f2fs: introduce mount option for fsync mode

Commit "0a007b97aad6"(f2fs: recover directory operations by fsync)
fixed xfstest generic/342 case, but it also increased the written
data and caused the performance degradation. In most cases, there's
no need to do so heavy fsync actually.

So we introduce new mount option "fsync_mode={posix,strict}" to
control the policy of fsync. "fsync_mode=posix" is set by default,
and means that f2fs uses a light fsync, which follows POSIX semantics.
And "fsync_mode=strict" means that it's a heavy fsync, which behaves
in line with xfs, ext4 and btrfs, where generic/342 will pass, but
the performance will regress.

Signed-off-by: Junling Zheng <zhengjunling@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>

5 files changed, 44 insertions, 5 deletions

diff --git a/fs/f2fs/dir.c b/fs/f2fs/dir.c
index 797eb05cb538..87c51709bc48 100644
--- a/fs/f2fs/dir.c
+++ b/fs/f2fs/dir.c
@@ -704,7 +704,8 @@ void f2fs_delete_entry(struct f2fs_dir_entry *dentry, struct page *page,
 
 	f2fs_update_time(F2FS_I_SB(dir), REQ_TIME);
 
-	add_ino_entry(F2FS_I_SB(dir), dir->i_ino, TRANS_DIR_INO);
+	if (F2FS_I_SB(dir)->fsync_mode == FSYNC_MODE_STRICT)
+		add_ino_entry(F2FS_I_SB(dir), dir->i_ino, TRANS_DIR_INO);
 
 	if (f2fs_has_inline_dentry(dir))
 		return f2fs_delete_inline_entry(dentry, page, dir, inode);
diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index 0e47fe93c718..3c79b3565cb3 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -1052,6 +1052,11 @@ enum {
 	ALLOC_MODE_REUSE,	/* reuse segments as much as possible */
 };
 
+enum fsync_mode {
+	FSYNC_MODE_POSIX,	/* fsync follows posix semantics */
+	FSYNC_MODE_STRICT,	/* fsync behaves in line with ext4 */
+};
+
 struct f2fs_sb_info {
 	struct super_block *sb;			/* pointer to VFS super block */
 	struct proc_dir_entry *s_proc;		/* proc entry */
@@ -1241,6 +1246,9 @@ struct f2fs_sb_info {
 
 	/* segment allocation policy */
 	int alloc_mode;
+
+	/* fsync policy */
+	int fsync_mode;
 };
 
 #ifdef CONFIG_F2FS_FAULT_INJECTION
diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
index 6a202e5a6386..99fa207fc310 100644
--- a/fs/f2fs/file.c
+++ b/fs/f2fs/file.c
@@ -165,7 +165,8 @@ static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
 		cp_reason = CP_FASTBOOT_MODE;
 	else if (sbi->active_logs == 2)
 		cp_reason = CP_SPEC_LOG_NUM;
-	else if (need_dentry_mark(sbi, inode->i_ino) &&
+	else if (sbi->fsync_mode == FSYNC_MODE_STRICT &&
+		need_dentry_mark(sbi, inode->i_ino) &&
 		exist_written_data(sbi, F2FS_I(inode)->i_pino, TRANS_DIR_INO))
 		cp_reason = CP_RECOVER_DIR;
 
diff --git a/fs/f2fs/namei.c b/fs/f2fs/namei.c
index 78ea0156f027..e9bb3a3a8b0a 100644
--- a/fs/f2fs/namei.c
+++ b/fs/f2fs/namei.c
@@ -970,7 +970,8 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry,
 			f2fs_put_page(old_dir_page, 0);
 		f2fs_i_links_write(old_dir, false);
 	}
-	add_ino_entry(sbi, new_dir->i_ino, TRANS_DIR_INO);
+	if (sbi->fsync_mode == FSYNC_MODE_STRICT)
+		add_ino_entry(sbi, new_dir->i_ino, TRANS_DIR_INO);
 
 	f2fs_unlock_op(sbi);
 
@@ -1120,8 +1121,10 @@ static int f2fs_cross_rename(struct inode *old_dir, struct dentry *old_dentry,
 	}
 	f2fs_mark_inode_dirty_sync(new_dir, false);
 
-	add_ino_entry(sbi, old_dir->i_ino, TRANS_DIR_INO);
-	add_ino_entry(sbi, new_dir->i_ino, TRANS_DIR_INO);
+	if (sbi->fsync_mode == FSYNC_MODE_STRICT) {
+		add_ino_entry(sbi, old_dir->i_ino, TRANS_DIR_INO);
+		add_ino_entry(sbi, new_dir->i_ino, TRANS_DIR_INO);
+	}
 
 	f2fs_unlock_op(sbi);
 
diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
index 36c5e582c1f2..2516dfa26a98 100644
--- a/fs/f2fs/super.c
+++ b/fs/f2fs/super.c
@@ -131,6 +131,7 @@ enum {
 	Opt_jqfmt_vfsv1,
 	Opt_whint,
 	Opt_alloc,
+	Opt_fsync,
 	Opt_err,
 };
 
@@ -186,6 +187,7 @@ static match_table_t f2fs_tokens = {
 	{Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
 	{Opt_whint, "whint_mode=%s"},
 	{Opt_alloc, "alloc_mode=%s"},
+	{Opt_fsync, "fsync_mode=%s"},
 	{Opt_err, NULL},
 };
 
@@ -719,6 +721,22 @@ static int parse_options(struct super_block *sb, char *options)
 			}
 			kfree(name);
 			break;
+		case Opt_fsync:
+			name = match_strdup(&args[0]);
+			if (!name)
+				return -ENOMEM;
+			if (strlen(name) == 5 &&
+					!strncmp(name, "posix", 5)) {
+				sbi->fsync_mode = FSYNC_MODE_POSIX;
+			} else if (strlen(name) == 6 &&
+					!strncmp(name, "strict", 6)) {
+				sbi->fsync_mode = FSYNC_MODE_STRICT;
+			} else {
+				kfree(name);
+				return -EINVAL;
+			}
+			kfree(name);
+			break;
 		default:
 			f2fs_msg(sb, KERN_ERR,
 				"Unrecognized mount option \"%s\" or missing value",
@@ -1287,6 +1305,11 @@ static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
 		seq_printf(seq, ",alloc_mode=%s", "default");
 	else if (sbi->alloc_mode == ALLOC_MODE_REUSE)
 		seq_printf(seq, ",alloc_mode=%s", "reuse");
+
+	if (sbi->fsync_mode == FSYNC_MODE_POSIX)
+		seq_printf(seq, ",fsync_mode=%s", "posix");
+	else if (sbi->fsync_mode == FSYNC_MODE_STRICT)
+		seq_printf(seq, ",fsync_mode=%s", "strict");
 	return 0;
 }
 
@@ -1297,6 +1320,7 @@ static void default_options(struct f2fs_sb_info *sbi)
 	sbi->inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
 	sbi->whint_mode = WHINT_MODE_OFF;
 	sbi->alloc_mode = ALLOC_MODE_DEFAULT;
+	sbi->fsync_mode = FSYNC_MODE_POSIX;
 	sbi->readdir_ra = 1;
 
 	set_opt(sbi, BG_GC);
@@ -1340,6 +1364,7 @@ static int f2fs_remount(struct super_block *sb, int *flags, char *data)
 	bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
 	int old_whint_mode = sbi->whint_mode;
 	int old_alloc_mode = sbi->alloc_mode;
+	int old_fsync_mode = sbi->fsync_mode;
 	int old_inline_xattr_size = sbi->inline_xattr_size;
 	block_t old_root_reserved_blocks = sbi->root_reserved_blocks;
 	kuid_t old_resuid = sbi->s_resuid;
@@ -1500,6 +1525,7 @@ restore_opts:
 	sbi->root_reserved_blocks = old_root_reserved_blocks;
 	sbi->inline_xattr_size = old_inline_xattr_size;
 	sbi->alloc_mode = old_alloc_mode;
+	sbi->fsync_mode = old_fsync_mode;
 	sbi->whint_mode = old_whint_mode;
 	sbi->mount_opt = org_mount_opt;
 	sbi->active_logs = active_logs;