PRD: resume super fast read support for Row Repair

Change-Id: I7735becb3a6b8caf5d52d3ce9be7e6b1f50b4f21
RTC: 196073
Reviewed-on: http://rchgit01.rchland.ibm.com/gerrit1/64072
Reviewed-by: Caleb N. Palmer <cnpalmer@us.ibm.com>
Tested-by: Jenkins Server <pfd-jenkins+hostboot@us.ibm.com>
Reviewed-by: Brian J. Stegmiller <bjs@us.ibm.com>
Reviewed-by: Zane C. Shelley <zshelle@us.ibm.com>
Reviewed-on: http://rchgit01.rchland.ibm.com/gerrit1/64556
Tested-by: Jenkins OP Build CI <op-jenkins+hostboot@us.ibm.com>
Tested-by: FSP CI Jenkins <fsp-CI-jenkins+hostboot@us.ibm.com>
Tested-by: Jenkins OP HW <op-hw-jenkins+hostboot@us.ibm.com>

5 files changed, 300 insertions, 71 deletions

diff --git a/src/usr/diag/prdf/plat/prdfPlatServices.C b/src/usr/diag/prdf/plat/prdfPlatServices.C
index ad742e30d..96729796b 100644
--- a/src/usr/diag/prdf/plat/prdfPlatServices.C
+++ b/src/usr/diag/prdf/plat/prdfPlatServices.C
@@ -37,6 +37,7 @@
 #include <prdfErrlUtil.H>
 #include <prdfTrace.H>
 #include <prdfAssert.h>
+#include <prdfRegisterCache.H>
 
 #include <prdfCenMbaDataBundle.H>
 #include <prdfMemScrubUtils.H>
@@ -927,6 +928,103 @@ uint32_t startTdScrub<TYPE_MBA>( ExtensibleChip * i_chip,
     #undef PRDF_FUNC
 }
 
+//------------------------------------------------------------------------------
+
+template<>
+uint32_t incMaintAddr<TYPE_MBA>( ExtensibleChip * i_chip,
+                                 MemAddr & o_addr )
+{
+    #define PRDF_FUNC "[PlatServices::incMaintAddr<TYPE_MBA>] "
+
+    PRDF_ASSERT( nullptr != i_chip );
+    PRDF_ASSERT( TYPE_MBA == i_chip->getType() );
+
+    uint32_t o_rc = SUCCESS;
+
+    fapi2::Target<fapi2::TARGET_TYPE_MBA> fapiTrgt ( i_chip->getTrgt() );
+    errlHndl_t errl = nullptr;
+
+    do
+    {
+        // Manually clear the CE counters based on the error type and clear the
+        // maintenance FIRs. Note that we only want to clear counters that are
+        // at attention to allow the other CE types the opportunity to reach
+        // threshold, if possible.
+        o_rc = conditionallyClearEccCounters<TYPE_MBA>( i_chip );
+        if ( SUCCESS != o_rc )
+        {
+            PRDF_ERR( PRDF_FUNC "conditionallyClearEccCounters(0x%08x) failed",
+                      i_chip->getHuid() );
+            break;
+        }
+
+        o_rc = clearEccFirs<TYPE_MBA>( i_chip );
+        if ( SUCCESS != o_rc )
+        {
+            PRDF_ERR( PRDF_FUNC "clearEccFirs(0x%08x) failed",
+                      i_chip->getHuid() );
+            break;
+        }
+
+        o_rc = clearCmdCompleteAttn<TYPE_MBA>( i_chip );
+        if ( SUCCESS != o_rc )
+        {
+            PRDF_ERR( PRDF_FUNC "clearCmdCompleteAttn(0x%08x) failed",
+                      i_chip->getHuid() );
+            break;
+        }
+
+        // Increment the current maintenance address.
+        mss_IncrementAddress incCmd { fapiTrgt };
+        FAPI_INVOKE_HWP( errl, incCmd.setupAndExecuteCmd );
+        if ( nullptr != errl )
+        {
+            PRDF_ERR( PRDF_FUNC "mss_IncrementAddress setupAndExecuteCmd() on "
+                      "0x%08x failed", i_chip->getHuid() );
+            PRDF_COMMIT_ERRL( errl, ERRL_ACTION_REPORT );
+            o_rc = FAIL;
+            break;
+        }
+
+        // Clear the maintenance FIRs again. This time do not clear the CE
+        // counters.
+        o_rc = clearEccFirs<TYPE_MBA>( i_chip );
+        if ( SUCCESS != o_rc )
+        {
+            PRDF_ERR( PRDF_FUNC "clearEccFirs(0x%08x) failed",
+                      i_chip->getHuid() );
+            break;
+        }
+
+        o_rc = clearCmdCompleteAttn<TYPE_MBA>( i_chip );
+        if ( SUCCESS != o_rc )
+        {
+            PRDF_ERR( PRDF_FUNC "clearCmdCompleteAttn(0x%08x) failed",
+                      i_chip->getHuid() );
+            break;
+        }
+
+        // The address register has been updated so we need to clear our cache
+        // to ensure we can do a new read.
+        SCAN_COMM_REGISTER_CLASS * reg = i_chip->getRegister( "MBMACA" );
+        RegDataCache::getCachedRegisters().flush( i_chip, reg );
+
+        // Read the new start address from hardware.
+        o_rc = getMemMaintAddr<TYPE_MBA>( i_chip, o_addr );
+        if ( SUCCESS != o_rc )
+        {
+            PRDF_ERR( PRDF_FUNC "getMemMaintAddr(0x%08x) failed",
+                      i_chip->getHuid() );
+            break;
+        }
+
+    } while (0);
+
+    return o_rc;
+
+    #undef PRDF_FUNC
+}
+
 //##############################################################################
 //##                  Core/cache trace array functions
 //##############################################################################
diff --git a/src/usr/diag/prdf/plat/prdfPlatServices.H b/src/usr/diag/prdf/plat/prdfPlatServices.H
index bd20bd239..28dd3d536 100644
--- a/src/usr/diag/prdf/plat/prdfPlatServices.H
+++ b/src/usr/diag/prdf/plat/prdfPlatServices.H
@@ -186,6 +186,15 @@ template<TARGETING::TYPE TT, typename SCT>
 uint32_t startTdScrub( ExtensibleChip * i_chip, const MemRank & i_rank,
                        AddrRangeType i_rangeType, SCT i_stopCond );
 
+/**
+ * @brief  Increment the current maintenance address.
+ * @param  i_chip      MBA chip.
+ * @param  o_newAddr   The new address from the maintenance address register.
+ * @return Non-SUCCESS if an internal function fails, otherwise SUCCESS.
+ */
+template<TARGETING::TYPE T>
+uint32_t incMaintAddr( ExtensibleChip * i_chip, MemAddr & o_newAddr );
+
 //##############################################################################
 //##                  Core/cache trace array functions
 //##############################################################################
diff --git a/src/usr/diag/prdf/plat/prdfPlatServices_ipl.C b/src/usr/diag/prdf/plat/prdfPlatServices_ipl.C
index d3c0729a9..e8a22e1af 100644
--- a/src/usr/diag/prdf/plat/prdfPlatServices_ipl.C
+++ b/src/usr/diag/prdf/plat/prdfPlatServices_ipl.C
@@ -558,6 +558,83 @@ uint32_t startTdSfRead<TYPE_MBA>( ExtensibleChip * i_chip,
 //------------------------------------------------------------------------------
 
 template<>
+uint32_t resumeTdSfRead<TYPE_MBA>( ExtensibleChip * i_chip,
+                                   AddrRangeType i_rangeType,
+                                   uint32_t i_stopCond )
+{
+    #define PRDF_FUNC "[PlatServices::startTdSfRead<TYPE_MBA>] "
+
+    PRDF_ASSERT( isInMdiaMode() ); // MDIA must be running.
+
+    PRDF_ASSERT( nullptr != i_chip );
+    PRDF_ASSERT( TYPE_MBA == i_chip->getType() );
+
+    uint32_t o_rc = SUCCESS;
+
+    // Make sure there is a command complete attention when the command stops.
+    i_stopCond |= mss_MaintCmd::ENABLE_CMD_COMPLETE_ATTENTION;
+
+    // Make sure the command stops on the end address if there are no errors.
+    i_stopCond |= mss_MaintCmd::STOP_ON_END_ADDRESS;
+
+    fapi2::Target<fapi2::TARGET_TYPE_MBA> fapiTrgt ( i_chip->getTrgt() );
+    errlHndl_t errl = nullptr;
+
+    do
+    {
+        // Increment the address that is currently in hardware. Note that the CE
+        // counters will be conditionally cleared. Also, all of the appropriate
+        // attentions will be cleared as well.
+        MemAddr memSaddr;
+        o_rc = incMaintAddr<TYPE_MBA>( i_chip, memSaddr );
+        if ( SUCCESS != o_rc )
+        {
+            PRDF_ERR( PRDF_FUNC "incMaintAddr(0x%08x) failed",
+                      i_chip->getHuid() );
+            break;
+        }
+        fapi2::buffer<uint64_t> saddr = memSaddr.toMaintAddr<TYPE_MBA>();
+
+        // Get the address range of the given rank.
+        fapi2::buffer<uint64_t> junk, eaddr;
+        MemRank rank = memSaddr.getRank();
+        o_rc = getMemAddrRange<TYPE_MBA>( i_chip, rank, junk, eaddr,
+                                          i_rangeType );
+        if ( SUCCESS != o_rc )
+        {
+            PRDF_ERR( PRDF_FUNC "getMemAddrRange(0x%08x,0x%2x) failed",
+                      i_chip->getHuid(), rank.getKey() );
+            break;
+        }
+
+        // Create the new command. Store a pointer to the command in the MBA
+        // data bundle so that we can call the cleanup function after the
+        // command has completed.
+        MbaDataBundle * db = getMbaDataBundle( i_chip );
+        PRDF_ASSERT( nullptr == db->iv_sfCmd ); // Code bug.
+        db->iv_sfCmd = new mss_SuperFastRead { fapiTrgt, saddr, eaddr,
+                                               i_stopCond, false };
+
+        // Start the super fast read command.
+        FAPI_INVOKE_HWP( errl, db->iv_sfCmd->setupAndExecuteCmd );
+        if ( nullptr != errl )
+        {
+            PRDF_ERR( PRDF_FUNC "setupAndExecuteCmd() on 0x%08x,0x%02x failed",
+                      i_chip->getHuid(), rank.getKey() );
+            PRDF_COMMIT_ERRL( errl, ERRL_ACTION_REPORT );
+            o_rc = FAIL; break;
+        }
+
+    } while (0);
+
+    return o_rc;
+
+    #undef PRDF_FUNC
+}
+
+//------------------------------------------------------------------------------
+
+template<>
 uint32_t startTdSteerCleanup<TYPE_MBA>( ExtensibleChip * i_chip,
                                         const MemRank & i_rank,
                                         AddrRangeType i_rangeType,
@@ -616,7 +693,7 @@ uint32_t startTdSteerCleanup<TYPE_MBA>( ExtensibleChip * i_chip,
         // Get the MBA fapi target.
         fapi2::Target<fapi2::TARGET_TYPE_MBA> fapiTrgt ( i_chip->getTrgt() );
 
-        // Start the background scrub command.
+        // Start the steer cleanup command.
         mss_TimeBaseSteerCleanup cmd { fapiTrgt, saddr, eaddr, cmdSpeed,
                                        i_stopCond, false };
         errlHndl_t errl = nullptr;
@@ -638,6 +715,90 @@ uint32_t startTdSteerCleanup<TYPE_MBA>( ExtensibleChip * i_chip,
 
 //------------------------------------------------------------------------------
 
+template<>
+uint32_t resumeTdSteerCleanup<TYPE_MBA>( ExtensibleChip * i_chip,
+                                         AddrRangeType i_rangeType,
+                                         uint32_t i_stopCond )
+{
+    #define PRDF_FUNC "[PlatServices::startTdSteerCleanup<TYPE_MBA>] "
+
+    PRDF_ASSERT( isInMdiaMode() ); // MDIA must be running.
+
+    PRDF_ASSERT( nullptr != i_chip );
+    PRDF_ASSERT( TYPE_MBA == i_chip->getType() );
+
+    uint32_t o_rc = SUCCESS;
+
+    // Make sure there is a command complete attention when the command stops.
+    i_stopCond |= mss_MaintCmd::ENABLE_CMD_COMPLETE_ATTENTION;
+
+    // Make sure the command stops on the end address if there are no errors.
+    i_stopCond |= mss_MaintCmd::STOP_ON_END_ADDRESS;
+
+    // Default speed is to run as fast as possible.
+    mss_MaintCmd::TimeBaseSpeed cmdSpeed = mss_MaintCmd::FAST_MAX_BW_IMPACT;
+
+    // IUEs (reported via RCE ETE) are reported as UEs during read operations.
+    // Therefore, we will treat IUEs like UEs for scrub operations simply to
+    // maintain consistency during all of Memory Diagnostics. Note that since we
+    // set the stop on RCE ETE flag, this requires a threshold in the MBSTR.
+    // Fortunately, MDIA sets the threshold to 1 when it starts the first
+    // command on this MBA and that threshold should never change throughout all
+    // of Memory Diagnostics.
+
+    i_stopCond |= mss_MaintCmd::STOP_ON_RETRY_CE_ETE;
+
+    fapi2::Target<fapi2::TARGET_TYPE_MBA> fapiTrgt ( i_chip->getTrgt() );
+    errlHndl_t errl = nullptr;
+
+    do
+    {
+        // Increment the address that is currently in hardware. Note that the CE
+        // counters will be conditionally cleared. Also, all of the appropriate
+        // attentions will be cleared as well.
+        MemAddr newAddr;
+        o_rc = incMaintAddr<TYPE_MBA>( i_chip, newAddr );
+        if ( SUCCESS != o_rc )
+        {
+            PRDF_ERR( PRDF_FUNC "incMaintAddr(0x%08x) failed",
+                      i_chip->getHuid() );
+            break;
+        }
+        fapi2::buffer<uint64_t> saddr = newAddr.toMaintAddr<TYPE_MBA>();
+
+        // Get the address range of the given rank.
+        fapi2::buffer<uint64_t> junk, eaddr;
+        MemRank rank = newAddr.getRank();
+        o_rc = getMemAddrRange<TYPE_MBA>( i_chip, rank, junk, eaddr,
+                                          i_rangeType );
+        if ( SUCCESS != o_rc )
+        {
+            PRDF_ERR( PRDF_FUNC "getMemAddrRange(0x%08x,0x%2x) failed",
+                      i_chip->getHuid(), rank.getKey() );
+            break;
+        }
+
+        // Start the steer cleanup command.
+        mss_TimeBaseSteerCleanup cmd { fapiTrgt, saddr, eaddr, cmdSpeed,
+                                       i_stopCond, false };
+        FAPI_INVOKE_HWP( errl, cmd.setupAndExecuteCmd );
+        if ( nullptr != errl )
+        {
+            PRDF_ERR( PRDF_FUNC "setupAndExecuteCmd() on 0x%08x,0x%02x failed",
+                      i_chip->getHuid(), rank.getKey() );
+            PRDF_COMMIT_ERRL( errl, ERRL_ACTION_REPORT );
+            o_rc = FAIL; break;
+        }
+
+    } while (0);
+
+    return o_rc;
+
+    #undef PRDF_FUNC
+}
+
+//------------------------------------------------------------------------------
+
 } // end namespace PlatServices
 
 } // end namespace PRDF
diff --git a/src/usr/diag/prdf/plat/prdfPlatServices_ipl.H b/src/usr/diag/prdf/plat/prdfPlatServices_ipl.H
index 6bf884795..90724f3a3 100644
--- a/src/usr/diag/prdf/plat/prdfPlatServices_ipl.H
+++ b/src/usr/diag/prdf/plat/prdfPlatServices_ipl.H
@@ -145,6 +145,17 @@ uint32_t startTdSfRead( ExtensibleChip * i_chip, const MemRank & i_rank,
                         AddrRangeType i_rangeType, VT i_stopCond );
 
 /**
+ * @brief  Resumes TD super fast read after it has paused on error.
+ * @param  i_chip      MBA chip.
+ * @param  i_rangeType See enum AddrRangeType.
+ * @param  i_stopCond  The stop conditions for the targeted scrub.
+ * @return Non-SUCCESS if an internal function fails, SUCCESS otherwise.
+ */
+template<TARGETING::TYPE T, typename SCT>
+uint32_t resumeTdSfRead( ExtensibleChip * i_chip, AddrRangeType i_rangeType,
+                         SCT i_stopCond );
+
+/**
  * @brief  Starts a steer cleanup scrub command on the target rank.
  * @param  i_chip      MBA chip.
  * @param  i_rank      Target rank.
@@ -156,6 +167,18 @@ template<TARGETING::TYPE TT, typename VT>
 uint32_t startTdSteerCleanup( ExtensibleChip * i_chip, const MemRank & i_rank,
                               AddrRangeType i_rangeType, VT i_stopCond );
 
+/**
+ * @brief  Resumes TD steer cleanup after it has paused on error.
+ * @param  i_chip      MBA chip.
+ * @param  i_rangeType See enum AddrRangeType.
+ * @param  i_stopCond  The stop conditions for the targeted scrub.
+ * @return Non-SUCCESS if an internal function fails, SUCCESS otherwise.
+ */
+template<TARGETING::TYPE T, typename SCT>
+uint32_t resumeTdSteerCleanup( ExtensibleChip * i_chip,
+                               AddrRangeType i_rangeType,
+                               SCT i_stopCond );
+
 } // end namespace PlatServices
 
 } // end namespace PRDF
diff --git a/src/usr/diag/prdf/plat/prdfPlatServices_rt.C b/src/usr/diag/prdf/plat/prdfPlatServices_rt.C
index eb2db3e78..e04eff9c5 100644
--- a/src/usr/diag/prdf/plat/prdfPlatServices_rt.C
+++ b/src/usr/diag/prdf/plat/prdfPlatServices_rt.C
@@ -34,7 +34,6 @@
 // Framework includes
 #include <prdfErrlUtil.H>
 #include <prdfTrace.H>
-#include <prdfRegisterCache.H>
 
 // Platform includes
 #include <prdfCenMbaDataBundle.H>
@@ -290,83 +289,22 @@ uint32_t __resumeScrub<TYPE_MBA>( ExtensibleChip * i_chip,
 
     do
     {
-        // Manually clear the CE counters based on the error type and clear the
-        // maintenance FIRs. Note that we only want to clear counters that are
-        // at attention to allow the other CE types the opportunity to reach
-        // threshold, if possible.
-        o_rc = conditionallyClearEccCounters<TYPE_MBA>( i_chip );
+        // Increment the address that is currently in hardware. Note that the CE
+        // counters will be conditionally cleared. Also, all of the appropriate
+        // attentions will be cleared as well.
+        MemAddr memSaddr;
+        o_rc = incMaintAddr<TYPE_MBA>( i_chip, memSaddr );
         if ( SUCCESS != o_rc )
         {
-            PRDF_ERR( PRDF_FUNC "conditionallyClearEccCounters(0x%08x) failed",
+            PRDF_ERR( PRDF_FUNC "incMaintAddr(0x%08x) failed",
                       i_chip->getHuid() );
             break;
         }
-
-        o_rc = clearEccFirs<TYPE_MBA>( i_chip );
-        if ( SUCCESS != o_rc )
-        {
-            PRDF_ERR( PRDF_FUNC "clearEccFirs(0x%08x) failed",
-                      i_chip->getHuid() );
-            break;
-        }
-
-        o_rc = clearCmdCompleteAttn<TYPE_MBA>( i_chip );
-        if ( SUCCESS != o_rc )
-        {
-            PRDF_ERR( PRDF_FUNC "clearCmdCompleteAttn(0x%08x) failed",
-                      i_chip->getHuid() );
-            break;
-        }
-
-        // Increment the current maintenance address.
-        mss_IncrementAddress incCmd { fapiTrgt };
-        FAPI_INVOKE_HWP( errl, incCmd.setupAndExecuteCmd );
-        if ( nullptr != errl )
-        {
-            PRDF_ERR( PRDF_FUNC "mss_IncrementAddress setupAndExecuteCmd() on "
-                      "0x%08x failed", i_chip->getHuid() );
-            PRDF_COMMIT_ERRL( errl, ERRL_ACTION_REPORT );
-            o_rc = FAIL;
-            break;
-        }
-
-        // Clear the maintenance FIRs again. This time do not clear the CE
-        // counters.
-        o_rc = clearEccFirs<TYPE_MBA>( i_chip );
-        if ( SUCCESS != o_rc )
-        {
-            PRDF_ERR( PRDF_FUNC "clearEccFirs(0x%08x) failed",
-                      i_chip->getHuid() );
-            break;
-        }
-
-        o_rc = clearCmdCompleteAttn<TYPE_MBA>( i_chip );
-        if ( SUCCESS != o_rc )
-        {
-            PRDF_ERR( PRDF_FUNC "clearCmdCompleteAttn(0x%08x) failed",
-                      i_chip->getHuid() );
-            break;
-        }
-
-        // The address register has been updated so we need to clear our cache
-        // to ensure we can do a new read.
-        SCAN_COMM_REGISTER_CLASS * reg = i_chip->getRegister( "MBMACA" );
-        RegDataCache::getCachedRegisters().flush( i_chip, reg );
-
-        // Read the new start address from hardware.
-        MemAddr addr;
-        o_rc = getMemMaintAddr<TYPE_MBA>( i_chip, addr );
-        if ( SUCCESS != o_rc )
-        {
-            PRDF_ERR( PRDF_FUNC "getMemMaintAddr(0x%08x) failed",
-                      i_chip->getHuid() );
-            break;
-        }
-        fapi2::buffer<uint64_t> saddr = addr.toMaintAddr<TYPE_MBA>();
+        fapi2::buffer<uint64_t> saddr = memSaddr.toMaintAddr<TYPE_MBA>();
 
         // Get the end address of the current rank.
         fapi2::buffer<uint64_t> eaddr, junk;
-        MemRank rank = addr.getRank();
+        MemRank rank = memSaddr.getRank();
         o_rc = getMemAddrRange<TYPE_MBA>( i_chip, rank, junk, eaddr,
                                           i_rangeType );
         if ( SUCCESS != o_rc )