PRD: don't query per-symbol counters in broadcast mode

Change-Id: I42c8348df5f4b5ae846a2fb7d821d5eb07c3949e
Reviewed-on: http://ralgit01.raleigh.ibm.com/gerrit1/44760
Tested-by: Jenkins Server <pfd-jenkins+hostboot@us.ibm.com>
Reviewed-by: Caleb N. Palmer <cnpalmer@us.ibm.com>
Reviewed-by: Benjamin J. Weisenbeck <bweisenb@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/gerrit1/45000
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>

3 files changed, 22 insertions, 19 deletions

diff --git a/src/usr/diag/prdf/plat/mem/prdfMemScrubUtils.C b/src/usr/diag/prdf/plat/mem/prdfMemScrubUtils.C
index ebe116431..5e1efccc8 100644
--- a/src/usr/diag/prdf/plat/mem/prdfMemScrubUtils.C
+++ b/src/usr/diag/prdf/plat/mem/prdfMemScrubUtils.C
@@ -289,6 +289,8 @@ uint32_t checkEccFirs<TYPE_MCA>( ExtensibleChip * i_chip,
         if ( 0 != mcaeccfir->GetBitFieldJustified(20,8) )
             o_eccAttns |= MAINT_MPE;
 
+        if ( mcaeccfir->IsBitSet(28) ) o_eccAttns |= MAINT_NCE;
+        if ( mcaeccfir->IsBitSet(29) ) o_eccAttns |= MAINT_TCE;
         if ( mcaeccfir->IsBitSet(30) ) o_eccAttns |= MAINT_SCE;
         if ( mcaeccfir->IsBitSet(31) ) o_eccAttns |= MAINT_MCE;
         if ( mcaeccfir->IsBitSet(34) ) o_eccAttns |= MAINT_UE;
@@ -354,6 +356,7 @@ uint32_t checkEccFirs<TYPE_MBA>( ExtensibleChip * i_chip,
         if ( 0 != mbseccfir->GetBitFieldJustified(20,8) )
             o_eccAttns |= MAINT_MPE;
 
+        if ( mbseccfir->IsBitSet(36) ) o_eccAttns |= MAINT_NCE;
         if ( mbseccfir->IsBitSet(37) ) o_eccAttns |= MAINT_SCE;
         if ( mbseccfir->IsBitSet(38) ) o_eccAttns |= MAINT_MCE;
         if ( mbseccfir->IsBitSet(41) ) o_eccAttns |= MAINT_UE;
diff --git a/src/usr/diag/prdf/plat/mem/prdfMemScrubUtils.H b/src/usr/diag/prdf/plat/mem/prdfMemScrubUtils.H
index 881a5b1fd..c81c6b4a7 100644
--- a/src/usr/diag/prdf/plat/mem/prdfMemScrubUtils.H
+++ b/src/usr/diag/prdf/plat/mem/prdfMemScrubUtils.H
@@ -126,6 +126,10 @@ enum MaintEccAttns
     // Nimbus only
     MAINT_IUE           = 0x0080, ///< Intermittent UE
     MAINT_IMPE          = 0x0040, ///< Intermittent chip mark
+
+    // Maintenance CEs
+    MAINT_NCE           = 0x0020, ///< CE
+    MAINT_TCE           = 0x0010, ///< CE on two symbols (Nimbus only)
 };
 
 /**
diff --git a/src/usr/diag/prdf/plat/mem/prdfMemTdCtlr_ipl.C b/src/usr/diag/prdf/plat/mem/prdfMemTdCtlr_ipl.C
index 64719bf5f..e6d769b6c 100644
--- a/src/usr/diag/prdf/plat/mem/prdfMemTdCtlr_ipl.C
+++ b/src/usr/diag/prdf/plat/mem/prdfMemTdCtlr_ipl.C
@@ -190,28 +190,24 @@ uint32_t __checkEcc( ExtensibleChip * i_chip, TdQueue & io_queue,
             }
         }
         else if ( isMfgCeCheckingEnabled() &&
-                  (0 != (eccAttns & MAINT_HARD_NCE_ETE)) )
+                  (0 != (eccAttns & MAINT_HARD_NCE_ETE)) &&
+                  ( (0 != (eccAttns & MAINT_NCE)) ||
+                    (0 != (eccAttns & MAINT_TCE))    ) )
         {
-            io_sc.service_data->AddSignatureList( trgt, PRDFSIG_MaintHARD_CTE );
+            // NOTE: The MAINT_HARD_NCE_ETE attention is reported on the MCBIST.
+            //       If the command is run in broadcast mode, we may end up
+            //       doing TPS on all four ports when only one port has the CE.
+            //       Therefore, we must check for MAINT_NCE or MAINT_TCE, which
+            //       are found on the MCA, to determine if this MCA need a TPS
+            //       procedure.
 
-            // Query the per-symbol counters for the hard CE symbol.
-            MemUtils::MaintSymbols symData; MemSymbol junk;
-            o_rc = MemUtils::collectCeStats<T>( i_chip, rank, symData, junk );
-            if ( SUCCESS != o_rc )
-            {
-                PRDF_ERR( PRDF_FUNC "MemUtils::collectCeStats(0x%08x,m%ds%d) "
-                          "failed", i_chip->GetId(), rank.getMaster(),
-                          rank.getSlave() );
-                break;
-            }
+            io_sc.service_data->AddSignatureList( trgt, PRDFSIG_MaintHARD_CTE );
 
-            // The command will have finished at the end of the rank so there
-            // may be more than one symbol. Add all to the callout list.
-            for ( auto & s : symData )
-            {
-                MemoryMru memmru ( trgt, rank, s.symbol );
-                io_sc.service_data->SetCallout( memmru );
-            }
+            // NOTE: Normally we would add each symbol in the per-symbol
+            //       counters to the callout list for FFDC, but the results are
+            //       a bit undefined in broadcast mode. Since this should not be
+            //       a predictive error log, it is fine to have no callouts for
+            //       now.
 
             // Add a TPS procedure to the queue.
             TdEntry * e = new TpsEvent<T>{ i_chip, rank };