HP/Compaq NC6120 ACPI problems

HP/Compaq NC6120 ACPI problems

[Andrej Prsa]

Hi all,

I recently bought a new HP/Compaq NC6120 laptop which doesn't play well
with Linux. It took me a while to figure out that the problems are
probably ACPI-related.

The symptoms are as follows: computer starts fine, but within 20-30
seconds it gradually slows down to the point where it can no longer be
used - it takes 5 minutes to start the terminal. The fan doesn't work,
so the machine overheats. Using 'top' shows that 55% of CPU power is
consumed by kacpid.

If on the other hand I pass 'acpi=off' option to GRUB command line, the
computer hangs. In a rare event that it doesn't hang (I still can't
figure out what causes this behavior) it works in full speed without any
problems! A totally unreliable and probably a misleading thought is that
PCI interrupts are messed up without ACPI; this assumption is based
on the fact that the system hangs prior to the statement:

ACPI: PCI interrupt 0000:02:06.2[C] -> GSI 22 (level, low) -> IRQ 217

in dmesg (which occurs if ACPI is left on). When ACPI is turned off,
the fan works normally and there is no overheating whatsoever.

I hope I'm not too cryptic and that all of this makes at least some
sense. I also tried recompiling the 2.6.11.4 kernel and fiddling with
ACPI settings, but the result is kernel panic with VFS related problem
(although I created ramdisk by hand and I'm sure everything else is ok).

Used configuration: Debian testing (Sarge), kernel 2.6.8

I'm posting the original DSDT table:

	http://www.fiz.uni-lj.si/~prsa/acpi/dsdt.aml

The disassembled DSDT table using iasl -d:

	http://www.fiz.uni-lj.si/~prsa/acpi/dsdt.dsl

The precompiled DSDT table using iasl -tc:

	http://www.fiz.uni-lj.si/~prsa/acpi/dsdt.hex

Although there are 0 errors and 2 warnings, I still tend to blame ACPI
for my problems.

Does anyone have any suggestions on how to proceed and what to do next?

Thanks,

Andrej


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click

Re: HP/Compaq NC6120 ACPI problems

[Bjorn Helgaas]

On Thu, 2005-03-17 at 20:00 +0100, Andrej Prsa wrote:
> The symptoms are as follows: computer starts fine, but within 20-30
> seconds it gradually slows down to the point where it can no longer be
> used - it takes 5 minutes to start the terminal. The fan doesn't work,
> so the machine overheats. Using 'top' shows that 55% of CPU power is
> consumed by kacpid.

Take a look at /proc/interrupts a couple times, and see if there's
a device that's spewing interrupts.

> A totally unreliable and probably a misleading thought is that
> PCI interrupts are messed up without ACPI; this assumption is based
> on the fact that the system hangs prior to the statement:
> 
> ACPI: PCI interrupt 0000:02:06.2[C] -> GSI 22 (level, low) -> IRQ 217
> 
> in dmesg (which occurs if ACPI is left on). When ACPI is turned off,
> the fan works normally and there is no overheating whatsoever.

Try booting with "pci=routeirq".  If that makes a difference,
please post the dmesg log and output of "lspci".  If you can
also capture the dmesg log without "pci=routeirq", that would
also be useful.

> I also tried recompiling the 2.6.11.4 kernel and fiddling with
> ACPI settings, but the result is kernel panic with VFS related problem
> (although I created ramdisk by hand and I'm sure everything else is ok).

You can avoid the hassle of a ramdisk by building the drivers
and filesystems you need static (instead of as modules).

Bjorn




-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click

Re: HP/Compaq NC6120 ACPI problems

[Andrej Prsa]

Hi,

> > The symptoms are as follows: computer starts fine, but within 20-30
> > seconds it gradually slows down to the point where it can no longer
> > be used - it takes 5 minutes to start the terminal. The fan doesn't
> > work, so the machine overheats. Using 'top' shows that 55% of CPU
> > power is consumed by kacpid.
> 
> Take a look at /proc/interrupts a couple times, and see if there's
> a device that's spewing interrupts.

Observational update: it does *NOT* gradually slow down, it slows down
*instantly* after 2-3 minutes of running; the time of kacpid blowup is
somewhat arbitrary, it sometimes happens even before the boot-up is
complete. If I was to make another wild guess, it might be connected
with the fan - when the computer is cool, the blowup occurs later, if it
is warm, it occurs sooner. Is it possible that the call to turn the fan
on is triggering this kacpid blowup? How does one go about to test this?

For the first time I was now running 'top' from the moment a computer
came up to the moment kacpid boosted CPU consumption. Another erratum:
it is not 55% of CPU power for kacpid, it is 98%-99% of CPU power
reported in 'top'.

I did 'cat /proc/interrupts' before and after this discrete change in
behavior and I didn't notice anything out of the ordinary. For the sake
of completion I'm posting the contents 1-2 minutes after the kacpid
blowup here:

	http://www.fiz.uni-lj.si/~prsa/acpi/procint.txt

Also, I had 'tail -f /var/log/messages' and 'tail -f /var/log/syslog'
when the blowup occured - there were no outputs to those two files at
the time of blowup.

> > A totally unreliable and probably a misleading thought is that
> > PCI interrupts are messed up without ACPI; this assumption is based
> > on the fact that the system hangs prior to the statement:
> > 
> > ACPI: PCI interrupt 0000:02:06.2[C] -> GSI 22 (level, low) -> IRQ
> > 217
> > 
> > in dmesg (which occurs if ACPI is left on). When ACPI is turned off,
> > the fan works normally and there is no overheating whatsoever.
> 
> Try booting with "pci=routeirq".  If that makes a difference,
> please post the dmesg log and output of "lspci".  If you can
> also capture the dmesg log without "pci=routeirq", that would
> also be useful.

The option pci=routeirq doesn't help, hence I'm omitting the dmesg
output. If you find it useful, I could still put it to www.

This is the dmesg output of a standard boot:

	http://www.fiz.uni-lj.si/~prsa/acpi/dmesg.acpi_on.out

And here you may find the lspci outputs:

	http://www.fiz.uni-lj.si/~prsa/acpi/lspci.out
	http://www.fiz.uni-lj.si/~prsa/acpi/lspci-v.out
	http://www.fiz.uni-lj.si/~prsa/acpi/lspci-vv.out

> > I also tried recompiling the 2.6.11.4 kernel and fiddling with
> > ACPI settings, but the result is kernel panic with VFS related
> > problem (although I created ramdisk by hand and I'm sure everything
> > else is ok).
> 
> You can avoid the hassle of a ramdisk by building the drivers
> and filesystems you need static (instead of as modules).

I thought I needed initrd functionality to use the corrected DSDT tables
- that's what I did on my old laptop, because I recompile my kernel
often and I don't feel like constantly patching the kernel for
custom-made DSDT. But if you can suggest any better approach, I'd be
happy to hear about it! :)

Thanks,

Andrej


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click

Re: HP/Compaq NC6120 ACPI problems

[Kristoffer Sjoberg]

[-- Attachment #1: Type: text/plain, Size: 1660 bytes --]


On tor, 2005-03-17 at 21:56 +0100, Andrej Prsa wrote:
> Hi,
> 
> > > The symptoms are as follows: computer starts fine, but within 20-30
> > > seconds it gradually slows down to the point where it can no longer
> > > be used - it takes 5 minutes to start the terminal. The fan doesn't
> > > work, so the machine overheats. Using 'top' shows that 55% of CPU
> > > power is consumed by kacpid.
> > 
> > Take a look at /proc/interrupts a couple times, and see if there's
> > a device that's spewing interrupts.
> 
> Observational update: it does *NOT* gradually slow down, it slows down
> *instantly* after 2-3 minutes of running; the time of kacpid blowup is
> somewhat arbitrary, it sometimes happens even before the boot-up is
> complete. If I was to make another wild guess, it might be connected
> with the fan - when the computer is cool, the blowup occurs later, if it
> is warm, it occurs sooner. Is it possible that the call to turn the fan
> on is triggering this kacpid blowup? How does one go about to test this?
> 
> For the first time I was now running 'top' from the moment a computer
> came up to the moment kacpid boosted CPU consumption. Another erratum:
> it is not 55% of CPU power for kacpid, it is 98%-99% of CPU power
> reported in 'top'.

Hi,

I have a very similar problem on my NC6000, but only after resume from
S4 (swsusp2 2.1.8.2). It simply won't turn on the fan, and after a while
running 'yes' in a terminal, everything comes to a halt until I abort
the scrolling. Running 2.6.11.2 - I think that 2.6.11-rc1 worked fine
though (although it had a few other problems).


BR

Kristoffer Sjöberg

[-- Attachment #2: smime.p7s --]
[-- Type: application/x-pkcs7-signature, Size: 3134 bytes --]

Re: HP/Compaq NC6120 ACPI problems

[Matthew Garrett]

On Thu, 2005-03-17 at 21:56 +0100, Andrej Prsa wrote:

> Observational update: it does *NOT* gradually slow down, it slows down
> *instantly* after 2-3 minutes of running; the time of kacpid blowup is
> somewhat arbitrary, it sometimes happens even before the boot-up is
> complete. If I was to make another wild guess, it might be connected
> with the fan - when the computer is cool, the blowup occurs later, if it
> is warm, it occurs sooner. Is it possible that the call to turn the fan
> on is triggering this kacpid blowup? How does one go about to test this?

Ah, right - I've seen similar behaviour on a range of HPs (inluding the
6120). Try using 
http://bugzilla.kernel.org/attachment.cgi?id=4516&action=view and see if
it makes a difference.
-- 
Matthew Garrett | mjg59-1xO5oi07KQx4cg9Nei1l7Q@public.gmane.org



-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click

Re: HP/Compaq NC6120 ACPI problems

[Andrej Prsa]

Hi,

> > Observational update: it does *NOT* gradually slow down, it slows
> > down *instantly* after 2-3 minutes of running; the time of kacpid
> > blowup is somewhat arbitrary, it sometimes happens even before the
> > boot-up is complete. If I was to make another wild guess, it might
> > be connected with the fan - when the computer is cool, the blowup
> > occurs later, if it is warm, it occurs sooner. Is it possible that
> > the call to turn the fan on is triggering this kacpid blowup? How
> > does one go about to test this?
> 
> Ah, right - I've seen similar behaviour on a range of HPs (inluding
> the 6120). Try using 
> http://bugzilla.kernel.org/attachment.cgi?id=4516&action=view and see
> if it makes a difference.

Tried it, unfortunately the outcome is still the same; I used the patch
against 2.6.11.4 kernel *without* any other ACPI patch. Should I use the
patch you provided over the ACPI patch or was it supposed to fix the fan
issue by itself? Seemingly the problem is the fan; is there a way to
test this to be sure?

Is there a chance that another kernel (i.e. 2.4) would do a better job?
It's kinda frustrating to have a brand new laptop that sits turned off
on my desk... :(

Also, would you be so kind to provide the bug ID number so I can take a
look at what's causing this behavior? I can't figure it out from only a
link to the patch.

Any further ideas are very welcome! :)

Thanks,

Andrej


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click

Re: HP/Compaq NC6120 ACPI problems

[Matthew Garrett]

On Fri, 2005-03-18 at 19:23 +0100, Andrej Prsa wrote:

> Tried it, unfortunately the outcome is still the same; I used the patch
> against 2.6.11.4 kernel *without* any other ACPI patch. Should I use the
> patch you provided over the ACPI patch or was it supposed to fix the fan
> issue by itself? Seemingly the problem is the fan; is there a way to
> test this to be sure?

Hmm. Yes, I can now repeat this with a 6220, though most of the time I
manage it it seems to be linked to a suspend to disk beforehand. I
suspect this /may/ be linked to the fact that on the 6120, only the
first lid event is received. I'll see if I can track this down over the
weekend.

-- 
Matthew Garrett | mjg59-1xO5oi07KQx4cg9Nei1l7Q@public.gmane.org



-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click

Re: HP/Compaq NC6120 ACPI problems

[Andrej Prsa]

Hi,

just a small observational update:

* applying the latest ACPI patch + the patch you recommended doesn't
  solve the issue
* during boot-up, the fan.ko module doesn't get loaded automatically
  and hence there is no /proc/acpi/fan entry. If I modprobe it manually,
  the fan/ directory is created with four subdirectories C253 through
  C256, each having only a state entry; all are set to off
* if I try to change them to on by 'echo 0 > ...' or 'echo on > ...',
  it doesn't work - they all remain set to "off".

> Hmm. Yes, I can now repeat this with a 6220, though most of the time I
> manage it it seems to be linked to a suspend to disk beforehand. I
> suspect this /may/ be linked to the fact that on the 6120, only the
> first lid event is received. I'll see if I can track this down over
> the weekend.

Thanks; if there's anything I can do to help, please let me know!

Bye,

Andrej


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click

Re: HP/Compaq NC6120 ACPI problems

[Stefan Seyfried]

On Sat, Mar 19, 2005 at 10:35:50AM +0100, Andrej Prsa wrote:
> Hi,
> 
> just a small observational update:
> 
> * applying the latest ACPI patch + the patch you recommended doesn't
>   solve the issue
> * during boot-up, the fan.ko module doesn't get loaded automatically
>   and hence there is no /proc/acpi/fan entry. If I modprobe it manually,
>   the fan/ directory is created with four subdirectories C253 through
>   C256, each having only a state entry; all are set to off

on my HP / Compaqs, each "fan" is for one speed level of the fan which
has 4 speeds (on nx500 and on an Armada E500)

> * if I try to change them to on by 'echo 0 > ...' or 'echo on > ...',
>   it doesn't work - they all remain set to "off".

echo 0 > ... => off
echo 3 > ... => on
-- 
Stefan Seyfried



-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click

Re: HP/Compaq NC6120 ACPI problems

[Matthew Garrett]

On Sat, 2005-03-19 at 03:19 +0000, Matthew Garrett wrote:

> Hmm. Yes, I can now repeat this with a 6220, though most of the time I
> manage it it seems to be linked to a suspend to disk beforehand. I
> suspect this /may/ be linked to the fact that on the 6120, only the
> first lid event is received. I'll see if I can track this down over the
> weekend.

Ha. Yes, progress, of sorts. On the 6120, you need to do

setpci -s 1f.0 bb.b=84

after every lid event in order to get any more. I /think/ this is a bug
with the HP BIOS (the 6220 doesn't do this) - I'm in touch with HP over
it now. As for the high kacpid load, I'm only able to trigger this after
a suspend to disk, and using the platform option (to ensure the _WAK
methods are called) /seems/ to avoid it. Could you try building a kernel
with acpi debug support, set the output level fairly high (echo -n
0x0000ffff >/proc/acpi/debug_level) and see what sort of output you get?

Warning - there'll be huge amounts of it. Really huge amounts. The stuff
we're interested in is whatever appears after kacpid has begun taking up
all the processor time.

-- 
Matthew Garrett | mjg59-1xO5oi07KQx4cg9Nei1l7Q@public.gmane.org



-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click

Re: HP/Compaq NC6120 ACPI problems

[Andrej Prsa]

Hi!

> > Hmm. Yes, I can now repeat this with a 6220, though most of the time
> > I manage it it seems to be linked to a suspend to disk beforehand. I
> > suspect this /may/ be linked to the fact that on the 6120, only the
> > first lid event is received. I'll see if I can track this down over
> > the weekend.
> 
> Ha. Yes, progress, of sorts. On the 6120, you need to do
> 
> setpci -s 1f.0 bb.b=84
> 
> after every lid event in order to get any more. I /think/ this is a
> bug with the HP BIOS (the 6220 doesn't do this) - I'm in touch with HP
> over it now.

I see; please let me know when you get any additional information and of
course if there's anything I can do to help! If it's not too much
trouble, could you (or anyone else of course) please explain what
exactly does this setpci do?

> As for the high kacpid load, I'm only able to trigger this after
> a suspend to disk, and using the platform option (to ensure the _WAK
> methods are called) /seems/ to avoid it. Could you try building a
> kernel with acpi debug support, set the output level fairly high (echo
> -n 0x0000ffff >/proc/acpi/debug_level) and see what sort of output you
> get?

Intriguingly, I recompiled the kernel with ACPI support built in (in
contrast to having a modular ACPI support) and this time kacpid
doesn't overload the processor and the fan is working!

Determined to come to the bottom of the problem, I recompiled the kernel
again, this time with ACPI support modular and voila - the same
malfunctioning behavior popped up. I 'tail -f'd both syslog and
messages; before kacpid blowup there was no significant output, but at
the time of kacpid blowup both logs started filling up like crazy to the
point that even init 0 didn't work anymore - all I got was a console
with ACPI debug messages filling screen after screen and I had to force
computer to a halt.

> Warning - there'll be huge amounts of it. Really huge amounts. The
> stuff we're interested in is whatever appears after kacpid has begun
> taking up all the processor time.

Please find the messages.gz file at (3.1MB compressed size!):

	http://www.fiz.uni-lj.si/~prsa/messages.gz

(sorry for the enormous size (~50MB uncompressed), but I didn't want to
cut out anything relevant!) The start of the messages file is the start
of kacpid going crazy.


The solution for me is of course ACPI built into the kernel, for it
seems that the fan is working that way. Still, I would like to help
out in resolving this issue, so let me know if there's anything else I
could do!

Thanks & best wishes,

Andrej


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click

Re: HP/Compaq NC6120 ACPI problems

[Matthew Garrett]

On Mon, 2005-03-21 at 11:40 +0100, Andrej Prsa wrote:

> I see; please let me know when you get any additional information and of
> course if there's anything I can do to help! If it's not too much
> trouble, could you (or anyone else of course) please explain what
> exactly does this setpci do?

The setpci sets a register that allows the GPE to fire again. The BIOS
ought to be doing it itself - I've reported it to HP.

> Intriguingly, I recompiled the kernel with ACPI support built in (in
> contrast to having a modular ACPI support) and this time kacpid
> doesn't overload the processor and the fan is working!

Ah. That /may/ fit with my observations from suspend to disk - the
hardware seems unhappy if Linux and the hardware disagree about the fan
state. My suspicion is that if there's a thermal trip event between
Linux switching the hardware to ACPI mode and the fan module being
loaded, you'll get the "mad kacpid" behaviour. I'd strongly recommend
having the fan module built in - there's the risk of hardware damage
otherwise.

> > Warning - there'll be huge amounts of it. Really huge amounts. The
> > stuff we're interested in is whatever appears after kacpid has begun
> > taking up all the processor time.
> 
> Please find the messages.gz file at (3.1MB compressed size!):
> 
> 	http://www.fiz.uni-lj.si/~prsa/messages.gz
> 
> (sorry for the enormous size (~50MB uncompressed), but I didn't want to
> cut out anything relevant!) The start of the messages file is the start
> of kacpid going crazy.

Great, thanks! I'll take a look at this later on. When I traced it on
the 6220, I found that it seemed to be related to a thermal zone event,
which would fit with observations.

-- 
Matthew Garrett | mjg59-1xO5oi07KQx4cg9Nei1l7Q@public.gmane.org



-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click

Re: HP/Compaq NC6120 ACPI problems

[Andrej Prsa]

Hi,

> > > Warning - there'll be huge amounts of it. Really huge amounts. The
> > > stuff we're interested in is whatever appears after kacpid has begun
> > > taking up all the processor time.
> > 
> > Please find the messages.gz file at (3.1MB compressed size!):
> > 
> > 	http://www.fiz.uni-lj.si/~prsa/messages.gz
> > 
> > (sorry for the enormous size (~50MB uncompressed), but I didn't want to
> > cut out anything relevant!) The start of the messages file is the start
> > of kacpid going crazy.
> 
> Great, thanks! I'll take a look at this later on. When I traced it on
> the 6220, I found that it seemed to be related to a thermal zone event,
> which would fit with observations.

any news from HP on this matter?

Andrej


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click

Re: HP/Compaq NC6120 ACPI problems

[Pavel Machek]

Hi!

> > Hmm. Yes, I can now repeat this with a 6220, though most of the time I
> > manage it it seems to be linked to a suspend to disk beforehand. I
> > suspect this /may/ be linked to the fact that on the 6120, only the
> > first lid event is received. I'll see if I can track this down over the
> > weekend.
> 
> Ha. Yes, progress, of sorts. On the 6120, you need to do
> 
> setpci -s 1f.0 bb.b=84
> 
> after every lid event in order to get any more. I /think/ this is a bug
> with the HP BIOS (the 6220 doesn't do this) - I'm in touch with HP over
> it now. As for the high kacpid load, I'm only able to trigger this after
> a suspend to disk, and using the platform option (to ensure the _WAK
> methods are called) /seems/ to avoid it. Could you try building a kernel

Well, running with "shutdown" is not really correct on acpi-enabled system.
Perhaps we should change the default?

				Pavel
-- 
64 bytes from 195.113.31.123: icmp_seq=28 ttl=51 time=448769.1 ms         



-------------------------------------------------------
This SF.net email is sponsored by Demarc:
A global provider of Threat Management Solutions.
Download our HomeAdmin security software for free today!
http://www.demarc.com/Info/Sentarus/hamr30

Re: HP/Compaq NC6120 ACPI problems

[Stefan Seyfried]

On Thu, Mar 31, 2005 at 10:27:24PM +0200, Pavel Machek wrote:
 
> Well, running with "shutdown" is not really correct on acpi-enabled system.
> Perhaps we should change the default?

i vote for it. I have not seen a single "platform"-mode failure for a long
time. At least on i386 :-)
-- 
Stefan Seyfried



-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click