[lm-sensors] Question on my mainboard

[lm-sensors] Question on my mainboard

[Dieter Jurzitza]

Dear listmembers,
I am using several Tyan Tiger 2460 motherboards since quite a while. And I am 
observing a strange behaviour someone in the list (might) have a clue about.

I am reading the sensors data periodically every 10 seconds. After an 
undefined time (maybe one week, maybe one day, maybe three weeks) sensors 
"hangs" after start and cannot be killed any more. A manual call to sensors 
ends in a hang as well (of sensors, the board is running on).

What is my configuration:
I am loading 
MODULE_0=i2c-amd756
MODULE_1=i2c-isa
MODULE_2=w83781d
MODULE_3îprom
MODULE_4=w83627hf
(in this sequence).

By recommendation from Tyan I use the line:
options w83781d force_w83782d=0,0x2d force_subclients=0,0x2d,0x48,0x49 
force_w83627hf=0,0x2c force_subclients=0,0x2c,0x4a,0x4b init=0

in /etc/modprobe.conf.local

I have a script watchtemp that is started at boot that calls sensors every 10 
seconds; at boot it is called once with the parameter DAEMON, later on only 
once with watchtemp what gives me a temperature display on the screen.

If someone finds the time to look into it and may give me a hint - greatly 
appreciated. I cannot force the behaviour described above - it just happens 
to occur after a while.

Besides the many "SuSEisms" in the script it (might) be useful for someone 
because it is very simple. If of interest I could provide a rpm / srpm for 
SuSE installing all the required components at the right places.

I am using sensors version 2.10.0 with libsensors version 2.10.0 on SuSE 9.3 
SMP.

Many thanks for your efforts in advance,
take care



Dieter Jurzitza




-- 
-----------------------------------------------------------

                               |
                                \
                 /\_/\           |
                | ~x~ |/-----\   /
                 \   /-       \_/
  ^^__   _        /  _  ____   /
 <??__ \- \_/     |  |/    |  |
  ||  ||         _| _|    _| _|

if you really want to see the pictures above - use some font
with constant spacing like courier! :-)
-----------------------------------------------------------
-------------- next part --------------
A non-text attachment was scrubbed...
Name: watchtemp
Type: application/x-shellscript
Size: 2731 bytes
Desc: not available
Url : http://lists.lm-sensors.org/pipermail/lm-sensors/attachments/20060328/72b1c2ad/watchtemp.bin

[lm-sensors] Question on my mainboard

[Rudolf Marek]

Hello Dieter,

> Dear listmembers,
> I am using several Tyan Tiger 2460 motherboards since quite a while. And I am 
> observing a strange behaviour someone in the list (might) have a clue about.
> 
> I am reading the sensors data periodically every 10 seconds. After an 
> undefined time (maybe one week, maybe one day, maybe three weeks) sensors 
> "hangs" after start and cannot be killed any more. A manual call to sensors 
> ends in a hang as well (of sensors, the board is running on).
> 

Well perhaps I will have future questions:

Hmm this seems some race condition in the kernel maybe? If you do ps ax
can you see the D state of the process? Any OOPs message in the log?

Do you use preempt? (perhaps not)
Does cat hang if you do for example cat some value from /sys/bus/i2c/devices.../temp1_input

Regards
Rudolf

[lm-sensors] Question on my mainboard

[Jean Delvare]

Hi Dieter,

> I am using several Tyan Tiger 2460 motherboards since quite a while. And I am 
> observing a strange behaviour someone in the list (might) have a clue about.
> 
> I am reading the sensors data periodically every 10 seconds. After an 
> undefined time (maybe one week, maybe one day, maybe three weeks) sensors 
> "hangs" after start and cannot be killed any more. A manual call to sensors 
> ends in a hang as well (of sensors, the board is running on).
> 
> What is my configuration:
> I am loading 
> MODULE_0=i2c-amd756
> MODULE_1=i2c-isa
> MODULE_2=w83781d
> MODULE_3îprom
> MODULE_4=w83627hf
> (in this sequence).
> 
> By recommendation from Tyan I use the line:
> options w83781d force_w83782d=0,0x2d force_subclients=0,0x2d,0x48,0x49 
> force_w83627hf=0,0x2c force_subclients=0,0x2c,0x4a,0x4b init=0
> 
> in /etc/modprobe.conf.local

This doesn't make much sense. You ask the w83781d driver to take care
of the W83627HF chip, but also load the w83627hf driver. Both drivers
will attempt to handle the same chip, this might explain the deadlock.

I'd suggest that you try without loading the w83627hf driver. If the
w83781d driver is able to handle both chips (the W83782D and the
W83627HF) then the w83627hf driver is not needed at all.

-- 
Jean Delvare

[lm-sensors] Question on my mainboard

[Steven Timm]

On Fri, 31 Mar 2006, Jean Delvare wrote:

> Hi Dieter,
>
>> I am using several Tyan Tiger 2460 motherboards since quite a while. And I am
>> observing a strange behaviour someone in the list (might) have a clue about.
>>
>> I am reading the sensors data periodically every 10 seconds. After an
>> undefined time (maybe one week, maybe one day, maybe three weeks) sensors
>> "hangs" after start and cannot be killed any more. A manual call to sensors
>> ends in a hang as well (of sensors, the board is running on).
>>
>> What is my configuration:
>> I am loading
>> MODULE_0=i2c-amd756
>> MODULE_1=i2c-isa
>> MODULE_2=w83781d
>> MODULE_3îprom
>> MODULE_4=w83627hf
>> (in this sequence).
>>
>> By recommendation from Tyan I use the line:
>> options w83781d force_w83782d=0,0x2d force_subclients=0,0x2d,0x48,0x49
>> force_w83627hf=0,0x2c force_subclients=0,0x2c,0x4a,0x4b init=0
>>
>> in /etc/modprobe.conf.local
>
> This doesn't make much sense. You ask the w83781d driver to take care
> of the W83627HF chip, but also load the w83627hf driver. Both drivers
> will attempt to handle the same chip, this might explain the deadlock.

This board and other Tyan boards of that era have actually got two chips, 
one of which is the W83627HF, and all of them have got some strange
subclients like that which are necessary.  When I was running that
board there was a custom sensors.conf that was necessary too to
interpret the output correctly, the stock one that comes with
the source doesn't cut it.  At that time Tyan had people who
made it for us, not sure if those folks are still with Tyan or not.
On the closely-related tyan 2466 board, I was following what
Jean said and only loading the W83781d driver.  That should work.

Steve


>
> I'd suggest that you try without loading the w83627hf driver. If the
> w83781d driver is able to handle both chips (the W83782D and the
> W83627HF) then the w83627hf driver is not needed at all.
>
>

-- 
------------------------------------------------------------------
Steven C. Timm, Ph.D  (630) 840-8525  timm at fnal.gov  http://home.fnal.gov/~timm/
Fermilab Computing Div/Core Support Services Dept./Scientific Computing Section
Assistant Group Leader, Farms and Clustered Systems Group
Lead of Computing Farms Team

[lm-sensors] Question on my mainboard

[Dieter Jurzitza]

Dear listmembers,
I tried to load the 83781d only according to your suggestions, and, yes, 
everything is working like before. The driver detects the sensor chips and 
reads all values from both chips. And, yes, the 83627 is read out even though 
the corresponding module is not loaded. As readily mentioned, the deadlock 
takes a while, so I'll wait and report what is happening.

This, however, leads me to the question why there is a different module for 
the w83781d and the w83627hf at all - if the first can handle both?
Thanks again
take care



Dieter

-- 
-----------------------------------------------------------

                               |
                                \
                 /\_/\           |
                | ~x~ |/-----\   /
                 \   /-       \_/
  ^^__   _        /  _  ____   /
 <??__ \- \_/     |  |/    |  |
  ||  ||         _| _|    _| _|

if you really want to see the pictures above - use some font
with constant spacing like courier! :-)
-----------------------------------------------------------

[lm-sensors] Question on my mainboard

[Dieter Jurzitza]

Dear Rudolf, dear listmembers,
unfortunately it was not the problem that both drivers (w83781d and w83627hf) 
are running (apparently). Today the same thing happened again even though the 
83627hf is not loaded any more.

The "D"-state:
fred/Downloads> ps ax | grep sensors
 2980 ?        D      0:00 /usr/bin/sensors -c /etc/sensors.tempconf

and cat hangs when trying a
fred/Downloads> cat /sys/bus/i2c/devices/0-002c/alarms 
but only for a while, then comes back with -257
(strange value, isn't it?)

fred/Downloads> cat /sys/bus/i2c/devices/0-002c/temp1_input
hangs for a while too, then comes back with -1000 

The same happens for the devices named 0-002d/XXX. 

By the way, this directory looks like:
lrwxrwxrwx  1 root root 0 2006-04-02 11:20 0-002c 
-> ../../../devices/pci0000:00/0000:00:07.3/i2c-0/0-002c
lrwxrwxrwx  1 root root 0 2006-04-02 11:20 0-002d 
-> ../../../devices/pci0000:00/0000:00:07.3/i2c-0/0-002d
lrwxrwxrwx  1 root root 0 2006-04-02 11:20 0-0048 
-> ../../../devices/pci0000:00/0000:00:07.3/i2c-0/0-0048
lrwxrwxrwx  1 root root 0 2006-04-02 11:20 0-0049 
-> ../../../devices/pci0000:00/0000:00:07.3/i2c-0/0-0049
lrwxrwxrwx  1 root root 0 2006-04-02 11:20 0-004a 
-> ../../../devices/pci0000:00/0000:00:07.3/i2c-0/0-004a
lrwxrwxrwx  1 root root 0 2006-04-02 11:20 0-004b 
-> ../../../devices/pci0000:00/0000:00:07.3/i2c-0/0-004b
lrwxrwxrwx  1 root root 0 2006-04-02 11:20 0-0050 
-> ../../../devices/pci0000:00/0000:00:07.3/i2c-0/0-0050
lrwxrwxrwx  1 root root 0 2006-04-02 11:20 0-0051 
-> ../../../devices/pci0000:00/0000:00:07.3/i2c-0/0-0051
lrwxrwxrwx  1 root root 0 2006-04-02 11:20 0-0052 
-> ../../../devices/pci0000:00/0000:00:07.3/i2c-0/0-0052
lrwxrwxrwx  1 root root 0 2006-04-02 11:20 0-0054 
-> ../../../devices/pci0000:00/0000:00:07.3/i2c-0/0-0054
lrwxrwxrwx  1 root root 0 2006-04-02 11:20 1-0050 
-> ../../../devices/pci0000:00/0000:00:0c.0/i2c-1/1-0050
lrwxrwxrwx  1 root root 0 2006-04-02 11:20 1-0061 
-> ../../../devices/pci0000:00/0000:00:0c.0/i2c-1/1-0061

By the way: what is preempt? How would it show?

There is one thing I found in addition: /usr/bin/sensors does not really hang 
"forever". It only hangs for a certain time, than comes back - but the values 
returned are meaningless. Just like cat <one of the values in /sys/XXX> hangs 
for about 80 seconds, I think sensors hangs for N*80seconds, where N is the 
number of values it reads.
Thank you for any suggestion,
take care



Dieter


Am Donnerstag, 30. M?rz 2006 23:45 schrieben Sie:
*******
>
> Hmm this seems some race condition in the kernel maybe? If you do ps ax
> can you see the D state of the process? Any OOPs message in the log?
>
> Do you use preempt? (perhaps not)
> Does cat hang if you do for example cat some value from
> /sys/bus/i2c/devices.../temp1_input
>
> Regards
> Rudolf
********

-- 
-----------------------------------------------------------

                               |
                                \
                 /\_/\           |
                | ~x~ |/-----\   /
                 \   /-       \_/
  ^^__   _        /  _  ____   /
 <??__ \- \_/     |  |/    |  |
  ||  ||         _| _|    _| _|

if you really want to see the pictures above - use some font
with constant spacing like courier! :-)
-----------------------------------------------------------

[lm-sensors] Question on my mainboard

[Rudolf Marek]

Dieter Jurzitza wrote:
> Dear Rudolf, dear listmembers,
> unfortunately it was not the problem that both drivers (w83781d and w83627hf) 
> are running (apparently). Today the same thing happened again even though the 
> 83627hf is not loaded any more.
> 
> The "D"-state:
> fred/Downloads> ps ax | grep sensors
>  2980 ?        D      0:00 /usr/bin/sensors -c /etc/sensors.tempconf
> 
> and cat hangs when trying a
> fred/Downloads> cat /sys/bus/i2c/devices/0-002c/alarms 
> but only for a while, then comes back with -257
> (strange value, isn't it?)
> 

This sounds like bus problem to me, If you use only the w83627hf, I think the
issue should disappear.

If you wish to help solve the issue continue using the w93781d and
tick in the config of i2c the debug stuff. (Bus driver debug) So we will see if
there are some problems with transaction.

Regards
Rudolf

[lm-sensors] Question on my mainboard

[Dieter Jurzitza]

Dear Rudolf,
would you kindly tell me where activate this debugging stuff? I looked a bit 
through the sensors-sources but I haven't got a clue.
Thank you very much,
take care


Dieter
 
Am Mittwoch, 5. April 2006 10:28 schrieb Rudolf Marek:
*******
> If you wish to help solve the issue continue using the w93781d and
> tick in the config of i2c the debug stuff. (Bus driver debug) So we will
*******
-- 
-----------------------------------------------------------

                               |
                                \
                 /\_/\           |
                | ~x~ |/-----\   /
                 \   /-       \_/
  ^^__   _        /  _  ____   /
 <??__ \- \_/     |  |/    |  |
  ||  ||         _| _|    _| _|

if you really want to see the pictures above - use some font
with constant spacing like courier! :-)
-----------------------------------------------------------

[lm-sensors] Question on my mainboard

[Rudolf Marek]

Dieter Jurzitza wrote:
> Dear Rudolf,
> would you kindly tell me where activate this debugging stuff? I looked a bit 
> through the sensors-sources but I haven't got a clue.


If you have the 2.6 kernel and I think you have so it should be
in kernel menu in i2c section.

Regards
Rudolf

[lm-sensors] Question on my mainboard

[Dieter Jurzitza]

Dear Rudolf,
dear listmembers,
after building the kernel with debugging active, I flood my /var/log/messages 
file (something like 100MByte per day).
What I get is attached. What is apparent to me is the fact that there are 
error messages every other time (I just attached several seconds) and non 
regular accesses (should happen every ten seconds) but no hang like before. 
So I have to wait a little more. If I am unlucky the delays introduced by the 
tons of printfs prevent sensors from crashing my sensor chip. But maybe some 
of you guys can understand why the errors are being reported.
Thank you in advance,
take care


Dieter Jurzitza


Tyan Tiger 2460, SuSE 9.3, kernel 2.6.11-SMP (have seen similar issues with 
older kernels, too).

-- 
-----------------------------------------------------------

                               |
                                \
                 /\_/\           |
                | ~x~ |/-----\   /
                 \   /-       \_/
  ^^__   _        /  _  ____   /
 <??__ \- \_/     |  |/    |  |
  ||  ||         _| _|    _| _|

if you really want to see the pictures above - use some font
with constant spacing like courier! :-)
-----------------------------------------------------------
-------------- next part --------------
A non-text attachment was scrubbed...
Name: logpart.bz2
Type: application/x-bzip2
Size: 1336 bytes
Desc: not available
Url : http://lists.lm-sensors.org/pipermail/lm-sensors/attachments/20060418/c2de2d8c/logpart.bz2

[lm-sensors] Question on my mainboard

[Rudolf Marek]

Dieter Jurzitza wrote:
> Dear Rudolf,
> dear listmembers,
> after building the kernel with debugging active, I flood my /var/log/messages 
> file (something like 100MByte per day).
> What I get is attached. What is apparent to me is the fact that there are 
> error messages every other time (I just attached several seconds) and non 
> regular accesses (should happen every ten seconds) but no hang like before. 
> So I have to wait a little more. If I am unlucky the delays introduced by the 
> tons of printfs prevent sensors from crashing my sensor chip. But maybe some 
> of you guys can understand why the errors are being reported.
> Thank you in advance,
> take care

Yep we can.

There are are errors reading addresses 0x4A and 0x4B
And 0x2C

options w83781d force_w83782d=0,0x2d force_subclients=0,0x2d,0x48,0x49
force_w83627hf=0,0x2c force_subclients=0,0x2c,0x4a,0x4b init=0

So you left the second line there??? and only loading w83627hf with
force_subclients=0,0x2c,0x4a,0x4b ?

I'm asking because there are more addresses access
(cat logpart | awk -FADD '{print $2}' |awk '{print $1}' | sort -u)

Regards
Rudolf

[lm-sensors] Question on my mainboard

[Dieter Jurzitza]

Dear Rudolf,
dear listmembers,
this is what I get when performing the awk - script Rudolf kindly provided 
parsing /var/log/messages in total (~300 MByte):
\058,
\059,
\05a,
\05b,
\091,
\093,
\095,
\097,
\0a0,
\0a1,
\0a2,
\0a3,
\0a4,
\0a5,
\0a8,
\0a9,
> options w83781d force_w83782d=0,0x2d force_subclients=0,0x2d,0x48,0x49
> force_w83627hf=0,0x2c force_subclients=0,0x2c,0x4a,0x4b init=0
>
> So you left the second line there??? and only loading w83627hf with
> force_subclients=0,0x2c,0x4a,0x4b ?

Yes, I left the second line there. I had had taken out the loading of the 
w83627hf driver but sensors crashed anyway, so I thought it might be a good 
start from where I had had come from - what may be wrong. 

> I'm asking because there are more addresses access

This:
> (cat logpart | awk -FADD '{print $2}' |awk '{print $1}' | sort -u)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
gives me the list as shown above. 


I left that there because the folks from Tyan told me to use exactly this line 
- and the last replies from the lm_sensors list referred to that I should 
unload the w83627hf driver and did not refer to the "force_subclients=XXX" 
lines. After removing the w83627hf driver there was no change, so I had had 
put it back in.

Thanks for your efforts - I am not the I?C - man.
Take care


Dieter


-- 
-----------------------------------------------------------

                               |
                                \
                 /\_/\           |
                | ~x~ |/-----\   /
                 \   /-       \_/
  ^^__   _        /  _  ____   /
 <??__ \- \_/     |  |/    |  |
  ||  ||         _| _|    _| _|

if you really want to see the pictures above - use some font
with constant spacing like courier! :-)
-----------------------------------------------------------