int32_t CenMbaTdCtlr::analyzeCmdComplete( STEP_CODE_DATA_STRUCT & io_sc, const CenAddr & i_stopAddr, const CenAddr & i_endAddr ) { #define PRDF_FUNC "[CenMbaTdCtlr::analyzeCmdComplete] " int32_t o_rc = SUCCESS; do { if ( NO_OP != iv_tdState ) { PRDF_ERR( PRDF_FUNC"Invalid state machine configuration" ); o_rc = FAIL; break; } // Initialize iv_rank. This must be done before calling other // functions as they require iv_rank to be accurate. iv_rank = CenRank( i_stopAddr.getRank() ); // 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 ) { // 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 if ( isMfgCeCheckingEnabled() && (eccErrorMask & HARD_CTE) ) { // During MNFG IPL CE, we will get this condition. // During SF read, all CE are reported as Hard CE. // So we will only check for Hard CE threshold. o_rc = handleMnfgCeEte( io_sc ); if ( SUCCESS != o_rc ) { PRDF_ERR( PRDF_FUNC"handleMnfgCeEte() failed" ); break; } } else { // If maint cmd completed with no error, don't commit error log. io_sc.service_data->DontCommitErrorLog(); } } 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 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 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[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 }