PRD: consistent handling for IMPE and VCM

Change-Id: Ic30829c48e54448c9a4f828dc24afe0e4d4d6bf0
CQ: SW394364
Reviewed-on: http://ralgit01.raleigh.ibm.com/gerrit1/43611
Tested-by: Jenkins Server <pfd-jenkins+hostboot@us.ibm.com>
Reviewed-by: Brian J. Stegmiller <bjs@us.ibm.com>
Reviewed-by: Benjamin J. Weisenbeck <bweisenb@us.ibm.com>
Reviewed-by: Caleb N. Palmer <cnpalmer@us.ibm.com>
Reviewed-by: Zane C. Shelley <zshelle@us.ibm.com>
Reviewed-on: http://ralgit01.raleigh.ibm.com/gerrit1/43798
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>

4 files changed, 308 insertions, 77 deletions

diff --git a/src/usr/diag/prdf/plat/mem/prdfMemTdFalseAlarm.H b/src/usr/diag/prdf/plat/mem/prdfMemTdFalseAlarm.H
index fd38ff832..9b497f8be 100644
--- a/src/usr/diag/prdf/plat/mem/prdfMemTdFalseAlarm.H
+++ b/src/usr/diag/prdf/plat/mem/prdfMemTdFalseAlarm.H
@@ -26,8 +26,6 @@
 #ifndef __prdfMemTdFalseAlarm_H
 #define __prdfMemTdFalseAlarm_H
 
-#ifdef __HOSTBOOT_RUNTIME // All of this code is runtime only
-
 // Framework includes
 #include <iipServiceDataCollector.h>
 #include <prdfThresholdUtils.H>
@@ -38,58 +36,213 @@
 namespace PRDF
 {
 
-/**
- * @brief At runtime, we have to keep a false alarm threshold for Targeted
- *        Diagnostics to avoid flooding of intermittent errors.
- *
- * This class is intented to be a static class variable for each TD event class
- * that requires this type of thresholding. It will contain a map for each chip
- * and unique key within each chip to the threshold container. Note that the key
- * could be different per TD event class. For example, VCM events will use only
- * the master rank, where TPS events will use both the master and slave rank.
- */
+/** @brief This is used at runtime to keep track of false alarm for Targeted
+ *         Diagnostics to avoid flooding of intermittent errors. */
 class TdFalseAlarm
 {
-  public:
+  public: // functions
+
+    /** @brief Default destructor */
+    virtual ~TdFalseAlarm()  = default;
+
+    /**
+     * @brief  Queries if the threshold timer has expired.
+     * @param  i_rank The rank with the false alarm.
+     * @param  io_sc  The step code data struct.
+     * @return True if the threshold timer has expired, false otherwise.
+     */
+    bool elapsed( MemRank i_rank, STEP_CODE_DATA_STRUCT & io_sc )
+    {
+        uint32_t key = init( i_rank, io_sc );
+        return iv_thMap[key].timeElapsed( io_sc );
+    }
+
+    /**
+     * @brief  Queries the current false alarm count.
+     * @param  i_rank The rank with the false alarm.
+     * @param  io_sc  The step code data struct.
+     * @return The current false alarm count.
+     */
+    uint8_t count( MemRank i_rank, STEP_CODE_DATA_STRUCT & io_sc )
+    {
+        uint32_t key = init( i_rank, io_sc );
+        return iv_thMap[key].getCount();
+    }
+
+  protected: // functions
 
     /**
      * @brief Constructor.
-     * @param i_th  Threshold value for all entries in the map.
-     * @param i_int Threshold interval for all entries in the map.
+     * @param i_th Threshold for each entry in the threshold map.
      */
-    TdFalseAlarm( uint8_t i_th, uint32_t i_int ) :
-        iv_thVal(i_th), iv_thInt(i_int)
-    {}
+    explicit TdFalseAlarm( TimeBasedThreshold i_th ) : iv_th(i_th) {}
+
+    /**
+     * @brief  The key could be different per TD procedure. For example, VCM
+     *         events will use only the master rank, where TPS events will use
+     *         both the master and slave rank.
+     * @param  i_rank The rank with the false alarm.
+     * @return The key value for this rank.
+     */
+    virtual uint32_t getKey( MemRank i_rank ) const = 0;
+
+    /**
+     * @brief  Initializes data specific to each child class.
+     * @param  i_rank The rank with the false alarm.
+     * @param  io_sc  The step code data struct.
+     * @return The key for this rank (see getKey()).
+     * @note   This function should be called before accessing the instance
+     *         variables to ensure this key is initialized with the correct
+     *         threshold.
+     * @note   The default is to call initThMap(). Each child class that
+     *         overloads this function should also call initThMap() to ensure
+     *         the threshold map is initialized properly.
+     */
+    virtual uint32_t init( MemRank i_rank, STEP_CODE_DATA_STRUCT & io_sc )
+    {
+        return initThMap( i_rank );
+    }
+
+    /**
+     * @brief  Initializes this key in the threshold map.
+     * @param  i_rank The rank with the false alarm.
+     * @return The key for this rank (see getKey()).
+     * @note   This function should only be called from init().
+     */
+    uint32_t initThMap( MemRank i_rank )
+    {
+        uint32_t key = getKey( i_rank );
+
+        // Create a new entry if an entry does not exist.
+        if ( iv_thMap.end() == iv_thMap.find(key) ) iv_thMap[key] = iv_th;
+
+        return key;
+    }
 
     /**
      * @brief  Increments the false alarm count.
-     * @param  i_chip Target chip.
-     * @param  i_key  Key relative to the chip.
+     * @param  i_rank The rank with the false alarm.
      * @param  io_sc  The step code data struct.
      * @return True if false alarm count has reached threshold, false otherwise.
      */
-    bool inc( ExtensibleChip * i_chip, uint32_t i_key,
-              STEP_CODE_DATA_STRUCT & io_sc )
+    bool incThMap( MemRank i_rank, STEP_CODE_DATA_STRUCT & io_sc )
     {
-        // Create a new entry if an entry does not exist.
-        if ( iv_map[i_chip].end() == iv_map[i_chip].find(i_key) )
-            iv_map[i_chip][i_key] = TimeBasedThreshold( iv_thVal, iv_thInt );
+        uint32_t key = init( i_rank, io_sc );
+        return iv_thMap[key].inc( io_sc );
+    }
+
+  private: // instance variables
+
+    /** Threshold for each entries in the map. */
+    const TimeBasedThreshold iv_th;
 
-        return iv_map[i_chip][i_key].inc( io_sc );
+    /** A map containing the thresholds for each key. */
+    std::map<uint32_t, TimeBasedThreshold> iv_thMap;
+};
+
+/** @brief A false alarm class specific to VCM procedures. */
+class VcmFalseAlarm : public TdFalseAlarm
+{
+  public:
+
+    /**
+     * @brief Constructor.
+     * @param i_th Threshold for each entry in the map.
+     */
+    VcmFalseAlarm( TimeBasedThreshold i_th ) : TdFalseAlarm(i_th) {}
+
+    /**
+     * @brief  Increments the false alarm count and stores the DRAM.
+     * @param  i_rank The rank with the false alarm.
+     * @param  i_dram The DRAM with the false alarm.
+     * @param  io_sc  The step code data struct.
+     * @return True if false alarm count has reached threshold, false otherwise.
+     */
+    bool inc( MemRank i_rank, uint8_t i_dram, STEP_CODE_DATA_STRUCT & io_sc )
+    {
+        // Increment the count and determine whether threshold is reached or
+        // not. Note that incThMap() calls init() and initializes all maps.
+        bool isTh = incThMap( i_rank, io_sc );
+
+        // Add the DRAM to the list. Note that the value for each DRAM is not
+        // important. The only reason to use a map verses a vector is to ensure
+        // unique entries in the list.
+        iv_dramMap[getKey(i_rank)][i_dram] = 1;
+
+        return isTh;
+    }
+
+    /**
+     * @param  i_rank The rank with the false alarm.
+     * @param  io_sc  The step code data struct.
+     * @return True there is more than one DRAM on this rank and the timer has
+     *         not elapsed, false otherwise.
+     */
+    bool queryDrams( MemRank i_rank, STEP_CODE_DATA_STRUCT & io_sc )
+    {
+        uint32_t key = init( i_rank, io_sc ); // will clear list if time elapsed
+
+        // Return true if there is more than one DRAM on this rank.
+        return ( 1 < iv_dramMap[key].size() );
+    }
+
+  private: // functions
+
+    // Overloaded from parent class.
+    uint32_t getKey( MemRank i_rank ) const
+    {
+        return MemRank(i_rank.getMaster()).getKey(); // master only
+    }
+
+    // Overloaded from parent class.
+    uint32_t init( MemRank i_rank, STEP_CODE_DATA_STRUCT & io_sc )
+    {
+        uint32_t key = initThMap( i_rank );
+
+        // Clear out the list of DRAMs if the threshold time has elapsed.
+        if ( elapsed(i_rank, io_sc) ) iv_dramMap[key].clear();
+
+        return key;
     }
 
-  private:
+  private: // instance variables
+
+    /** A map to keep track of which DRAMs per rank have reported chip marks. */
+    std::map< uint32_t, std::map<uint8_t, uint8_t> > iv_dramMap;
+};
+
+/** @brief A false alarm class specific to TPS procedures. */
+class TpsFalseAlarm : public TdFalseAlarm
+{
+  public:
+
+    /**
+     * @brief Constructor.
+     * @param i_th Threshold for each entry in the map.
+     */
+    TpsFalseAlarm( TimeBasedThreshold i_th ) : TdFalseAlarm(i_th) {}
+
+    /**
+     * @brief  Increments the false alarm count.
+     * @param  i_rank The rank with the false alarm.
+     * @param  io_sc  The step code data struct.
+     * @return True if false alarm count has reached threshold, false otherwise.
+     */
+    bool inc( MemRank i_rank, STEP_CODE_DATA_STRUCT & io_sc )
+    {
+        return incThMap( i_rank, io_sc );
+    }
 
-    const uint8_t  iv_thVal; ///< Threshold value for all entries in the map.
-    const uint32_t iv_thInt; ///< Threshold interval for all entries in the map.
+  private: // functions
 
-    /** A nested map containing the thresholds for each chip and key. */
-    std::map< ExtensibleChip *, std::map<uint32_t,TimeBasedThreshold> > iv_map;
+    // Overloaded from parent class.
+    uint32_t getKey( MemRank i_rank ) const
+    {
+        return i_rank.getKey(); // both master and slave
+    }
 };
 
 } // end namespace PRDF
 
-#endif // __HOSTBOOT_RUNTIME
-
 #endif // __prdfMemTdFalseAlarm_H
 
diff --git a/src/usr/diag/prdf/plat/mem/prdfMemVcm.H b/src/usr/diag/prdf/plat/mem/prdfMemVcm.H
index 9c39040cd..c42598935 100644
--- a/src/usr/diag/prdf/plat/mem/prdfMemVcm.H
+++ b/src/usr/diag/prdf/plat/mem/prdfMemVcm.H
@@ -260,32 +260,20 @@ class VcmEvent : public TdEntry
         io_sc.service_data->setSignature( iv_chip->getHuid(),
                                           PRDFSIG_VcmVerified );
 
-        do
+        if ( PlatServices::areDramRepairsDisabled() )
         {
-            // If DRAM repairs are disabled, make the error log predictive.
-            if ( PlatServices::areDramRepairsDisabled() )
-            {
-                io_sc.service_data->setServiceCall();
-                break; // Nothing more to do.
-            }
-
-            // If there is a symbol mark on the same DRAM as the newly verified
-            // chip mark, remove the symbol mark.
-            o_rc = MarkStore::balance<T>( iv_chip, iv_rank, io_sc );
+            // Make the error log predictive, nothing else to do.
+            io_sc.service_data->setServiceCall();
+        }
+        else
+        {
+            // Leave the chip mark in place and do any necessary cleanup.
+            o_rc = cleanup( io_sc );
             if ( SUCCESS != o_rc )
             {
-                PRDF_ERR( PRDF_FUNC "MarkStore::balance(0x%08x,0x%02x) failed",
-                          iv_chip->getHuid(), getKey() );
-                break;
+                PRDF_ERR( PRDF_FUNC "cleanup() failed" );
             }
-
-            // Set the entire chip in DRAM Repairs VPD.
-            // TODO: RTC 169939
-
-            // Add a DRAM sparing procedure to the queue, if supported.
-            // TODO: RTC 157888
-
-        } while (0);
+        }
 
         return o_rc;
 
@@ -293,6 +281,16 @@ class VcmEvent : public TdEntry
     }
 
     /**
+     * @brief  Cleanup required when a chip mark is left in place (i.e. chip
+     *         mark verified or false alarm threshold). Will balance the chip
+     *         and symbol marks, set VPD, and initiate DRAM sparing if
+     *         supported.
+     * @param  io_sc  The step code data struct.
+     * @return Non-SUCCESS if an internal function fails, SUCCESS otherwise.
+     */
+    uint32_t cleanup( STEP_CODE_DATA_STRUCT & io_sc );
+
+    /**
      * @brief  Verification failed. Do additional processing such as removing
      *         the chip mark and false alarm threshold handling.
      * @param  io_sc  The step code data struct.
@@ -303,11 +301,6 @@ class VcmEvent : public TdEntry
   private: // instance variables
 
     const MemMark iv_mark; ///< The chip mark from hardware.
-
-    #ifdef __HOSTBOOT_RUNTIME
-    /** False alarm counter for all instances of this class. */
-    static TdFalseAlarm cv_falseAlarm;
-    #endif
 };
 
 } // end namespace PRDF
diff --git a/src/usr/diag/prdf/plat/mem/prdfMemVcm_ipl.C b/src/usr/diag/prdf/plat/mem/prdfMemVcm_ipl.C
index e331d736c..50cfec50f 100644
--- a/src/usr/diag/prdf/plat/mem/prdfMemVcm_ipl.C
+++ b/src/usr/diag/prdf/plat/mem/prdfMemVcm_ipl.C
@@ -81,6 +81,40 @@ uint32_t VcmEvent<T>::falseAlarm( STEP_CODE_DATA_STRUCT & io_sc )
 
 //------------------------------------------------------------------------------
 
+template<TARGETING::TYPE T>
+uint32_t VcmEvent<T>::cleanup( STEP_CODE_DATA_STRUCT & io_sc )
+{
+    #define PRDF_FUNC "[VcmEvent::cleanup] "
+
+    uint32_t o_rc = SUCCESS;
+
+    do
+    {
+        // If there is a symbol mark on the same DRAM as the newly verified chip
+        // mark, remove the symbol mark.
+        o_rc = MarkStore::balance<T>( iv_chip, iv_rank, io_sc );
+        if ( SUCCESS != o_rc )
+        {
+            PRDF_ERR( PRDF_FUNC "MarkStore::balance(0x%08x,0x%02x) failed",
+                      iv_chip->getHuid(), getKey() );
+            break;
+        }
+
+        // Set the entire chip in DRAM Repairs VPD.
+        // TODO: RTC 169939
+
+        // Add a DRAM sparing procedure to the queue, if supported.
+        // TODO: RTC 157888
+
+    } while (0);
+
+    return o_rc;
+
+    #undef PRDF_FUNC
+}
+
+//------------------------------------------------------------------------------
+
 // Avoid linker errors with the template.
 template class VcmEvent<TYPE_MCA>;
 template class VcmEvent<TYPE_MBA>;
diff --git a/src/usr/diag/prdf/plat/mem/prdfMemVcm_rt.C b/src/usr/diag/prdf/plat/mem/prdfMemVcm_rt.C
index 835a54cfe..bce497fcd 100644
--- a/src/usr/diag/prdf/plat/mem/prdfMemVcm_rt.C
+++ b/src/usr/diag/prdf/plat/mem/prdfMemVcm_rt.C
@@ -38,6 +38,29 @@ using namespace PlatServices;
 
 //##############################################################################
 //
+//                               Helper functions
+//
+//##############################################################################
+
+template<TARGETING::TYPE T>
+VcmFalseAlarm * __getFalseAlarmCounter( ExtensibleChip * i_chip );
+
+template<>
+VcmFalseAlarm * __getFalseAlarmCounter<TYPE_MCA>( ExtensibleChip * i_chip )
+{
+    return getMcaDataBundle(i_chip)->getVcmFalseAlarmCounter();
+}
+
+template<>
+VcmFalseAlarm * __getFalseAlarmCounter<TYPE_MBA>( ExtensibleChip * i_chip )
+{
+    // TODO: RTC 157888
+    //return getMbaDataBundle(i_chip)->getVcmFalseAlarmCounter();
+    return nullptr;
+}
+
+//##############################################################################
+//
 //                          Generic template functions
 //
 //##############################################################################
@@ -65,10 +88,10 @@ uint32_t VcmEvent<T>::falseAlarm( STEP_CODE_DATA_STRUCT & io_sc )
         }
 
         // Increment the false alarm counter and check threshold.
-        if ( cv_falseAlarm.inc(iv_chip, getKey(), io_sc) )
+        uint8_t dram = iv_mark.getSymbol().getDram();
+        if ( __getFalseAlarmCounter<T>(iv_chip)->inc(iv_rank, dram, io_sc) )
         {
-            // False alarm threshold has been reached. Leave the mark in place
-            // and treat the chip mark as verified.
+            // False alarm threshold has been reached.
 
             io_sc.service_data->setSignature( iv_chip->getHuid(),
                                               PRDFSIG_VcmFalseAlarmTH );
@@ -76,10 +99,11 @@ uint32_t VcmEvent<T>::falseAlarm( STEP_CODE_DATA_STRUCT & io_sc )
             PRDF_TRAC( PRDF_FUNC "False alarm threshold: 0x%08x,0x%02x",
                        iv_chip->getHuid(), getKey() );
 
-            o_rc = verified( io_sc );
+            // Leave the chip mark in place and do any necessary cleanup.
+            o_rc = cleanup( io_sc );
             if ( SUCCESS != o_rc )
             {
-                PRDF_ERR( PRDF_FUNC "verified() failed" );
+                PRDF_ERR( PRDF_FUNC "cleanup() failed" );
                 break;
             }
         }
@@ -104,6 +128,45 @@ uint32_t VcmEvent<T>::falseAlarm( STEP_CODE_DATA_STRUCT & io_sc )
 
 //------------------------------------------------------------------------------
 
+template<TARGETING::TYPE T>
+uint32_t VcmEvent<T>::cleanup( STEP_CODE_DATA_STRUCT & io_sc )
+{
+    #define PRDF_FUNC "[VcmEvent::cleanup] "
+
+    uint32_t o_rc = SUCCESS;
+
+    do
+    {
+        // If there is a symbol mark on the same DRAM as the newly verified chip
+        // mark, remove the symbol mark.
+        o_rc = MarkStore::balance<T>( iv_chip, iv_rank, io_sc );
+        if ( SUCCESS != o_rc )
+        {
+            PRDF_ERR( PRDF_FUNC "MarkStore::balance(0x%08x,0x%02x) failed",
+                      iv_chip->getHuid(), getKey() );
+            break;
+        }
+
+        // Set the entire chip in DRAM Repairs VPD.
+        // TODO: RTC 169939
+
+        // Add a DRAM sparing procedure to the queue, if supported.
+        // TODO: RTC 157888
+
+        // If there was more than one DRAM on this rank with a false alarm, make
+        // the error log predictive.
+        if ( __getFalseAlarmCounter<T>(iv_chip)->queryDrams(iv_rank, io_sc) )
+            io_sc.service_data->setServiceCall();
+
+    } while (0);
+
+    return o_rc;
+
+    #undef PRDF_FUNC
+}
+
+//------------------------------------------------------------------------------
+
 // Avoid linker errors with the template.
 template class VcmEvent<TYPE_MCA>;
 template class VcmEvent<TYPE_MBA>;
@@ -115,12 +178,6 @@ template class VcmEvent<TYPE_MBA>;
 //##############################################################################
 
 template<>
-TdFalseAlarm VcmEvent<TYPE_MCA>::cv_falseAlarm
-                            = TdFalseAlarm { 4, ThresholdResolution::ONE_DAY };
-
-//------------------------------------------------------------------------------
-
-template<>
 uint32_t VcmEvent<TYPE_MCA>::checkEcc( const uint32_t & i_eccAttns,
                                        STEP_CODE_DATA_STRUCT & io_sc,
                                        bool & o_done )
@@ -203,12 +260,6 @@ uint32_t VcmEvent<TYPE_MCA>::checkEcc( const uint32_t & i_eccAttns,
 //##############################################################################
 
 template<>
-TdFalseAlarm VcmEvent<TYPE_MBA>::cv_falseAlarm
-                        = TdFalseAlarm { 4, 7 * ThresholdResolution::ONE_DAY };
-
-//------------------------------------------------------------------------------
-
-template<>
 uint32_t VcmEvent<TYPE_MBA>::checkEcc( const uint32_t & i_eccAttns,
                                        STEP_CODE_DATA_STRUCT & io_sc,
                                        bool & o_done )