Re: HT bug in some Intel CPUs ?

[Greg Oster <oster%cs.usask.ca@localhost>]

Greg Oster writes:
> Manuel Bouyer writes:
> > On Thu, Aug 06, 2009 at 08:55:39AM -0600, Greg Oster wrote:
> > > Manuel Bouyer writes:
> > > > Hi,
> [snip]
> > > 
> > > They don't show up as hyperthreaded in 3.0, but do in 5.0.1.
> > 
> > Did you try a SMP 3.0 kernel ? Maube a ACPI issue ?
> 
> Yes.. the box has been running GENERIC.MP for ages without issue.
> 
> > > > [...]
> > > > Did someone else notice something similar, or have informations about
> > > > such bug ?
> > > 
> > > I'd like to think that whatever issue you're seeing is the same one I 
> > > have... Would disabling hyperthreading help at all to provide a 
> > > datapoint?
> > 
> > That would, of course. Now there may be other issues :)
> 
> I'll see if I can schedule a quick downtime and turn off 
> hyperthreading... But yes, it's the "may be other issues" I'm mostly 
> concerned about :-}

So the answer to this is "no, turning off hyperthreading didn't help".

> > > (After an uptime of 330+ days, having 3 hangs in a week 
> > > isn't giving me warm, fuzzy feelings :-/ The machine in question 
> > > is an IBM x336 that has been rock-stable under 3.0)
> > 
> > I also have a similar box: same motherboard, with two
> > cpu0: Intel (686-class), 3000.23 MHz, id 0xf43
> > 3Ghz Xeon that is also due for upgrade. I'll see if it can run
> > SMP without hyperthreading. I'll probably test this this in a few
> > days.
> 
> Cool... I'm going to try turning off hyperthreading in a few hours... 
> of course, it's taken anywhere from 2 days to 2 minutes to see the 
> hang show up :-/

After turning off hyperthreading, the box was up for ~28 hours before 
falling into tstile land.  Unfortunately, this is a production box, 
and getting full tracebacks on the 40+ processes stuck in tstile just 
wasn't practical when users were beginning to phone me the instant the
box hung... :( :(

Later...

Greg Oster