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_PAGE_ALLOC] = "page alloc",
47 [FAULT_ALLOC_NID] = "alloc nid",
48 [FAULT_ORPHAN] = "orphan",
49 [FAULT_BLOCK] = "no more block",
50 [FAULT_DIR_DEPTH] = "too big dir depth",
51 [FAULT_EVICT_INODE] = "evict_inode fail",
52 [FAULT_TRUNCATE] = "truncate fail",
53 [FAULT_IO] = "IO error",
54 [FAULT_CHECKPOINT] = "checkpoint error",
57 static void f2fs_build_fault_attr(struct f2fs_sb_info *sbi,
60 struct f2fs_fault_info *ffi = &sbi->fault_info;
63 atomic_set(&ffi->inject_ops, 0);
64 ffi->inject_rate = rate;
65 ffi->inject_type = (1 << FAULT_MAX) - 1;
67 memset(ffi, 0, sizeof(struct f2fs_fault_info));
72 /* f2fs-wide shrinker description */
73 static struct shrinker f2fs_shrinker_info = {
74 .scan_objects = f2fs_shrink_scan,
75 .count_objects = f2fs_shrink_count,
76 .seeks = DEFAULT_SEEKS,
81 Opt_disable_roll_forward,
92 Opt_disable_ext_identify,
128 static match_table_t f2fs_tokens = {
129 {Opt_gc_background, "background_gc=%s"},
130 {Opt_disable_roll_forward, "disable_roll_forward"},
131 {Opt_norecovery, "norecovery"},
132 {Opt_discard, "discard"},
133 {Opt_nodiscard, "nodiscard"},
134 {Opt_noheap, "no_heap"},
136 {Opt_user_xattr, "user_xattr"},
137 {Opt_nouser_xattr, "nouser_xattr"},
139 {Opt_noacl, "noacl"},
140 {Opt_active_logs, "active_logs=%u"},
141 {Opt_disable_ext_identify, "disable_ext_identify"},
142 {Opt_inline_xattr, "inline_xattr"},
143 {Opt_noinline_xattr, "noinline_xattr"},
144 {Opt_inline_data, "inline_data"},
145 {Opt_inline_dentry, "inline_dentry"},
146 {Opt_noinline_dentry, "noinline_dentry"},
147 {Opt_flush_merge, "flush_merge"},
148 {Opt_noflush_merge, "noflush_merge"},
149 {Opt_nobarrier, "nobarrier"},
150 {Opt_fastboot, "fastboot"},
151 {Opt_extent_cache, "extent_cache"},
152 {Opt_noextent_cache, "noextent_cache"},
153 {Opt_noinline_data, "noinline_data"},
154 {Opt_data_flush, "data_flush"},
155 {Opt_mode, "mode=%s"},
156 {Opt_io_size_bits, "io_bits=%u"},
157 {Opt_fault_injection, "fault_injection=%u"},
158 {Opt_lazytime, "lazytime"},
159 {Opt_nolazytime, "nolazytime"},
160 {Opt_quota, "quota"},
161 {Opt_noquota, "noquota"},
162 {Opt_usrquota, "usrquota"},
163 {Opt_grpquota, "grpquota"},
164 {Opt_prjquota, "prjquota"},
165 {Opt_usrjquota, "usrjquota=%s"},
166 {Opt_grpjquota, "grpjquota=%s"},
167 {Opt_prjjquota, "prjjquota=%s"},
168 {Opt_offusrjquota, "usrjquota="},
169 {Opt_offgrpjquota, "grpjquota="},
170 {Opt_offprjjquota, "prjjquota="},
171 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
172 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
173 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
177 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
179 struct va_format vaf;
185 printk_ratelimited("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
189 static void init_once(void *foo)
191 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
193 inode_init_once(&fi->vfs_inode);
197 static const char * const quotatypes[] = INITQFNAMES;
198 #define QTYPE2NAME(t) (quotatypes[t])
199 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
202 struct f2fs_sb_info *sbi = F2FS_SB(sb);
206 if (sb_any_quota_loaded(sb) && !sbi->s_qf_names[qtype]) {
207 f2fs_msg(sb, KERN_ERR,
208 "Cannot change journaled "
209 "quota options when quota turned on");
212 qname = match_strdup(args);
214 f2fs_msg(sb, KERN_ERR,
215 "Not enough memory for storing quotafile name");
218 if (sbi->s_qf_names[qtype]) {
219 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
222 f2fs_msg(sb, KERN_ERR,
223 "%s quota file already specified",
227 if (strchr(qname, '/')) {
228 f2fs_msg(sb, KERN_ERR,
229 "quotafile must be on filesystem root");
232 sbi->s_qf_names[qtype] = qname;
240 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
242 struct f2fs_sb_info *sbi = F2FS_SB(sb);
244 if (sb_any_quota_loaded(sb) && sbi->s_qf_names[qtype]) {
245 f2fs_msg(sb, KERN_ERR, "Cannot change journaled quota options"
246 " when quota turned on");
249 kfree(sbi->s_qf_names[qtype]);
250 sbi->s_qf_names[qtype] = NULL;
254 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
257 * We do the test below only for project quotas. 'usrquota' and
258 * 'grpquota' mount options are allowed even without quota feature
259 * to support legacy quotas in quota files.
261 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi->sb)) {
262 f2fs_msg(sbi->sb, KERN_ERR, "Project quota feature not enabled. "
263 "Cannot enable project quota enforcement.");
266 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA] ||
267 sbi->s_qf_names[PRJQUOTA]) {
268 if (test_opt(sbi, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
269 clear_opt(sbi, USRQUOTA);
271 if (test_opt(sbi, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
272 clear_opt(sbi, GRPQUOTA);
274 if (test_opt(sbi, PRJQUOTA) && sbi->s_qf_names[PRJQUOTA])
275 clear_opt(sbi, PRJQUOTA);
277 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
278 test_opt(sbi, PRJQUOTA)) {
279 f2fs_msg(sbi->sb, KERN_ERR, "old and new quota "
284 if (!sbi->s_jquota_fmt) {
285 f2fs_msg(sbi->sb, KERN_ERR, "journaled quota format "
294 static int parse_options(struct super_block *sb, char *options)
296 struct f2fs_sb_info *sbi = F2FS_SB(sb);
297 struct request_queue *q;
298 substring_t args[MAX_OPT_ARGS];
308 while ((p = strsep(&options, ",")) != NULL) {
313 * Initialize args struct so we know whether arg was
314 * found; some options take optional arguments.
316 args[0].to = args[0].from = NULL;
317 token = match_token(p, f2fs_tokens, args);
320 case Opt_gc_background:
321 name = match_strdup(&args[0]);
325 if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
327 clear_opt(sbi, FORCE_FG_GC);
328 } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
329 clear_opt(sbi, BG_GC);
330 clear_opt(sbi, FORCE_FG_GC);
331 } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
333 set_opt(sbi, FORCE_FG_GC);
340 case Opt_disable_roll_forward:
341 set_opt(sbi, DISABLE_ROLL_FORWARD);
344 /* this option mounts f2fs with ro */
345 set_opt(sbi, DISABLE_ROLL_FORWARD);
346 if (!f2fs_readonly(sb))
350 q = bdev_get_queue(sb->s_bdev);
351 if (blk_queue_discard(q)) {
352 set_opt(sbi, DISCARD);
353 } else if (!f2fs_sb_mounted_blkzoned(sb)) {
354 f2fs_msg(sb, KERN_WARNING,
355 "mounting with \"discard\" option, but "
356 "the device does not support discard");
360 if (f2fs_sb_mounted_blkzoned(sb)) {
361 f2fs_msg(sb, KERN_WARNING,
362 "discard is required for zoned block devices");
365 clear_opt(sbi, DISCARD);
368 set_opt(sbi, NOHEAP);
371 clear_opt(sbi, NOHEAP);
373 #ifdef CONFIG_F2FS_FS_XATTR
375 set_opt(sbi, XATTR_USER);
377 case Opt_nouser_xattr:
378 clear_opt(sbi, XATTR_USER);
380 case Opt_inline_xattr:
381 set_opt(sbi, INLINE_XATTR);
383 case Opt_noinline_xattr:
384 clear_opt(sbi, INLINE_XATTR);
388 f2fs_msg(sb, KERN_INFO,
389 "user_xattr options not supported");
391 case Opt_nouser_xattr:
392 f2fs_msg(sb, KERN_INFO,
393 "nouser_xattr options not supported");
395 case Opt_inline_xattr:
396 f2fs_msg(sb, KERN_INFO,
397 "inline_xattr options not supported");
399 case Opt_noinline_xattr:
400 f2fs_msg(sb, KERN_INFO,
401 "noinline_xattr options not supported");
404 #ifdef CONFIG_F2FS_FS_POSIX_ACL
406 set_opt(sbi, POSIX_ACL);
409 clear_opt(sbi, POSIX_ACL);
413 f2fs_msg(sb, KERN_INFO, "acl options not supported");
416 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
419 case Opt_active_logs:
420 if (args->from && match_int(args, &arg))
422 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
424 sbi->active_logs = arg;
426 case Opt_disable_ext_identify:
427 set_opt(sbi, DISABLE_EXT_IDENTIFY);
429 case Opt_inline_data:
430 set_opt(sbi, INLINE_DATA);
432 case Opt_inline_dentry:
433 set_opt(sbi, INLINE_DENTRY);
435 case Opt_noinline_dentry:
436 clear_opt(sbi, INLINE_DENTRY);
438 case Opt_flush_merge:
439 set_opt(sbi, FLUSH_MERGE);
441 case Opt_noflush_merge:
442 clear_opt(sbi, FLUSH_MERGE);
445 set_opt(sbi, NOBARRIER);
448 set_opt(sbi, FASTBOOT);
450 case Opt_extent_cache:
451 set_opt(sbi, EXTENT_CACHE);
453 case Opt_noextent_cache:
454 clear_opt(sbi, EXTENT_CACHE);
456 case Opt_noinline_data:
457 clear_opt(sbi, INLINE_DATA);
460 set_opt(sbi, DATA_FLUSH);
463 name = match_strdup(&args[0]);
467 if (strlen(name) == 8 &&
468 !strncmp(name, "adaptive", 8)) {
469 if (f2fs_sb_mounted_blkzoned(sb)) {
470 f2fs_msg(sb, KERN_WARNING,
471 "adaptive mode is not allowed with "
472 "zoned block device feature");
476 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
477 } else if (strlen(name) == 3 &&
478 !strncmp(name, "lfs", 3)) {
479 set_opt_mode(sbi, F2FS_MOUNT_LFS);
486 case Opt_io_size_bits:
487 if (args->from && match_int(args, &arg))
489 if (arg > __ilog2_u32(BIO_MAX_PAGES)) {
490 f2fs_msg(sb, KERN_WARNING,
491 "Not support %d, larger than %d",
492 1 << arg, BIO_MAX_PAGES);
495 sbi->write_io_size_bits = arg;
497 case Opt_fault_injection:
498 if (args->from && match_int(args, &arg))
500 #ifdef CONFIG_F2FS_FAULT_INJECTION
501 f2fs_build_fault_attr(sbi, arg);
502 set_opt(sbi, FAULT_INJECTION);
504 f2fs_msg(sb, KERN_INFO,
505 "FAULT_INJECTION was not selected");
509 sb->s_flags |= MS_LAZYTIME;
512 sb->s_flags &= ~MS_LAZYTIME;
517 set_opt(sbi, USRQUOTA);
520 set_opt(sbi, GRPQUOTA);
523 set_opt(sbi, PRJQUOTA);
526 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
531 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
536 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
540 case Opt_offusrjquota:
541 ret = f2fs_clear_qf_name(sb, USRQUOTA);
545 case Opt_offgrpjquota:
546 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
550 case Opt_offprjjquota:
551 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
555 case Opt_jqfmt_vfsold:
556 sbi->s_jquota_fmt = QFMT_VFS_OLD;
558 case Opt_jqfmt_vfsv0:
559 sbi->s_jquota_fmt = QFMT_VFS_V0;
561 case Opt_jqfmt_vfsv1:
562 sbi->s_jquota_fmt = QFMT_VFS_V1;
565 clear_opt(sbi, QUOTA);
566 clear_opt(sbi, USRQUOTA);
567 clear_opt(sbi, GRPQUOTA);
568 clear_opt(sbi, PRJQUOTA);
578 case Opt_offusrjquota:
579 case Opt_offgrpjquota:
580 case Opt_offprjjquota:
581 case Opt_jqfmt_vfsold:
582 case Opt_jqfmt_vfsv0:
583 case Opt_jqfmt_vfsv1:
585 f2fs_msg(sb, KERN_INFO,
586 "quota operations not supported");
590 f2fs_msg(sb, KERN_ERR,
591 "Unrecognized mount option \"%s\" or missing value",
597 if (f2fs_check_quota_options(sbi))
601 if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) {
602 f2fs_msg(sb, KERN_ERR,
603 "Should set mode=lfs with %uKB-sized IO",
604 F2FS_IO_SIZE_KB(sbi));
610 static struct inode *f2fs_alloc_inode(struct super_block *sb)
612 struct f2fs_inode_info *fi;
614 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
618 init_once((void *) fi);
620 /* Initialize f2fs-specific inode info */
621 fi->vfs_inode.i_version = 1;
622 atomic_set(&fi->dirty_pages, 0);
623 fi->i_current_depth = 1;
625 init_rwsem(&fi->i_sem);
626 INIT_LIST_HEAD(&fi->dirty_list);
627 INIT_LIST_HEAD(&fi->gdirty_list);
628 INIT_LIST_HEAD(&fi->inmem_pages);
629 mutex_init(&fi->inmem_lock);
630 init_rwsem(&fi->dio_rwsem[READ]);
631 init_rwsem(&fi->dio_rwsem[WRITE]);
632 init_rwsem(&fi->i_mmap_sem);
633 init_rwsem(&fi->i_xattr_sem);
636 memset(&fi->i_dquot, 0, sizeof(fi->i_dquot));
637 fi->i_reserved_quota = 0;
639 /* Will be used by directory only */
640 fi->i_dir_level = F2FS_SB(sb)->dir_level;
642 return &fi->vfs_inode;
645 static int f2fs_drop_inode(struct inode *inode)
649 * This is to avoid a deadlock condition like below.
650 * writeback_single_inode(inode)
651 * - f2fs_write_data_page
652 * - f2fs_gc -> iput -> evict
653 * - inode_wait_for_writeback(inode)
655 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
656 if (!inode->i_nlink && !is_bad_inode(inode)) {
657 /* to avoid evict_inode call simultaneously */
658 atomic_inc(&inode->i_count);
659 spin_unlock(&inode->i_lock);
661 /* some remained atomic pages should discarded */
662 if (f2fs_is_atomic_file(inode))
663 drop_inmem_pages(inode);
665 /* should remain fi->extent_tree for writepage */
666 f2fs_destroy_extent_node(inode);
668 sb_start_intwrite(inode->i_sb);
669 f2fs_i_size_write(inode, 0);
671 if (F2FS_HAS_BLOCKS(inode))
672 f2fs_truncate(inode);
674 sb_end_intwrite(inode->i_sb);
676 fscrypt_put_encryption_info(inode, NULL);
677 spin_lock(&inode->i_lock);
678 atomic_dec(&inode->i_count);
680 trace_f2fs_drop_inode(inode, 0);
683 ret = generic_drop_inode(inode);
684 trace_f2fs_drop_inode(inode, ret);
688 int f2fs_inode_dirtied(struct inode *inode, bool sync)
690 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
693 spin_lock(&sbi->inode_lock[DIRTY_META]);
694 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
697 set_inode_flag(inode, FI_DIRTY_INODE);
698 stat_inc_dirty_inode(sbi, DIRTY_META);
700 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
701 list_add_tail(&F2FS_I(inode)->gdirty_list,
702 &sbi->inode_list[DIRTY_META]);
703 inc_page_count(sbi, F2FS_DIRTY_IMETA);
705 spin_unlock(&sbi->inode_lock[DIRTY_META]);
709 void f2fs_inode_synced(struct inode *inode)
711 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
713 spin_lock(&sbi->inode_lock[DIRTY_META]);
714 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
715 spin_unlock(&sbi->inode_lock[DIRTY_META]);
718 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
719 list_del_init(&F2FS_I(inode)->gdirty_list);
720 dec_page_count(sbi, F2FS_DIRTY_IMETA);
722 clear_inode_flag(inode, FI_DIRTY_INODE);
723 clear_inode_flag(inode, FI_AUTO_RECOVER);
724 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
725 spin_unlock(&sbi->inode_lock[DIRTY_META]);
729 * f2fs_dirty_inode() is called from __mark_inode_dirty()
731 * We should call set_dirty_inode to write the dirty inode through write_inode.
733 static void f2fs_dirty_inode(struct inode *inode, int flags)
735 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
737 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
738 inode->i_ino == F2FS_META_INO(sbi))
741 if (flags == I_DIRTY_TIME)
744 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
745 clear_inode_flag(inode, FI_AUTO_RECOVER);
747 f2fs_inode_dirtied(inode, false);
750 static void f2fs_i_callback(struct rcu_head *head)
752 struct inode *inode = container_of(head, struct inode, i_rcu);
753 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
756 static void f2fs_destroy_inode(struct inode *inode)
758 call_rcu(&inode->i_rcu, f2fs_i_callback);
761 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
763 percpu_counter_destroy(&sbi->alloc_valid_block_count);
764 percpu_counter_destroy(&sbi->total_valid_inode_count);
767 static void destroy_device_list(struct f2fs_sb_info *sbi)
771 for (i = 0; i < sbi->s_ndevs; i++) {
772 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
773 #ifdef CONFIG_BLK_DEV_ZONED
774 kfree(FDEV(i).blkz_type);
780 static void f2fs_put_super(struct super_block *sb)
782 struct f2fs_sb_info *sbi = F2FS_SB(sb);
785 /* unregister procfs/sysfs entries in advance to avoid race case */
786 f2fs_unregister_sysfs(sbi);
788 f2fs_quota_off_umount(sb);
790 /* prevent remaining shrinker jobs */
791 mutex_lock(&sbi->umount_mutex);
794 * We don't need to do checkpoint when superblock is clean.
795 * But, the previous checkpoint was not done by umount, it needs to do
796 * clean checkpoint again.
798 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
799 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
800 struct cp_control cpc = {
803 write_checkpoint(sbi, &cpc);
806 /* be sure to wait for any on-going discard commands */
807 f2fs_wait_discard_bios(sbi, true);
809 if (f2fs_discard_en(sbi) && !sbi->discard_blks) {
810 struct cp_control cpc = {
811 .reason = CP_UMOUNT | CP_TRIMMED,
813 write_checkpoint(sbi, &cpc);
816 /* write_checkpoint can update stat informaion */
817 f2fs_destroy_stats(sbi);
820 * normally superblock is clean, so we need to release this.
821 * In addition, EIO will skip do checkpoint, we need this as well.
823 release_ino_entry(sbi, true);
825 f2fs_leave_shrinker(sbi);
826 mutex_unlock(&sbi->umount_mutex);
828 /* our cp_error case, we can wait for any writeback page */
829 f2fs_flush_merged_writes(sbi);
831 iput(sbi->node_inode);
832 iput(sbi->meta_inode);
834 /* destroy f2fs internal modules */
835 destroy_node_manager(sbi);
836 destroy_segment_manager(sbi);
840 sb->s_fs_info = NULL;
841 if (sbi->s_chksum_driver)
842 crypto_free_shash(sbi->s_chksum_driver);
843 kfree(sbi->raw_super);
845 destroy_device_list(sbi);
846 mempool_destroy(sbi->write_io_dummy);
848 for (i = 0; i < MAXQUOTAS; i++)
849 kfree(sbi->s_qf_names[i]);
851 destroy_percpu_info(sbi);
852 for (i = 0; i < NR_PAGE_TYPE; i++)
853 kfree(sbi->write_io[i]);
857 int f2fs_sync_fs(struct super_block *sb, int sync)
859 struct f2fs_sb_info *sbi = F2FS_SB(sb);
862 trace_f2fs_sync_fs(sb, sync);
864 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
868 struct cp_control cpc;
870 cpc.reason = __get_cp_reason(sbi);
872 mutex_lock(&sbi->gc_mutex);
873 err = write_checkpoint(sbi, &cpc);
874 mutex_unlock(&sbi->gc_mutex);
876 f2fs_trace_ios(NULL, 1);
881 static int f2fs_freeze(struct super_block *sb)
883 if (f2fs_readonly(sb))
886 /* IO error happened before */
887 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
890 /* must be clean, since sync_filesystem() was already called */
891 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
896 static int f2fs_unfreeze(struct super_block *sb)
902 static int f2fs_statfs_project(struct super_block *sb,
903 kprojid_t projid, struct kstatfs *buf)
910 qid = make_kqid_projid(projid);
911 dquot = dqget(sb, qid);
913 return PTR_ERR(dquot);
914 spin_lock(&dq_data_lock);
916 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
917 dquot->dq_dqb.dqb_bhardlimit);
919 limit >>= sb->s_blocksize_bits;
921 if (limit && buf->f_blocks > limit) {
922 curblock = (dquot->dq_dqb.dqb_curspace +
923 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
924 buf->f_blocks = limit;
925 buf->f_bfree = buf->f_bavail =
926 (buf->f_blocks > curblock) ?
927 (buf->f_blocks - curblock) : 0;
930 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
931 dquot->dq_dqb.dqb_ihardlimit);
933 if (limit && buf->f_files > limit) {
934 buf->f_files = limit;
936 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
937 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
940 spin_unlock(&dq_data_lock);
946 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
948 struct super_block *sb = dentry->d_sb;
949 struct f2fs_sb_info *sbi = F2FS_SB(sb);
950 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
951 block_t total_count, user_block_count, start_count, ovp_count;
952 u64 avail_node_count;
954 total_count = le64_to_cpu(sbi->raw_super->block_count);
955 user_block_count = sbi->user_block_count;
956 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
957 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
958 buf->f_type = F2FS_SUPER_MAGIC;
959 buf->f_bsize = sbi->blocksize;
961 buf->f_blocks = total_count - start_count;
962 buf->f_bfree = user_block_count - valid_user_blocks(sbi) + ovp_count;
963 buf->f_bavail = user_block_count - valid_user_blocks(sbi) -
964 sbi->reserved_blocks;
966 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
968 if (avail_node_count > user_block_count) {
969 buf->f_files = user_block_count;
970 buf->f_ffree = buf->f_bavail;
972 buf->f_files = avail_node_count;
973 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
977 buf->f_namelen = F2FS_NAME_LEN;
978 buf->f_fsid.val[0] = (u32)id;
979 buf->f_fsid.val[1] = (u32)(id >> 32);
982 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
983 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
984 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
990 static inline void f2fs_show_quota_options(struct seq_file *seq,
991 struct super_block *sb)
994 struct f2fs_sb_info *sbi = F2FS_SB(sb);
996 if (sbi->s_jquota_fmt) {
999 switch (sbi->s_jquota_fmt) {
1010 seq_printf(seq, ",jqfmt=%s", fmtname);
1013 if (sbi->s_qf_names[USRQUOTA])
1014 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1016 if (sbi->s_qf_names[GRPQUOTA])
1017 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1019 if (sbi->s_qf_names[PRJQUOTA])
1020 seq_show_option(seq, "prjjquota", sbi->s_qf_names[PRJQUOTA]);
1024 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1026 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1028 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1029 if (test_opt(sbi, FORCE_FG_GC))
1030 seq_printf(seq, ",background_gc=%s", "sync");
1032 seq_printf(seq, ",background_gc=%s", "on");
1034 seq_printf(seq, ",background_gc=%s", "off");
1036 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1037 seq_puts(seq, ",disable_roll_forward");
1038 if (test_opt(sbi, DISCARD))
1039 seq_puts(seq, ",discard");
1040 if (test_opt(sbi, NOHEAP))
1041 seq_puts(seq, ",no_heap");
1043 seq_puts(seq, ",heap");
1044 #ifdef CONFIG_F2FS_FS_XATTR
1045 if (test_opt(sbi, XATTR_USER))
1046 seq_puts(seq, ",user_xattr");
1048 seq_puts(seq, ",nouser_xattr");
1049 if (test_opt(sbi, INLINE_XATTR))
1050 seq_puts(seq, ",inline_xattr");
1052 seq_puts(seq, ",noinline_xattr");
1054 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1055 if (test_opt(sbi, POSIX_ACL))
1056 seq_puts(seq, ",acl");
1058 seq_puts(seq, ",noacl");
1060 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1061 seq_puts(seq, ",disable_ext_identify");
1062 if (test_opt(sbi, INLINE_DATA))
1063 seq_puts(seq, ",inline_data");
1065 seq_puts(seq, ",noinline_data");
1066 if (test_opt(sbi, INLINE_DENTRY))
1067 seq_puts(seq, ",inline_dentry");
1069 seq_puts(seq, ",noinline_dentry");
1070 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1071 seq_puts(seq, ",flush_merge");
1072 if (test_opt(sbi, NOBARRIER))
1073 seq_puts(seq, ",nobarrier");
1074 if (test_opt(sbi, FASTBOOT))
1075 seq_puts(seq, ",fastboot");
1076 if (test_opt(sbi, EXTENT_CACHE))
1077 seq_puts(seq, ",extent_cache");
1079 seq_puts(seq, ",noextent_cache");
1080 if (test_opt(sbi, DATA_FLUSH))
1081 seq_puts(seq, ",data_flush");
1083 seq_puts(seq, ",mode=");
1084 if (test_opt(sbi, ADAPTIVE))
1085 seq_puts(seq, "adaptive");
1086 else if (test_opt(sbi, LFS))
1087 seq_puts(seq, "lfs");
1088 seq_printf(seq, ",active_logs=%u", sbi->active_logs);
1089 if (F2FS_IO_SIZE_BITS(sbi))
1090 seq_printf(seq, ",io_size=%uKB", F2FS_IO_SIZE_KB(sbi));
1091 #ifdef CONFIG_F2FS_FAULT_INJECTION
1092 if (test_opt(sbi, FAULT_INJECTION))
1093 seq_printf(seq, ",fault_injection=%u",
1094 sbi->fault_info.inject_rate);
1097 if (test_opt(sbi, QUOTA))
1098 seq_puts(seq, ",quota");
1099 if (test_opt(sbi, USRQUOTA))
1100 seq_puts(seq, ",usrquota");
1101 if (test_opt(sbi, GRPQUOTA))
1102 seq_puts(seq, ",grpquota");
1103 if (test_opt(sbi, PRJQUOTA))
1104 seq_puts(seq, ",prjquota");
1106 f2fs_show_quota_options(seq, sbi->sb);
1111 static void default_options(struct f2fs_sb_info *sbi)
1113 /* init some FS parameters */
1114 sbi->active_logs = NR_CURSEG_TYPE;
1116 set_opt(sbi, BG_GC);
1117 set_opt(sbi, INLINE_XATTR);
1118 set_opt(sbi, INLINE_DATA);
1119 set_opt(sbi, INLINE_DENTRY);
1120 set_opt(sbi, EXTENT_CACHE);
1121 set_opt(sbi, NOHEAP);
1122 sbi->sb->s_flags |= MS_LAZYTIME;
1123 set_opt(sbi, FLUSH_MERGE);
1124 if (f2fs_sb_mounted_blkzoned(sbi->sb)) {
1125 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1126 set_opt(sbi, DISCARD);
1128 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1131 #ifdef CONFIG_F2FS_FS_XATTR
1132 set_opt(sbi, XATTR_USER);
1134 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1135 set_opt(sbi, POSIX_ACL);
1138 #ifdef CONFIG_F2FS_FAULT_INJECTION
1139 f2fs_build_fault_attr(sbi, 0);
1143 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1145 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1146 struct f2fs_mount_info org_mount_opt;
1147 unsigned long old_sb_flags;
1148 int err, active_logs;
1149 bool need_restart_gc = false;
1150 bool need_stop_gc = false;
1151 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1152 #ifdef CONFIG_F2FS_FAULT_INJECTION
1153 struct f2fs_fault_info ffi = sbi->fault_info;
1157 char *s_qf_names[MAXQUOTAS];
1162 * Save the old mount options in case we
1163 * need to restore them.
1165 org_mount_opt = sbi->mount_opt;
1166 old_sb_flags = sb->s_flags;
1167 active_logs = sbi->active_logs;
1170 s_jquota_fmt = sbi->s_jquota_fmt;
1171 for (i = 0; i < MAXQUOTAS; i++) {
1172 if (sbi->s_qf_names[i]) {
1173 s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
1175 if (!s_qf_names[i]) {
1176 for (j = 0; j < i; j++)
1177 kfree(s_qf_names[j]);
1181 s_qf_names[i] = NULL;
1186 /* recover superblocks we couldn't write due to previous RO mount */
1187 if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1188 err = f2fs_commit_super(sbi, false);
1189 f2fs_msg(sb, KERN_INFO,
1190 "Try to recover all the superblocks, ret: %d", err);
1192 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1195 default_options(sbi);
1197 /* parse mount options */
1198 err = parse_options(sb, data);
1203 * Previous and new state of filesystem is RO,
1204 * so skip checking GC and FLUSH_MERGE conditions.
1206 if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
1209 if (!f2fs_readonly(sb) && (*flags & MS_RDONLY)) {
1210 err = dquot_suspend(sb, -1);
1214 /* dquot_resume needs RW */
1215 sb->s_flags &= ~MS_RDONLY;
1216 dquot_resume(sb, -1);
1219 /* disallow enable/disable extent_cache dynamically */
1220 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1222 f2fs_msg(sbi->sb, KERN_WARNING,
1223 "switch extent_cache option is not allowed");
1228 * We stop the GC thread if FS is mounted as RO
1229 * or if background_gc = off is passed in mount
1230 * option. Also sync the filesystem.
1232 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
1233 if (sbi->gc_thread) {
1234 stop_gc_thread(sbi);
1235 need_restart_gc = true;
1237 } else if (!sbi->gc_thread) {
1238 err = start_gc_thread(sbi);
1241 need_stop_gc = true;
1244 if (*flags & MS_RDONLY) {
1245 writeback_inodes_sb(sb, WB_REASON_SYNC);
1248 set_sbi_flag(sbi, SBI_IS_DIRTY);
1249 set_sbi_flag(sbi, SBI_IS_CLOSE);
1250 f2fs_sync_fs(sb, 1);
1251 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1255 * We stop issue flush thread if FS is mounted as RO
1256 * or if flush_merge is not passed in mount option.
1258 if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1259 clear_opt(sbi, FLUSH_MERGE);
1260 destroy_flush_cmd_control(sbi, false);
1262 err = create_flush_cmd_control(sbi);
1268 /* Release old quota file names */
1269 for (i = 0; i < MAXQUOTAS; i++)
1270 kfree(s_qf_names[i]);
1272 /* Update the POSIXACL Flag */
1273 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1274 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1278 if (need_restart_gc) {
1279 if (start_gc_thread(sbi))
1280 f2fs_msg(sbi->sb, KERN_WARNING,
1281 "background gc thread has stopped");
1282 } else if (need_stop_gc) {
1283 stop_gc_thread(sbi);
1287 sbi->s_jquota_fmt = s_jquota_fmt;
1288 for (i = 0; i < MAXQUOTAS; i++) {
1289 kfree(sbi->s_qf_names[i]);
1290 sbi->s_qf_names[i] = s_qf_names[i];
1293 sbi->mount_opt = org_mount_opt;
1294 sbi->active_logs = active_logs;
1295 sb->s_flags = old_sb_flags;
1296 #ifdef CONFIG_F2FS_FAULT_INJECTION
1297 sbi->fault_info = ffi;
1303 /* Read data from quotafile */
1304 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1305 size_t len, loff_t off)
1307 struct inode *inode = sb_dqopt(sb)->files[type];
1308 struct address_space *mapping = inode->i_mapping;
1309 block_t blkidx = F2FS_BYTES_TO_BLK(off);
1310 int offset = off & (sb->s_blocksize - 1);
1313 loff_t i_size = i_size_read(inode);
1320 if (off + len > i_size)
1323 while (toread > 0) {
1324 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1326 page = read_mapping_page(mapping, blkidx, NULL);
1328 return PTR_ERR(page);
1332 if (unlikely(page->mapping != mapping)) {
1333 f2fs_put_page(page, 1);
1336 if (unlikely(!PageUptodate(page))) {
1337 f2fs_put_page(page, 1);
1341 kaddr = kmap_atomic(page);
1342 memcpy(data, kaddr + offset, tocopy);
1343 kunmap_atomic(kaddr);
1344 f2fs_put_page(page, 1);
1354 /* Write to quotafile */
1355 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1356 const char *data, size_t len, loff_t off)
1358 struct inode *inode = sb_dqopt(sb)->files[type];
1359 struct address_space *mapping = inode->i_mapping;
1360 const struct address_space_operations *a_ops = mapping->a_ops;
1361 int offset = off & (sb->s_blocksize - 1);
1362 size_t towrite = len;
1368 while (towrite > 0) {
1369 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1372 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1377 kaddr = kmap_atomic(page);
1378 memcpy(kaddr + offset, data, tocopy);
1379 kunmap_atomic(kaddr);
1380 flush_dcache_page(page);
1382 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1394 inode->i_mtime = inode->i_ctime = current_time(inode);
1395 f2fs_mark_inode_dirty_sync(inode, false);
1396 return len - towrite;
1399 static struct dquot **f2fs_get_dquots(struct inode *inode)
1401 return F2FS_I(inode)->i_dquot;
1404 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1406 return &F2FS_I(inode)->i_reserved_quota;
1409 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1411 return dquot_quota_on_mount(sbi->sb, sbi->s_qf_names[type],
1412 sbi->s_jquota_fmt, type);
1415 void f2fs_enable_quota_files(struct f2fs_sb_info *sbi)
1419 for (i = 0; i < MAXQUOTAS; i++) {
1420 if (sbi->s_qf_names[i]) {
1421 ret = f2fs_quota_on_mount(sbi, i);
1423 f2fs_msg(sbi->sb, KERN_ERR,
1424 "Cannot turn on journaled "
1425 "quota: error %d", ret);
1430 static int f2fs_quota_sync(struct super_block *sb, int type)
1432 struct quota_info *dqopt = sb_dqopt(sb);
1436 ret = dquot_writeback_dquots(sb, type);
1441 * Now when everything is written we can discard the pagecache so
1442 * that userspace sees the changes.
1444 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
1445 if (type != -1 && cnt != type)
1447 if (!sb_has_quota_active(sb, cnt))
1450 ret = filemap_write_and_wait(dqopt->files[cnt]->i_mapping);
1454 inode_lock(dqopt->files[cnt]);
1455 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
1456 inode_unlock(dqopt->files[cnt]);
1461 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
1462 const struct path *path)
1464 struct inode *inode;
1467 err = f2fs_quota_sync(sb, type);
1471 err = dquot_quota_on(sb, type, format_id, path);
1475 inode = d_inode(path->dentry);
1478 F2FS_I(inode)->i_flags |= FS_NOATIME_FL | FS_IMMUTABLE_FL;
1479 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1480 S_NOATIME | S_IMMUTABLE);
1481 inode_unlock(inode);
1482 f2fs_mark_inode_dirty_sync(inode, false);
1487 static int f2fs_quota_off(struct super_block *sb, int type)
1489 struct inode *inode = sb_dqopt(sb)->files[type];
1492 if (!inode || !igrab(inode))
1493 return dquot_quota_off(sb, type);
1495 err = f2fs_quota_sync(sb, type);
1499 err = dquot_quota_off(sb, type);
1504 F2FS_I(inode)->i_flags &= ~(FS_NOATIME_FL | FS_IMMUTABLE_FL);
1505 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1506 inode_unlock(inode);
1507 f2fs_mark_inode_dirty_sync(inode, false);
1513 void f2fs_quota_off_umount(struct super_block *sb)
1518 for (type = 0; type < MAXQUOTAS; type++) {
1519 err = f2fs_quota_off(sb, type);
1521 int ret = dquot_quota_off(sb, type);
1523 f2fs_msg(sb, KERN_ERR,
1524 "Fail to turn off disk quota "
1525 "(type: %d, err: %d, ret:%d), Please "
1526 "run fsck to fix it.", type, err, ret);
1527 set_sbi_flag(F2FS_SB(sb), SBI_NEED_FSCK);
1531 * In case of checkpoint=disable, we must flush quota blocks.
1532 * This can cause NULL exception for node_inode in end_io, since
1533 * put_super already dropped it.
1535 sync_filesystem(sb);
1538 int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
1540 *projid = F2FS_I(inode)->i_projid;
1544 static const struct dquot_operations f2fs_quota_operations = {
1545 .get_reserved_space = f2fs_get_reserved_space,
1546 .write_dquot = dquot_commit,
1547 .acquire_dquot = dquot_acquire,
1548 .release_dquot = dquot_release,
1549 .mark_dirty = dquot_mark_dquot_dirty,
1550 .write_info = dquot_commit_info,
1551 .alloc_dquot = dquot_alloc,
1552 .destroy_dquot = dquot_destroy,
1553 .get_projid = f2fs_get_projid,
1554 .get_next_id = dquot_get_next_id,
1557 static const struct quotactl_ops f2fs_quotactl_ops = {
1558 .quota_on = f2fs_quota_on,
1559 .quota_off = f2fs_quota_off,
1560 .quota_sync = f2fs_quota_sync,
1561 .get_state = dquot_get_state,
1562 .set_info = dquot_set_dqinfo,
1563 .get_dqblk = dquot_get_dqblk,
1564 .set_dqblk = dquot_set_dqblk,
1565 .get_nextdqblk = dquot_get_next_dqblk,
1568 void f2fs_quota_off_umount(struct super_block *sb)
1573 static const struct super_operations f2fs_sops = {
1574 .alloc_inode = f2fs_alloc_inode,
1575 .drop_inode = f2fs_drop_inode,
1576 .destroy_inode = f2fs_destroy_inode,
1577 .write_inode = f2fs_write_inode,
1578 .dirty_inode = f2fs_dirty_inode,
1579 .show_options = f2fs_show_options,
1581 .quota_read = f2fs_quota_read,
1582 .quota_write = f2fs_quota_write,
1583 .get_dquots = f2fs_get_dquots,
1585 .evict_inode = f2fs_evict_inode,
1586 .put_super = f2fs_put_super,
1587 .sync_fs = f2fs_sync_fs,
1588 .freeze_fs = f2fs_freeze,
1589 .unfreeze_fs = f2fs_unfreeze,
1590 .statfs = f2fs_statfs,
1591 .remount_fs = f2fs_remount,
1594 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1595 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1597 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1598 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1602 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1605 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1606 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1607 ctx, len, fs_data, XATTR_CREATE);
1610 static unsigned f2fs_max_namelen(struct inode *inode)
1612 return S_ISLNK(inode->i_mode) ?
1613 inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1616 static const struct fscrypt_operations f2fs_cryptops = {
1617 .key_prefix = "f2fs:",
1618 .get_context = f2fs_get_context,
1619 .set_context = f2fs_set_context,
1620 .is_encrypted = f2fs_encrypted_inode,
1621 .empty_dir = f2fs_empty_dir,
1622 .max_namelen = f2fs_max_namelen,
1625 static const struct fscrypt_operations f2fs_cryptops = {
1626 .is_encrypted = f2fs_encrypted_inode,
1630 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1631 u64 ino, u32 generation)
1633 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1634 struct inode *inode;
1636 if (check_nid_range(sbi, ino))
1637 return ERR_PTR(-ESTALE);
1640 * f2fs_iget isn't quite right if the inode is currently unallocated!
1641 * However f2fs_iget currently does appropriate checks to handle stale
1642 * inodes so everything is OK.
1644 inode = f2fs_iget(sb, ino);
1646 return ERR_CAST(inode);
1647 if (unlikely(generation && inode->i_generation != generation)) {
1648 /* we didn't find the right inode.. */
1650 return ERR_PTR(-ESTALE);
1655 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1656 int fh_len, int fh_type)
1658 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1659 f2fs_nfs_get_inode);
1662 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1663 int fh_len, int fh_type)
1665 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1666 f2fs_nfs_get_inode);
1669 static const struct export_operations f2fs_export_ops = {
1670 .fh_to_dentry = f2fs_fh_to_dentry,
1671 .fh_to_parent = f2fs_fh_to_parent,
1672 .get_parent = f2fs_get_parent,
1675 static loff_t max_file_blocks(void)
1678 loff_t leaf_count = ADDRS_PER_BLOCK;
1681 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
1682 * F2FS_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
1683 * space in inode.i_addr, it will be more safe to reassign
1687 /* two direct node blocks */
1688 result += (leaf_count * 2);
1690 /* two indirect node blocks */
1691 leaf_count *= NIDS_PER_BLOCK;
1692 result += (leaf_count * 2);
1694 /* one double indirect node block */
1695 leaf_count *= NIDS_PER_BLOCK;
1696 result += leaf_count;
1701 static int __f2fs_commit_super(struct buffer_head *bh,
1702 struct f2fs_super_block *super)
1706 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
1707 set_buffer_uptodate(bh);
1708 set_buffer_dirty(bh);
1711 /* it's rare case, we can do fua all the time */
1712 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
1715 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
1716 struct buffer_head *bh)
1718 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1719 (bh->b_data + F2FS_SUPER_OFFSET);
1720 struct super_block *sb = sbi->sb;
1721 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
1722 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
1723 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
1724 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
1725 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
1726 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
1727 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
1728 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
1729 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
1730 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
1731 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1732 u32 segment_count = le32_to_cpu(raw_super->segment_count);
1733 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1734 u64 main_end_blkaddr = main_blkaddr +
1735 (segment_count_main << log_blocks_per_seg);
1736 u64 seg_end_blkaddr = segment0_blkaddr +
1737 (segment_count << log_blocks_per_seg);
1739 if (segment0_blkaddr != cp_blkaddr) {
1740 f2fs_msg(sb, KERN_INFO,
1741 "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1742 segment0_blkaddr, cp_blkaddr);
1746 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
1748 f2fs_msg(sb, KERN_INFO,
1749 "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1750 cp_blkaddr, sit_blkaddr,
1751 segment_count_ckpt << log_blocks_per_seg);
1755 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
1757 f2fs_msg(sb, KERN_INFO,
1758 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1759 sit_blkaddr, nat_blkaddr,
1760 segment_count_sit << log_blocks_per_seg);
1764 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
1766 f2fs_msg(sb, KERN_INFO,
1767 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1768 nat_blkaddr, ssa_blkaddr,
1769 segment_count_nat << log_blocks_per_seg);
1773 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
1775 f2fs_msg(sb, KERN_INFO,
1776 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1777 ssa_blkaddr, main_blkaddr,
1778 segment_count_ssa << log_blocks_per_seg);
1782 if (main_end_blkaddr > seg_end_blkaddr) {
1783 f2fs_msg(sb, KERN_INFO,
1784 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1787 (segment_count << log_blocks_per_seg),
1788 segment_count_main << log_blocks_per_seg);
1790 } else if (main_end_blkaddr < seg_end_blkaddr) {
1794 /* fix in-memory information all the time */
1795 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
1796 segment0_blkaddr) >> log_blocks_per_seg);
1798 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
1799 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1802 err = __f2fs_commit_super(bh, NULL);
1803 res = err ? "failed" : "done";
1805 f2fs_msg(sb, KERN_INFO,
1806 "Fix alignment : %s, start(%u) end(%u) block(%u)",
1809 (segment_count << log_blocks_per_seg),
1810 segment_count_main << log_blocks_per_seg);
1817 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
1818 struct buffer_head *bh)
1820 block_t segment_count, segs_per_sec, secs_per_zone;
1821 block_t total_sections, blocks_per_seg;
1822 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1823 (bh->b_data + F2FS_SUPER_OFFSET);
1824 struct super_block *sb = sbi->sb;
1825 unsigned int blocksize;
1827 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
1828 f2fs_msg(sb, KERN_INFO,
1829 "Magic Mismatch, valid(0x%x) - read(0x%x)",
1830 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
1834 /* Currently, support only 4KB page cache size */
1835 if (F2FS_BLKSIZE != PAGE_SIZE) {
1836 f2fs_msg(sb, KERN_INFO,
1837 "Invalid page_cache_size (%lu), supports only 4KB\n",
1839 return -EFSCORRUPTED;
1842 /* Currently, support only 4KB block size */
1843 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
1844 if (blocksize != F2FS_BLKSIZE) {
1845 f2fs_msg(sb, KERN_INFO,
1846 "Invalid blocksize (%u), supports only 4KB\n",
1848 return -EFSCORRUPTED;
1851 /* check log blocks per segment */
1852 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
1853 f2fs_msg(sb, KERN_INFO,
1854 "Invalid log blocks per segment (%u)\n",
1855 le32_to_cpu(raw_super->log_blocks_per_seg));
1856 return -EFSCORRUPTED;
1859 /* Currently, support 512/1024/2048/4096 bytes sector size */
1860 if (le32_to_cpu(raw_super->log_sectorsize) >
1861 F2FS_MAX_LOG_SECTOR_SIZE ||
1862 le32_to_cpu(raw_super->log_sectorsize) <
1863 F2FS_MIN_LOG_SECTOR_SIZE) {
1864 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
1865 le32_to_cpu(raw_super->log_sectorsize));
1866 return -EFSCORRUPTED;
1868 if (le32_to_cpu(raw_super->log_sectors_per_block) +
1869 le32_to_cpu(raw_super->log_sectorsize) !=
1870 F2FS_MAX_LOG_SECTOR_SIZE) {
1871 f2fs_msg(sb, KERN_INFO,
1872 "Invalid log sectors per block(%u) log sectorsize(%u)",
1873 le32_to_cpu(raw_super->log_sectors_per_block),
1874 le32_to_cpu(raw_super->log_sectorsize));
1875 return -EFSCORRUPTED;
1878 segment_count = le32_to_cpu(raw_super->segment_count);
1879 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1880 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1881 total_sections = le32_to_cpu(raw_super->section_count);
1883 /* blocks_per_seg should be 512, given the above check */
1884 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
1886 if (segment_count > F2FS_MAX_SEGMENT ||
1887 segment_count < F2FS_MIN_SEGMENTS) {
1888 f2fs_msg(sb, KERN_INFO,
1889 "Invalid segment count (%u)",
1891 return -EFSCORRUPTED;
1894 if (total_sections > segment_count ||
1895 total_sections < F2FS_MIN_SEGMENTS ||
1896 segs_per_sec > segment_count || !segs_per_sec) {
1897 f2fs_msg(sb, KERN_INFO,
1898 "Invalid segment/section count (%u, %u x %u)",
1899 segment_count, total_sections, segs_per_sec);
1900 return -EFSCORRUPTED;
1903 if ((segment_count / segs_per_sec) < total_sections) {
1904 f2fs_msg(sb, KERN_INFO,
1905 "Small segment_count (%u < %u * %u)",
1906 segment_count, segs_per_sec, total_sections);
1907 return -EFSCORRUPTED;
1910 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
1911 f2fs_msg(sb, KERN_INFO,
1912 "Wrong segment_count / block_count (%u > %llu)",
1913 segment_count, le64_to_cpu(raw_super->block_count));
1914 return -EFSCORRUPTED;
1917 if (secs_per_zone > total_sections || !secs_per_zone) {
1918 f2fs_msg(sb, KERN_INFO,
1919 "Wrong secs_per_zone / total_sections (%u, %u)",
1920 secs_per_zone, total_sections);
1921 return -EFSCORRUPTED;
1923 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION) {
1924 f2fs_msg(sb, KERN_INFO,
1925 "Corrupted extension count (%u > %u)",
1926 le32_to_cpu(raw_super->extension_count),
1927 F2FS_MAX_EXTENSION);
1928 return -EFSCORRUPTED;
1931 if (le32_to_cpu(raw_super->cp_payload) >
1932 (blocks_per_seg - F2FS_CP_PACKS)) {
1933 f2fs_msg(sb, KERN_INFO,
1934 "Insane cp_payload (%u > %u)",
1935 le32_to_cpu(raw_super->cp_payload),
1936 blocks_per_seg - F2FS_CP_PACKS);
1937 return -EFSCORRUPTED;
1940 /* check reserved ino info */
1941 if (le32_to_cpu(raw_super->node_ino) != 1 ||
1942 le32_to_cpu(raw_super->meta_ino) != 2 ||
1943 le32_to_cpu(raw_super->root_ino) != 3) {
1944 f2fs_msg(sb, KERN_INFO,
1945 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1946 le32_to_cpu(raw_super->node_ino),
1947 le32_to_cpu(raw_super->meta_ino),
1948 le32_to_cpu(raw_super->root_ino));
1949 return -EFSCORRUPTED;
1952 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1953 if (sanity_check_area_boundary(sbi, bh))
1954 return -EFSCORRUPTED;
1959 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1961 unsigned int total, fsmeta;
1962 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1963 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1964 unsigned int ovp_segments, reserved_segments;
1965 unsigned int main_segs, blocks_per_seg;
1966 unsigned int sit_segs, nat_segs;
1967 unsigned int sit_bitmap_size, nat_bitmap_size;
1968 unsigned int log_blocks_per_seg;
1969 unsigned int segment_count_main;
1970 unsigned int cp_pack_start_sum, cp_payload;
1971 block_t user_block_count;
1974 total = le32_to_cpu(raw_super->segment_count);
1975 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
1976 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
1978 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
1980 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
1981 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
1983 if (unlikely(fsmeta >= total))
1986 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
1987 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
1989 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
1990 ovp_segments == 0 || reserved_segments == 0)) {
1991 f2fs_msg(sbi->sb, KERN_ERR,
1992 "Wrong layout: check mkfs.f2fs version");
1996 user_block_count = le64_to_cpu(ckpt->user_block_count);
1997 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1998 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1999 if (!user_block_count || user_block_count >=
2000 segment_count_main << log_blocks_per_seg) {
2001 f2fs_msg(sbi->sb, KERN_ERR,
2002 "Wrong user_block_count: %u", user_block_count);
2006 main_segs = le32_to_cpu(raw_super->segment_count_main);
2007 blocks_per_seg = sbi->blocks_per_seg;
2009 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2010 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2011 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2013 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2014 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2015 le32_to_cpu(ckpt->cur_node_segno[j])) {
2016 f2fs_msg(sbi->sb, KERN_ERR,
2017 "Node segment (%u, %u) has the same "
2019 le32_to_cpu(ckpt->cur_node_segno[i]));
2024 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2025 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2026 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2028 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2029 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2030 le32_to_cpu(ckpt->cur_data_segno[j])) {
2031 f2fs_msg(sbi->sb, KERN_ERR,
2032 "Data segment (%u, %u) has the same "
2034 le32_to_cpu(ckpt->cur_data_segno[i]));
2039 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2040 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
2041 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2042 le32_to_cpu(ckpt->cur_data_segno[j])) {
2043 f2fs_msg(sbi->sb, KERN_ERR,
2044 "Node segment (%u) and Data segment (%u)"
2045 " has the same segno: %u", i, j,
2046 le32_to_cpu(ckpt->cur_node_segno[i]));
2052 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2053 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2055 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2056 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2057 f2fs_msg(sbi->sb, KERN_ERR,
2058 "Wrong bitmap size: sit: %u, nat:%u",
2059 sit_bitmap_size, nat_bitmap_size);
2063 cp_pack_start_sum = __start_sum_addr(sbi);
2064 cp_payload = __cp_payload(sbi);
2065 if (cp_pack_start_sum < cp_payload + 1 ||
2066 cp_pack_start_sum > blocks_per_seg - 1 -
2068 f2fs_msg(sbi->sb, KERN_ERR,
2069 "Wrong cp_pack_start_sum: %u",
2074 if (unlikely(f2fs_cp_error(sbi))) {
2075 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
2081 static void init_sb_info(struct f2fs_sb_info *sbi)
2083 struct f2fs_super_block *raw_super = sbi->raw_super;
2086 sbi->log_sectors_per_block =
2087 le32_to_cpu(raw_super->log_sectors_per_block);
2088 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2089 sbi->blocksize = 1 << sbi->log_blocksize;
2090 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2091 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2092 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2093 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2094 sbi->total_sections = le32_to_cpu(raw_super->section_count);
2095 sbi->total_node_count =
2096 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2097 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2098 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2099 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2100 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2101 sbi->cur_victim_sec = NULL_SECNO;
2102 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2104 sbi->dir_level = DEF_DIR_LEVEL;
2105 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2106 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2107 clear_sbi_flag(sbi, SBI_NEED_FSCK);
2109 for (i = 0; i < NR_COUNT_TYPE; i++)
2110 atomic_set(&sbi->nr_pages[i], 0);
2112 atomic_set(&sbi->wb_sync_req, 0);
2114 INIT_LIST_HEAD(&sbi->s_list);
2115 mutex_init(&sbi->umount_mutex);
2116 for (i = 0; i < NR_PAGE_TYPE - 1; i++)
2117 for (j = HOT; j < NR_TEMP_TYPE; j++)
2118 mutex_init(&sbi->wio_mutex[i][j]);
2119 spin_lock_init(&sbi->cp_lock);
2122 static int init_percpu_info(struct f2fs_sb_info *sbi)
2126 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2130 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
2133 percpu_counter_destroy(&sbi->alloc_valid_block_count);
2138 #ifdef CONFIG_BLK_DEV_ZONED
2139 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2141 struct block_device *bdev = FDEV(devi).bdev;
2142 sector_t nr_sectors = bdev->bd_part->nr_sects;
2143 sector_t sector = 0;
2144 struct blk_zone *zones;
2145 unsigned int i, nr_zones;
2149 if (!f2fs_sb_mounted_blkzoned(sbi->sb))
2152 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2153 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2155 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2156 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2157 __ilog2_u32(sbi->blocks_per_blkz))
2159 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2160 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2161 sbi->log_blocks_per_blkz;
2162 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2163 FDEV(devi).nr_blkz++;
2165 FDEV(devi).blkz_type = kmalloc(FDEV(devi).nr_blkz, GFP_KERNEL);
2166 if (!FDEV(devi).blkz_type)
2169 #define F2FS_REPORT_NR_ZONES 4096
2171 zones = kcalloc(F2FS_REPORT_NR_ZONES, sizeof(struct blk_zone),
2176 /* Get block zones type */
2177 while (zones && sector < nr_sectors) {
2179 nr_zones = F2FS_REPORT_NR_ZONES;
2180 err = blkdev_report_zones(bdev, sector,
2190 for (i = 0; i < nr_zones; i++) {
2191 FDEV(devi).blkz_type[n] = zones[i].type;
2192 sector += zones[i].len;
2204 * Read f2fs raw super block.
2205 * Because we have two copies of super block, so read both of them
2206 * to get the first valid one. If any one of them is broken, we pass
2207 * them recovery flag back to the caller.
2209 static int read_raw_super_block(struct f2fs_sb_info *sbi,
2210 struct f2fs_super_block **raw_super,
2211 int *valid_super_block, int *recovery)
2213 struct super_block *sb = sbi->sb;
2215 struct buffer_head *bh;
2216 struct f2fs_super_block *super;
2219 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2223 for (block = 0; block < 2; block++) {
2224 bh = sb_bread(sb, block);
2226 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
2232 /* sanity checking of raw super */
2233 err = sanity_check_raw_super(sbi, bh);
2235 f2fs_msg(sb, KERN_ERR,
2236 "Can't find valid F2FS filesystem in %dth superblock",
2243 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2245 *valid_super_block = block;
2251 /* Fail to read any one of the superblocks*/
2255 /* No valid superblock */
2264 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
2266 struct buffer_head *bh;
2269 if ((recover && f2fs_readonly(sbi->sb)) ||
2270 bdev_read_only(sbi->sb->s_bdev)) {
2271 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2275 /* write back-up superblock first */
2276 bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1);
2279 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2282 /* if we are in recovery path, skip writing valid superblock */
2286 /* write current valid superblock */
2287 bh = sb_getblk(sbi->sb, sbi->valid_super_block);
2290 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2295 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
2297 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2298 unsigned int max_devices = MAX_DEVICES;
2301 /* Initialize single device information */
2302 if (!RDEV(0).path[0]) {
2303 if (!bdev_is_zoned(sbi->sb->s_bdev))
2309 * Initialize multiple devices information, or single
2310 * zoned block device information.
2312 sbi->devs = kcalloc(max_devices, sizeof(struct f2fs_dev_info),
2317 for (i = 0; i < max_devices; i++) {
2319 if (i > 0 && !RDEV(i).path[0])
2322 if (max_devices == 1) {
2323 /* Single zoned block device mount */
2325 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
2326 sbi->sb->s_mode, sbi->sb->s_type);
2328 /* Multi-device mount */
2329 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
2330 FDEV(i).total_segments =
2331 le32_to_cpu(RDEV(i).total_segments);
2333 FDEV(i).start_blk = 0;
2334 FDEV(i).end_blk = FDEV(i).start_blk +
2335 (FDEV(i).total_segments <<
2336 sbi->log_blocks_per_seg) - 1 +
2337 le32_to_cpu(raw_super->segment0_blkaddr);
2339 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
2340 FDEV(i).end_blk = FDEV(i).start_blk +
2341 (FDEV(i).total_segments <<
2342 sbi->log_blocks_per_seg) - 1;
2344 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
2345 sbi->sb->s_mode, sbi->sb->s_type);
2347 if (IS_ERR(FDEV(i).bdev))
2348 return PTR_ERR(FDEV(i).bdev);
2350 /* to release errored devices */
2351 sbi->s_ndevs = i + 1;
2353 #ifdef CONFIG_BLK_DEV_ZONED
2354 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
2355 !f2fs_sb_mounted_blkzoned(sbi->sb)) {
2356 f2fs_msg(sbi->sb, KERN_ERR,
2357 "Zoned block device feature not enabled\n");
2360 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
2361 if (init_blkz_info(sbi, i)) {
2362 f2fs_msg(sbi->sb, KERN_ERR,
2363 "Failed to initialize F2FS blkzone information");
2366 if (max_devices == 1)
2368 f2fs_msg(sbi->sb, KERN_INFO,
2369 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
2371 FDEV(i).total_segments,
2372 FDEV(i).start_blk, FDEV(i).end_blk,
2373 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
2374 "Host-aware" : "Host-managed");
2378 f2fs_msg(sbi->sb, KERN_INFO,
2379 "Mount Device [%2d]: %20s, %8u, %8x - %8x",
2381 FDEV(i).total_segments,
2382 FDEV(i).start_blk, FDEV(i).end_blk);
2384 f2fs_msg(sbi->sb, KERN_INFO,
2385 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
2389 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
2391 struct f2fs_sb_info *sbi;
2392 struct f2fs_super_block *raw_super;
2395 bool retry = true, need_fsck = false;
2396 char *options = NULL;
2397 int recovery, i, valid_super_block;
2398 struct curseg_info *seg_i;
2403 valid_super_block = -1;
2406 /* allocate memory for f2fs-specific super block info */
2407 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
2413 /* Load the checksum driver */
2414 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
2415 if (IS_ERR(sbi->s_chksum_driver)) {
2416 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
2417 err = PTR_ERR(sbi->s_chksum_driver);
2418 sbi->s_chksum_driver = NULL;
2422 /* set a block size */
2423 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
2424 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
2428 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
2433 sb->s_fs_info = sbi;
2434 sbi->raw_super = raw_super;
2436 /* precompute checksum seed for metadata */
2437 if (f2fs_sb_has_inode_chksum(sb))
2438 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
2439 sizeof(raw_super->uuid));
2442 * The BLKZONED feature indicates that the drive was formatted with
2443 * zone alignment optimization. This is optional for host-aware
2444 * devices, but mandatory for host-managed zoned block devices.
2446 #ifndef CONFIG_BLK_DEV_ZONED
2447 if (f2fs_sb_mounted_blkzoned(sb)) {
2448 f2fs_msg(sb, KERN_ERR,
2449 "Zoned block device support is not enabled\n");
2454 default_options(sbi);
2455 /* parse mount options */
2456 options = kstrdup((const char *)data, GFP_KERNEL);
2457 if (data && !options) {
2462 err = parse_options(sb, options);
2466 sbi->max_file_blocks = max_file_blocks();
2467 sb->s_maxbytes = sbi->max_file_blocks <<
2468 le32_to_cpu(raw_super->log_blocksize);
2469 sb->s_max_links = F2FS_LINK_MAX;
2470 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2473 sb->dq_op = &f2fs_quota_operations;
2474 sb->s_qcop = &f2fs_quotactl_ops;
2475 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
2478 sb->s_op = &f2fs_sops;
2479 sb->s_cop = &f2fs_cryptops;
2480 sb->s_xattr = f2fs_xattr_handlers;
2481 sb->s_export_op = &f2fs_export_ops;
2482 sb->s_magic = F2FS_SUPER_MAGIC;
2483 sb->s_time_gran = 1;
2484 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2485 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
2486 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
2488 /* init f2fs-specific super block info */
2489 sbi->valid_super_block = valid_super_block;
2490 mutex_init(&sbi->gc_mutex);
2491 mutex_init(&sbi->cp_mutex);
2492 init_rwsem(&sbi->node_write);
2493 init_rwsem(&sbi->node_change);
2495 /* disallow all the data/node/meta page writes */
2496 set_sbi_flag(sbi, SBI_POR_DOING);
2497 spin_lock_init(&sbi->stat_lock);
2499 /* init iostat info */
2500 spin_lock_init(&sbi->iostat_lock);
2501 sbi->iostat_enable = false;
2503 for (i = 0; i < NR_PAGE_TYPE; i++) {
2504 int n = (i == META) ? 1: NR_TEMP_TYPE;
2507 sbi->write_io[i] = kmalloc(n * sizeof(struct f2fs_bio_info),
2509 if (!sbi->write_io[i]) {
2514 for (j = HOT; j < n; j++) {
2515 init_rwsem(&sbi->write_io[i][j].io_rwsem);
2516 sbi->write_io[i][j].sbi = sbi;
2517 sbi->write_io[i][j].bio = NULL;
2518 spin_lock_init(&sbi->write_io[i][j].io_lock);
2519 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
2523 init_rwsem(&sbi->cp_rwsem);
2524 init_waitqueue_head(&sbi->cp_wait);
2527 err = init_percpu_info(sbi);
2531 if (F2FS_IO_SIZE(sbi) > 1) {
2532 sbi->write_io_dummy =
2533 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
2534 if (!sbi->write_io_dummy) {
2540 /* get an inode for meta space */
2541 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
2542 if (IS_ERR(sbi->meta_inode)) {
2543 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
2544 err = PTR_ERR(sbi->meta_inode);
2548 err = get_valid_checkpoint(sbi);
2550 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
2551 goto free_meta_inode;
2554 /* Initialize device list */
2555 err = f2fs_scan_devices(sbi);
2557 f2fs_msg(sb, KERN_ERR, "Failed to find devices");
2561 sbi->total_valid_node_count =
2562 le32_to_cpu(sbi->ckpt->valid_node_count);
2563 percpu_counter_set(&sbi->total_valid_inode_count,
2564 le32_to_cpu(sbi->ckpt->valid_inode_count));
2565 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
2566 sbi->total_valid_block_count =
2567 le64_to_cpu(sbi->ckpt->valid_block_count);
2568 sbi->last_valid_block_count = sbi->total_valid_block_count;
2569 sbi->reserved_blocks = 0;
2571 for (i = 0; i < NR_INODE_TYPE; i++) {
2572 INIT_LIST_HEAD(&sbi->inode_list[i]);
2573 spin_lock_init(&sbi->inode_lock[i]);
2576 init_extent_cache_info(sbi);
2578 init_ino_entry_info(sbi);
2580 /* setup f2fs internal modules */
2581 err = build_segment_manager(sbi);
2583 f2fs_msg(sb, KERN_ERR,
2584 "Failed to initialize F2FS segment manager");
2587 err = build_node_manager(sbi);
2589 f2fs_msg(sb, KERN_ERR,
2590 "Failed to initialize F2FS node manager");
2594 /* For write statistics */
2595 if (sb->s_bdev->bd_part)
2596 sbi->sectors_written_start =
2597 (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
2599 /* Read accumulated write IO statistics if exists */
2600 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
2601 if (__exist_node_summaries(sbi))
2602 sbi->kbytes_written =
2603 le64_to_cpu(seg_i->journal->info.kbytes_written);
2605 build_gc_manager(sbi);
2607 /* get an inode for node space */
2608 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
2609 if (IS_ERR(sbi->node_inode)) {
2610 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
2611 err = PTR_ERR(sbi->node_inode);
2615 f2fs_join_shrinker(sbi);
2617 err = f2fs_build_stats(sbi);
2621 /* read root inode and dentry */
2622 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
2624 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
2625 err = PTR_ERR(root);
2626 goto free_node_inode;
2628 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2631 goto free_node_inode;
2634 sb->s_root = d_make_root(root); /* allocate root dentry */
2637 goto free_root_inode;
2640 err = f2fs_register_sysfs(sbi);
2642 goto free_root_inode;
2644 /* if there are nt orphan nodes free them */
2645 err = recover_orphan_inodes(sbi);
2649 /* recover fsynced data */
2650 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
2652 * mount should be failed, when device has readonly mode, and
2653 * previous checkpoint was not done by clean system shutdown.
2655 if (bdev_read_only(sb->s_bdev) &&
2656 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
2662 set_sbi_flag(sbi, SBI_NEED_FSCK);
2667 err = recover_fsync_data(sbi, false);
2670 f2fs_msg(sb, KERN_ERR,
2671 "Cannot recover all fsync data errno=%d", err);
2675 err = recover_fsync_data(sbi, true);
2677 if (!f2fs_readonly(sb) && err > 0) {
2679 f2fs_msg(sb, KERN_ERR,
2680 "Need to recover fsync data");
2685 /* recover_fsync_data() cleared this already */
2686 clear_sbi_flag(sbi, SBI_POR_DOING);
2689 * If filesystem is not mounted as read-only then
2690 * do start the gc_thread.
2692 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
2693 /* After POR, we can run background GC thread.*/
2694 err = start_gc_thread(sbi);
2700 /* recover broken superblock */
2702 err = f2fs_commit_super(sbi, true);
2703 f2fs_msg(sb, KERN_INFO,
2704 "Try to recover %dth superblock, ret: %d",
2705 sbi->valid_super_block ? 1 : 2, err);
2708 f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx",
2709 cur_cp_version(F2FS_CKPT(sbi)));
2710 f2fs_update_time(sbi, CP_TIME);
2711 f2fs_update_time(sbi, REQ_TIME);
2715 f2fs_sync_inode_meta(sbi);
2717 * Some dirty meta pages can be produced by recover_orphan_inodes()
2718 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
2719 * followed by write_checkpoint() through f2fs_write_node_pages(), which
2720 * falls into an infinite loop in sync_meta_pages().
2722 truncate_inode_pages_final(META_MAPPING(sbi));
2724 f2fs_unregister_sysfs(sbi);
2729 truncate_inode_pages_final(NODE_MAPPING(sbi));
2730 mutex_lock(&sbi->umount_mutex);
2731 release_ino_entry(sbi, true);
2732 f2fs_leave_shrinker(sbi);
2733 iput(sbi->node_inode);
2734 mutex_unlock(&sbi->umount_mutex);
2735 f2fs_destroy_stats(sbi);
2737 destroy_node_manager(sbi);
2739 destroy_segment_manager(sbi);
2741 destroy_device_list(sbi);
2744 make_bad_inode(sbi->meta_inode);
2745 iput(sbi->meta_inode);
2747 mempool_destroy(sbi->write_io_dummy);
2749 for (i = 0; i < NR_PAGE_TYPE; i++)
2750 kfree(sbi->write_io[i]);
2751 destroy_percpu_info(sbi);
2753 for (i = 0; i < MAXQUOTAS; i++)
2754 kfree(sbi->s_qf_names[i]);
2760 if (sbi->s_chksum_driver)
2761 crypto_free_shash(sbi->s_chksum_driver);
2764 /* give only one another chance */
2767 shrink_dcache_sb(sb);
2773 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
2774 const char *dev_name, void *data)
2776 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
2779 static void kill_f2fs_super(struct super_block *sb)
2782 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
2783 stop_gc_thread(F2FS_SB(sb));
2784 stop_discard_thread(F2FS_SB(sb));
2786 kill_block_super(sb);
2789 static struct file_system_type f2fs_fs_type = {
2790 .owner = THIS_MODULE,
2792 .mount = f2fs_mount,
2793 .kill_sb = kill_f2fs_super,
2794 .fs_flags = FS_REQUIRES_DEV,
2796 MODULE_ALIAS_FS("f2fs");
2798 static int __init init_inodecache(void)
2800 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
2801 sizeof(struct f2fs_inode_info), 0,
2802 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
2803 if (!f2fs_inode_cachep)
2808 static void destroy_inodecache(void)
2811 * Make sure all delayed rcu free inodes are flushed before we
2815 kmem_cache_destroy(f2fs_inode_cachep);
2818 static int __init init_f2fs_fs(void)
2822 if (PAGE_SIZE != F2FS_BLKSIZE) {
2823 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
2824 PAGE_SIZE, F2FS_BLKSIZE);
2828 f2fs_build_trace_ios();
2830 err = init_inodecache();
2833 err = create_node_manager_caches();
2835 goto free_inodecache;
2836 err = create_segment_manager_caches();
2838 goto free_node_manager_caches;
2839 err = create_checkpoint_caches();
2841 goto free_segment_manager_caches;
2842 err = create_extent_cache();
2844 goto free_checkpoint_caches;
2845 err = f2fs_init_sysfs();
2847 goto free_extent_cache;
2848 err = register_shrinker(&f2fs_shrinker_info);
2851 err = register_filesystem(&f2fs_fs_type);
2854 err = f2fs_create_root_stats();
2856 goto free_filesystem;
2860 unregister_filesystem(&f2fs_fs_type);
2862 unregister_shrinker(&f2fs_shrinker_info);
2866 destroy_extent_cache();
2867 free_checkpoint_caches:
2868 destroy_checkpoint_caches();
2869 free_segment_manager_caches:
2870 destroy_segment_manager_caches();
2871 free_node_manager_caches:
2872 destroy_node_manager_caches();
2874 destroy_inodecache();
2879 static void __exit exit_f2fs_fs(void)
2881 f2fs_destroy_root_stats();
2882 unregister_filesystem(&f2fs_fs_type);
2883 unregister_shrinker(&f2fs_shrinker_info);
2885 destroy_extent_cache();
2886 destroy_checkpoint_caches();
2887 destroy_segment_manager_caches();
2888 destroy_node_manager_caches();
2889 destroy_inodecache();
2890 f2fs_destroy_trace_ios();
2893 module_init(init_f2fs_fs)
2894 module_exit(exit_f2fs_fs)
2896 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
2897 MODULE_DESCRIPTION("Flash Friendly File System");
2898 MODULE_LICENSE("GPL");
2899 MODULE_SOFTDEP("pre: crc32");