HP laptop CPU temps not seen by kernel

[67GTA]

My laptop's CPU temps are not read by the kernel. This also causes the thermal trip points to not be triggered, makes the laptop overheat, and causes the fan to run constantly. This has been present in every kernel version. The cause is a buggy dsdt. I have used a custom dsdt to fix this. The dsdt bug is referenced here for a different model: https://wiki.edubuntu.org/LaptopTestingTeam/HPdv5z I have also attached my original and custom dsdt's. The offending lines in the original are 7714-7730.

ProblemType: Bug
Architecture: i386
DistroRelease: Ubuntu 9.04
HibernationDevice: RESUME=UUID=4fb40fc9-3020-4da4-836b-3e209ff74240
Lsusb:
 Bus 002 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
 Bus 004 Device 001: ID 1d6b:0001 Linux Foundation 1.1 root hub
 Bus 001 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
 Bus 003 Device 002: ID 045e:00e1 Microsoft Corp. Wireless Laser Mouse 6000 Reciever
 Bus 003 Device 001: ID 1d6b:0001 Linux Foundation 1.1 root hub
MachineType: Hewlett-Packard HP Pavilion dv6700 Notebook PC
NonfreeKernelModules: nvidia
Package: linux-image-2.6.28-15-generic 2.6.28-15.48
ProcCmdLine: root=UUID=7d423cf8-a1c7-4889-b791-9a2b308908fa ro quiet splash acpi_osi="Linux" rootflags=data=writeback
ProcEnviron:
 LANG=en_US.UTF-8
 SHELL=/bin/bash
ProcVersionSignature: Ubuntu 2.6.28-15.48-generic
SourcePackage: linux

[67GTA (67gta)]

I will add that my dsdt was compiled by Nvidia with the MSFT compiler which is not ACPI compliant. The IASL compiler makes 1130 optimizations on top of the errors I fixed.

[67GTA (67gta)]

dmesg line without custom dsdt: [ 1.714689] ACPI Exception (thermal-0377): AE_OK, No or invalid critical threshold [20080926]

dmesg line with custom dsdt: [ 1.710351] ACPI: Thermal Zone [THRM] (51 C)

[67GTA (67gta)]

dmesg line without custom dsdt: [ 1.714689] ACPI Exception (thermal-0377): AE_OK, No or invalid critical threshold [20080926]

dmesg line with custom dsdt: [ 1.710351] ACPI: Thermal Zone [THRM] (51 C)

Without the custom dsdt, /proc/acpi/thermal-zone/THRM is empty.

With the custom dsdt it is populated with trip points, temperature, cooling_mode, polling_frequency, and state.

[Clarkt-cnsp (clarkt-cnsp)]

Created an attachment (id=313114)
dmesg output from 11.1 showing thermal is seen and used

User-Agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9.1.2) Gecko/20090730 SUSE/3.5.2-1.1 Firefox/3.5.2

HP updated the bios for several dv series laptops. This enabled thermal management and other acpi fixes for linux. In the 2.6.27 kernel, cpu temp is read as a valid value. In the 11.2 kernels (30 and 31) this feature is broken.

Reproducible: Always

Steps to Reproduce:
1.install 11.1 on a HP laptop with the bios update for linux thermal and acpi
2.cpu temp is valid and shown. thermal events work.
3.install 11.2 MS5.
4. thermal events are not seen and acpi reports that the value in dsdt is invalid
Actual Results:
11.1 -- thermal management works.
11.2 -- thermal management does not work

Expected Results:
with bios update, the cpu temp should be seen and thermal management should work.

[Clarkt-cnsp (clarkt-cnsp)]

Created an attachment (id=313115)
dmsg output from 11.2 showing thermal is seen as invalid

dmesg output from 11.2 MS5 showing that the current kernel cannot see thermal values and that it considers the dsdt values to be invalid.

[Clarkt-cnsp (clarkt-cnsp)]

Created an attachment (id=313116)
dsdt dump

dsdt dump of the bios with the thermal update

[67GTA (67gta)]

This affects me also. I have already filed a bug on launchpad https://bugs.launchpad.net/ubuntu/+source/linux/+bug/412167 This is caused by a buggy DSDT. BIOS updates didn't change mine. I have to use a custom DSDT to fix. Here is a link to the exact problem within the DSDT https://wiki.edubuntu.org/LaptopTestingTeam/HPdv5z Now, custom DSDT support is being removed in every distro, so I'm filing a bug.

[Clarkt-cnsp (clarkt-cnsp)]

Vendors fix the bios for thermal then the kernel team breaks it. What do we have to do to get the thermal readings back in the newer kernels?

[Jdelvare (jdelvare)]

This seems to be related to ACPI only, then it's not something for me, sorry.

Clark, it would help if you would let us know where you were reading the CPU temperature from before it broke.

[Clarkt-cnsp (clarkt-cnsp)]

From the dmesg output that is attached to this bug:

    2.453455] processor LNXCPU:01: registered as cooling_device1
[ 2.464542] ACPI Exception: AE_OK, No or invalid critical threshold 20090521 thermal-385

Nothing in kde4 will not see temps, and the system is running hotter in 11.2 than 11.1

[Trenn-novell (trenn-novell)]

This:
[ 2.464542] ACPI Exception: AE_OK, No or invalid critical threshold 20090521
thermal-385
means that the critical temperature is below 0 degree Celsius.
Please attach acpidump output.

> Nothing in kde4 will not see temps
This may be a kde4, but not a kernel bug, let's see..

What does:
cat /proc/acpi/thermal_zone/*/temperature
or better:
cat /proc/acpi/thermal_zone/*/*
tell us?

[Clarkt-cnsp (clarkt-cnsp)]

linux-hwky:~ # cat /proc/acpi/thermal_zone/*/temperature
cat: /proc/acpi/thermal_zone/*/temperature: No such file or directory
linux-hwky:~ # cat /proc/acpi/thermal_zone/*/*
cat: /proc/acpi/thermal_zone/*/*: No such file or directory

[Clarkt-cnsp (clarkt-cnsp)]

Created an attachment (id=320492)
acpi dump

acpidump from 11.2 2.6.31

[Clarkt-cnsp (clarkt-cnsp)]

output from 11.1 2.6.29

linux-g9wc:/home/clark # cat /proc/acpi/thermal_zone/*/temperature
temperature: 57 C
linux-g9wc:/home/clark # cat /proc/acpi/thermal_zone/*/*
<setting not supported>
<polling disabled>
state: ok
temperature: 56 C
critical (S5): 429496456 C
hot (S4): 100 C
passive: 95 C: tc1=2 tc2=3 tsp=30 devices=CPU0 CPU1

[Clarkt-cnsp (clarkt-cnsp)]

Created an attachment (id=320494)
acpidump from 11.1

acpidump output from 11.1 2.6.29

[Jdelvare (jdelvare)]

What a terrific critical temperature :p

[Trenn-novell (trenn-novell)]

I am very sorry to come back that late...
> What a terrific critical temperature
Yep, this is a BIOS bug, but nevertheless the thermal zone shouldn't get ignored totally.
I added a patch which should show up in some hours here:
ftp://ftp.suse.com/pub/projects/kernel/kotd/master/x86_64
Best double check if the change is included:
rpm -qp --changelog kernel-default.rpm |less
Fri Feb 12 16:57:20 CET 2010 - <email address hidden>

- patches.fixes/acpi_fix_no_critical_tp.patch: ACPI thermal:
  Don't invalidate thermal zone if critical trip point is bad
  (bnc#531547).

Then install the kernel additionally (like that you can just test boot it and you can fall back to your old kernel matching the distri):
rpm -ivh kernel-default.rpm kernel-base.rpm # eventually kernel-desktop,
                                              # doesn't matter
Eventually edit /boot/grub/menu.lst to make sure the kernel gets booted or selected on next reboot in grub boot menu.

[Trenn-novell (trenn-novell)]

If you can confirm that the thermal zone now shows valid temperatures again, I will submit the patch mainline. Thanks.

[Clarkt-cnsp (clarkt-cnsp)]

The files in the kotd have not been updated since the 10th.

[Trenn-novell (trenn-novell)]

Yep, kotd seem to be broken for current master branch.
Possibly Michael can help, but this may not have highest prio atm.

[Trenn-novell (trenn-novell)]

From <email address hidden> (yesterday):
> Is it a known problem that KOTD HEAD wasn't updated since 10-Feb-2010 (
> version 2.6.33-2.99.14.0943949) ?

Yes. It was caused by a stale NFS handle. Michal fixed that a few hours ago,
the kernels should be building now.

[Clarkt-cnsp (clarkt-cnsp)]

Installed the default kernel and it's now working. When do we expect to see this pushed into the normal kernel updates?

[Trenn-novell (trenn-novell)]

Let's get this mainline first.
I'll add you with
Tested-by: <email address hidden>
in the patch and send it together with some other stuff off the next days.
IMO this issue is not worth backporting. From the thermal zone, there is only one passive (95C) trip point, no active ones, thus I expect you don't have any bad side effects but you cannot see the CPU temperature. If your fans are not strong enough, you may see thermal shutdowns instead of passive cooling at 95C.

[Len Brown (len-brown)]

upstream in 2.6.34-rc1:

commit fa80945269f312bc609e8384302f58b03c916e12
Author: Thomas Renninger <email address hidden>
Date: Sat Feb 20 11:44:27 2010 +0100

    ACPI thermal: Don't invalidate thermal zone if critical trip point is bad

    V2: Corrected integer/long conversion.

[Len Brown (len-brown)]

actually that is 2.6.34-rc2, if anybody is checking...

[Johan Kiviniemi (ion)]

fa80945269f312bc609e8384302f58b03c916e12 from Linus’ branch applies to the current lucid branch cleanly when cherry-picked.

[67GTA (67gta)]

This bug is fixed in Opensuse 11.3 M4 with 2.6.33-1.

[Trenn-novell (trenn-novell)]

Fixed mainline and should be included in 11.3 and 11.4.

[Vitaly Zawullon Katraev (zawullon)]

Bug still present in Natty (Ubuntu 11.04)

[Brad Figg (brad-figg)]

This bug was filed against a series that is no longer supported and so is being marked as Won't Fix. If this issue still exists in a supported series, please file a new bug.

This change has been made by an automated script, maintained by the Ubuntu Kernel Team.