On Wed, 2014-05-07 at 11:02 +0200, Johannes Berg wrote:

On Tue, 2014-05-06 at 18:14 +0200, Loic Poulain wrote:

quoted

+	if (gpiod_cansleep(rfkill->shutdown_gpio))
+		gpiod_set_value_cansleep(rfkill->shutdown_gpio, !blocked);
+	else
+		gpiod_set_value(rfkill->shutdown_gpio, !blocked);
+
+	if (gpiod_cansleep(rfkill->reset_gpio))
+		gpiod_set_value_cansleep(rfkill->reset_gpio, !blocked);
+	else
+		gpiod_set_value(rfkill->reset_gpio, !blocked);

Really? I mean, there's not even a fallback where the cansleep() API
calls the non-sleeping one if the cansleep isn't support? This is really
ugly.

Actually, it looks like unconditionally using the cansleep API should be
fine since it just calls the same internal helper function. Please
respin.

johannes