Re: Clockport devices configuration layer [was: Re: USB stack vs bus_dma_tag_t]

[Radek Kujawa <radoslaw.kujawa%c0ff33.net@localhost>]

Wiadomość napisana w dniu 2010-10-18, o godz. 23:04, przez Frank Wille:

> Radek Kujawa wrote:
>
> > > > clockport0 at mainbus addr 0xd80000 # A1200 mainboard clockport
> > > > clockport* at zbus? # various Zorro clockports
> > > > tduhc0 at clockport? # E3B SUBWAY USB Controller
> > >
> > > [...]
> > > But what is a Zorro clockport? Does it really exist? Do you have
> > > examples?
> >
> > It is simply a clockport located on a Zorro card. Best known Zorro
> > cards with clockports include Individual Computers Buddha and X-Surf.
>
> Ok, but then the clockport should attach to that specific Zorro card  
> and not
> to the zbus directly. For example:
>
> ne* at zbus0
> clockport* at ne?        # X-surf clockport

I understand what you mean but... ne is a cross-platform driver for a  
NE2000 chip and it does know absolutely nothing about clockports. Why  
should it know? On a hardware level, clockport on these cards is  
attached to Zorro bus directly ;). Connecting any device to such  
clockport is like connecting it to super simplified non- 
autoconfiguring zbus. Clockport does share the PCB with ne or wdc, but  
these functions are totally separate. We could even support clockports  
even on Zorro devices where main functionality is unsupported (think  
VarIO, ZIV busboard etc.).

Also, doing it like you suggest means I have to include clockport  
support code in every driver/MD attachment code (right?). I'm not sure  
that is justified, considering simple nature of clockport. I think  
that all trivial clockports could be handled by just one driver, as  
the only difference is base address.

> > In some cases base address of such clockport is static, always the
> > same, and in some cases it is configured dynamically. I think most
> > clockports on Zorro cards are static, but I know at least one that is
> > configurable - clockport on a Deneb USB controller. It is probably
> > possible to change its base address via software.
>
> But the Deneb driver (when it ever exists) would probably know a way  
> to
> determine this address, so there is no problem.

Yes, but Deneb is a special case. Here it would certainly make sense  
to configure clockport at deneb, as in:

deneb* at zbus0         # Deneb card
clockport* at deneb?    # Deneb's configurable clockport
dehci at deneb?         # Deneb EHCI USB 2.0 Controller
usb* at dehci?          # USB bus

But this case is totally different than Buddha, X-Surf etc. Here we  
need special code to prepare clockport and use it (and it is handled  
through card's FPGA).

Neverthless, I'll start with implementing clockport at mainbus.

Best regards,
Radoslaw Kujawa