Re: gmac Ethernet

[Benjamin Herrenschmidt <benh@kernel.crashing.org>]

>   02:0f.0 Class ffff: Apple Computer Inc. UniNorth GMAC (rev ff)
>   00: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
>   10: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
>   20: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
>   30: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
>   
>   Anything that might work? Is it toast?
>
>All 1's means the PCI device isn't driving the data/addr
>lines for config space accesses.  It looks bad
>but I wonder how any correct PCI config information at
>all is being obtained.

The chip return those values when it's clock is shut down
(that is when it's power managed). The driver should bring
it back up when the interface is up. Do you get sensible
lspci values then ?

If that doesn't work, look into pmac_feature.c for the
core99_gmac_enable() function and hack it so that after
enabling the cell clock, it waits a few ns, then write
the bit again. I've spotted some code in darwin that
suggests some revs of the uninorth ASIC may "miss" the
write to this bit occasionally. 

Ben.


-- 
To UNSUBSCRIBE, email to debian-powerpc-request@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org