Updates Centaur training to continue on fails for FW

Change-Id: I295fbcdcee0691215a8b45ff951842801775b6b3
CQ:SW426968
RTC:192763
Reviewed-on: http://ralgit01.raleigh.ibm.com/gerrit1/58712
Tested-by: FSP CI Jenkins <fsp-CI-jenkins+hostboot@us.ibm.com>
Dev-Ready: STEPHEN GLANCY <sglancy@us.ibm.com>
Tested-by: Jenkins Server <pfd-jenkins+hostboot@us.ibm.com>
Reviewed-by: Louis Stermole <stermole@us.ibm.com>
Reviewed-by: ANDRE A. MARIN <aamarin@us.ibm.com>
Reviewed-by: Dean Sanner <dsanner@us.ibm.com>
Reviewed-on: http://ralgit01.raleigh.ibm.com/gerrit1/58718
Tested-by: Jenkins OP Build CI <op-jenkins+hostboot@us.ibm.com>
Tested-by: Jenkins OP HW <op-hw-jenkins+hostboot@us.ibm.com>

3 files changed, 42 insertions, 10 deletions

diff --git a/src/import/chips/centaur/procedures/hwp/memory/p9c_mss_draminit_training.C b/src/import/chips/centaur/procedures/hwp/memory/p9c_mss_draminit_training.C
index 5f46f29ad..c079e6d13 100755
--- a/src/import/chips/centaur/procedures/hwp/memory/p9c_mss_draminit_training.C
+++ b/src/import/chips/centaur/procedures/hwp/memory/p9c_mss_draminit_training.C
@@ -663,6 +663,8 @@ extern "C" {
             uint8_t& io_dqs_try
                                             )
     {
+        // Used to determine if the error should be logged as recovered or not
+        bool l_training_error = false;
         fapi2::buffer<uint64_t> l_cal_error_buffer_64;
         fapi2::buffer<uint64_t> l_disable_bit_data_for_dp18_buffer_64;
         uint8_t l_mbaPosition = 0;
@@ -682,7 +684,7 @@ extern "C" {
         if(l_cal_error_buffer_64.getBit<CEN_MBA_DDRPHY_PC_INIT_CAL_ERROR_P0_ERROR_RANK_PAIR, CEN_MBA_DDRPHY_PC_INIT_CAL_ERROR_P0_ERROR_RANK_PAIR_LEN>())
         {
             io_status = MSS_INIT_CAL_FAIL;
-
+            l_training_error = true;
             FAPI_ASSERT(!l_cal_error_buffer_64.getBit<CEN_MBA_DDRPHY_PC_INIT_CAL_ERROR_P0_ERROR_WR_LEVEL>(),
                         fapi2::CEN_MSS_DRAMINIT_TRAINING_WR_LVL_ERROR().
                         set_MBA_POSITION(l_mbaPosition).
@@ -697,6 +699,7 @@ extern "C" {
                 // DQS Alignment Work Around:
                 if (io_dqs_try < MAX_DQS_RETRY)
                 {
+                    l_training_error = false;
                     ++io_dqs_try;
                     --io_cur_cal_step;
                     FAPI_INF( "+++ DQS Alignment recovery attempt %d on %s port: %d rank group: %d! +++", io_dqs_try, mss::c_str(i_target),
@@ -721,6 +724,7 @@ extern "C" {
                 } // if dqs_try < max
                 else
                 {
+                    l_training_error = true;
                     FAPI_ASSERT(false,
                                 fapi2::CEN_MSS_DRAMINIT_TRAINING_DQS_ALIGNMENT_ERROR().
                                 set_TARGET_MBA_ERROR(i_target).
@@ -732,6 +736,7 @@ extern "C" {
                 }
             } // if getBit<50>
 
+            l_training_error = true;
             FAPI_ASSERT(!l_cal_error_buffer_64.getBit<CEN_MBA_DDRPHY_PC_INIT_CAL_ERROR_P0_ERROR_RDCLK_ALIGN>(),
                         fapi2::CEN_MSS_DRAMINIT_TRAINING_RD_CLK_SYS_CLK_ALIGNMENT_ERROR().
                         set_TARGET_MBA_ERROR(i_target).
@@ -837,7 +842,26 @@ extern "C" {
             io_status = MSS_INIT_CAL_PASS;
         }
 
+        return fapi2::FAPI2_RC_SUCCESS;
     fapi_try_exit:
+#ifdef __HOSTBOOT_MODULE
+
+        // If we took a training fail, log it as recovered - memdiags will sort it out
+        if(l_training_error)
+        {
+            auto l_temp_rc = fapi2::current_err;
+            fapi2::logError(l_temp_rc, fapi2::FAPI2_ERRL_SEV_RECOVERED);
+            fapi2::current_err = fapi2::FAPI2_RC_SUCCESS;
+        }
+
+#else
+
+        if(l_training_error)
+        {
+            FAPI_ERR("%s error was caused by a training error", mss::c_str(i_target));
+        }
+
+#endif
         return fapi2::current_err;
     }
 
diff --git a/src/import/chips/centaur/procedures/hwp/memory/p9c_mss_generic_shmoo.C b/src/import/chips/centaur/procedures/hwp/memory/p9c_mss_generic_shmoo.C
index e22376c1f..e1d70e29d 100755
--- a/src/import/chips/centaur/procedures/hwp/memory/p9c_mss_generic_shmoo.C
+++ b/src/import/chips/centaur/procedures/hwp/memory/p9c_mss_generic_shmoo.C
@@ -542,6 +542,20 @@ extern "C"
         }
 
         //////////////// changed the check condition ... The error call out need to gard the dimm=l_faulted_dimm(0 or 1) //// port=l_faulted_port(0 or 1) target=i_target ...
+
+#ifdef __HOSTBOOT_MODULE
+        FAPI_ASSERT_NOEXIT(!l_memory_health,
+                           fapi2::CEN_MSS_GENERIC_SHMOO_MCBIST_FAILED().
+                           set_MBA_TARGET(i_target).
+                           set_MBA_PORT_NUMBER(l_faulted_port).
+                           set_MBA_DIMM_NUMBER(l_faulted_dimm),
+                           "generic_shmoo:sanity_check failed !! MCBIST failed on %s initial run , memory is not in good state needs investigation port=%d rank=%d dimm=%d",
+                           mss::c_str(i_target),
+                           l_faulted_port,
+                           l_faulted_rank,
+                           l_faulted_dimm);
+
+#else
         FAPI_ASSERT(!l_memory_health,
                     fapi2::CEN_MSS_GENERIC_SHMOO_MCBIST_FAILED().
                     set_MBA_TARGET(i_target).
@@ -553,6 +567,8 @@ extern "C"
                     l_faulted_rank,
                     l_faulted_dimm);
 
+#endif
+
     fapi_try_exit:
         return fapi2::current_err;
     }
diff --git a/src/import/chips/centaur/procedures/xml/error_info/p9c_memory_mss_generic_shmoo_errors.xml b/src/import/chips/centaur/procedures/xml/error_info/p9c_memory_mss_generic_shmoo_errors.xml
index e84b96714..2bd50d9f2 100644
--- a/src/import/chips/centaur/procedures/xml/error_info/p9c_memory_mss_generic_shmoo_errors.xml
+++ b/src/import/chips/centaur/procedures/xml/error_info/p9c_memory_mss_generic_shmoo_errors.xml
@@ -5,7 +5,7 @@
 <!--                                                                        -->
 <!-- OpenPOWER HostBoot Project                                             -->
 <!--                                                                        -->
-<!-- Contributors Listed Below - COPYRIGHT 2016,2017                        -->
+<!-- Contributors Listed Below - COPYRIGHT 2016,2018                        -->
 <!-- [+] International Business Machines Corp.                              -->
 <!--                                                                        -->
 <!--                                                                        -->
@@ -50,13 +50,5 @@
     </childTargets>
     <priority>HIGH</priority>
   </callout>
-  <deconfigure>
-    <childTargets>
-      <parent>MBA_TARGET</parent>
-      <childType>TARGET_TYPE_DIMM</childType>
-      <childPort>MBA_PORT_NUMBER</childPort>
-      <childNumber>MBA_DIMM_NUMBER</childNumber>
-    </childTargets>
-  </deconfigure>
 </hwpError>
 </hwpErrors>