On Sun, Nov 24, 2024 at 11:48:45PM +0100, Mark Kettenis wrote:
> > From: David Gwynne <david@gwynne.id.au>
> > Date: Thu, 21 Nov 2024 21:19:09 +1000
> > 
> > > On 19 Nov 2024, at 08:45, Paul Fertser <fercerpav@gmail.com> wrote:
> > > 
> > > On Mon, Nov 18, 2024 at 02:56:42AM +0300, Paul Fertser wrote:
> > >> However with changes to rkusbphy(4) added back it manages to attach
> > >> and continues to boot so they must be doing something important but
> > >> not quite right, full speed device attached to xHCI hubs
> > >> (TYPEC{0,1}_D* signals) works, for EHCI and OHCI (HOST{0,1}_D*)
> > >> there's zero reaction. This can be fixed later.
> > > 
> > > Later happened to be today, and with the attached diff I confirm all
> > > ports work in all modes both after booting from a modern U-Boot with
> > > "usb start" and after booting from the stock OpenBSD U-Boot which
> > > doesn't touch USB at all.
> > 
> > this looks right to me.
> 
> I dropped the comment about OpenBSD only supporting host mode.  That
> is true, but the sentence seemed garbled and we are (mostly) using the
> same values as the Linux driver I'm looking at.  Although there is a
> funny difference between the two PHYs for the RK3568 that I need to
> look into...

For the reference, the values I observed booting with U-boot that
doesn't touch USB matched exactly what the SoC datasheet specified as
reset values, and if "usb start" was performed the lowest two bits
were changed from 2 to 1. I didn't really dig the ULPI + controller
docs to fully understand what the datasheet was trying to tell me but
the testing results were pretty convincing in that 1 there disables
the USB2.0 functionality on the corresponding port while 2
enables. Comment I didn't really fully understood so just copied from
your earlier patch :)

Thank you for merging!