Re-read SBE doorbell register in simics if PSU interrupt is found

We use PSU interrupts to signal between the host and the SBE if there
is a PSU message available for each other. Bit 0 in the doorbell reg
tells hostboot that the SBE has put a message in the communication
registers. Once hostboot sees this it is responsible for clearing that
bit so the SBE knows it can send another msg. We are seeing a problem
in simics where hostboot is not clearing this bit, we think its because
when hostboot checks if the bit is set, the simics has not set the value
yet, so hostboot never clears it. This change adds a retry in simics so
we will re-read the doorbell register if we take a PSU interrupt and
the doorbell register is saying its empty. We are hoping this wait+retry
will give the simics models enough time to catch up if this is indeed
the problem. This issue has not been observed in real HW.

Change-Id: I85c63fe545d473d55f6ec0fb554b6c760ae5cbf8
CQ: SW444734
Reviewed-on: http://rchgit01.rchland.ibm.com/gerrit1/66622
Reviewed-by: Daniel M. Crowell <dcrowell@us.ibm.com>
Tested-by: Jenkins Server <pfd-jenkins+hostboot@us.ibm.com>
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>
Reviewed-by: William G. Hoffa <wghoffa@us.ibm.com>

1 files changed, 23 insertions, 1 deletions

diff --git a/src/usr/sbeio/sbe_psudd.C b/src/usr/sbeio/sbe_psudd.C
index ab53d0d3e..3e23c9642 100644
--- a/src/usr/sbeio/sbe_psudd.C
+++ b/src/usr/sbeio/sbe_psudd.C
@@ -54,6 +54,7 @@
 #include <sys/time.h>
 #include <errl/errludprintk.H>
 #include <vfs/vfs.H> // module_is_loaded
+#include <util/misc.H>
 
 trace_desc_t* g_trac_sbeio;
 TRAC_INIT(&g_trac_sbeio, SBEIO_COMP_NAME, 6*KILOBYTE, TRACE::BUFFER_SLOW);
@@ -201,7 +202,7 @@ void SbePsu::msgHandler()
                     {
                         //Handle the interrupt message -- pass the PIR of the
                         // proc causing the interrupt
-                        SBE_TRACD("SbePsu::msgHandler got MSG_INTR message");
+                        SBE_TRACF("SbePsu::msgHandler got MSG_INTR message");
                         l_err = handleInterrupt(msg->data[1]);
 
                         if (l_err)
@@ -492,6 +493,7 @@ errlHndl_t SbePsu::handleInterrupt(PIR_t i_pir)
     errlHndl_t errl = nullptr;
     SBE_TRACD(ENTER_MRK "SbePsu::handleInterrupt");
     bool l_responseAvailable = false;
+    bool l_simicsRunning = Util::isSimicsRunning();
 
     do
     {
@@ -525,6 +527,20 @@ errlHndl_t SbePsu::handleInterrupt(PIR_t i_pir)
         if (errl)
         { break; }
 
+
+        if ( l_simicsRunning &&
+             HOST_RESPONSE_WAITING != (l_doorbellVal & HOST_RESPONSE_WAITING) )
+        {
+            // wait a second and try again
+            nanosleep(1,0);
+
+            //read the door bell again to see if the simics model caught up
+            errl = readScom(l_intrChip,PSU_HOST_DOORBELL_REG_RW,&l_doorbellVal);
+            if (errl)
+            { break; }
+
+        }
+
         if (l_doorbellVal & HOST_RESPONSE_WAITING)
         {
             //In this case the doorbell reg indicated a response is
@@ -552,6 +568,12 @@ errlHndl_t SbePsu::handleInterrupt(PIR_t i_pir)
             if (errl)
             { break; }
         }
+        else if(l_simicsRunning)
+        {
+            SBE_TRACF(ENTER_MRK "SbePsu::handleInterrupt interrupt found but no Doorbell is set 0x%llx", l_doorbellVal);
+            // If we are in simics we want to break out here
+            MAGIC_INSTRUCTION(MAGIC_BREAK_ON_ERROR);
+        }
 
         //Clear the rest of the PSU Scom Reg Interrupt Status register
         //  This clears the PSU interrupt condition so the PSU interrupt