[Freeipmi-devel] Re: ipmi-dcmi testing results

[Al Chu]

Hey Holger,

On Mon, 2009-12-14 at 17:01 +0100, Liebig, Holger wrote:
> Hi everyone,
> 
> here are some very first results and patch proposals from testing the
> ipmi-dcmi commands from freeipmi 0.8.1 (thanks for adding dcmi support
> into freeipmi).
> 
> --get-asset-tag:
> Assumes asset tag to be multiple of 16bytes (max read size specified in DCMI 
> spec).

Yup, that seems to be a bug.

> Also assumes, that there is an additional length/encoding information
> present (1st byte) as in the raw FRU information - which is not
> specified in the DCMI spec. Please note, that the direct reference to
> the FRU Product Data Spec. has been removed in Errata/Clarification
> E18 of the DCMI Spec. 1.0 Revision 1.0.
> Basically, the asset tag can be stored somewhere else in the BMC and
> only needs to be retrievable via IPMI command (it's interesting that
> no set command is not specified).

Ahhh.  Ok.  The wording of the spec suggested that it returned the
asset-tag field of the FRU, but instead it's just english.  Now that you
mention it, it seems so obvious and I'm not sure why I read it that way
to begin with.

> --get-power-limit:
> The exception action is not a bit field and OEM actions are only defined for 
> values 0x02 ... 0x10.

Thanks.  I actually had a note in my printed copy of the spec with a
question mark next to it b/c I couldn't tell if it was values or
bitmasks.  I obviously guessed wrong :-)

> --get-system-power-statistics:
> Stops after asking for 'power_measurement' instead of 
> 'power_reading_state.power_measurement'.
>
> --get-enhanced-system-power-statistics:
> The template is slightly incorrect (it asks for a fixed/required number of 
> 2048 required rolling average time periods in the response).
>
> --get-sensor-info:
> The number of parsed instances is not incremented, therefore the code loops 
> forever.
>
> --activate-deactivate-power-limit: 
> currently core dumps in argp_parse().

Yup, looks like these are all bugs.  Thanks for the catches.

Thanks for the patch.  I'll roll it in, w/ some tweaks (no need to keep
my old code commented out since they're just wrong).

Al

> Thanks
> Holger
> 
> --
> Holger L i e b i g                       
> address@hidden        Fujitsu Technology Solutions GmbH
> Senior Developer                                     Otto Hahn Ring 6
> TSP x86 E SW 4                                           81739 Munich
> Telephone: +49 89 636 43820                                   GERMANY
> Telefax  : +49 89 636 41215
> Internet                                        http://*ts.fujitsu.com
> Company Details                 http://*de.ts.fujitsu.com/imprint.html 
> 
-- 
Albert Chu
address@hidden
Computer Scientist
High Performance Systems Division
Lawrence Livermore National Laboratory