Re: [Freeipmi-users] Wrong values from a Winbond WPCM450 BMC

[Albert Chu]

On Wed, 2014-10-29 at 23:12 +0100, Paul van der Vlis wrote:
> op 29-10-14 21:27, Albert Chu schreef:
> > On Wed, 2014-10-29 at 19:47 +0100, Paul van der Vlis wrote:
> >> Helle Albert and others,
> >>
> >> op 27-10-14 17:48, Albert Chu schreef:
> >>> On Sat, 2014-10-25 at 20:11 +0200, Paul van der Vlis wrote:
> >>>> Hello Albert,
> >>>>
> >>>> op 25-10-14 19:47, Albert Chu schreef:
> >>>>> Hi Paul,
> >>>>>
> >>>>> I doubt there was an issue with reading the wrong values from the BMC.  
> >>>>> I
> >>>>> have seen many BMCs where the vendor populates the BMC with "poor" 
> >>>>> choices
> >>>>> of default values.  In fact, sometimes I've seen vendors populate the 
> >>>>> BMC
> >>>>> with illegal values (e.g. 0x1-0x4 are the only legal values, but the
> >>>>> default value populated in the BMC is 0x7).
> >>>>
> >>>> Doesn't bmc-config read the actual used values, but something else?
> >>>
> >>> I'm not quite sure what you're asking here.  It reads the actual values
> >>> stored in the BMC during a --checkout.
> >>
> >> The BMC did work with me at home. At home I used the bios to configure
> >> the IP.
> >>
> >> After changing the IP in the datacenter with bmc-config the BMC did not
> >> work anymore. I did only change the network settings (and later ARP
> >> settings).
> >>
> >> Then it's strange I have to edit the Volatile_Access_Mode and
> >> Non_Volatile_Access_Mode. True?
> > 
> > It is strange that the motherboard had these as "false" by default.
> 
> It did work before I changed the settings. So when this was the setting,
> it should work with this setting. But it did not.

If you used vendor proprietary software (including the bios) to
configure, it's possible they enable this temporarily then disable it
afterwards.  Or there could be a mechanism besides this that is special
to this BMC.

Al

> >>>> Is it possible to use a BMC from another vendor?
> >>>
> >>> Unlikely, given the BMC is part of the motherboard.
> >>
> >> I can buy the motherboard I use with and without a BMC:
> >> http://www.supermicro.com/products/motherboard/QPI/5500/X8DTU-F.cfm
> >> http://www.supermicro.com/products/motherboard/QPI/5500/X8DTU.cfm
> >>
> >> CoreIPM says on the website: "coreIPM-LINUX provides a ready to use,
> >> extremely compact drop in solution for platform management. It is
> >> specifically targeted towards shelf and appliance management."
> > 
> > Personally, I've never heard of a 3rd party BMC available.
> 
> Maybe it has to do with OPMA, what's mostly used with AMD:
> http://en.wikipedia.org/wiki/Open_Platform_Management_Architecture
> 
> I saw Intel implementations too, e.g.:
> http://www.tyan.com/Motherboards_S5393_S5393WG2NR
> 
> Not sure this is the only way, the OPMA article also tell's about PCI
> based management cards.
> 
> >>>> Which vendors make nice and good BMC's?
> >>>
> >>> To be very honest, most vendors solutions are fine.  They all make
> >>> mistakes though.  You can see the history of bugs I've found in vendor
> >>> solutions here:
> >>>
> >>> http://www.gnu.org/software/freeipmi/freeipmi-bugs-issues-and-workarounds.txt
> >>
> >> When I see lists like this, I think: "most solutions are far from
> >> perfect, they maybe work with their proprietary software, but give many
> >> problems with standards compliant software."
> > 
> > I should note that most of the bugs are simply that, bugs.  I wouldn't
> > say it's an attempt from the vendor to make IPMI only work with their
> > proprietary software.  They may perform the majority of QA against their
> > proprietary software, which is part of the problem.
> 
> I think that's correct.
> 
> With regards,
> Paul van der Vlis.
> 
> 
> 
-- 
Albert Chu
address@hidden
Computer Scientist
High Performance Systems Division
Lawrence Livermore National Laboratory