background_gc=%s"},
141 {Opt_disable_roll_forward, "disable_roll_forward"},
142 {Opt_norecovery, "norecovery"},
143 {Opt_discard, "discard"},
144 {Opt_nodiscard, "nodiscard"},
145 {Opt_noheap, "no_heap"},
147 {Opt_user_xattr, "user_xattr"},
148 {Opt_nouser_xattr, "nouser_xattr"},
150 {Opt_noacl, "noacl"},
151 {Opt_active_logs, "active_logs=%u"},
152 {Opt_disable_ext_identify, "disable_ext_identify"},
153 {Opt_inline_xattr, "inline_xattr"},
154 {Opt_noinline_xattr, "noinline_xattr"},
155 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
156 {Opt_inline_data, "inline_data"},
157 {Opt_inline_dentry, "inline_dentry"},
158 {Opt_noinline_dentry, "noinline_dentry"},
159 {Opt_flush_merge, "flush_merge"},
160 {Opt_noflush_merge, "noflush_merge"},
161 {Opt_nobarrier, "nobarrier"},
162 {Opt_fastboot, "fastboot"},
163 {Opt_extent_cache, "extent_cache"},
164 {Opt_noextent_cache, "noextent_cache"},
165 {Opt_noinline_data, "noinline_data"},
166 {Opt_data_flush, "data_flush"},
167 {Opt_reserve_root, "reserve_root=%u"},
168 {Opt_resgid, "resgid=%u"},
169 {Opt_resuid, "resuid=%u"},
170 {Opt_mode, "mode=%s"},
171 {Opt_io_size_bits, "io_bits=%u"},
172 {Opt_fault_injection, "fault_injection=%u"},
173 {Opt_lazytime, "lazytime"},
174 {Opt_nolazytime, "nolazytime"},
175 {Opt_quota, "quota"},
176 {Opt_noquota, "noquota"},
177 {Opt_usrquota, "usrquota"},
178 {Opt_grpquota, "grpquota"},
179 {Opt_prjquota, "prjquota"},
180 {Opt_usrjquota, "usrjquota=%s"},
181 {Opt_grpjquota, "grpjquota=%s"},
182 {Opt_prjjquota, "prjjquota=%s"},
183 {Opt_offusrjquota, "usrjquota="},
184 {Opt_offgrpjquota, "grpjquota="},
185 {Opt_offprjjquota, "prjjquota="},
186 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
187 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
188 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
189 {Opt_whint, "whint_mode=%s"},
190 {Opt_alloc, "alloc_mode=%s"},
191 {Opt_fsync, "fsync_mode=%s"},
192 {Opt_test_dummy_encryption, "test_dummy_encryption"},
196 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
198 struct va_format vaf;
204 printk_ratelimited("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
208 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
210 block_t limit = (sbi->user_block_count << 1) / 1000;
213 if (test_opt(sbi, RESERVE_ROOT) &&
214 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
215 F2FS_OPTION(sbi).root_reserved_blocks = limit;
216 f2fs_msg(sbi->sb, KERN_INFO,
217 "Reduce reserved blocks for root = %u",
218 F2FS_OPTION(sbi).root_reserved_blocks);
220 if (!test_opt(sbi, RESERVE_ROOT) &&
221 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
222 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
223 !gid_eq(F2FS_OPTION(sbi).s_resgid,
224 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
225 f2fs_msg(sbi->sb, KERN_INFO,
226 "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
227 from_kuid_munged(&init_user_ns,
228 F2FS_OPTION(sbi).s_resuid),
229 from_kgid_munged(&init_user_ns,
230 F2FS_OPTION(sbi).s_resgid));
233 static void init_once(void *foo)
235 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
237 inode_init_once(&fi->vfs_inode);
241 static const char * const quotatypes[] = INITQFNAMES;
242 #define QTYPE2NAME(t) (quotatypes[t])
243 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
246 struct f2fs_sb_info *sbi = F2FS_SB(sb);
250 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
251 f2fs_msg(sb, KERN_ERR,
252 "Cannot change journaled "
253 "quota options when quota turned on");
256 if (f2fs_sb_has_quota_ino(sb)) {
257 f2fs_msg(sb, KERN_INFO,
258 "QUOTA feature is enabled, so ignore qf_name");
262 qname = match_strdup(args);
264 f2fs_msg(sb, KERN_ERR,
265 "Not enough memory for storing quotafile name");
268 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
269 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
272 f2fs_msg(sb, KERN_ERR,
273 "%s quota file already specified",
277 if (strchr(qname, '/')) {
278 f2fs_msg(sb, KERN_ERR,
279 "quotafile must be on filesystem root");
282 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
290 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
292 struct f2fs_sb_info *sbi = F2FS_SB(sb);
294 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
295 f2fs_msg(sb, KERN_ERR, "Cannot change journaled quota options"
296 " when quota turned on");
299 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
300 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
304 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
307 * We do the test below only for project quotas. 'usrquota' and
308 * 'grpquota' mount options are allowed even without quota feature
309 * to support legacy quotas in quota files.
311 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi->sb)) {
312 f2fs_msg(sbi->sb, KERN_ERR, "Project quota feature not enabled. "
313 "Cannot enable project quota enforcement.");
316 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
317 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
318 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
319 if (test_opt(sbi, USRQUOTA) &&
320 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
321 clear_opt(sbi, USRQUOTA);
323 if (test_opt(sbi, GRPQUOTA) &&
324 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
325 clear_opt(sbi, GRPQUOTA);
327 if (test_opt(sbi, PRJQUOTA) &&
328 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
329 clear_opt(sbi, PRJQUOTA);
331 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
332 test_opt(sbi, PRJQUOTA)) {
333 f2fs_msg(sbi->sb, KERN_ERR, "old and new quota "
338 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
339 f2fs_msg(sbi->sb, KERN_ERR, "journaled quota format "
345 if (f2fs_sb_has_quota_ino(sbi->sb) && F2FS_OPTION(sbi).s_jquota_fmt) {
346 f2fs_msg(sbi->sb, KERN_INFO,
347 "QUOTA feature is enabled, so ignore jquota_fmt");
348 F2FS_OPTION(sbi).s_jquota_fmt = 0;
350 if (f2fs_sb_has_quota_ino(sbi->sb) && f2fs_readonly(sbi->sb)) {
351 f2fs_msg(sbi->sb, KERN_INFO,
352 "Filesystem with quota feature cannot be mounted RDWR "
353 "without CONFIG_QUOTA");
360 static int parse_options(struct super_block *sb, char *options)
362 struct f2fs_sb_info *sbi = F2FS_SB(sb);
363 struct request_queue *q;
364 substring_t args[MAX_OPT_ARGS];
376 while ((p = strsep(&options, ",")) != NULL) {
381 * Initialize args struct so we know whether arg was
382 * found; some options take optional arguments.
384 args[0].to = args[0].from = NULL;
385 token = match_token(p, f2fs_tokens, args);
388 case Opt_gc_background:
389 name = match_strdup(&args[0]);
393 if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
395 clear_opt(sbi, FORCE_FG_GC);
396 } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
397 clear_opt(sbi, BG_GC);
398 clear_opt(sbi, FORCE_FG_GC);
399 } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
401 set_opt(sbi, FORCE_FG_GC);
408 case Opt_disable_roll_forward:
409 set_opt(sbi, DISABLE_ROLL_FORWARD);
412 /* this option mounts f2fs with ro */
413 set_opt(sbi, DISABLE_ROLL_FORWARD);
414 if (!f2fs_readonly(sb))
418 q = bdev_get_queue(sb->s_bdev);
419 if (blk_queue_discard(q)) {
420 set_opt(sbi, DISCARD);
421 } else if (!f2fs_sb_has_blkzoned(sb)) {
422 f2fs_msg(sb, KERN_WARNING,
423 "mounting with \"discard\" option, but "
424 "the device does not support discard");
428 if (f2fs_sb_has_blkzoned(sb)) {
429 f2fs_msg(sb, KERN_WARNING,
430 "discard is required for zoned block devices");
433 clear_opt(sbi, DISCARD);
436 set_opt(sbi, NOHEAP);
439 clear_opt(sbi, NOHEAP);
441 #ifdef CONFIG_F2FS_FS_XATTR
443 set_opt(sbi, XATTR_USER);
445 case Opt_nouser_xattr:
446 clear_opt(sbi, XATTR_USER);
448 case Opt_inline_xattr:
449 set_opt(sbi, INLINE_XATTR);
451 case Opt_noinline_xattr:
452 clear_opt(sbi, INLINE_XATTR);
454 case Opt_inline_xattr_size:
455 if (args->from && match_int(args, &arg))
457 set_opt(sbi, INLINE_XATTR_SIZE);
458 F2FS_OPTION(sbi).inline_xattr_size = arg;
462 f2fs_msg(sb, KERN_INFO,
463 "user_xattr options not supported");
465 case Opt_nouser_xattr:
466 f2fs_msg(sb, KERN_INFO,
467 "nouser_xattr options not supported");
469 case Opt_inline_xattr:
470 f2fs_msg(sb, KERN_INFO,
471 "inline_xattr options not supported");
473 case Opt_noinline_xattr:
474 f2fs_msg(sb, KERN_INFO,
475 "noinline_xattr options not supported");
478 #ifdef CONFIG_F2FS_FS_POSIX_ACL
480 set_opt(sbi, POSIX_ACL);
483 clear_opt(sbi, POSIX_ACL);
487 f2fs_msg(sb, KERN_INFO, "acl options not supported");
490 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
493 case Opt_active_logs:
494 if (args->from && match_int(args, &arg))
496 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
498 F2FS_OPTION(sbi).active_logs = arg;
500 case Opt_disable_ext_identify:
501 set_opt(sbi, DISABLE_EXT_IDENTIFY);
503 case Opt_inline_data:
504 set_opt(sbi, INLINE_DATA);
506 case Opt_inline_dentry:
507 set_opt(sbi, INLINE_DENTRY);
509 case Opt_noinline_dentry:
510 clear_opt(sbi, INLINE_DENTRY);
512 case Opt_flush_merge:
513 set_opt(sbi, FLUSH_MERGE);
515 case Opt_noflush_merge:
516 clear_opt(sbi, FLUSH_MERGE);
519 set_opt(sbi, NOBARRIER);
522 set_opt(sbi, FASTBOOT);
524 case Opt_extent_cache:
525 set_opt(sbi, EXTENT_CACHE);
527 case Opt_noextent_cache:
528 clear_opt(sbi, EXTENT_CACHE);
530 case Opt_noinline_data:
531 clear_opt(sbi, INLINE_DATA);
534 set_opt(sbi, DATA_FLUSH);
536 case Opt_reserve_root:
537 if (args->from && match_int(args, &arg))
539 if (test_opt(sbi, RESERVE_ROOT)) {
540 f2fs_msg(sb, KERN_INFO,
541 "Preserve previous reserve_root=%u",
542 F2FS_OPTION(sbi).root_reserved_blocks);
544 F2FS_OPTION(sbi).root_reserved_blocks = arg;
545 set_opt(sbi, RESERVE_ROOT);
549 if (args->from && match_int(args, &arg))
551 uid = make_kuid(current_user_ns(), arg);
552 if (!uid_valid(uid)) {
553 f2fs_msg(sb, KERN_ERR,
554 "Invalid uid value %d", arg);
557 F2FS_OPTION(sbi).s_resuid = uid;
560 if (args->from && match_int(args, &arg))
562 gid = make_kgid(current_user_ns(), arg);
563 if (!gid_valid(gid)) {
564 f2fs_msg(sb, KERN_ERR,
565 "Invalid gid value %d", arg);
568 F2FS_OPTION(sbi).s_resgid = gid;
571 name = match_strdup(&args[0]);
575 if (strlen(name) == 8 &&
576 !strncmp(name, "adaptive", 8)) {
577 if (f2fs_sb_has_blkzoned(sb)) {
578 f2fs_msg(sb, KERN_WARNING,
579 "adaptive mode is not allowed with "
580 "zoned block device feature");
584 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
585 } else if (strlen(name) == 3 &&
586 !strncmp(name, "lfs", 3)) {
587 set_opt_mode(sbi, F2FS_MOUNT_LFS);
594 case Opt_io_size_bits:
595 if (args->from && match_int(args, &arg))
597 if (arg > __ilog2_u32(BIO_MAX_PAGES)) {
598 f2fs_msg(sb, KERN_WARNING,
599 "Not support %d, larger than %d",
600 1 << arg, BIO_MAX_PAGES);
603 F2FS_OPTION(sbi).write_io_size_bits = arg;
605 case Opt_fault_injection:
606 if (args->from && match_int(args, &arg))
608 #ifdef CONFIG_F2FS_FAULT_INJECTION
609 f2fs_build_fault_attr(sbi, arg);
610 set_opt(sbi, FAULT_INJECTION);
612 f2fs_msg(sb, KERN_INFO,
613 "FAULT_INJECTION was not selected");
617 sb->s_flags |= SB_LAZYTIME;
620 sb->s_flags &= ~SB_LAZYTIME;
625 set_opt(sbi, USRQUOTA);
628 set_opt(sbi, GRPQUOTA);
631 set_opt(sbi, PRJQUOTA);
634 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
639 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
644 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
648 case Opt_offusrjquota:
649 ret = f2fs_clear_qf_name(sb, USRQUOTA);
653 case Opt_offgrpjquota:
654 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
658 case Opt_offprjjquota:
659 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
663 case Opt_jqfmt_vfsold:
664 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
666 case Opt_jqfmt_vfsv0:
667 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
669 case Opt_jqfmt_vfsv1:
670 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
673 clear_opt(sbi, QUOTA);
674 clear_opt(sbi, USRQUOTA);
675 clear_opt(sbi, GRPQUOTA);
676 clear_opt(sbi, PRJQUOTA);
686 case Opt_offusrjquota:
687 case Opt_offgrpjquota:
688 case Opt_offprjjquota:
689 case Opt_jqfmt_vfsold:
690 case Opt_jqfmt_vfsv0:
691 case Opt_jqfmt_vfsv1:
693 f2fs_msg(sb, KERN_INFO,
694 "quota operations not supported");
698 name = match_strdup(&args[0]);
701 if (strlen(name) == 10 &&
702 !strncmp(name, "user-based", 10)) {
703 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
704 } else if (strlen(name) == 3 &&
705 !strncmp(name, "off", 3)) {
706 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
707 } else if (strlen(name) == 8 &&
708 !strncmp(name, "fs-based", 8)) {
709 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
717 name = match_strdup(&args[0]);
721 if (strlen(name) == 7 &&
722 !strncmp(name, "default", 7)) {
723 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
724 } else if (strlen(name) == 5 &&
725 !strncmp(name, "reuse", 5)) {
726 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
734 name = match_strdup(&args[0]);
737 if (strlen(name) == 5 &&
738 !strncmp(name, "posix", 5)) {
739 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
740 } else if (strlen(name) == 6 &&
741 !strncmp(name, "strict", 6)) {
742 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
743 } else if (strlen(name) == 9 &&
744 !strncmp(name, "nobarrier", 9)) {
745 F2FS_OPTION(sbi).fsync_mode =
746 FSYNC_MODE_NOBARRIER;
753 case Opt_test_dummy_encryption:
754 #ifdef CONFIG_F2FS_FS_ENCRYPTION
755 if (!f2fs_sb_has_encrypt(sb)) {
756 f2fs_msg(sb, KERN_ERR, "Encrypt feature is off");
760 F2FS_OPTION(sbi).test_dummy_encryption = true;
761 f2fs_msg(sb, KERN_INFO,
762 "Test dummy encryption mode enabled");
764 f2fs_msg(sb, KERN_INFO,
765 "Test dummy encryption mount option ignored");
769 f2fs_msg(sb, KERN_ERR,
770 "Unrecognized mount option \"%s\" or missing value",
776 if (f2fs_check_quota_options(sbi))
780 if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) {
781 f2fs_msg(sb, KERN_ERR,
782 "Should set mode=lfs with %uKB-sized IO",
783 F2FS_IO_SIZE_KB(sbi));
787 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
788 if (!f2fs_sb_has_extra_attr(sb) ||
789 !f2fs_sb_has_flexible_inline_xattr(sb)) {
790 f2fs_msg(sb, KERN_ERR,
791 "extra_attr or flexible_inline_xattr "
795 if (!test_opt(sbi, INLINE_XATTR)) {
796 f2fs_msg(sb, KERN_ERR,
797 "inline_xattr_size option should be "
798 "set with inline_xattr option");
801 if (!F2FS_OPTION(sbi).inline_xattr_size ||
802 F2FS_OPTION(sbi).inline_xattr_size >=
803 DEF_ADDRS_PER_INODE -
804 F2FS_TOTAL_EXTRA_ATTR_SIZE -
805 DEF_INLINE_RESERVED_SIZE -
806 DEF_MIN_INLINE_SIZE) {
807 f2fs_msg(sb, KERN_ERR,
808 "inline xattr size is out of range");
813 /* Not pass down write hints if the number of active logs is lesser
814 * than NR_CURSEG_TYPE.
816 if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
817 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
821 static struct inode *f2fs_alloc_inode(struct super_block *sb)
823 struct f2fs_inode_info *fi;
825 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
829 init_once((void *) fi);
831 /* Initialize f2fs-specific inode info */
832 atomic_set(&fi->dirty_pages, 0);
833 init_rwsem(&fi->i_sem);
834 INIT_LIST_HEAD(&fi->dirty_list);
835 INIT_LIST_HEAD(&fi->gdirty_list);
836 INIT_LIST_HEAD(&fi->inmem_ilist);
837 INIT_LIST_HEAD(&fi->inmem_pages);
838 mutex_init(&fi->inmem_lock);
839 init_rwsem(&fi->i_gc_rwsem[READ]);
840 init_rwsem(&fi->i_gc_rwsem[WRITE]);
841 init_rwsem(&fi->i_mmap_sem);
842 init_rwsem(&fi->i_xattr_sem);
844 /* Will be used by directory only */
845 fi->i_dir_level = F2FS_SB(sb)->dir_level;
847 return &fi->vfs_inode;
850 static int f2fs_drop_inode(struct inode *inode)
854 * This is to avoid a deadlock condition like below.
855 * writeback_single_inode(inode)
856 * - f2fs_write_data_page
857 * - f2fs_gc -> iput -> evict
858 * - inode_wait_for_writeback(inode)
860 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
861 if (!inode->i_nlink && !is_bad_inode(inode)) {
862 /* to avoid evict_inode call simultaneously */
863 atomic_inc(&inode->i_count);
864 spin_unlock(&inode->i_lock);
866 /* some remained atomic pages should discarded */
867 if (f2fs_is_atomic_file(inode))
868 f2fs_drop_inmem_pages(inode);
870 /* should remain fi->extent_tree for writepage */
871 f2fs_destroy_extent_node(inode);
873 sb_start_intwrite(inode->i_sb);
874 f2fs_i_size_write(inode, 0);
876 if (F2FS_HAS_BLOCKS(inode))
877 f2fs_truncate(inode);
879 sb_end_intwrite(inode->i_sb);
881 spin_lock(&inode->i_lock);
882 atomic_dec(&inode->i_count);
884 trace_f2fs_drop_inode(inode, 0);
887 ret = generic_drop_inode(inode);
888 trace_f2fs_drop_inode(inode, ret);
892 int f2fs_inode_dirtied(struct inode *inode, bool sync)
894 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
897 spin_lock(&sbi->inode_lock[DIRTY_META]);
898 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
901 set_inode_flag(inode, FI_DIRTY_INODE);
902 stat_inc_dirty_inode(sbi, DIRTY_META);
904 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
905 list_add_tail(&F2FS_I(inode)->gdirty_list,
906 &sbi->inode_list[DIRTY_META]);
907 inc_page_count(sbi, F2FS_DIRTY_IMETA);
909 spin_unlock(&sbi->inode_lock[DIRTY_META]);
913 void f2fs_inode_synced(struct inode *inode)
915 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
917 spin_lock(&sbi->inode_lock[DIRTY_META]);
918 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
919 spin_unlock(&sbi->inode_lock[DIRTY_META]);
922 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
923 list_del_init(&F2FS_I(inode)->gdirty_list);
924 dec_page_count(sbi, F2FS_DIRTY_IMETA);
926 clear_inode_flag(inode, FI_DIRTY_INODE);
927 clear_inode_flag(inode, FI_AUTO_RECOVER);
928 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
929 spin_unlock(&sbi->inode_lock[DIRTY_META]);
933 * f2fs_dirty_inode() is called from __mark_inode_dirty()
935 * We should call set_dirty_inode to write the dirty inode through write_inode.
937 static void f2fs_dirty_inode(struct inode *inode, int flags)
939 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
941 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
942 inode->i_ino == F2FS_META_INO(sbi))
945 if (flags == I_DIRTY_TIME)
948 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
949 clear_inode_flag(inode, FI_AUTO_RECOVER);
951 f2fs_inode_dirtied(inode, false);
954 static void f2fs_i_callback(struct rcu_head *head)
956 struct inode *inode = container_of(head, struct inode, i_rcu);
957 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
960 static void f2fs_destroy_inode(struct inode *inode)
962 call_rcu(&inode->i_rcu, f2fs_i_callback);
965 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
967 percpu_counter_destroy(&sbi->alloc_valid_block_count);
968 percpu_counter_destroy(&sbi->total_valid_inode_count);
971 static void destroy_device_list(struct f2fs_sb_info *sbi)
975 for (i = 0; i < sbi->s_ndevs; i++) {
976 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
977 #ifdef CONFIG_BLK_DEV_ZONED
978 kfree(FDEV(i).blkz_type);
984 static void f2fs_put_super(struct super_block *sb)
986 struct f2fs_sb_info *sbi = F2FS_SB(sb);
990 f2fs_quota_off_umount(sb);
992 /* prevent remaining shrinker jobs */
993 mutex_lock(&sbi->umount_mutex);
996 * We don't need to do checkpoint when superblock is clean.
997 * But, the previous checkpoint was not done by umount, it needs to do
998 * clean checkpoint again.
1000 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1001 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
1002 struct cp_control cpc = {
1003 .reason = CP_UMOUNT,
1005 f2fs_write_checkpoint(sbi, &cpc);
1008 /* be sure to wait for any on-going discard commands */
1009 dropped = f2fs_wait_discard_bios(sbi);
1011 if (f2fs_discard_en(sbi) && !sbi->discard_blks && !dropped) {
1012 struct cp_control cpc = {
1013 .reason = CP_UMOUNT | CP_TRIMMED,
1015 f2fs_write_checkpoint(sbi, &cpc);
1018 /* f2fs_write_checkpoint can update stat informaion */
1019 f2fs_destroy_stats(sbi);
1022 * normally superblock is clean, so we need to release this.
1023 * In addition, EIO will skip do checkpoint, we need this as well.
1025 f2fs_release_ino_entry(sbi, true);
1027 f2fs_leave_shrinker(sbi);
1028 mutex_unlock(&sbi->umount_mutex);
1030 /* our cp_error case, we can wait for any writeback page */
1031 f2fs_flush_merged_writes(sbi);
1033 iput(sbi->node_inode);
1034 iput(sbi->meta_inode);
1036 /* destroy f2fs internal modules */
1037 f2fs_destroy_node_manager(sbi);
1038 f2fs_destroy_segment_manager(sbi);
1042 f2fs_unregister_sysfs(sbi);
1044 sb->s_fs_info = NULL;
1045 if (sbi->s_chksum_driver)
1046 crypto_free_shash(sbi->s_chksum_driver);
1047 kfree(sbi->raw_super);
1049 destroy_device_list(sbi);
1050 mempool_destroy(sbi->write_io_dummy);
1052 for (i = 0; i < MAXQUOTAS; i++)
1053 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1055 destroy_percpu_info(sbi);
1056 for (i = 0; i < NR_PAGE_TYPE; i++)
1057 kfree(sbi->write_io[i]);
1061 int f2fs_sync_fs(struct super_block *sb, int sync)
1063 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1066 if (unlikely(f2fs_cp_error(sbi)))
1069 trace_f2fs_sync_fs(sb, sync);
1071 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1075 struct cp_control cpc;
1077 cpc.reason = __get_cp_reason(sbi);
1079 mutex_lock(&sbi->gc_mutex);
1080 err = f2fs_write_checkpoint(sbi, &cpc);
1081 mutex_unlock(&sbi->gc_mutex);
1083 f2fs_trace_ios(NULL, 1);
1088 static int f2fs_freeze(struct super_block *sb)
1090 if (f2fs_readonly(sb))
1093 /* IO error happened before */
1094 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1097 /* must be clean, since sync_filesystem() was already called */
1098 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1103 static int f2fs_unfreeze(struct super_block *sb)
1109 static int f2fs_statfs_project(struct super_block *sb,
1110 kprojid_t projid, struct kstatfs *buf)
1113 struct dquot *dquot;
1117 qid = make_kqid_projid(projid);
1118 dquot = dqget(sb, qid);
1120 return PTR_ERR(dquot);
1121 spin_lock(&dq_data_lock);
1123 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
1124 dquot->dq_dqb.dqb_bsoftlimit :
1125 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
1126 if (limit && buf->f_blocks > limit) {
1127 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1128 buf->f_blocks = limit;
1129 buf->f_bfree = buf->f_bavail =
1130 (buf->f_blocks > curblock) ?
1131 (buf->f_blocks - curblock) : 0;
1134 limit = dquot->dq_dqb.dqb_isoftlimit ?
1135 dquot->dq_dqb.dqb_isoftlimit :
1136 dquot->dq_dqb.dqb_ihardlimit;
1137 if (limit && buf->f_files > limit) {
1138 buf->f_files = limit;
1140 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1141 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1144 spin_unlock(&dq_data_lock);
1150 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1152 struct super_block *sb = dentry->d_sb;
1153 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1154 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1155 block_t total_count, user_block_count, start_count;
1156 u64 avail_node_count;
1158 total_count = le64_to_cpu(sbi->raw_super->block_count);
1159 user_block_count = sbi->user_block_count;
1160 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1161 buf->f_type = F2FS_SUPER_MAGIC;
1162 buf->f_bsize = sbi->blocksize;
1164 buf->f_blocks = total_count - start_count;
1165 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1166 sbi->current_reserved_blocks;
1167 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1168 buf->f_bavail = buf->f_bfree -
1169 F2FS_OPTION(sbi).root_reserved_blocks;
1173 avail_node_count = sbi->total_node_count - sbi->nquota_files -
1174 F2FS_RESERVED_NODE_NUM;
1176 if (avail_node_count > user_block_count) {
1177 buf->f_files = user_block_count;
1178 buf->f_ffree = buf->f_bavail;
1180 buf->f_files = avail_node_count;
1181 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1185 buf->f_namelen = F2FS_NAME_LEN;
1186 buf->f_fsid.val[0] = (u32)id;
1187 buf->f_fsid.val[1] = (u32)(id >> 32);
1190 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1191 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1192 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1198 static inline void f2fs_show_quota_options(struct seq_file *seq,
1199 struct super_block *sb)
1202 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1204 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1207 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1218 seq_printf(seq, ",jqfmt=%s", fmtname);
1221 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1222 seq_show_option(seq, "usrjquota",
1223 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1225 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1226 seq_show_option(seq, "grpjquota",
1227 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1229 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1230 seq_show_option(seq, "prjjquota",
1231 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1235 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1237 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1239 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1240 if (test_opt(sbi, FORCE_FG_GC))
1241 seq_printf(seq, ",background_gc=%s", "sync");
1243 seq_printf(seq, ",background_gc=%s", "on");
1245 seq_printf(seq, ",background_gc=%s", "off");
1247 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1248 seq_puts(seq, ",disable_roll_forward");
1249 if (test_opt(sbi, DISCARD))
1250 seq_puts(seq, ",discard");
1251 if (test_opt(sbi, NOHEAP))
1252 seq_puts(seq, ",no_heap");
1254 seq_puts(seq, ",heap");
1255 #ifdef CONFIG_F2FS_FS_XATTR
1256 if (test_opt(sbi, XATTR_USER))
1257 seq_puts(seq, ",user_xattr");
1259 seq_puts(seq, ",nouser_xattr");
1260 if (test_opt(sbi, INLINE_XATTR))
1261 seq_puts(seq, ",inline_xattr");
1263 seq_puts(seq, ",noinline_xattr");
1264 if (test_opt(sbi, INLINE_XATTR_SIZE))
1265 seq_printf(seq, ",inline_xattr_size=%u",
1266 F2FS_OPTION(sbi).inline_xattr_size);
1268 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1269 if (test_opt(sbi, POSIX_ACL))
1270 seq_puts(seq, ",acl");
1272 seq_puts(seq, ",noacl");
1274 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1275 seq_puts(seq, ",disable_ext_identify");
1276 if (test_opt(sbi, INLINE_DATA))
1277 seq_puts(seq, ",inline_data");
1279 seq_puts(seq, ",noinline_data");
1280 if (test_opt(sbi, INLINE_DENTRY))
1281 seq_puts(seq, ",inline_dentry");
1283 seq_puts(seq, ",noinline_dentry");
1284 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1285 seq_puts(seq, ",flush_merge");
1286 if (test_opt(sbi, NOBARRIER))
1287 seq_puts(seq, ",nobarrier");
1288 if (test_opt(sbi, FASTBOOT))
1289 seq_puts(seq, ",fastboot");
1290 if (test_opt(sbi, EXTENT_CACHE))
1291 seq_puts(seq, ",extent_cache");
1293 seq_puts(seq, ",noextent_cache");
1294 if (test_opt(sbi, DATA_FLUSH))
1295 seq_puts(seq, ",data_flush");
1297 seq_puts(seq, ",mode=");
1298 if (test_opt(sbi, ADAPTIVE))
1299 seq_puts(seq, "adaptive");
1300 else if (test_opt(sbi, LFS))
1301 seq_puts(seq, "lfs");
1302 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1303 if (test_opt(sbi, RESERVE_ROOT))
1304 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1305 F2FS_OPTION(sbi).root_reserved_blocks,
1306 from_kuid_munged(&init_user_ns,
1307 F2FS_OPTION(sbi).s_resuid),
1308 from_kgid_munged(&init_user_ns,
1309 F2FS_OPTION(sbi).s_resgid));
1310 if (F2FS_IO_SIZE_BITS(sbi))
1311 seq_printf(seq, ",io_size=%uKB", F2FS_IO_SIZE_KB(sbi));
1312 #ifdef CONFIG_F2FS_FAULT_INJECTION
1313 if (test_opt(sbi, FAULT_INJECTION))
1314 seq_printf(seq, ",fault_injection=%u",
1315 F2FS_OPTION(sbi).fault_info.inject_rate);
1318 if (test_opt(sbi, QUOTA))
1319 seq_puts(seq, ",quota");
1320 if (test_opt(sbi, USRQUOTA))
1321 seq_puts(seq, ",usrquota");
1322 if (test_opt(sbi, GRPQUOTA))
1323 seq_puts(seq, ",grpquota");
1324 if (test_opt(sbi, PRJQUOTA))
1325 seq_puts(seq, ",prjquota");
1327 f2fs_show_quota_options(seq, sbi->sb);
1328 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1329 seq_printf(seq, ",whint_mode=%s", "user-based");
1330 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1331 seq_printf(seq, ",whint_mode=%s", "fs-based");
1332 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1333 if (F2FS_OPTION(sbi).test_dummy_encryption)
1334 seq_puts(seq, ",test_dummy_encryption");
1337 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1338 seq_printf(seq, ",alloc_mode=%s", "default");
1339 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1340 seq_printf(seq, ",alloc_mode=%s", "reuse");
1342 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1343 seq_printf(seq, ",fsync_mode=%s", "posix");
1344 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1345 seq_printf(seq, ",fsync_mode=%s", "strict");
1349 static void default_options(struct f2fs_sb_info *sbi)
1351 /* init some FS parameters */
1352 F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1353 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1354 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1355 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1356 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1357 F2FS_OPTION(sbi).test_dummy_encryption = false;
1358 sbi->readdir_ra = 1;
1360 set_opt(sbi, BG_GC);
1361 set_opt(sbi, INLINE_XATTR);
1362 set_opt(sbi, INLINE_DATA);
1363 set_opt(sbi, INLINE_DENTRY);
1364 set_opt(sbi, EXTENT_CACHE);
1365 set_opt(sbi, NOHEAP);
1366 sbi->sb->s_flags |= SB_LAZYTIME;
1367 set_opt(sbi, FLUSH_MERGE);
1368 if (f2fs_sb_has_blkzoned(sbi->sb)) {
1369 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1370 set_opt(sbi, DISCARD);
1372 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1375 #ifdef CONFIG_F2FS_FS_XATTR
1376 set_opt(sbi, XATTR_USER);
1378 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1379 set_opt(sbi, POSIX_ACL);
1382 #ifdef CONFIG_F2FS_FAULT_INJECTION
1383 f2fs_build_fault_attr(sbi, 0);
1388 static int f2fs_enable_quotas(struct super_block *sb);
1390 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1392 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1393 struct f2fs_mount_info org_mount_opt;
1394 unsigned long old_sb_flags;
1396 bool need_restart_gc = false;
1397 bool need_stop_gc = false;
1398 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1404 * Save the old mount options in case we
1405 * need to restore them.
1407 org_mount_opt = sbi->mount_opt;
1408 old_sb_flags = sb->s_flags;
1411 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1412 for (i = 0; i < MAXQUOTAS; i++) {
1413 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1414 org_mount_opt.s_qf_names[i] =
1415 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1417 if (!org_mount_opt.s_qf_names[i]) {
1418 for (j = 0; j < i; j++)
1419 kfree(org_mount_opt.s_qf_names[j]);
1423 org_mount_opt.s_qf_names[i] = NULL;
1428 /* recover superblocks we couldn't write due to previous RO mount */
1429 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1430 err = f2fs_commit_super(sbi, false);
1431 f2fs_msg(sb, KERN_INFO,
1432 "Try to recover all the superblocks, ret: %d", err);
1434 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1437 default_options(sbi);
1439 /* parse mount options */
1440 err = parse_options(sb, data);
1445 * Previous and new state of filesystem is RO,
1446 * so skip checking GC and FLUSH_MERGE conditions.
1448 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1452 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1453 err = dquot_suspend(sb, -1);
1456 } else if (f2fs_readonly(sb) && !(*flags & MS_RDONLY)) {
1457 /* dquot_resume needs RW */
1458 sb->s_flags &= ~SB_RDONLY;
1459 if (sb_any_quota_suspended(sb)) {
1460 dquot_resume(sb, -1);
1461 } else if (f2fs_sb_has_quota_ino(sb)) {
1462 err = f2fs_enable_quotas(sb);
1468 /* disallow enable/disable extent_cache dynamically */
1469 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1471 f2fs_msg(sbi->sb, KERN_WARNING,
1472 "switch extent_cache option is not allowed");
1477 * We stop the GC thread if FS is mounted as RO
1478 * or if background_gc = off is passed in mount
1479 * option. Also sync the filesystem.
1481 if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) {
1482 if (sbi->gc_thread) {
1483 f2fs_stop_gc_thread(sbi);
1484 need_restart_gc = true;
1486 } else if (!sbi->gc_thread) {
1487 err = f2fs_start_gc_thread(sbi);
1490 need_stop_gc = true;
1493 if (*flags & SB_RDONLY ||
1494 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1495 writeback_inodes_sb(sb, WB_REASON_SYNC);
1498 set_sbi_flag(sbi, SBI_IS_DIRTY);
1499 set_sbi_flag(sbi, SBI_IS_CLOSE);
1500 f2fs_sync_fs(sb, 1);
1501 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1505 * We stop issue flush thread if FS is mounted as RO
1506 * or if flush_merge is not passed in mount option.
1508 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1509 clear_opt(sbi, FLUSH_MERGE);
1510 f2fs_destroy_flush_cmd_control(sbi, false);
1512 err = f2fs_create_flush_cmd_control(sbi);
1518 /* Release old quota file names */
1519 for (i = 0; i < MAXQUOTAS; i++)
1520 kfree(org_mount_opt.s_qf_names[i]);
1522 /* Update the POSIXACL Flag */
1523 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1524 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1526 limit_reserve_root(sbi);
1529 if (need_restart_gc) {
1530 if (f2fs_start_gc_thread(sbi))
1531 f2fs_msg(sbi->sb, KERN_WARNING,
1532 "background gc thread has stopped");
1533 } else if (need_stop_gc) {
1534 f2fs_stop_gc_thread(sbi);
1538 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1539 for (i = 0; i < MAXQUOTAS; i++) {
1540 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1541 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1544 sbi->mount_opt = org_mount_opt;
1545 sb->s_flags = old_sb_flags;
1550 /* Read data from quotafile */
1551 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1552 size_t len, loff_t off)
1554 struct inode *inode = sb_dqopt(sb)->files[type];
1555 struct address_space *mapping = inode->i_mapping;
1556 block_t blkidx = F2FS_BYTES_TO_BLK(off);
1557 int offset = off & (sb->s_blocksize - 1);
1560 loff_t i_size = i_size_read(inode);
1567 if (off + len > i_size)
1570 while (toread > 0) {
1571 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1573 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1575 if (PTR_ERR(page) == -ENOMEM) {
1576 congestion_wait(BLK_RW_ASYNC, HZ/50);
1579 return PTR_ERR(page);
1584 if (unlikely(page->mapping != mapping)) {
1585 f2fs_put_page(page, 1);
1588 if (unlikely(!PageUptodate(page))) {
1589 f2fs_put_page(page, 1);
1593 kaddr = kmap_atomic(page);
1594 memcpy(data, kaddr + offset, tocopy);
1595 kunmap_atomic(kaddr);
1596 f2fs_put_page(page, 1);
1606 /* Write to quotafile */
1607 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1608 const char *data, size_t len, loff_t off)
1610 struct inode *inode = sb_dqopt(sb)->files[type];
1611 struct address_space *mapping = inode->i_mapping;
1612 const struct address_space_operations *a_ops = mapping->a_ops;
1613 int offset = off & (sb->s_blocksize - 1);
1614 size_t towrite = len;
1620 while (towrite > 0) {
1621 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1624 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1626 if (unlikely(err)) {
1627 if (err == -ENOMEM) {
1628 congestion_wait(BLK_RW_ASYNC, HZ/50);
1634 kaddr = kmap_atomic(page);
1635 memcpy(kaddr + offset, data, tocopy);
1636 kunmap_atomic(kaddr);
1637 flush_dcache_page(page);
1639 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1650 inode->i_mtime = inode->i_ctime = current_time(inode);
1651 f2fs_mark_inode_dirty_sync(inode, false);
1652 return len - towrite;
1655 static struct dquot **f2fs_get_dquots(struct inode *inode)
1657 return F2FS_I(inode)->i_dquot;
1660 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1662 return &F2FS_I(inode)->i_reserved_quota;
1665 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1667 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
1668 F2FS_OPTION(sbi).s_jquota_fmt, type);
1671 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1676 if (f2fs_sb_has_quota_ino(sbi->sb) && rdonly) {
1677 err = f2fs_enable_quotas(sbi->sb);
1679 f2fs_msg(sbi->sb, KERN_ERR,
1680 "Cannot turn on quota_ino: %d", err);
1686 for (i = 0; i < MAXQUOTAS; i++) {
1687 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1688 err = f2fs_quota_on_mount(sbi, i);
1693 f2fs_msg(sbi->sb, KERN_ERR,
1694 "Cannot turn on quotas: %d on %d", err, i);
1700 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
1703 struct inode *qf_inode;
1704 unsigned long qf_inum;
1707 BUG_ON(!f2fs_sb_has_quota_ino(sb));
1709 qf_inum = f2fs_qf_ino(sb, type);
1713 qf_inode = f2fs_iget(sb, qf_inum);
1714 if (IS_ERR(qf_inode)) {
1715 f2fs_msg(sb, KERN_ERR,
1716 "Bad quota inode %u:%lu", type, qf_inum);
1717 return PTR_ERR(qf_inode);
1720 /* Don't account quota for quota files to avoid recursion */
1721 qf_inode->i_flags |= S_NOQUOTA;
1722 err = dquot_enable(qf_inode, type, format_id, flags);
1727 static int f2fs_enable_quotas(struct super_block *sb)
1730 unsigned long qf_inum;
1731 bool quota_mopt[MAXQUOTAS] = {
1732 test_opt(F2FS_SB(sb), USRQUOTA),
1733 test_opt(F2FS_SB(sb), GRPQUOTA),
1734 test_opt(F2FS_SB(sb), PRJQUOTA),
1737 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
1738 for (type = 0; type < MAXQUOTAS; type++) {
1739 qf_inum = f2fs_qf_ino(sb, type);
1741 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
1742 DQUOT_USAGE_ENABLED |
1743 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
1745 f2fs_msg(sb, KERN_ERR,
1746 "Failed to enable quota tracking "
1747 "(type=%d, err=%d). Please run "
1748 "fsck to fix.", type, err);
1749 for (type--; type >= 0; type--)
1750 dquot_quota_off(sb, type);
1758 static int f2fs_quota_sync(struct super_block *sb, int type)
1760 struct quota_info *dqopt = sb_dqopt(sb);
1764 ret = dquot_writeback_dquots(sb, type);
1769 * Now when everything is written we can discard the pagecache so
1770 * that userspace sees the changes.
1772 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
1773 if (type != -1 && cnt != type)
1775 if (!sb_has_quota_active(sb, cnt))
1778 ret = filemap_write_and_wait(dqopt->files[cnt]->i_mapping);
1782 inode_lock(dqopt->files[cnt]);
1783 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
1784 inode_unlock(dqopt->files[cnt]);
1789 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
1790 const struct path *path)
1792 struct inode *inode;
1795 err = f2fs_quota_sync(sb, type);
1799 err = dquot_quota_on(sb, type, format_id, path);
1803 inode = d_inode(path->dentry);
1806 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
1807 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1808 S_NOATIME | S_IMMUTABLE);
1809 inode_unlock(inode);
1810 f2fs_mark_inode_dirty_sync(inode, false);
1815 static int f2fs_quota_off(struct super_block *sb, int type)
1817 struct inode *inode = sb_dqopt(sb)->files[type];
1820 if (!inode || !igrab(inode))
1821 return dquot_quota_off(sb, type);
1823 f2fs_quota_sync(sb, type);
1825 err = dquot_quota_off(sb, type);
1826 if (err || f2fs_sb_has_quota_ino(sb))
1830 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
1831 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1832 inode_unlock(inode);
1833 f2fs_mark_inode_dirty_sync(inode, false);
1839 void f2fs_quota_off_umount(struct super_block *sb)
1843 for (type = 0; type < MAXQUOTAS; type++)
1844 f2fs_quota_off(sb, type);
1847 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
1849 *projid = F2FS_I(inode)->i_projid;
1853 static const struct dquot_operations f2fs_quota_operations = {
1854 .get_reserved_space = f2fs_get_reserved_space,
1855 .write_dquot = dquot_commit,
1856 .acquire_dquot = dquot_acquire,
1857 .release_dquot = dquot_release,
1858 .mark_dirty = dquot_mark_dquot_dirty,
1859 .write_info = dquot_commit_info,
1860 .alloc_dquot = dquot_alloc,
1861 .destroy_dquot = dquot_destroy,
1862 .get_projid = f2fs_get_projid,
1863 .get_next_id = dquot_get_next_id,
1866 static const struct quotactl_ops f2fs_quotactl_ops = {
1867 .quota_on = f2fs_quota_on,
1868 .quota_off = f2fs_quota_off,
1869 .quota_sync = f2fs_quota_sync,
1870 .get_state = dquot_get_state,
1871 .set_info = dquot_set_dqinfo,
1872 .get_dqblk = dquot_get_dqblk,
1873 .set_dqblk = dquot_set_dqblk,
1874 .get_nextdqblk = dquot_get_next_dqblk,
1877 void f2fs_quota_off_umount(struct super_block *sb)
1882 static const struct super_operations f2fs_sops = {
1883 .alloc_inode = f2fs_alloc_inode,
1884 .drop_inode = f2fs_drop_inode,
1885 .destroy_inode = f2fs_destroy_inode,
1886 .write_inode = f2fs_write_inode,
1887 .dirty_inode = f2fs_dirty_inode,
1888 .show_options = f2fs_show_options,
1890 .quota_read = f2fs_quota_read,
1891 .quota_write = f2fs_quota_write,
1892 .get_dquots = f2fs_get_dquots,
1894 .evict_inode = f2fs_evict_inode,
1895 .put_super = f2fs_put_super,
1896 .sync_fs = f2fs_sync_fs,
1897 .freeze_fs = f2fs_freeze,
1898 .unfreeze_fs = f2fs_unfreeze,
1899 .statfs = f2fs_statfs,
1900 .remount_fs = f2fs_remount,
1903 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1904 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1906 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1907 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1911 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1914 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1917 * Encrypting the root directory is not allowed because fsck
1918 * expects lost+found directory to exist and remain unencrypted
1919 * if LOST_FOUND feature is enabled.
1922 if (f2fs_sb_has_lost_found(sbi->sb) &&
1923 inode->i_ino == F2FS_ROOT_INO(sbi))
1926 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1927 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1928 ctx, len, fs_data, XATTR_CREATE);
1931 static bool f2fs_dummy_context(struct inode *inode)
1933 return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
1936 static const struct fscrypt_operations f2fs_cryptops = {
1937 .key_prefix = "f2fs:",
1938 .get_context = f2fs_get_context,
1939 .set_context = f2fs_set_context,
1940 .dummy_context = f2fs_dummy_context,
1941 .empty_dir = f2fs_empty_dir,
1942 .max_namelen = F2FS_NAME_LEN,
1946 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1947 u64 ino, u32 generation)
1949 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1950 struct inode *inode;
1952 if (f2fs_check_nid_range(sbi, ino))
1953 return ERR_PTR(-ESTALE);
1956 * f2fs_iget isn't quite right if the inode is currently unallocated!
1957 * However f2fs_iget currently does appropriate checks to handle stale
1958 * inodes so everything is OK.
1960 inode = f2fs_iget(sb, ino);
1962 return ERR_CAST(inode);
1963 if (unlikely(generation && inode->i_generation != generation)) {
1964 /* we didn't find the right inode.. */
1966 return ERR_PTR(-ESTALE);
1971 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1972 int fh_len, int fh_type)
1974 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1975 f2fs_nfs_get_inode);
1978 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1979 int fh_len, int fh_type)
1981 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1982 f2fs_nfs_get_inode);
1985 static const struct export_operations f2fs_export_ops = {
1986 .fh_to_dentry = f2fs_fh_to_dentry,
1987 .fh_to_parent = f2fs_fh_to_parent,
1988 .get_parent = f2fs_get_parent,
1991 static loff_t max_file_blocks(void)
1994 loff_t leaf_count = ADDRS_PER_BLOCK;
1997 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
1998 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
1999 * space in inode.i_addr, it will be more safe to reassign
2003 /* two direct node blocks */
2004 result += (leaf_count * 2);
2006 /* two indirect node blocks */
2007 leaf_count *= NIDS_PER_BLOCK;
2008 result += (leaf_count * 2);
2010 /* one double indirect node block */
2011 leaf_count *= NIDS_PER_BLOCK;
2012 result += leaf_count;
2017 static int __f2fs_commit_super(struct buffer_head *bh,
2018 struct f2fs_super_block *super)
2022 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2023 set_buffer_dirty(bh);
2026 /* it's rare case, we can do fua all the time */
2027 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2030 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2031 struct buffer_head *bh)
2033 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2034 (bh->b_data + F2FS_SUPER_OFFSET);
2035 struct super_block *sb = sbi->sb;
2036 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2037 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2038 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2039 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2040 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2041 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2042 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2043 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2044 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2045 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2046 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2047 u32 segment_count = le32_to_cpu(raw_super->segment_count);
2048 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2049 u64 main_end_blkaddr = main_blkaddr +
2050 (segment_count_main << log_blocks_per_seg);
2051 u64 seg_end_blkaddr = segment0_blkaddr +
2052 (segment_count << log_blocks_per_seg);
2054 if (segment0_blkaddr != cp_blkaddr) {
2055 f2fs_msg(sb, KERN_INFO,
2056 "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2057 segment0_blkaddr, cp_blkaddr);
2061 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2063 f2fs_msg(sb, KERN_INFO,
2064 "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2065 cp_blkaddr, sit_blkaddr,
2066 segment_count_ckpt << log_blocks_per_seg);
2070 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2072 f2fs_msg(sb, KERN_INFO,
2073 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2074 sit_blkaddr, nat_blkaddr,
2075 segment_count_sit << log_blocks_per_seg);
2079 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2081 f2fs_msg(sb, KERN_INFO,
2082 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2083 nat_blkaddr, ssa_blkaddr,
2084 segment_count_nat << log_blocks_per_seg);
2088 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2090 f2fs_msg(sb, KERN_INFO,
2091 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2092 ssa_blkaddr, main_blkaddr,
2093 segment_count_ssa << log_blocks_per_seg);
2097 if (main_end_blkaddr > seg_end_blkaddr) {
2098 f2fs_msg(sb, KERN_INFO,
2099 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2102 (segment_count << log_blocks_per_seg),
2103 segment_count_main << log_blocks_per_seg);
2105 } else if (main_end_blkaddr < seg_end_blkaddr) {
2109 /* fix in-memory information all the time */
2110 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2111 segment0_blkaddr) >> log_blocks_per_seg);
2113 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2114 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2117 err = __f2fs_commit_super(bh, NULL);
2118 res = err ? "failed" : "done";
2120 f2fs_msg(sb, KERN_INFO,
2121 "Fix alignment : %s, start(%u) end(%u) block(%u)",
2124 (segment_count << log_blocks_per_seg),
2125 segment_count_main << log_blocks_per_seg);
2132 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2133 struct buffer_head *bh)
2135 block_t segment_count, segs_per_sec, secs_per_zone;
2136 block_t total_sections, blocks_per_seg;
2137 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2138 (bh->b_data + F2FS_SUPER_OFFSET);
2139 struct super_block *sb = sbi->sb;
2140 unsigned int blocksize;
2142 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
2143 f2fs_msg(sb, KERN_INFO,
2144 "Magic Mismatch, valid(0x%x) - read(0x%x)",
2145 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2149 /* Currently, support only 4KB page cache size */
2150 if (F2FS_BLKSIZE != PAGE_SIZE) {
2151 f2fs_msg(sb, KERN_INFO,
2152 "Invalid page_cache_size (%lu), supports only 4KB\n",
2157 /* Currently, support only 4KB block size */
2158 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2159 if (blocksize != F2FS_BLKSIZE) {
2160 f2fs_msg(sb, KERN_INFO,
2161 "Invalid blocksize (%u), supports only 4KB\n",
2166 /* check log blocks per segment */
2167 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2168 f2fs_msg(sb, KERN_INFO,
2169 "Invalid log blocks per segment (%u)\n",
2170 le32_to_cpu(raw_super->log_blocks_per_seg));
2174 /* Currently, support 512/1024/2048/4096 bytes sector size */
2175 if (le32_to_cpu(raw_super->log_sectorsize) >
2176 F2FS_MAX_LOG_SECTOR_SIZE ||
2177 le32_to_cpu(raw_super->log_sectorsize) <
2178 F2FS_MIN_LOG_SECTOR_SIZE) {
2179 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
2180 le32_to_cpu(raw_super->log_sectorsize));
2183 if (le32_to_cpu(raw_super->log_sectors_per_block) +
2184 le32_to_cpu(raw_super->log_sectorsize) !=
2185 F2FS_MAX_LOG_SECTOR_SIZE) {
2186 f2fs_msg(sb, KERN_INFO,
2187 "Invalid log sectors per block(%u) log sectorsize(%u)",
2188 le32_to_cpu(raw_super->log_sectors_per_block),
2189 le32_to_cpu(raw_super->log_sectorsize));
2193 segment_count = le32_to_cpu(raw_super->segment_count);
2194 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2195 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2196 total_sections = le32_to_cpu(raw_super->section_count);
2198 /* blocks_per_seg should be 512, given the above check */
2199 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2201 if (segment_count > F2FS_MAX_SEGMENT ||
2202 segment_count < F2FS_MIN_SEGMENTS) {
2203 f2fs_msg(sb, KERN_INFO,
2204 "Invalid segment count (%u)",
2209 if (total_sections > segment_count ||
2210 total_sections < F2FS_MIN_SEGMENTS ||
2211 segs_per_sec > segment_count || !segs_per_sec) {
2212 f2fs_msg(sb, KERN_INFO,
2213 "Invalid segment/section count (%u, %u x %u)",
2214 segment_count, total_sections, segs_per_sec);
2218 if ((segment_count / segs_per_sec) < total_sections) {
2219 f2fs_msg(sb, KERN_INFO,
2220 "Small segment_count (%u < %u * %u)",
2221 segment_count, segs_per_sec, total_sections);
2225 if (segment_count > (le32_to_cpu(raw_super->block_count) >> 9)) {
2226 f2fs_msg(sb, KERN_INFO,
2227 "Wrong segment_count / block_count (%u > %u)",
2228 segment_count, le32_to_cpu(raw_super->block_count));
2232 if (secs_per_zone > total_sections) {
2233 f2fs_msg(sb, KERN_INFO,
2234 "Wrong secs_per_zone (%u > %u)",
2235 secs_per_zone, total_sections);
2238 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2239 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2240 (le32_to_cpu(raw_super->extension_count) +
2241 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2242 f2fs_msg(sb, KERN_INFO,
2243 "Corrupted extension count (%u + %u > %u)",
2244 le32_to_cpu(raw_super->extension_count),
2245 raw_super->hot_ext_count,
2246 F2FS_MAX_EXTENSION);
2250 if (le32_to_cpu(raw_super->cp_payload) >
2251 (blocks_per_seg - F2FS_CP_PACKS)) {
2252 f2fs_msg(sb, KERN_INFO,
2253 "Insane cp_payload (%u > %u)",
2254 le32_to_cpu(raw_super->cp_payload),
2255 blocks_per_seg - F2FS_CP_PACKS);
2259 /* check reserved ino info */
2260 if (le32_to_cpu(raw_super->node_ino) != 1 ||
2261 le32_to_cpu(raw_super->meta_ino) != 2 ||
2262 le32_to_cpu(raw_super->root_ino) != 3) {
2263 f2fs_msg(sb, KERN_INFO,
2264 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2265 le32_to_cpu(raw_super->node_ino),
2266 le32_to_cpu(raw_super->meta_ino),
2267 le32_to_cpu(raw_super->root_ino));
2271 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2272 if (sanity_check_area_boundary(sbi, bh))
2278 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2280 unsigned int total, fsmeta;
2281 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2282 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2283 unsigned int ovp_segments, reserved_segments;
2284 unsigned int main_segs, blocks_per_seg;
2287 total = le32_to_cpu(raw_super->segment_count);
2288 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2289 fsmeta += le32_to_cpu(raw_super->segment_count_sit);
2290 fsmeta += le32_to_cpu(raw_super->segment_count_nat);
2291 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2292 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2294 if (unlikely(fsmeta >= total))
2297 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2298 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2300 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2301 ovp_segments == 0 || reserved_segments == 0)) {
2302 f2fs_msg(sbi->sb, KERN_ERR,
2303 "Wrong layout: check mkfs.f2fs version");
2307 main_segs = le32_to_cpu(raw_super->segment_count_main);
2308 blocks_per_seg = sbi->blocks_per_seg;
2310 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2311 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2312 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2315 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2316 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2317 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2321 if (unlikely(f2fs_cp_error(sbi))) {
2322 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
2328 static void init_sb_info(struct f2fs_sb_info *sbi)
2330 struct f2fs_super_block *raw_super = sbi->raw_super;
2333 sbi->log_sectors_per_block =
2334 le32_to_cpu(raw_super->log_sectors_per_block);
2335 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2336 sbi->blocksize = 1 << sbi->log_blocksize;
2337 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2338 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2339 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2340 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2341 sbi->total_sections = le32_to_cpu(raw_super->section_count);
2342 sbi->total_node_count =
2343 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2344 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2345 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2346 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2347 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2348 sbi->cur_victim_sec = NULL_SECNO;
2349 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2351 sbi->dir_level = DEF_DIR_LEVEL;
2352 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2353 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2354 clear_sbi_flag(sbi, SBI_NEED_FSCK);
2356 for (i = 0; i < NR_COUNT_TYPE; i++)
2357 atomic_set(&sbi->nr_pages[i], 0);
2359 for (i = 0; i < META; i++)
2360 atomic_set(&sbi->wb_sync_req[i], 0);
2362 INIT_LIST_HEAD(&sbi->s_list);
2363 mutex_init(&sbi->umount_mutex);
2364 for (i = 0; i < NR_PAGE_TYPE - 1; i++)
2365 for (j = HOT; j < NR_TEMP_TYPE; j++)
2366 mutex_init(&sbi->wio_mutex[i][j]);
2367 init_rwsem(&sbi->io_order_lock);
2368 spin_lock_init(&sbi->cp_lock);
2370 sbi->dirty_device = 0;
2371 spin_lock_init(&sbi->dev_lock);
2373 init_rwsem(&sbi->sb_lock);
2376 static int init_percpu_info(struct f2fs_sb_info *sbi)
2380 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2384 return percpu_counter_init(&sbi->total_valid_inode_count, 0,
2388 #ifdef CONFIG_BLK_DEV_ZONED
2389 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2391 struct block_device *bdev = FDEV(devi).bdev;
2392 sector_t nr_sectors = bdev->bd_part->nr_sects;
2393 sector_t sector = 0;
2394 struct blk_zone *zones;
2395 unsigned int i, nr_zones;
2399 if (!f2fs_sb_has_blkzoned(sbi->sb))
2402 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2403 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2405 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2406 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2407 __ilog2_u32(sbi->blocks_per_blkz))
2409 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2410 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2411 sbi->log_blocks_per_blkz;
2412 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2413 FDEV(devi).nr_blkz++;
2415 FDEV(devi).blkz_type = f2fs_kmalloc(sbi, FDEV(devi).nr_blkz,
2417 if (!FDEV(devi).blkz_type)
2420 #define F2FS_REPORT_NR_ZONES 4096
2422 zones = f2fs_kzalloc(sbi,
2423 array_size(F2FS_REPORT_NR_ZONES,
2424 sizeof(struct blk_zone)),
2429 /* Get block zones type */
2430 while (zones && sector < nr_sectors) {
2432 nr_zones = F2FS_REPORT_NR_ZONES;
2433 err = blkdev_report_zones(bdev, sector,
2443 for (i = 0; i < nr_zones; i++) {
2444 FDEV(devi).blkz_type[n] = zones[i].type;
2445 sector += zones[i].len;
2457 * Read f2fs raw super block.
2458 * Because we have two copies of super block, so read both of them
2459 * to get the first valid one. If any one of them is broken, we pass
2460 * them recovery flag back to the caller.
2462 static int read_raw_super_block(struct f2fs_sb_info *sbi,
2463 struct f2fs_super_block **raw_super,
2464 int *valid_super_block, int *recovery)
2466 struct super_block *sb = sbi->sb;
2468 struct buffer_head *bh;
2469 struct f2fs_super_block *super;
2472 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2476 for (block = 0; block < 2; block++) {
2477 bh = sb_bread(sb, block);
2479 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
2485 /* sanity checking of raw super */
2486 if (sanity_check_raw_super(sbi, bh)) {
2487 f2fs_msg(sb, KERN_ERR,
2488 "Can't find valid F2FS filesystem in %dth superblock",
2496 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2498 *valid_super_block = block;
2504 /* Fail to read any one of the superblocks*/
2508 /* No valid superblock */
2517 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
2519 struct buffer_head *bh;
2522 if ((recover && f2fs_readonly(sbi->sb)) ||
2523 bdev_read_only(sbi->sb->s_bdev)) {
2524 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2528 /* write back-up superblock first */
2529 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
2532 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2535 /* if we are in recovery path, skip writing valid superblock */
2539 /* write current valid superblock */
2540 bh = sb_bread(sbi->sb, sbi->valid_super_block);
2543 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2548 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
2550 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2551 unsigned int max_devices = MAX_DEVICES;
2554 /* Initialize single device information */
2555 if (!RDEV(0).path[0]) {
2556 if (!bdev_is_zoned(sbi->sb->s_bdev))
2562 * Initialize multiple devices information, or single
2563 * zoned block device information.
2565 sbi->devs = f2fs_kzalloc(sbi,
2566 array_size(max_devices,
2567 sizeof(struct f2fs_dev_info)),
2572 for (i = 0; i < max_devices; i++) {
2574 if (i > 0 && !RDEV(i).path[0])
2577 if (max_devices == 1) {
2578 /* Single zoned block device mount */
2580 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
2581 sbi->sb->s_mode, sbi->sb->s_type);
2583 /* Multi-device mount */
2584 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
2585 FDEV(i).total_segments =
2586 le32_to_cpu(RDEV(i).total_segments);
2588 FDEV(i).start_blk = 0;
2589 FDEV(i).end_blk = FDEV(i).start_blk +
2590 (FDEV(i).total_segments <<
2591 sbi->log_blocks_per_seg) - 1 +
2592 le32_to_cpu(raw_super->segment0_blkaddr);
2594 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
2595 FDEV(i).end_blk = FDEV(i).start_blk +
2596 (FDEV(i).total_segments <<
2597 sbi->log_blocks_per_seg) - 1;
2599 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
2600 sbi->sb->s_mode, sbi->sb->s_type);
2602 if (IS_ERR(FDEV(i).bdev))
2603 return PTR_ERR(FDEV(i).bdev);
2605 /* to release errored devices */
2606 sbi->s_ndevs = i + 1;
2608 #ifdef CONFIG_BLK_DEV_ZONED
2609 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
2610 !f2fs_sb_has_blkzoned(sbi->sb)) {
2611 f2fs_msg(sbi->sb, KERN_ERR,
2612 "Zoned block device feature not enabled\n");
2615 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
2616 if (init_blkz_info(sbi, i)) {
2617 f2fs_msg(sbi->sb, KERN_ERR,
2618 "Failed to initialize F2FS blkzone information");
2621 if (max_devices == 1)
2623 f2fs_msg(sbi->sb, KERN_INFO,
2624 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
2626 FDEV(i).total_segments,
2627 FDEV(i).start_blk, FDEV(i).end_blk,
2628 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
2629 "Host-aware" : "Host-managed");
2633 f2fs_msg(sbi->sb, KERN_INFO,
2634 "Mount Device [%2d]: %20s, %8u, %8x - %8x",
2636 FDEV(i).total_segments,
2637 FDEV(i).start_blk, FDEV(i).end_blk);
2639 f2fs_msg(sbi->sb, KERN_INFO,
2640 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
2644 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
2646 struct f2fs_sm_info *sm_i = SM_I(sbi);
2648 /* adjust parameters according to the volume size */
2649 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
2650 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
2651 sm_i->dcc_info->discard_granularity = 1;
2652 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
2656 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
2658 struct f2fs_sb_info *sbi;
2659 struct f2fs_super_block *raw_super;
2662 bool retry = true, need_fsck = false;
2663 char *options = NULL;
2664 int recovery, i, valid_super_block;
2665 struct curseg_info *seg_i;
2670 valid_super_block = -1;
2673 /* allocate memory for f2fs-specific super block info */
2674 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
2680 /* Load the checksum driver */
2681 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
2682 if (IS_ERR(sbi->s_chksum_driver)) {
2683 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
2684 err = PTR_ERR(sbi->s_chksum_driver);
2685 sbi->s_chksum_driver = NULL;
2689 /* set a block size */
2690 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
2691 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
2695 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
2700 sb->s_fs_info = sbi;
2701 sbi->raw_super = raw_super;
2703 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
2704 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
2706 /* precompute checksum seed for metadata */
2707 if (f2fs_sb_has_inode_chksum(sb))
2708 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
2709 sizeof(raw_super->uuid));
2712 * The BLKZONED feature indicates that the drive was formatted with
2713 * zone alignment optimization. This is optional for host-aware
2714 * devices, but mandatory for host-managed zoned block devices.
2716 #ifndef CONFIG_BLK_DEV_ZONED
2717 if (f2fs_sb_has_blkzoned(sb)) {
2718 f2fs_msg(sb, KERN_ERR,
2719 "Zoned block device support is not enabled\n");
2724 default_options(sbi);
2725 /* parse mount options */
2726 options = kstrdup((const char *)data, GFP_KERNEL);
2727 if (data && !options) {
2732 err = parse_options(sb, options);
2736 sbi->max_file_blocks = max_file_blocks();
2737 sb->s_maxbytes = sbi->max_file_blocks <<
2738 le32_to_cpu(raw_super->log_blocksize);
2739 sb->s_max_links = F2FS_LINK_MAX;
2740 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2743 sb->dq_op = &f2fs_quota_operations;
2744 if (f2fs_sb_has_quota_ino(sb))
2745 sb->s_qcop = &dquot_quotactl_sysfile_ops;
2747 sb->s_qcop = &f2fs_quotactl_ops;
2748 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
2750 if (f2fs_sb_has_quota_ino(sbi->sb)) {
2751 for (i = 0; i < MAXQUOTAS; i++) {
2752 if (f2fs_qf_ino(sbi->sb, i))
2753 sbi->nquota_files++;
2758 sb->s_op = &f2fs_sops;
2759 #ifdef CONFIG_F2FS_FS_ENCRYPTION
2760 sb->s_cop = &f2fs_cryptops;
2762 sb->s_xattr = f2fs_xattr_handlers;
2763 sb->s_export_op = &f2fs_export_ops;
2764 sb->s_magic = F2FS_SUPER_MAGIC;
2765 sb->s_time_gran = 1;
2766 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2767 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2768 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
2769 sb->s_iflags |= SB_I_CGROUPWB;
2771 /* init f2fs-specific super block info */
2772 sbi->valid_super_block = valid_super_block;
2773 mutex_init(&sbi->gc_mutex);
2774 mutex_init(&sbi->cp_mutex);
2775 init_rwsem(&sbi->node_write);
2776 init_rwsem(&sbi->node_change);
2778 /* disallow all the data/node/meta page writes */
2779 set_sbi_flag(sbi, SBI_POR_DOING);
2780 spin_lock_init(&sbi->stat_lock);
2782 /* init iostat info */
2783 spin_lock_init(&sbi->iostat_lock);
2784 sbi->iostat_enable = false;
2786 for (i = 0; i < NR_PAGE_TYPE; i++) {
2787 int n = (i == META) ? 1: NR_TEMP_TYPE;
2793 sizeof(struct f2fs_bio_info)),
2795 if (!sbi->write_io[i]) {
2800 for (j = HOT; j < n; j++) {
2801 init_rwsem(&sbi->write_io[i][j].io_rwsem);
2802 sbi->write_io[i][j].sbi = sbi;
2803 sbi->write_io[i][j].bio = NULL;
2804 spin_lock_init(&sbi->write_io[i][j].io_lock);
2805 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
2809 init_rwsem(&sbi->cp_rwsem);
2810 init_waitqueue_head(&sbi->cp_wait);
2813 err = init_percpu_info(sbi);
2817 if (F2FS_IO_SIZE(sbi) > 1) {
2818 sbi->write_io_dummy =
2819 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
2820 if (!sbi->write_io_dummy) {
2826 /* get an inode for meta space */
2827 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
2828 if (IS_ERR(sbi->meta_inode)) {
2829 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
2830 err = PTR_ERR(sbi->meta_inode);
2834 err = f2fs_get_valid_checkpoint(sbi);
2836 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
2837 goto free_meta_inode;
2840 /* Initialize device list */
2841 err = f2fs_scan_devices(sbi);
2843 f2fs_msg(sb, KERN_ERR, "Failed to find devices");
2847 sbi->total_valid_node_count =
2848 le32_to_cpu(sbi->ckpt->valid_node_count);
2849 percpu_counter_set(&sbi->total_valid_inode_count,
2850 le32_to_cpu(sbi->ckpt->valid_inode_count));
2851 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
2852 sbi->total_valid_block_count =
2853 le64_to_cpu(sbi->ckpt->valid_block_count);
2854 sbi->last_valid_block_count = sbi->total_valid_block_count;
2855 sbi->reserved_blocks = 0;
2856 sbi->current_reserved_blocks = 0;
2857 limit_reserve_root(sbi);
2859 for (i = 0; i < NR_INODE_TYPE; i++) {
2860 INIT_LIST_HEAD(&sbi->inode_list[i]);
2861 spin_lock_init(&sbi->inode_lock[i]);
2864 f2fs_init_extent_cache_info(sbi);
2866 f2fs_init_ino_entry_info(sbi);
2868 /* setup f2fs internal modules */
2869 err = f2fs_build_segment_manager(sbi);
2871 f2fs_msg(sb, KERN_ERR,
2872 "Failed to initialize F2FS segment manager");
2875 err = f2fs_build_node_manager(sbi);
2877 f2fs_msg(sb, KERN_ERR,
2878 "Failed to initialize F2FS node manager");
2882 /* For write statistics */
2883 if (sb->s_bdev->bd_part)
2884 sbi->sectors_written_start =
2885 (u64)part_stat_read(sb->s_bdev->bd_part,
2886 sectors[STAT_WRITE]);
2888 /* Read accumulated write IO statistics if exists */
2889 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
2890 if (__exist_node_summaries(sbi))
2891 sbi->kbytes_written =
2892 le64_to_cpu(seg_i->journal->info.kbytes_written);
2894 f2fs_build_gc_manager(sbi);
2896 /* get an inode for node space */
2897 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
2898 if (IS_ERR(sbi->node_inode)) {
2899 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
2900 err = PTR_ERR(sbi->node_inode);
2904 err = f2fs_build_stats(sbi);
2906 goto free_node_inode;
2908 /* read root inode and dentry */
2909 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
2911 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
2912 err = PTR_ERR(root);
2915 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2918 goto free_node_inode;
2921 sb->s_root = d_make_root(root); /* allocate root dentry */
2924 goto free_root_inode;
2927 err = f2fs_register_sysfs(sbi);
2929 goto free_root_inode;
2933 * Turn on quotas which were not enabled for read-only mounts if
2934 * filesystem has quota feature, so that they are updated correctly.
2936 if (f2fs_sb_has_quota_ino(sb) && !f2fs_readonly(sb)) {
2937 err = f2fs_enable_quotas(sb);
2939 f2fs_msg(sb, KERN_ERR,
2940 "Cannot turn on quotas: error %d", err);
2945 /* if there are nt orphan nodes free them */
2946 err = f2fs_recover_orphan_inodes(sbi);
2950 /* recover fsynced data */
2951 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
2953 * mount should be failed, when device has readonly mode, and
2954 * previous checkpoint was not done by clean system shutdown.
2956 if (bdev_read_only(sb->s_bdev) &&
2957 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
2963 set_sbi_flag(sbi, SBI_NEED_FSCK);
2968 err = f2fs_recover_fsync_data(sbi, false);
2971 f2fs_msg(sb, KERN_ERR,
2972 "Cannot recover all fsync data errno=%d", err);
2976 err = f2fs_recover_fsync_data(sbi, true);
2978 if (!f2fs_readonly(sb) && err > 0) {
2980 f2fs_msg(sb, KERN_ERR,
2981 "Need to recover fsync data");
2986 /* f2fs_recover_fsync_data() cleared this already */
2987 clear_sbi_flag(sbi, SBI_POR_DOING);
2990 * If filesystem is not mounted as read-only then
2991 * do start the gc_thread.
2993 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
2994 /* After POR, we can run background GC thread.*/
2995 err = f2fs_start_gc_thread(sbi);
3001 /* recover broken superblock */
3003 err = f2fs_commit_super(sbi, true);
3004 f2fs_msg(sb, KERN_INFO,
3005 "Try to recover %dth superblock, ret: %d",
3006 sbi->valid_super_block ? 1 : 2, err);
3009 f2fs_join_shrinker(sbi);
3011 f2fs_tuning_parameters(sbi);
3013 f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx",
3014 cur_cp_version(F2FS_CKPT(sbi)));
3015 f2fs_update_time(sbi, CP_TIME);
3016 f2fs_update_time(sbi, REQ_TIME);
3021 if (f2fs_sb_has_quota_ino(sb) && !f2fs_readonly(sb))
3022 f2fs_quota_off_umount(sbi->sb);
3024 f2fs_sync_inode_meta(sbi);
3026 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3027 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3028 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3029 * falls into an infinite loop in f2fs_sync_meta_pages().
3031 truncate_inode_pages_final(META_MAPPING(sbi));
3035 f2fs_unregister_sysfs(sbi);
3040 f2fs_destroy_stats(sbi);
3042 f2fs_release_ino_entry(sbi, true);
3043 truncate_inode_pages_final(NODE_MAPPING(sbi));
3044 iput(sbi->node_inode);
3046 f2fs_destroy_node_manager(sbi);
3048 f2fs_destroy_segment_manager(sbi);
3050 destroy_device_list(sbi);
3053 make_bad_inode(sbi->meta_inode);
3054 iput(sbi->meta_inode);
3056 mempool_destroy(sbi->write_io_dummy);
3058 destroy_percpu_info(sbi);
3060 for (i = 0; i < NR_PAGE_TYPE; i++)
3061 kfree(sbi->write_io[i]);
3064 for (i = 0; i < MAXQUOTAS; i++)
3065 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
3071 if (sbi->s_chksum_driver)
3072 crypto_free_shash(sbi->s_chksum_driver);
3075 /* give only one another chance */
3078 shrink_dcache_sb(sb);
3084 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3085 const char *dev_name, void *data)
3087 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3090 static void kill_f2fs_super(struct super_block *sb)
3093 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
3094 f2fs_stop_gc_thread(F2FS_SB(sb));
3095 f2fs_stop_discard_thread(F2FS_SB(sb));
3097 kill_block_super(sb);
3100 static struct file_system_type f2fs_fs_type = {
3101 .owner = THIS_MODULE,
3103 .mount = f2fs_mount,
3104 .kill_sb = kill_f2fs_super,
3105 .fs_flags = FS_REQUIRES_DEV,
3107 MODULE_ALIAS_FS("f2fs");
3109 static int __init init_inodecache(void)
3111 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3112 sizeof(struct f2fs_inode_info), 0,
3113 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3114 if (!f2fs_inode_cachep)
3119 static void destroy_inodecache(void)
3122 * Make sure all delayed rcu free inodes are flushed before we
3126 kmem_cache_destroy(f2fs_inode_cachep);
3129 static int __init init_f2fs_fs(void)
3133 if (PAGE_SIZE != F2FS_BLKSIZE) {
3134 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3135 PAGE_SIZE, F2FS_BLKSIZE);
3139 f2fs_build_trace_ios();
3141 err = init_inodecache();
3144 err = f2fs_create_node_manager_caches();
3146 goto free_inodecache;
3147 err = f2fs_create_segment_manager_caches();
3149 goto free_node_manager_caches;
3150 err = f2fs_create_checkpoint_caches();
3152 goto free_segment_manager_caches;
3153 err = f2fs_create_extent_cache();
3155 goto free_checkpoint_caches;
3156 err = f2fs_init_sysfs();
3158 goto free_extent_cache;
3159 err = register_shrinker(&f2fs_shrinker_info);
3162 err = register_filesystem(&f2fs_fs_type);
3165 err = f2fs_create_root_stats();
3167 goto free_filesystem;
3168 err = f2fs_init_post_read_processing();
3170 goto free_root_stats;
3174 f2fs_destroy_root_stats();
3176 unregister_filesystem(&f2fs_fs_type);
3178 unregister_shrinker(&f2fs_shrinker_info);
3182 f2fs_destroy_extent_cache();
3183 free_checkpoint_caches:
3184 f2fs_destroy_checkpoint_caches();
3185 free_segment_manager_caches:
3186 f2fs_destroy_segment_manager_caches();
3187 free_node_manager_caches:
3188 f2fs_destroy_node_manager_caches();
3190 destroy_inodecache();
3195 static void __exit exit_f2fs_fs(void)
3197 f2fs_destroy_post_read_processing();
3198 f2fs_destroy_root_stats();
3199 unregister_filesystem(&f2fs_fs_type);
3200 unregister_shrinker(&f2fs_shrinker_info);
3202 f2fs_destroy_extent_cache();
3203 f2fs_destroy_checkpoint_caches();
3204 f2fs_destroy_segment_manager_caches();
3205 f2fs_destroy_node_manager_caches();
3206 destroy_inodecache();
3207 f2fs_destroy_trace_ios();
3210 module_init(init_f2fs_fs)
3211 module_exit(exit_f2fs_fs)
3213 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3214 MODULE_DESCRIPTION("Flash Friendly File System");
3215 MODULE_LICENSE("GPL");