Enable buttons GPIO passthrough

[Wang, Kuiying]

Hi Linus,
User space supporting is necessary in OpenBMC.
OpenBMC has to support IPMI command(user space network command) to dynamically disable/enable all the buttons in the front panel of a server, which need pass-through to support be disabled/enabled dynamically in the user space. 
Once pass-through of power/reset button is enabled, IPMI command cannot handle the power any more.
So correct logic should be that pass-through of buttons is enabled by default, once OpenBMC got the related ipmi command (like power on/off/reset/disable front panel command), disable pass-through firstly, do the corresponding operation and then enable pass-through back.


Thanks,
Kwin.


-----Original Message-----
From: Linus Walleij [mailto:linus.walleij at linaro.org] 
Sent: Monday, February 11, 2019 4:08 PM
To: Andrew Jeffery <andrew at aj.id.au>
Cc: Wang, Kuiying <kuiying.wang at intel.com>; Joel Stanley <joel at jms.id.au>; Thomas Petazzoni <thomas.petazzoni at bootlin.com>; open list:GPIO SUBSYSTEM <linux-gpio at vger.kernel.org>; Andrew Geissler <geissonator at gmail.com>; OpenBMC Maillist <openbmc at lists.ozlabs.org>; Mauery, Vernon <vernon.mauery at intel.com>; Feist, James <james.feist at intel.com>; Yoo, Jae Hyun <jae.hyun.yoo at intel.com>; Nguyen, Hai V <hai.v.nguyen at intel.com>; Khetan, Sharad <sharad.khetan at intel.com>; Benjamin Herrenschmidt <benh at kernel.crashing.org>
Subject: Re: Enable buttons GPIO passthrough

On Mon, Feb 11, 2019 at 5:54 AM Andrew Jeffery <andrew at aj.id.au> wrote:
> On Fri, 8 Feb 2019, at 22:31, Linus Walleij wrote:

> > > +static int aspeed_gpio_pass_through(struct gpio_chip *chip, unsigned int offset,
> > > +                                   unsigned long usecs) {
> > > +       printk("kwin::aspeed_gpio_pass_through is called");
> > > +       return 0;
> > > +}
> >
> > I guess you want to actually implement this too :D
>
> Something to keep in mind is that the ASPEED pinctrl already supports 
> configuring the passthrough bits. What we don't want is to cause some 
> inconsistent state between the GPIO and pinmux subsystems by just whacking the bits directly.
>
> Looks like we should be able to make it work with PIN_MAP_MUX_GROUP() 
> and using some custom name for the state (maybe "passthrough")? It's a 
> half-baked thought though so I'm not sure if it's feasible let alone 
> acceptable, but hopefully Linus can comment.

Since its a real odd thing this "passthrough" I would certainly prefer if you could keep it as a custom thing on the pin control side of things. Then I suppose that you set it up using the device tree and all is fine?

What I'd like to hear from Kwin is a convincing story why Aspeed SoCs need to do this from userspace, as that seems to be his goal.

Yours,
Linus Walleij