Re: [Qemu-devel] [PATCH v6 7/7] hw/pci-bridge: format SeaBIOS-compliant OFW device node for PXB

[Kevin O'Connor]

On Wed, Jun 17, 2015 at 08:16:30PM +0200, Laszlo Ersek wrote:
> As discussed earlier, there are two questions to consider about the OFW
> devpath pattern
> 
>   /address@hidden/address@hidden/...
> 
> that SeaBIOS currently recognizes for devices that reside behind extra
> PCI root buses.
> 
> Q1: everything in that pattern that is not "N"
> Q2: what goes into N
> 
> These are independent questions.
> 
> Marcel's patch for SeaBIOS intended to change SeaBIOS's behavior for
> both Q1 and Q2:
> 
> - For Q1, the proposed OFW devpath fragment was
> 
>   /address@hidden/...
> 
>   ie. dropping the second node ("address@hidden").
> 
> - For Q2, the proposed change was: instead of making N a *serial number*
> (where N stood for the N'th extra root bus discovered by SeaBIOS), make
> N an actual bus number.
> 
> However, applying these changes unconditionally would have broken the
> interface between Coreboot and SeaBIOS, in physical hardware
> environments (because Coreboot agrees with SeaBIOS on the current syntax
> (Q1) and interpretation (Q2) of the devpath pattern). Therefore there
> was an idea to make both SeaBIOS changes conditional on runningOnQemu().

Just to be clear, there is no syntax requirement between coreboot and
SeaBIOS with respect to Q1.  The only requirement is with respect to
Q2.  Coreboot code doesn't read or write the bootorder file - it's
treated as a static user provided config file.

> As far as I remember, Kevin was more or less okay with that, but (again,
> as far as I remember) he did express a mild dislike for such tweaks.

I agree with previous comments about this email discussion going on
too long.  I apologize if my earlier comments drove excess
development.  My only requirement is that we not break a
seabios/coreboot feature (ie, not change Q2 if not on qemu).

The rest of your email seems like an accurate synopsis to me.

-Kevin