PRD: fixed TPS action in MNFG for memory CEs.

Change-Id: Ie72e5c8feafbcf8cb0ffb0316df2c1072e5fe5a8
Reviewed-on: http://ralgit01.raleigh.ibm.com/gerrit1/38219
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/gerrit1/38413
Tested-by: Jenkins OP Build CI <op-jenkins+hostboot@us.ibm.com>
Tested-by: FSP CI Jenkins <fsp-CI-jenkins+hostboot@us.ibm.com>

2 files changed, 22 insertions, 5 deletions

diff --git a/src/usr/diag/prdf/common/plat/mem/prdfMemCeTable.H b/src/usr/diag/prdf/common/plat/mem/prdfMemCeTable.H
index 18a9fa12f..0f21ff457 100644
--- a/src/usr/diag/prdf/common/plat/mem/prdfMemCeTable.H
+++ b/src/usr/diag/prdf/common/plat/mem/prdfMemCeTable.H
@@ -65,6 +65,9 @@ class MemCeTable
         MNFG_TH_DRAM     = 0x08, ///< MNFG DRAM threshold reached.
         MNFG_TH_RANK     = 0x10, ///< MNFG rank threshold reached.
         MNFG_TH_DIMM     = 0x20, ///< MNFG DIMM threshold reached.
+
+        FIELD_TH_ALL     = ENTRY_TH_REACHED | RANK_TH_REACHED | TABLE_FULL,
+        MNFG_TH_ALL      = MNFG_TH_DRAM     | MNFG_TH_RANK    | MNFG_TH_DIMM,
     };
 
   private: // constants, enums
diff --git a/src/usr/diag/prdf/common/plat/mem/prdfMemEccAnalysis.C b/src/usr/diag/prdf/common/plat/mem/prdfMemEccAnalysis.C
index 7036888fe..a43f50131 100644
--- a/src/usr/diag/prdf/common/plat/mem/prdfMemEccAnalysis.C
+++ b/src/usr/diag/prdf/common/plat/mem/prdfMemEccAnalysis.C
@@ -257,6 +257,7 @@ uint32_t __analyzeFetchNceTce( ExtensibleChip * i_chip, const MemAddr & i_addr,
     // Add data to the CE table.
     D db = static_cast<D>(i_chip->getDataBundle());
     uint32_t ceTableRc = db->iv_ceTable.addEntry( i_addr, i_symbol );
+    bool doTps = false;
 
     // Check MNFG thresholds, if needed.
     if ( mfgMode() )
@@ -265,16 +266,19 @@ uint32_t __analyzeFetchNceTce( ExtensibleChip * i_chip, const MemAddr & i_addr,
         {
             io_sc.service_data->AddSignatureList( trgt, PRDFSIG_MnfgDramCte );
             io_sc.service_data->setServiceCall();
+            doTps = true;
         }
         else if ( 0 != (MemCeTable<T>::MNFG_TH_RANK & ceTableRc) )
         {
             io_sc.service_data->AddSignatureList( trgt, PRDFSIG_MnfgRankCte );
             io_sc.service_data->setServiceCall();
+            doTps = true;
         }
         else if ( 0 != (MemCeTable<T>::MNFG_TH_DIMM & ceTableRc) )
         {
             io_sc.service_data->AddSignatureList( trgt, PRDFSIG_MnfgDimmCte );
             io_sc.service_data->setServiceCall();
+            doTps = true;
         }
         else if ( 0 != (MemCeTable<T>::TABLE_FULL & ceTableRc) )
         {
@@ -289,6 +293,7 @@ uint32_t __analyzeFetchNceTce( ExtensibleChip * i_chip, const MemAddr & i_addr,
             io_sc.service_data->SetCallout( all_mm, MRU_MEDA );
             io_sc.service_data->SetCallout( trgt,   MRU_MEDA );
             io_sc.service_data->setServiceCall();
+            doTps = true;
         }
         else if ( 0 != (MemCeTable<T>::ENTRY_TH_REACHED & ceTableRc) )
         {
@@ -298,14 +303,23 @@ uint32_t __analyzeFetchNceTce( ExtensibleChip * i_chip, const MemAddr & i_addr,
             // has been met. This is a potential flooding issue. So make
             // the DIMM callout predictive.
             io_sc.service_data->setServiceCall();
+            doTps = true;
         }
     }
-
-    #ifdef __HOSTBOOT_RUNTIME
+    else // field thresholds
+    {
+        // It is possible that the MNFG thresholds are higher than the field
+        // thresholds because of the scaling due to DRAM side. Therefore, we
+        // cannot simply trigger TPS on any threshold. The field and MNFG
+        // thresholds must be handled separately.
+        doTps = ( 0 != (MemCeTable<T>::FIELD_TH_ALL & ceTableRc) );
+    }
 
     // Initiate a TPS procedure, if needed.
-    if ( MemCeTable<T>::NO_TH_REACHED != ceTableRc )
+    if ( doTps )
     {
+        #ifdef __HOSTBOOT_RUNTIME
+
         // If a MNFG threshold has been reached (predictive callout), we
         // will still try to start TPS just in case MNFG disables the
         // termination policy.
@@ -316,9 +330,9 @@ uint32_t __analyzeFetchNceTce( ExtensibleChip * i_chip, const MemAddr & i_addr,
             PRDF_ERR( PRDF_FUNC "addTpsEvent(0x%08x, m%ds%d) failed",
                       i_chip->getHuid(), rank.getMaster(), rank.getSlave() );
         }
-    }
 
-    #endif
+        #endif
+    }
 
     return o_rc;