PRD: Stop commands during MDIA when handling RCD parity error

Change-Id: Ic896637293a398d241b0c9e43c1559ac9869d196
CQ: SW397959
Reviewed-on: http://ralgit01.raleigh.ibm.com/gerrit1/44194
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: Zane C. Shelley <zshelle@us.ibm.com>
Reviewed-on: http://ralgit01.raleigh.ibm.com/gerrit1/44122
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>

2 files changed, 49 insertions, 2 deletions

diff --git a/src/usr/diag/prdf/common/plat/p9/p9_mcbist_regs.rule b/src/usr/diag/prdf/common/plat/p9/p9_mcbist_regs.rule
index 0b6c0465f..4200737bb 100644
--- a/src/usr/diag/prdf/common/plat/p9/p9_mcbist_regs.rule
+++ b/src/usr/diag/prdf/common/plat/p9/p9_mcbist_regs.rule
@@ -29,6 +29,7 @@
     ###########################################################################
     # P9 MCBIST target MCBISTFIR
     ###########################################################################
+
     register MCBISTFIR_AND
     {
         name        "P9 MCBIST target MCBISTFIR AND";
@@ -37,6 +38,14 @@
         access      write_only;
     };
 
+    register MCBISTFIR_MASK_OR
+    {
+        name        "P9 MCBIST target MCBISTFIR MASK OR";
+        scomaddr    0x07012305;
+        capture     group never;
+        access      write_only;
+    };
+
     ############################################################################
     # P9 Memory ECC Address Registers
     ############################################################################
diff --git a/src/usr/diag/prdf/plat/mem/prdfP9Mca.C b/src/usr/diag/prdf/plat/mem/prdfP9Mca.C
index 17161c03d..70e91a0f5 100644
--- a/src/usr/diag/prdf/plat/mem/prdfP9Mca.C
+++ b/src/usr/diag/prdf/plat/mem/prdfP9Mca.C
@@ -112,6 +112,8 @@ int32_t RcdParityError( ExtensibleChip * i_mcaChip,
     if ( CHECK_STOP == io_sc.service_data->getPrimaryAttnType() )
         return SUCCESS;
 
+    ExtensibleChip * mcbChip = getConnectedParent( i_mcaChip, TYPE_MCBIST );
+
     #ifdef __HOSTBOOT_RUNTIME // TPS only supported at runtime.
 
     // Recovery is always enabled during runtime. If the threshold is reached,
@@ -121,8 +123,6 @@ int32_t RcdParityError( ExtensibleChip * i_mcaChip,
     {
         io_sc.service_data->setServiceCall();
 
-        ExtensibleChip * mcbChip = getConnectedParent( i_mcaChip, TYPE_MCBIST );
-
         McbistDataBundle * mcbdb = getMcbistDataBundle( mcbChip );
 
         std::vector<MemRank> list;
@@ -162,6 +162,44 @@ int32_t RcdParityError( ExtensibleChip * i_mcaChip,
         // predictive if threshold is reached.
         if ( rcdParityErrorReconfigLoop(i_mcaChip->getTrgt()) )
             io_sc.service_data->setServiceCall();
+
+        if ( isInMdiaMode() )
+        {
+            uint32_t l_rc = SUCCESS;
+            SCAN_COMM_REGISTER_CLASS * mask = nullptr;
+
+            // Stop any further commands on this MCBIST to avoid subsequent RCD
+            // errors or potential AUEs.
+            l_rc = mdiaSendEventMsg( mcbChip->getTrgt(), MDIA::STOP_TESTING );
+            if ( SUCCESS != l_rc )
+            {
+                PRDF_ERR( PRDF_FUNC "mdiaSendEventMsg(STOP_TESTING) failed" );
+            }
+
+            // Mask the maintenance AUE/IAUE attentions on this MCA because they
+            // are potential side-effects of the RCD parity errors.
+            mask = i_mcaChip->getRegister( "MCAECCFIR_MASK_OR" );
+            mask->SetBit(33); // maintenance AUE
+            mask->SetBit(36); // maintenance IAUE
+            l_rc = mask->Write();
+            if ( SUCCESS != l_rc )
+            {
+                PRDF_ERR( PRDF_FUNC "Write() failed on MCAECCFIR_MASK_OR: "
+                          "i_mcaChip=0x%08x", i_mcaChip->getHuid() );
+            }
+
+            // Mask the maintenance command complete bits to avoid false
+            // attentions.
+            mask = mcbChip->getRegister( "MCBISTFIR_MASK_OR" );
+            mask->SetBit(10); // Command complete
+            mask->SetBit(12); // WAT workaround
+            l_rc = mask->Write();
+            if ( SUCCESS != l_rc )
+            {
+                PRDF_ERR( PRDF_FUNC "Write() failed on MCBISTFIR_MASK_OR: "
+                          "mcbChip=0x%08x", mcbChip->getHuid() );
+            }
+        }
     }
     else
     {