The 10/14/2021 13:47, Philipp Zabel wrote:

Hi Horatiu,

Hi Philipp

quoted

+     reset = devm_reset_control_get_shared(&pdev->dev, "switch");

Please use devm_reset_control_get_optional_shared() for optional resets
and handle errors. That will return NULL in case the optional reset is
not specified in the device tree.

I will do that.

It seems weird to me that the reset input to the GPIO controller is
called "switch" reset. You can request a single unnamed reset with

        reset = devm_reset_control_get_shared(&pdev->dev, NULL);

although that would limit future extendability in case this driver will
ever require to handle multiple separate resets. If you decide to
request the reset control by name, the yaml binding should specify the
same name.

I think this requires a little bit more explanation from my side. On
lan966x we are facing the following issue. When we try to reset just the
switch core then also the sgpio device was reset and there was no way
from HW perspective to prevent this.

So our solutions was to create a reset driver[1] that will be triggered
only one time, by the sgpio driver or by the switch driver. That is the
reason why it was called "switch" reset. And that is the purpose of this
patch to allow the sgpio driver to reset the switch in case is probed
before the switch driver so it would not get reset after that.

quoted

+     if (!IS_ERR(reset))
+             reset_control_reset(reset);

With optional resets, this can be just:

        reset_control_reset(reset);

Great I will do that.

regards
Philipp

[1] https://lore.kernel.org/lkml/20211013073807.2282230-1-horatiu.vultur@microchip.com/ (local)

-- 
/Horatiu

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel