// Move the OCCs to active state or log unrecoverable error and
// stay in safe mode
void processOccStartStatus(const bool i_startCompleted,
TARGETING::Target * i_failedOccTarget)
{
TMGT_INF(">>processOccStartStatus(%d,0x%p)",
i_startCompleted, i_failedOccTarget);
errlHndl_t l_err = NULL;
uint32_t l_huid = 0;
if (i_failedOccTarget)
{
l_huid = TARGETING::get_huid(i_failedOccTarget);
}
TMGT_INF("processOccStartStatus(Start Success=%c, failedOcc=0x%08X)",
i_startCompleted?'y':'n', l_huid);
if (i_startCompleted)
{
// Query functional OCCs
l_err = OccManager::buildOccs();
if (NULL == l_err)
{
if (NULL != OccManager::getMasterOcc())
{
do
{
#ifndef __HOSTBOOT_RUNTIME
// Build pstate tables (once per IPL)
l_err = genPstateTables();
if(l_err)
{
break;
}
// Calc memory throttles (once per IPL)
calcMemThrottles();
#endif
// Make sure OCCs are ready for communication
OccManager::waitForOccCheckpoint();
#ifdef __HOSTBOOT_RUNTIME
// TODO RTC 124738 Final solution TBD
// Perhapse POLL scom 0x6a214 until bit 31 is set?
nanosleep(1,0);
#endif
// Send poll to establish comm
TMGT_INF("Send initial poll to all OCCs to"
" establish comm");
l_err = OccManager::sendOccPoll();
if (l_err)
{
// Continue even if failed (poll will be retried)
ERRORLOG::errlCommit(l_err, HTMGT_COMP_ID);
}
// Send ALL config data
sendOccConfigData();
// Set the User PCAP
l_err = sendOccUserPowerCap();
if (l_err)
{
break;
}
// Wait for all OCCs to go to the target state
l_err = waitForOccState();
if ( l_err )
{
break;
}
// Set active sensors for all OCCs,
// so BMC can start communication with OCCs
l_err = setOccActiveSensors(true);
if (l_err)
{
// Continue even if failed to update sensor
ERRORLOG::errlCommit(l_err, HTMGT_COMP_ID);
}
} while(0);
}
else
{
TMGT_ERR("Unable to find any Master capable OCCs");
/*@
* @errortype
* @reasoncode HTMGT_RC_OCC_MASTER_NOT_FOUND
* @moduleid HTMGT_MOD_LOAD_START_STATUS
* @userdata1[0:7] number of OCCs
* @devdesc No OCC master was found
*/
bldErrLog(l_err, HTMGT_MOD_LOAD_START_STATUS,
HTMGT_RC_OCC_MASTER_NOT_FOUND,
OccManager::getNumOccs(), 0, 0, 0,
ERRORLOG::ERRL_SEV_INFORMATIONAL);
}
}
else
{
// Failed to find functional OCCs, no need to try again
// Set original error log as unrecoverable and commit
l_err->setSev(ERRORLOG::ERRL_SEV_UNRECOVERABLE);
ERRORLOG::errlCommit(l_err, HTMGT_COMP_ID);
}
}
else
{
TMGT_ERR("All OCCs were not loaded/started successfully");
/*@
* @errortype
* @reasoncode HTMGT_RC_OCC_START_FAIL
* @moduleid HTMGT_MOD_LOAD_START_STATUS
* @userdata1 Failing OCC HUID
* @devdesc OCCs were not loaded/started successfully
*/
bldErrLog(l_err, HTMGT_MOD_LOAD_START_STATUS,
HTMGT_RC_OCC_START_FAIL,
l_huid, 0, 0, 0,
ERRORLOG::ERRL_SEV_INFORMATIONAL);
}
if (NULL != l_err)
{
TMGT_ERR("OCCs not all active. Attempting OCC Reset");
TMGT_CONSOLE("OCCs are not active (rc=0x%04X). "
"Attempting OCC Reset",
l_err->reasonCode());
TMGT_INF("Calling resetOccs");
errlHndl_t err2 = OccManager::resetOccs(NULL);
if(err2)
{
TMGT_ERR("OccManager::resetOccs failed with 0x%04X",
err2->reasonCode());
// Set original error log as unrecoverable and commit
l_err->setSev(ERRORLOG::ERRL_SEV_UNRECOVERABLE);
ERRORLOG::errlCommit(l_err, HTMGT_COMP_ID);
// Commit occReset error
ERRORLOG::errlCommit(err2, HTMGT_COMP_ID);
}
else
{
// retry worked - commit original error as informational
l_err->setSev(ERRORLOG::ERRL_SEV_INFORMATIONAL);
ERRORLOG::errlCommit(l_err, HTMGT_COMP_ID);
}
}
TMGT_INF("<<processOccStartStatus()");
} // end processOccStartStatus()
// Set the OCC state
errlHndl_t enableOccActuation(bool i_occActivation)
{
TMGT_INF(">>enableOccActuation(%c)", i_occActivation?'Y':'N');
errlHndl_t l_err = NULL;
TARGETING::Target* sys = NULL;
// If the system is already in safemode then can't talk to OCCs
TARGETING::targetService().getTopLevelTarget(sys);
uint8_t safeMode = 0;
if(sys)
{
sys->tryGetAttr<TARGETING::ATTR_HTMGT_SAFEMODE>(safeMode);
}
if (0 == safeMode)
{
occStateId targetState = OCC_STATE_ACTIVE;
if (false == i_occActivation)
{
targetState = OCC_STATE_OBSERVATION;
}
// Set state for all OCCs
l_err = OccManager::setOccState(targetState);
if (NULL == l_err)
{
TMGT_INF("enableOccActuation: OCC states updated to 0x%02X",
targetState);
}
if (OccManager::occNeedsReset())
{
if (l_err)
{
// Commit setOccState elog since OCCs will be reset
// and recovery attempted.
ERRORLOG::errlCommit(l_err, HTMGT_COMP_ID);
}
TMGT_ERR("enableOccActuation(): OCCs need to be reset");
// Don't pass failed target as OCC should have already
// been marked as failed during the poll.
l_err = OccManager::resetOccs(NULL);
// NOTE: If the system exceeded its reset count and ended up
// in safe mode an error may not be returned here (if a
// failure happened after the first reset attempt).
// This is because the resets are recursive:
// HTMGT calls back into HBRT to initiate the reset, then
// HBRT calls into HTMGT when reset completed
// To detected this condition we need to check for safe mode
// after the recovery attempts and return error if in safe.
if(sys)
{
sys->tryGetAttr<TARGETING::ATTR_HTMGT_SAFEMODE>(safeMode);
}
}
}
if ((NULL == l_err) && safeMode)
{
// Create an elog so the user knows the cmd failed.
TMGT_ERR("enableOccActuation(): System is in safe mode");
/*@
* @errortype
* @reasoncode HTMGT_RC_OCC_CRIT_FAILURE
* @moduleid HTMGT_MOD_ENABLE_OCC_ACTUATION
* @userdata1 OCC activate [1==true][0==false]
* @devdesc Operation not allowed, system is in safe mode
*/
bldErrLog(l_err,
HTMGT_MOD_ENABLE_OCC_ACTUATION,
HTMGT_RC_OCC_CRIT_FAILURE,
0, i_occActivation, 0, safeMode,
ERRORLOG::ERRL_SEV_UNRECOVERABLE);
}
TMGT_INF("<<enableOccActuation() returning 0x%04X",
(l_err==NULL) ? 0 : l_err->reasonCode());
return l_err;
} // end enableOccActuation()
// Notify HTMGT that an OCC has failed and needs to be reset
void processOccReset(TARGETING::Target * i_proc)
{
TMGT_INF(">>processOccReset(0x%p)", i_proc);
errlHndl_t errl = NULL;
TARGETING::Target * failedOccTarget = NULL;
TARGETING::Target* sys = NULL;
TARGETING::targetService().getTopLevelTarget(sys);
uint8_t safeMode = 0;
// If the system is in safemode then ignore request to reset OCCs
if(sys &&
sys->tryGetAttr<TARGETING::ATTR_HTMGT_SAFEMODE>(safeMode) &&
safeMode)
{
return;
}
// Get functional OCC (one per proc)
TARGETING::TargetHandleList pOccs;
getChildChiplets(pOccs, i_proc, TARGETING::TYPE_OCC);
if (pOccs.size() > 0)
{
failedOccTarget = pOccs[0];
}
if(NULL != failedOccTarget)
{
uint32_t huid = failedOccTarget->getAttr<TARGETING::ATTR_HUID>();
TMGT_INF("processOccReset(HUID=0x%08X) called", huid);
}
else
{
uint32_t huid = i_proc->getAttr<TARGETING::ATTR_HUID>();
TMGT_INF("processOccReset: Invalid OCC target (proc huid=0x08X)"
"resetting OCCs anyway",
huid);
/*@
* @errortype
* @reasoncode HTMGT_RC_INVALID_PARAMETER
* @moduleid HTMGT_MOD_PROCESS_OCC_RESET
* @userdata1[0:7] Processor HUID
* @devdesc No OCC target found for proc Target,
*/
bldErrLog(errl,
HTMGT_MOD_PROCESS_OCC_RESET,
HTMGT_RC_INVALID_PARAMETER,
huid, 0, 0, 1,
ERRORLOG::ERRL_SEV_INFORMATIONAL);
// Add HB firmware callout
errl->addProcedureCallout(HWAS::EPUB_PRC_HB_CODE,
HWAS::SRCI_PRIORITY_MED);
ERRORLOG::errlCommit(errl, HTMGT_COMP_ID); // sets errl to NULL
}
errl = OccManager::resetOccs(failedOccTarget);
if(errl)
{
ERRORLOG::errlCommit(errl, HTMGT_COMP_ID); // sets errl to NULL
}
TMGT_INF("<<processOccReset()");
} // end processOccReset()
// Process elog entry from OCC poll response
void occProcessElog(Occ * i_occ,
const uint8_t i_id,
const uint32_t i_address,
const uint16_t i_length)
{
errlHndl_t l_errlHndl = NULL;
// Read data from SRAM (length must be multiple of 8 bytes)
const uint16_t l_length = (i_length + 8) & 0xFFF8;
uint8_t l_sram_data[8 + l_length];
ecmdDataBufferBase l_buffer(l_length*8); // convert to bits
// HBOCC is only defined for HTMGT
#ifdef CONFIG_HTMGT
l_errlHndl = HBOCC::readSRAM(i_occ->getTarget(), i_address, l_buffer);
#endif
if (NULL == l_errlHndl)
{
const uint32_t l_flatSize = l_buffer.flattenSize();
l_buffer.flatten(l_sram_data, l_flatSize);
// Skip 8 byte ecmd header
const occErrlEntry_t *l_occElog=(occErrlEntry_t *)&l_sram_data[8];
TMGT_BIN("OCC ELOG", l_occElog, 64);
const uint32_t l_occSrc = OCCC_COMP_ID | l_occElog->reasonCode;
ERRORLOG::errlSeverity_t l_errlSeverity =
ERRORLOG::ERRL_SEV_INFORMATIONAL;
#if 0
// TODO: RTC 109224 - determine correct severity/actions
// Process Severity
const uint8_t l_occSeverity = l_occElog->severity;
const uint8_t l_occActions = l_occElog->actions;
if (l_occSeverity < OCC_SEV_ACTION_XLATE_SIZE)
{
l_errlSeverity =
occSeverityErrorActionXlate[l_occSeverity].occErrlSeverity;
}
else
{
TMGT_ERR("occProcessElog: Severity translate failure"
" (severity = 0x%02X)", l_occElog->severity);
}
// Process elog Actions
bool l_occReset = false;
elogProcessActions(l_occActions, l_occReset, l_errlSeverity);
if (l_occReset == true)
{
iv_needsReset = true;
UPDATE_SAFE_MODE_REASON(l_occSrc, iv_huid, true);
}
#endif
// Create OCC error log
// NOTE: word 4 (used by extended reason code) to save off OCC
// sub component value which is needed to correctly parse
// srcs which have similar uniqueness
// NOTE: SRC tags are NOT required here as these logs will get
// parsed with the OCC src tags
const occErrlUsrDtls_t *l_usrDtls_ptr = (occErrlUsrDtls_t *)
((uint8_t*)l_occElog+sizeof(occErrlEntry_t)+
(l_occElog->numCallouts * sizeof(occErrlCallout_t)) );
bldErrLog(l_errlHndl,
(htmgtModuleId)(l_usrDtls_ptr->modId & 0x00FF),
(htmgtReasonCode)l_occSrc, // occ reason code
l_usrDtls_ptr->userData1,
l_usrDtls_ptr->userData2,
l_usrDtls_ptr->userData3,
((l_usrDtls_ptr->modId & 0xFF00) << 16 ) |
l_occElog->userData4, // extended reason code
l_errlSeverity);
#if 0
// TODO: RTC 109224
// Add callout information
bool l_bad_fru_data = false;
uint8_t l_callout_num = 0;
if (! ((ERRL_SEV_INFORMATIONAL == l_errlSeverity) &&
(TMGT_ERRL_ACTIONS_MANUFACTURING_ERROR & l_occActions)) )
{
// Only add callouts if this is MFG error and system not in
// MFG (in MFG severity would not be Info)
uint8_t l_index = 0;
uint8_t l_count = 1;
const uint8_t l_max_callout = l_occElog->numCallouts;
// The beginning address of callout data
l_index = sizeof(occErrlEntry_t);
do {
occErrlCallout_t *l_callout_ptr = NULL;
l_callout_ptr = (occErrlCallout_t *)
((uint8_t*)l_occElog+l_index);
if (l_callout_ptr->type != 0)
{
srciPriority l_priority;
bool l_success = true;
l_success =
elogXlateSrciPriority(l_callout_ptr->priority,
l_priority);
if (l_success == true)
{
l_success = elogAddCallout(l_errlHndl,
l_errlSeverity,
l_priority,
*l_callout_ptr,
l_callout_num);
if (l_success == false)
{
l_bad_fru_data = true;
}
}
else
{
l_bad_fru_data = true;
TMGT_ERR("occProcessElog: Priority translate"
" failure (priority = 0x%02X)",
l_callout_ptr->priority);
}
l_index += sizeof(occErrlCallout_t);
} // if (l_type != 0)
else
{ // make sure all the remaining callout data are zeros,
// otherwise mark bad fru data
uint8_t *l_ptr = (uint8_t*)l_occElog+l_index;
uint8_t l_len = (l_max_callout-l_count+1)*
sizeof(occErrlCallout_t);
while (l_len != 0)
{
if (*l_ptr != 0x00)
{
TMGT_ERR("occProcessElog: The remaining"
" callout data should be all zeros");
l_bad_fru_data = true;
break;
}
l_len--;
l_ptr++;
}
break;
}
l_count++;
} while (l_count <= l_max_callout);
}
else
{
TMGT_ERR("MFG error found outside MFG; callouts will not be"
" added to log (OCC severity=0x%02X, actions=0x%02X)",
l_occSeverity, l_occActions);
const uint8_t l_callout_length = l_occElog->numCallouts * 12;
const char *l_callout_ptr = (char *)((uint8_t*)l_occElog+
sizeof(occErrlEntry_t));
// Add raw callout data from the OCC
l_errlHndl->addUsrDtls(l_callout_ptr,
l_callout_length,
TMGT_COMP_ID,
TMGT_VERSION,
TMGT_ERROR_DATA_TYPE);
}
// Any bad fru data found ?
errlHndl_t l_errlHndl2 = NULL;
if (l_bad_fru_data == true)
{
/*@
* @errortype
* @refcode LIC_REFCODE
* @subsys EPUB_FIRMWARE_SP
* @reasoncode HTMGT_RC_OCC_ERROR_LOG
* @moduleid HTMGT_MOD_BAD_FRU_CALLOUTS
* @userdata1 OCC elog id
* @userdata2 Number of good callouts
* @devdesc Bad FRU data received in OCC error log
*/
bldErrLog(l_errlHndl2, HTMGT_MOD_BAD_FRU_CALLOUTS,
HTMGT_RC_OCC_ERROR_LOG,
i_id, l_callout_num, 0, 0, ERRL_SEV_INFORMATIONAL);
ERRORLOG::errlCommit(l_errlHndl2, HTMGT_COMP_ID);
}
// Check callout number and severity
if ((l_callout_num == 0) &&
(l_errlSeverity != ERRL_SEV_INFORMATIONAL))
{
TMGT_ERR("occProcessElog: No FRU callouts found for OCC%d"
" elog_id:0x%02X, severity:0x%0X",
iv_instance, i_id, l_errlSeverity);
/*@
* @errortype
* @refcode LIC_REFCODE
* @subsys EPUB_FIRMWARE_SP
* @reasoncode HTMGT_RC_OCC_ERROR_LOG
* @moduleid HTMGT_MOD_MISMATCHING_SEVERITY
* @userdata1 OCC elog id
* @userdata2 OCC severity
* @userdata3
* @userdata4
* @devdesc No FRU callouts found for non-info OCC Error Log
*/
bldErrLog(l_errlHndl2, HTMGT_MOD_MISMATCHING_SEVERITY,
HTMGT_RC_OCC_ERROR_LOG,
i_id, l_errlSeverity, 0, 0, ERRL_SEV_INFORMATIONAL);
ERRORLOG::errlCommit(l_errlHndl2, HTMGT_COMP_ID);
}
#endif
// Add full OCC error log data as a User Details section
l_errlHndl->addFFDC(OCCC_COMP_ID,
l_occElog,
i_length,
1, // version
0); // subsection
#if 0
// TODO: RTC 109224
// Add additional data
addTmgtElogData(l_errlHndl);
addThermalElogData(l_errlHndl);
#endif
// Commit Error (or terminate if required)
ERRORLOG::errlCommit(l_errlHndl, HTMGT_COMP_ID);
// Clear elog
const uint8_t l_cmdData[1] = {i_id};
OccCmd l_cmd(i_occ, OCC_CMD_CLEAR_ERROR_LOG,
sizeof(l_cmdData), l_cmdData);
l_errlHndl = l_cmd.sendOccCmd();
if (l_errlHndl != NULL)
{
TMGT_ERR("occProcessElog: Failed to clear elog id %d to"
" OCC%d (rc=0x%04X)",
i_id, i_occ, l_errlHndl->reasonCode());
ERRORLOG::errlCommit(l_errlHndl, HTMGT_COMP_ID);
}
}
else
{
TMGT_ERR("occProcessElog: Unable to read elog %d from SRAM"
" address (0x%08X) length (0x%04X), rc=0x%04X",
i_id, i_address, i_length, l_errlHndl->reasonCode());
ERRORLOG::errlCommit(l_errlHndl, HTMGT_COMP_ID);
}
} // end Occ::occProcessElog()
// Send pass-thru command to HTMGT
errlHndl_t passThruCommand(uint16_t i_cmdLength,
uint8_t * i_cmdData,
uint16_t & o_rspLength,
uint8_t * o_rspData)
{
errlHndl_t err = NULL;
htmgtReasonCode failingSrc = HTMGT_RC_NO_ERROR;
o_rspLength = 0;
if ((i_cmdLength > 0) && (NULL != i_cmdData))
{
switch (i_cmdData[0])
{
case PASSTHRU_OCC_STATUS:
TMGT_INF("passThruCommand: OCC Status");
OccManager::getOccData(o_rspLength, o_rspData);
break;
case PASSTHRU_GENERATE_MFG_PSTATE:
if (i_cmdLength == 1)
{
TMGT_INF("passThruCommand: Generate MFG pstate tables",
i_cmdData[1]);
err = genPstateTables(false);
}
else
{
TMGT_ERR("passThruCommand: invalid generate pstate "
"command length %d", i_cmdLength);
/*@
* @errortype
* @reasoncode HTMGT_RC_INVALID_LENGTH
* @moduleid HTMGT_MOD_PASS_THRU
* @userdata1 command data[0-7]
* @userdata2 command data length
* @devdesc Invalid pass thru command data length
*/
failingSrc = HTMGT_RC_INVALID_LENGTH;
}
break;
case PASSTHRU_LOAD_PSTATE:
if (i_cmdLength == 2)
{
const uint8_t pstateType = i_cmdData[1];
if ((0 == pstateType) || (1 == pstateType))
{
TMGT_INF("passThruCommand: Load pstate tables "
"(type: %d)", pstateType);
// 0 = Normal Pstate Tables
err = OccManager::loadPstates(0 == pstateType);
}
else
{
TMGT_ERR("passThruCommand: invalid pstate type "
"specified: %d", pstateType);
/*@
* @errortype
* @reasoncode HTMGT_RC_INVALID_PARAMETER
* @moduleid HTMGT_MOD_PASS_THRU
* @userdata1 command data[0-7]
* @userdata2 command data length
* @devdesc Invalid load pstate table type
*/
failingSrc = HTMGT_RC_INVALID_PARAMETER;
}
}
else
{
TMGT_ERR("passThruCommand: invalid load pstate "
"command length %d", i_cmdLength);
failingSrc = HTMGT_RC_INVALID_LENGTH;
}
break;
default:
TMGT_ERR("passThruCommand: Invalid command 0x%08X "
"(%d bytes)", UINT32_GET(i_cmdData), i_cmdLength);
/*@
* @errortype
* @reasoncode HTMGT_RC_INVALID_DATA
* @moduleid HTMGT_MOD_PASS_THRU
* @userdata1 command data[0-7]
* @userdata2 command data length
* @devdesc Invalid pass thru command
*/
failingSrc = HTMGT_RC_INVALID_DATA;
break;
}
if ((HTMGT_RC_NO_ERROR != failingSrc) && (NULL == err))
{
bldErrLog(err, HTMGT_MOD_PASS_THRU,
failingSrc,
UINT32_GET(i_cmdData),
UINT32_GET(&i_cmdData[4]),
0, i_cmdLength,
ERRORLOG::ERRL_SEV_INFORMATIONAL);
}
}
return err;
} // end passThruCommand()
// Send a poll command to all OCCs
errlHndl_t sendOccPoll(const bool i_flushAllErrors)
{
errlHndl_t l_err = NULL;
uint8_t * l_poll_rsp = NULL;
// Loop through all functional OCCs
std::vector<Occ*> occList = occMgr::instance().getOccArray();
for (std::vector<Occ*>::iterator itr = occList.begin();
(itr < occList.end()) && (NULL == l_err);
++itr)
{
Occ * occ = (*itr);
const uint8_t occInstance = occ->getInstance();
bool continuePolling = false;
do
{
// create 1 byte buffer for poll command data
const uint8_t l_cmdData[1] = { 0x10 /*version*/ };
OccCmd cmd(occ, OCC_CMD_POLL, sizeof(l_cmdData), l_cmdData);
l_err = cmd.sendOccCmd();
if (l_err != NULL)
{
// Poll failed
TMGT_ERR("sendOccPoll: OCC%d poll failed with rc=0x%04X",
occInstance, l_err->reasonCode());
}
else
{
// Poll succeeded, check response
uint32_t l_poll_rsp_size = cmd.getResponseData(l_poll_rsp);
if (l_poll_rsp_size >= OCC_POLL_DATA_MIN_SIZE)
{
if (i_flushAllErrors)
{
const occPollRspStruct_t *currentPollRsp =
(occPollRspStruct_t *) l_poll_rsp;
if (currentPollRsp->errorId != 0)
{
// An error was returned, keep polling OCC
continuePolling = true;
}
else
{
continuePolling = false;
}
}
occ->pollRspHandler(l_poll_rsp, l_poll_rsp_size);
}
else
{
TMGT_ERR("sendOccPoll: OCC%d poll command response "
"failed with invalid data length %d",
occInstance, l_poll_rsp_size);
/*@
* @errortype
* @reasoncode HTMGT_RC_INVALID_LENGTH
* @moduleid HTMGT_MOD_OCC_POLL
* @userdata1 OCC instance
* @devdesc Invalid POLL response length
*/
bldErrLog(l_err, HTMGT_MOD_OCC_POLL,
HTMGT_RC_INVALID_LENGTH,
occInstance, 0, 0, 0,
ERRORLOG::ERRL_SEV_INFORMATIONAL);
}
}
}
while (continuePolling);
} // for each OCC
return l_err;
} // end sendOccPoll()