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 = &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 static match_table_t f2fs_tokens = {
136 {Opt_gc_background, "background_gc=%s"},
137 {Opt_disable_roll_forward, "disable_roll_forward"},
138 {Opt_norecovery, "norecovery"},
139 {Opt_discard, "discard"},
140 {Opt_nodiscard, "nodiscard"},
141 {Opt_noheap, "no_heap"},
143 {Opt_user_xattr, "user_xattr"},
144 {Opt_nouser_xattr, "nouser_xattr"},
146 {Opt_noacl, "noacl"},
147 {Opt_active_logs, "active_logs=%u"},
148 {Opt_disable_ext_identify, "disable_ext_identify"},
149 {Opt_inline_xattr, "inline_xattr"},
150 {Opt_noinline_xattr, "noinline_xattr"},
151 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
152 {Opt_inline_data, "inline_data"},
153 {Opt_inline_dentry, "inline_dentry"},
154 {Opt_noinline_dentry, "noinline_dentry"},
155 {Opt_flush_merge, "flush_merge"},
156 {Opt_noflush_merge, "noflush_merge"},
157 {Opt_nobarrier, "nobarrier"},
158 {Opt_fastboot, "fastboot"},
159 {Opt_extent_cache, "extent_cache"},
160 {Opt_noextent_cache, "noextent_cache"},
161 {Opt_noinline_data, "noinline_data"},
162 {Opt_data_flush, "data_flush"},
163 {Opt_reserve_root, "reserve_root=%u"},
164 {Opt_resgid, "resgid=%u"},
165 {Opt_resuid, "resuid=%u"},
166 {Opt_mode, "mode=%s"},
167 {Opt_io_size_bits, "io_bits=%u"},
168 {Opt_fault_injection, "fault_injection=%u"},
169 {Opt_lazytime, "lazytime"},
170 {Opt_nolazytime, "nolazytime"},
171 {Opt_quota, "quota"},
172 {Opt_noquota, "noquota"},
173 {Opt_usrquota, "usrquota"},
174 {Opt_grpquota, "grpquota"},
175 {Opt_prjquota, "prjquota"},
176 {Opt_usrjquota, "usrjquota=%s"},
177 {Opt_grpjquota, "grpjquota=%s"},
178 {Opt_prjjquota, "prjjquota=%s"},
179 {Opt_offusrjquota, "usrjquota="},
180 {Opt_offgrpjquota, "grpjquota="},
181 {Opt_offprjjquota, "prjjquota="},
182 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
183 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
184 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
188 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
190 struct va_format vaf;
196 printk_ratelimited("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
200 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
202 block_t limit = (sbi->user_block_count << 1) / 1000;
205 if (test_opt(sbi, RESERVE_ROOT) && sbi->root_reserved_blocks > limit) {
206 sbi->root_reserved_blocks = limit;
207 f2fs_msg(sbi->sb, KERN_INFO,
208 "Reduce reserved blocks for root = %u",
209 sbi->root_reserved_blocks);
211 if (!test_opt(sbi, RESERVE_ROOT) &&
212 (!uid_eq(sbi->s_resuid,
213 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
214 !gid_eq(sbi->s_resgid,
215 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
216 f2fs_msg(sbi->sb, KERN_INFO,
217 "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
218 from_kuid_munged(&init_user_ns, sbi->s_resuid),
219 from_kgid_munged(&init_user_ns, sbi->s_resgid));
222 static void init_once(void *foo)
224 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
226 inode_init_once(&fi->vfs_inode);
230 static const char * const quotatypes[] = INITQFNAMES;
231 #define QTYPE2NAME(t) (quotatypes[t])
232 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
235 struct f2fs_sb_info *sbi = F2FS_SB(sb);
239 if (sb_any_quota_loaded(sb) && !sbi->s_qf_names[qtype]) {
240 f2fs_msg(sb, KERN_ERR,
241 "Cannot change journaled "
242 "quota options when quota turned on");
245 if (f2fs_sb_has_quota_ino(sb)) {
246 f2fs_msg(sb, KERN_INFO,
247 "QUOTA feature is enabled, so ignore qf_name");
251 qname = match_strdup(args);
253 f2fs_msg(sb, KERN_ERR,
254 "Not enough memory for storing quotafile name");
257 if (sbi->s_qf_names[qtype]) {
258 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
261 f2fs_msg(sb, KERN_ERR,
262 "%s quota file already specified",
266 if (strchr(qname, '/')) {
267 f2fs_msg(sb, KERN_ERR,
268 "quotafile must be on filesystem root");
271 sbi->s_qf_names[qtype] = qname;
279 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
281 struct f2fs_sb_info *sbi = F2FS_SB(sb);
283 if (sb_any_quota_loaded(sb) && sbi->s_qf_names[qtype]) {
284 f2fs_msg(sb, KERN_ERR, "Cannot change journaled quota options"
285 " when quota turned on");
288 kfree(sbi->s_qf_names[qtype]);
289 sbi->s_qf_names[qtype] = NULL;
293 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
296 * We do the test below only for project quotas. 'usrquota' and
297 * 'grpquota' mount options are allowed even without quota feature
298 * to support legacy quotas in quota files.
300 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi->sb)) {
301 f2fs_msg(sbi->sb, KERN_ERR, "Project quota feature not enabled. "
302 "Cannot enable project quota enforcement.");
305 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA] ||
306 sbi->s_qf_names[PRJQUOTA]) {
307 if (test_opt(sbi, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
308 clear_opt(sbi, USRQUOTA);
310 if (test_opt(sbi, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
311 clear_opt(sbi, GRPQUOTA);
313 if (test_opt(sbi, PRJQUOTA) && sbi->s_qf_names[PRJQUOTA])
314 clear_opt(sbi, PRJQUOTA);
316 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
317 test_opt(sbi, PRJQUOTA)) {
318 f2fs_msg(sbi->sb, KERN_ERR, "old and new quota "
323 if (!sbi->s_jquota_fmt) {
324 f2fs_msg(sbi->sb, KERN_ERR, "journaled quota format "
330 if (f2fs_sb_has_quota_ino(sbi->sb) && sbi->s_jquota_fmt) {
331 f2fs_msg(sbi->sb, KERN_INFO,
332 "QUOTA feature is enabled, so ignore jquota_fmt");
333 sbi->s_jquota_fmt = 0;
335 if (f2fs_sb_has_quota_ino(sbi->sb) && sb_rdonly(sbi->sb)) {
336 f2fs_msg(sbi->sb, KERN_INFO,
337 "Filesystem with quota feature cannot be mounted RDWR "
338 "without CONFIG_QUOTA");
345 static int parse_options(struct super_block *sb, char *options)
347 struct f2fs_sb_info *sbi = F2FS_SB(sb);
348 struct request_queue *q;
349 substring_t args[MAX_OPT_ARGS];
361 while ((p = strsep(&options, ",")) != NULL) {
366 * Initialize args struct so we know whether arg was
367 * found; some options take optional arguments.
369 args[0].to = args[0].from = NULL;
370 token = match_token(p, f2fs_tokens, args);
373 case Opt_gc_background:
374 name = match_strdup(&args[0]);
378 if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
380 clear_opt(sbi, FORCE_FG_GC);
381 } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
382 clear_opt(sbi, BG_GC);
383 clear_opt(sbi, FORCE_FG_GC);
384 } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
386 set_opt(sbi, FORCE_FG_GC);
393 case Opt_disable_roll_forward:
394 set_opt(sbi, DISABLE_ROLL_FORWARD);
397 /* this option mounts f2fs with ro */
398 set_opt(sbi, DISABLE_ROLL_FORWARD);
399 if (!f2fs_readonly(sb))
403 q = bdev_get_queue(sb->s_bdev);
404 if (blk_queue_discard(q)) {
405 set_opt(sbi, DISCARD);
406 } else if (!f2fs_sb_mounted_blkzoned(sb)) {
407 f2fs_msg(sb, KERN_WARNING,
408 "mounting with \"discard\" option, but "
409 "the device does not support discard");
413 if (f2fs_sb_mounted_blkzoned(sb)) {
414 f2fs_msg(sb, KERN_WARNING,
415 "discard is required for zoned block devices");
418 clear_opt(sbi, DISCARD);
421 set_opt(sbi, NOHEAP);
424 clear_opt(sbi, NOHEAP);
426 #ifdef CONFIG_F2FS_FS_XATTR
428 set_opt(sbi, XATTR_USER);
430 case Opt_nouser_xattr:
431 clear_opt(sbi, XATTR_USER);
433 case Opt_inline_xattr:
434 set_opt(sbi, INLINE_XATTR);
436 case Opt_noinline_xattr:
437 clear_opt(sbi, INLINE_XATTR);
439 case Opt_inline_xattr_size:
440 if (args->from && match_int(args, &arg))
442 set_opt(sbi, INLINE_XATTR_SIZE);
443 sbi->inline_xattr_size = arg;
447 f2fs_msg(sb, KERN_INFO,
448 "user_xattr options not supported");
450 case Opt_nouser_xattr:
451 f2fs_msg(sb, KERN_INFO,
452 "nouser_xattr options not supported");
454 case Opt_inline_xattr:
455 f2fs_msg(sb, KERN_INFO,
456 "inline_xattr options not supported");
458 case Opt_noinline_xattr:
459 f2fs_msg(sb, KERN_INFO,
460 "noinline_xattr options not supported");
463 #ifdef CONFIG_F2FS_FS_POSIX_ACL
465 set_opt(sbi, POSIX_ACL);
468 clear_opt(sbi, POSIX_ACL);
472 f2fs_msg(sb, KERN_INFO, "acl options not supported");
475 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
478 case Opt_active_logs:
479 if (args->from && match_int(args, &arg))
481 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
483 sbi->active_logs = arg;
485 case Opt_disable_ext_identify:
486 set_opt(sbi, DISABLE_EXT_IDENTIFY);
488 case Opt_inline_data:
489 set_opt(sbi, INLINE_DATA);
491 case Opt_inline_dentry:
492 set_opt(sbi, INLINE_DENTRY);
494 case Opt_noinline_dentry:
495 clear_opt(sbi, INLINE_DENTRY);
497 case Opt_flush_merge:
498 set_opt(sbi, FLUSH_MERGE);
500 case Opt_noflush_merge:
501 clear_opt(sbi, FLUSH_MERGE);
504 set_opt(sbi, NOBARRIER);
507 set_opt(sbi, FASTBOOT);
509 case Opt_extent_cache:
510 set_opt(sbi, EXTENT_CACHE);
512 case Opt_noextent_cache:
513 clear_opt(sbi, EXTENT_CACHE);
515 case Opt_noinline_data:
516 clear_opt(sbi, INLINE_DATA);
519 set_opt(sbi, DATA_FLUSH);
521 case Opt_reserve_root:
522 if (args->from && match_int(args, &arg))
524 if (test_opt(sbi, RESERVE_ROOT)) {
525 f2fs_msg(sb, KERN_INFO,
526 "Preserve previous reserve_root=%u",
527 sbi->root_reserved_blocks);
529 sbi->root_reserved_blocks = arg;
530 set_opt(sbi, RESERVE_ROOT);
534 if (args->from && match_int(args, &arg))
536 uid = make_kuid(current_user_ns(), arg);
537 if (!uid_valid(uid)) {
538 f2fs_msg(sb, KERN_ERR,
539 "Invalid uid value %d", arg);
545 if (args->from && match_int(args, &arg))
547 gid = make_kgid(current_user_ns(), arg);
548 if (!gid_valid(gid)) {
549 f2fs_msg(sb, KERN_ERR,
550 "Invalid gid value %d", arg);
556 name = match_strdup(&args[0]);
560 if (strlen(name) == 8 &&
561 !strncmp(name, "adaptive", 8)) {
562 if (f2fs_sb_mounted_blkzoned(sb)) {
563 f2fs_msg(sb, KERN_WARNING,
564 "adaptive mode is not allowed with "
565 "zoned block device feature");
569 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
570 } else if (strlen(name) == 3 &&
571 !strncmp(name, "lfs", 3)) {
572 set_opt_mode(sbi, F2FS_MOUNT_LFS);
579 case Opt_io_size_bits:
580 if (args->from && match_int(args, &arg))
582 if (arg > __ilog2_u32(BIO_MAX_PAGES)) {
583 f2fs_msg(sb, KERN_WARNING,
584 "Not support %d, larger than %d",
585 1 << arg, BIO_MAX_PAGES);
588 sbi->write_io_size_bits = arg;
590 case Opt_fault_injection:
591 if (args->from && match_int(args, &arg))
593 #ifdef CONFIG_F2FS_FAULT_INJECTION
594 f2fs_build_fault_attr(sbi, arg);
595 set_opt(sbi, FAULT_INJECTION);
597 f2fs_msg(sb, KERN_INFO,
598 "FAULT_INJECTION was not selected");
602 sb->s_flags |= SB_LAZYTIME;
605 sb->s_flags &= ~SB_LAZYTIME;
610 set_opt(sbi, USRQUOTA);
613 set_opt(sbi, GRPQUOTA);
616 set_opt(sbi, PRJQUOTA);
619 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
624 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
629 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
633 case Opt_offusrjquota:
634 ret = f2fs_clear_qf_name(sb, USRQUOTA);
638 case Opt_offgrpjquota:
639 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
643 case Opt_offprjjquota:
644 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
648 case Opt_jqfmt_vfsold:
649 sbi->s_jquota_fmt = QFMT_VFS_OLD;
651 case Opt_jqfmt_vfsv0:
652 sbi->s_jquota_fmt = QFMT_VFS_V0;
654 case Opt_jqfmt_vfsv1:
655 sbi->s_jquota_fmt = QFMT_VFS_V1;
658 clear_opt(sbi, QUOTA);
659 clear_opt(sbi, USRQUOTA);
660 clear_opt(sbi, GRPQUOTA);
661 clear_opt(sbi, PRJQUOTA);
671 case Opt_offusrjquota:
672 case Opt_offgrpjquota:
673 case Opt_offprjjquota:
674 case Opt_jqfmt_vfsold:
675 case Opt_jqfmt_vfsv0:
676 case Opt_jqfmt_vfsv1:
678 f2fs_msg(sb, KERN_INFO,
679 "quota operations not supported");
683 f2fs_msg(sb, KERN_ERR,
684 "Unrecognized mount option \"%s\" or missing value",
690 if (f2fs_check_quota_options(sbi))
694 if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) {
695 f2fs_msg(sb, KERN_ERR,
696 "Should set mode=lfs with %uKB-sized IO",
697 F2FS_IO_SIZE_KB(sbi));
701 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
702 if (!test_opt(sbi, INLINE_XATTR)) {
703 f2fs_msg(sb, KERN_ERR,
704 "inline_xattr_size option should be "
705 "set with inline_xattr option");
708 if (!sbi->inline_xattr_size ||
709 sbi->inline_xattr_size >= DEF_ADDRS_PER_INODE -
710 F2FS_TOTAL_EXTRA_ATTR_SIZE -
711 DEF_INLINE_RESERVED_SIZE -
712 DEF_MIN_INLINE_SIZE) {
713 f2fs_msg(sb, KERN_ERR,
714 "inline xattr size is out of range");
721 static struct inode *f2fs_alloc_inode(struct super_block *sb)
723 struct f2fs_inode_info *fi;
725 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
729 init_once((void *) fi);
731 /* Initialize f2fs-specific inode info */
732 atomic_set(&fi->dirty_pages, 0);
733 fi->i_current_depth = 1;
735 init_rwsem(&fi->i_sem);
736 INIT_LIST_HEAD(&fi->dirty_list);
737 INIT_LIST_HEAD(&fi->gdirty_list);
738 INIT_LIST_HEAD(&fi->inmem_ilist);
739 INIT_LIST_HEAD(&fi->inmem_pages);
740 mutex_init(&fi->inmem_lock);
741 init_rwsem(&fi->dio_rwsem[READ]);
742 init_rwsem(&fi->dio_rwsem[WRITE]);
743 init_rwsem(&fi->i_mmap_sem);
744 init_rwsem(&fi->i_xattr_sem);
747 memset(&fi->i_dquot, 0, sizeof(fi->i_dquot));
748 fi->i_reserved_quota = 0;
750 /* Will be used by directory only */
751 fi->i_dir_level = F2FS_SB(sb)->dir_level;
753 return &fi->vfs_inode;
756 static int f2fs_drop_inode(struct inode *inode)
760 * This is to avoid a deadlock condition like below.
761 * writeback_single_inode(inode)
762 * - f2fs_write_data_page
763 * - f2fs_gc -> iput -> evict
764 * - inode_wait_for_writeback(inode)
766 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
767 if (!inode->i_nlink && !is_bad_inode(inode)) {
768 /* to avoid evict_inode call simultaneously */
769 atomic_inc(&inode->i_count);
770 spin_unlock(&inode->i_lock);
772 /* some remained atomic pages should discarded */
773 if (f2fs_is_atomic_file(inode))
774 drop_inmem_pages(inode);
776 /* should remain fi->extent_tree for writepage */
777 f2fs_destroy_extent_node(inode);
779 sb_start_intwrite(inode->i_sb);
780 f2fs_i_size_write(inode, 0);
782 if (F2FS_HAS_BLOCKS(inode))
783 f2fs_truncate(inode);
785 sb_end_intwrite(inode->i_sb);
787 spin_lock(&inode->i_lock);
788 atomic_dec(&inode->i_count);
790 trace_f2fs_drop_inode(inode, 0);
793 ret = generic_drop_inode(inode);
794 trace_f2fs_drop_inode(inode, ret);
798 int f2fs_inode_dirtied(struct inode *inode, bool sync)
800 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
803 spin_lock(&sbi->inode_lock[DIRTY_META]);
804 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
807 set_inode_flag(inode, FI_DIRTY_INODE);
808 stat_inc_dirty_inode(sbi, DIRTY_META);
810 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
811 list_add_tail(&F2FS_I(inode)->gdirty_list,
812 &sbi->inode_list[DIRTY_META]);
813 inc_page_count(sbi, F2FS_DIRTY_IMETA);
815 spin_unlock(&sbi->inode_lock[DIRTY_META]);
819 void f2fs_inode_synced(struct inode *inode)
821 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
823 spin_lock(&sbi->inode_lock[DIRTY_META]);
824 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
825 spin_unlock(&sbi->inode_lock[DIRTY_META]);
828 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
829 list_del_init(&F2FS_I(inode)->gdirty_list);
830 dec_page_count(sbi, F2FS_DIRTY_IMETA);
832 clear_inode_flag(inode, FI_DIRTY_INODE);
833 clear_inode_flag(inode, FI_AUTO_RECOVER);
834 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
835 spin_unlock(&sbi->inode_lock[DIRTY_META]);
839 * f2fs_dirty_inode() is called from __mark_inode_dirty()
841 * We should call set_dirty_inode to write the dirty inode through write_inode.
843 static void f2fs_dirty_inode(struct inode *inode, int flags)
845 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
847 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
848 inode->i_ino == F2FS_META_INO(sbi))
851 if (flags == I_DIRTY_TIME)
854 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
855 clear_inode_flag(inode, FI_AUTO_RECOVER);
857 f2fs_inode_dirtied(inode, false);
860 static void f2fs_i_callback(struct rcu_head *head)
862 struct inode *inode = container_of(head, struct inode, i_rcu);
863 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
866 static void f2fs_destroy_inode(struct inode *inode)
868 call_rcu(&inode->i_rcu, f2fs_i_callback);
871 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
873 percpu_counter_destroy(&sbi->alloc_valid_block_count);
874 percpu_counter_destroy(&sbi->total_valid_inode_count);
877 static void destroy_device_list(struct f2fs_sb_info *sbi)
881 for (i = 0; i < sbi->s_ndevs; i++) {
882 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
883 #ifdef CONFIG_BLK_DEV_ZONED
884 kfree(FDEV(i).blkz_type);
890 static void f2fs_put_super(struct super_block *sb)
892 struct f2fs_sb_info *sbi = F2FS_SB(sb);
896 f2fs_quota_off_umount(sb);
898 /* prevent remaining shrinker jobs */
899 mutex_lock(&sbi->umount_mutex);
902 * We don't need to do checkpoint when superblock is clean.
903 * But, the previous checkpoint was not done by umount, it needs to do
904 * clean checkpoint again.
906 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
907 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
908 struct cp_control cpc = {
911 write_checkpoint(sbi, &cpc);
914 /* be sure to wait for any on-going discard commands */
915 dropped = f2fs_wait_discard_bios(sbi);
917 if (f2fs_discard_en(sbi) && !sbi->discard_blks && !dropped) {
918 struct cp_control cpc = {
919 .reason = CP_UMOUNT | CP_TRIMMED,
921 write_checkpoint(sbi, &cpc);
924 /* write_checkpoint can update stat informaion */
925 f2fs_destroy_stats(sbi);
928 * normally superblock is clean, so we need to release this.
929 * In addition, EIO will skip do checkpoint, we need this as well.
931 release_ino_entry(sbi, true);
933 f2fs_leave_shrinker(sbi);
934 mutex_unlock(&sbi->umount_mutex);
936 /* our cp_error case, we can wait for any writeback page */
937 f2fs_flush_merged_writes(sbi);
939 iput(sbi->node_inode);
940 iput(sbi->meta_inode);
942 /* destroy f2fs internal modules */
943 destroy_node_manager(sbi);
944 destroy_segment_manager(sbi);
948 f2fs_unregister_sysfs(sbi);
950 sb->s_fs_info = NULL;
951 if (sbi->s_chksum_driver)
952 crypto_free_shash(sbi->s_chksum_driver);
953 kfree(sbi->raw_super);
955 destroy_device_list(sbi);
956 mempool_destroy(sbi->write_io_dummy);
958 for (i = 0; i < MAXQUOTAS; i++)
959 kfree(sbi->s_qf_names[i]);
961 destroy_percpu_info(sbi);
962 for (i = 0; i < NR_PAGE_TYPE; i++)
963 kfree(sbi->write_io[i]);
967 int f2fs_sync_fs(struct super_block *sb, int sync)
969 struct f2fs_sb_info *sbi = F2FS_SB(sb);
972 if (unlikely(f2fs_cp_error(sbi)))
975 trace_f2fs_sync_fs(sb, sync);
977 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
981 struct cp_control cpc;
983 cpc.reason = __get_cp_reason(sbi);
985 mutex_lock(&sbi->gc_mutex);
986 err = write_checkpoint(sbi, &cpc);
987 mutex_unlock(&sbi->gc_mutex);
989 f2fs_trace_ios(NULL, 1);
994 static int f2fs_freeze(struct super_block *sb)
996 if (f2fs_readonly(sb))
999 /* IO error happened before */
1000 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1003 /* must be clean, since sync_filesystem() was already called */
1004 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1009 static int f2fs_unfreeze(struct super_block *sb)
1015 static int f2fs_statfs_project(struct super_block *sb,
1016 kprojid_t projid, struct kstatfs *buf)
1019 struct dquot *dquot;
1023 qid = make_kqid_projid(projid);
1024 dquot = dqget(sb, qid);
1026 return PTR_ERR(dquot);
1027 spin_lock(&dq_data_lock);
1029 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
1030 dquot->dq_dqb.dqb_bsoftlimit :
1031 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
1032 if (limit && buf->f_blocks > limit) {
1033 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1034 buf->f_blocks = limit;
1035 buf->f_bfree = buf->f_bavail =
1036 (buf->f_blocks > curblock) ?
1037 (buf->f_blocks - curblock) : 0;
1040 limit = dquot->dq_dqb.dqb_isoftlimit ?
1041 dquot->dq_dqb.dqb_isoftlimit :
1042 dquot->dq_dqb.dqb_ihardlimit;
1043 if (limit && buf->f_files > limit) {
1044 buf->f_files = limit;
1046 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1047 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1050 spin_unlock(&dq_data_lock);
1056 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1058 struct super_block *sb = dentry->d_sb;
1059 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1060 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1061 block_t total_count, user_block_count, start_count;
1062 u64 avail_node_count;
1064 total_count = le64_to_cpu(sbi->raw_super->block_count);
1065 user_block_count = sbi->user_block_count;
1066 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1067 buf->f_type = F2FS_SUPER_MAGIC;
1068 buf->f_bsize = sbi->blocksize;
1070 buf->f_blocks = total_count - start_count;
1071 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1072 sbi->current_reserved_blocks;
1073 if (buf->f_bfree > sbi->root_reserved_blocks)
1074 buf->f_bavail = buf->f_bfree - sbi->root_reserved_blocks;
1078 avail_node_count = sbi->total_node_count - sbi->nquota_files -
1079 F2FS_RESERVED_NODE_NUM;
1081 if (avail_node_count > user_block_count) {
1082 buf->f_files = user_block_count;
1083 buf->f_ffree = buf->f_bavail;
1085 buf->f_files = avail_node_count;
1086 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1090 buf->f_namelen = F2FS_NAME_LEN;
1091 buf->f_fsid.val[0] = (u32)id;
1092 buf->f_fsid.val[1] = (u32)(id >> 32);
1095 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1096 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1097 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1103 static inline void f2fs_show_quota_options(struct seq_file *seq,
1104 struct super_block *sb)
1107 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1109 if (sbi->s_jquota_fmt) {
1112 switch (sbi->s_jquota_fmt) {
1123 seq_printf(seq, ",jqfmt=%s", fmtname);
1126 if (sbi->s_qf_names[USRQUOTA])
1127 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1129 if (sbi->s_qf_names[GRPQUOTA])
1130 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1132 if (sbi->s_qf_names[PRJQUOTA])
1133 seq_show_option(seq, "prjjquota", sbi->s_qf_names[PRJQUOTA]);
1137 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1139 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1141 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1142 if (test_opt(sbi, FORCE_FG_GC))
1143 seq_printf(seq, ",background_gc=%s", "sync");
1145 seq_printf(seq, ",background_gc=%s", "on");
1147 seq_printf(seq, ",background_gc=%s", "off");
1149 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1150 seq_puts(seq, ",disable_roll_forward");
1151 if (test_opt(sbi, DISCARD))
1152 seq_puts(seq, ",discard");
1153 if (test_opt(sbi, NOHEAP))
1154 seq_puts(seq, ",no_heap");
1156 seq_puts(seq, ",heap");
1157 #ifdef CONFIG_F2FS_FS_XATTR
1158 if (test_opt(sbi, XATTR_USER))
1159 seq_puts(seq, ",user_xattr");
1161 seq_puts(seq, ",nouser_xattr");
1162 if (test_opt(sbi, INLINE_XATTR))
1163 seq_puts(seq, ",inline_xattr");
1165 seq_puts(seq, ",noinline_xattr");
1166 if (test_opt(sbi, INLINE_XATTR_SIZE))
1167 seq_printf(seq, ",inline_xattr_size=%u",
1168 sbi->inline_xattr_size);
1170 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1171 if (test_opt(sbi, POSIX_ACL))
1172 seq_puts(seq, ",acl");
1174 seq_puts(seq, ",noacl");
1176 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1177 seq_puts(seq, ",disable_ext_identify");
1178 if (test_opt(sbi, INLINE_DATA))
1179 seq_puts(seq, ",inline_data");
1181 seq_puts(seq, ",noinline_data");
1182 if (test_opt(sbi, INLINE_DENTRY))
1183 seq_puts(seq, ",inline_dentry");
1185 seq_puts(seq, ",noinline_dentry");
1186 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1187 seq_puts(seq, ",flush_merge");
1188 if (test_opt(sbi, NOBARRIER))
1189 seq_puts(seq, ",nobarrier");
1190 if (test_opt(sbi, FASTBOOT))
1191 seq_puts(seq, ",fastboot");
1192 if (test_opt(sbi, EXTENT_CACHE))
1193 seq_puts(seq, ",extent_cache");
1195 seq_puts(seq, ",noextent_cache");
1196 if (test_opt(sbi, DATA_FLUSH))
1197 seq_puts(seq, ",data_flush");
1199 seq_puts(seq, ",mode=");
1200 if (test_opt(sbi, ADAPTIVE))
1201 seq_puts(seq, "adaptive");
1202 else if (test_opt(sbi, LFS))
1203 seq_puts(seq, "lfs");
1204 seq_printf(seq, ",active_logs=%u", sbi->active_logs);
1205 if (test_opt(sbi, RESERVE_ROOT))
1206 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1207 sbi->root_reserved_blocks,
1208 from_kuid_munged(&init_user_ns, sbi->s_resuid),
1209 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1210 if (F2FS_IO_SIZE_BITS(sbi))
1211 seq_printf(seq, ",io_size=%uKB", F2FS_IO_SIZE_KB(sbi));
1212 #ifdef CONFIG_F2FS_FAULT_INJECTION
1213 if (test_opt(sbi, FAULT_INJECTION))
1214 seq_printf(seq, ",fault_injection=%u",
1215 sbi->fault_info.inject_rate);
1218 if (test_opt(sbi, QUOTA))
1219 seq_puts(seq, ",quota");
1220 if (test_opt(sbi, USRQUOTA))
1221 seq_puts(seq, ",usrquota");
1222 if (test_opt(sbi, GRPQUOTA))
1223 seq_puts(seq, ",grpquota");
1224 if (test_opt(sbi, PRJQUOTA))
1225 seq_puts(seq, ",prjquota");
1227 f2fs_show_quota_options(seq, sbi->sb);
1232 static void default_options(struct f2fs_sb_info *sbi)
1234 /* init some FS parameters */
1235 sbi->active_logs = NR_CURSEG_TYPE;
1236 sbi->inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1238 set_opt(sbi, BG_GC);
1239 set_opt(sbi, INLINE_XATTR);
1240 set_opt(sbi, INLINE_DATA);
1241 set_opt(sbi, INLINE_DENTRY);
1242 set_opt(sbi, EXTENT_CACHE);
1243 set_opt(sbi, NOHEAP);
1244 sbi->sb->s_flags |= SB_LAZYTIME;
1245 set_opt(sbi, FLUSH_MERGE);
1246 if (f2fs_sb_mounted_blkzoned(sbi->sb)) {
1247 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1248 set_opt(sbi, DISCARD);
1250 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1253 #ifdef CONFIG_F2FS_FS_XATTR
1254 set_opt(sbi, XATTR_USER);
1256 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1257 set_opt(sbi, POSIX_ACL);
1260 #ifdef CONFIG_F2FS_FAULT_INJECTION
1261 f2fs_build_fault_attr(sbi, 0);
1266 static int f2fs_enable_quotas(struct super_block *sb);
1268 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1270 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1271 struct f2fs_mount_info org_mount_opt;
1272 unsigned long old_sb_flags;
1273 int err, active_logs;
1274 bool need_restart_gc = false;
1275 bool need_stop_gc = false;
1276 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1277 #ifdef CONFIG_F2FS_FAULT_INJECTION
1278 struct f2fs_fault_info ffi = sbi->fault_info;
1282 char *s_qf_names[MAXQUOTAS];
1287 * Save the old mount options in case we
1288 * need to restore them.
1290 org_mount_opt = sbi->mount_opt;
1291 old_sb_flags = sb->s_flags;
1292 active_logs = sbi->active_logs;
1295 s_jquota_fmt = sbi->s_jquota_fmt;
1296 for (i = 0; i < MAXQUOTAS; i++) {
1297 if (sbi->s_qf_names[i]) {
1298 s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
1300 if (!s_qf_names[i]) {
1301 for (j = 0; j < i; j++)
1302 kfree(s_qf_names[j]);
1306 s_qf_names[i] = NULL;
1311 /* recover superblocks we couldn't write due to previous RO mount */
1312 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1313 err = f2fs_commit_super(sbi, false);
1314 f2fs_msg(sb, KERN_INFO,
1315 "Try to recover all the superblocks, ret: %d", err);
1317 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1320 default_options(sbi);
1322 /* parse mount options */
1323 err = parse_options(sb, data);
1328 * Previous and new state of filesystem is RO,
1329 * so skip checking GC and FLUSH_MERGE conditions.
1331 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1335 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1336 err = dquot_suspend(sb, -1);
1339 } else if (f2fs_readonly(sb) && !(*flags & MS_RDONLY)) {
1340 /* dquot_resume needs RW */
1341 sb->s_flags &= ~SB_RDONLY;
1342 if (sb_any_quota_suspended(sb)) {
1343 dquot_resume(sb, -1);
1344 } else if (f2fs_sb_has_quota_ino(sb)) {
1345 err = f2fs_enable_quotas(sb);
1351 /* disallow enable/disable extent_cache dynamically */
1352 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1354 f2fs_msg(sbi->sb, KERN_WARNING,
1355 "switch extent_cache option is not allowed");
1360 * We stop the GC thread if FS is mounted as RO
1361 * or if background_gc = off is passed in mount
1362 * option. Also sync the filesystem.
1364 if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) {
1365 if (sbi->gc_thread) {
1366 stop_gc_thread(sbi);
1367 need_restart_gc = true;
1369 } else if (!sbi->gc_thread) {
1370 err = start_gc_thread(sbi);
1373 need_stop_gc = true;
1376 if (*flags & SB_RDONLY) {
1377 writeback_inodes_sb(sb, WB_REASON_SYNC);
1380 set_sbi_flag(sbi, SBI_IS_DIRTY);
1381 set_sbi_flag(sbi, SBI_IS_CLOSE);
1382 f2fs_sync_fs(sb, 1);
1383 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1387 * We stop issue flush thread if FS is mounted as RO
1388 * or if flush_merge is not passed in mount option.
1390 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1391 clear_opt(sbi, FLUSH_MERGE);
1392 destroy_flush_cmd_control(sbi, false);
1394 err = create_flush_cmd_control(sbi);
1400 /* Release old quota file names */
1401 for (i = 0; i < MAXQUOTAS; i++)
1402 kfree(s_qf_names[i]);
1404 /* Update the POSIXACL Flag */
1405 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1406 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1408 limit_reserve_root(sbi);
1411 if (need_restart_gc) {
1412 if (start_gc_thread(sbi))
1413 f2fs_msg(sbi->sb, KERN_WARNING,
1414 "background gc thread has stopped");
1415 } else if (need_stop_gc) {
1416 stop_gc_thread(sbi);
1420 sbi->s_jquota_fmt = s_jquota_fmt;
1421 for (i = 0; i < MAXQUOTAS; i++) {
1422 kfree(sbi->s_qf_names[i]);
1423 sbi->s_qf_names[i] = s_qf_names[i];
1426 sbi->mount_opt = org_mount_opt;
1427 sbi->active_logs = active_logs;
1428 sb->s_flags = old_sb_flags;
1429 #ifdef CONFIG_F2FS_FAULT_INJECTION
1430 sbi->fault_info = ffi;
1436 /* Read data from quotafile */
1437 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1438 size_t len, loff_t off)
1440 struct inode *inode = sb_dqopt(sb)->files[type];
1441 struct address_space *mapping = inode->i_mapping;
1442 block_t blkidx = F2FS_BYTES_TO_BLK(off);
1443 int offset = off & (sb->s_blocksize - 1);
1446 loff_t i_size = i_size_read(inode);
1453 if (off + len > i_size)
1456 while (toread > 0) {
1457 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1459 page = read_mapping_page(mapping, blkidx, NULL);
1461 if (PTR_ERR(page) == -ENOMEM) {
1462 congestion_wait(BLK_RW_ASYNC, HZ/50);
1465 return PTR_ERR(page);
1470 if (unlikely(page->mapping != mapping)) {
1471 f2fs_put_page(page, 1);
1474 if (unlikely(!PageUptodate(page))) {
1475 f2fs_put_page(page, 1);
1479 kaddr = kmap_atomic(page);
1480 memcpy(data, kaddr + offset, tocopy);
1481 kunmap_atomic(kaddr);
1482 f2fs_put_page(page, 1);
1492 /* Write to quotafile */
1493 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1494 const char *data, size_t len, loff_t off)
1496 struct inode *inode = sb_dqopt(sb)->files[type];
1497 struct address_space *mapping = inode->i_mapping;
1498 const struct address_space_operations *a_ops = mapping->a_ops;
1499 int offset = off & (sb->s_blocksize - 1);
1500 size_t towrite = len;
1506 while (towrite > 0) {
1507 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1510 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1512 if (unlikely(err)) {
1513 if (err == -ENOMEM) {
1514 congestion_wait(BLK_RW_ASYNC, HZ/50);
1520 kaddr = kmap_atomic(page);
1521 memcpy(kaddr + offset, data, tocopy);
1522 kunmap_atomic(kaddr);
1523 flush_dcache_page(page);
1525 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1536 inode->i_mtime = inode->i_ctime = current_time(inode);
1537 f2fs_mark_inode_dirty_sync(inode, false);
1538 return len - towrite;
1541 static struct dquot **f2fs_get_dquots(struct inode *inode)
1543 return F2FS_I(inode)->i_dquot;
1546 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1548 return &F2FS_I(inode)->i_reserved_quota;
1551 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1553 return dquot_quota_on_mount(sbi->sb, sbi->s_qf_names[type],
1554 sbi->s_jquota_fmt, type);
1557 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1562 if (f2fs_sb_has_quota_ino(sbi->sb) && rdonly) {
1563 err = f2fs_enable_quotas(sbi->sb);
1565 f2fs_msg(sbi->sb, KERN_ERR,
1566 "Cannot turn on quota_ino: %d", err);
1572 for (i = 0; i < MAXQUOTAS; i++) {
1573 if (sbi->s_qf_names[i]) {
1574 err = f2fs_quota_on_mount(sbi, i);
1579 f2fs_msg(sbi->sb, KERN_ERR,
1580 "Cannot turn on quotas: %d on %d", err, i);
1586 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
1589 struct inode *qf_inode;
1590 unsigned long qf_inum;
1593 BUG_ON(!f2fs_sb_has_quota_ino(sb));
1595 qf_inum = f2fs_qf_ino(sb, type);
1599 qf_inode = f2fs_iget(sb, qf_inum);
1600 if (IS_ERR(qf_inode)) {
1601 f2fs_msg(sb, KERN_ERR,
1602 "Bad quota inode %u:%lu", type, qf_inum);
1603 return PTR_ERR(qf_inode);
1606 /* Don't account quota for quota files to avoid recursion */
1607 qf_inode->i_flags |= S_NOQUOTA;
1608 err = dquot_enable(qf_inode, type, format_id, flags);
1613 static int f2fs_enable_quotas(struct super_block *sb)
1616 unsigned long qf_inum;
1617 bool quota_mopt[MAXQUOTAS] = {
1618 test_opt(F2FS_SB(sb), USRQUOTA),
1619 test_opt(F2FS_SB(sb), GRPQUOTA),
1620 test_opt(F2FS_SB(sb), PRJQUOTA),
1623 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
1624 for (type = 0; type < MAXQUOTAS; type++) {
1625 qf_inum = f2fs_qf_ino(sb, type);
1627 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
1628 DQUOT_USAGE_ENABLED |
1629 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
1631 f2fs_msg(sb, KERN_ERR,
1632 "Failed to enable quota tracking "
1633 "(type=%d, err=%d). Please run "
1634 "fsck to fix.", type, err);
1635 for (type--; type >= 0; type--)
1636 dquot_quota_off(sb, type);
1644 static int f2fs_quota_sync(struct super_block *sb, int type)
1646 struct quota_info *dqopt = sb_dqopt(sb);
1650 ret = dquot_writeback_dquots(sb, type);
1655 * Now when everything is written we can discard the pagecache so
1656 * that userspace sees the changes.
1658 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
1659 if (type != -1 && cnt != type)
1661 if (!sb_has_quota_active(sb, cnt))
1664 ret = filemap_write_and_wait(dqopt->files[cnt]->i_mapping);
1668 inode_lock(dqopt->files[cnt]);
1669 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
1670 inode_unlock(dqopt->files[cnt]);
1675 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
1676 const struct path *path)
1678 struct inode *inode;
1681 err = f2fs_quota_sync(sb, type);
1685 err = dquot_quota_on(sb, type, format_id, path);
1689 inode = d_inode(path->dentry);
1692 F2FS_I(inode)->i_flags |= FS_NOATIME_FL | FS_IMMUTABLE_FL;
1693 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1694 S_NOATIME | S_IMMUTABLE);
1695 inode_unlock(inode);
1696 f2fs_mark_inode_dirty_sync(inode, false);
1701 static int f2fs_quota_off(struct super_block *sb, int type)
1703 struct inode *inode = sb_dqopt(sb)->files[type];
1706 if (!inode || !igrab(inode))
1707 return dquot_quota_off(sb, type);
1709 f2fs_quota_sync(sb, type);
1711 err = dquot_quota_off(sb, type);
1712 if (err || f2fs_sb_has_quota_ino(sb))
1716 F2FS_I(inode)->i_flags &= ~(FS_NOATIME_FL | FS_IMMUTABLE_FL);
1717 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1718 inode_unlock(inode);
1719 f2fs_mark_inode_dirty_sync(inode, false);
1725 void f2fs_quota_off_umount(struct super_block *sb)
1729 for (type = 0; type < MAXQUOTAS; type++)
1730 f2fs_quota_off(sb, type);
1733 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
1735 *projid = F2FS_I(inode)->i_projid;
1739 static const struct dquot_operations f2fs_quota_operations = {
1740 .get_reserved_space = f2fs_get_reserved_space,
1741 .write_dquot = dquot_commit,
1742 .acquire_dquot = dquot_acquire,
1743 .release_dquot = dquot_release,
1744 .mark_dirty = dquot_mark_dquot_dirty,
1745 .write_info = dquot_commit_info,
1746 .alloc_dquot = dquot_alloc,
1747 .destroy_dquot = dquot_destroy,
1748 .get_projid = f2fs_get_projid,
1749 .get_next_id = dquot_get_next_id,
1752 static const struct quotactl_ops f2fs_quotactl_ops = {
1753 .quota_on = f2fs_quota_on,
1754 .quota_off = f2fs_quota_off,
1755 .quota_sync = f2fs_quota_sync,
1756 .get_state = dquot_get_state,
1757 .set_info = dquot_set_dqinfo,
1758 .get_dqblk = dquot_get_dqblk,
1759 .set_dqblk = dquot_set_dqblk,
1760 .get_nextdqblk = dquot_get_next_dqblk,
1763 void f2fs_quota_off_umount(struct super_block *sb)
1768 static const struct super_operations f2fs_sops = {
1769 .alloc_inode = f2fs_alloc_inode,
1770 .drop_inode = f2fs_drop_inode,
1771 .destroy_inode = f2fs_destroy_inode,
1772 .write_inode = f2fs_write_inode,
1773 .dirty_inode = f2fs_dirty_inode,
1774 .show_options = f2fs_show_options,
1776 .quota_read = f2fs_quota_read,
1777 .quota_write = f2fs_quota_write,
1778 .get_dquots = f2fs_get_dquots,
1780 .evict_inode = f2fs_evict_inode,
1781 .put_super = f2fs_put_super,
1782 .sync_fs = f2fs_sync_fs,
1783 .freeze_fs = f2fs_freeze,
1784 .unfreeze_fs = f2fs_unfreeze,
1785 .statfs = f2fs_statfs,
1786 .remount_fs = f2fs_remount,
1789 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1790 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1792 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1793 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1797 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1800 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1801 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1802 ctx, len, fs_data, XATTR_CREATE);
1805 static unsigned f2fs_max_namelen(struct inode *inode)
1807 return S_ISLNK(inode->i_mode) ?
1808 inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1811 static const struct fscrypt_operations f2fs_cryptops = {
1812 .key_prefix = "f2fs:",
1813 .get_context = f2fs_get_context,
1814 .set_context = f2fs_set_context,
1815 .empty_dir = f2fs_empty_dir,
1816 .max_namelen = f2fs_max_namelen,
1820 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1821 u64 ino, u32 generation)
1823 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1824 struct inode *inode;
1826 if (check_nid_range(sbi, ino))
1827 return ERR_PTR(-ESTALE);
1830 * f2fs_iget isn't quite right if the inode is currently unallocated!
1831 * However f2fs_iget currently does appropriate checks to handle stale
1832 * inodes so everything is OK.
1834 inode = f2fs_iget(sb, ino);
1836 return ERR_CAST(inode);
1837 if (unlikely(generation && inode->i_generation != generation)) {
1838 /* we didn't find the right inode.. */
1840 return ERR_PTR(-ESTALE);
1845 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1846 int fh_len, int fh_type)
1848 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1849 f2fs_nfs_get_inode);
1852 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1853 int fh_len, int fh_type)
1855 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1856 f2fs_nfs_get_inode);
1859 static const struct export_operations f2fs_export_ops = {
1860 .fh_to_dentry = f2fs_fh_to_dentry,
1861 .fh_to_parent = f2fs_fh_to_parent,
1862 .get_parent = f2fs_get_parent,
1865 static loff_t max_file_blocks(void)
1868 loff_t leaf_count = ADDRS_PER_BLOCK;
1871 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
1872 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
1873 * space in inode.i_addr, it will be more safe to reassign
1877 /* two direct node blocks */
1878 result += (leaf_count * 2);
1880 /* two indirect node blocks */
1881 leaf_count *= NIDS_PER_BLOCK;
1882 result += (leaf_count * 2);
1884 /* one double indirect node block */
1885 leaf_count *= NIDS_PER_BLOCK;
1886 result += leaf_count;
1891 static int __f2fs_commit_super(struct buffer_head *bh,
1892 struct f2fs_super_block *super)
1896 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
1897 set_buffer_uptodate(bh);
1898 set_buffer_dirty(bh);
1901 /* it's rare case, we can do fua all the time */
1902 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
1905 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
1906 struct buffer_head *bh)
1908 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1909 (bh->b_data + F2FS_SUPER_OFFSET);
1910 struct super_block *sb = sbi->sb;
1911 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
1912 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
1913 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
1914 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
1915 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
1916 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
1917 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
1918 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
1919 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
1920 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
1921 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1922 u32 segment_count = le32_to_cpu(raw_super->segment_count);
1923 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1924 u64 main_end_blkaddr = main_blkaddr +
1925 (segment_count_main << log_blocks_per_seg);
1926 u64 seg_end_blkaddr = segment0_blkaddr +
1927 (segment_count << log_blocks_per_seg);
1929 if (segment0_blkaddr != cp_blkaddr) {
1930 f2fs_msg(sb, KERN_INFO,
1931 "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1932 segment0_blkaddr, cp_blkaddr);
1936 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
1938 f2fs_msg(sb, KERN_INFO,
1939 "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1940 cp_blkaddr, sit_blkaddr,
1941 segment_count_ckpt << log_blocks_per_seg);
1945 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
1947 f2fs_msg(sb, KERN_INFO,
1948 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1949 sit_blkaddr, nat_blkaddr,
1950 segment_count_sit << log_blocks_per_seg);
1954 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
1956 f2fs_msg(sb, KERN_INFO,
1957 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1958 nat_blkaddr, ssa_blkaddr,
1959 segment_count_nat << log_blocks_per_seg);
1963 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
1965 f2fs_msg(sb, KERN_INFO,
1966 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1967 ssa_blkaddr, main_blkaddr,
1968 segment_count_ssa << log_blocks_per_seg);
1972 if (main_end_blkaddr > seg_end_blkaddr) {
1973 f2fs_msg(sb, KERN_INFO,
1974 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1977 (segment_count << log_blocks_per_seg),
1978 segment_count_main << log_blocks_per_seg);
1980 } else if (main_end_blkaddr < seg_end_blkaddr) {
1984 /* fix in-memory information all the time */
1985 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
1986 segment0_blkaddr) >> log_blocks_per_seg);
1988 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
1989 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1992 err = __f2fs_commit_super(bh, NULL);
1993 res = err ? "failed" : "done";
1995 f2fs_msg(sb, KERN_INFO,
1996 "Fix alignment : %s, start(%u) end(%u) block(%u)",
1999 (segment_count << log_blocks_per_seg),
2000 segment_count_main << log_blocks_per_seg);
2007 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2008 struct buffer_head *bh)
2010 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2011 (bh->b_data + F2FS_SUPER_OFFSET);
2012 struct super_block *sb = sbi->sb;
2013 unsigned int blocksize;
2015 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
2016 f2fs_msg(sb, KERN_INFO,
2017 "Magic Mismatch, valid(0x%x) - read(0x%x)",
2018 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2022 /* Currently, support only 4KB page cache size */
2023 if (F2FS_BLKSIZE != PAGE_SIZE) {
2024 f2fs_msg(sb, KERN_INFO,
2025 "Invalid page_cache_size (%lu), supports only 4KB\n",
2030 /* Currently, support only 4KB block size */
2031 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2032 if (blocksize != F2FS_BLKSIZE) {
2033 f2fs_msg(sb, KERN_INFO,
2034 "Invalid blocksize (%u), supports only 4KB\n",
2039 /* check log blocks per segment */
2040 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2041 f2fs_msg(sb, KERN_INFO,
2042 "Invalid log blocks per segment (%u)\n",
2043 le32_to_cpu(raw_super->log_blocks_per_seg));
2047 /* Currently, support 512/1024/2048/4096 bytes sector size */
2048 if (le32_to_cpu(raw_super->log_sectorsize) >
2049 F2FS_MAX_LOG_SECTOR_SIZE ||
2050 le32_to_cpu(raw_super->log_sectorsize) <
2051 F2FS_MIN_LOG_SECTOR_SIZE) {
2052 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
2053 le32_to_cpu(raw_super->log_sectorsize));
2056 if (le32_to_cpu(raw_super->log_sectors_per_block) +
2057 le32_to_cpu(raw_super->log_sectorsize) !=
2058 F2FS_MAX_LOG_SECTOR_SIZE) {
2059 f2fs_msg(sb, KERN_INFO,
2060 "Invalid log sectors per block(%u) log sectorsize(%u)",
2061 le32_to_cpu(raw_super->log_sectors_per_block),
2062 le32_to_cpu(raw_super->log_sectorsize));
2066 /* check reserved ino info */
2067 if (le32_to_cpu(raw_super->node_ino) != 1 ||
2068 le32_to_cpu(raw_super->meta_ino) != 2 ||
2069 le32_to_cpu(raw_super->root_ino) != 3) {
2070 f2fs_msg(sb, KERN_INFO,
2071 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2072 le32_to_cpu(raw_super->node_ino),
2073 le32_to_cpu(raw_super->meta_ino),
2074 le32_to_cpu(raw_super->root_ino));
2078 if (le32_to_cpu(raw_super->segment_count) > F2FS_MAX_SEGMENT) {
2079 f2fs_msg(sb, KERN_INFO,
2080 "Invalid segment count (%u)",
2081 le32_to_cpu(raw_super->segment_count));
2085 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2086 if (sanity_check_area_boundary(sbi, bh))
2092 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
2094 unsigned int total, fsmeta;
2095 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2096 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2097 unsigned int ovp_segments, reserved_segments;
2098 unsigned int main_segs, blocks_per_seg;
2101 total = le32_to_cpu(raw_super->segment_count);
2102 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2103 fsmeta += le32_to_cpu(raw_super->segment_count_sit);
2104 fsmeta += le32_to_cpu(raw_super->segment_count_nat);
2105 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2106 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2108 if (unlikely(fsmeta >= total))
2111 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2112 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2114 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2115 ovp_segments == 0 || reserved_segments == 0)) {
2116 f2fs_msg(sbi->sb, KERN_ERR,
2117 "Wrong layout: check mkfs.f2fs version");
2121 main_segs = le32_to_cpu(raw_super->segment_count_main);
2122 blocks_per_seg = sbi->blocks_per_seg;
2124 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2125 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2126 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2129 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2130 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2131 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2135 if (unlikely(f2fs_cp_error(sbi))) {
2136 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
2142 static void init_sb_info(struct f2fs_sb_info *sbi)
2144 struct f2fs_super_block *raw_super = sbi->raw_super;
2147 sbi->log_sectors_per_block =
2148 le32_to_cpu(raw_super->log_sectors_per_block);
2149 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2150 sbi->blocksize = 1 << sbi->log_blocksize;
2151 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2152 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2153 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2154 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2155 sbi->total_sections = le32_to_cpu(raw_super->section_count);
2156 sbi->total_node_count =
2157 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2158 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2159 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2160 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2161 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2162 sbi->cur_victim_sec = NULL_SECNO;
2163 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2165 sbi->dir_level = DEF_DIR_LEVEL;
2166 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2167 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2168 clear_sbi_flag(sbi, SBI_NEED_FSCK);
2170 for (i = 0; i < NR_COUNT_TYPE; i++)
2171 atomic_set(&sbi->nr_pages[i], 0);
2173 atomic_set(&sbi->wb_sync_req, 0);
2175 INIT_LIST_HEAD(&sbi->s_list);
2176 mutex_init(&sbi->umount_mutex);
2177 for (i = 0; i < NR_PAGE_TYPE - 1; i++)
2178 for (j = HOT; j < NR_TEMP_TYPE; j++)
2179 mutex_init(&sbi->wio_mutex[i][j]);
2180 spin_lock_init(&sbi->cp_lock);
2182 sbi->dirty_device = 0;
2183 spin_lock_init(&sbi->dev_lock);
2186 static int init_percpu_info(struct f2fs_sb_info *sbi)
2190 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2194 return percpu_counter_init(&sbi->total_valid_inode_count, 0,
2198 #ifdef CONFIG_BLK_DEV_ZONED
2199 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2201 struct block_device *bdev = FDEV(devi).bdev;
2202 sector_t nr_sectors = bdev->bd_part->nr_sects;
2203 sector_t sector = 0;
2204 struct blk_zone *zones;
2205 unsigned int i, nr_zones;
2209 if (!f2fs_sb_mounted_blkzoned(sbi->sb))
2212 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2213 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2215 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2216 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2217 __ilog2_u32(sbi->blocks_per_blkz))
2219 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2220 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2221 sbi->log_blocks_per_blkz;
2222 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2223 FDEV(devi).nr_blkz++;
2225 FDEV(devi).blkz_type = f2fs_kmalloc(sbi, FDEV(devi).nr_blkz,
2227 if (!FDEV(devi).blkz_type)
2230 #define F2FS_REPORT_NR_ZONES 4096
2232 zones = f2fs_kzalloc(sbi, sizeof(struct blk_zone) *
2233 F2FS_REPORT_NR_ZONES, GFP_KERNEL);
2237 /* Get block zones type */
2238 while (zones && sector < nr_sectors) {
2240 nr_zones = F2FS_REPORT_NR_ZONES;
2241 err = blkdev_report_zones(bdev, sector,
2251 for (i = 0; i < nr_zones; i++) {
2252 FDEV(devi).blkz_type[n] = zones[i].type;
2253 sector += zones[i].len;
2265 * Read f2fs raw super block.
2266 * Because we have two copies of super block, so read both of them
2267 * to get the first valid one. If any one of them is broken, we pass
2268 * them recovery flag back to the caller.
2270 static int read_raw_super_block(struct f2fs_sb_info *sbi,
2271 struct f2fs_super_block **raw_super,
2272 int *valid_super_block, int *recovery)
2274 struct super_block *sb = sbi->sb;
2276 struct buffer_head *bh;
2277 struct f2fs_super_block *super;
2280 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2284 for (block = 0; block < 2; block++) {
2285 bh = sb_bread(sb, block);
2287 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
2293 /* sanity checking of raw super */
2294 if (sanity_check_raw_super(sbi, bh)) {
2295 f2fs_msg(sb, KERN_ERR,
2296 "Can't find valid F2FS filesystem in %dth superblock",
2304 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2306 *valid_super_block = block;
2312 /* Fail to read any one of the superblocks*/
2316 /* No valid superblock */
2325 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
2327 struct buffer_head *bh;
2330 if ((recover && f2fs_readonly(sbi->sb)) ||
2331 bdev_read_only(sbi->sb->s_bdev)) {
2332 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2336 /* write back-up superblock first */
2337 bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1);
2340 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2343 /* if we are in recovery path, skip writing valid superblock */
2347 /* write current valid superblock */
2348 bh = sb_getblk(sbi->sb, sbi->valid_super_block);
2351 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2356 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
2358 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2359 unsigned int max_devices = MAX_DEVICES;
2362 /* Initialize single device information */
2363 if (!RDEV(0).path[0]) {
2364 if (!bdev_is_zoned(sbi->sb->s_bdev))
2370 * Initialize multiple devices information, or single
2371 * zoned block device information.
2373 sbi->devs = f2fs_kzalloc(sbi, sizeof(struct f2fs_dev_info) *
2374 max_devices, GFP_KERNEL);
2378 for (i = 0; i < max_devices; i++) {
2380 if (i > 0 && !RDEV(i).path[0])
2383 if (max_devices == 1) {
2384 /* Single zoned block device mount */
2386 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
2387 sbi->sb->s_mode, sbi->sb->s_type);
2389 /* Multi-device mount */
2390 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
2391 FDEV(i).total_segments =
2392 le32_to_cpu(RDEV(i).total_segments);
2394 FDEV(i).start_blk = 0;
2395 FDEV(i).end_blk = FDEV(i).start_blk +
2396 (FDEV(i).total_segments <<
2397 sbi->log_blocks_per_seg) - 1 +
2398 le32_to_cpu(raw_super->segment0_blkaddr);
2400 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
2401 FDEV(i).end_blk = FDEV(i).start_blk +
2402 (FDEV(i).total_segments <<
2403 sbi->log_blocks_per_seg) - 1;
2405 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
2406 sbi->sb->s_mode, sbi->sb->s_type);
2408 if (IS_ERR(FDEV(i).bdev))
2409 return PTR_ERR(FDEV(i).bdev);
2411 /* to release errored devices */
2412 sbi->s_ndevs = i + 1;
2414 #ifdef CONFIG_BLK_DEV_ZONED
2415 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
2416 !f2fs_sb_mounted_blkzoned(sbi->sb)) {
2417 f2fs_msg(sbi->sb, KERN_ERR,
2418 "Zoned block device feature not enabled\n");
2421 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
2422 if (init_blkz_info(sbi, i)) {
2423 f2fs_msg(sbi->sb, KERN_ERR,
2424 "Failed to initialize F2FS blkzone information");
2427 if (max_devices == 1)
2429 f2fs_msg(sbi->sb, KERN_INFO,
2430 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
2432 FDEV(i).total_segments,
2433 FDEV(i).start_blk, FDEV(i).end_blk,
2434 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
2435 "Host-aware" : "Host-managed");
2439 f2fs_msg(sbi->sb, KERN_INFO,
2440 "Mount Device [%2d]: %20s, %8u, %8x - %8x",
2442 FDEV(i).total_segments,
2443 FDEV(i).start_blk, FDEV(i).end_blk);
2445 f2fs_msg(sbi->sb, KERN_INFO,
2446 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
2450 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
2452 struct f2fs_sb_info *sbi;
2453 struct f2fs_super_block *raw_super;
2456 bool retry = true, need_fsck = false;
2457 char *options = NULL;
2458 int recovery, i, valid_super_block;
2459 struct curseg_info *seg_i;
2464 valid_super_block = -1;
2467 /* allocate memory for f2fs-specific super block info */
2468 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
2474 /* Load the checksum driver */
2475 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
2476 if (IS_ERR(sbi->s_chksum_driver)) {
2477 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
2478 err = PTR_ERR(sbi->s_chksum_driver);
2479 sbi->s_chksum_driver = NULL;
2483 /* set a block size */
2484 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
2485 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
2489 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
2494 sb->s_fs_info = sbi;
2495 sbi->raw_super = raw_super;
2497 sbi->s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
2498 sbi->s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
2500 /* precompute checksum seed for metadata */
2501 if (f2fs_sb_has_inode_chksum(sb))
2502 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
2503 sizeof(raw_super->uuid));
2506 * The BLKZONED feature indicates that the drive was formatted with
2507 * zone alignment optimization. This is optional for host-aware
2508 * devices, but mandatory for host-managed zoned block devices.
2510 #ifndef CONFIG_BLK_DEV_ZONED
2511 if (f2fs_sb_mounted_blkzoned(sb)) {
2512 f2fs_msg(sb, KERN_ERR,
2513 "Zoned block device support is not enabled\n");
2518 default_options(sbi);
2519 /* parse mount options */
2520 options = kstrdup((const char *)data, GFP_KERNEL);
2521 if (data && !options) {
2526 err = parse_options(sb, options);
2530 sbi->max_file_blocks = max_file_blocks();
2531 sb->s_maxbytes = sbi->max_file_blocks <<
2532 le32_to_cpu(raw_super->log_blocksize);
2533 sb->s_max_links = F2FS_LINK_MAX;
2534 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2537 sb->dq_op = &f2fs_quota_operations;
2538 if (f2fs_sb_has_quota_ino(sb))
2539 sb->s_qcop = &dquot_quotactl_sysfile_ops;
2541 sb->s_qcop = &f2fs_quotactl_ops;
2542 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
2544 if (f2fs_sb_has_quota_ino(sbi->sb)) {
2545 for (i = 0; i < MAXQUOTAS; i++) {
2546 if (f2fs_qf_ino(sbi->sb, i))
2547 sbi->nquota_files++;
2552 sb->s_op = &f2fs_sops;
2553 #ifdef CONFIG_F2FS_FS_ENCRYPTION
2554 sb->s_cop = &f2fs_cryptops;
2556 sb->s_xattr = f2fs_xattr_handlers;
2557 sb->s_export_op = &f2fs_export_ops;
2558 sb->s_magic = F2FS_SUPER_MAGIC;
2559 sb->s_time_gran = 1;
2560 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2561 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2562 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
2563 sb->s_iflags |= SB_I_CGROUPWB;
2565 /* init f2fs-specific super block info */
2566 sbi->valid_super_block = valid_super_block;
2567 mutex_init(&sbi->gc_mutex);
2568 mutex_init(&sbi->cp_mutex);
2569 init_rwsem(&sbi->node_write);
2570 init_rwsem(&sbi->node_change);
2572 /* disallow all the data/node/meta page writes */
2573 set_sbi_flag(sbi, SBI_POR_DOING);
2574 spin_lock_init(&sbi->stat_lock);
2576 /* init iostat info */
2577 spin_lock_init(&sbi->iostat_lock);
2578 sbi->iostat_enable = false;
2580 for (i = 0; i < NR_PAGE_TYPE; i++) {
2581 int n = (i == META) ? 1: NR_TEMP_TYPE;
2584 sbi->write_io[i] = f2fs_kmalloc(sbi,
2585 n * sizeof(struct f2fs_bio_info),
2587 if (!sbi->write_io[i]) {
2592 for (j = HOT; j < n; j++) {
2593 init_rwsem(&sbi->write_io[i][j].io_rwsem);
2594 sbi->write_io[i][j].sbi = sbi;
2595 sbi->write_io[i][j].bio = NULL;
2596 spin_lock_init(&sbi->write_io[i][j].io_lock);
2597 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
2601 init_rwsem(&sbi->cp_rwsem);
2602 init_waitqueue_head(&sbi->cp_wait);
2605 err = init_percpu_info(sbi);
2609 if (F2FS_IO_SIZE(sbi) > 1) {
2610 sbi->write_io_dummy =
2611 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
2612 if (!sbi->write_io_dummy) {
2618 /* get an inode for meta space */
2619 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
2620 if (IS_ERR(sbi->meta_inode)) {
2621 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
2622 err = PTR_ERR(sbi->meta_inode);
2626 err = get_valid_checkpoint(sbi);
2628 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
2629 goto free_meta_inode;
2632 /* Initialize device list */
2633 err = f2fs_scan_devices(sbi);
2635 f2fs_msg(sb, KERN_ERR, "Failed to find devices");
2639 sbi->total_valid_node_count =
2640 le32_to_cpu(sbi->ckpt->valid_node_count);
2641 percpu_counter_set(&sbi->total_valid_inode_count,
2642 le32_to_cpu(sbi->ckpt->valid_inode_count));
2643 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
2644 sbi->total_valid_block_count =
2645 le64_to_cpu(sbi->ckpt->valid_block_count);
2646 sbi->last_valid_block_count = sbi->total_valid_block_count;
2647 sbi->reserved_blocks = 0;
2648 sbi->current_reserved_blocks = 0;
2649 limit_reserve_root(sbi);
2651 for (i = 0; i < NR_INODE_TYPE; i++) {
2652 INIT_LIST_HEAD(&sbi->inode_list[i]);
2653 spin_lock_init(&sbi->inode_lock[i]);
2656 init_extent_cache_info(sbi);
2658 init_ino_entry_info(sbi);
2660 /* setup f2fs internal modules */
2661 err = build_segment_manager(sbi);
2663 f2fs_msg(sb, KERN_ERR,
2664 "Failed to initialize F2FS segment manager");
2667 err = build_node_manager(sbi);
2669 f2fs_msg(sb, KERN_ERR,
2670 "Failed to initialize F2FS node manager");
2674 /* For write statistics */
2675 if (sb->s_bdev->bd_part)
2676 sbi->sectors_written_start =
2677 (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
2679 /* Read accumulated write IO statistics if exists */
2680 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
2681 if (__exist_node_summaries(sbi))
2682 sbi->kbytes_written =
2683 le64_to_cpu(seg_i->journal->info.kbytes_written);
2685 build_gc_manager(sbi);
2687 /* get an inode for node space */
2688 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
2689 if (IS_ERR(sbi->node_inode)) {
2690 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
2691 err = PTR_ERR(sbi->node_inode);
2695 err = f2fs_build_stats(sbi);
2697 goto free_node_inode;
2699 /* read root inode and dentry */
2700 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
2702 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
2703 err = PTR_ERR(root);
2706 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2709 goto free_node_inode;
2712 sb->s_root = d_make_root(root); /* allocate root dentry */
2715 goto free_root_inode;
2718 err = f2fs_register_sysfs(sbi);
2720 goto free_root_inode;
2724 * Turn on quotas which were not enabled for read-only mounts if
2725 * filesystem has quota feature, so that they are updated correctly.
2727 if (f2fs_sb_has_quota_ino(sb) && !sb_rdonly(sb)) {
2728 err = f2fs_enable_quotas(sb);
2730 f2fs_msg(sb, KERN_ERR,
2731 "Cannot turn on quotas: error %d", err);
2736 /* if there are nt orphan nodes free them */
2737 err = recover_orphan_inodes(sbi);
2741 /* recover fsynced data */
2742 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
2744 * mount should be failed, when device has readonly mode, and
2745 * previous checkpoint was not done by clean system shutdown.
2747 if (bdev_read_only(sb->s_bdev) &&
2748 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
2754 set_sbi_flag(sbi, SBI_NEED_FSCK);
2759 err = recover_fsync_data(sbi, false);
2762 f2fs_msg(sb, KERN_ERR,
2763 "Cannot recover all fsync data errno=%d", err);
2767 err = recover_fsync_data(sbi, true);
2769 if (!f2fs_readonly(sb) && err > 0) {
2771 f2fs_msg(sb, KERN_ERR,
2772 "Need to recover fsync data");
2777 /* recover_fsync_data() cleared this already */
2778 clear_sbi_flag(sbi, SBI_POR_DOING);
2781 * If filesystem is not mounted as read-only then
2782 * do start the gc_thread.
2784 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
2785 /* After POR, we can run background GC thread.*/
2786 err = start_gc_thread(sbi);
2792 /* recover broken superblock */
2794 err = f2fs_commit_super(sbi, true);
2795 f2fs_msg(sb, KERN_INFO,
2796 "Try to recover %dth superblock, ret: %d",
2797 sbi->valid_super_block ? 1 : 2, err);
2800 f2fs_join_shrinker(sbi);
2802 f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx",
2803 cur_cp_version(F2FS_CKPT(sbi)));
2804 f2fs_update_time(sbi, CP_TIME);
2805 f2fs_update_time(sbi, REQ_TIME);
2810 if (f2fs_sb_has_quota_ino(sb) && !sb_rdonly(sb))
2811 f2fs_quota_off_umount(sbi->sb);
2813 f2fs_sync_inode_meta(sbi);
2815 * Some dirty meta pages can be produced by recover_orphan_inodes()
2816 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
2817 * followed by write_checkpoint() through f2fs_write_node_pages(), which
2818 * falls into an infinite loop in sync_meta_pages().
2820 truncate_inode_pages_final(META_MAPPING(sbi));
2824 f2fs_unregister_sysfs(sbi);
2829 f2fs_destroy_stats(sbi);
2831 release_ino_entry(sbi, true);
2832 truncate_inode_pages_final(NODE_MAPPING(sbi));
2833 iput(sbi->node_inode);
2835 destroy_node_manager(sbi);
2837 destroy_segment_manager(sbi);
2839 destroy_device_list(sbi);
2842 make_bad_inode(sbi->meta_inode);
2843 iput(sbi->meta_inode);
2845 mempool_destroy(sbi->write_io_dummy);
2847 destroy_percpu_info(sbi);
2849 for (i = 0; i < NR_PAGE_TYPE; i++)
2850 kfree(sbi->write_io[i]);
2853 for (i = 0; i < MAXQUOTAS; i++)
2854 kfree(sbi->s_qf_names[i]);
2860 if (sbi->s_chksum_driver)
2861 crypto_free_shash(sbi->s_chksum_driver);
2864 /* give only one another chance */
2867 shrink_dcache_sb(sb);
2873 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
2874 const char *dev_name, void *data)
2876 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
2879 static void kill_f2fs_super(struct super_block *sb)
2882 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
2883 stop_gc_thread(F2FS_SB(sb));
2884 stop_discard_thread(F2FS_SB(sb));
2886 kill_block_super(sb);
2889 static struct file_system_type f2fs_fs_type = {
2890 .owner = THIS_MODULE,
2892 .mount = f2fs_mount,
2893 .kill_sb = kill_f2fs_super,
2894 .fs_flags = FS_REQUIRES_DEV,
2896 MODULE_ALIAS_FS("f2fs");
2898 static int __init init_inodecache(void)
2900 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
2901 sizeof(struct f2fs_inode_info), 0,
2902 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
2903 if (!f2fs_inode_cachep)
2908 static void destroy_inodecache(void)
2911 * Make sure all delayed rcu free inodes are flushed before we
2915 kmem_cache_destroy(f2fs_inode_cachep);
2918 static int __init init_f2fs_fs(void)
2922 f2fs_build_trace_ios();
2924 err = init_inodecache();
2927 err = create_node_manager_caches();
2929 goto free_inodecache;
2930 err = create_segment_manager_caches();
2932 goto free_node_manager_caches;
2933 err = create_checkpoint_caches();
2935 goto free_segment_manager_caches;
2936 err = create_extent_cache();
2938 goto free_checkpoint_caches;
2939 err = f2fs_init_sysfs();
2941 goto free_extent_cache;
2942 err = register_shrinker(&f2fs_shrinker_info);
2945 err = register_filesystem(&f2fs_fs_type);
2948 err = f2fs_create_root_stats();
2950 goto free_filesystem;
2954 unregister_filesystem(&f2fs_fs_type);
2956 unregister_shrinker(&f2fs_shrinker_info);
2960 destroy_extent_cache();
2961 free_checkpoint_caches:
2962 destroy_checkpoint_caches();
2963 free_segment_manager_caches:
2964 destroy_segment_manager_caches();
2965 free_node_manager_caches:
2966 destroy_node_manager_caches();
2968 destroy_inodecache();
2973 static void __exit exit_f2fs_fs(void)
2975 f2fs_destroy_root_stats();
2976 unregister_filesystem(&f2fs_fs_type);
2977 unregister_shrinker(&f2fs_shrinker_info);
2979 destroy_extent_cache();
2980 destroy_checkpoint_caches();
2981 destroy_segment_manager_caches();
2982 destroy_node_manager_caches();
2983 destroy_inodecache();
2984 f2fs_destroy_trace_ios();
2987 module_init(init_f2fs_fs)
2988 module_exit(exit_f2fs_fs)
2990 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
2991 MODULE_DESCRIPTION("Flash Friendly File System");
2992 MODULE_LICENSE("GPL");