int32_t chnlCsCleanup( ExtensibleChip *i_mbChip,
STEP_CODE_DATA_STRUCT & i_sc )
{
#define PRDF_FUNC "[MemUtils::chnlCsCleanup] "
int32_t o_rc = SUCCESS;
do
{
if( ( NULL == i_mbChip ) ||
( TYPE_MEMBUF != getTargetType( i_mbChip->GetChipHandle() )))
{
PRDF_ERR( PRDF_FUNC "Invalid parameters" );
o_rc = FAIL; break;
}
if (( ! i_sc.service_data->IsUnitCS() ) ||
(CHECK_STOP == i_sc.service_data->getPrimaryAttnType()) )
break;
CenMembufDataBundle * mbdb = getMembufDataBundle( i_mbChip );
if ( !mbdb->iv_doChnlFailCleanup )
break; // Cleanup has already been done.
// Set it as SUE generation point.
i_sc.service_data->SetFlag( ServiceDataCollector::UERE );
ExtensibleChip * mcsChip = mbdb->getMcsChip();
if ( NULL == mcsChip )
{
PRDF_ERR( PRDF_FUNC "MCS chip is NULL for Membuf:0x%08X",
i_mbChip->GetId() );
o_rc = FAIL; break;
}
TargetHandle_t mcs = mcsChip->GetChipHandle();
ExtensibleChip * procChip = NULL;
uint8_t pos = getTargetPosition( mcs );
TargetHandle_t proc = getParentChip ( mcs );
if ( NULL == proc )
{
PRDF_ERR( PRDF_FUNC "Proc is NULL for Mcs:0x%08X", getHuid( mcs ) );
o_rc = FAIL; break;
}
procChip = (ExtensibleChip *)systemPtr->GetChip( proc );
if( NULL == procChip )
{
PRDF_ERR( PRDF_FUNC "Can not find Proc chip for HUID:0x%08X",
getHuid( proc) );
o_rc = FAIL; break;
}
// This is a cleanup function. If we get any error from scom
// operations, we will still continue with cleanup.
SCAN_COMM_REGISTER_CLASS * l_tpMask =
procChip->getRegister("TP_CHIPLET_FIR_MASK");
o_rc |= l_tpMask->Read();
if ( SUCCESS == o_rc )
{
// Bits 5-12 maps to attentions from MCS0-MCS7.
l_tpMask->SetBit( 5 + pos );
o_rc |= l_tpMask->Write();
}
// Mask attentions from the Centaur
const char *iomcFirMask = ( pos < 4 )?
"IOMCFIR_0_MASK_OR":"IOMCFIR_1_MASK_OR";
SCAN_COMM_REGISTER_CLASS * iomcMask =
procChip->getRegister( iomcFirMask);
if ( pos >= 4 ) pos -= 4;
// 8 bits are reserved for each Centaur in IOMCFIR.
// There are total 4 ( for P system ) centaur supported
// in MCS. Bits for first centaur starts from bit 8.
iomcMask->SetBitFieldJustified( 8+ ( pos*8 ), 8, 0xff);
o_rc |= iomcMask->Write();
SCAN_COMM_REGISTER_CLASS * l_tpfirmask = NULL;
SCAN_COMM_REGISTER_CLASS * l_nestfirmask = NULL;
SCAN_COMM_REGISTER_CLASS * l_memfirmask = NULL;
SCAN_COMM_REGISTER_CLASS * l_memspamask = NULL;
l_tpfirmask = i_mbChip->getRegister("TP_CHIPLET_FIR_MASK");
l_nestfirmask = i_mbChip->getRegister("NEST_CHIPLET_FIR_MASK");
l_memfirmask = i_mbChip->getRegister("MEM_CHIPLET_FIR_MASK");
l_memspamask = i_mbChip->getRegister("MEM_CHIPLET_SPA_MASK");
l_tpfirmask->setAllBits(); o_rc |= l_tpfirmask->Write();
l_nestfirmask->setAllBits(); o_rc |= l_nestfirmask->Write();
l_memfirmask->setAllBits(); o_rc |= l_memfirmask->Write();
l_memspamask->setAllBits(); o_rc |= l_memspamask->Write();
for ( uint32_t i = 0; i < MAX_MBA_PER_MEMBUF; i++ )
{
ExtensibleChip * mbaChip = mbdb->getMbaChip( i );
if( NULL != mbaChip )
{
TargetHandle_t mba = mbaChip->GetChipHandle();
if ( NULL != mba )
{
#if defined(__HOSTBOOT_MODULE) && \
!defined(__HOSTBOOT_RUNTIME)
// This is very small platform specific code. So not
// creating a separate file for this.
int32_t l_rc = mdiaSendEventMsg( mba, MDIA::SKIP_MBA );
if ( SUCCESS != l_rc )
{
PRDF_ERR( PRDF_FUNC "mdiaSendEventMsg(0x%08x, SKIP_MBA) "
"failed", getHuid( mba ) );
o_rc |= l_rc;
}
#else
int32_t l_rc = DEALLOC::mbaGard( mbaChip );
if ( SUCCESS != l_rc )
{
PRDF_ERR( PRDF_FUNC "mbaGard failed. HUID: 0x%08x",
getHuid( mba ) );
o_rc |= l_rc;
}
#endif // __HOSTBOOT_MODULE
}
}
}
// Clean up complete an is no longer required.
mbdb->iv_doChnlFailCleanup = false;
} while(0);
return o_rc;
#undef PRDF_FUNC
}
/**
* @brief MBSECCFIR[19] - Fetch UE.
* @param i_membChip A Centaur chip.
* @param i_sc The step code data struct.
* @param i_mbaPos The MBA position.
* @return SUCCESS
*/
int32_t AnalyzeFetchUe( ExtensibleChip * i_membChip,
STEP_CODE_DATA_STRUCT & i_sc, uint32_t i_mbaPos )
{
#define PRDF_FUNC "[AnalyzeFetchUe] "
int32_t l_rc = SUCCESS;
ExtensibleChip * mbaChip = NULL;
do
{
// All memory UEs should be customer viewable. Normally, this would be
// done by setting the threshold to 1, but we do not want to mask UEs
// on the first occurrence.
i_sc.service_data->SetServiceCall();
CenMembufDataBundle * membdb = getMembufDataBundle( i_membChip );
mbaChip = membdb->getMbaChip( i_mbaPos );
if ( NULL == mbaChip )
{
PRDF_ERR( PRDF_FUNC"getMbaChip() returned NULL" );
l_rc = FAIL; break;
}
CenAddr addr;
l_rc = getCenReadAddr( i_membChip, i_mbaPos, READ_UE_ADDR, addr );
if ( SUCCESS != l_rc )
{
PRDF_ERR( PRDF_FUNC"getCenReadAddr() failed" );
break;
}
CenRank rank = addr.getRank();
// Add address to UE table.
CenMbaDataBundle * mbadb = getMbaDataBundle( mbaChip );
mbadb->iv_ueTable.addEntry( UE_TABLE::FETCH_UE, addr );
// Callout the rank.
MemoryMru memmru ( mbaChip->GetChipHandle(), rank,
MemoryMruData::CALLOUT_RANK );
i_sc.service_data->SetCallout( memmru );
// Add a TPS request to the TD queue and ban any further TPS requests
// for this rank.
l_rc = mbadb->iv_tdCtlr.handleTdEvent( i_sc, rank,
CenMbaTdCtlrCommon::TPS_EVENT,
true );
if ( SUCCESS != l_rc )
{
PRDF_ERR( PRDF_FUNC"handleTdEvent() failed: rank=m%ds%d",
rank.getMaster(), rank.getSlave() );
// We are not adding break here as we still want to do lmbGard
// If you want to add any code after this which depends on result
// of handleTdEvent result, add the code judicially.
}
#ifndef __HOSTBOOT_MODULE
// Send lmb gard message to PHYP.
int32_t lmbRc = DEALLOC::lmbGard( mbaChip, addr );
if ( SUCCESS != lmbRc )
{
PRDF_ERR( PRDF_FUNC"lmbGard() failed" );
l_rc = lmbRc; break;
}
#endif
} while (0);
// Add ECC capture data for FFDC.
if ( NULL != mbaChip )
CenMbaCaptureData::addMemEccData( mbaChip, i_sc );
if ( SUCCESS != l_rc )
{
PRDF_ERR( PRDF_FUNC"Failed: i_membChip=0x%08x i_mbaPos=%d",
i_membChip->GetId(), i_mbaPos );
CalloutUtil::defaultError( i_sc );
}
return SUCCESS; // Intentionally return SUCCESS for this plugin
#undef PRDF_FUNC
}
/**
* @brief Fetch Retry CE / Prefetch UE Errors.
* @param i_membChip A Centaur chip.
* @param i_sc The step code data struct.
* @param i_mbaPos The MBA position.
* @param i_isRceError True for RCE error false otherwise.
* @return SUCCESS
*/
int32_t AnalyzeFetchRcePue( ExtensibleChip * i_membChip,
STEP_CODE_DATA_STRUCT & i_sc, uint32_t i_mbaPos,
bool i_isRceError )
{
#define PRDF_FUNC "[AnalyzeFetchRcePue] "
int32_t l_rc = SUCCESS;
ExtensibleChip * mbaChip = NULL;
do
{
CenMembufDataBundle * membdb = getMembufDataBundle( i_membChip );
mbaChip = membdb->getMbaChip( i_mbaPos );
if ( NULL == mbaChip )
{
PRDF_ERR( PRDF_FUNC"getMbaChip() returned NULL" );
l_rc = FAIL; break;
}
CenMbaDataBundle * mbadb = getMbaDataBundle( mbaChip );
CenAddr addr;
if ( i_isRceError )
l_rc = getCenReadAddr( i_membChip, i_mbaPos, READ_RCE_ADDR, addr );
else
l_rc = getCenReadAddr( i_membChip, i_mbaPos, READ_UE_ADDR, addr );
if ( SUCCESS != l_rc )
{
PRDF_ERR( PRDF_FUNC"getCenReadAddr() failed" );
break;
}
CenRank rank = addr.getRank();
// Callout the rank.
MemoryMru memmru ( mbaChip->GetChipHandle(), rank,
MemoryMruData::CALLOUT_RANK );
i_sc.service_data->SetCallout( memmru );
// Add an entry to the RCE table.
if ( mbadb->iv_rceTable.addEntry(rank, i_sc) )
{
// Add a TPS request to the queue TD queue.
l_rc = mbadb->iv_tdCtlr.handleTdEvent( i_sc, rank,
CenMbaTdCtlrCommon::TPS_EVENT );
if ( SUCCESS != l_rc )
{
PRDF_ERR( PRDF_FUNC"handleTdEvent() failed." );
break;
}
}
} while (0);
// Add ECC capture data for FFDC.
if ( NULL != mbaChip )
CenMbaCaptureData::addMemEccData( mbaChip, i_sc );
if ( SUCCESS != l_rc )
{
PRDF_ERR( PRDF_FUNC"Failed: i_membChip=0x%08x i_mbaPos=%d "
"i_isRceError=%c", i_membChip->GetId(), i_mbaPos,
i_isRceError ? 'T' : 'F' );
CalloutUtil::defaultError( i_sc );
}
return SUCCESS; // Intentionally return SUCCESS for this plugin
#undef PRDF_FUNC
}
/**
* @brief MBSECCFIR[16] - Fetch New CE (NCE).
* @param i_membChip A Centaur chip.
* @param i_sc The step code data struct.
* @param i_mbaPos The MBA position.
* @return SUCCESS
*/
int32_t AnalyzeFetchNce( ExtensibleChip * i_membChip,
STEP_CODE_DATA_STRUCT & i_sc, uint32_t i_mbaPos )
{
#define PRDF_FUNC "[AnalyzeFetchNce] "
int32_t l_rc = SUCCESS;
ExtensibleChip * mbaChip = NULL;
do
{
CenMembufDataBundle * membdb = getMembufDataBundle( i_membChip );
mbaChip = membdb->getMbaChip( i_mbaPos );
if ( NULL == mbaChip )
{
PRDF_ERR( PRDF_FUNC"getMbaChip() returned NULL" );
l_rc = FAIL; break;
}
TargetHandle_t mbaTrgt = mbaChip->GetChipHandle();
CenAddr addr;
l_rc = getCenReadAddr( i_membChip, i_mbaPos, READ_NCE_ADDR, addr );
if ( SUCCESS != l_rc )
{
PRDF_ERR( PRDF_FUNC"getCenReadAddr() failed" );
break;
}
CenRank rank = addr.getRank();
if ( 0x20 > getChipLevel(i_membChip->GetChipHandle()) )
{
// There is a bug in DD1.x where the value of MBSEVR cannot be
// trusted. The workaround is too complicated for its value so
// callout the rank instead.
MemoryMru memmru ( mbaTrgt, rank, MemoryMruData::CALLOUT_RANK );
i_sc.service_data->SetCallout( memmru );
}
else // DD2.0+
{
// Get the failing symbol
const char * reg_str = (0 == i_mbaPos) ? "MBA0_MBSEVR"
: "MBA1_MBSEVR";
SCAN_COMM_REGISTER_CLASS * reg = i_membChip->getRegister(reg_str);
l_rc = reg->Read();
if ( SUCCESS != l_rc )
{
PRDF_ERR( PRDF_FUNC"Read() failed on %s", reg_str );
break;
}
uint8_t galois = reg->GetBitFieldJustified( 40, 8 );
uint8_t mask = reg->GetBitFieldJustified( 32, 8 );
CenSymbol symbol = CenSymbol::fromGalois( mbaTrgt, rank, galois,
mask );
if ( !symbol.isValid() )
{
PRDF_ERR( PRDF_FUNC"Failed to create symbol: galois=0x%02x "
"mask=0x%02x", galois, mask );
break;
}
// Check if this symbol is on any of the spares.
CenSymbol sp0, sp1, ecc;
l_rc = mssGetSteerMux( mbaTrgt, rank, sp0, sp1, ecc );
if ( SUCCESS != l_rc )
{
PRDF_ERR( PRDF_FUNC"mssGetSteerMux() failed. HUID: 0x%08x "
"rank: %d", getHuid(mbaTrgt), rank.getMaster() );
break;
}
if ( (sp0.isValid() && (sp0.getDram() == symbol.getDram())) ||
(sp1.isValid() && (sp1.getDram() == symbol.getDram())) )
{
symbol.setDramSpared();
}
if ( ecc.isValid() && (ecc.getDram() == symbol.getDram()) )
{
symbol.setEccSpared();
}
// Add the DIMM to the callout list
MemoryMru memmru ( mbaTrgt, rank, symbol );
i_sc.service_data->SetCallout( memmru, MRU_MEDA );
// Add to CE table
CenMbaDataBundle * mbadb = getMbaDataBundle( mbaChip );
uint32_t ceTableRc = mbadb->iv_ceTable.addEntry( addr, symbol );
bool doTps = ( CenMbaCeTable::NO_TH_REACHED != ceTableRc );
// Check MNFG thresholds, if needed.
if ( mfgMode() )
{
// Get the MNFG CE thresholds.
uint16_t dramTh, hrTh, dimmTh;
l_rc = getMnfgMemCeTh( mbaChip, rank, dramTh, hrTh, dimmTh );
if ( SUCCESS != l_rc )
{
PRDF_ERR( PRDF_FUNC"getMnfgMemCeTh() failed: rank=m%ds%d",
rank.getMaster(), rank.getSlave() );
break;
}
// Get counts from CE table.
uint32_t dramCount, hrCount, dimmCount;
mbadb->iv_ceTable.getMnfgCounts( addr.getRank(), symbol,
dramCount, hrCount,
dimmCount );
if ( dramTh < dramCount )
{
i_sc.service_data->AddSignatureList( mbaTrgt,
PRDFSIG_MnfgDramCte );
i_sc.service_data->SetServiceCall();
doTps = true;
}
else if ( hrTh < hrCount )
{
i_sc.service_data->AddSignatureList( mbaTrgt,
PRDFSIG_MnfgHrCte );
i_sc.service_data->SetServiceCall();
doTps = true;
}
else if ( dimmTh < dimmCount )
{
i_sc.service_data->AddSignatureList( mbaTrgt,
PRDFSIG_MnfgDimmCte );
i_sc.service_data->SetServiceCall();
doTps = true;
}
else if ( 0 != (CenMbaCeTable::TABLE_FULL & ceTableRc) )
{
i_sc.service_data->AddSignatureList( mbaTrgt,
PRDFSIG_MnfgTableFull);
// The table is full and no other threshold has been met.
// We are in a state where we may never hit a MNFG
// threshold. Callout all memory behind the MBA. Also, since
// the counts are all over the place, there may be a problem
// with the MBA. So call it out as well.
MemoryMru all_mm ( mbaTrgt, rank,
MemoryMruData::CALLOUT_ALL_MEM );
i_sc.service_data->SetCallout( all_mm, MRU_MEDA );
i_sc.service_data->SetCallout( mbaTrgt, MRU_MEDA );
i_sc.service_data->SetServiceCall();
}
else if ( 0 != (CenMbaCeTable::ENTRY_TH_REACHED & ceTableRc) )
{
i_sc.service_data->AddSignatureList( mbaTrgt,
PRDFSIG_MnfgEntryCte );
// There is a single entry threshold and no other threshold
// has been met. This is a potential flooding issue, so make
// the DIMM callout predictive.
i_sc.service_data->SetServiceCall();
}
}
// Initiate a TPS procedure, if needed.
if ( doTps )
{
// If a MNFG threshold has been reached (predictive callout), we
// will still try to start TPS just in case MNFG disables the
// termination policy.
// Will not be able to do TPS during hostboot. Note that we will
// still call handleTdEvent() so we can get the trace statement
// indicating TPS was requested during Hostboot.
l_rc = mbadb->iv_tdCtlr.handleTdEvent( i_sc, rank,
CenMbaTdCtlrCommon::TPS_EVENT );
if ( SUCCESS != l_rc )
{
PRDF_ERR( PRDF_FUNC"handleTdEvent() failed: rank=m%ds%d",
rank.getMaster(), rank.getSlave() );
break;
}
}
}
} while (0);
// Add ECC capture data for FFDC.
if ( NULL != mbaChip )
CenMbaCaptureData::addMemEccData( mbaChip, i_sc );
if ( SUCCESS != l_rc )
{
PRDF_ERR( PRDF_FUNC"Failed: i_membChip=0x%08x i_mbaPos=%d",
i_membChip->GetId(), i_mbaPos );
CalloutUtil::defaultError( i_sc );
}
return SUCCESS; // Intentionally return SUCCESS for this plugin
#undef PRDF_FUNC
}
/**
* @brief MBSECCFIR[0:7] - Fetch Mark Placed Event (MPE).
* @param i_membChip A Centaur chip.
* @param i_sc The step code data struct.
* @param i_mbaPos The MBA position.
* @param i_rank The target rank.
* @return SUCCESS
*/
int32_t AnalyzeFetchMpe( ExtensibleChip * i_membChip,
STEP_CODE_DATA_STRUCT & i_sc,
uint32_t i_mbaPos, uint8_t i_rank )
{
#define PRDF_FUNC "[AnalyzeFetchMpe] "
int32_t l_rc = SUCCESS;
ExtensibleChip * mbaChip = NULL;
do
{
CenMembufDataBundle * membdb = getMembufDataBundle( i_membChip );
mbaChip = membdb->getMbaChip( i_mbaPos );
if ( NULL == mbaChip )
{
PRDF_ERR( PRDF_FUNC"getMbaChip() returned NULL" );
l_rc = FAIL; break;
}
CenMbaDataBundle * mbadb = getMbaDataBundle( mbaChip );
TargetHandle_t mbaTrgt = mbaChip->GetChipHandle();
CenAddr addr;
l_rc = getCenReadAddr( i_membChip, i_mbaPos, READ_MPE_ADDR, addr );
if ( SUCCESS != l_rc )
{
PRDF_ERR( PRDF_FUNC"getCenReadAddr() failed" );
break;
}
// If the address does not match the rank that reported the attention,
// there are multiple MPE attentions and the address was overwritten.
// In this case, add an invalid dummy address to the UE table.
if ( addr.getRank().getMaster() != i_rank )
{
addr = CenAddr( i_rank, 0, 0xffffffff, 0xffffffff, 0xffffffff );
}
mbadb->iv_ueTable.addEntry( UE_TABLE::FETCH_MPE, addr );
// Get the current mark in hardware.
CenRank rank ( addr.getRank() );
CenMark mark;
l_rc = mssGetMarkStore( mbaTrgt, rank, mark );
if ( SUCCESS != l_rc )
{
PRDF_ERR( PRDF_FUNC"mssGetMarkStore() failed");
break;
}
if ( !mark.getCM().isValid() )
{
PRDF_ERR( PRDF_FUNC"FIR bit set but no valid chip mark" );
l_rc = FAIL; break;
}
// Callout the mark.
CalloutUtil::calloutMark( mbaTrgt, rank, mark, i_sc );
// Tell TD controller to handle VCM event.
l_rc = mbadb->iv_tdCtlr.handleTdEvent( i_sc, rank,
CenMbaTdCtlrCommon::VCM_EVENT );
if ( SUCCESS != l_rc )
{
PRDF_ERR( PRDF_FUNC"handleTdEvent() failed." );
break;
}
} while (0);
// Add ECC capture data for FFDC.
if ( NULL != mbaChip )
CenMbaCaptureData::addMemEccData( mbaChip, i_sc );
if ( SUCCESS != l_rc )
{
PRDF_ERR( PRDF_FUNC"Failed: i_membChip=0x%08x i_mbaPos=%d i_rank=%d",
i_membChip->GetId(), i_mbaPos, i_rank );
CalloutUtil::defaultError( i_sc );
}
return SUCCESS; // Intentionally return SUCCESS for this plugin
#undef PRDF_FUNC
}
int32_t cleanupSecondaryFirBits( ExtensibleChip * i_chip,
TYPE i_busType,
uint32_t i_busPos )
{
int32_t l_rc = SUCCESS;
TargetHandle_t mcsTgt = NULL;
TargetHandle_t mbTgt = NULL;
ExtensibleChip * mcsChip = NULL;
ExtensibleChip * mbChip = NULL;
//In case of spare deployed attention for DMI bus, we need to clear
// secondary MBIFIR[10] and MCIFIR[10] bits.
do
{
if ( i_busType == TYPE_MCS )
{
mcsTgt = getConnectedChild( i_chip->GetChipHandle(),
TYPE_MCS,
i_busPos);
if (!mcsTgt) break;
mcsChip = ( ExtensibleChip * )systemPtr->GetChip( mcsTgt );
if (!mcsChip) break;
mbChip = getMcsDataBundle( mcsChip )->getMembChip();
if (!mbChip) break;
mbTgt = mbChip->GetChipHandle();
if (!mbTgt) break;
}
else if ( i_busType == TYPE_MEMBUF )
{
mbTgt = i_chip->GetChipHandle();
if (!mbTgt) break;
mcsChip = getMembufDataBundle( i_chip )->getMcsChip();
if (!mcsChip) break;
mcsTgt = mcsChip->GetChipHandle();
if (!mcsTgt) break;
mbChip = i_chip;
}
else
{
// We only need to clean secondary FIR bits for DMI bus
l_rc = SUCCESS;
break;
}
SCAN_COMM_REGISTER_CLASS * mciAnd = mcsChip->getRegister("MCIFIR_AND");
SCAN_COMM_REGISTER_CLASS * mbiAnd = mbChip->getRegister( "MBIFIR_AND");
mciAnd->setAllBits(); mciAnd->ClearBit(10);
mbiAnd->setAllBits(); mbiAnd->ClearBit(10);
l_rc = mciAnd->Write();
l_rc |= mbiAnd->Write();
if ( SUCCESS != l_rc )
{
PRDF_ERR( "[cleanupSecondaryFirBits] Write() failed on "
"MCIFIR/MBIFIR: MCS=0x%08x MEMB=0x%08x",
mcsChip->GetId(), mbChip->GetId() );
break;
}
} while (0);
return l_rc;
}
