2006-12-05 16:30:15

by raz

[permalink] [raw]
Subject: irq/0/smp_affinity =3 doesn't seem to work

hello.

I have a dual cpu AMD machine, I noticed that
only one timer0 is working in /proc/interrutps.
setting proc/irq/0/smp_affinity to 3 does make
any difference.
setting smp_affinity to 2 does move the inetrrupts.
the above applys with or withour irq_balancer .


thank you
--
Raz


2006-12-05 16:32:32

by Arjan van de Ven

[permalink] [raw]
Subject: Re: irq/0/smp_affinity =3 doesn't seem to work

On Tue, 2006-12-05 at 18:30 +0200, Raz Ben-Jehuda(caro) wrote:
> hello.
>
> I have a dual cpu AMD machine, I noticed that
> only one timer0 is working in /proc/interrutps.
> setting proc/irq/0/smp_affinity to 3 does make
> any difference.

if you set it to 3 then the chipset gets to decide where the irq goes.
Many decide to send it to the first cpu.

(not that it matters, the timer is so low frequency that it can go
anywhere without problems)

--
if you want to mail me at work (you don't), use arjan (at) linux.intel.com
Test the interaction between Linux and your BIOS via http://www.linuxfirmwarekit.org

2006-12-05 19:14:59

by raz

[permalink] [raw]
Subject: Re: irq/0/smp_affinity =3 doesn't seem to work

On 12/5/06, Arjan van de Ven <[email protected]> wrote:
> On Tue, 2006-12-05 at 18:30 +0200, Raz Ben-Jehuda(caro) wrote:
> > hello.
> >
> > I have a dual cpu AMD machine, I noticed that
> > only one timer0 is working in /proc/interrutps.
> > setting proc/irq/0/smp_affinity to 3 does make
> > any difference.
>
> if you set it to 3 then the chipset gets to decide where the irq goes.
> Many decide to send it to the first cpu.
>
> (not that it matters, the timer is so low frequency that it can go
> anywhere without problems)
>
> --
> if you want to mail me at work (you don't), use arjan (at) linux.intel.com
> Test the interaction between Linux and your BIOS via http://www.linuxfirmwarekit.org
>
>
thanks Arjan.

Yet something is not clear to me.
To the best of my knowledge , each cpu run its own schedule routine.
So, if only cpu0 gets timer0 interrupts, how does cpu1
gets to run the schedule function ?
what interrupts cpu1' current task ?

raz

2006-12-05 19:19:11

by Arjan van de Ven

[permalink] [raw]
Subject: Re: irq/0/smp_affinity =3 doesn't seem to work

On Tue, 2006-12-05 at 21:14 +0200, Raz Ben-Jehuda(caro) wrote:
> On 12/5/06, Arjan van de Ven <[email protected]> wrote:
> > On Tue, 2006-12-05 at 18:30 +0200, Raz Ben-Jehuda(caro) wrote:
> > > hello.
> > >
> > > I have a dual cpu AMD machine, I noticed that
> > > only one timer0 is working in /proc/interrutps.
> > > setting proc/irq/0/smp_affinity to 3 does make
> > > any difference.
> >
> > if you set it to 3 then the chipset gets to decide where the irq goes.
> > Many decide to send it to the first cpu.
> >
> > (not that it matters, the timer is so low frequency that it can go
> > anywhere without problems)
> >
> > --
> > if you want to mail me at work (you don't), use arjan (at) linux.intel.com
> > Test the interaction between Linux and your BIOS via http://www.linuxfirmwarekit.org
> >
> >
> thanks Arjan.
>
> Yet something is not clear to me.
> To the best of my knowledge , each cpu run its own schedule routine.
> So, if only cpu0 gets timer0 interrupts, how does cpu1
> gets to run the schedule function
> what interrupts cpu1' current task ?

the scheduler code uses IPI's for that...