Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email hr@neohapsis.com
Re: Backout mclgeti for vr(4).

From: Stuart Henderson (stuspacehopper.org)
Date: Tue Aug 31 2010 - 16:02:32 CDT

On 2010/08/30 23:46, Thordur I Bjornsson wrote:
> On Mon, Aug 30, 2010 at 06:46:53PM -0400, Brynet wrote:
> > Evening,
> >
> > I have two machines with vr(4) interfaces running 4.7, and I can't seem
> > to find any problem running ping -f against them.
> >
> > vr0 at pci0 dev 12 function 0 "VIA VT6105 RhineIII" rev 0x86: apic 2 int
> > 19 (irq 10), address 00:19:5b:82:a1:e0
> > vr0 at pci0 dev 16 function 0 "VIA Rhine/RhineII" rev 0x06: irq 9,
> > address 00:50:ba:bd:89:4d
> >
> > Is it possible that this bug only effects a few models?
> Possible. I can't remember what model I had (as I no longer have access
> to the machines) but it was a soekris.

tried some more... 3x ping -f from two fast sources causes my
system with VT6105M to crash within seconds. 3x ping -f from a single
source causes it to crash in ~30 seconds.

on a hunch I tried switching over to copying buffers (i.e. adding
the VR_Q_NEEDALIGN quirk that's needed for older ICs) but no
difference there.

on another hunch, I tried a snapshot kernel from June 30th
(pre gcc 4). I haven't been able to crash it yet...