int32_t analyzeIplCEStats( TargetHandle_t i_mba, bool &o_calloutMade )
{
#define PRDF_FUNC "PRDF::analyzeIplCEStats"
PRDF_ENTER( PRDF_FUNC"(0x%08x)", getHuid(i_mba) );
// will unlock when going out of scope
PRDF_SYSTEM_SCOPELOCK;
int32_t o_rc = SUCCESS;
o_calloutMade = false;
ExtensibleChip * mbaChip = (ExtensibleChip *)systemPtr->GetChip( i_mba );
CenMbaDataBundle * mbadb = getMbaDataBundle( mbaChip );
o_rc = mbadb->getIplCeStats()->analyzeStats( o_calloutMade );
if ( SUCCESS != o_rc )
{
PRDF_ERR( "["PRDF_FUNC"] analyzeStats() failed");
// Get user data
uint64_t ud12 = PRDF_GET_UINT64_FROM_UINT32( getHuid(i_mba), 0 );
uint64_t ud34 = PRDF_GET_UINT64_FROM_UINT32( PRDFSIG_MnfgIplFail, 0 );
// Create error log
errlHndl_t errl = new ERRORLOG::ErrlEntry(
ERRORLOG::ERRL_SEV_PREDICTIVE, // severity
PRDF_MNFG_IPL_CE_ANALYSIS, // module ID
PRDF_DETECTED_FAIL_SOFTWARE, // reason code
ud12, ud34 ); // user data 1-4
// Add 2nd level support
errl->addProcedureCallout( EPUB_PRC_LVL_SUPP, SRCI_PRIORITY_HIGH );
// Add traces
errl->collectTrace( PRDF_COMP_NAME, 512 );
// Commit the error log
ERRORLOG::errlCommit( errl, PRDF_COMP_ID );
}
PRDF_EXIT( PRDF_FUNC"(0x%08x), o_calloutMade:%u",
getHuid(i_mba), o_calloutMade );
return o_rc;
#undef PRDF_FUNC
}
/**
* @brief Plugin to mask the side effects of an RCD parity error
* @param i_mbaChip A Centaur MBA chip.
* @param i_sc The step code data struct.
* @return SUCCESS
*/
int32_t maskRcdParitySideEffects( ExtensibleChip * i_mbaChip,
STEP_CODE_DATA_STRUCT & i_sc )
{
#define PRDF_FUNC "[maskRcdParitySideEffects] "
int32_t l_rc = SUCCESS;
do
{
//use a data bundle to get the membuf chip
CenMbaDataBundle * mbadb = getMbaDataBundle( i_mbaChip );
ExtensibleChip * membChip = mbadb->getMembChip();
if (NULL == membChip)
{
PRDF_ERR(PRDF_FUNC "getMembChip() failed");
break;
}
//get the masks for each FIR
SCAN_COMM_REGISTER_CLASS * mbsFirMaskOr =
membChip->getRegister("MBSFIR_MASK_OR");
SCAN_COMM_REGISTER_CLASS * mbaCalMaskOr =
i_mbaChip->getRegister("MBACALFIR_MASK_OR");
SCAN_COMM_REGISTER_CLASS * mbaFirMaskOr =
i_mbaChip->getRegister("MBAFIR_MASK_OR");
mbaFirMaskOr->SetBit(2);
mbaCalMaskOr->SetBit(2);
mbaCalMaskOr->SetBit(17);
mbsFirMaskOr->SetBit(4);
l_rc = mbaFirMaskOr->Write();
l_rc |= mbaCalMaskOr->Write();
l_rc |= mbsFirMaskOr->Write();
if (SUCCESS != l_rc)
{
PRDF_ERR(PRDF_FUNC "MBAFIR_MASK_OR/MBACALFIR_MASK_OR/MBSFIR_MASK_OR"
" write failed for 0x%08x", i_mbaChip->GetId());
break;
}
}while(0);
return SUCCESS;
#undef PRDF_FUNC
}
int32_t CenMbaTdCtlrCommon::initialize()
{
#define PRDF_FUNC "[CenMbaTdCtlrCommon::initialize] "
int32_t o_rc = SUCCESS;
do
{
// Set iv_mbaTrgt
iv_mbaTrgt = iv_mbaChip->GetChipHandle();
// Validate iv_mbaChip.
if ( TYPE_MBA != getTargetType(iv_mbaTrgt) )
{
PRDF_ERR( PRDF_FUNC "iv_mbaChip is not TYPE_MBA" );
o_rc = FAIL; break;
}
// Set iv_membChip.
CenMbaDataBundle * mbadb = getMbaDataBundle( iv_mbaChip );
iv_membChip = mbadb->getMembChip();
if ( NULL == iv_membChip )
{
PRDF_ERR( PRDF_FUNC "getMembChip() failed" );
o_rc = FAIL; break;
}
// Set iv_mbaPos.
iv_mbaPos = getTargetPosition( iv_mbaTrgt );
if ( MAX_MBA_PER_MEMBUF <= iv_mbaPos )
{
PRDF_ERR( PRDF_FUNC "iv_mbaPos=%d is invalid", iv_mbaPos );
o_rc = FAIL; break;
}
// Set iv_x4Dimm.
iv_x4Dimm = isDramWidthX4(iv_mbaTrgt);
} while (0);
return o_rc;
#undef PRDF_FUNC
}
int32_t getDramSize( ExtensibleChip *i_mbaChip, uint8_t & o_size )
{
#define PRDF_FUNC "[MemUtils::getDramSize] "
int32_t o_rc = SUCCESS;
o_size = SIZE_2GB;
do
{
TargetHandle_t mbaTrgt = i_mbaChip->GetChipHandle();
CenMbaDataBundle * mbadb = getMbaDataBundle( i_mbaChip );
ExtensibleChip * membufChip = mbadb->getMembChip();
if ( NULL == membufChip )
{
PRDF_ERR( PRDF_FUNC "getMembChip() failed: MBA=0x%08x",
getHuid(mbaTrgt) );
o_rc = FAIL; break;
}
uint32_t pos = getTargetPosition(mbaTrgt);
const char * reg_str = (0 == pos) ? "MBA0_MBAXCR" : "MBA1_MBAXCR";
SCAN_COMM_REGISTER_CLASS * reg = membufChip->getRegister( reg_str );
o_rc = reg->Read();
if ( SUCCESS != o_rc )
{
PRDF_ERR( PRDF_FUNC "Read() failed on %s. Target=0x%08x",
reg_str, getHuid(mbaTrgt) );
break;
}
o_size = reg->GetBitFieldJustified( 6, 2 );
} while(0);
return o_rc;
#undef PRDF_FUNC
}
int32_t collectCeStats( ExtensibleChip * i_mbaChip, const CenRank & i_rank,
MaintSymbols & o_maintStats, CenSymbol & o_chipMark,
uint8_t i_thr )
{
#define PRDF_FUNC "[MemUtils::collectCeStats] "
int32_t o_rc = SUCCESS;
o_chipMark = CenSymbol(); // Initially invalid.
do
{
if ( 0 == i_thr ) // Must be non-zero
{
PRDF_ERR( PRDF_FUNC "i_thr %d is invalid", i_thr );
o_rc = FAIL; break;
}
TargetHandle_t mbaTrgt = i_mbaChip->GetChipHandle();
CenMbaDataBundle * mbadb = getMbaDataBundle( i_mbaChip );
ExtensibleChip * membufChip = mbadb->getMembChip();
if ( NULL == membufChip )
{
PRDF_ERR( PRDF_FUNC "getMembChip() failed" );
o_rc = FAIL; break;
}
uint8_t mbaPos = getTargetPosition( mbaTrgt );
if ( MAX_MBA_PER_MEMBUF <= mbaPos )
{
PRDF_ERR( PRDF_FUNC "mbaPos %d is invalid", mbaPos );
o_rc = FAIL; break;
}
const bool isX4 = isDramWidthX4(mbaTrgt);
// Get the current spares on this rank.
CenSymbol sp0, sp1, ecc;
o_rc = mssGetSteerMux( mbaTrgt, i_rank, sp0, sp1, ecc );
if ( SUCCESS != o_rc )
{
PRDF_ERR( PRDF_FUNC "mssGetSteerMux() failed." );
break;
}
// Use this map to keep track of the total counts per DRAM.
DramCountMap dramCounts;
const char * reg_str = NULL;
SCAN_COMM_REGISTER_CLASS * reg = NULL;
for ( uint8_t regIdx = 0; regIdx < CE_REGS_PER_MBA; regIdx++ )
{
reg_str = mbsCeStatReg[mbaPos][regIdx];
reg = membufChip->getRegister( reg_str );
o_rc = reg->Read();
if ( SUCCESS != o_rc )
{
PRDF_ERR( PRDF_FUNC "Read() failed on %s", reg_str );
break;
}
uint8_t baseSymbol = SYMBOLS_PER_CE_REG * regIdx;
for ( uint8_t i = 0; i < SYMBOLS_PER_CE_REG; i++ )
{
uint8_t count = reg->GetBitFieldJustified( (i*8), 8 );
if ( 0 == count ) continue; // nothing to do
uint8_t sym = baseSymbol + i;
uint8_t dram = symbol2Dram( sym, isX4 );
// Keep track of the total DRAM counts.
dramCounts[dram].totalCount += count;
// Add any symbols that have exceeded threshold to the list.
if ( i_thr <= count )
{
// Keep track of the total number of symbols per DRAM that
// have exceeded threshold.
dramCounts[dram].symbolCount++;
SymbolData symData;
symData.symbol = CenSymbol::fromSymbol( mbaTrgt, i_rank,
sym, CEN_SYMBOL::BOTH_SYMBOL_DQS );
if ( !symData.symbol.isValid() )
{
PRDF_ERR( PRDF_FUNC "CenSymbol() failed: symbol=%d",
sym );
o_rc = FAIL;
break;
}
else
{
// Check if this symbol is on any of the spares.
if ( ( sp0.isValid() &&
(sp0.getDram() == symData.symbol.getDram()) ) ||
( sp1.isValid() &&
(sp1.getDram() == symData.symbol.getDram()) ) )
{
symData.symbol.setDramSpared();
}
if ( ecc.isValid() &&
(ecc.getDram() == symData.symbol.getDram()) )
{
symData.symbol.setEccSpared();
}
// Add the symbol to the list.
symData.count = count;
o_maintStats.push_back( symData );
}
}
}
if ( SUCCESS != o_rc ) break;
}
if ( SUCCESS != o_rc ) break;
if ( o_maintStats.empty() ) break; // no need to continue
// Sort the list of symbols.
std::sort( o_maintStats.begin(), o_maintStats.end(), sortSymDataCount );
// Get the DRAM with the highest count.
uint32_t highestDram = 0;
uint32_t highestCount = 0;
const uint32_t symbolTH = isX4 ? 1 : 2;
for ( DramCountMap::iterator it = dramCounts.begin();
it != dramCounts.end(); ++it )
{
if ( (symbolTH <= it->second.symbolCount) &&
(highestCount < it->second.totalCount ) )
{
highestDram = it->first;
highestCount = it->second.totalCount;
}
}
if ( 0 != highestCount )
{
uint8_t sym = dram2Symbol( highestDram, isX4 );
o_chipMark = CenSymbol::fromSymbol( mbaTrgt, i_rank, sym );
// Check if this symbol is on any of the spares.
if ( ( sp0.isValid() && (sp0.getDram() == o_chipMark.getDram()) ) ||
( sp1.isValid() && (sp1.getDram() == o_chipMark.getDram()) ) )
{
o_chipMark.setDramSpared();
}
if ( ecc.isValid() && (ecc.getDram() == o_chipMark.getDram()) )
{
o_chipMark.setEccSpared();
}
}
} while(0);
if ( SUCCESS != o_rc )
{
PRDF_ERR( PRDF_FUNC "Failed: i_mbaChip=0x%08x i_rank=m%ds%d i_thr=%d",
i_mbaChip->GetId(), i_rank.getMaster(), i_rank.getSlave(),
i_thr );
}
return o_rc;
#undef PRDF_FUNC
}
int32_t CenMbaTdCtlr::analyzeTpsPhase2( STEP_CODE_DATA_STRUCT & io_sc,
const CenAddr & i_stopAddr,
const CenAddr & i_endAddr )
{
#define PRDF_FUNC "[CenMbaTdCtlr::analyzeTpsPhase2] "
int32_t o_rc = SUCCESS;
do
{
if ( TPS_PHASE_2 != iv_tdState )
{
PRDF_ERR( PRDF_FUNC"Invalid state machine configuration" );
o_rc = FAIL; break;
}
CenMbaDataBundle * mbadb = getMbaDataBundle( iv_mbaChip );
o_rc = mbadb->getIplCeStats()->calloutHardCes( iv_rank );
if ( SUCCESS != o_rc )
{
PRDF_ERR( PRDF_FUNC"calloutHardCes() failed");
break;
}
// Get error condition which caused command to stop
uint16_t eccErrorMask = NO_ERROR;
o_rc = checkEccErrors( eccErrorMask, io_sc );
if ( SUCCESS != o_rc )
{
PRDF_ERR( PRDF_FUNC"checkEccErrors() failed" );
break;
}
if ( ( eccErrorMask & UE ) || ( eccErrorMask & RETRY_CTE ))
{
// Handle UE. Highest priority
o_rc = handleUE( io_sc );
if ( SUCCESS != o_rc )
{
PRDF_ERR( PRDF_FUNC"handleUE() failed" );
break;
}
}
else if ( eccErrorMask & MPE )
{
o_rc = handleMPE( io_sc );
if ( SUCCESS != o_rc )
{
PRDF_ERR( PRDF_FUNC"handleMPE() failed");
break;
}
}
else
{
// No error found so add rank to callout list, just in case.
MemoryMru memmru (iv_mbaTrgt, iv_rank, MemoryMruData::CALLOUT_RANK);
io_sc.service_data->SetCallout( memmru );
io_sc.service_data->AddSignatureList( iv_mbaTrgt,
PRDFSIG_EndTpsPhase2 );
iv_tdState = NO_OP;
}
} while (0);
return o_rc;
#undef PRDF_FUNC
}
int32_t CenMbaTdCtlr::handleUE( STEP_CODE_DATA_STRUCT & io_sc )
{
#define PRDF_FUNC "[CenMbaTdCtlr::handleUE] "
using namespace CalloutUtil;
int32_t o_rc = SUCCESS;
iv_tdState = NO_OP; // Abort the TD procedure.
setTdSignature( io_sc, PRDFSIG_MaintUE );
io_sc.service_data->SetServiceCall();
CenMbaDataBundle * mbadb = getMbaDataBundle( iv_mbaChip );
do
{
// Clean up the maintenance command. This is needed just in case the UE
// isolation procedure is modified to use maintenance commands.
o_rc = cleanupPrevCmd();
if ( SUCCESS != o_rc )
{
PRDF_ERR( PRDF_FUNC"cleanupPrevCmd() failed" );
break;
}
// Look for all failing bits on this rank.
CenDqBitmap bitmap;
o_rc = mssIplUeIsolation( iv_mbaTrgt, iv_rank, bitmap );
if ( SUCCESS != o_rc )
{
PRDF_ERR( PRDF_FUNC"mssIplUeIsolation() failed" );
break;
}
// Add UE data to capture data.
bitmap.getCaptureData( io_sc.service_data->GetCaptureData() );
// Callout the failing DIMMs.
TargetHandleList callouts;
for ( int32_t ps = 0; ps < PORT_SLCT_PER_MBA; ps++ )
{
bool badDqs = false;
o_rc = bitmap.badDqs( ps, badDqs );
if ( SUCCESS != o_rc )
{
PRDF_ERR( PRDF_FUNC"badDqs(%d) failed", ps );
break;
}
if ( !badDqs ) continue; // nothing to do.
TargetHandleList dimms = getConnectedDimms(iv_mbaTrgt, iv_rank, ps);
if ( 0 == dimms.size() )
{
PRDF_ERR( PRDF_FUNC"getConnectedDimms(%d) failed", ps );
o_rc = FAIL; break;
}
callouts.insert( callouts.end(), dimms.begin(), dimms.end() );
if ( isMfgCeCheckingEnabled() )
{
// As we are doing callout for UE, we dont need to do callout
// during CE for this rank on given port
mbadb->getIplCeStats()->banAnalysis( iv_rank, ps );
}
}
if ( SUCCESS != o_rc ) break;
if ( 0 == callouts.size() )
{
// It is possible the scrub counters have rolled over to zero due to
// a known DD1.0 hardware bug. In this case, the best we can do is
// callout both DIMMs, because at minimum we know there was a UE, we
// just don't know where.
// NOTE: If this condition happens because of a DD2.0+ bug, the
// mssIplUeIsolation procedure will callout the Centaur.
callouts = getConnectedDimms( iv_mbaTrgt, iv_rank );
if ( 0 == callouts.size() )
{
PRDF_ERR( PRDF_FUNC"getConnectedDimms() failed" );
o_rc = FAIL; break;
}
if ( isMfgCeCheckingEnabled() )
{
// As we are doing callout for UE, we dont need to do callout
// during CE for this rank on both port
mbadb->getIplCeStats()->banAnalysis( iv_rank);
}
}
// Callout all DIMMs in the list.
for ( TargetHandleList::iterator i = callouts.begin();
i != callouts.end(); i++ )
{
io_sc.service_data->SetCallout( *i, MRU_HIGH );
}
} while(0);
return o_rc;
#undef PRDF_FUNC
}
