HTMGT: Prevent communication with OCC during OCC reset if comm not established

- during OCC reset, prevent sending poll/reset prep to OCC if communications
has never been successfully completed.
- fix elog leak

Change-Id: Ib154718834995401b585d584b8e2686c3b468873
CQ: SW390460
Reviewed-on: http://ralgit01.raleigh.ibm.com/gerrit1/42495
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: Martha Broyles <mbroyles@us.ibm.com>
Reviewed-by: Sheldon R. Bailey <baileysh@us.ibm.com>
Reviewed-by: Nicholas E. Bofferding <bofferdn@us.ibm.com>
Reviewed-by: Daniel M. Crowell <dcrowell@us.ibm.com>

4 files changed, 56 insertions, 21 deletions

diff --git a/src/usr/htmgt/htmgt.C b/src/usr/htmgt/htmgt.C
index 7107d3e44..d7675bb56 100644
--- a/src/usr/htmgt/htmgt.C
+++ b/src/usr/htmgt/htmgt.C
@@ -56,6 +56,7 @@ namespace HTMGT
                  i_startCompleted, i_failedOccTarget);
         errlHndl_t l_err = nullptr;
         uint32_t l_huid = 0;
+        bool skip_comm = true;
         if (i_failedOccTarget)
         {
             l_huid = TARGETING::get_huid(i_failedOccTarget);
@@ -116,6 +117,7 @@ namespace HTMGT
                                     ERRORLOG::errlCommit(l_err, HTMGT_COMP_ID);
                                 }
                             }
+                            skip_comm = false;
 
                             // Send ALL config data
                             sendOccConfigData();
@@ -193,7 +195,11 @@ namespace HTMGT
                              "Attempting OCC Reset",
                              l_err->reasonCode());
                 TMGT_INF("processOccStartStatus: Calling resetOccs");
-                errlHndl_t err2 = OccManager::resetOccs(nullptr);
+                // Reset ALL OCCs, don't skip incrementing reset count, and
+                // if comm has not been established, don't try to talk to OCCs
+                errlHndl_t err2 = OccManager::resetOccs(nullptr,
+                                                        false,
+                                                        skip_comm);
                 if(err2)
                 {
                     TMGT_ERR("OccManager::resetOccs failed with 0x%04X",
diff --git a/src/usr/htmgt/htmgt_occ.C b/src/usr/htmgt/htmgt_occ.C
index 0d47926b6..c6d230fad 100644
--- a/src/usr/htmgt/htmgt_occ.C
+++ b/src/usr/htmgt/htmgt_occ.C
@@ -206,24 +206,28 @@ namespace HTMGT
             cmdData[1] = OCC_RESET_FAIL_OTHER_OCC;
         }
 
-        OccCmd cmd(this, OCC_CMD_RESET_PREP, sizeof(cmdData), cmdData);
-        err = cmd.sendOccCmd();
-        if(err)
+        if (iv_commEstablished)
         {
-            // log error and keep going
-            TMGT_ERR("OCC::resetPrep: OCC%d resetPrep failed with rc = 0x%04x",
-                     iv_instance,
-                     err->reasonCode());
+            OccCmd cmd(this, OCC_CMD_RESET_PREP, sizeof(cmdData), cmdData);
+            err = cmd.sendOccCmd();
+            if(err)
+            {
+                // log error and keep going
+                TMGT_ERR("OCC::resetPrep: OCC%d resetPrep failed, rc=0x%04x",
+                         iv_instance,
+                         err->reasonCode());
 
-            ERRORLOG::errlCommit(err, HTMGT_COMP_ID);
-        }
+                ERRORLOG::errlCommit(err, HTMGT_COMP_ID);
+            }
 
-        // poll and flush error logs from OCC - Check Ex return code
-        err = pollForErrors(true);
-        if(err)
-        {
-            ERRORLOG::errlCommit(err, HTMGT_COMP_ID);
+            // poll and flush error logs from OCC - Check Ex return code
+            err = pollForErrors(true);
+            if(err)
+            {
+                ERRORLOG::errlCommit(err, HTMGT_COMP_ID);
+            }
         }
+        // else comm to OCC has not been established yet
 
         return atThreshold;
     }
@@ -841,8 +845,11 @@ namespace HTMGT
 
                 if (false == i_skipComm)
                 {
-                    // Send poll cmd to all OCCs to establish comm
-                    err = _sendOccPoll(false,nullptr);
+                    // Send poll cmd to all OCCs
+                    err = _sendOccPoll(false, // don't flush errors
+                                       nullptr, // send to all OCCs
+                                       true); // only poll if communications
+                    // has been established
                     if (err)
                     {
                         TMGT_ERR("_resetOccs: Poll OCCs failed.");
diff --git a/src/usr/htmgt/htmgt_occ.H b/src/usr/htmgt/htmgt_occ.H
index 1c33cbbe4..73f081411 100644
--- a/src/usr/htmgt/htmgt_occ.H
+++ b/src/usr/htmgt/htmgt_occ.H
@@ -710,9 +710,13 @@ namespace HTMGT
                                   bool i_skipComm = false);
 
             /** See sendOccPoll() above */
+            /* @param[in] i_onlyIfEstablished: If true, only send poll if
+             *                 communications has already been established
+             *                 with that OCC */
             errlHndl_t
                 _sendOccPoll(const bool i_flushAllErrors,
-                            TARGETING::Target * i_occTarget);
+                             TARGETING::Target * i_occTarget,
+                             const bool i_onlyIfEstablished = false);
 
             /** See updateSafeModeReason() above */
             void _updateSafeModeReason(uint32_t i_src,
diff --git a/src/usr/htmgt/htmgt_poll.C b/src/usr/htmgt/htmgt_poll.C
index a3d5d2238..2457db373 100644
--- a/src/usr/htmgt/htmgt_poll.C
+++ b/src/usr/htmgt/htmgt_poll.C
@@ -44,9 +44,10 @@ namespace HTMGT
 {
 
     errlHndl_t OccManager::_sendOccPoll(const bool i_flushAllErrors,
-                                        TARGETING::Target * i_occTarget)
+                                        TARGETING::Target * i_occTarget,
+                                        const bool onlyIfEstablished)
     {
-        errlHndl_t l_err = NULL;
+        errlHndl_t l_err = nullptr;
 
         TMGT_INF("sendOccPoll(flush=%c)", i_flushAllErrors?'y':'n');
 
@@ -54,7 +55,24 @@ namespace HTMGT
         {
             if(NULL == i_occTarget || l_occ->iv_target == i_occTarget)
             {
-                l_err = l_occ->pollForErrors(i_flushAllErrors);
+                if ((l_occ->iv_commEstablished) ||
+                    (onlyIfEstablished == false))
+                {
+                    errlHndl_t poll_err=l_occ->pollForErrors(i_flushAllErrors);
+                    if (poll_err != nullptr)
+                    {
+                        if (l_err == nullptr)
+                        {
+                            // Only return 1st error (continue to poll others)
+                            l_err = poll_err;
+                            poll_err = nullptr;
+                        }
+                        else
+                        {
+                            ERRORLOG::errlCommit(poll_err, HTMGT_COMP_ID);
+                        }
+                    }
+                }
             }
         }