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 |= MS_LAZYTIME;
620 sb->s_flags &= ~MS_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 fi->i_current_depth = 1;
834 init_rwsem(&fi->i_sem);
835 INIT_LIST_HEAD(&fi->dirty_list);
836 INIT_LIST_HEAD(&fi->gdirty_list);
837 INIT_LIST_HEAD(&fi->inmem_ilist);
838 INIT_LIST_HEAD(&fi->inmem_pages);
839 mutex_init(&fi->inmem_lock);
840 init_rwsem(&fi->dio_rwsem[READ]);
841 init_rwsem(&fi->dio_rwsem[WRITE]);
842 init_rwsem(&fi->i_mmap_sem);
843 init_rwsem(&fi->i_xattr_sem);
845 /* Will be used by directory only */
846 fi->i_dir_level = F2FS_SB(sb)->dir_level;
848 return &fi->vfs_inode;
851 static int f2fs_drop_inode(struct inode *inode)
855 * This is to avoid a deadlock condition like below.
856 * writeback_single_inode(inode)
857 * - f2fs_write_data_page
858 * - f2fs_gc -> iput -> evict
859 * - inode_wait_for_writeback(inode)
861 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
862 if (!inode->i_nlink && !is_bad_inode(inode)) {
863 /* to avoid evict_inode call simultaneously */
864 atomic_inc(&inode->i_count);
865 spin_unlock(&inode->i_lock);
867 /* some remained atomic pages should discarded */
868 if (f2fs_is_atomic_file(inode))
869 drop_inmem_pages(inode);
871 /* should remain fi->extent_tree for writepage */
872 f2fs_destroy_extent_node(inode);
874 sb_start_intwrite(inode->i_sb);
875 f2fs_i_size_write(inode, 0);
877 if (F2FS_HAS_BLOCKS(inode))
878 f2fs_truncate(inode);
880 sb_end_intwrite(inode->i_sb);
882 spin_lock(&inode->i_lock);
883 atomic_dec(&inode->i_count);
885 trace_f2fs_drop_inode(inode, 0);
888 ret = generic_drop_inode(inode);
889 trace_f2fs_drop_inode(inode, ret);
893 int f2fs_inode_dirtied(struct inode *inode, bool sync)
895 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
898 spin_lock(&sbi->inode_lock[DIRTY_META]);
899 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
902 set_inode_flag(inode, FI_DIRTY_INODE);
903 stat_inc_dirty_inode(sbi, DIRTY_META);
905 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
906 list_add_tail(&F2FS_I(inode)->gdirty_list,
907 &sbi->inode_list[DIRTY_META]);
908 inc_page_count(sbi, F2FS_DIRTY_IMETA);
910 spin_unlock(&sbi->inode_lock[DIRTY_META]);
914 void f2fs_inode_synced(struct inode *inode)
916 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
918 spin_lock(&sbi->inode_lock[DIRTY_META]);
919 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
920 spin_unlock(&sbi->inode_lock[DIRTY_META]);
923 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
924 list_del_init(&F2FS_I(inode)->gdirty_list);
925 dec_page_count(sbi, F2FS_DIRTY_IMETA);
927 clear_inode_flag(inode, FI_DIRTY_INODE);
928 clear_inode_flag(inode, FI_AUTO_RECOVER);
929 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
930 spin_unlock(&sbi->inode_lock[DIRTY_META]);
934 * f2fs_dirty_inode() is called from __mark_inode_dirty()
936 * We should call set_dirty_inode to write the dirty inode through write_inode.
938 static void f2fs_dirty_inode(struct inode *inode, int flags)
940 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
942 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
943 inode->i_ino == F2FS_META_INO(sbi))
946 if (flags == I_DIRTY_TIME)
949 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
950 clear_inode_flag(inode, FI_AUTO_RECOVER);
952 f2fs_inode_dirtied(inode, false);
955 static void f2fs_i_callback(struct rcu_head *head)
957 struct inode *inode = container_of(head, struct inode, i_rcu);
958 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
961 static void f2fs_destroy_inode(struct inode *inode)
963 call_rcu(&inode->i_rcu, f2fs_i_callback);
966 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
968 percpu_counter_destroy(&sbi->alloc_valid_block_count);
969 percpu_counter_destroy(&sbi->total_valid_inode_count);
972 static void destroy_device_list(struct f2fs_sb_info *sbi)
976 for (i = 0; i < sbi->s_ndevs; i++) {
977 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
978 #ifdef CONFIG_BLK_DEV_ZONED
979 kfree(FDEV(i).blkz_type);
985 static void f2fs_put_super(struct super_block *sb)
987 struct f2fs_sb_info *sbi = F2FS_SB(sb);
991 f2fs_quota_off_umount(sb);
993 /* prevent remaining shrinker jobs */
994 mutex_lock(&sbi->umount_mutex);
997 * We don't need to do checkpoint when superblock is clean.
998 * But, the previous checkpoint was not done by umount, it needs to do
999 * clean checkpoint again.
1001 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1002 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
1003 struct cp_control cpc = {
1004 .reason = CP_UMOUNT,
1006 write_checkpoint(sbi, &cpc);
1009 /* be sure to wait for any on-going discard commands */
1010 dropped = f2fs_wait_discard_bios(sbi);
1012 if (f2fs_discard_en(sbi) && !sbi->discard_blks && !dropped) {
1013 struct cp_control cpc = {
1014 .reason = CP_UMOUNT | CP_TRIMMED,
1016 write_checkpoint(sbi, &cpc);
1019 /* write_checkpoint can update stat informaion */
1020 f2fs_destroy_stats(sbi);
1023 * normally superblock is clean, so we need to release this.
1024 * In addition, EIO will skip do checkpoint, we need this as well.
1026 release_ino_entry(sbi, true);
1028 f2fs_leave_shrinker(sbi);
1029 mutex_unlock(&sbi->umount_mutex);
1031 /* our cp_error case, we can wait for any writeback page */
1032 f2fs_flush_merged_writes(sbi);
1034 iput(sbi->node_inode);
1035 iput(sbi->meta_inode);
1037 /* destroy f2fs internal modules */
1038 destroy_node_manager(sbi);
1039 destroy_segment_manager(sbi);
1043 f2fs_unregister_sysfs(sbi);
1045 sb->s_fs_info = NULL;
1046 if (sbi->s_chksum_driver)
1047 crypto_free_shash(sbi->s_chksum_driver);
1048 kfree(sbi->raw_super);
1050 destroy_device_list(sbi);
1051 if (sbi->write_io_dummy)
1052 mempool_destroy(sbi->write_io_dummy);
1054 for (i = 0; i < MAXQUOTAS; i++)
1055 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1057 destroy_percpu_info(sbi);
1058 for (i = 0; i < NR_PAGE_TYPE; i++)
1059 kfree(sbi->write_io[i]);
1063 int f2fs_sync_fs(struct super_block *sb, int sync)
1065 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1068 if (unlikely(f2fs_cp_error(sbi)))
1071 trace_f2fs_sync_fs(sb, sync);
1073 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1077 struct cp_control cpc;
1079 cpc.reason = __get_cp_reason(sbi);
1081 mutex_lock(&sbi->gc_mutex);
1082 err = write_checkpoint(sbi, &cpc);
1083 mutex_unlock(&sbi->gc_mutex);
1085 f2fs_trace_ios(NULL, 1);
1090 static int f2fs_freeze(struct super_block *sb)
1092 if (f2fs_readonly(sb))
1095 /* IO error happened before */
1096 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1099 /* must be clean, since sync_filesystem() was already called */
1100 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1105 static int f2fs_unfreeze(struct super_block *sb)
1111 static int f2fs_statfs_project(struct super_block *sb,
1112 kprojid_t projid, struct kstatfs *buf)
1115 struct dquot *dquot;
1119 qid = make_kqid_projid(projid);
1120 dquot = dqget(sb, qid);
1122 return PTR_ERR(dquot);
1123 spin_lock(&dq_data_lock);
1125 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
1126 dquot->dq_dqb.dqb_bsoftlimit :
1127 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
1128 if (limit && buf->f_blocks > limit) {
1129 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1130 buf->f_blocks = limit;
1131 buf->f_bfree = buf->f_bavail =
1132 (buf->f_blocks > curblock) ?
1133 (buf->f_blocks - curblock) : 0;
1136 limit = dquot->dq_dqb.dqb_isoftlimit ?
1137 dquot->dq_dqb.dqb_isoftlimit :
1138 dquot->dq_dqb.dqb_ihardlimit;
1139 if (limit && buf->f_files > limit) {
1140 buf->f_files = limit;
1142 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1143 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1146 spin_unlock(&dq_data_lock);
1152 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1154 struct super_block *sb = dentry->d_sb;
1155 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1156 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1157 block_t total_count, user_block_count, start_count;
1158 u64 avail_node_count;
1160 total_count = le64_to_cpu(sbi->raw_super->block_count);
1161 user_block_count = sbi->user_block_count;
1162 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1163 buf->f_type = F2FS_SUPER_MAGIC;
1164 buf->f_bsize = sbi->blocksize;
1166 buf->f_blocks = total_count - start_count;
1167 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1168 sbi->current_reserved_blocks;
1169 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1170 buf->f_bavail = buf->f_bfree -
1171 F2FS_OPTION(sbi).root_reserved_blocks;
1175 avail_node_count = sbi->total_node_count - sbi->nquota_files -
1176 F2FS_RESERVED_NODE_NUM;
1178 if (avail_node_count > user_block_count) {
1179 buf->f_files = user_block_count;
1180 buf->f_ffree = buf->f_bavail;
1182 buf->f_files = avail_node_count;
1183 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1187 buf->f_namelen = F2FS_NAME_LEN;
1188 buf->f_fsid.val[0] = (u32)id;
1189 buf->f_fsid.val[1] = (u32)(id >> 32);
1192 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1193 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1194 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1200 static inline void f2fs_show_quota_options(struct seq_file *seq,
1201 struct super_block *sb)
1204 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1206 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1209 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1220 seq_printf(seq, ",jqfmt=%s", fmtname);
1223 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1224 seq_show_option(seq, "usrjquota",
1225 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1227 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1228 seq_show_option(seq, "grpjquota",
1229 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1231 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1232 seq_show_option(seq, "prjjquota",
1233 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1237 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1239 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1241 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1242 if (test_opt(sbi, FORCE_FG_GC))
1243 seq_printf(seq, ",background_gc=%s", "sync");
1245 seq_printf(seq, ",background_gc=%s", "on");
1247 seq_printf(seq, ",background_gc=%s", "off");
1249 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1250 seq_puts(seq, ",disable_roll_forward");
1251 if (test_opt(sbi, DISCARD))
1252 seq_puts(seq, ",discard");
1253 if (test_opt(sbi, NOHEAP))
1254 seq_puts(seq, ",no_heap");
1256 seq_puts(seq, ",heap");
1257 #ifdef CONFIG_F2FS_FS_XATTR
1258 if (test_opt(sbi, XATTR_USER))
1259 seq_puts(seq, ",user_xattr");
1261 seq_puts(seq, ",nouser_xattr");
1262 if (test_opt(sbi, INLINE_XATTR))
1263 seq_puts(seq, ",inline_xattr");
1265 seq_puts(seq, ",noinline_xattr");
1266 if (test_opt(sbi, INLINE_XATTR_SIZE))
1267 seq_printf(seq, ",inline_xattr_size=%u",
1268 F2FS_OPTION(sbi).inline_xattr_size);
1270 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1271 if (test_opt(sbi, POSIX_ACL))
1272 seq_puts(seq, ",acl");
1274 seq_puts(seq, ",noacl");
1276 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1277 seq_puts(seq, ",disable_ext_identify");
1278 if (test_opt(sbi, INLINE_DATA))
1279 seq_puts(seq, ",inline_data");
1281 seq_puts(seq, ",noinline_data");
1282 if (test_opt(sbi, INLINE_DENTRY))
1283 seq_puts(seq, ",inline_dentry");
1285 seq_puts(seq, ",noinline_dentry");
1286 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1287 seq_puts(seq, ",flush_merge");
1288 if (test_opt(sbi, NOBARRIER))
1289 seq_puts(seq, ",nobarrier");
1290 if (test_opt(sbi, FASTBOOT))
1291 seq_puts(seq, ",fastboot");
1292 if (test_opt(sbi, EXTENT_CACHE))
1293 seq_puts(seq, ",extent_cache");
1295 seq_puts(seq, ",noextent_cache");
1296 if (test_opt(sbi, DATA_FLUSH))
1297 seq_puts(seq, ",data_flush");
1299 seq_puts(seq, ",mode=");
1300 if (test_opt(sbi, ADAPTIVE))
1301 seq_puts(seq, "adaptive");
1302 else if (test_opt(sbi, LFS))
1303 seq_puts(seq, "lfs");
1304 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1305 if (test_opt(sbi, RESERVE_ROOT))
1306 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1307 F2FS_OPTION(sbi).root_reserved_blocks,
1308 from_kuid_munged(&init_user_ns,
1309 F2FS_OPTION(sbi).s_resuid),
1310 from_kgid_munged(&init_user_ns,
1311 F2FS_OPTION(sbi).s_resgid));
1312 if (F2FS_IO_SIZE_BITS(sbi))
1313 seq_printf(seq, ",io_size=%uKB", F2FS_IO_SIZE_KB(sbi));
1314 #ifdef CONFIG_F2FS_FAULT_INJECTION
1315 if (test_opt(sbi, FAULT_INJECTION))
1316 seq_printf(seq, ",fault_injection=%u",
1317 F2FS_OPTION(sbi).fault_info.inject_rate);
1320 if (test_opt(sbi, QUOTA))
1321 seq_puts(seq, ",quota");
1322 if (test_opt(sbi, USRQUOTA))
1323 seq_puts(seq, ",usrquota");
1324 if (test_opt(sbi, GRPQUOTA))
1325 seq_puts(seq, ",grpquota");
1326 if (test_opt(sbi, PRJQUOTA))
1327 seq_puts(seq, ",prjquota");
1329 f2fs_show_quota_options(seq, sbi->sb);
1330 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1331 seq_printf(seq, ",whint_mode=%s", "user-based");
1332 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1333 seq_printf(seq, ",whint_mode=%s", "fs-based");
1334 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1335 if (F2FS_OPTION(sbi).test_dummy_encryption)
1336 seq_puts(seq, ",test_dummy_encryption");
1339 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1340 seq_printf(seq, ",alloc_mode=%s", "default");
1341 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1342 seq_printf(seq, ",alloc_mode=%s", "reuse");
1344 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1345 seq_printf(seq, ",fsync_mode=%s", "posix");
1346 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1347 seq_printf(seq, ",fsync_mode=%s", "strict");
1351 static void default_options(struct f2fs_sb_info *sbi)
1353 /* init some FS parameters */
1354 F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1355 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1356 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1357 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1358 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1359 F2FS_OPTION(sbi).test_dummy_encryption = false;
1360 sbi->readdir_ra = 1;
1362 set_opt(sbi, BG_GC);
1363 set_opt(sbi, INLINE_XATTR);
1364 set_opt(sbi, INLINE_DATA);
1365 set_opt(sbi, INLINE_DENTRY);
1366 set_opt(sbi, EXTENT_CACHE);
1367 set_opt(sbi, NOHEAP);
1368 sbi->sb->s_flags |= MS_LAZYTIME;
1369 set_opt(sbi, FLUSH_MERGE);
1370 if (f2fs_sb_has_blkzoned(sbi->sb)) {
1371 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1372 set_opt(sbi, DISCARD);
1374 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1377 #ifdef CONFIG_F2FS_FS_XATTR
1378 set_opt(sbi, XATTR_USER);
1380 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1381 set_opt(sbi, POSIX_ACL);
1384 #ifdef CONFIG_F2FS_FAULT_INJECTION
1385 f2fs_build_fault_attr(sbi, 0);
1390 static int f2fs_enable_quotas(struct super_block *sb);
1392 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1394 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1395 struct f2fs_mount_info org_mount_opt;
1396 unsigned long old_sb_flags;
1398 bool need_restart_gc = false;
1399 bool need_stop_gc = false;
1400 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1406 * Save the old mount options in case we
1407 * need to restore them.
1409 org_mount_opt = sbi->mount_opt;
1410 old_sb_flags = sb->s_flags;
1413 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1414 for (i = 0; i < MAXQUOTAS; i++) {
1415 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1416 org_mount_opt.s_qf_names[i] =
1417 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1419 if (!org_mount_opt.s_qf_names[i]) {
1420 for (j = 0; j < i; j++)
1421 kfree(org_mount_opt.s_qf_names[j]);
1425 org_mount_opt.s_qf_names[i] = NULL;
1430 /* recover superblocks we couldn't write due to previous RO mount */
1431 if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1432 err = f2fs_commit_super(sbi, false);
1433 f2fs_msg(sb, KERN_INFO,
1434 "Try to recover all the superblocks, ret: %d", err);
1436 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1439 default_options(sbi);
1441 /* parse mount options */
1442 err = parse_options(sb, data);
1447 * Previous and new state of filesystem is RO,
1448 * so skip checking GC and FLUSH_MERGE conditions.
1450 if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
1454 if (!f2fs_readonly(sb) && (*flags & MS_RDONLY)) {
1455 err = dquot_suspend(sb, -1);
1458 } else if (f2fs_readonly(sb) && !(*flags & MS_RDONLY)) {
1459 /* dquot_resume needs RW */
1460 sb->s_flags &= ~MS_RDONLY;
1461 if (sb_any_quota_suspended(sb)) {
1462 dquot_resume(sb, -1);
1463 } else if (f2fs_sb_has_quota_ino(sb)) {
1464 err = f2fs_enable_quotas(sb);
1470 /* disallow enable/disable extent_cache dynamically */
1471 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1473 f2fs_msg(sbi->sb, KERN_WARNING,
1474 "switch extent_cache option is not allowed");
1479 * We stop the GC thread if FS is mounted as RO
1480 * or if background_gc = off is passed in mount
1481 * option. Also sync the filesystem.
1483 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
1484 if (sbi->gc_thread) {
1485 stop_gc_thread(sbi);
1486 need_restart_gc = true;
1488 } else if (!sbi->gc_thread) {
1489 err = start_gc_thread(sbi);
1492 need_stop_gc = true;
1495 if (*flags & MS_RDONLY ||
1496 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1497 writeback_inodes_sb(sb, WB_REASON_SYNC);
1500 set_sbi_flag(sbi, SBI_IS_DIRTY);
1501 set_sbi_flag(sbi, SBI_IS_CLOSE);
1502 f2fs_sync_fs(sb, 1);
1503 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1507 * We stop issue flush thread if FS is mounted as RO
1508 * or if flush_merge is not passed in mount option.
1510 if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1511 clear_opt(sbi, FLUSH_MERGE);
1512 destroy_flush_cmd_control(sbi, false);
1514 err = create_flush_cmd_control(sbi);
1520 /* Release old quota file names */
1521 for (i = 0; i < MAXQUOTAS; i++)
1522 kfree(org_mount_opt.s_qf_names[i]);
1524 /* Update the POSIXACL Flag */
1525 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1526 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1528 limit_reserve_root(sbi);
1531 if (need_restart_gc) {
1532 if (start_gc_thread(sbi))
1533 f2fs_msg(sbi->sb, KERN_WARNING,
1534 "background gc thread has stopped");
1535 } else if (need_stop_gc) {
1536 stop_gc_thread(sbi);
1540 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1541 for (i = 0; i < MAXQUOTAS; i++) {
1542 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1543 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1546 sbi->mount_opt = org_mount_opt;
1547 sb->s_flags = old_sb_flags;
1552 /* Read data from quotafile */
1553 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1554 size_t len, loff_t off)
1556 struct inode *inode = sb_dqopt(sb)->files[type];
1557 struct address_space *mapping = inode->i_mapping;
1558 block_t blkidx = F2FS_BYTES_TO_BLK(off);
1559 int offset = off & (sb->s_blocksize - 1);
1562 loff_t i_size = i_size_read(inode);
1569 if (off + len > i_size)
1572 while (toread > 0) {
1573 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1575 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1577 if (PTR_ERR(page) == -ENOMEM) {
1578 congestion_wait(BLK_RW_ASYNC, HZ/50);
1581 return PTR_ERR(page);
1586 if (unlikely(page->mapping != mapping)) {
1587 f2fs_put_page(page, 1);
1590 if (unlikely(!PageUptodate(page))) {
1591 f2fs_put_page(page, 1);
1595 kaddr = kmap_atomic(page);
1596 memcpy(data, kaddr + offset, tocopy);
1597 kunmap_atomic(kaddr);
1598 f2fs_put_page(page, 1);
1608 /* Write to quotafile */
1609 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1610 const char *data, size_t len, loff_t off)
1612 struct inode *inode = sb_dqopt(sb)->files[type];
1613 struct address_space *mapping = inode->i_mapping;
1614 const struct address_space_operations *a_ops = mapping->a_ops;
1615 int offset = off & (sb->s_blocksize - 1);
1616 size_t towrite = len;
1622 while (towrite > 0) {
1623 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1626 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1628 if (unlikely(err)) {
1629 if (err == -ENOMEM) {
1630 congestion_wait(BLK_RW_ASYNC, HZ/50);
1636 kaddr = kmap_atomic(page);
1637 memcpy(kaddr + offset, data, tocopy);
1638 kunmap_atomic(kaddr);
1639 flush_dcache_page(page);
1641 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1652 inode->i_mtime = inode->i_ctime = current_time(inode);
1653 f2fs_mark_inode_dirty_sync(inode, false);
1654 return len - towrite;
1657 static struct dquot **f2fs_get_dquots(struct inode *inode)
1659 return F2FS_I(inode)->i_dquot;
1662 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1664 return &F2FS_I(inode)->i_reserved_quota;
1667 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1669 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
1670 F2FS_OPTION(sbi).s_jquota_fmt, type);
1673 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1678 if (f2fs_sb_has_quota_ino(sbi->sb) && rdonly) {
1679 err = f2fs_enable_quotas(sbi->sb);
1681 f2fs_msg(sbi->sb, KERN_ERR,
1682 "Cannot turn on quota_ino: %d", err);
1688 for (i = 0; i < MAXQUOTAS; i++) {
1689 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1690 err = f2fs_quota_on_mount(sbi, i);
1695 f2fs_msg(sbi->sb, KERN_ERR,
1696 "Cannot turn on quotas: %d on %d", err, i);
1702 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
1705 struct inode *qf_inode;
1706 unsigned long qf_inum;
1709 BUG_ON(!f2fs_sb_has_quota_ino(sb));
1711 qf_inum = f2fs_qf_ino(sb, type);
1715 qf_inode = f2fs_iget(sb, qf_inum);
1716 if (IS_ERR(qf_inode)) {
1717 f2fs_msg(sb, KERN_ERR,
1718 "Bad quota inode %u:%lu", type, qf_inum);
1719 return PTR_ERR(qf_inode);
1722 /* Don't account quota for quota files to avoid recursion */
1723 qf_inode->i_flags |= S_NOQUOTA;
1724 err = dquot_enable(qf_inode, type, format_id, flags);
1729 static int f2fs_enable_quotas(struct super_block *sb)
1732 unsigned long qf_inum;
1733 bool quota_mopt[MAXQUOTAS] = {
1734 test_opt(F2FS_SB(sb), USRQUOTA),
1735 test_opt(F2FS_SB(sb), GRPQUOTA),
1736 test_opt(F2FS_SB(sb), PRJQUOTA),
1739 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
1740 for (type = 0; type < MAXQUOTAS; type++) {
1741 qf_inum = f2fs_qf_ino(sb, type);
1743 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
1744 DQUOT_USAGE_ENABLED |
1745 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
1747 f2fs_msg(sb, KERN_ERR,
1748 "Failed to enable quota tracking "
1749 "(type=%d, err=%d). Please run "
1750 "fsck to fix.", type, err);
1751 for (type--; type >= 0; type--)
1752 dquot_quota_off(sb, type);
1760 static int f2fs_quota_sync(struct super_block *sb, int type)
1762 struct quota_info *dqopt = sb_dqopt(sb);
1766 ret = dquot_writeback_dquots(sb, type);
1771 * Now when everything is written we can discard the pagecache so
1772 * that userspace sees the changes.
1774 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
1775 if (type != -1 && cnt != type)
1777 if (!sb_has_quota_active(sb, cnt))
1780 ret = filemap_write_and_wait(dqopt->files[cnt]->i_mapping);
1784 inode_lock(dqopt->files[cnt]);
1785 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
1786 inode_unlock(dqopt->files[cnt]);
1791 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
1794 struct inode *inode;
1797 err = f2fs_quota_sync(sb, type);
1801 err = dquot_quota_on(sb, type, format_id, path);
1805 inode = d_inode(path->dentry);
1808 F2FS_I(inode)->i_flags |= FS_NOATIME_FL | FS_IMMUTABLE_FL;
1809 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1810 S_NOATIME | S_IMMUTABLE);
1811 inode_unlock(inode);
1812 f2fs_mark_inode_dirty_sync(inode, false);
1817 static int f2fs_quota_off(struct super_block *sb, int type)
1819 struct inode *inode = sb_dqopt(sb)->files[type];
1822 if (!inode || !igrab(inode))
1823 return dquot_quota_off(sb, type);
1825 f2fs_quota_sync(sb, type);
1827 err = dquot_quota_off(sb, type);
1828 if (err || f2fs_sb_has_quota_ino(sb))
1832 F2FS_I(inode)->i_flags &= ~(FS_NOATIME_FL | FS_IMMUTABLE_FL);
1833 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1834 inode_unlock(inode);
1835 f2fs_mark_inode_dirty_sync(inode, false);
1841 void f2fs_quota_off_umount(struct super_block *sb)
1845 for (type = 0; type < MAXQUOTAS; type++)
1846 f2fs_quota_off(sb, type);
1850 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
1852 *projid = F2FS_I(inode)->i_projid;
1857 static const struct dquot_operations f2fs_quota_operations = {
1858 .get_reserved_space = f2fs_get_reserved_space,
1859 .write_dquot = dquot_commit,
1860 .acquire_dquot = dquot_acquire,
1861 .release_dquot = dquot_release,
1862 .mark_dirty = dquot_mark_dquot_dirty,
1863 .write_info = dquot_commit_info,
1864 .alloc_dquot = dquot_alloc,
1865 .destroy_dquot = dquot_destroy,
1867 .get_projid = f2fs_get_projid,
1868 .get_next_id = dquot_get_next_id,
1872 static const struct quotactl_ops f2fs_quotactl_ops = {
1873 .quota_on = f2fs_quota_on,
1874 .quota_off = f2fs_quota_off,
1875 .quota_sync = f2fs_quota_sync,
1876 .get_state = dquot_get_state,
1877 .set_info = dquot_set_dqinfo,
1878 .get_dqblk = dquot_get_dqblk,
1879 .set_dqblk = dquot_set_dqblk,
1882 void f2fs_quota_off_umount(struct super_block *sb)
1887 static const struct super_operations f2fs_sops = {
1888 .alloc_inode = f2fs_alloc_inode,
1889 .drop_inode = f2fs_drop_inode,
1890 .destroy_inode = f2fs_destroy_inode,
1891 .write_inode = f2fs_write_inode,
1892 .dirty_inode = f2fs_dirty_inode,
1893 .show_options = f2fs_show_options,
1895 .quota_read = f2fs_quota_read,
1896 .quota_write = f2fs_quota_write,
1897 .get_dquots = f2fs_get_dquots,
1899 .evict_inode = f2fs_evict_inode,
1900 .put_super = f2fs_put_super,
1901 .sync_fs = f2fs_sync_fs,
1902 .freeze_fs = f2fs_freeze,
1903 .unfreeze_fs = f2fs_unfreeze,
1904 .statfs = f2fs_statfs,
1905 .remount_fs = f2fs_remount,
1908 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1909 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1911 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1912 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1916 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1919 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1922 * Encrypting the root directory is not allowed because fsck
1923 * expects lost+found directory to exist and remain unencrypted
1924 * if LOST_FOUND feature is enabled.
1927 if (f2fs_sb_has_lost_found(sbi->sb) &&
1928 inode->i_ino == F2FS_ROOT_INO(sbi))
1931 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1932 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1933 ctx, len, fs_data, XATTR_CREATE);
1936 static bool f2fs_dummy_context(struct inode *inode)
1938 return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
1941 static unsigned f2fs_max_namelen(struct inode *inode)
1943 return S_ISLNK(inode->i_mode) ?
1944 inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1947 static const struct fscrypt_operations f2fs_cryptops = {
1948 .key_prefix = "f2fs:",
1949 .get_context = f2fs_get_context,
1950 .set_context = f2fs_set_context,
1951 .dummy_context = f2fs_dummy_context,
1952 .empty_dir = f2fs_empty_dir,
1953 .max_namelen = f2fs_max_namelen,
1957 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1958 u64 ino, u32 generation)
1960 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1961 struct inode *inode;
1963 if (check_nid_range(sbi, ino))
1964 return ERR_PTR(-ESTALE);
1967 * f2fs_iget isn't quite right if the inode is currently unallocated!
1968 * However f2fs_iget currently does appropriate checks to handle stale
1969 * inodes so everything is OK.
1971 inode = f2fs_iget(sb, ino);
1973 return ERR_CAST(inode);
1974 if (unlikely(generation && inode->i_generation != generation)) {
1975 /* we didn't find the right inode.. */
1977 return ERR_PTR(-ESTALE);
1982 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1983 int fh_len, int fh_type)
1985 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1986 f2fs_nfs_get_inode);
1989 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1990 int fh_len, int fh_type)
1992 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1993 f2fs_nfs_get_inode);
1996 static const struct export_operations f2fs_export_ops = {
1997 .fh_to_dentry = f2fs_fh_to_dentry,
1998 .fh_to_parent = f2fs_fh_to_parent,
1999 .get_parent = f2fs_get_parent,
2002 static loff_t max_file_blocks(void)
2005 loff_t leaf_count = ADDRS_PER_BLOCK;
2008 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2009 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2010 * space in inode.i_addr, it will be more safe to reassign
2014 /* two direct node blocks */
2015 result += (leaf_count * 2);
2017 /* two indirect node blocks */
2018 leaf_count *= NIDS_PER_BLOCK;
2019 result += (leaf_count * 2);
2021 /* one double indirect node block */
2022 leaf_count *= NIDS_PER_BLOCK;
2023 result += leaf_count;
2028 static int __f2fs_commit_super(struct buffer_head *bh,
2029 struct f2fs_super_block *super)
2033 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2034 set_buffer_dirty(bh);
2037 /* it's rare case, we can do fua all the time */
2038 return __sync_dirty_buffer(bh, WRITE_FLUSH_FUA);
2041 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2042 struct buffer_head *bh)
2044 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2045 (bh->b_data + F2FS_SUPER_OFFSET);
2046 struct super_block *sb = sbi->sb;
2047 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2048 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2049 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2050 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2051 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2052 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2053 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2054 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2055 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2056 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2057 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2058 u32 segment_count = le32_to_cpu(raw_super->segment_count);
2059 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2060 u64 main_end_blkaddr = main_blkaddr +
2061 (segment_count_main << log_blocks_per_seg);
2062 u64 seg_end_blkaddr = segment0_blkaddr +
2063 (segment_count << log_blocks_per_seg);
2065 if (segment0_blkaddr != cp_blkaddr) {
2066 f2fs_msg(sb, KERN_INFO,
2067 "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2068 segment0_blkaddr, cp_blkaddr);
2072 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2074 f2fs_msg(sb, KERN_INFO,
2075 "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2076 cp_blkaddr, sit_blkaddr,
2077 segment_count_ckpt << log_blocks_per_seg);
2081 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2083 f2fs_msg(sb, KERN_INFO,
2084 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2085 sit_blkaddr, nat_blkaddr,
2086 segment_count_sit << log_blocks_per_seg);
2090 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2092 f2fs_msg(sb, KERN_INFO,
2093 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2094 nat_blkaddr, ssa_blkaddr,
2095 segment_count_nat << log_blocks_per_seg);
2099 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2101 f2fs_msg(sb, KERN_INFO,
2102 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2103 ssa_blkaddr, main_blkaddr,
2104 segment_count_ssa << log_blocks_per_seg);
2108 if (main_end_blkaddr > seg_end_blkaddr) {
2109 f2fs_msg(sb, KERN_INFO,
2110 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2113 (segment_count << log_blocks_per_seg),
2114 segment_count_main << log_blocks_per_seg);
2116 } else if (main_end_blkaddr < seg_end_blkaddr) {
2120 /* fix in-memory information all the time */
2121 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2122 segment0_blkaddr) >> log_blocks_per_seg);
2124 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2125 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2128 err = __f2fs_commit_super(bh, NULL);
2129 res = err ? "failed" : "done";
2131 f2fs_msg(sb, KERN_INFO,
2132 "Fix alignment : %s, start(%u) end(%u) block(%u)",
2135 (segment_count << log_blocks_per_seg),
2136 segment_count_main << log_blocks_per_seg);
2143 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2144 struct buffer_head *bh)
2146 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2147 (bh->b_data + F2FS_SUPER_OFFSET);
2148 struct super_block *sb = sbi->sb;
2149 unsigned int blocksize;
2151 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
2152 f2fs_msg(sb, KERN_INFO,
2153 "Magic Mismatch, valid(0x%x) - read(0x%x)",
2154 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2158 /* Currently, support only 4KB page cache size */
2159 if (F2FS_BLKSIZE != PAGE_SIZE) {
2160 f2fs_msg(sb, KERN_INFO,
2161 "Invalid page_cache_size (%lu), supports only 4KB\n",
2166 /* Currently, support only 4KB block size */
2167 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2168 if (blocksize != F2FS_BLKSIZE) {
2169 f2fs_msg(sb, KERN_INFO,
2170 "Invalid blocksize (%u), supports only 4KB\n",
2175 /* check log blocks per segment */
2176 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2177 f2fs_msg(sb, KERN_INFO,
2178 "Invalid log blocks per segment (%u)\n",
2179 le32_to_cpu(raw_super->log_blocks_per_seg));
2183 /* Currently, support 512/1024/2048/4096 bytes sector size */
2184 if (le32_to_cpu(raw_super->log_sectorsize) >
2185 F2FS_MAX_LOG_SECTOR_SIZE ||
2186 le32_to_cpu(raw_super->log_sectorsize) <
2187 F2FS_MIN_LOG_SECTOR_SIZE) {
2188 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
2189 le32_to_cpu(raw_super->log_sectorsize));
2192 if (le32_to_cpu(raw_super->log_sectors_per_block) +
2193 le32_to_cpu(raw_super->log_sectorsize) !=
2194 F2FS_MAX_LOG_SECTOR_SIZE) {
2195 f2fs_msg(sb, KERN_INFO,
2196 "Invalid log sectors per block(%u) log sectorsize(%u)",
2197 le32_to_cpu(raw_super->log_sectors_per_block),
2198 le32_to_cpu(raw_super->log_sectorsize));
2202 /* check reserved ino info */
2203 if (le32_to_cpu(raw_super->node_ino) != 1 ||
2204 le32_to_cpu(raw_super->meta_ino) != 2 ||
2205 le32_to_cpu(raw_super->root_ino) != 3) {
2206 f2fs_msg(sb, KERN_INFO,
2207 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2208 le32_to_cpu(raw_super->node_ino),
2209 le32_to_cpu(raw_super->meta_ino),
2210 le32_to_cpu(raw_super->root_ino));
2214 if (le32_to_cpu(raw_super->segment_count) > F2FS_MAX_SEGMENT) {
2215 f2fs_msg(sb, KERN_INFO,
2216 "Invalid segment count (%u)",
2217 le32_to_cpu(raw_super->segment_count));
2221 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2222 if (sanity_check_area_boundary(sbi, bh))
2228 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
2230 unsigned int total, fsmeta;
2231 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2232 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2233 unsigned int ovp_segments, reserved_segments;
2234 unsigned int main_segs, blocks_per_seg;
2235 unsigned int sit_segs, nat_segs;
2236 unsigned int sit_bitmap_size, nat_bitmap_size;
2237 unsigned int log_blocks_per_seg;
2240 total = le32_to_cpu(raw_super->segment_count);
2241 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2242 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2244 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2246 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2247 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2249 if (unlikely(fsmeta >= total))
2252 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2253 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2255 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2256 ovp_segments == 0 || reserved_segments == 0)) {
2257 f2fs_msg(sbi->sb, KERN_ERR,
2258 "Wrong layout: check mkfs.f2fs version");
2262 main_segs = le32_to_cpu(raw_super->segment_count_main);
2263 blocks_per_seg = sbi->blocks_per_seg;
2265 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2266 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2267 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2269 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2270 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2271 le32_to_cpu(ckpt->cur_node_segno[j])) {
2272 f2fs_msg(sbi->sb, KERN_ERR,
2273 "Node segment (%u, %u) has the same "
2275 le32_to_cpu(ckpt->cur_node_segno[i]));
2280 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2281 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2282 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2284 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2285 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2286 le32_to_cpu(ckpt->cur_data_segno[j])) {
2287 f2fs_msg(sbi->sb, KERN_ERR,
2288 "Data segment (%u, %u) has the same "
2290 le32_to_cpu(ckpt->cur_data_segno[i]));
2295 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2296 for (j = i; j < NR_CURSEG_DATA_TYPE; j++) {
2297 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2298 le32_to_cpu(ckpt->cur_data_segno[j])) {
2299 f2fs_msg(sbi->sb, KERN_ERR,
2300 "Data segment (%u) and Data segment (%u)"
2301 " has the same segno: %u", i, j,
2302 le32_to_cpu(ckpt->cur_node_segno[i]));
2308 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2309 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2310 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2312 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2313 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2314 f2fs_msg(sbi->sb, KERN_ERR,
2315 "Wrong bitmap size: sit: %u, nat:%u",
2316 sit_bitmap_size, nat_bitmap_size);
2320 if (unlikely(f2fs_cp_error(sbi))) {
2321 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
2327 static void init_sb_info(struct f2fs_sb_info *sbi)
2329 struct f2fs_super_block *raw_super = sbi->raw_super;
2332 sbi->log_sectors_per_block =
2333 le32_to_cpu(raw_super->log_sectors_per_block);
2334 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2335 sbi->blocksize = 1 << sbi->log_blocksize;
2336 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2337 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2338 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2339 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2340 sbi->total_sections = le32_to_cpu(raw_super->section_count);
2341 sbi->total_node_count =
2342 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2343 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2344 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2345 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2346 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2347 sbi->cur_victim_sec = NULL_SECNO;
2348 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2350 sbi->dir_level = DEF_DIR_LEVEL;
2351 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2352 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2353 clear_sbi_flag(sbi, SBI_NEED_FSCK);
2355 for (i = 0; i < NR_COUNT_TYPE; i++)
2356 atomic_set(&sbi->nr_pages[i], 0);
2358 atomic_set(&sbi->wb_sync_req, 0);
2360 INIT_LIST_HEAD(&sbi->s_list);
2361 mutex_init(&sbi->umount_mutex);
2362 for (i = 0; i < NR_PAGE_TYPE - 1; i++)
2363 for (j = HOT; j < NR_TEMP_TYPE; j++)
2364 mutex_init(&sbi->wio_mutex[i][j]);
2365 spin_lock_init(&sbi->cp_lock);
2367 sbi->dirty_device = 0;
2368 spin_lock_init(&sbi->dev_lock);
2370 init_rwsem(&sbi->sb_lock);
2373 static int init_percpu_info(struct f2fs_sb_info *sbi)
2377 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2381 return percpu_counter_init(&sbi->total_valid_inode_count, 0,
2385 #ifdef CONFIG_BLK_DEV_ZONED
2386 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2388 struct block_device *bdev = FDEV(devi).bdev;
2389 sector_t nr_sectors = bdev->bd_part->nr_sects;
2390 sector_t sector = 0;
2391 struct blk_zone *zones;
2392 unsigned int i, nr_zones;
2396 if (!f2fs_sb_has_blkzoned(sbi->sb))
2399 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2400 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2402 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2403 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2404 __ilog2_u32(sbi->blocks_per_blkz))
2406 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2407 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2408 sbi->log_blocks_per_blkz;
2409 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2410 FDEV(devi).nr_blkz++;
2412 FDEV(devi).blkz_type = f2fs_kmalloc(sbi, FDEV(devi).nr_blkz,
2414 if (!FDEV(devi).blkz_type)
2417 #define F2FS_REPORT_NR_ZONES 4096
2419 zones = f2fs_kzalloc(sbi, sizeof(struct blk_zone) *
2420 F2FS_REPORT_NR_ZONES, GFP_KERNEL);
2424 /* Get block zones type */
2425 while (zones && sector < nr_sectors) {
2427 nr_zones = F2FS_REPORT_NR_ZONES;
2428 err = blkdev_report_zones(bdev, sector,
2438 for (i = 0; i < nr_zones; i++) {
2439 FDEV(devi).blkz_type[n] = zones[i].type;
2440 sector += zones[i].len;
2452 * Read f2fs raw super block.
2453 * Because we have two copies of super block, so read both of them
2454 * to get the first valid one. If any one of them is broken, we pass
2455 * them recovery flag back to the caller.
2457 static int read_raw_super_block(struct f2fs_sb_info *sbi,
2458 struct f2fs_super_block **raw_super,
2459 int *valid_super_block, int *recovery)
2461 struct super_block *sb = sbi->sb;
2463 struct buffer_head *bh;
2464 struct f2fs_super_block *super;
2467 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2471 for (block = 0; block < 2; block++) {
2472 bh = sb_bread(sb, block);
2474 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
2480 /* sanity checking of raw super */
2481 if (sanity_check_raw_super(sbi, bh)) {
2482 f2fs_msg(sb, KERN_ERR,
2483 "Can't find valid F2FS filesystem in %dth superblock",
2491 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2493 *valid_super_block = block;
2499 /* Fail to read any one of the superblocks*/
2503 /* No valid superblock */
2512 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
2514 struct buffer_head *bh;
2517 if ((recover && f2fs_readonly(sbi->sb)) ||
2518 bdev_read_only(sbi->sb->s_bdev)) {
2519 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2523 /* write back-up superblock first */
2524 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
2527 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2530 /* if we are in recovery path, skip writing valid superblock */
2534 /* write current valid superblock */
2535 bh = sb_bread(sbi->sb, sbi->valid_super_block);
2538 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2543 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
2545 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2546 unsigned int max_devices = MAX_DEVICES;
2549 /* Initialize single device information */
2550 if (!RDEV(0).path[0]) {
2551 #ifdef CONFIG_BLK_DEV_ZONED
2552 if (!bdev_is_zoned(sbi->sb->s_bdev))
2561 * Initialize multiple devices information, or single
2562 * zoned block device information.
2564 sbi->devs = f2fs_kzalloc(sbi, sizeof(struct f2fs_dev_info) *
2565 max_devices, GFP_KERNEL);
2569 for (i = 0; i < max_devices; i++) {
2571 if (i > 0 && !RDEV(i).path[0])
2574 if (max_devices == 1) {
2575 /* Single zoned block device mount */
2577 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
2578 sbi->sb->s_mode, sbi->sb->s_type);
2580 /* Multi-device mount */
2581 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
2582 FDEV(i).total_segments =
2583 le32_to_cpu(RDEV(i).total_segments);
2585 FDEV(i).start_blk = 0;
2586 FDEV(i).end_blk = FDEV(i).start_blk +
2587 (FDEV(i).total_segments <<
2588 sbi->log_blocks_per_seg) - 1 +
2589 le32_to_cpu(raw_super->segment0_blkaddr);
2591 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
2592 FDEV(i).end_blk = FDEV(i).start_blk +
2593 (FDEV(i).total_segments <<
2594 sbi->log_blocks_per_seg) - 1;
2596 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
2597 sbi->sb->s_mode, sbi->sb->s_type);
2599 if (IS_ERR(FDEV(i).bdev))
2600 return PTR_ERR(FDEV(i).bdev);
2602 /* to release errored devices */
2603 sbi->s_ndevs = i + 1;
2605 #ifdef CONFIG_BLK_DEV_ZONED
2606 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
2607 !f2fs_sb_has_blkzoned(sbi->sb)) {
2608 f2fs_msg(sbi->sb, KERN_ERR,
2609 "Zoned block device feature not enabled\n");
2612 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
2613 if (init_blkz_info(sbi, i)) {
2614 f2fs_msg(sbi->sb, KERN_ERR,
2615 "Failed to initialize F2FS blkzone information");
2618 if (max_devices == 1)
2620 f2fs_msg(sbi->sb, KERN_INFO,
2621 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
2623 FDEV(i).total_segments,
2624 FDEV(i).start_blk, FDEV(i).end_blk,
2625 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
2626 "Host-aware" : "Host-managed");
2630 f2fs_msg(sbi->sb, KERN_INFO,
2631 "Mount Device [%2d]: %20s, %8u, %8x - %8x",
2633 FDEV(i).total_segments,
2634 FDEV(i).start_blk, FDEV(i).end_blk);
2636 f2fs_msg(sbi->sb, KERN_INFO,
2637 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
2641 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
2643 struct f2fs_sm_info *sm_i = SM_I(sbi);
2645 /* adjust parameters according to the volume size */
2646 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
2647 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
2648 sm_i->dcc_info->discard_granularity = 1;
2649 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
2653 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
2655 struct f2fs_sb_info *sbi;
2656 struct f2fs_super_block *raw_super;
2659 bool retry = true, need_fsck = false;
2660 char *options = NULL;
2661 int recovery, i, valid_super_block;
2662 struct curseg_info *seg_i;
2667 valid_super_block = -1;
2670 /* allocate memory for f2fs-specific super block info */
2671 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
2677 /* Load the checksum driver */
2678 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
2679 if (IS_ERR(sbi->s_chksum_driver)) {
2680 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
2681 err = PTR_ERR(sbi->s_chksum_driver);
2682 sbi->s_chksum_driver = NULL;
2686 /* set a block size */
2687 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
2688 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
2692 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
2697 sb->s_fs_info = sbi;
2698 sbi->raw_super = raw_super;
2700 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
2701 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
2703 /* precompute checksum seed for metadata */
2704 if (f2fs_sb_has_inode_chksum(sb))
2705 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
2706 sizeof(raw_super->uuid));
2709 * The BLKZONED feature indicates that the drive was formatted with
2710 * zone alignment optimization. This is optional for host-aware
2711 * devices, but mandatory for host-managed zoned block devices.
2713 #ifndef CONFIG_BLK_DEV_ZONED
2714 if (f2fs_sb_has_blkzoned(sb)) {
2715 f2fs_msg(sb, KERN_ERR,
2716 "Zoned block device support is not enabled\n");
2721 default_options(sbi);
2722 /* parse mount options */
2723 options = kstrdup((const char *)data, GFP_KERNEL);
2724 if (data && !options) {
2729 err = parse_options(sb, options);
2733 sbi->max_file_blocks = max_file_blocks();
2734 sb->s_maxbytes = sbi->max_file_blocks <<
2735 le32_to_cpu(raw_super->log_blocksize);
2736 sb->s_max_links = F2FS_LINK_MAX;
2737 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2740 sb->dq_op = &f2fs_quota_operations;
2741 if (f2fs_sb_has_quota_ino(sb))
2742 sb->s_qcop = &dquot_quotactl_sysfile_ops;
2744 sb->s_qcop = &f2fs_quotactl_ops;
2745 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
2747 if (f2fs_sb_has_quota_ino(sbi->sb)) {
2748 for (i = 0; i < MAXQUOTAS; i++) {
2749 if (f2fs_qf_ino(sbi->sb, i))
2750 sbi->nquota_files++;
2755 sb->s_op = &f2fs_sops;
2756 #ifdef CONFIG_F2FS_FS_ENCRYPTION
2757 sb->s_cop = &f2fs_cryptops;
2759 sb->s_xattr = f2fs_xattr_handlers;
2760 sb->s_export_op = &f2fs_export_ops;
2761 sb->s_magic = F2FS_SUPER_MAGIC;
2762 sb->s_time_gran = 1;
2763 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2764 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
2765 memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
2766 sb->s_iflags |= SB_I_CGROUPWB;
2768 /* init f2fs-specific super block info */
2769 sbi->valid_super_block = valid_super_block;
2770 mutex_init(&sbi->gc_mutex);
2771 mutex_init(&sbi->cp_mutex);
2772 init_rwsem(&sbi->node_write);
2773 init_rwsem(&sbi->node_change);
2775 /* disallow all the data/node/meta page writes */
2776 set_sbi_flag(sbi, SBI_POR_DOING);
2777 spin_lock_init(&sbi->stat_lock);
2779 /* init iostat info */
2780 spin_lock_init(&sbi->iostat_lock);
2781 sbi->iostat_enable = false;
2783 for (i = 0; i < NR_PAGE_TYPE; i++) {
2784 int n = (i == META) ? 1: NR_TEMP_TYPE;
2787 sbi->write_io[i] = f2fs_kmalloc(sbi,
2788 n * sizeof(struct f2fs_bio_info),
2790 if (!sbi->write_io[i]) {
2795 for (j = HOT; j < n; j++) {
2796 init_rwsem(&sbi->write_io[i][j].io_rwsem);
2797 sbi->write_io[i][j].sbi = sbi;
2798 sbi->write_io[i][j].bio = NULL;
2799 spin_lock_init(&sbi->write_io[i][j].io_lock);
2800 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
2804 init_rwsem(&sbi->cp_rwsem);
2805 init_waitqueue_head(&sbi->cp_wait);
2808 err = init_percpu_info(sbi);
2812 if (F2FS_IO_SIZE(sbi) > 1) {
2813 sbi->write_io_dummy =
2814 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
2815 if (!sbi->write_io_dummy) {
2821 /* get an inode for meta space */
2822 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
2823 if (IS_ERR(sbi->meta_inode)) {
2824 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
2825 err = PTR_ERR(sbi->meta_inode);
2829 err = get_valid_checkpoint(sbi);
2831 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
2832 goto free_meta_inode;
2835 /* Initialize device list */
2836 err = f2fs_scan_devices(sbi);
2838 f2fs_msg(sb, KERN_ERR, "Failed to find devices");
2842 sbi->total_valid_node_count =
2843 le32_to_cpu(sbi->ckpt->valid_node_count);
2844 percpu_counter_set(&sbi->total_valid_inode_count,
2845 le32_to_cpu(sbi->ckpt->valid_inode_count));
2846 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
2847 sbi->total_valid_block_count =
2848 le64_to_cpu(sbi->ckpt->valid_block_count);
2849 sbi->last_valid_block_count = sbi->total_valid_block_count;
2850 sbi->reserved_blocks = 0;
2851 sbi->current_reserved_blocks = 0;
2852 limit_reserve_root(sbi);
2854 for (i = 0; i < NR_INODE_TYPE; i++) {
2855 INIT_LIST_HEAD(&sbi->inode_list[i]);
2856 spin_lock_init(&sbi->inode_lock[i]);
2859 init_extent_cache_info(sbi);
2861 init_ino_entry_info(sbi);
2863 /* setup f2fs internal modules */
2864 err = build_segment_manager(sbi);
2866 f2fs_msg(sb, KERN_ERR,
2867 "Failed to initialize F2FS segment manager");
2870 err = build_node_manager(sbi);
2872 f2fs_msg(sb, KERN_ERR,
2873 "Failed to initialize F2FS node manager");
2877 /* For write statistics */
2878 if (sb->s_bdev->bd_part)
2879 sbi->sectors_written_start =
2880 (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
2882 /* Read accumulated write IO statistics if exists */
2883 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
2884 if (__exist_node_summaries(sbi))
2885 sbi->kbytes_written =
2886 le64_to_cpu(seg_i->journal->info.kbytes_written);
2888 build_gc_manager(sbi);
2890 /* get an inode for node space */
2891 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
2892 if (IS_ERR(sbi->node_inode)) {
2893 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
2894 err = PTR_ERR(sbi->node_inode);
2898 err = f2fs_build_stats(sbi);
2900 goto free_node_inode;
2902 /* read root inode and dentry */
2903 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
2905 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
2906 err = PTR_ERR(root);
2909 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2912 goto free_node_inode;
2915 sb->s_root = d_make_root(root); /* allocate root dentry */
2918 goto free_root_inode;
2921 err = f2fs_register_sysfs(sbi);
2923 goto free_root_inode;
2927 * Turn on quotas which were not enabled for read-only mounts if
2928 * filesystem has quota feature, so that they are updated correctly.
2930 if (f2fs_sb_has_quota_ino(sb) && !f2fs_readonly(sb)) {
2931 err = f2fs_enable_quotas(sb);
2933 f2fs_msg(sb, KERN_ERR,
2934 "Cannot turn on quotas: error %d", err);
2939 /* if there are nt orphan nodes free them */
2940 err = recover_orphan_inodes(sbi);
2944 /* recover fsynced data */
2945 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
2947 * mount should be failed, when device has readonly mode, and
2948 * previous checkpoint was not done by clean system shutdown.
2950 if (bdev_read_only(sb->s_bdev) &&
2951 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
2957 set_sbi_flag(sbi, SBI_NEED_FSCK);
2962 err = recover_fsync_data(sbi, false);
2965 f2fs_msg(sb, KERN_ERR,
2966 "Cannot recover all fsync data errno=%d", err);
2970 err = recover_fsync_data(sbi, true);
2972 if (!f2fs_readonly(sb) && err > 0) {
2974 f2fs_msg(sb, KERN_ERR,
2975 "Need to recover fsync data");
2980 /* recover_fsync_data() cleared this already */
2981 clear_sbi_flag(sbi, SBI_POR_DOING);
2984 * If filesystem is not mounted as read-only then
2985 * do start the gc_thread.
2987 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
2988 /* After POR, we can run background GC thread.*/
2989 err = start_gc_thread(sbi);
2995 /* recover broken superblock */
2997 err = f2fs_commit_super(sbi, true);
2998 f2fs_msg(sb, KERN_INFO,
2999 "Try to recover %dth superblock, ret: %d",
3000 sbi->valid_super_block ? 1 : 2, err);
3003 f2fs_join_shrinker(sbi);
3005 f2fs_tuning_parameters(sbi);
3007 f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx",
3008 cur_cp_version(F2FS_CKPT(sbi)));
3009 f2fs_update_time(sbi, CP_TIME);
3010 f2fs_update_time(sbi, REQ_TIME);
3015 if (f2fs_sb_has_quota_ino(sb) && !f2fs_readonly(sb))
3016 f2fs_quota_off_umount(sbi->sb);
3018 f2fs_sync_inode_meta(sbi);
3020 * Some dirty meta pages can be produced by recover_orphan_inodes()
3021 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3022 * followed by write_checkpoint() through f2fs_write_node_pages(), which
3023 * falls into an infinite loop in sync_meta_pages().
3025 truncate_inode_pages_final(META_MAPPING(sbi));
3029 f2fs_unregister_sysfs(sbi);
3034 f2fs_destroy_stats(sbi);
3036 release_ino_entry(sbi, true);
3037 truncate_inode_pages_final(NODE_MAPPING(sbi));
3038 iput(sbi->node_inode);
3040 destroy_node_manager(sbi);
3042 destroy_segment_manager(sbi);
3044 destroy_device_list(sbi);
3047 make_bad_inode(sbi->meta_inode);
3048 iput(sbi->meta_inode);
3050 mempool_destroy(sbi->write_io_dummy);
3052 destroy_percpu_info(sbi);
3054 for (i = 0; i < NR_PAGE_TYPE; i++)
3055 kfree(sbi->write_io[i]);
3058 for (i = 0; i < MAXQUOTAS; i++)
3059 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
3065 if (sbi->s_chksum_driver)
3066 crypto_free_shash(sbi->s_chksum_driver);
3069 /* give only one another chance */
3072 shrink_dcache_sb(sb);
3078 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3079 const char *dev_name, void *data)
3081 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3084 static void kill_f2fs_super(struct super_block *sb)
3087 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
3088 stop_gc_thread(F2FS_SB(sb));
3089 stop_discard_thread(F2FS_SB(sb));
3091 kill_block_super(sb);
3094 static struct file_system_type f2fs_fs_type = {
3095 .owner = THIS_MODULE,
3097 .mount = f2fs_mount,
3098 .kill_sb = kill_f2fs_super,
3099 .fs_flags = FS_REQUIRES_DEV,
3101 MODULE_ALIAS_FS("f2fs");
3103 static int __init init_inodecache(void)
3105 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3106 sizeof(struct f2fs_inode_info), 0,
3107 SLAB_RECLAIM_ACCOUNT, NULL);
3108 if (!f2fs_inode_cachep)
3113 static void destroy_inodecache(void)
3116 * Make sure all delayed rcu free inodes are flushed before we
3120 kmem_cache_destroy(f2fs_inode_cachep);
3123 static int __init init_f2fs_fs(void)
3127 if (PAGE_SIZE != F2FS_BLKSIZE) {
3128 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3129 PAGE_SIZE, F2FS_BLKSIZE);
3133 f2fs_build_trace_ios();
3135 err = init_inodecache();
3138 err = create_node_manager_caches();
3140 goto free_inodecache;
3141 err = create_segment_manager_caches();
3143 goto free_node_manager_caches;
3144 err = create_checkpoint_caches();
3146 goto free_segment_manager_caches;
3147 err = create_extent_cache();
3149 goto free_checkpoint_caches;
3150 err = f2fs_init_sysfs();
3152 goto free_extent_cache;
3153 err = register_shrinker(&f2fs_shrinker_info);
3156 err = register_filesystem(&f2fs_fs_type);
3159 err = f2fs_create_root_stats();
3161 goto free_filesystem;
3162 err = f2fs_init_post_read_processing();
3164 goto free_root_stats;
3168 f2fs_destroy_root_stats();
3170 unregister_filesystem(&f2fs_fs_type);
3172 unregister_shrinker(&f2fs_shrinker_info);
3176 destroy_extent_cache();
3177 free_checkpoint_caches:
3178 destroy_checkpoint_caches();
3179 free_segment_manager_caches:
3180 destroy_segment_manager_caches();
3181 free_node_manager_caches:
3182 destroy_node_manager_caches();
3184 destroy_inodecache();
3189 static void __exit exit_f2fs_fs(void)
3191 f2fs_destroy_post_read_processing();
3192 f2fs_destroy_root_stats();
3193 unregister_filesystem(&f2fs_fs_type);
3194 unregister_shrinker(&f2fs_shrinker_info);
3196 destroy_extent_cache();
3197 destroy_checkpoint_caches();
3198 destroy_segment_manager_caches();
3199 destroy_node_manager_caches();
3200 destroy_inodecache();
3201 f2fs_destroy_trace_ios();
3204 module_init(init_f2fs_fs)
3205 module_exit(exit_f2fs_fs)
3207 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3208 MODULE_DESCRIPTION("Flash Friendly File System");
3209 MODULE_LICENSE("GPL");