PRD: UE callout rank if no dimm callout

Change-Id: Ic8254be5e5d33b18f433f2e308aa23cd88687c7d
CQ: SW427493
Reviewed-on: http://ralgit01.raleigh.ibm.com/gerrit1/58709
Tested-by: Jenkins Server <pfd-jenkins+hostboot@us.ibm.com>
Reviewed-by: Benjamin J. Weisenbeck <bweisenb@us.ibm.com>
Reviewed-by: Matt Derksen <mderkse1@us.ibm.com>
Reviewed-by: Brian J. Stegmiller <bjs@us.ibm.com>
Reviewed-by: Zane C. Shelley <zshelle@us.ibm.com>
Reviewed-on: http://ralgit01.raleigh.ibm.com/gerrit1/58724
Tested-by: Jenkins OP Build CI <op-jenkins+hostboot@us.ibm.com>
Tested-by: Jenkins OP HW <op-hw-jenkins+hostboot@us.ibm.com>
Tested-by: FSP CI Jenkins <fsp-CI-jenkins+hostboot@us.ibm.com>

1 files changed, 24 insertions, 1 deletions

diff --git a/src/usr/diag/prdf/common/plat/mem/prdfMemEccAnalysis.C b/src/usr/diag/prdf/common/plat/mem/prdfMemEccAnalysis.C
index 4c5153e81..9cb7f5b42 100644
--- a/src/usr/diag/prdf/common/plat/mem/prdfMemEccAnalysis.C
+++ b/src/usr/diag/prdf/common/plat/mem/prdfMemEccAnalysis.C
@@ -178,6 +178,7 @@ uint32_t handleMemUe<TYPE_MBA>( ExtensibleChip * i_chip, const MemAddr & i_addr,
             l_dqBitmap.getCaptureData( io_sc.service_data->GetCaptureData() );
 
             // Add all DIMMs with bad bits to the callout list.
+            TargetHandleList callouts;
             for ( uint8_t ps = 0; ps < DIMMS_PER_RANK::MBA; ps++ )
             {
                 bool badDqs = false;
@@ -194,7 +195,7 @@ uint32_t handleMemUe<TYPE_MBA>( ExtensibleChip * i_chip, const MemAddr & i_addr,
                                                           rank, ps );
                 if ( l_dimm == nullptr ) continue;
 
-                io_sc.service_data->SetCallout( l_dimm, MRU_HIGH );
+                callouts.push_back( l_dimm );
 
                 if ( isMfgCeCheckingEnabled() )
                 {
@@ -204,6 +205,28 @@ uint32_t handleMemUe<TYPE_MBA>( ExtensibleChip * i_chip, const MemAddr & i_addr,
                 }
             }
 
+            if ( 0 == callouts.size() )
+            {
+                // It is possible the scrub counters have rolled over to zero
+                // due to a known DD1.0 hardware bug. In this case, the best
+                // we can do is callout both DIMMs, because at minimum we know
+                // there was a UE, we just don't know where.
+                // NOTE: If this condition happens because of a DD2.0+ bug, the
+                //       mssIplUeIsolation procedure will callout the Centaur.
+                callouts = getConnectedDimms( i_chip->getTrgt(), rank );
+                if ( 0 == callouts.size() )
+                {
+                    PRDF_ERR( PRDF_FUNC "getConnectedDimms() failed" );
+                    o_rc = FAIL; break;
+                }
+            }
+
+            // Callout all DIMMs in the list.
+            for ( auto & dimm : callouts )
+            {
+                io_sc.service_data->SetCallout( dimm, MRU_HIGH );
+            }
+
             // Make the error log predictive.
             io_sc.service_data->setServiceCall();