You can simply groupby the 3 levels and add dropna at the end:

print(t["c1"].groupby(level=[0,1,2]).first().dropna())

1  1  1    1.0
   2  9    2.0
2  1  7    4.0
4  2  2    6.0