> On 18 Oct 1999, Jes Sorensen wrote:
>
> > It is not quite clear to me if this is related, however I have an HP
> > NetServer box as well (not sure which model number) which also has two
> > NCR controllers on the motherboard. What I found is that all PCI
> > devices on the motherboard are detected twice and the SCSI driver for
> > the NCR then tries to probe each twice (in parallel) and boom. I
> > hacked the sym53c8xx.c driver to get around this quickly and get the
> > system booted.
>
> The devices that appeared to be detected twice seem to be the ones that
> reside on secondary PCI BUSes. At the time we have had to look into this
Peer PCI busses to be more accurate. The problem is that the PCI probe
code will probe the peer busses that are behind a 451NX chipset without
needing a bridge device to guide them there (which is correct) but in
order to make legacy PCI probe code from other OSes work on the system, HP
put some "fake" bridge devices on pci bus 0 that *also* point to the same
busses as the peer bus configuration registers on the 451NX chipset. Net
result, we probe the bus twice. There is a patch that fixes this in the
2.2.12 kernel we shipped, and I also gave that patch to Martin and it has
been forward ported to his 2.3 stuff. It handles more than just the 451NX
chipset, it also does the two RCC chipsets that have peer busses and it
handles the Compaq hot plug PCI chipset in use on most 8way x86 boxes
today.
Doug Ledford <dledford@redhat.com>
Opinions expressed are my own, but
they should be everybody's.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/