Re: [PATCH v17 1/7] usb: misc: Add onboard_usb_hub driver

[Matthias Kaehlcke <mka@chromium.org>]

On Wed, Nov 17, 2021 at 04:11:34PM -0800, Doug Anderson wrote:
> Hi,
> 
> On Tue, Nov 16, 2021 at 12:07 PM Matthias Kaehlcke <mka@chromium.org> wrote:
> >
> > --- a/drivers/usb/misc/Kconfig
> > +++ b/drivers/usb/misc/Kconfig
> > @@ -284,3 +284,20 @@ config BRCM_USB_PINMAP
> >           This option enables support for remapping some USB external
> >           signals, which are typically on dedicated pins on the chip,
> >           to any gpio.
> > +
> > +config USB_ONBOARD_HUB
> > +       tristate "Onboard USB hub support"
> 
> Aren't you back to shenanigans now that you're being called straight
> from the USB core? What if you're a module and the USB core is
> builtin? It can't call you, right? ...or what if you're builtin but
> the USB core is a module (yeah, I know that sounds insane but I don't
> think anything technically prevents it)?

Indeed, a dependency involving USB host mode is needed, as previously
with xhci_plat.

> Can you just add a dependency here such that if the USB core is a
> module that you're a module and if the USB core is builtin that you're
> builtin?

I couldn't find a way to specify that in the config options of the driver
itself. I fear the dependency has to be specified in CONFIG_USB, like it
was done previously with USB_XHCI_PLATFORM:

https://patchwork.kernel.org/project/linux-usb/patch/20210813125146.v16.6.I7a3a7d9d2126c34079b1cab87aa0b2ec3030f9b7@changeid/

Hope that isn't controversial.

> > +void onboard_hub_create_pdevs(struct usb_device *parent_hub, struct list_head *pdev_list)
> > +{
> > +       int i;
> > +       struct device_node *np, *npc;
> > +       struct platform_device *pdev;
> > +       struct pdev_list_entry *pdle;
> > +
> > +       INIT_LIST_HEAD(pdev_list);
> > +
> > +       for (i = 1; i <= parent_hub->maxchild; i++) {
> > +               np = usb_of_get_device_node(parent_hub, i);
> > +               if (!np)
> > +                       continue;
> > +
> > +               if (!of_is_onboard_usb_hub(np))
> > +                       goto node_put;
> > +
> > +               npc = of_parse_phandle(np, "companion-hub", 0);
> > +               if (!npc)
> > +                       goto create_pdev;
> > +
> > +               pdev = of_find_device_by_node(npc);
> > +               of_node_put(npc);
> > +
> > +               if (pdev) {
> > +                       /* the companion hub already has a platform device, nothing to do here */
> > +                       put_device(&pdev->dev);
> > +                       goto node_put;
> > +               }
> > +
> > +create_pdev:
> 
> I don't really like this "goto". I'd rather just use an "if" test for
> the few lines even if the indentation gets to be a bit much.

Ok, I'll remove the "goto" in the next version.