Re: 2.6.9-rc2 and Hyperthreading. (SMT)

[Troy McFerrron]

On Wed, 15 Sep 2004 11:11:14 +1000, Darren Williams
<dsw@xxxxxxxxxxxxxxxxxx> wrote:
> On Tue, 14 Sep 2004, hotdog day wrote:
> 
> > Does anyone have any other suggestions on this issue? I know others
> > who are experincing the same thing.
> >
> >
> > On Tue, 14 Sep 2004 00:40:57 -0500, hotdog day <hotdogday@xxxxxxxxx> wrote:
> > > Actually, it just hardlocked again. Is there anything else that could
> > > be done, or am I stuck without SMP?
> > >
> > >
> > >
> > >
> > > On Mon, 13 Sep 2004 23:21:05 -0500, hotdog day <hotdogday@xxxxxxxxx> wrote:
> > > > Turning off CONFIG_SCHED_SMT has apparently fixed the issue.
> > > >
> > > > Three Q's:
> > > >
> > > > 1) Am I taking some kind of performance hit by doing this?
> > > >
> > > > 2) Is this something we can look forward to seeing fixed?
> > > >
> > > > 3) Do you need any info from me to help you?
> > > >
> > > > Thanks,
> > > >
> > > > Troy McFerron
> > > >
> > > >
> > > >
> > > >
> > > > On Tue, 14 Sep 2004 12:07:00 +1000, Nick Piggin <nickpiggin@xxxxxxxxxxxx> wrote:
> > > > >
> > > > >
> > > > > hotdog day wrote:
> > > > > > I have been testing the 2.6.9-rc1, and 2.6.9-rc2 kernel patches over
> > > > > > the past couple days and have been having some issues with
> > > > > > hyperthreading (SMT) turned on.
> I have tested 2.6.9-rc2 on a 3.0 Ghz HT and all seams OK, results of LTP
> message, and meminfo at:
> http://quasar.cse.unsw.edu.au/~dsw/public-files/x86
> 
> Not extensive though no lockups and left running overnight.
> 
> > > > > >
> > > > > > This problem first exhibited itself when I was testing
> > > > > > 2.6.9-rc2-mm2-love2. I noticed the following quirks that ONLY show
> > > > > > themselves with hyperthreading enabled on my 3.0C Pentium 4.
> > > > > >
> > > > > > Random HARD LOCKS. No messages from the kernel. Just a good swift hard lock.
> > > > > >
> > > > > > Hard locks when mounting two cdrom drives in quick succession.
> I'll try this tonight.
> 
> > > > > >
> > > > > > Turning off hyperthreading solves these issues.  Going back to 2.6.8.1
> > > > > > solves these issues.
> > > > > >
> > > > > > I then tried 2.6.9-rc1 with no mm or love patches. I had the exact same issues.
> > > > > >
> > > > > > Today I downloaded the prepatch to 2.6.9-rc2 and applied it to clean
> > > > > > 2.6.8 source. The issues are still there.
> > > > > >
> > > > > > I hope someone is paying attention to the way scheduler tweaks and
> > > > > > changes are affecting SMT enabled kernels. I don't think anyone wants
> > > > > > to disable features of their hardware in order to run an optimized
> > > > > > scheduler.
> > > > >
> > > > > Try turning off CONFIG_SCHED_SMT and see how you go. Thanks.
> > > > >
> > > >
> > >
> > -
> > 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/
> --------------------------------------------------
> Darren Williams <dsw AT gelato.unsw.edu.au>
> Gelato@UNSW <www.gelato.unsw.edu.au>
> --------------------------------------------------
> 

Well, I took your suggestion Nick. I cnanot recreate the lockup when I
am not running X. It happens with both the fglrx module AND the radeon
module.

Darren, could you send me your kernel .config so I can do a diff and
see what you might be doing differently?

-- 
Troy McFerron
Kernel Ricer and Linux Hobbyist Extrodinaire.
-
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/