[lm-sensors] Kernel module asus_atk0110

[lm-sensors] Kernel module asus_atk0110

[Charles]

Hello :-)

According to http://hansdegoede.livejournal.com/7932.html and
http://pavelshub.com/blog/2010/12/fatal-error-inserting-w83627ehf/, the
proper module to use on ASUS motherboards equipped with the Nuvoton
NCT6776F is asus_atk0110, not w83627ehf, but there is an open lm_sensors
ticket (http://www.lm-sensors.org/ticket/2374) about sensors-detect not
supporting is asus_atk0110.

Running Slackware64 13.1 with a 2.6.38.7 kernel on an ASUS P8H67-V
motherboard with an NCT6776F and asus_atk0110 modeprobed, sensors-detect
3.3.1 was not successful.  Digging deeper, looking for directories under
/sys/bus/acpi/drivers/ATK0110 found nothing:

root@CW8:~# ls -l /sys/bus/acpi/drivers/ATK0110
total 0
--w------- 1 root root 4.0K 2011-12-02 18:49 bind
--w------- 1 root root 4.0K 2011-12-02 18:49 uevent
--w------- 1 root root 4.0K 2011-12-02 18:49 unbind

Others have been able to use asus_atk0110 with lm_sensors (examples:
http://pastebin.com/8gREkwXZ,
http://comments.gmane.org/gmane.comp.gnome.apps.gkrellm/1556,
http://forums.opensuse.org/english/get-technical-help-here/pre-release-beta/423075-does-your-pwmconfig-find-modules-run.html;
there are many more).

The issue was worked around by using kernel option
acpi_enforce_resources=lax but I understand this will not work with
later kernels.

What to do for best?

Best

Charles

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

[lm-sensors] Kernel module asus_atk0110

[Charles]

Hello :-)

According to http://hansdegoede.livejournal.com/7932.html and
http://pavelshub.com/blog/2010/12/fatal-error-inserting-w83627ehf/, the
proper module to use on ASUS motherboards equipped with the Nuvoton
NCT6776F is asus_atk0110, not w83627ehf, but there is an open lm_sensors
ticket (http://www.lm-sensors.org/ticket/2374) about sensors-detect not
supporting is asus_atk0110.

Running Slackware64 13.1 with a 2.6.38.7 kernel on an ASUS P8H67-V
motherboard with an NCT6776F and asus_atk0110 modeprobed, sensors-detect
3.3.1 was not successful.  Digging deeper, looking for directories under
/sys/bus/acpi/drivers/ATK0110 found nothing:

root@CW8:~# ls -l /sys/bus/acpi/drivers/ATK0110
total 0
--w------- 1 root root 4.0K 2011-12-02 18:49 bind
--w------- 1 root root 4.0K 2011-12-02 18:49 uevent
--w------- 1 root root 4.0K 2011-12-02 18:49 unbind

Others have been able to use asus_atk0110 with lm_sensors (examples:
http://pastebin.com/8gREkwXZ,
http://comments.gmane.org/gmane.comp.gnome.apps.gkrellm/1556,
http://forums.opensuse.org/english/get-technical-help-here/pre-release-beta/423075-does-your-pwmconfig-find-modules-run.html;
there are many more).

The issue was worked around by using kernel option
acpi_enforce_resources=lax but I understand this will not work with
later kernels.

What to do for best?

Best

Charles

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] Kernel module asus_atk0110

[Luca Tettamanti]
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Re: [lm-sensors] Kernel module asus_atk0110

[Charles]

On 05/12/11 18:33, Luca Tettamanti wrote:
> Hi,
> 
[snip]
> 
> It's not easy to do, proper detection would require the full ACPI
> stack in userspace.
> 
[snip]
> You shouldn't need to load the driver manually; if it's not
> auto-loaded then your board is not recognized by the driver.
> ASUS has changed the ATK0110 interface from time to time, so hopefully
> a few tweaks is all that's needed.
> Please send me a copy of this file:
> /sys/firmware/acpi/tables/DSDT
> 
> Luca

Thanks Luca :)

Slackware loads (auto-loads?) modules via the /etc/rc.d/modules symlink.
 One problem may be that I am using a symlink target manually edited
from the 2.6.33.4 version instead of the 2.6.38.7 version.

ACPI modules in the original 2.6.33.4 version, commented out ready for
uncommenting by the sysadmin include (I have omitted "obviously"
irrelevant ones): button, fan, processor and thermal.  I disabled
acpi-cpufreq in favour of cpufreq_ondemand.  Would you like me to try
any changes?

I temporarily commented out w83627ehf and added asus_atk0110 then
removed kernel option acpi_enforce_resources=lax, rebooted and created a
file by catting /sys/firmware/acpi/tables/DSDT which I am sending you
off list.

Best

Charles





_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] Kernel module asus_atk0110

[Jean Delvare]

On Mon, 05 Dec 2011 19:44:30 +0530, Charles wrote:
> On 05/12/11 18:33, Luca Tettamanti wrote:
> > Hi,
> > 
> [snip]
> > 
> > It's not easy to do, proper detection would require the full ACPI
> > stack in userspace.
> > 
> [snip]
> > You shouldn't need to load the driver manually; if it's not
> > auto-loaded then your board is not recognized by the driver.
> > ASUS has changed the ATK0110 interface from time to time, so hopefully
> > a few tweaks is all that's needed.
> > Please send me a copy of this file:
> > /sys/firmware/acpi/tables/DSDT
> 
> Thanks Luca :)
> 
> Slackware loads (auto-loads?) modules via the /etc/rc.d/modules symlink.
>  One problem may be that I am using a symlink target manually edited
> from the 2.6.33.4 version instead of the 2.6.38.7 version.

The asus_atk0110 driver should be loaded by udev, this is not
distribution-dependent.

> ACPI modules in the original 2.6.33.4 version, commented out ready for
> uncommenting by the sysadmin include (I have omitted "obviously"
> irrelevant ones): button, fan, processor and thermal.  I disabled
> acpi-cpufreq in favour of cpufreq_ondemand.  Would you like me to try
> any changes?

acpi-cpufreq is a cpufreq driver, cpufreq_ondemand is a governor, you
can't trade one for the other. Not sure what you were trying to
achieve... This is not related to asus_atk0110 anyway.

> I temporarily commented out w83627ehf and added asus_atk0110 then
> removed kernel option acpi_enforce_resources=lax, rebooted and created a
> file by catting /sys/firmware/acpi/tables/DSDT which I am sending you
> off list.

The question is, does "sensors" output anything with only the
asus_atk0110 driver loaded? If not, hopefully Luca will be able to
figure out why from the DSDT.

-- 
Jean Delvare

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] Kernel module asus_atk0110

[Luca Tettamanti]

On Mon, Dec 5, 2011 at 3:43 PM, Jean Delvare <khali@linux-fr.org> wrote:
> On Mon, 05 Dec 2011 19:44:30 +0530, Charles wrote:
>> I temporarily commented out w83627ehf and added asus_atk0110 then
>> removed kernel option acpi_enforce_resources=lax, rebooted and created a
>> file by catting /sys/firmware/acpi/tables/DSDT which I am sending you
>> off list.
>
> The question is, does "sensors" output anything with only the
> asus_atk0110 driver loaded? If not, hopefully Luca will be able to
> figure out why from the DSDT.

ATK0110 is not present in the DSDT. The IO range is reserved and used
to check a single temperature against high and hysteresis thresholds
(see method _L16) and notify WMI (yeah...). The method also reads a
bit at 0x41 (AUIS) which is not used in the linux driver; the value is
discarded, it might be a read-to-clear alarm bit; Jean do you know
what that might be?
Anyway, the board does not expose a monitoring framework :-(

Luca
[1]  The disassembled DSDT: http://pastebin.com/YZX2uY3a

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] Kernel module asus_atk0110

[Charles]

On 05/12/11 20:58, Luca Tettamanti wrote:
> On Mon, Dec 5, 2011 at 3:43 PM, Jean Delvare <khali@linux-fr.org> wrote:
>>
>> The question is, does "sensors" output anything with only the
>> asus_atk0110 driver loaded? If not, hopefully Luca will be able to
>> figure out why from the DSDT.
> 
> ATK0110 is not present in the DSDT. The IO range is reserved and used
> to check a single temperature against high and hysteresis thresholds
> (see method _L16) and notify WMI (yeah...). The method also reads a
> bit at 0x41 (AUIS) which is not used in the linux driver; the value is
> discarded, it might be a read-to-clear alarm bit; Jean do you know
> what that might be?
> Anyway, the board does not expose a monitoring framework :-(
> 
> Luca
> [1]  The disassembled DSDT: http://pastebin.com/YZX2uY3a

Thanks Luca :)

That sounds pretty final for using asus_atk0110 on the ASUS P8H67-V :(

Not surprisingly, then, sensors only finds coretemps when asus_atk0110
is modprobed.

It's great that an ordinary and naive user can get this level of suport
from real experts and so quickly.  Many thanks to Jean, Guenter and Luca :)

Best

Charles


_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] Kernel module asus_atk0110

[Jean Delvare]

On Mon, 5 Dec 2011 16:28:19 +0100, Luca Tettamanti wrote:
> On Mon, Dec 5, 2011 at 3:43 PM, Jean Delvare <khali@linux-fr.org> wrote:
> > On Mon, 05 Dec 2011 19:44:30 +0530, Charles wrote:
> >> I temporarily commented out w83627ehf and added asus_atk0110 then
> >> removed kernel option acpi_enforce_resources=lax, rebooted and created a
> >> file by catting /sys/firmware/acpi/tables/DSDT which I am sending you
> >> off list.
> >
> > The question is, does "sensors" output anything with only the
> > asus_atk0110 driver loaded? If not, hopefully Luca will be able to
> > figure out why from the DSDT.
> 
> ATK0110 is not present in the DSDT. The IO range is reserved and used
> to check a single temperature against high and hysteresis thresholds
> (see method _L16) and notify WMI (yeah...). The method also reads a
> bit at 0x41 (AUIS) which is not used in the linux driver; the value is
> discarded, it might be a read-to-clear alarm bit; Jean do you know
> what that might be?

Register 0x41 is named Interrupt Status Register 1 in the datasheet,
bit 5 is named SOURCE2_SMI:
	A one indicates the high limit of SMIOVT SOURCE2 temperature has been exceeded.
	(CPUTIN is default temperature)

> Anyway, the board does not expose a monitoring framework :-(

But it exposes a register map (HWRE)... Might this be used to access
the hardware monitoring chip without interfering with ACPI? I'm not
saying that would be particularly nice, but I don't think we have any
other option, do we?

-- 
Jean Delvare

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] Kernel module asus_atk0110

[Luca Tettamanti]

On Mon, Dec 5, 2011 at 5:04 PM, Jean Delvare <khali@linux-fr.org> wrote:
> On Mon, 5 Dec 2011 16:28:19 +0100, Luca Tettamanti wrote:
[cut]
>> Anyway, the board does not expose a monitoring framework :-(
>
> But it exposes a register map (HWRE)... Might this be used to access
> the hardware monitoring chip without interfering with ACPI?

Not directly: other boards had read/write methods with byte
granularity that could be exploited to jury rig a native driver to
access the hardware via ACPI. In this case there are no such methods,
the only option that I see would be injecting new methods into the
DSDT and use those; it can be done but it's a bit... distasteful ;-)
And fragile.
I wonder that ASUS does under Windows...

Luca

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors