RE: NForce2 pseudoscience stability testing (2.6.0-test11)

From: b@xxxxxxxxxxxxx
Date: Thu Dec 04 2003 - 04:10:44 EST


Prakash Cheemplavam wrote
>>>Thanks everyone for your continued interest in this, I'll
>>>try and test the no-onboard-PATA + UP LAPIC and IOAPIC and
>>>add-in-card-PATA with no onboard PATA + + UP LAPIC and IOAPIC
>>>when I get a spare moment which is rare.
>
>I don't think that the AMD IDE is the problem. I have compiled
>it in, as
>well, but I am using the onboard SATA. Since this can be
>considered as >an pci-card (the chip is connected to the pci
> bridge) I think ölocking occurs on high traffic on PCI bus.
> Like now I get over 60mb/s with my >HD. Formerly I got only
> 25mb/s. before I could do some rounds of hdparm -t, before it
> locks. Now it locks immediately when doing hdparm -t
> when APIC is enabled.
>
>SO, I think it is not IDE specific. Does anybody have gigabit
> network card? Maybe that we should try to push something big
> through it (without reading from hd). If that leads to lock
> up we have a semi proof that it is due to high traffic on
> pci-bus.

I was thinking that myself (PCI activity triggering this).
The first time I hit this problem was with a card with two
ACENics/tigon2 (acenic.o) sniffing traffic at high rates
(100,000pps+) with most file i/o going over NFS on the
integrated 3com interface.

I ran for three days 200,000+pps sniffing with tethereal on
windows2000 on both acenics and never a lockup.

The AMD-Nvidia PATA does seem to be a very common in this
problem, but everyone at least has a CD-ROM and most
have a PATA hard disk so its going to be there every time
a problems crops up. I think we need to prove that an
solid add-in PCI PATA card that takes the CD and the
and the PATA disk and shut off the onboard ATA and torture
test again. I havent had time to yet.




-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/