pstore: Actually give up during locking failure

Without a return after the pr_err(), dumps will collide when two threads
call pstore_dump() at the same time.
Signed-off-by: NLiu Hailong <liuhailong5@huawei.com>
Signed-off-by: NLi Pengcheng <lipengcheng8@huawei.com>
Signed-off-by: NLi Zhong <lizhong11@hisilicon.com>
[kees: improved commit message]
Signed-off-by: NKees Cook <keescook@chromium.org>

fs/pstore/platform.c

@@ -493,6 +493,7 @@ static void pstore_dump(struct kmsg_dumper *dumper,
 		if (!is_locked) {
 			pr_err("pstore dump routine blocked in %s path, may corrupt error record\n"
 				       , in_nmi() ? "NMI" : why);
+			return;
 		}
 	} else {
 		spin_lock_irqsave(&psinfo->buf_lock, flags);