[lm-sensors] coretemp - digital temperature driver for Intel Core

[lm-sensors] coretemp - digital temperature driver for Intel Core

[Rudolf Marek]

Hello all,

I had today some time and coded the driver for Core 2 processors which will read 
the temperature from the CPU. (precisely I worked on it from 16:36 to 23:59 with 
some preemptions ;)

I dont have such CPU so I'm looking for brave testers!

The code is "virgin" it compiles but that is all I know. Testers should expect 
oops, panic or simple crash... Or maybe if I did not make any mistake (hehe) the 
temperature in /sys/class/hwmon/correcthwmondir/temp1_input

To compile:
make -C /lib/modules/`uname -r`/build SUBDIRS=$PWD  modules

Again this should be for either brave testers or persons with the knowledge of C 
... I will review the code tomorrow again.

Good luck and thanks!

regards
Rudolf
-------------- next part --------------
A non-text attachment was scrubbed...
Name: coretemp.c
Type: application/octet
Size: 8780 bytes
Desc: not available
Url : http://lists.lm-sensors.org/pipermail/lm-sensors/attachments/20060927/5ad49cd3/attachment.bin 
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: Makefile
Url: http://lists.lm-sensors.org/pipermail/lm-sensors/attachments/20060927/5ad49cd3/attachment.pl 

[lm-sensors] coretemp - digital temperature driver for Intel

[Rudolf Marek]

Hi again,


Sorry for one more email. I should really go sleep...
It should work on Core Duo/Solo without "2" too.

cat /proc/cpuinfo should show:
cpu family      : 6
model           : 14

or:

cpu family      : 6
model           : 15

It does not support CPU hotplug yet...

Rudolf

[lm-sensors] coretemp - digital temperature driver for Intel

[Michael Nelson]

On Wed, Sep 27, 2006 at 12:12:17AM +0200, Rudolf Marek wrote:
> Hello all,
> 
> I had today some time and coded the driver for Core 2 processors which will 
> read the temperature from the CPU. (precisely I worked on it from 16:36 to 
> 23:59 with some preemptions ;)
> 
> I dont have such CPU so I'm looking for brave testers!
> 
> The code is "virgin" it compiles but that is all I know. Testers should 
> expect oops, panic or simple crash... Or maybe if I did not make any 
> mistake (hehe) the temperature in 
> /sys/class/hwmon/correcthwmondir/temp1_input
> 
> To compile:
> make -C /lib/modules/`uname -r`/build SUBDIRS=$PWD  modules
> 
> Again this should be for either brave testers or persons with the knowledge 
> of C ... I will review the code tomorrow again.

Doesn't seem to work here with a Core2 Duo 6400 and kernel 2.6.18:

  cat /sys/class/hwmon/hwmon1/device/temp1_input
0

Thanks for the effort though!

Michael

-- 

San Francisco, CA

[lm-sensors] coretemp - digital temperature driver for Intel

[Rudolf Marek]

Hello,

Thanks for the test. It is cool but it does not work ;)

> Doesn't seem to work here with a Core2 Duo 6400 and kernel 2.6.18:
> 
>   cat /sys/class/hwmon/hwmon1/device/temp1_input
> 0

Ehm.


Please add this two printk to this code:

              data->therm_status = eax;
              printk("EAX from  MSR %x\n", eax);

                 /* update only if data has been valid */
                 if (eax & 0x80000000) {
			printk("in temp condition\n");

Run it again, to see the 0 but check the log...

Thanks,

regards
Rudolf

[lm-sensors] coretemp - digital temperature driver for Intel

[Michael Nelson]

On Wed, Sep 27, 2006 at 08:20:51AM +0200, Rudolf Marek wrote:

> Please add this two printk to this code:
> 
>               data->therm_status = eax;
>               printk("EAX from  MSR %x\n", eax);
> 
>                  /* update only if data has been valid */
>                  if (eax & 0x80000000) {
> 			printk("in temp condition\n");

Uhhhh... add them WHERE?

BTW, the hw monitoring chip on my motherboard (Asus P5B Deluxe WiFi) already
does report CPU temp.

Michael

-- 

San Francisco, CA

[lm-sensors] coretemp - digital temperature driver for Intel

[Rudolf Marek]

Michael Nelson wrote:
> On Wed, Sep 27, 2006 at 08:20:51AM +0200, Rudolf Marek wrote:
>
>   
>> Please add this two printk to this code:
>>
>>               data->therm_status = eax;
>>               printk("EAX from  MSR %x\n", eax);
>>
>>                  /* update only if data has been valid */
>>                  if (eax & 0x80000000) {
>> 			printk("in temp condition\n");
>>     
>
> Uhhhh... add them WHERE?
>
>   
Find this line:

  data->therm_status = eax;

on next line add

 printk("EAX from  MSR %x\n", eax);

and find this line:
if (eax & 0x80000000) {
and on next line add

printk("in temp condition\n");


Thanks.

> BTW, the hw monitoring chip on my motherboard (Asus P5B Deluxe WiFi) already
> does report CPU temp.
>   
Yes but I need to test this ;)

Thanks,
Rudolf

[lm-sensors] coretemp - digital temperature driver for Intel

[Michael Nelson]

On Wed, Sep 27, 2006 at 12:38:30PM +0200, Rudolf Marek wrote:

> Find this line:
> 
>   data->therm_status = eax;
> 
> on next line add
> 
>  printk("EAX from  MSR %x\n", eax);
> 
> and find this line:
> if (eax & 0x80000000) {
> and on next line add
> 
> printk("in temp condition\n");

Ok, I did that, recompiled it using 

make -C /lib/modules/`uname -r`/build SUBDIRS=$PWD  modules

Then copied it to:

/lib/modules/2.6.18/kernel/drivers/hwmon/coretemp.ko

depmod -a

# modprobe coretemp
# lsmod | grep core
#

Hmmmm

The end of dmesg says:

kobject_add failed for coretemp.0 with -EEXIST, don't try to register things
with the same name in the same directory.
 [<c01b32fd>] kobject_add+0x10d/0x180
 [<c02109f6>] device_add+0x86/0x2c0
 [<c0213c24>] platform_device_add+0xe4/0x150
 [<f92640c5>] coretemp_init+0xc5/0x15f [coretemp]
 [<c0135d4c>] sys_init_module+0x12c/0x1ac0
 [<c0102c5d>] sysenter_past_esp+0x56/0x79
coretemp: Device addition failed (-17)

What's next?

Michael

-- 

San Francisco, CA

[lm-sensors] coretemp - digital temperature driver for Intel

[Rudolf Marek]

Michael Nelson wrote:
> On Wed, Sep 27, 2006 at 12:38:30PM +0200, Rudolf Marek wrote:
>
>   
>> Find this line:
>>
>>   data->therm_status = eax;
>>
>> on next line add
>>
>>  printk("EAX from  MSR %x\n", eax);
>>
>> and find this line:
>> if (eax & 0x80000000) {
>> and on next line add
>>
>> printk("in temp condition\n");
>>     
>
> Ok, I did that, recompiled it using 
>
> make -C /lib/modules/`uname -r`/build SUBDIRS=$PWD  modules
>
> Then copied it to:
>
> /lib/modules/2.6.18/kernel/drivers/hwmon/coretemp.ko
>
> depmod -a
>
> # modprobe coretemp
> # lsmod | grep core
> #
>
> Hmmmm
>
> The end of dmesg says:
>
> kobject_add failed for coretemp.0 with -EEXIST, don't try to register things
> with the same name in the same directory.
>  [<c01b32fd>] kobject_add+0x10d/0x180
>  [<c02109f6>] device_add+0x86/0x2c0
>  [<c0213c24>] platform_device_add+0xe4/0x150
>  [<f92640c5>] coretemp_init+0xc5/0x15f [coretemp]
>  [<c0135d4c>] sys_init_module+0x12c/0x1ac0
>  [<c0102c5d>] sysenter_past_esp+0x56/0x79
> coretemp: Device addition failed (-17)
>
> What's next?
>   
I think something has been left out in kernel. Did you rmmoded the 
driver before inserting the new one?
If you rmmoded, there is some problem. Please reboot and try to load the 
driver again.

Thanks
Regards
Rudolf

[lm-sensors] coretemp - digital temperature driver for Intel

[Michael Nelson]

On Wed, Sep 27, 2006 at 03:49:25PM +0200, Rudolf Marek wrote:

> I think something has been left out in kernel. Did you rmmoded the 
> driver before inserting the new one?
> If you rmmoded, there is some problem. Please reboot and try to load the 
> driver again.

OK I did, and it loaded.  

seahunt:/var/log # cat /sys/class/hwmon/hwmon0/device/temp1_input
29000
seahunt:/var/log # cat /sys/class/hwmon/hwmon1/device/temp1_input
0
seahunt:/var/log # cat /sys/class/hwmon/hwmon2/device/temp1_input
0

So does that 29000 mean it's working?  The hwmon chip says 53.0C right now.

Michael

-- 

San Francisco, CA

[lm-sensors] coretemp - digital temperature driver for Intel

[Michael Nelson]

I also get this:

# grep EAX /var/log/messages
Sep 27 07:15:46 seahunt kernel: EAX from  MSR 2
Sep 27 07:16:07 seahunt kernel: EAX from  MSR 2
Sep 27 07:16:12 seahunt kernel: EAX from  MSR 2


-- 

San Francisco, CA

[lm-sensors] coretemp - digital temperature driver for Intel

[Rudolf Marek]

Michael Nelson wrote:
> I also get this:
>
> # grep EAX /var/log/messages
> Sep 27 07:15:46 seahunt kernel: EAX from  MSR 2
> Sep 27 07:16:07 seahunt kernel: EAX from  MSR 2
>   
Did you add the second printk too?

Is it displayed?

I think there is something wrong I will take a look maybe in the evening...

regards
Rudolf

[lm-sensors] coretemp - digital temperature driver for Intel

[Michael Nelson]

On Wed, Sep 27, 2006 at 04:28:59PM +0200, Rudolf Marek wrote:

> Did you add the second printk too?

I did.

> Is it displayed?

No:

# grep condition /var/log/messages
# 

> I think there is something wrong I will take a look maybe in the evening...

OK, I will be glad to test further.

Thank you
Michael

-- 

San Francisco, CA

[lm-sensors] coretemp - digital temperature driver for Intel

[Rudolf Marek]

> OK, I will be glad to test further.
>   

Ok I think I found it.:

Please change this:
             do_rdmsr(data->id, MSR_IA32_THERM_CONTROL, &eax, &edx);

Into this:

             do_rdmsr(data->id, MSR_IA32_THERM_STATUS, &eax, &edx);

It should work ...

Thanks,

Rudolf

[lm-sensors] coretemp - digital temperature driver for Intel

[Michael Nelson]

On Wed, Sep 27, 2006 at 06:49:00PM +0200, Rudolf Marek wrote:
> 
> > OK, I will be glad to test further.
> >   
> 
> Ok I think I found it.:
> 
> Please change this:
>              do_rdmsr(data->id, MSR_IA32_THERM_CONTROL, &eax, &edx);
> 
> Into this:
> 
>              do_rdmsr(data->id, MSR_IA32_THERM_STATUS, &eax, &edx);
> 
> It should work ...

I made the change and recompiled, and even after verifying that the previous
one had been rmmodded, the new one won't insert, gives the same error as
before when I try to modprobe it.  

Can you please fix it so it can be modprobed and rmmodded without having to
reboot?  I was going to test this while ssh'd in from work but don't want to
reboot the machine.

Thank you
Michael

-- 

San Francisco, CA

[lm-sensors] coretemp - digital temperature driver for Intel

[Rudolf Marek]

t.
> 
> Can you please fix it so it can be modprobed and rmmodded without having to
> reboot?  I was going to test this while ssh'd in from work but don't want to
> reboot the machine.

I think I found that, please use attached version.

Thanks,
Rudolf
-------------- next part --------------
A non-text attachment was scrubbed...
Name: coretemp.c
Type: application/octet
Size: 8867 bytes
Desc: not available
Url : http://lists.lm-sensors.org/pipermail/lm-sensors/attachments/20060928/1a9f30ad/attachment.bin 

[lm-sensors] coretemp - digital temperature driver for Intel

[Michael Nelson]

On Thu, Sep 28, 2006 at 12:11:30AM +0200, Rudolf Marek wrote:

> I think I found that, please use attached version.

OK, here's the latest:

seahunt:/var/log # cat /sys/class/hwmon/hwmon0/device/temp1_input
31000
seahunt:/var/log # cat /sys/class/hwmon/hwmon1/device/temp1_input
58000
seahunt:/var/log # cat /sys/class/hwmon/hwmon2/device/temp1_input
58000
seahunt:/var/log # sensors | grep Temp
MB Temp:     +31??C  (high =   +35??C, hyst =   +40??C)
CPU Temp:  +55.5??C  (high = +70.0??C, hyst = +40.0??C)

Assuming 58000 means 58 degrees C and 31000 means 31 degrees C, it looks OK?

Thanks
Michael

-- 

San Francisco, CA

[lm-sensors] coretemp - digital temperature driver for Intel

[Rudolf Marek]

Hello all,

Please check the new version on http://assembler.cz/download/coretemp/

It has:

1) code improvements
2) support for critical temperature alarm
3) shows the TjMax temperature
4) hotplug CPU support
5) new bugs? ;)

there is also a patch for userspace - should be applied on top of the SVN HEAD.

Please test it and let me know how the output of "sensors" command looks like. I
believe this version is less dangerous to use than a first one, but of course I
cannot guarantee that. All I know is that it compiles...

Any testing is appreciated.

Thanks,

Rudolf

[lm-sensors] coretemp - digital temperature driver for

[Matthias Hentges]

Hello Rudolf,

I'm running a C2D (E6600) system here and would be happy to be of
assistance.

I've downloaded and compiled your latest version (I think heh) but the
resulting temperature is a bit off I think.

The CPU temp as reported by the SuperIO chip is 23.5 degrees,
you driver reports about 37000 (37 degrees? and which core? or is that
supposed to be the m/b diode?)
a little c program to read the coretemp from the CPU I have found
elsewhere on the net reports ~51 degrees for both cores on the same idle
system (the little program is attached, it's header says it's GPL).

If you need any testing done let me know. A hwmon sysfs interface to
both(!) coretemps would be most welcome =)



-- 
Matthias Hentges 

My OS: Debian SID. Geek by Nature, Linux by Choice
-------------- next part --------------
A non-text attachment was scrubbed...
Name: coretemp.c
Type: text/x-csrc
Size: 2435 bytes
Desc: not available
Url : http://lists.lm-sensors.org/pipermail/lm-sensors/attachments/20060930/8e36379b/attachment.bin 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Dies ist ein digital signierter Nachrichtenteil
Url : http://lists.lm-sensors.org/pipermail/lm-sensors/attachments/20060930/8e36379b/attachment-0001.bin 

[lm-sensors] coretemp - digital temperature driver for Intel

[Rudolf Marek]

Hello Matthias,

> I'm running a C2D (E6600) system here and would be happy to be of
> assistance.

Good thanks. What motherboard is it. Is it the desktop/laptop?

> I've downloaded and compiled your latest version (I think heh) but the
> resulting temperature is a bit off I think.
> 
> The CPU temp as reported by the SuperIO chip is 23.5 degrees,
> you driver reports about 37000 (37 degrees? and which core? or is that
> supposed to be the m/b diode?)

This is even better than the diode, it is the value from special build-in
digital temperature sensor. The value from superio could be

1) from thermistor bellow the CPU socket
2) thermal diode value

Can you paste the output of "sensors" command please?

> a little c program to read the coretemp from the CPU I have found
> elsewhere on the net reports ~51 degrees for both cores on the same idle
> system (the little program is attached, it's header says it's GPL).

Well it there is a bit problem. To calculate the temperature, you need to know
so called "TjMax" which that program sets fixed to 100C, however my
implementation detect this TjMax and set it to 100C or 85C.

85C should have systems for other than LGA socket.

Toutput = TjMax - registerValue

So If the TjMax is wrong, output temperature is also wrong (offset 15C)

> If you need any testing done let me know. A hwmon sysfs interface to
> both(!) coretemps would be most welcome =)

I think it very good to have such driver so I "donated" my time for this driver
development ;)

Well it should be for both cores. Have you used this version?
http://assembler.cz/download/coretemp/

There is a patch for lm-sensors so you just need to run "sensors" and it should
show the temps for both cores. Please can you cut&paste the output of sensors
command to the mail?

>     return 100 - data[2];
It uses fixed 100 value.

Please can you provide output of following files:
cat /sys/class/hwmon/hwmon0/device/temp1_input
cat /sys/class/hwmon/hwmon0/device/temp1_crit

cat /sys/class/hwmon/hwmon1/device/temp1_input
cat /sys/class/hwmon/hwmon1/device/temp1_crit

cat /sys/class/hwmon/hwmon2/device/temp1_input
cat /sys/class/hwmon/hwmon2/device/temp1_crit

I think we will see the crit temperature as 85000 (so it is 85C).

Question is what is correct. Please can you try to reboot and check the BIOS?
Also how much the temperature grows when you run in linux when running following
command:
dd if=/dev/zero of=/dev/null bs=1M count@00

I'm also on IRC irc.freenode.net #linux-sensors

Thank you for help,
Regards
Rudolf