Re: 2.6.0 kernel: Bind interrupt question.

Re: 2.6.0 kernel: Bind interrupt question.

[Dong V Nguyen]

Anton,
You're right. By defaut, the CONFIG_NR_CPUS is set to 32.
I need to reset that and rebuild the kernel to try the interrupt binding
again.
Thanks,



Anton Blanchard <anton@samba.org> on 11/10/2003 04:48:22 PM

To:    Dong V Nguyen/Austin/IBM@IBMUS
cc:    linux-kernel@vger.kernel.org
Subject:    Re: 2.6.0 kernel: Bind interrupt question.




Hi,

> Have you seen any problems with interrupt binding on 2.6.0-drv45003 ?
> I tried this command to bind interrupt, but it does not work:
> ============================
> cat  /proc/irq/165/smp_affinity
> ffffffff00000000
> echo 01 > /proc/irq/165/smp_affinity
> cat  /proc/irq/165/smp_affinity
> ffffffff00000000

This is probably a ppc64 specific issue, we can continue this on
linuxppc64-dev@lists.linuxppc.org

> There is nothing changed after binding.
> One thing I see is it shows 16 digits "ffffffff00000000" on 2.6.0 while
> only 8 digits in 2.4 .

Its part of the support for > 32way machines, but it looks like its
broken for some configurations (Im guessing you have CONFIG_NR_CPUS set
to 32).

 Anton

2.6.0 kernel: Bind interrupt question.

[Dong V Nguyen]

Have you seen any problems with interrupt binding on 2.6.0-drv45003 ?
I tried this command to bind interrupt, but it does not work:
============================
cat  /proc/irq/165/smp_affinity
ffffffff00000000
echo 01 > /proc/irq/165/smp_affinity
cat  /proc/irq/165/smp_affinity
ffffffff00000000
===========================
There is nothing changed after binding.
One thing I see is it shows 16 digits "ffffffff00000000" on 2.6.0 while
only 8 digits in 2.4 .
Do I need any special ways to bind interrupt ?


Thanks.,

Re: 2.6.0 kernel: Bind interrupt question.

[Anton Blanchard]

Hi,

> Have you seen any problems with interrupt binding on 2.6.0-drv45003 ?
> I tried this command to bind interrupt, but it does not work:
> ============================
> cat  /proc/irq/165/smp_affinity
> ffffffff00000000
> echo 01 > /proc/irq/165/smp_affinity
> cat  /proc/irq/165/smp_affinity
> ffffffff00000000

This is probably a ppc64 specific issue, we can continue this on
linuxppc64-dev@lists.linuxppc.org

> There is nothing changed after binding.
> One thing I see is it shows 16 digits "ffffffff00000000" on 2.6.0 while
> only 8 digits in 2.4 .

Its part of the support for > 32way machines, but it looks like its
broken for some configurations (Im guessing you have CONFIG_NR_CPUS set
to 32).

Anton

Re: 2.6.0 kernel: Bind interrupt question.

[Zwane Mwaikambo]

On Tue, 11 Nov 2003, Anton Blanchard wrote:

> > Have you seen any problems with interrupt binding on 2.6.0-drv45003 ?
> > I tried this command to bind interrupt, but it does not work:
> > ============================
> > cat  /proc/irq/165/smp_affinity
> > ffffffff00000000
> > echo 01 > /proc/irq/165/smp_affinity
> > cat  /proc/irq/165/smp_affinity
> > ffffffff00000000
> 
> This is probably a ppc64 specific issue, we can continue this on
> linuxppc64-dev@lists.linuxppc.org
> 
> > There is nothing changed after binding.
> > One thing I see is it shows 16 digits "ffffffff00000000" on 2.6.0 while
> > only 8 digits in 2.4 .
> 
> Its part of the support for > 32way machines, but it looks like its
> broken for some configurations (Im guessing you have CONFIG_NR_CPUS set
> to 32).

There was a fix which went in for something similar on i386 a while back.

Re: 2.6.0 kernel: Bind interrupt question.

[Zwane Mwaikambo]

On Tue, 11 Nov 2003, Zwane Mwaikambo wrote:

> > Its part of the support for > 32way machines, but it looks like its
> > broken for some configurations (Im guessing you have CONFIG_NR_CPUS set
> > to 32).
> 
> There was a fix which went in for something similar on i386 a while back.

And here is a link; PPC64 appears to have the older version

http://groups.google.com/groups?hl=en&lr=&ie=UTF-8&oe=UTF-8&threadm=uoUN.V9.15%40gated-at.bofh.it&rnum=1&prev=/groups%3Fq%3Dsmp_affinity%2Bi386%2BZwane%26hl%3Den%26lr%3D%26ie%3DUTF-8%26oe%3DUTF-8%26selm%3DuoUN.V9.15%2540gated-at.bofh.it%26rnum%3D1