ot_gpio_eg: remember input lines between resets

[jwnrt]

GPIO has a chardev interface for driving the inputs into QEMU, however we're currently clearing our memory of these lines at reset.

For bootstrapping Earlgrey we need to assert some pins and then reset the chip. I have restored the input lines on reset so we remember which pins were asserted.

[rivos-eblot]

I start to think that we are trying to solve these reset issues the wrong way.
I do not think it is the GPIO that should not be reset. There should be some persistent state, but that is not part of the GPIO that does get reset if I'm not mistaken.

The state is preserved on the board/chip package where some signal are maintained. On reset, the GPIO gets reset as other IPs, then samples what's on the external pins. The more we deviate from the actual HW, the more we'll get into trouble.

I think what is really needed here is a padring which should be part of the SoC and that the GPIO must be connected to it.

[jwnrt]

Yes, I think you're right. The padring should have the chardev interface that GPIO currently has and we should route those pads through pinmux and to the GPIO block. It would also be nice to have a "pin" for reset in the padring too.

Would you be willing to accept this patch to the GPIO block before we start moving everything to padring? We'd like to get bootstrapping working in the shorter term if possible.

[rivos-eblot]

Would you be willing to accept this patch to the GPIO block before we start moving everything to padring?

Can you only keep it for EG? I'm pretty sure if we use the DJ version we will get into trouble ( + add a new ticket for the padring work)

[rivos-eblot]

Can you update the format of the PR title as well? Thanks.

[rivos-eblot]

LGTM. Can you update the PR title with the "new" syntax? Thanks.