Re: [PATCH] usb: musb-new: Extend Allwinner quirk to newer SoCs

[Jagan Teki <jagan@amarulasolutions.com>]

On Tue, Apr 27, 2021 at 1:41 PM Andre Przywara <andre.przywara@arm.com> wrote:
>
> On Tue, 27 Apr 2021 02:37:20 +0200
> Marek Vasut <marex@denx.de> wrote:
>
> Hi,
>
> > On 4/27/21 2:03 AM, Andre Przywara wrote:
> > > As the comment in musb_regs.h describes, Allwinner saves the
> > > MUSB_CONFIGDATA register, which always return 0 on those SoCs.
> > >
> > > This is also true for the H6 and H616, so extend the quirk to those
> > > controllers as well.
> > >
> > > This fixes USB peripheral mode on H6 and H616 boards.
> > >
> > > Signed-off-by: Andre Przywara <andre.przywara@arm.com>
> > > ---
> > >   drivers/usb/musb-new/musb_regs.h | 3 ++-
> > >   1 file changed, 2 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/usb/musb-new/musb_regs.h b/drivers/usb/musb-new/musb_regs.h
> > > index c4d7203b851..bee1b715a95 100644
> > > --- a/drivers/usb/musb-new/musb_regs.h
> > > +++ b/drivers/usb/musb-new/musb_regs.h
> > > @@ -432,7 +432,8 @@ static inline u8 musb_read_ulpi_buscontrol(void __iomem *mbase)
> > >   static inline u8 musb_read_configdata(void __iomem *mbase)
> > >   {
> > >   #if defined CONFIG_MACH_SUN8I_A33 || defined CONFIG_MACH_SUN8I_A83T || \
> > > -           defined CONFIG_MACH_SUNXI_H3_H5 || defined CONFIG_MACH_SUN50I
> > > +   defined CONFIG_MACH_SUNXI_H3_H5 || defined CONFIG_MACH_SUN50I || \
> > > +   defined CONFIG_SUN50I_GEN_H6
> >
> > Isn't there some better solution then ever-growing list of macros to
> > check, like e.g. a single CONFIG_MACH_SUNXI ?
>
> I was wondering the same, but I think this does not apply to the older
> SoCs (we use ARCH_SUNXI in the two functions above and below, so I
> guess the differentiation here is deliberate). I will test this later.
>
> So we could probably use the quirk also for the older, working(?) SoCs,
> but I am not sure we should do that. CONFIG_SUN50I_GEN_H6 is already a
> symbol covering multiple SoCs, so ideally we won't need to add many
> more.
> I can have a look if we have other checks like that in the code, then
> maybe define a collective symbol for newer SoCs?

The better approach is to support config_read via musb_platform_ops.
If so, we can identify the offsets to endpoint registers using SoC
musb reg space and return the relevant 16-bit config value.

Jagan.