Re: [Freeipmi-users] Disabled temp sensors

[Al Chu]

Hey Eric,

Ahhhh.  That would do it.  The slave address is probably wrong in the
SDR.  When you run w/ bridging, the sensors attempts to bridge to an
address that is probably non-functional/illegal.

LMK what your final patch looks like.  I can work it into a workaround
of some sort for ipmi-sensors.  (e.g.
--workaround-flags=assumebmcslaveaddr).

How does ipmi-sel look like?  I'm wondering if SEL events are reporting
right/wrong slave addresses and sensor related outputs are outputting
correctly or not.

> My fans still look messed up.

It certainly depends on if the SDR is correct or not.  From the output
below, it looks as though the Fans are "transition" fans.  They only
report the transition state instead of fan instead of an RPM.  If they
aren't "transition" fans, then the SDR might be wrong which is leading
to this kind of output.

BTW, you forgot the debug output from your previous e-mail.

Al

On Mon, 2010-05-03 at 21:32 -0700, Eric Pooch wrote:
> OK,  I think I found the problem on my computer's implementation of IPMI
> I edited:
> /freeipmi-0.8.5/libfreeipmi/src/sensor-read/ipmi-sensor-read.c
> 
>    if (slave_address == IPMI_SLAVE_ADDRESS_BMC)*/
>    if (slave_address != IPMI_SLAVE_ADDRESS_BMC)
> And received what looks like good data:
> 
> 1  | Fan 1            | Fan                      | N/A        | N/A    
> | 'transition to Off Line'
> 2  | Fan 2            | Fan                      | N/A        | N/A    
> | 'transition to Running' 'transition to On Line'
> 3  | Fan 3            | Fan                      | N/A        | N/A    
> | 'transition to Running' 'transition to On Line'
> 4  | Fan 4            | Fan                      | N/A        | N/A    
> | 'transition to Running' 'transition to On Line'
> 5  | PCI Fan          | Fan                      | N/A        | N/A    
> | 'transition to Off Line'
> 6  | Memory           | Memory                   | N/A        | N/A    
> | 'OK'
> 7  | CPU 1            | Processor                | N/A        | N/A    
> | 'Processor Presence detected'
> 8  | CPU 2            | Processor                | N/A        | N/A    
> | 'Processor Presence detected'
> 9  | VRM              | Voltage                  | N/A        | N/A    
> | 'OK'
> 10 | CPU1 Temperature | Temperature              | 35.00      | C      
> | 'OK'
> 11 | CPU2 Temperature | Temperature              | 33.00      | C      
> | 'OK'
> 12 | Thermal Trip     | Temperature              | N/A        | N/A    
> | 'OK'
> 13 | Sys Temperature  | Temperature              | 31.00      | C      
> | 'OK'
> 14 | DDR 1.25V        | Voltage                  | 1.25       | V      
> | 'OK'
> 15 | Sys 3.3V         | Voltage                  | 3.25       | V      
> | 'OK'
> 16 | Sys 5V           | Voltage                  | 5.00       | V      
> | 'OK'
> 17 | CIOBE 1.2V       | Voltage                  | 1.21       | V      
> | 'OK'
> 18 | CIOBE 2.5V       | Voltage                  | 2.52       | V      
> | 'OK'
> 19 | BIOS Progress    | System Firmware Progress | N/A        | N/A    
> | N/A
> 20 | Watchdog         | Watchdog 2               | N/A        | N/A    
> | N/A
> 
> This is much better, and I get info for almost all of the sensors  
> that just showed N/A before. My fans still look messed up.  I will  
> figure out more details, make it a bit cleaner and send a patch for  
> users of this flawed IPMI implementation
> Thanks
> --Eric
> 
> On May 3, 2010, at 8:42 PM, Eric Pooch wrote:
> 
> > Ok, I updated to 0.8.5 and attached an archive of the debug log from:
> > $ sudo ipmi-sensors --debug
> > see below:
> > On May 3, 2010, at 5:23 PM, Al Chu wrote:
> >
> >> Hey Eric,
> >>
> >>> Also, bmc-info returns IPMI version 1.0 that is probably not
> >>> supported by FreeIPMI, but ipmi-locate, returns "IPMI Version: 1.5"
> >>> for all of the devices.
> >>
> >> Doing a quick online search, this machine appears to be pretty  
> >> old.  It
> >> is possible that it does not support IPMI 1.5.  The output from
> >> ipmi-locate you're seeing may be the defaults and not actual outputs
> >> from the machine (this is confusing many people so I'm changing this
> >> output for the next 0.9.1 release).  If it is only IPMI 1.0, there's
> >> probably not much I can do to help you, since many of the IPMI  
> >> commands
> >> will just not be supported on your motherboard.
> >>
> >
> > Ok.  I understand
> >>> First, all my sensor values come back as [NA] even though most work
> >>> properly under ipmitool.
> >>
> >> I assume you're using FreeIPMI 0.7.X b/c the newest one (0.8.X line)
> >> does not have "[NA]" output.  There have certainly been fixes since
> >> then, so you may wish to upgrade.  My initial guess was bridging, but
> >> you seem to have tried that.
> >>
> >> I've noticed on some motherboards that there are issues b/c I find
> >> errors/problems in other parts of IPMI that ipmitool doesn't, thus I
> >> output errors and they don't.  We need to dig into the core of the
> >> errors on your board to figure out what they/I are doing
> >> wrong/differently.  Can you provide --debug output?
> >>
> >>> So, I think maybe there something that is  disabling the Temp sensor
> >>> at another level.  I noticed on the HP lightsout user guide that  
> >>> they
> >>> have a setting "o PEF Control—Enables or disables the sensor. "
> >>
> >> Based on some of the error messages you posted from ipmitool (BTW, in
> >> the future could you indicate what tools the error messages came  
> >> from, I
> >> thought you were indicating FreeIPMI errors and couldn't find them at
> >> first),
> > Sorry, I thought I was listing FreeIPMI errors, but I guess I  
> > posted errors from the wrong log.
> >> my guess is bridging is not supported on your motherboard and/or
> >> there is a firmware issue w/ bridging, so the temp sensors can't be
> >> reached.
> > I would agree, except that the standard IMPI raw "get sensor  
> > reading" command works fine.  It is almost like ipmi-sensors and  
> > ipmitool are finding something they don't like in the sdr and not  
> > trying to read the sensor at all.
> >
> > $ sudo ipmi-raw 0 04 2d 0A
> > rcvd: 2D 00 23 C0 00 00
> >
> > 0x23 =  35 degrees celsius, which seems right for my processor  
> > temp.  As I mentioned before, it varies proportionally with server  
> > load, seems like the value I need, and is the correct command as  
> > far as I can tell from the IPMI v 1.5 specs
> >
> >> It's hard to say.  If you can provide me --debug output from
> >> ipmi-sensors, I can maybe analyze it deeper.
> >
> >
> > $ sudo ipmi-sensors --debug
> > see attachment
> >
> > $ sudo ipmi-sensors --bridge
> > ipmi_sensor_read: internal IPMI error
> >
> >> Does any HP specific software work for you for all these sensors?  If
> >> their software does, and ipmitool/FreeIPMI does not, it indicates  
> >> there
> >> is something kooky on your motherboard.
> >>
> > I don't know, I don't have access to Windows.  If it won't work  
> > with FreeIPMI, I understand that my motherboard is old, but it just  
> > seems strange that I can get the sensor reading using ipmi-raw, but  
> > not ipmi-sensors.
> >
> > Thanks a lot for your help.
> > --Eric
> >
> >> Al
> >>
> >> On Sun, 2010-05-02 at 10:10 -0700, Eric Pooch wrote:
> >>> I am having several problems on my HP proliant dl140 G1
> >>>
> >>> First, all my sensor values come back as [NA] even though most work
> >>> properly under ipmitool.
> >>> I get the debug errors from ipmi-sensors:
> >>>
> >>> Error reading event status for sensor #09: Invalid command
> >>> ...
> >>> Error reading event enable for sensor #09: Invalid command
> >>>
> >>> When I try ipmi-raw to send those commands, I also get the same
> >>> error, so I think the commands are not supported on the sensors.   
> >>> The
> >>> sensors are returning the proper information when I send a raw
> >>> command to get their readings. (see below)
> >>>
> >>> However, none of my temp sensors work properly in either freeipmi or
> >>> ipmitool and I get a debug error:
> >>> Error reading sensor CPU1 Temperature (#0a): Destination unavailable
> >>>
> >>> I get the same "destination unavailable message from event status  
> >>> and
> >>> event enable. However, when I enter the raw ipmi command to read the
> >>> temp sensor:
> >>> sudo ipmi-raw 0 04 2d 0A
> >>>
> >>> it responds correctly:
> >>> rcvd: 2D 00 1B C0 00 00
> >>>
> >>> The 1B is the correct temperature in Celsius that rises with
> >>> processor load.  It is definitely the correct temperature.
> >>> I have tried the bridge mode but I get an error also.
> >>> It seems like the sensor is responding correctly, but is disabled as
> >>> far as the sdr is concerned?  I can't enable it through a raw  
> >>> command
> >>> because none of the sensors respond to the "event status" or "event
> >>> enable" commands.  So, I think maybe there something that is
> >>> disabling the Temp sensor at another level.  I noticed on the HP
> >>> lightsout user guide that they have a setting "o PEF Control—Enables
> >>> or disables the sensor. "
> >>> I am not really sure how to make a change that would cause the  
> >>> sensor
> >>> to be enabled.
> >>>
> >>> Also, bmc-info returns IPMI version 1.0 that is probably not
> >>> supported by FreeIPMI, but ipmi-locate, returns "IPMI Version: 1.5"
> >>> for all of the devices.
> >>>
> >>> Thanks for any help!_______________________________________________
> >>> Freeipmi-users mailing list
> >>> address@hidden
> >>> http://**lists.gnu.org/mailman/listinfo/freeipmi-users
> >>>
> >> -- 
> >> Albert Chu
> >> address@hidden
> >> Computer Scientist
> >> High Performance Systems Division
> >> Lawrence Livermore National Laboratory
> >>
> >
> > _______________________________________________
> > Freeipmi-users mailing list
> > address@hidden
> > http://*lists.gnu.org/mailman/listinfo/freeipmi-users
> 
> 
> 
> _______________________________________________
> Freeipmi-users mailing list
> address@hidden
> http://*lists.gnu.org/mailman/listinfo/freeipmi-users
> 
-- 
Albert Chu
address@hidden
Computer Scientist
High Performance Systems Division
Lawrence Livermore National Laboratory