Re: [RFC PATCH 3/3] gpio: ej1x8: Add GPIO driver for Etron Tech Inc. EJ168/EJ188/EJ198

[Linus Walleij]

On Mon, Dec 21, 2020 at 4:28 PM Martin Blumenstingl
<martin.blumenstingl@xxxxxxxxxxxxxx> wrote:
> On Wed, Oct 7, 2020 at 9:44 PM Martin Blumenstingl
> <martin.blumenstingl@xxxxxxxxxxxxxx> wrote:
> [...]
> > > As noted on the earlier patches I think this should be folded into the
> > > existing XHCI USB driver in drivers/usb/host/xhci-pci.c or, if that
> > > gets messy, as a separate bolt-on, something like
> > > xhci-pci-gpio.[c|h] in the drivers/usb/host/* directory.
> > > You can use a Kconfig symbol for the GPIO portions or not.
> > OK, I will do that if there are no objections from other developers
> > I am intending to place the relevant code in xhci-pci-etron.c, similar
> > to what we already have with xhci-pci-renesas.c
>
> I tried this and unfortunately there's a catch.
> the nice thing about having a separate GPIO driver means that the
> xhci-pci driver doesn't need to know about it.

Since PCI devices have device-wide power management and things
like that I think that is a really dangerous idea.

What if the GPIO driver starts poking around in this PCI device
when the main driver is also probed and has put the device
into sleep state?

This type of set-up needs to be discussed with
the PCI maintainer to make sure it is safe.

> I implemented xhci-pci-etron.c and gave it a Kconfig option.
> xhci-pci is then calling into xhci-pci-etron (through some
> etron_xhci_pci_probe function).

This sounds about right.

> unfortunately this means that xhci-pci now depends on xhci-pci-etron.
> for xhci-pci-renesas this is fine (I think) because that part of the
> code is needed to get the xHCI controller going
> but for xhci-pci-etron this is a different story: the GPIO controller
> is entirely optional and only used on few devices

I might be naive but should it not be the other way around?
That xhci-pci-etron is dependent on xhci-pci? I imagine
it would be an optional add-on.

> my goal is (at some point in the future) to have the GPIO driver in OpenWrt.
> I am not sure if they would accept a patch where xhci-pci would then
> pull in the dependencies for that Etron controller, even though most
> boards don't need it.

Make sure the etron part is an additional module that can be
loaded after xhci-pci.

OpenWrt support optional modules to be compiled per-system.

Yours,
Linus Walleij