Re: Patch needed for ASUS EEEPC 701 2.6.27 (perhaps a regession)

Re: Patch needed for ASUS EEEPC 701 2.6.27 (perhaps a regession)

[Alan Jenkins]

Robert Goldner wrote:

> Hi,
>
> I am a happy user of an asus eeepc 701 with an self-made kernel 2.6.27.
> Its working quite good, but there is one patch needed to avoid problems
> related to ACPI and Hotkeys.
> After using a few hotkeys there comes a message in syslog like:
> ACPI: EC: GPE storm detected, disabling EC GPE
> and no hotkey event is recognized any more. This is quite bad on a nettop.

Hi, I reported the original bug.  It is a regression, though the issue
has been present for a few kernel versions now.

I recommend that you avoid reproducing the problem.  I now have to add
"noapic" to the boot options on my EeePC, otherwise it randomly hangs /
reboots when I press hotkeys - I think it  damaged the hardware.

BTW you can avoid reproducing the issue on buggy kernels by not pressing
hotkeys quickly - and *never* holding them down; autorepeat is the main
problem.  Wait for the effect of the previous hotkey press to take
effect (e.g. decrementing the screen brightness) before you press a
hotkey again.  But it is safest to just not use them.
 
> The needed patch and the bug description is available at:
> http://bugzilla.kernel.org/show_bug.cgi?id=10919
> The patch is called 'patch vs 2.6.27-rc7'
>
> The patch fits also to 2.6.27 without any problems and works very fine.
> Thanks to everyone who was involved.
>
> I would suggest to include this patch in the mainline kernel very soon
> (e.g. 2.6.28 or if possible in 2.6.27.1).

The patch is part of Len's ACPI patch queue for the current merge
window, so it should appear in 2.6.28.  However it's not marked "CC:
linux-stable@vger.kernel.org", so it will not be automatically included
in 2.6.27.1.

I think it's up to the ACPI maintainers, if they think there's a risk it
might introduce other regressions.  As you say, it's been tested on a
range of problematic laptops, but OTOH this area has been a minefield. 
Maybe it needs to wait, at least until it's been in an -rc release or two.

Alan

Re: Patch needed for ASUS EEEPC 701 2.6.27 (perhaps a regession)

[Rafael J. Wysocki]

On Saturday, 11 of October 2008, Robert Goldner wrote:
> Hi,
> 
> I am a happy user of an asus eeepc 701 with an self-made kernel 2.6.27. 
> Its working quite good, but there is one patch needed to avoid problems 
> related to ACPI and Hotkeys.
> After using a few hotkeys there comes a message in syslog like:
> ACPI: EC: GPE storm detected, disabling EC GPE
> and no hotkey event is recognized any more. This is quite bad on a nettop.
> 
> The needed patch and the bug description is available at:
> http://bugzilla.kernel.org/show_bug.cgi?id=10919
> The patch is called 'patch vs 2.6.27-rc7'
> 
> The patch fits also to 2.6.27 without any problems and works very fine. 
> Thanks to everyone who was involved.
> 
> I would suggest to include this patch in the mainline kernel very soon 
> (e.g. 2.6.28 or if possible in 2.6.27.1).

This patch is scheduled for 2.6.28 AFAICS and will be included into -stable if
it doesn't break anything in the mainline.

Thanks,
Rafael

Re: Patch needed for ASUS EEEPC 701 2.6.27 (perhaps a regession)

[Pavel Machek]

On Sat 2008-10-11 15:50:21, Alan Jenkins wrote:
> Robert Goldner wrote:
> 
> > Hi,
> >
> > I am a happy user of an asus eeepc 701 with an self-made kernel 2.6.27.
> > Its working quite good, but there is one patch needed to avoid problems
> > related to ACPI and Hotkeys.
> > After using a few hotkeys there comes a message in syslog like:
> > ACPI: EC: GPE storm detected, disabling EC GPE
> > and no hotkey event is recognized any more. This is quite bad on a nettop.
> 
> Hi, I reported the original bug.  It is a regression, though the issue
> has been present for a few kernel versions now.
> 
> I recommend that you avoid reproducing the problem.  I now have to add
> "noapic" to the boot options on my EeePC, otherwise it randomly hangs /
> reboots when I press hotkeys - I think it  damaged the hardware.

I guess your hw went bad for some other reason. Restore original
system and if it is still broken, RMA?

-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: Patch needed for ASUS EEEPC 701 2.6.27 (perhaps a regession)

[Alan Jenkins]

Pavel Machek wrote:
> On Sat 2008-10-11 15:50:21, Alan Jenkins wrote:
>> Robert Goldner wrote:
>>
>>> Hi,
>>>
>>> I am a happy user of an asus eeepc 701 with an self-made kernel 2.6.27.
>>> Its working quite good, but there is one patch needed to avoid problems
>>> related to ACPI and Hotkeys.
>>> After using a few hotkeys there comes a message in syslog like:
>>> ACPI: EC: GPE storm detected, disabling EC GPE
>>> and no hotkey event is recognized any more. This is quite bad on a nettop.
>> Hi, I reported the original bug.  It is a regression, though the issue
>> has been present for a few kernel versions now.
>>
>> I recommend that you avoid reproducing the problem.  I now have to add
>> "noapic" to the boot options on my EeePC, otherwise it randomly hangs /
>> reboots when I press hotkeys - I think it  damaged the hardware.
> 
> I guess your hw went bad for some other reason. Restore original
> system and if it is still broken, RMA?
> 

It could be a coincidence.  But it's suspicious enough to advise caution.  We know what the bug is, and we have a very nice workaround queued up now.  There's no reason to test any more of these specific systems to destruction :).

I did try re-installing off the vendor DVD and it was still broken.  I fear I bought it from the wrong place to get sympathetic _frontline_ support / warrantee.  I'm not interested in RMA back to Asus - too much work and downtime for a cheap system, when I have an easy workaround.  I haven't noticed any problems with "noapic".


The symptoms strongly suggest overflow in an event buffer or counter maintained by the Embedded Controller.  So the EC firmware may have a bug, of the sort that results in "unspecified behaviour".  A bug in a special purpose (read: not subject to wide testing) subsystem which has direct connections to things like frequency and voltage control.

Hopefully I'm wrong, and I don't really know what I'm talking about here.

Regards
Alan

Re: Patch needed for ASUS EEEPC 701 2.6.27 (perhaps a regession)

[Alexey Starikovskiy]

Alan Jenkins wrote:
> It could be a coincidence. But it's suspicious enough to advise 
> caution. We know what the bug is, and we have a very nice workaround 
> queued up now. There's no reason to test any more of these specific 
> systems to destruction :).
> I did try re-installing off the vendor DVD and it was still broken.  I fear I bought it from the wrong place to get sympathetic _frontline_ support / warrantee.  I'm not interested in RMA back to Asus - too much work and downtime for a cheap system, when I have an easy workaround.  I haven't noticed any problems with "noapic".
>
>
> The symptoms strongly suggest overflow in an event buffer or counter maintained by the Embedded Controller.  So the EC firmware may have a bug, of the sort that results in "unspecified behaviour".  A bug in a special purpose (read: not subject to wide testing) subsystem which has direct connections to things like frequency and voltage control.
>
> Hopefully I'm wrong, and I don't really know what I'm talking about here.
>
> Regards
> Alan
>   
Did you try to disconnect all power resources (e.g. AC adapter and all 
batteries and wait for some time)?
May be there are some default settings in the BIOS?
This looks small enough to be a HW failure...

Regards,
Alex.

Re: Patch needed for ASUS EEEPC 701 2.6.27 (perhaps a regession)

[Alan Jenkins]

Alexey Starikovskiy wrote:
> Alan Jenkins wrote:
>> It could be a coincidence. But it's suspicious enough to advise
>> caution. We know what the bug is, and we have a very nice workaround
>> queued up now. There's no reason to test any more of these specific
>> systems to destruction :).
>> I did try re-installing off the vendor DVD and it was still broken. 
>> I fear I bought it from the wrong place to get sympathetic
>> _frontline_ support / warrantee.  I'm not interested in RMA back to
>> Asus - too much work and downtime for a cheap system, when I have an
>> easy workaround.  I haven't noticed any problems with "noapic".
>>
>>
>> The symptoms strongly suggest overflow in an event buffer or counter
>> maintained by the Embedded Controller.  So the EC firmware may have a
>> bug, of the sort that results in "unspecified behaviour".  A bug in a
>> special purpose (read: not subject to wide testing) subsystem which
>> has direct connections to things like frequency and voltage control.
>>
>> Hopefully I'm wrong, and I don't really know what I'm talking about
>> here.
>>
>> Regards
>> Alan
>>   
> Did you try to disconnect all power resources (e.g. AC adapter and all
> batteries and wait for some time)?

Yes, I tried leaving it with AC + battery unplugged overnight.

> May be there are some default settings in the BIOS?
> This looks small enough to be a HW failure...

I also upgraded the BIOS to the latest version.  And tried the BIOS
setup option to restore defaults.  Maybe there is some more state, but
it's not like a desktop where I can easily remove the CMOS battery.

Thanks
Alan