Avoid C2/C3 idle state

Avoid C2/C3 idle state

[Alex]

I run Ubuntu 8.04.3 LTS server 32bit in a small low power motherboard (VIA LN10000EG) equipped with VIA C7 1GHz CPU.
Kernel 2.6.24

I've some problems with serial device: sometimes I've overrun errors when receive data through /dev/ttyS0

After some days of hard work I've discovered that when CPU enter in C2 or C3 state and data come in from serial line, I've the overrun.

I've searched for a solution and I've found that kernel parameter "idle=halt" can avoid CPU enter C2/C3 state, but this parameter is now not supported anymore (it was in the past).
So I've completely disabled the ACPI support ("acpi=off" kernel parameter).
But I lack support for power button, poweroff command and so on.
I wish to re-enable ACPI.

The question is: Howto avoid CPU enter C2/C3 state and re-activate ACPI so CPU can go in C1 state ? C1 is better than nothing at all ....

I have no problem to modify kernel source and recompile it.

Thanks fo suggestions


      

Re: Avoid C2/C3 idle state

[Dominik Brodowski]

Hey,

you can set the maximum C-state to use using the
"processor.max_cstate=<value>" parameter. No need to recompile.

Best,
	Dominik

On Tue, Nov 24, 2009 at 09:50:01PM +0000, Alex wrote:
> I run Ubuntu 8.04.3 LTS server 32bit in a small low power motherboard (VIA LN10000EG) equipped with VIA C7 1GHz CPU.
> Kernel 2.6.24
> 
> I've some problems with serial device: sometimes I've overrun errors when receive data through /dev/ttyS0
> 
> After some days of hard work I've discovered that when CPU enter in C2 or C3 state and data come in from serial line, I've the overrun.
> 
> I've searched for a solution and I've found that kernel parameter "idle=halt" can avoid CPU enter C2/C3 state, but this parameter is now not supported anymore (it was in the past).
> So I've completely disabled the ACPI support ("acpi=off" kernel parameter).
> But I lack support for power button, poweroff command and so on.
> I wish to re-enable ACPI.
> 
> The question is: Howto avoid CPU enter C2/C3 state and re-activate ACPI so CPU can go in C1 state ? C1 is better than nothing at all ....
> 
> I have no problem to modify kernel source and recompile it.
> 
> Thanks fo suggestions
> 
> 
>       
> --
> To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Avoid C2/C3 idle state

[Alex]

--- Mar 24/11/09, Dominik Brodowski <linux@dominikbrodowski.net> ha scritto:

> Da: Dominik Brodowski <linux@dominikbrodowski.net>
> Oggetto: Re: Avoid C2/C3 idle state
> A: "Alex" <ambroxa@yahoo.it>
> Cc: linux-acpi@vger.kernel.org
> Data: Martedì 24 novembre 2009, 23:29
> Hey,
> 
> you can set the maximum C-state to use using the
> "processor.max_cstate=<value>" parameter. No need to
> recompile.

Thanks for your kind reply.

A stupid qustion:
<value> is a numeric value as 1 , 2 , 3  (i.e. processor.max_cstate=1 )
or is a string like C1 , C2 , C3 (i.e. processor.max_cstate=C1 ) ?





      
--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Avoid C2/C3 idle state

[Dominik Brodowski]

On Wed, Nov 25, 2009 at 06:53:14AM +0000, Alex wrote:
> --- Mar 24/11/09, Dominik Brodowski <linux@dominikbrodowski.net> ha scritto:
> 
> > Da: Dominik Brodowski <linux@dominikbrodowski.net>
> > Oggetto: Re: Avoid C2/C3 idle state
> > A: "Alex" <ambroxa@yahoo.it>
> > Cc: linux-acpi@vger.kernel.org
> > Data: Martedì 24 novembre 2009, 23:29
> > Hey,
> > 
> > you can set the maximum C-state to use using the
> > "processor.max_cstate=<value>" parameter. No need to
> > recompile.
> 
> Thanks for your kind reply.
> 
> A stupid qustion:
> <value> is a numeric value as 1 , 2 , 3  (i.e. processor.max_cstate=1 )
> or is a string like C1 , C2 , C3 (i.e. processor.max_cstate=C1 ) ?

It's the former. I'd suggest you try max_cstate=2 first, as that might save
power, reduce heat and still not interfere with the serial device.

Best,
	Dominik
--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Avoid C2/C3 idle state

[Alex]

> > > Hey,
> > > 
> > > you can set the maximum C-state to use using the
> > > "processor.max_cstate=<value>" parameter.
> No need to
> > > recompile.
> > 
> > Thanks for your kind reply.
> > 
> > A stupid qustion:
> > <value> is a numeric value as 1 , 2 , 3 
> (i.e. processor.max_cstate=1 )
> > or is a string like C1 , C2 , C3 (i.e.
> processor.max_cstate=C1 ) ?
> 
> It's the former. I'd suggest you try max_cstate=2 first, as
> that might save
> power, reduce heat and still not interfere with the serial
> device.


MMhh , there is a problem in my kernel 2.6.24 (Ubuntu 8.04.3 original)

[    0.000000] Kernel command line: root=UUID=27d1e179-62d3-4996-878e-e4a456bd45f4 ro quiet splash acpi=force processor.max_cstate=2
[    0.000000] Unknown boot option `processor.max_cstate=2': ignoring

This is a known problem related to Ubuntu 8.04, solved here:
http://www.inliniac.net/blog/2008/07/25/fixing-noise-on-ubuntu-hardy-804-aka-setting-max_cstate.html

Really my idea is that my serial problems are involved with C3 state: 800 microsec. latency and cache loss.
So I agree with your suggestion to use max_cstate=2.
But it doesn't work: serial overrun still there.
max_cstate=1 works fine : no overrun at all.
C2 state is buggy too.
I don't know if it's a hardware problem or acpi code problem, but with C1 it works fine and I'm happy :-)





      
--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html