2.6.0-test3 smp irq balance

2.6.0-test3 smp irq balance

[Jens Gecius]

Hi!

having switched to 2.6.0-t3 to test some of the new features, I
noticed more framedrops from bttv. sometimes the system feels a little
sluggish, although i expected the opposite.

trying to track down the "problem" I stumbled across this:

jens@maniac:/usr/src/linux$ cat /proc/interrupts
           CPU0       CPU1       
  0:   91133739         29    IO-APIC-edge  timer
  1:      43765          1    IO-APIC-edge  i8042
  2:          0          0          XT-PIC  cascade
  4:     774189          0    IO-APIC-edge  serial
  5:    9337660          1   IO-APIC-level  ohci1394, uhci-hcd, uhci-hcd
  8:          4          0    IO-APIC-edge  rtc
  9:   15340458          1   IO-APIC-level  Ensoniq AudioPCI, eth1
 10:    1233326          5   IO-APIC-level  ide3, eth0
 11:   21436308          1   IO-APIC-level  nvidia, bttv0
 14:     884901          0    IO-APIC-edge  ide0
 15:        126          2    IO-APIC-edge  ide1
NMI:          0          0 
LOC:   91137709   91138393 
ERR:          0
MIS:         12

irqs seem not to be distributed between cpus, having one to handle all
(even while building kernel on both cpus (according to gkrell), the
numbers for the second cpu don't change.

i could post any needed information (hopefully). The bootup process
did not show anything raising questions to me.

any hints?

-- 
Tschoe,                http://gecius.de/gpg-key.txt - Fingerprint:
 Jens                  1AAB 67A2 1068 77CA 6B0A  41A4 18D4 A89B 28D0 F097

Re: 2.6.0-test3 smp irq balance

[Arjan van de Ven]

[-- Attachment #1: Type: text/plain, Size: 340 bytes --]

On Wed, 2003-08-20 at 17:03, Jens Gecius wrote:

> 
> irqs seem not to be distributed between cpus, having one to handle all
> (even while building kernel on both cpus (according to gkrell), the
> numbers for the second cpu don't change.

just install and run the irqbalance daemon from:
http://people.redhat.com/arjanv/irqbalance

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: 2.6.0-test3 smp irq balance

[Jens Gecius]

Arjan van de Ven <arjanv@redhat.com> writes:

>> irqs seem not to be distributed between cpus, having one to handle all
>> (even while building kernel on both cpus (according to gkrell), the
>> numbers for the second cpu don't change.
>
> just install and run the irqbalance daemon from:
> http://people.redhat.com/arjanv/irqbalance

I did. Didn't change anything. The numbers are the same - the only
CPU2 irq increasing is LOC. Any other hints?

-- 
Tschoe,                http://gecius.de/gpg-key.txt - Fingerprint:
 Jens                  1AAB 67A2 1068 77CA 6B0A  41A4 18D4 A89B 28D0 F097

Re: 2.6.0-test3 smp irq balance

[Herbert Pötzl]

On Thu, Aug 21, 2003 at 06:21:39PM +0200, Jens Gecius wrote:
> Arjan van de Ven <arjanv@redhat.com> writes:
> 
> >> irqs seem not to be distributed between cpus, having one to handle all
> >> (even while building kernel on both cpus (according to gkrell), the
> >> numbers for the second cpu don't change.
> >
> > just install and run the irqbalance daemon from:
> > http://people.redhat.com/arjanv/irqbalance
> 
> I did. Didn't change anything. The numbers are the same - the only
> CPU2 irq increasing is LOC. Any other hints?

I remember some strange case (somewhere around 2.4.18,
yes I know you are running 2.6.0 whatever) where I needed
to set the smp_affinity manually after boot, otherwise
the second cpu would not receive any interrupts on some
dual athlon system ...

echo ff >/proc/irq/xx/smp_affinity 

did the trick, everything else was working as expected ...

just a weird idea, don't blame me if it actually works ;)

best,
Herbert

> -- 
> Tschoe,                http://gecius.de/gpg-key.txt - Fingerprint:
>  Jens                  1AAB 67A2 1068 77CA 6B0A  41A4 18D4 A89B 28D0 F097
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

RE: 2.6.0-test3 smp irq balance

[Pallipadi, Venkatesh]

This lkml thread may be of help.
http://www.ussg.iu.edu/hypermail/linux/kernel/0301.3/0197.html

-Venkatesh

> -----Original Message-----
> From: Jens Gecius [mailto:jens@gecius.de] 
> Sent: Thursday, August 21, 2003 9:22 AM
> To: linux-kernel@vger.kernel.org
> Subject: Re: 2.6.0-test3 smp irq balance
> 
> 
> Arjan van de Ven <arjanv@redhat.com> writes:
> 
> >> irqs seem not to be distributed between cpus, having one 
> to handle all
> >> (even while building kernel on both cpus (according to gkrell), the
> >> numbers for the second cpu don't change.
> >
> > just install and run the irqbalance daemon from:
> > http://people.redhat.com/arjanv/irqbalance
> 
> I did. Didn't change anything. The numbers are the same - the only
> CPU2 irq increasing is LOC. Any other hints?
> 
> -- 
> Tschoe,                http://gecius.de/gpg-key.txt - Fingerprint:
>  Jens                  1AAB 67A2 1068 77CA 6B0A  41A4 18D4 
> A89B 28D0 F097
> 
> -
> To unsubscribe from this list: send the line "unsubscribe 
> linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
>