Re: [lm-sensors] CPU fan with full speed

[Guenter Roeck <linux@roeck-us.net>]

On Mon, Jul 02, 2012 at 03:28:19AM +0200, Philipp Kraus wrote:
> On 2012-07-02 03:14:03 +0200, Philipp Kraus said:
> 
> >On 2012-06-30 21:24:59 +0200, Guenter Roeck said:
> >
> >>On Sat, Jun 30, 2012 at 07:27:49PM +0200, Philipp Kraus wrote:
> >>>On 2012-06-30 19:03:44 +0200, Guenter Roeck said:
> >>>
> >>>>On Sat, Jun 30, 2012 at 06:04:53PM +0200, Philipp Kraus wrote:
> >>>>>On 2012-06-30 17:33:38 +0200, Guenter Roeck said:
> >>>>>
> >>>>>>On Sat, Jun 30, 2012 at 04:37:28PM +0200, Philipp Kraus wrote:
> >>>>>>>
> >>>>>>>the grep command shows
> >>>>>>>grep: /sys/class/hwmon/hwmon1/device/driver: Is a directory
> >>>>>>>grep: /sys/class/hwmon/hwmon1/device/hwmon: Is a directory
> >>>>>>>grep: /sys/class/hwmon/hwmon1/device/power: Is a directory
> >>>>>>>grep: /sys/class/hwmon/hwmon1/device/subsystem: Is a directory
> >>>>>>>
> >>>>>>>and within the directory are the in*, pwm* temp* files
> >>>>>>>
> >>>>>>Not sure I understand.
> >>>>>>
> >>>>>>I'll need the names and output for the pwm* and temp* files. They
> >>>>>>_should_ be in
> >>>>>>/sys/class/hwmon/hwmon1/device/pwm* and /sys/class/hwmon/hwmon1/device/temp*.
> >>>>>
> >>>>>the command grep /sys.... does not work, I have run an ls -l and
> >>>>
> >>>>Phil,
> >>>>
> >>>>It wasn't "grep /sys/...", it was "grep . /sys/class/hwmon/hwmon1/device/*".
> >>>>The "." is important. Your output gives me the file names, but I need names and
> >>>>content, not names and permissions.
> >>>
> >>>Sorry, my mistake, I haven't seen that it ist "grep . <space> /sys..."
> >>>Hope this the correct output now
> >>>
> >>>/sys/class/hwmon/hwmon1/device/pwm1:252
> >>
> >>Looks like all pwm values report the "default" pwm value.
> >>
> >>>/sys/class/hwmon/hwmon1/device/temp1_auto_channels_pwm:1
> >>
> >>If I understand the code and the data sheet correectly, this attribute
> >>is a bit map used to map a temperature source to one or multiple fans.
> >>And it looks like all temperature sources are mapped to fan1, which
> >>in turn means that effectively temp4 controls the speed of fan1 (the highest
> >>temperature determines the fan speed), and all other fans are not mapped to
> >>a temperature input.
> >>
> >>What you should probably do is to find the association between tempX and pwmX
> >>and program tempX_auto_channels_pwm such that the two map.
> >>For example, if temp1
> >>is the CPU1 temperature, and pwm4 (fan4) is connected to the CPU1 fan, write
> >>0b00001000 = 0x08 into /sys/class/hwmon/hwmon1/device/temp1_auto_channels_pwm.
> >>Similar, if temp2 is the CPU2 temperature and fan5/pwm5 controls its fan,
> >>write 0x10 (\x16) into /sys/class/hwmon/hwmon1/device/temp2_auto_channels_pwm.
> >
> >You are write, temp1 and temp2 are the values for the CPU 1 and CPU 2.
> >The fans for the CPU
> >are fan7 = CPU 1 and fan8 = CPU 2
> >
> >Can you explane me in detail how you create the map 0b00001000 = 0x08 ?
> >I think 0x08 = 0x8 = 8 is a hexadecimal value but is this 0b00001000
> >also a hexadecimal value, in this case both values are not equal !?
> >Should I write with echo -e "\x8" >
> >/sys/class/hwmon/hwmon1/device/temp1_auto_channels_pwm ?
> >
> >IMHO I must enable (1) the n-th bit on the auto_channels_pwm value for
> >the n-th fan, so in my configuration the 7 and 8 bit (begin = 0)?
> 
> I have set with
> echo 128 >  /sys/class/hwmon/hwmon1/device/temp1_auto_channels_pwm
> echo 256 >  /sys/class/hwmon/hwmon1/device/temp2_auto_channels_pwm
> 
> the 7th and 8th fan-bit but nothing is changed
> 
Does pwm7 and pwm8 still report 252 ? If so, we must be missing something.

What do you get if you read back
/sys/class/hwmon/hwmon1/device/temp1_auto_channels_pwm and
/sys/class/hwmon/hwmon1/device/temp2_auto_channels_pwm ?

Thanks,
Guenter

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