void ServiceDataCollector::SetCallout( PRDcallout mru,
PRDpriority priority )
{
bool found = false;
if ( PRDcalloutData::TYPE_TARGET == mru.getType() )
{
// Ensuring target is not NULL
if ( NULL == mru.getTarget() )
{
PRDF_ERR( "[ServiceDataCollector::SetCallout] "
"skipping NULL callout" );
return;
}
}
for ( SDC_MRU_LIST::iterator i = xMruList.begin();
i != xMruList.end() && found == false; ++i )
{
if ( i->callout == mru )
{
found = true;
if ( priority > i->priority )
{
i->priority = priority;
}
}
}
if ( found == false )
{
xMruList.push_back( SdcCallout(mru, priority) );
}
}
errlHndl_t ErrDataService::GenerateSrcPfa( ATTENTION_TYPE i_attnType,
ServiceDataCollector & io_sdc,
bool & o_initiateHwudump,
TargetHandle_t & o_dumpTrgt,
errlHndl_t & o_dumpErrl,
uint32_t & o_dumpErrlActions)
{
#define PRDF_FUNC "[ErrDataService::GenerateSrcPfa] "
o_initiateHwudump = false;
o_dumpTrgt = NULL;
o_dumpErrl = NULL;
o_dumpErrlActions = 0;
// First, check if an error log should be committed. Note that there should
// always be an error log if there was a system or unit checkstop.
if ( io_sdc.queryDontCommitErrl() &&
MACHINE_CHECK != i_attnType && !io_sdc.IsUnitCS() )
{
// User did not want this error log committed. No need to continue. So
// delete it and exit.
delete iv_errl; iv_errl = NULL;
return NULL;
}
#ifdef __HOSTBOOT_MODULE
using namespace ERRORLOG;
using namespace HWAS;
#else
uint8_t sdcSaveFlags = SDC_NO_SAVE_FLAGS;
size_t sz_uint8 = sizeof(uint8_t);
#endif
epubProcedureID thisProcedureID;
bool ForceTerminate = false;
bool iplDiagMode = false;
++iv_serviceActionCounter;
uint16_t PRD_Reason_Code = 0;
//**************************************************************
// Callout loop to set up Reason code and SRC word 9
//**************************************************************
// Must go thru callout list to look for RIOPORT procedure callouts,
// since they require the port info to be in SRC Word 9
bool HW = false;
bool SW = false;
bool SW_High = false;
bool SecondLevel = false;
uint32_t SrcWord7 = 0;
uint32_t SrcWord9 = 0;
// Should not gard hardware if there is a hardware callout at LOW priority
// and a symbolic FRU indicating a possibility of a software error at MED or
// HIGH priority.
bool sappSwNoGardReq = false, sappHwNoGardReq = false;
const SDC_MRU_LIST & mruList = io_sdc.getMruList();
int32_t calloutsPlusDimms = mruList.size();
for ( SDC_MRU_LIST::const_iterator it = mruList.begin();
it < mruList.end(); ++it )
{
PRDcallout thiscallout = it->callout;
if ( PRDcalloutData::TYPE_SYMFRU == thiscallout.getType() )
{
if ( (SP_CODE == thiscallout.flatten()) ||
(SYS_SW_CODE == thiscallout.flatten()) )
{
SW = true;
if ( MRU_LOW != it->priority )
{
sappSwNoGardReq = true;
}
if ( MRU_MED == it->priority )
{
SW_High = true;
}
}
else if ( LEVEL2_SUPPORT == thiscallout.flatten())
{
SecondLevel = true;
if ( MRU_LOW != it->priority )
{
sappSwNoGardReq = true;
}
}
}
else if ( PRDcalloutData::TYPE_MEMMRU == thiscallout.getType() )
{
MemoryMru memMru (thiscallout.flatten());
SrcWord9 = memMru.toUint32(); // Get MemMru value
TargetHandleList partList = memMru.getCalloutList();
uint32_t partCount = partList.size();
calloutsPlusDimms = calloutsPlusDimms + partCount -1;
HW = true; //hardware callout
if ( MRU_LOW == it->priority )
{
sappHwNoGardReq = true;
}
}
else // PRDcalloutData::TYPE_TARGET
{
HW = true; // Hardware callout
// Determines if all the hardware callouts have low priority.
if ( MRU_LOW == it->priority )
{
sappHwNoGardReq = true;
}
}
}
////////////////////////////////////////////////////////////////
//Set the PRD Reason Code based on the flags set in the above callout loop.
////////////////////////////////////////////////////////////////
if (HW == true && SW == true)
{
if (SW_High == true)
PRD_Reason_Code = PRDF_DETECTED_FAIL_SOFTWARE_PROBABLE;
else
PRD_Reason_Code = PRDF_DETECTED_FAIL_HARDWARE_PROBABLE;
}
else if (HW == true && SW == false && SecondLevel == true)
PRD_Reason_Code = PRDF_DETECTED_FAIL_HARDWARE_PROBABLE;
else if (HW == true && SW == false && SecondLevel == false)
PRD_Reason_Code = PRDF_DETECTED_FAIL_HARDWARE;
else if (HW == false && SW == true)
PRD_Reason_Code = PRDF_DETECTED_FAIL_SOFTWARE;
else
{
// If we get here both HW and SW flags were false. Callout may be
// Second Level Support only, or a procedure not checked in the SW
// flag code.
PRD_Reason_Code = PRDF_DETECTED_FAIL_HARDWARE_PROBABLE;
}
SrcWord7 = io_sdc.getPrimaryAttnType() << 8;
SrcWord7 |= io_sdc.getSecondaryAttnType();
//--------------------------------------------------------------------------
// Check for IPL Diag Mode
//--------------------------------------------------------------------------
#if defined(__HOSTBOOT_MODULE) && !defined(__HOSTBOOT_RUNTIME)
iplDiagMode = PlatServices::isInMdiaMode();
#endif
//**************************************************************
// Update Error Log with SRC
//**************************************************************
ErrorSignature * esig = io_sdc.GetErrorSignature();
updateSrc( esig->getChipId(), SrcWord7, esig->getSigId(),
SrcWord9, PRD_Reason_Code);
//**************************************************************
// Add SDC Capture data to Error Log User Data here only if
// there are 4 or more callouts,
// (including Dimm callouts in the MemoryMru).
//**************************************************************
bool capDataAdded = false;
if (calloutsPlusDimms > 3)
{
AddCapData( io_sdc.GetCaptureData(), iv_errl );
AddCapData( io_sdc.getTraceArrayData(), iv_errl );
capDataAdded = true;
}
//--------------------------------------------------------------------------
// Set the error log severity and get the error log action flags.
//--------------------------------------------------------------------------
// Let's assume the default is the action for a system checkstop.
#ifdef __HOSTBOOT_MODULE
errlSeverity_t errlSev = ERRL_SEV_UNRECOVERABLE;
#else
errlSeverity errlSev = ERRL_SEV_UNRECOVERABLE;
#endif
uint32_t errlAct = ERRL_ACTION_SA | // Service action required.
ERRL_ACTION_REPORT | // Report to HMC and hypervisor.
ERRL_ACTION_CALL_HOME; // Call home.
if ( MACHINE_CHECK != i_attnType ) // Anything other that a system checkstop
{
if ( io_sdc.queryServiceCall() ) // still a serviceable event
{
errlSev = ERRL_SEV_PREDICTIVE;
}
else // not a serviceable event
{
errlSev = io_sdc.queryLogging()
? ERRL_SEV_RECOVERED // should still be logged
: ERRL_SEV_INFORMATIONAL; // can be ignored
errlAct = ERRL_ACTION_HIDDEN;
}
}
// This needs to be done after setting the SRCs otherwise it will be
// overridden.
iv_errl->setSev( errlSev );
// Add procedure callout for SUE attentions. The intent is to make sure the
// customer looks for other service actions before replacing parts for this
// attention.
if ( io_sdc.IsSUE() )
{
PRDF_HW_ADD_PROC_CALLOUT( SUE_PREV_ERR, MRU_HIGH, iv_errl, errlSev );
}
//--------------------------------------------------------------------------
// Get the global gard policy.
//--------------------------------------------------------------------------
HWAS::GARD_ErrorType gardPolicy = HWAS::GARD_NULL;
// Gard only if the error is a serviceable event.
if ( io_sdc.queryServiceCall() )
{
// We will not Resource Recover on a checkstop attention.
gardPolicy = ( MACHINE_CHECK == i_attnType ) ? HWAS::GARD_Fatal
: HWAS::GARD_Predictive;
}
if ( io_sdc.IsSUE() && ( MACHINE_CHECK == i_attnType ) )
{
// If we are logging an error for an SUE consumed, we should not
// perform any GARD here. Appropriate resources should have already
// been GARDed for the original UE.
gardPolicy = HWAS::GARD_NULL;
}
// Apply special policies for OPAL.
if ( isHyprConfigOpal() && // OPAL is used
!isMfgAvpEnabled() && !isMfgHdatAvpEnabled() ) // No AVPs running
{
// OPAL has requested that we disable garding for predictive errors
// found at runtime.
if ( HWAS::GARD_Predictive == gardPolicy )
{
#if !defined(__HOSTBOOT_MODULE) // FSP only
if ( isHyprRunning() ) gardPolicy = HWAS::GARD_NULL;
#elif defined(__HOSTBOOT_RUNTIME) // HBRT only
gardPolicy = HWAS::GARD_NULL;
#endif
}
// OPAL has requested that we diable garding for fatal errors (system
// checkstops) that could have been caused by a software generated
// attention at runtime. This will be determined if there is a software
// callout with higher priority than a hardware callout.
else if ( HWAS::GARD_Fatal == gardPolicy &&
sappSwNoGardReq && sappHwNoGardReq ) // Gard requirements met
{
#if !defined(__HOSTBOOT_MODULE) // FSP only
if ( isHyprRunning() ) gardPolicy = HWAS::GARD_NULL;
#elif !defined(__HOSTBOOT_RUNTIME) // Hostboot only
#ifdef CONFIG_ENABLE_CHECKSTOP_ANALYSIS
// Checkstop analysis is only done at the beginning of the IPL,
// regardless if the checkstop actually occurred during the IPL
// or at runtime. We will need to check the IPL state in FIR
// data to determine when the checkstop occurred.
// Get access to IPL state info from the FIR data in the PNOR.
if ( !(PnorFirDataReader::getPnorFirDataReader().isIplState()) )
gardPolicy = HWAS::GARD_NULL;
#endif
#endif
}
}
//--------------------------------------------------------------------------
// Get the global deconfig policy (must be done after setting gard policy).
//--------------------------------------------------------------------------
HWAS::DeconfigEnum deconfigPolicy = HWAS::NO_DECONFIG;
bool deferDeconfig = false;
if ( HWAS::GARD_NULL != gardPolicy )
{
#if !defined(__HOSTBOOT_MODULE) // FSP only
// Change the deconfig state based the gard type. This is only required
// to control what the FSP does during the reconfig loop.
deconfigPolicy = HWSV::SvrError::isInHwReconfLoop() ? HWAS::DECONFIG
: HWAS::NO_DECONFIG;
#elif !defined(__HOSTBOOT_RUNTIME) // Hostboot only
// Deferred Deconfig should be used throughout all of Hostboot (both
// checkForIplAttns() and MDIA).
deconfigPolicy = HWAS::DECONFIG;
deferDeconfig = true;
#endif
}
//--------------------------------------------------------------------------
// Get the HCDB diagnostics policy.
//--------------------------------------------------------------------------
HWSV::hwsvDiagUpdate l_diagUpdate = HWSV::HWSV_DIAG_NEEDED;
if ( ERRL_ACTION_HIDDEN == errlAct )
{
// Diagnostics is not needed in the next IPL cycle for non-visible logs.
l_diagUpdate = HWSV::HWSV_DIAG_NOT_NEEDED;
}
//--------------------------------------------------------------------------
// Initialize the PFA data
//--------------------------------------------------------------------------
PfaData pfaData;
initPfaData( io_sdc, i_attnType, deferDeconfig, errlAct, errlSev,
gardPolicy, pfaData, o_dumpTrgt );
//--------------------------------------------------------------------------
// Add each mru/callout to the error log.
//--------------------------------------------------------------------------
for ( SDC_MRU_LIST::const_iterator it = mruList.begin();
it < mruList.end(); ++it )
{
PRDcallout thiscallout = it->callout;
PRDpriority thispriority = it->priority;
// Use the global gard/deconfig policies as default.
HWAS::GARD_ErrorType thisGard = gardPolicy;
HWAS::DeconfigEnum thisDeconfig = deconfigPolicy;
// Change the gard/deconfig actions if this MRU should not be garded.
if ( NO_GARD == it->gardState )
{
thisGard = HWAS::GARD_NULL;
thisDeconfig = HWAS::NO_DECONFIG;
}
// Add the callout to the PFA data
addCalloutToPfaData( pfaData, thiscallout, thispriority, thisGard );
// Add callout based on callout type.
if( PRDcalloutData::TYPE_TARGET == thiscallout.getType() )
{
PRDF_HW_ADD_CALLOUT(thiscallout.getTarget(),
thispriority,
thisDeconfig,
iv_errl,
thisGard,
errlSev,
l_diagUpdate);
}
else if(PRDcalloutData::TYPE_PROCCLK == thiscallout.getType() ||
PRDcalloutData::TYPE_PCICLK == thiscallout.getType())
{
PRDF_ADD_CLOCK_CALLOUT(iv_errl,
thiscallout.getTarget(),
thiscallout.getType(),
thispriority,
thisDeconfig,
thisGard);
}
else if ( PRDcalloutData::TYPE_MEMMRU == thiscallout.getType() )
{
MemoryMru memMru (thiscallout.flatten());
TargetHandleList partList = memMru.getCalloutList();
for ( TargetHandleList::iterator it = partList.begin();
it != partList.end(); it++ )
{
PRDF_HW_ADD_CALLOUT( *it,
thispriority,
thisDeconfig,
iv_errl,
thisGard,
errlSev,
l_diagUpdate );
}
}
else if ( PRDcalloutData::TYPE_SYMFRU == thiscallout.getType() )
{
thisProcedureID = epubProcedureID(thiscallout.flatten());
PRDF_DTRAC( PRDF_FUNC "thisProcedureID: %x, thispriority: %x, "
"errlSev: %x", thisProcedureID, thispriority,errlSev );
PRDF_HW_ADD_PROC_CALLOUT(thisProcedureID,
thispriority,
iv_errl,
errlSev);
// Use the flags set earlier to determine if the callout is just
// Software (SP code or Phyp Code). Add a Second Level Support
// procedure callout Low, for this case.
if (HW == false && SW == true && SecondLevel == false)
{
PRDF_DTRAC( PRDF_FUNC "thisProcedureID= %x, thispriority=%x, "
"errlSev=%x", LEVEL2_SUPPORT, MRU_LOW, errlSev );
PRDF_HW_ADD_PROC_CALLOUT( LEVEL2_SUPPORT, MRU_LOW, iv_errl,
errlSev );
SecondLevel = true;
}
}
}
// Send the dynamic memory Dealloc message for DIMMS for Predictive
// callouts.
// We can not check for ERRL severity here as there are some cases
// e.g. DD02 where we create a Predictive error log but callouts
// are not predictive.
if ( HWAS::GARD_Predictive == gardPolicy )
{
deallocateDimms( mruList );
}
//**************************************************************
// Check for Terminating the system for non mnfg conditions.
//**************************************************************
ForceTerminate = checkForceTerm( io_sdc, o_dumpTrgt, pfaData );
//*************************************************************
// Check for Manufacturing Mode terminate here and then do
// the needed overrides on ForceTerminate flag.
//*************************************************************
if ( PlatServices::mnfgTerminate() && !ForceTerminate )
{
ForceTerminate = true;
if ( !((errlSev == ERRL_SEV_RECOVERED) ||
(errlSev == ERRL_SEV_INFORMATIONAL)) &&
iplDiagMode &&
!HW )
{
//Terminate in Manufacturing Mode, in IPL mode, for visible log, with no HW callouts.
PRDF_SRC_WRITE_TERM_STATE_ON(iv_errl, SRCI_TERM_STATE_MNFG);
}
// Do not terminate if recoverable or informational.
// Do not terminate if deferred deconfig.
else if ( deferDeconfig ||
(errlSev == ERRL_SEV_RECOVERED ) ||
(errlSev == ERRL_SEV_INFORMATIONAL) )
{
ForceTerminate = false;
errlAct |= ERRL_ACTION_DONT_TERMINATE;
}
else
{
PRDF_SRC_WRITE_TERM_STATE_ON(iv_errl, SRCI_TERM_STATE_MNFG);
}
pfaData.errlActions = errlAct;
}
// Needed to move the errl add user data sections here because of some updates
// of the data required in the Aysnc section for the SMA dual reporting fix.
//**************************************************************
// Add the PFA data to Error Log User Data
//**************************************************************
UtilMem l_membuf;
l_membuf << pfaData;
PRDF_ADD_FFDC( iv_errl, (const char*)l_membuf.base(), l_membuf.size(),
ErrlVer1, ErrlSectPFA5_1 );
//**************************************************************
// Add SDC Capture data to Error Log User Data
//**************************************************************
// Pulled some code out to incorporate into AddCapData
// Check to make sure Capture Data wasn't added earlier.
if (!capDataAdded)
{
AddCapData( io_sdc.GetCaptureData(), iv_errl );
AddCapData( io_sdc.getTraceArrayData(), iv_errl );
}
//**************************************************************************
// Add extended MemoryMru error log sections (if needed).
//**************************************************************************
for ( SDC_MRU_LIST::const_iterator it = mruList.begin();
it < mruList.end(); ++it )
{
// Operate only on MemoryMru callouts.
if ( PRDcalloutData::TYPE_MEMMRU != it->callout.getType() ) continue;
/* TODO RTC 136125
// Only add single DIMM callouts. Otherwise, the parsed data is
// redundant.
MemoryMru memMru ( it->callout.flatten() );
if ( !memMru.getSymbol().isValid() ) continue;
// Add the MemoryMru to the capture data.
CenMbaCaptureData::addExtMemMruData( memMru, iv_errl );
*/
}
//**************************************************************************
// Additional FFDC
//**************************************************************************
// For OP checkstop analysis, add a string indicating a system checkstop
// occurred and when. This will be printed out in the console traces along
// with the error log.
#if defined(__HOSTBOOT_MODULE) && !defined(__HOSTBOOT_RUNTIME) // IPL only
#ifdef CONFIG_ENABLE_CHECKSTOP_ANALYSIS
if ( MACHINE_CHECK == i_attnType )
{
const char * const str =
PnorFirDataReader::getPnorFirDataReader().isIplState()
? "System checkstop occurred during IPL on previous boot"
: "System checkstop occurred during runtime on previous boot";
ErrlUserDetailsString(str).addToLog(iv_errl);
}
#endif
#endif
// Collect PRD traces.
// NOTE: Each line of a trace is on average 36 bytes so 768 bytes should get
// us around 21 lines of trace output.
PRDF_COLLECT_TRACE(iv_errl, 768);
//**************************************************************
// Commit the error log.
// This will also perform Gard and Deconfig actions.
// Do the Unit Dumps if needed.
//**************************************************************
// Add the MNFG trace information.
MnfgTrace( io_sdc.GetErrorSignature(), pfaData );
// If this is not a terminating condition, commit the error log. If the
// error log is not committed, the error log will be passed back to
// PRDF::main() and eventually ATTN.
if ( MACHINE_CHECK != pfaData.priAttnType && !ForceTerminate &&
!pfaData.TERMINATE )
{
// Handle any unit checkstop conditions, if needed (i.e. runtime
// deconfiguration, dump/FFDC collection, etc.
if ( io_sdc.IsUnitCS() && !io_sdc.IsUsingSavedSdc() )
{
handleUnitCS( io_sdc, o_dumpTrgt, o_initiateHwudump );
}
if ( true == o_initiateHwudump )
{
// the dump log will be deleted later in PRDF::main
// after the hwudump is initiated there.
o_dumpErrl = iv_errl;
iv_errl = NULL;
o_dumpErrlActions = errlAct;
PRDF_TRAC( PRDF_FUNC "for target: 0x%08x, i_errl: 0x%08x, "
"i_errlActions: 0x%08x", getHuid(o_dumpTrgt),
ERRL_GETRC_SAFE(o_dumpErrl), o_dumpErrlActions );
}
else
{
// Commit the error log.
commitErrLog( iv_errl, pfaData );
}
}
#ifndef __HOSTBOOT_MODULE
errlHndl_t reg_errl = UtilReg::read ("prdf/RasServices", &sdcSaveFlags, sz_uint8);
if (reg_errl)
{
PRDF_ERR( PRDF_FUNC "Failure in SDC Sync flag Registry read" );
PRDF_COMMIT_ERRL(reg_errl, ERRL_ACTION_REPORT);
}
else
{
//Turn off indicator that there is saved Sdc Analysis info
sdcSaveFlags &= ( ~SDC_ANALYSIS_SAVE_FLAG );
reg_errl = UtilReg::write ("prdf/RasServices", &sdcSaveFlags, sz_uint8);
if (reg_errl)
{
PRDF_ERR( PRDF_FUNC "Failure in SDC Sync flag Registry write" );
PRDF_COMMIT_ERRL(reg_errl, ERRL_ACTION_REPORT);
}
}
#endif
PRDF_INF( PRDF_FUNC "PRD called to analyze an error: 0x%08x 0x%08x",
esig->getChipId(), esig->getSigId() );
// Reset iv_errl to NULL. This is done to catch logical bug in our code.
// It enables us to assert in createInitialErrl function if iv_errl is
// not NULL which should catch any logical bug in initial stages of testing.
errlHndl_t o_errl = iv_errl;
iv_errl = NULL;
return o_errl;
#undef PRDF_FUNC
}