HB Attribute for IPL chkstop analysis

Change-Id: Ied553bd777704a25152ab4bb35b80167b71bf603
CQ:SW409689
Backport: release-op910
Reviewed-on: http://ralgit01.raleigh.ibm.com/gerrit1/50427
Tested-by: Jenkins Server <pfd-jenkins+hostboot@us.ibm.com>
Tested-by: FSP CI Jenkins <fsp-CI-jenkins+hostboot@us.ibm.com>
Tested-by: Jenkins OP Build CI <op-jenkins+hostboot@us.ibm.com>
Tested-by: Jenkins OP HW <op-hw-jenkins+hostboot@us.ibm.com>
Reviewed-by: Zane C. Shelley <zshelle@us.ibm.com>
Reviewed-by: Caleb N. Palmer <cnpalmer@us.ibm.com>
Reviewed-by: Daniel M. Crowell <dcrowell@us.ibm.com>

6 files changed, 143 insertions, 4 deletions

diff --git a/src/usr/diag/attn/ipl/attn.C b/src/usr/diag/attn/ipl/attn.C
index 6b203c613..60acbc328 100644
--- a/src/usr/diag/attn/ipl/attn.C
+++ b/src/usr/diag/attn/ipl/attn.C
@@ -270,6 +270,16 @@ errlHndl_t checkForCSAttentions()
         FileRegSvc fileRegs;
         fileRegs.installScomImpl();
 
+        TARGETING::Target   *l_sys = NULL;
+        TARGETING::targetService().getTopLevelTarget( l_sys );
+        assert(l_sys != NULL);
+        // Indicate to PRD that we are doing chkstop analysis
+        // on saved FIR data so you can read regs without HW issues.
+        uint8_t l_iplAnalysis = CHKSTOP_ANALYSIS_ON_STARTUP_ANALYZING_CHECKSTOP;
+        bool l_set =
+             l_sys->trySetAttr<ATTR_CHKSTOP_ANALYSIS_ON_STARTUP>(l_iplAnalysis);
+        ATTN_SLOW("Checkstop attribute set for PRD:%d", l_set);
+
         // Process the checkstop attention
         errl = Singleton<Service>::instance().processCheckstop();
         if ( NULL != errl )
@@ -278,6 +288,10 @@ errlHndl_t checkForCSAttentions()
             errlCommit( errl, ATTN_COMP_ID );
         }
 
+        // Ensure we don't leave analysis attribute set
+        l_iplAnalysis = CHKSTOP_ANALYSIS_ON_STARTUP_NOT_ANALYZING_DEFAULT;
+        l_sys->trySetAttr<ATTR_CHKSTOP_ANALYSIS_ON_STARTUP>(l_iplAnalysis);
+
         // Uninstall File scom implementation.
         Singleton<ScomImpl>::instance().installScomImpl();
 
diff --git a/src/usr/diag/prdf/common/framework/service/xspprdsdbug.C b/src/usr/diag/prdf/common/framework/service/xspprdsdbug.C
index ea7c1136e..855ce9bc1 100755
--- a/src/usr/diag/prdf/common/framework/service/xspprdsdbug.C
+++ b/src/usr/diag/prdf/common/framework/service/xspprdsdbug.C
@@ -189,6 +189,27 @@ uint32_t SYSTEM_DEBUG_CLASS::Reinitialize(const AttnList & i_attnList)
         {
             addChipToAttnList( (*i).targetHndl,(*i).attnType );
 
+            // We can have UNIT CHKSTOPs present along with a normal
+            // CheckStop.  We want to be sure we look at the UnitCs
+            // in this case as it may be the real issue -- example is
+            // OPAL forcing a checkstop since a Unit CS occurs which
+            // they can't handle.
+            if  ( (*i).attnType == CHECK_STOP )
+            {
+                bool l_ucsFound = false;
+                ExtensibleChip * l_chip =
+                    ( ExtensibleChip *) systemPtr->GetChip( (*i).targetHndl );
+                ExtensibleChipFunction * ef
+                           = l_chip->getExtensibleFunction("CheckForUnitCs");
+
+                (*ef)( l_chip, bindParm<bool &>( l_ucsFound ) );
+
+                if ( l_ucsFound )
+                {
+                    addChipToAttnList( (*i).targetHndl,UNIT_CS );
+                }
+            }
+
             // There can be a case where chip has both recoverable and Check
             // Stop. In that case chip shall report only Check Stop. In such a
             // case, we analyse the recoverable first and see if we can blame
diff --git a/src/usr/diag/prdf/common/plat/p9/prdfP9Proc.C b/src/usr/diag/prdf/common/plat/p9/prdfP9Proc.C
index 2593492d9..6bae441b4 100644
--- a/src/usr/diag/prdf/common/plat/p9/prdfP9Proc.C
+++ b/src/usr/diag/prdf/common/plat/p9/prdfP9Proc.C
@@ -322,11 +322,34 @@ PRDF_PLUGIN_DEFINE_NS( p9_cumulus, Proc, isHostAttnFirAccessible );
 
 int32_t isUcsFirAccessible(ExtensibleChip * i_chip, bool & o_isOkToAccess)
 {
-    o_isOkToAccess = (true == atRuntime()) ? true : false;
-
     // Host Processor side can always access the 'unitCS' reg
     // The FSP can not access it during IPL steps 15 thru 16.2
-    // We  don't really use unitCs at this time.
+    o_isOkToAccess = atRuntime();
+
+
+#ifdef CONFIG_ENABLE_CHECKSTOP_ANALYSIS
+    if (false == o_isOkToAccess)
+    {
+        // For checkstop analysis (non-FSP systems), we
+        // already collected registers in FIRDATA. Hence, we
+        // can attempt reading the unit CS FIR from this data.
+        Target   *l_sys = NULL;
+        targetService().getTopLevelTarget( l_sys );
+        assert(l_sys != NULL);
+
+        // ATTN code sets an attribute to indicate we are
+        // doing the analysis of FIR data from prior failure.
+        uint8_t l_doingAnalysis =
+                           CHKSTOP_ANALYSIS_ON_STARTUP_NOT_ANALYZING_DEFAULT;
+        l_sys->tryGetAttr<ATTR_CHKSTOP_ANALYSIS_ON_STARTUP>(l_doingAnalysis);
+
+        if (CHKSTOP_ANALYSIS_ON_STARTUP_ANALYZING_CHECKSTOP == l_doingAnalysis)
+        {
+            o_isOkToAccess = true;
+        }
+    }
+#endif  // CONFIG_ENABLE_CHECKSTOP_ANALYSIS
+
 
     return SUCCESS;
 }
@@ -353,6 +376,40 @@ PRDF_PLUGIN_DEFINE_NS( p9_cumulus, Proc, handleDeadmanTimer );
 
 //------------------------------------------------------------------------------
 
+/**
+ * @brief  Used when the chip has a CHECK_STOP to see if there
+ *         is also a UNIT_CS present
+ * @param  i_chip         A P9 chip.
+ * @param  o_hasUcs True if the chip has a UNIT CS attention.
+ * @return SUCCESS
+ */
+int32_t CheckForUnitCs( ExtensibleChip * i_chip,
+                        bool & o_hasUcs )
+{
+    o_hasUcs = false;
+
+    int32_t l_rc = SUCCESS;
+
+    SCAN_COMM_REGISTER_CLASS * l_gUcsR = i_chip->getRegister("GLOBAL_UCS_FIR");
+    l_rc = l_gUcsR->Read();
+
+    if ( SUCCESS != l_rc )
+    {
+        PRDF_ERR("[CheckForUnitCs] GLOBAL_UCS_FIR read failed"
+                 "for 0x%08x", i_chip->GetId());
+    }
+    else if ( 0 != l_gUcsR->GetBitFieldJustified(1,55) )
+    {
+        o_hasUcs = true;
+    }
+
+    return SUCCESS;
+}
+PRDF_PLUGIN_DEFINE_NS( p9_nimbus,  Proc, CheckForUnitCs );
+PRDF_PLUGIN_DEFINE_NS( p9_cumulus, Proc, CheckForUnitCs );
+
+//------------------------------------------------------------------------------
+
 } // end namespace Proc
 
 } // end namespace PRDF
diff --git a/src/usr/diag/prdf/occ_firdata/prdfWriteHomerFirData.C b/src/usr/diag/prdf/occ_firdata/prdfWriteHomerFirData.C
index 5182afa76..472c5a57d 100644
--- a/src/usr/diag/prdf/occ_firdata/prdfWriteHomerFirData.C
+++ b/src/usr/diag/prdf/occ_firdata/prdfWriteHomerFirData.C
@@ -943,7 +943,25 @@ static HOMER_ChipSpecAddr_t  s_ecDepProcRegisters[]
       0, 0x0000000005013C40ll }, // NPU1FIR DD2
 
     { HOMER_CHIP_NIMBUS, TRGT_PROC, REG_FIR, 0x20,
-      0, 0x000000005013C80ll }   // NPU2FIR DD2
+      0, 0x000000005013C80ll },  // NPU2FIR DD2
+
+    { HOMER_CHIP_NIMBUS, TRGT_PROC, REG_FIR, 0x21,
+      0, 0x0000000005013C00ll }, // NPU0FIR DD2.1
+
+    { HOMER_CHIP_NIMBUS, TRGT_PROC, REG_FIR, 0x21,
+      0, 0x0000000005013C40ll }, // NPU1FIR DD2.1
+
+    { HOMER_CHIP_NIMBUS, TRGT_PROC, REG_FIR, 0x21,
+      0, 0x000000005013C80ll },  // NPU2FIR DD2.1
+
+    { HOMER_CHIP_NIMBUS, TRGT_PROC, REG_FIR, 0x22,
+      0, 0x0000000005013C00ll }, // NPU0FIR DD2.2
+
+    { HOMER_CHIP_NIMBUS, TRGT_PROC, REG_FIR, 0x22,
+      0, 0x0000000005013C40ll }, // NPU1FIR DD2.2
+
+    { HOMER_CHIP_NIMBUS, TRGT_PROC, REG_FIR, 0x22,
+      0, 0x000000005013C80ll }   // NPU2FIR DD2.2
 };
 
 //------------------------------------------------------------------------------
diff --git a/src/usr/targeting/common/xmltohb/attribute_types_hb.xml b/src/usr/targeting/common/xmltohb/attribute_types_hb.xml
index ab13a4e5b..28e58dbe4 100755
--- a/src/usr/targeting/common/xmltohb/attribute_types_hb.xml
+++ b/src/usr/targeting/common/xmltohb/attribute_types_hb.xml
@@ -74,6 +74,34 @@
 </enumerationType>
 
 <attribute>
+  <description>
+      Are we doing checkstop analysis on
+      systems as we are booting back up based
+      on information captured in FIRDATA section.
+  </description>
+  <id>CHKSTOP_ANALYSIS_ON_STARTUP</id>
+  <persistency>volatile-zeroed</persistency>
+  <readable></readable>
+  <simpleType>
+    <uint8_t></uint8_t>
+  </simpleType>
+  <writeable></writeable>
+</attribute>
+
+<enumerationType>
+  <description>Enum for CHKSTOP_ANALYSIS_ON_STARTUP</description>
+  <enumerator>
+    <name>NOT_ANALYZING_DEFAULT</name>
+    <value>0</value>
+  </enumerator>
+  <enumerator>
+    <name>ANALYZING_CHECKSTOP</name>
+    <value>1</value>
+  </enumerator>
+  <id>CHKSTOP_ANALYSIS_ON_STARTUP</id>
+</enumerationType>
+
+<attribute>
     <id>IS_MPIPL_HB</id>
     <description>1 = in Memory Preserving IPL mode. 0 = in normal IPL mode.</description>
     <simpleType>
diff --git a/src/usr/targeting/common/xmltohb/target_types_hb.xml b/src/usr/targeting/common/xmltohb/target_types_hb.xml
index 3c60a832c..60b85e1d6 100644
--- a/src/usr/targeting/common/xmltohb/target_types_hb.xml
+++ b/src/usr/targeting/common/xmltohb/target_types_hb.xml
@@ -47,6 +47,7 @@
     <attribute><id>HB_RSV_MEM_NEXT_SECTION</id></attribute>
     <attribute><id>ALLOW_ATTR_OVERRIDES_IN_SECURE_MODE</id></attribute>
     <attribute><id>HIDDEN_ERRLOGS_ENABLE</id></attribute>
+    <attribute><id>CHKSTOP_ANALYSIS_ON_STARTUP</id></attribute>
 </targetTypeExtension>
 
 <targetTypeExtension>