Meaning of negative memory_channel value in rasdaemon message

Meaning of negative memory_channel value in rasdaemon message

[tibor.gyori]

Dear EDAC list,

when rasdaemon reports a memory error, what does it mean for the  
memory_channel value to be negative?

Example:
Feb 11 22:34:40 CRD-MILAN-3 rasdaemon[918]:            <...>-42552  
[001]     0.007845: mce_record:           2022-02-11 22:34:40 +0100  
Unified Memory Controller (bank=18), status= 9c2040000000011b,  
Corrected error, no action required., mci=CECC, mca= DRAM ECC error.
Feb 11 22:34:40 CRD-MILAN-3 rasdaemon[918]:  Memory Error 'mem-tx:  
generic read, tx: generic, level: L3/generic',  
memory_channel=-1,csrow=3, cpu_type= AMD Scalable MCA, cpu= 1,  
socketid= 0, misc= d01a0c7b01000000, addr= 2fa07780, synd=  
e4da80000a800603, ipid= 9600350f00, mcgstatus=0, mcgcap= 118, apicid= 1

Is -1 a magic value for memory_channel? On the AMD Milan platform, I  
would expect that memory channels would be numbered from 0 to 7. I  
googled for a bit, but could not find where this sort of stuff would  
be documented.

Best,
Tibor

----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.

Re: Meaning of negative memory_channel value in rasdaemon message

[Yazen Ghannam]

On Fri, Feb 11, 2022 at 11:40:58PM +0100, tibor.gyori@chem.u-szeged.hu wrote:
> Dear EDAC list,
> 
> when rasdaemon reports a memory error, what does it mean for the
> memory_channel value to be negative?
> 
> Example:
> Feb 11 22:34:40 CRD-MILAN-3 rasdaemon[918]:            <...>-42552 [001]
> 0.007845: mce_record:           2022-02-11 22:34:40 +0100 Unified Memory
> Controller (bank=18), status= 9c2040000000011b, Corrected error, no action
> required., mci=CECC, mca= DRAM ECC error.
> Feb 11 22:34:40 CRD-MILAN-3 rasdaemon[918]:  Memory Error 'mem-tx: generic
> read, tx: generic, level: L3/generic', memory_channel=-1,csrow=3, cpu_type=
> AMD Scalable MCA, cpu= 1, socketid= 0, misc= d01a0c7b01000000, addr=
> 2fa07780, synd= e4da80000a800603, ipid= 9600350f00, mcgstatus=0, mcgcap=
> 118, apicid= 1
> 
> Is -1 a magic value for memory_channel? On the AMD Milan platform, I would
> expect that memory channels would be numbered from 0 to 7. I googled for a
> bit, but could not find where this sort of stuff would be documented.
> 

Hi Tibor,

Which version of rasdaemon are you using?

You may be missing the following commit which is including v0.6.7.

https://github.com/mchehab/rasdaemon/commit/854364ba44aee9bc5646f6537fc744b0b54aff37

Thanks,
Yazen

Re: Meaning of negative memory_channel value in rasdaemon message

[tibor.gyori]

Idézet (Yazen Ghannam <yazen.ghannam@amd.com>):

> On Fri, Feb 11, 2022 at 11:40:58PM +0100, tibor.gyori@chem.u-szeged.hu wrote:
>> Dear EDAC list,
>>
>> when rasdaemon reports a memory error, what does it mean for the
>> memory_channel value to be negative?
>>
>> Example:
>> Feb 11 22:34:40 CRD-MILAN-3 rasdaemon[918]:            <...>-42552 [001]
>> 0.007845: mce_record:           2022-02-11 22:34:40 +0100 Unified Memory
>> Controller (bank=18), status= 9c2040000000011b, Corrected error, no action
>> required., mci=CECC, mca= DRAM ECC error.
>> Feb 11 22:34:40 CRD-MILAN-3 rasdaemon[918]:  Memory Error 'mem-tx: generic
>> read, tx: generic, level: L3/generic', memory_channel=-1,csrow=3, cpu_type=
>> AMD Scalable MCA, cpu= 1, socketid= 0, misc= d01a0c7b01000000, addr=
>> 2fa07780, synd= e4da80000a800603, ipid= 9600350f00, mcgstatus=0, mcgcap=
>> 118, apicid= 1
>>
>> Is -1 a magic value for memory_channel? On the AMD Milan platform, I would
>> expect that memory channels would be numbered from 0 to 7. I googled for a
>> bit, but could not find where this sort of stuff would be documented.
>>
>
> Hi Tibor,
>
> Which version of rasdaemon are you using?
>
> You may be missing the following commit which is including v0.6.7.
>
> https://github.com/mchehab/rasdaemon/commit/854364ba44aee9bc5646f6537fc744b0b54aff37
>
> Thanks,
> Yazen
>
>

Dear Yazen,

Thanks for taking the time to reply. Upon further investigation it  
looks like the rasdaemon package provided by Ubuntu 20.04 is indeed  
missing the commit you specified.
Building rasdaemon from source has resolved this particular issue,  
although I have to say it was not without bumps on the road.
For what its worth, looking at the GitHub repo, I am somewhat  
concerned that rasdaemon might be becoming unmaintained. No commits  
have been made since Aug 2021, and reasonable-looking PRs are not  
getting reviewed.

Best wishes,
Tibor Győri

----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.