The console display of dimm present do not match the hardware.

[MengLiMarkham]

P9 hostboot code will get and display the dimms which are present in the system as below:

        TARGETING::TYPE l_type = target->getAttr<TARGETING::ATTR_TYPE>();
        TARGETING::ATTR_FAPI_POS_type l_pos = 0;
        if( target->tryGetAttr<TARGETING::ATTR_FAPI_POS>(l_pos) )
        {
            l_presData[l_type] |= (0x8000000000000000 >> l_pos);
        }

In P8 we use ATTR_POSITION, that's ok for P8. But in P9 hostboot ATTR_FAPI_POS will over 64, and when shift over 64, overlapping happened, then these dimms can not be displayed in the console log as preset.
https://github.com/open-power/hostboot/blob/master/src/usr/isteps/istep06/host_discover_targets.C#L501

[dcrowell77]

This issue has been reported elsewhere too... The problem is that we don't really have a good position value to use that is consistent across all system types and all the target types that we want to display.

The best answer would be if we actually settled on a universal numbering scheme controlled via the MRW that actually matched the documentation for a particular system.

[ghost]

Dan <[email protected]> writes:

The best answer would be if we actually settled on a universal numbering scheme controlled via the MRW that actually matched the documentation for a particular system.

I'd be in favour of that if it helped us get something that matched what as silkscreened on the board. Mind you, it would require somebody to actually verify that it matches... which our experience with firestone and PCI slots is that somebody notices at least 12-18 months after shipping :/

…

-- Stewart Smith OPAL Architect, IBM.