Why chinese character gash appear when i use spark textFile?

[Jone Zhang]

var textFile = sc.textFile("xxx");
textFile.first();
res1: String = 1.0     100733314       18_?????:100733314
8919173c6d49abfab02853458247e584        1:129:18_?????:1.0


hadoop fs -cat xxx
1.0    100733314       18_百度输入法:100733314 8919173c6d49abfab02853458247e584
     1:129:18_百度输入法:1.0

Why  chinese character gash appear when i use spark textFile?
The code of hdfs file is utf-8.


Thanks