Convert __roundup_pow_of_two() to use fls_long() instead of fls() so
the code would be identical to that found in Linux. This fix also
allowed next patch(taken from Linux kernel verbatim) to apply cleanly.

Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>