ACPI-related bug caused poor responsitivity (2.6.26.x line)

ACPI-related bug caused poor responsitivity (2.6.26.x line)

[Zbigniew Baniewski]

[-- Attachment #1: Type: text/plain, Size: 1120 bytes --]

Hallo,

There are so many issues reported during last 3 months, that I'm not sure,
whether (or not) someone wrote about the problem. But few words anyway:

I noticed, that using new kernels (tested 2.6.26 and 2.6.26.2), there are
typical for ACPI-related problems troubles with responsitivity. For example:
if I'm opening "bloated" web page using Opera, one can't move the mouse
cursor smoothly around the screen; there is a "jitter" during the movement,
and any output (like GKrellm's diagrams created in real time) for that 2-3
seconds is "frozen". As I've found googling around, these are rather typical
ACPI-related problems, and indeed: forcing "acpi=off" was the cure.

Additionally I noticed, that there are no such problems while using older
kernel 2.6.25.2 - so something bad happened during the period "between"
2.6.25.2 and 2.6.26.

Attached gzipped dmesg's messages; those described "aon" are with ACPI
switched on, and - respectively - "aoff" with ACPI off.

Tested under Debian Etch, Pentium III 700, mobo Gigabyte GA-6BX7 with newest
BIOS available.
-- 
				pozdrawiam / regards

						Zbigniew Baniewski

[-- Attachment #2: 2.6.26.aon.txt.gz --]
[-- Type: application/octet-stream, Size: 5675 bytes --]

[-- Attachment #3: 2.6.26.aoff.txt.gz --]
[-- Type: application/octet-stream, Size: 5274 bytes --]

[-- Attachment #4: 2.6.26.2.aon.txt.gz --]
[-- Type: application/octet-stream, Size: 5677 bytes --]

[-- Attachment #5: 2.6.25.2.aon.txt.gz --]
[-- Type: application/octet-stream, Size: 5751 bytes --]

Re: ACPI-related bug caused poor responsitivity (2.6.26.x line)

[Andi Kleen]

Zbigniew Baniewski <zb@ispid.com.pl> writes:


> There are so many issues reported during last 3 months, that I'm not sure,
> whether (or not) someone wrote about the problem. But few words anyway:
>
> I noticed, that using new kernels (tested 2.6.26 and 2.6.26.2), there are
> typical for ACPI-related problems troubles with responsitivity. For example:
> if I'm opening "bloated" web page using Opera, one can't move the mouse
> cursor smoothly around the screen; there is a "jitter" during the movement,
> and any output (like GKrellm's diagrams created in real time) for that 2-3
> seconds is "frozen". As I've found googling around, these are rather typical
> ACPI-related problems, and indeed: forcing "acpi=off" was the cure.

I assume you double checked that? As in you removed the option again
and the problem reappeared? And you add it again and it disappeared
again? If not do so please.

If it's checked please put a full report into http://bugzilla.kernel.org

Might be some kind of power saving issue. Does it happen when 
you boot with processor.max_cstate=1? (without acpi=off) Does it happen when you
disable CPUIDLE in the config?

Thanks

-Andi

Re: ACPI-related bug caused poor responsitivity (2.6.26.x line)

[Zhang Rui]

On Thu, 2008-08-14 at 23:13 +0200, Zbigniew Baniewski wrote:
> Hallo,
> 
> There are so many issues reported during last 3 months, that I'm not sure,
> whether (or not) someone wrote about the problem. But few words anyway:
> 
> I noticed, that using new kernels (tested 2.6.26 and 2.6.26.2), there are
> typical for ACPI-related problems troubles with responsitivity. For example:
> if I'm opening "bloated" web page using Opera, one can't move the mouse
> cursor smoothly around the screen; there is a "jitter" during the movement,
> and any output (like GKrellm's diagrams created in real time) for that 2-3
> seconds is "frozen". As I've found googling around, these are rather typical
> ACPI-related problems, and indeed: forcing "acpi=off" was the cure.
> 
> Additionally I noticed, that there are no such problems while using older
> kernel 2.6.25.2 - so something bad happened during the period "between"
> 2.6.25.2 and 2.6.26.
> 
> Attached gzipped dmesg's messages; those described "aon" are with ACPI
> switched on, and - respectively - "aoff" with ACPI off.
> 
> Tested under Debian Etch, Pentium III 700, mobo Gigabyte GA-6BX7 with newest
> BIOS available.

Please check if it's a duplicate of bug
http://bugzilla.kernel.org/show_bug.cgi?id=11141

thanks,
rui

Re: ACPI-related bug caused poor responsitivity (2.6.26.x line)

[Zbigniew Baniewski]

On Fri, Aug 15, 2008 at 04:20:11AM +0200, Andi Kleen wrote:

> I assume you double checked that? As in you removed the option again
> and the problem reappeared? And you add it again and it disappeared
> again? If not do so please.

Checked it several times.

> If it's checked please put a full report into http://bugzilla.kernel.org

OK, I'll do that.

> Might be some kind of power saving issue. Does it happen when 
> you boot with processor.max_cstate=1? (without acpi=off)

It hasn't been accepted as the boot parameter - but I've inserted it into
/etc/modprobe.d/processor and updated modules. And it made the situation
a bit better - I mean: no "jitter" during mouse movement anymore, but
still "momentary freeze" of GKrellm output (which isn't the case when
using 2.6.25.2 kernel).

> Does it happen when you disable CPUIDLE in the config?

You mean: CONFIG_CPU_IDLE=n ?
-- 
				pozdrawiam / regards

						Zbigniew Baniewski

Re: ACPI-related bug caused poor responsitivity (2.6.26.x line)

[Zbigniew Baniewski]

On Fri, Aug 15, 2008 at 10:35:18AM +0800, Zhang Rui wrote:

> Please check if it's a duplicate of bug
> http://bugzilla.kernel.org/show_bug.cgi?id=11141

Seems to be somewhat similar - but reading all that comments I've got
a feeling, there is a different one described. But I can't be 100% sure.
-- 
				pozdrawiam / regards

						Zbigniew Baniewski