diff options
Diffstat (limited to 'src/occ_405/amec/amec_sensors_ocmb.c')
| -rw-r--r-- | src/occ_405/amec/amec_sensors_ocmb.c | 364 |
1 files changed, 215 insertions, 149 deletions
diff --git a/src/occ_405/amec/amec_sensors_ocmb.c b/src/occ_405/amec/amec_sensors_ocmb.c index a2cd4a6..8ffbefa 100644 --- a/src/occ_405/amec/amec_sensors_ocmb.c +++ b/src/occ_405/amec/amec_sensors_ocmb.c @@ -110,21 +110,21 @@ void amec_update_ocmb_sensors(uint8_t i_membuf) // End Function Specification void amec_update_ocmb_dimm_dts_sensors(OcmbMemData * i_sensor_cache, uint8_t i_membuf) { +// confirmed ok to use same values for all types (internal mc, dimm, external mc, pmic...) #define MIN_VALID_DIMM_TEMP 1 #define MAX_VALID_DIMM_TEMP 125 //according to Mike Pardiek 04/23/2019 -#define MAX_MEM_TEMP_CHANGE 2 +#define MAX_MEM_TEMP_CHANGE 4 - uint32_t k, l_hottest_dimm_temp; + uint32_t k; uint16_t l_dts[NUM_DIMMS_PER_OCMB] = {0}; - uint32_t l_hottest_dimm_loc = NUM_DIMMS_PER_OCMB; int32_t l_dimm_temp, l_prev_temp; static uint8_t L_ran_once[MAX_NUM_OCMBS] = {FALSE}; - static bool L_ot_traced[MAX_NUM_OCMBS][NUM_DIMMS_PER_OCMB] = {{false}}; - // Harvest thermal data for all dimms + // Harvest thermal data for memory thermal sensors that are enabled and being used for(k=0; k < NUM_DIMMS_PER_OCMB; k++) { - if(!CENTAUR_SENSOR_ENABLED(i_membuf, k)) + if( (!CENTAUR_SENSOR_ENABLED(i_membuf, k)) || + (g_amec->proc[0].memctl[i_membuf].centaur.dimm_temps[k].temp_fru_type == DATA_FRU_NOT_USED) ) { continue; } @@ -250,48 +250,17 @@ void amec_update_ocmb_dimm_dts_sensors(OcmbMemData * i_sensor_cache, uint8_t i_m } } - //Check if at or above the error temperature - if(l_dts[k] >= g_amec->thermaldimm.ot_error) - { - //Set a bit so that this dimm can be called out by the thermal thread - G_dimm_overtemp_bitmap.bytes[i_membuf] |= (DIMM_SENSOR0 >> k); - // trace first time OT per DIMM - if( !L_ot_traced[i_membuf][k] ) - { - TRAC_ERR("amec_update_ocmb_dimm_dts_sensors: Mem Buf[%d] DIMM[%d] reached error temp[%d]. current[%d]", - i_membuf, - k, - g_amec->thermaldimm.ot_error, - l_dts[k]); - L_ot_traced[i_membuf][k] = true; - } - } + //Check for over temperature must be done by type and will be checked + // in amec_update_ocmb_temp_sensors() which happens after all OCMBs have been read } - // Find hottest temperature from all DIMMs for this centaur - for(l_hottest_dimm_temp = 0, k = 0; k < NUM_DIMMS_PER_OCMB; k++) + // update the current temperatures + for(k = 0; k < NUM_DIMMS_PER_OCMB; k++) { - if(l_dts[k] > l_hottest_dimm_temp) - { - l_hottest_dimm_temp = l_dts[k]; - l_hottest_dimm_loc = k; - } g_amec->proc[0].memctl[i_membuf].centaur.dimm_temps[k].cur_temp = l_dts[k]; } - amec_centaur_t* l_centaur_ptr = &g_amec->proc[0].memctl[i_membuf].centaur; - - //only update location if hottest dimm temp is greater than previous maximum - if(l_hottest_dimm_temp > l_centaur_ptr->tempdimmax.sample_max) - { - sensor_update(&l_centaur_ptr->locdimmax, l_hottest_dimm_loc); - } - - //update the max dimm temperature sensor for this centaur - sensor_update(&l_centaur_ptr->tempdimmax, l_hottest_dimm_temp); - L_ran_once[i_membuf] = TRUE; - AMEC_DBG("Centaur[%d]: HotDimm=%d\n",i_membuf,l_hottest_dimm_temp); } // Function Specification @@ -317,129 +286,223 @@ void amec_update_ocmb_dts_sensors(OcmbMemData * i_sensor_cache, uint8_t i_membuf fru_temp_t* l_fru = &g_amec->proc[0].memctl[i_membuf].centaur.centaur_hottest; - l_prev_temp = l_fru->cur_temp; - if(!l_prev_temp) + // Internal DTS sensor is either for internal memctrl or not being used + // ignore the internal sensor if it isn't marked for internal memctrl + if(l_fru->temp_fru_type == DATA_FRU_CENTAUR) { - l_prev_temp = l_sens_temp; - } - - //Check DTS status bits - if(i_sensor_cache->status.fields.ubdts0_valid && - (!i_sensor_cache->status.fields.ubdts0_err)) - { - //make sure temperature is within a 'reasonable' range. - if(l_sens_temp < MIN_VALID_MEMBUF_TEMP || - l_sens_temp > MAX_VALID_MEMBUF_TEMP) - { - //set a flag so that if we end up logging an error we have something to debug why - l_fru->flags |= FRU_TEMP_OUT_OF_RANGE; - l_dts = l_prev_temp; - } - else - { - //don't allow temp to change more than is reasonable since last read - if(l_sens_temp > (l_prev_temp + MAX_MEM_TEMP_CHANGE)) - { - l_dts = l_prev_temp + MAX_MEM_TEMP_CHANGE; - if(!l_fru->flags) - { - TRAC_INFO("membuf temp rose faster than reasonable: membuf[%d] prev[%d] cur[%d]", - i_membuf, l_prev_temp, l_sens_temp); - l_fru->flags |= FRU_TEMP_FAST_CHANGE; - } - } - else if (l_sens_temp < (l_prev_temp - MAX_MEM_TEMP_CHANGE)) - { - l_dts = l_prev_temp - MAX_MEM_TEMP_CHANGE; - if(!l_fru->flags) - { - TRAC_INFO("membuf temp fell faster than reasonable: cent[%d] prev[%d] cur[%d]", - i_membuf, l_prev_temp, l_sens_temp); - l_fru->flags |= FRU_TEMP_FAST_CHANGE; - } - } - else //reasonable amount of change occurred - { - l_dts = l_sens_temp; - l_fru->flags &= ~FRU_TEMP_FAST_CHANGE; - } - - //Notify thermal thread that temperature has been updated - G_cent_temp_updated_bitmap |= (CENTAUR0_PRESENT_MASK >> i_membuf); - - //clear error flags - l_fru->flags &= FRU_TEMP_FAST_CHANGE; - } + l_prev_temp = l_fru->cur_temp; + if(!l_prev_temp) + { + l_prev_temp = l_sens_temp; + } + + //Check DTS status bits + if(i_sensor_cache->status.fields.ubdts0_valid && + (!i_sensor_cache->status.fields.ubdts0_err)) + { + //make sure temperature is within a 'reasonable' range. + if(l_sens_temp < MIN_VALID_MEMBUF_TEMP || + l_sens_temp > MAX_VALID_MEMBUF_TEMP) + { + //set a flag so that if we end up logging an error we have something to debug why + l_fru->flags |= FRU_TEMP_OUT_OF_RANGE; + l_dts = l_prev_temp; + } + else + { + //don't allow temp to change more than is reasonable since last read + if(l_sens_temp > (l_prev_temp + MAX_MEM_TEMP_CHANGE)) + { + l_dts = l_prev_temp + MAX_MEM_TEMP_CHANGE; + if(!l_fru->flags) + { + TRAC_INFO("membuf temp rose faster than reasonable: membuf[%d] prev[%d] cur[%d]", + i_membuf, l_prev_temp, l_sens_temp); + l_fru->flags |= FRU_TEMP_FAST_CHANGE; + } + } + else if (l_sens_temp < (l_prev_temp - MAX_MEM_TEMP_CHANGE)) + { + l_dts = l_prev_temp - MAX_MEM_TEMP_CHANGE; + if(!l_fru->flags) + { + TRAC_INFO("membuf temp fell faster than reasonable: cent[%d] prev[%d] cur[%d]", + i_membuf, l_prev_temp, l_sens_temp); + l_fru->flags |= FRU_TEMP_FAST_CHANGE; + } + } + else //reasonable amount of change occurred + { + l_dts = l_sens_temp; + l_fru->flags &= ~FRU_TEMP_FAST_CHANGE; + } + + //Notify thermal thread that temperature has been updated + G_cent_temp_updated_bitmap |= (CENTAUR0_PRESENT_MASK >> i_membuf); + + //clear error flags + l_fru->flags &= FRU_TEMP_FAST_CHANGE; + } + } + else //status was INVALID + { + if(L_ran_once[i_membuf]) + { + //Trace the error if we haven't traced it already for this sensor + if(!(l_fru->flags & FRU_SENSOR_STATUS_INVALID) && + i_sensor_cache->status.fields.ubdts0_err) + { + TRAC_ERR("Membuf %d temp sensor error.", i_membuf); + } + + l_fru->flags |= FRU_SENSOR_STATUS_INVALID; + } + + //use last temperature + l_dts = l_prev_temp; + } + + L_ran_once[i_membuf] = TRUE; + + //Check if at or above the error temperature + if(l_dts >= g_amec->thermalcent.ot_error) + { + //Set a bit so that this dimm can be called out by the thermal thread + G_cent_overtemp_bitmap |= (CENTAUR0_PRESENT_MASK >> i_membuf); + } + + // Update Interim Data - later this will get picked up to form centaur sensor + l_fru->cur_temp = l_dts; + + AMEC_DBG("Membuf[%d]: HotMembuf=%d\n",i_membuf,l_dts); } - else //status was INVALID + else // internal sensor not being used { - if(L_ran_once[i_membuf]) - { - //Trace the error if we haven't traced it already for this sensor - if(!(l_fru->flags & FRU_SENSOR_STATUS_INVALID) && - i_sensor_cache->status.fields.ubdts0_err) - { - TRAC_ERR("Membuf %d temp sensor error.", i_membuf); - } + // make sure temperature is 0 indicating not present + l_fru->cur_temp = 0; - l_fru->flags |= FRU_SENSOR_STATUS_INVALID; - } + //Notify thermal thread that temperature has been updated so no timeout error is logged + G_cent_temp_updated_bitmap |= CENTAUR0_PRESENT_MASK >> i_membuf; - //use last temperature - l_dts = l_prev_temp; + //clear error flags + l_fru->flags = 0; } - - L_ran_once[i_membuf] = TRUE; - - //Check if at or above the error temperature - if(l_dts >= g_amec->thermalcent.ot_error) - { - //Set a bit so that this dimm can be called out by the thermal thread - G_cent_overtemp_bitmap |= (CENTAUR0_PRESENT_MASK >> i_membuf); - } - - // Update Interim Data - later this will get picked up to form centaur sensor - g_amec->proc[0].memctl[i_membuf].centaur.centaur_hottest.cur_temp = l_dts; - - AMEC_DBG("Membuf[%d]: HotMembuf=%d\n",i_membuf,l_dts); } // Function Specification // // Name: amec_update_ocmb_temp_sensors // -// Description: Updates thermal sensors that have data grabbed by the centaur. +// Description: Updates thermal sensors to give summary (across all OCMBs) for each mem type // // Thread: RealTime Loop // // End Function Specification void amec_update_ocmb_temp_sensors(void) { - uint32_t k; + uint32_t k, l_dimm; uint32_t l_hot_dimm = 0; uint32_t l_hot_mb = 0; + uint32_t l_hot_mb_dimm = 0; + uint32_t l_hot_pmic = 0; + uint32_t l_hot_ext_mb = 0; + uint8_t l_ot_error = 0; + uint8_t l_cur_temp = 0; + uint8_t l_fru_type = DATA_FRU_NOT_USED; + static bool L_ot_traced[MAX_NUM_OCMBS][NUM_DIMMS_PER_OCMB] = {{false}}; - // ----------------------------------------------------------- - // Find hottest temperature from all membufs for this Proc chip - // Find hottest temperature from all DIMMs for this Proc chip - // ----------------------------------------------------------- for(k=0; k < MAX_NUM_OCMBS; k++) { - if(g_amec->proc[0].memctl[k].centaur.centaur_hottest.cur_temp > l_hot_mb) + // Find hottest temperature from all internal membufs for this Proc chip + // make sure the type is "CENTAUR" i.e. internal memory controller temp + if( (g_amec->proc[0].memctl[k].centaur.centaur_hottest.temp_fru_type == DATA_FRU_CENTAUR) && + (g_amec->proc[0].memctl[k].centaur.centaur_hottest.cur_temp > l_hot_mb) ) { l_hot_mb = g_amec->proc[0].memctl[k].centaur.centaur_hottest.cur_temp; } - if(g_amec->proc[0].memctl[k].centaur.tempdimmax.sample > l_hot_dimm) + + // process each of the thermal sensors (stored as "dimm" temps) + // based on what type they are for and finding the hottest for each type + for(l_dimm=0; l_dimm < NUM_DIMMS_PER_OCMB; l_dimm++) { - l_hot_dimm = g_amec->proc[0].memctl[k].centaur.tempdimmax.sample; - } - } + l_fru_type = g_amec->proc[0].memctl[k].centaur.dimm_temps[l_dimm].temp_fru_type; + l_cur_temp = g_amec->proc[0].memctl[k].centaur.dimm_temps[l_dimm].cur_temp; + + switch(l_fru_type) + { + case DATA_FRU_DIMM: + l_ot_error = g_amec->thermaldimm.ot_error; + if(g_amec->proc[0].memctl[k].centaur.dimm_temps[l_dimm].cur_temp > l_hot_dimm) + { + l_hot_dimm = g_amec->proc[0].memctl[k].centaur.dimm_temps[l_dimm].cur_temp; + } + break; + + case DATA_FRU_MEMCTRL_DRAM: + l_ot_error = g_amec->thermalmcdimm.ot_error; + if(g_amec->proc[0].memctl[k].centaur.dimm_temps[l_dimm].cur_temp > l_hot_mb_dimm) + { + l_hot_mb_dimm = g_amec->proc[0].memctl[k].centaur.dimm_temps[l_dimm].cur_temp; + } + break; + + case DATA_FRU_PMIC: + l_ot_error = g_amec->thermalpmic.ot_error; + if(g_amec->proc[0].memctl[k].centaur.dimm_temps[l_dimm].cur_temp > l_hot_pmic) + { + l_hot_pmic = g_amec->proc[0].memctl[k].centaur.dimm_temps[l_dimm].cur_temp; + } + break; + + case DATA_FRU_MEMCTRL_EXT: + l_ot_error = g_amec->thermalmcext.ot_error; + if(g_amec->proc[0].memctl[k].centaur.dimm_temps[l_dimm].cur_temp > l_hot_ext_mb) + { + l_hot_ext_mb = g_amec->proc[0].memctl[k].centaur.dimm_temps[l_dimm].cur_temp; + } + break; + + case DATA_FRU_NOT_USED: + default: + // ignore reading + l_ot_error = 0; + break; + } // end switch fru type + + // check if this "DIMM" sensor is over its error temperature + if( l_ot_error && (l_cur_temp >= l_ot_error) ) + { + //Set a bit so that this sensor can be called out by the thermal thread + G_dimm_overtemp_bitmap.bytes[k] |= (DIMM_SENSOR0 >> l_dimm); + // trace first time OT per DIMM DTS sensor + if( !L_ot_traced[k][l_dimm] ) + { + TRAC_ERR("amec_update_ocmb_temp_sensors: OCMB[%d] DTS[%d] type[0x%02X] reached error temp[%d]. current[%d]", + k, + l_dimm, + l_fru_type, + l_ot_error, + l_cur_temp); + L_ot_traced[k][l_dimm] = true; + } + } + } // end for each "dimm" thermal sensor + } // end for each OCMB + sensor_update(&g_amec->proc[0].tempcent,l_hot_mb); AMEC_DBG("HotMembuf=%d\n",l_hot_mb); sensor_update(&g_amec->proc[0].tempdimmthrm,l_hot_dimm); AMEC_DBG("HotDimm=%d\n",l_hot_dimm); + sensor_update(&g_amec->proc[0].tempmcdimmthrm,l_hot_mb_dimm); + AMEC_DBG("HotMCDimm=%d\n",l_hot_mb_dimm); + + sensor_update(&g_amec->proc[0].temppmicthrm,l_hot_pmic); + AMEC_DBG("HotPmic=%d\n",l_hot_pmic); + + sensor_update(&g_amec->proc[0].tempmcextthrm,l_hot_ext_mb); + AMEC_DBG("HotExternalMembuf=%d\n",l_hot_ext_mb); } @@ -510,24 +573,27 @@ void amec_perfcount_ocmb_getmc( OcmbMemData * i_sensor_cache, g_amec->proc[0].memctl[i_membuf].centaur.portpair[0].perf.memread2ms = tempreg; - // Go after second MC performance counter (power ups and activations) - tempu = l_sensor_cache->mba_act; - templ = l_sensor_cache->mba_powerups; - - // ------------------------------------------------------------ - // Sensor: MRDMx (0.01 Mrps) Memory read requests per sec - // ------------------------------------------------------------ - tempreg = g_amec->proc[0].memctl[i_membuf].centaur.portpair[0].perf.memread2ms; - tempreg += g_amec->proc[0].memctl[i_membuf].centaur.portpair[1].perf.memread2ms; - sensor_update( (&(g_amec->proc[0].memctl[i_membuf].mrd)), tempreg); - - // ------------------------------------------------------------- - // Sensor: MWRMx (0.01 Mrps) Memory write requests per sec - // ------------------------------------------------------------- - tempreg = g_amec->proc[0].memctl[i_membuf].centaur.portpair[0].perf.memwrite2ms; - tempreg += g_amec->proc[0].memctl[i_membuf].centaur.portpair[1].perf.memwrite2ms; - sensor_update( (&(g_amec->proc[0].memctl[i_membuf].mwr)), tempreg); - + // Due to limited SRAM only have sensor support for first 12 mem buffs + if(i_membuf < 12) + { + // Go after second MC performance counter (power ups and activations) + tempu = l_sensor_cache->mba_act; + templ = l_sensor_cache->mba_powerups; + + // ------------------------------------------------------------ + // Sensor: MRDMx (0.01 Mrps) Memory read requests per sec + // ------------------------------------------------------------ + tempreg = g_amec->proc[0].memctl[i_membuf].centaur.portpair[0].perf.memread2ms; + tempreg += g_amec->proc[0].memctl[i_membuf].centaur.portpair[1].perf.memread2ms; + sensor_update( (&(g_amec->proc[0].memctl[i_membuf].mrd)), tempreg); + + // ------------------------------------------------------------- + // Sensor: MWRMx (0.01 Mrps) Memory write requests per sec + // ------------------------------------------------------------- + tempreg = g_amec->proc[0].memctl[i_membuf].centaur.portpair[0].perf.memwrite2ms; + tempreg += g_amec->proc[0].memctl[i_membuf].centaur.portpair[1].perf.memwrite2ms; + sensor_update( (&(g_amec->proc[0].memctl[i_membuf].mwr)), tempreg); + } return; } |
