This reverts commit b8c2b10a.

This patch was in my queue at the same time that a conversion of
the same driver from bool --> tristate was pending and merged.

That is commit 337ea0fb ("pinctrl: Turn AMD support to tristate")

Normally the conflict would show up in the build coverage I do,
however in this case an avoidable instance of linux/module.h in
linux/gpio/driver.h (!) causes the build failure to be masked and
instead the tristate gets built-in even for selected "=m".

In working on removing module.h from driver.h this issue was then
revealed (along with other implicit module.h assumptions in gpio,
and mfd -- which will be fixed separately.)

Cc: linux-gpio@vger.kernel.org
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Acked-by: Jean Delvare <jdelvare@suse.de>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>