[PATCH 1/2] drm/amd/pm: update the gpu metrics data retrieving for Sienna Cichlid
Lazar, Lijo
lijo.lazar at amd.com
Tue Jun 29 13:38:27 UTC 2021
On 6/25/2021 1:42 PM, Evan Quan wrote:
> Due to the structure layout change: "uint32_t ThrottlerStatus" -> "
> uint8_t ThrottlingPercentage[THROTTLER_COUNT]".
>
> Change-Id: Id5c148b0584d972ae73fb9d7347a312944cec13d
> Signed-off-by: Evan Quan <evan.quan at amd.com>
> ---
> .../pm/inc/smu11_driver_if_sienna_cichlid.h | 63 ++++-
> .../amd/pm/swsmu/smu11/sienna_cichlid_ppt.c | 234 ++++++++++++------
> 2 files changed, 222 insertions(+), 75 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/pm/inc/smu11_driver_if_sienna_cichlid.h b/drivers/gpu/drm/amd/pm/inc/smu11_driver_if_sienna_cichlid.h
> index 61c87c39be80..0b916a1933df 100644
> --- a/drivers/gpu/drm/amd/pm/inc/smu11_driver_if_sienna_cichlid.h
> +++ b/drivers/gpu/drm/amd/pm/inc/smu11_driver_if_sienna_cichlid.h
> @@ -211,6 +211,7 @@ typedef enum {
> #define THROTTLER_FIT_BIT 17
> #define THROTTLER_PPM_BIT 18
> #define THROTTLER_APCC_BIT 19
> +#define THROTTLER_COUNT 20
>
> // FW DState Features Control Bits
> // FW DState Features Control Bits
> @@ -1406,7 +1407,67 @@ typedef struct {
> } SmuMetrics_t;
>
> typedef struct {
> - SmuMetrics_t SmuMetrics;
> + uint32_t CurrClock[PPCLK_COUNT];
> +
> + uint16_t AverageGfxclkFrequencyPreDs;
> + uint16_t AverageGfxclkFrequencyPostDs;
> + uint16_t AverageFclkFrequencyPreDs;
> + uint16_t AverageFclkFrequencyPostDs;
> + uint16_t AverageUclkFrequencyPreDs ;
> + uint16_t AverageUclkFrequencyPostDs ;
> +
> +
> + uint16_t AverageGfxActivity ;
> + uint16_t AverageUclkActivity ;
> + uint8_t CurrSocVoltageOffset ;
> + uint8_t CurrGfxVoltageOffset ;
> + uint8_t CurrMemVidOffset ;
> + uint8_t Padding8 ;
> + uint16_t AverageSocketPower ;
> + uint16_t TemperatureEdge ;
> + uint16_t TemperatureHotspot ;
> + uint16_t TemperatureMem ;
> + uint16_t TemperatureVrGfx ;
> + uint16_t TemperatureVrMem0 ;
> + uint16_t TemperatureVrMem1 ;
> + uint16_t TemperatureVrSoc ;
> + uint16_t TemperatureLiquid0 ;
> + uint16_t TemperatureLiquid1 ;
> + uint16_t TemperaturePlx ;
> + uint16_t Padding16 ;
> + uint32_t AccCnt ;
> + uint8_t ThrottlingPercentage[THROTTLER_COUNT];
> +
> +
> + uint8_t LinkDpmLevel;
> + uint8_t CurrFanPwm;
> + uint16_t CurrFanSpeed;
> +
> + //BACO metrics, PMFW-1721
> + //metrics for D3hot entry/exit and driver ARM msgs
> + uint8_t D3HotEntryCountPerMode[D3HOT_SEQUENCE_COUNT];
> + uint8_t D3HotExitCountPerMode[D3HOT_SEQUENCE_COUNT];
> + uint8_t ArmMsgReceivedCountPerMode[D3HOT_SEQUENCE_COUNT];
> +
> + //PMFW-4362
> + uint32_t EnergyAccumulator;
> + uint16_t AverageVclk0Frequency ;
> + uint16_t AverageDclk0Frequency ;
> + uint16_t AverageVclk1Frequency ;
> + uint16_t AverageDclk1Frequency ;
> + uint16_t VcnActivityPercentage ; //place holder, David N. to provide full sequence
> + uint8_t PcieRate ;
> + uint8_t PcieWidth ;
> + uint16_t AverageGfxclkFrequencyTarget;
> + uint16_t Padding16_2;
> +
> +} SmuMetrics_V2_t;
> +
> +typedef struct {
> + union {
> + SmuMetrics_t SmuMetrics;
> + SmuMetrics_V2_t SmuMetrics_V2;
> + };
> uint32_t Spare[1];
>
> // Padding - ignore
> diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu11/sienna_cichlid_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu11/sienna_cichlid_ppt.c
> index 0c3407025eb2..f882c6756bf0 100644
> --- a/drivers/gpu/drm/amd/pm/swsmu/smu11/sienna_cichlid_ppt.c
> +++ b/drivers/gpu/drm/amd/pm/swsmu/smu11/sienna_cichlid_ppt.c
> @@ -80,6 +80,13 @@
> (*member) = (smu->smu_table.driver_pptable + offsetof(PPTable_t, field));\
> } while(0)
>
> +#define GET_METRICS_MEMBER(field, member) do { \
> + if ((smu->adev->asic_type == CHIP_SIENNA_CICHLID) && (smu->smc_fw_version >= 0x3A4300)) \
> + (*member) = ((void *)(&(((SmuMetricsExternal_t *)(smu->smu_table.metrics_table))->SmuMetrics_V2)) + offsetof(SmuMetrics_V2_t, field)); \
> + else \
> + (*member) = ((void *)(&(((SmuMetricsExternal_t *)(smu->smu_table.metrics_table))->SmuMetrics)) + offsetof(SmuMetrics_t, field)); \
> +} while(0)
> +
> static int get_table_size(struct smu_context *smu)
> {
> if (smu->adev->asic_type == CHIP_BEIGE_GOBY)
> @@ -489,13 +496,33 @@ static int sienna_cichlid_tables_init(struct smu_context *smu)
> return -ENOMEM;
> }
>
> +static uint32_t sienna_cichlid_get_throttler_status_locked(struct smu_context *smu)
> +{
> + struct smu_table_context *smu_table= &smu->smu_table;
> + SmuMetricsExternal_t *metrics_ext =
> + (SmuMetricsExternal_t *)(smu_table->metrics_table);
> + uint32_t throttler_status = 0;
> + int i;
> +
> + if ((smu->adev->asic_type == CHIP_SIENNA_CICHLID) &&
> + (smu->smc_fw_version >= 0x3A4300)) {
> + for (i = 0; i < THROTTLER_COUNT; i++) {
> + if (metrics_ext->SmuMetrics_V2.ThrottlingPercentage[i])
> + throttler_status |= 1U << i;
> + }
> + } else {
> + throttler_status = metrics_ext->SmuMetrics.ThrottlerStatus;
> + }
> +
> + return throttler_status;
> +}
> +
> static int sienna_cichlid_get_smu_metrics_data(struct smu_context *smu,
> MetricsMember_t member,
> uint32_t *value)
> {
> - struct smu_table_context *smu_table= &smu->smu_table;
> - SmuMetrics_t *metrics =
> - &(((SmuMetricsExternal_t *)(smu_table->metrics_table))->SmuMetrics);
> + uint32_t *data_u32;
> + uint16_t *data_u16;
> int ret = 0;
>
> mutex_lock(&smu->metrics_lock);
> @@ -510,78 +537,100 @@ static int sienna_cichlid_get_smu_metrics_data(struct smu_context *smu,
>
> switch (member) {
> case METRICS_CURR_GFXCLK:
> - *value = metrics->CurrClock[PPCLK_GFXCLK];
> + GET_METRICS_MEMBER(CurrClock, &data_u32);
One problem with this style is the need to track the datatype of each
field. Why not use the old style?
metricsv1? metricsv1->field : metricsv2->field;
Thanks,
Lijo
> + *value = data_u32[PPCLK_GFXCLK];
> break;
> case METRICS_CURR_SOCCLK:
> - *value = metrics->CurrClock[PPCLK_SOCCLK];
> + GET_METRICS_MEMBER(CurrClock, &data_u32);
> + *value = data_u32[PPCLK_SOCCLK];
> break;
> case METRICS_CURR_UCLK:
> - *value = metrics->CurrClock[PPCLK_UCLK];
> + GET_METRICS_MEMBER(CurrClock, &data_u32);
> + *value = data_u32[PPCLK_UCLK];
> break;
> case METRICS_CURR_VCLK:
> - *value = metrics->CurrClock[PPCLK_VCLK_0];
> + GET_METRICS_MEMBER(CurrClock, &data_u32);
> + *value = data_u32[PPCLK_VCLK_0];
> break;
> case METRICS_CURR_VCLK1:
> - *value = metrics->CurrClock[PPCLK_VCLK_1];
> + GET_METRICS_MEMBER(CurrClock, &data_u32);
> + *value = data_u32[PPCLK_VCLK_1];
> break;
> case METRICS_CURR_DCLK:
> - *value = metrics->CurrClock[PPCLK_DCLK_0];
> + GET_METRICS_MEMBER(CurrClock, &data_u32);
> + *value = data_u32[PPCLK_DCLK_0];
> break;
> case METRICS_CURR_DCLK1:
> - *value = metrics->CurrClock[PPCLK_DCLK_1];
> + GET_METRICS_MEMBER(CurrClock, &data_u32);
> + *value = data_u32[PPCLK_DCLK_1];
> break;
> case METRICS_CURR_DCEFCLK:
> - *value = metrics->CurrClock[PPCLK_DCEFCLK];
> + GET_METRICS_MEMBER(CurrClock, &data_u32);
> + *value = data_u32[PPCLK_DCEFCLK];
> break;
> case METRICS_CURR_FCLK:
> - *value = metrics->CurrClock[PPCLK_FCLK];
> + GET_METRICS_MEMBER(CurrClock, &data_u32);
> + *value = data_u32[PPCLK_FCLK];
> break;
> case METRICS_AVERAGE_GFXCLK:
> - if (metrics->AverageGfxActivity <= SMU_11_0_7_GFX_BUSY_THRESHOLD)
> - *value = metrics->AverageGfxclkFrequencyPostDs;
> + GET_METRICS_MEMBER(AverageGfxActivity, &data_u16);
> + if (*data_u16 <= SMU_11_0_7_GFX_BUSY_THRESHOLD)
> + GET_METRICS_MEMBER(AverageGfxclkFrequencyPostDs, &data_u16);
> else
> - *value = metrics->AverageGfxclkFrequencyPreDs;
> + GET_METRICS_MEMBER(AverageGfxclkFrequencyPreDs, &data_u16);
> + *value = *data_u16;
> break;
> case METRICS_AVERAGE_FCLK:
> - *value = metrics->AverageFclkFrequencyPostDs;
> + GET_METRICS_MEMBER(AverageFclkFrequencyPostDs, &data_u16);
> + *value = *data_u16;
> break;
> case METRICS_AVERAGE_UCLK:
> - *value = metrics->AverageUclkFrequencyPostDs;
> + GET_METRICS_MEMBER(AverageUclkFrequencyPostDs, &data_u16);
> + *value = *data_u16;
> break;
> case METRICS_AVERAGE_GFXACTIVITY:
> - *value = metrics->AverageGfxActivity;
> + GET_METRICS_MEMBER(AverageGfxActivity, &data_u16);
> + *value = *data_u16;
> break;
> case METRICS_AVERAGE_MEMACTIVITY:
> - *value = metrics->AverageUclkActivity;
> + GET_METRICS_MEMBER(AverageUclkActivity, &data_u16);
> + *value = *data_u16;
> break;
> case METRICS_AVERAGE_SOCKETPOWER:
> - *value = metrics->AverageSocketPower << 8;
> + GET_METRICS_MEMBER(AverageSocketPower, &data_u16);
> + *value = *data_u16 << 8;
> break;
> case METRICS_TEMPERATURE_EDGE:
> - *value = metrics->TemperatureEdge *
> + GET_METRICS_MEMBER(TemperatureEdge, &data_u16);
> + *value = *data_u16 *
> SMU_TEMPERATURE_UNITS_PER_CENTIGRADES;
> break;
> case METRICS_TEMPERATURE_HOTSPOT:
> - *value = metrics->TemperatureHotspot *
> + GET_METRICS_MEMBER(TemperatureHotspot, &data_u16);
> + *value = *data_u16 *
> SMU_TEMPERATURE_UNITS_PER_CENTIGRADES;
> break;
> case METRICS_TEMPERATURE_MEM:
> - *value = metrics->TemperatureMem *
> + GET_METRICS_MEMBER(TemperatureMem, &data_u16);
> + *value = *data_u16 *
> SMU_TEMPERATURE_UNITS_PER_CENTIGRADES;
> break;
> case METRICS_TEMPERATURE_VRGFX:
> - *value = metrics->TemperatureVrGfx *
> + GET_METRICS_MEMBER(TemperatureVrGfx, &data_u16);
> + *value = *data_u16 *
> SMU_TEMPERATURE_UNITS_PER_CENTIGRADES;
> break;
> case METRICS_TEMPERATURE_VRSOC:
> - *value = metrics->TemperatureVrSoc *
> + GET_METRICS_MEMBER(TemperatureVrSoc, &data_u16);
> + *value = *data_u16 *
> SMU_TEMPERATURE_UNITS_PER_CENTIGRADES;
> break;
> case METRICS_THROTTLER_STATUS:
> - *value = metrics->ThrottlerStatus;
> + *value = sienna_cichlid_get_throttler_status_locked(smu);
> break;
> case METRICS_CURR_FANSPEED:
> - *value = metrics->CurrFanSpeed;
> + GET_METRICS_MEMBER(CurrFanSpeed, &data_u16);
> + *value = *data_u16;
> break;
> default:
> *value = UINT_MAX;
> @@ -3564,68 +3613,103 @@ static ssize_t sienna_cichlid_get_gpu_metrics(struct smu_context *smu,
> struct smu_table_context *smu_table = &smu->smu_table;
> struct gpu_metrics_v1_3 *gpu_metrics =
> (struct gpu_metrics_v1_3 *)smu_table->gpu_metrics_table;
> - SmuMetricsExternal_t metrics_external;
> - SmuMetrics_t *metrics =
> - &(metrics_external.SmuMetrics);
> - struct amdgpu_device *adev = smu->adev;
> - uint32_t smu_version;
> + uint32_t *data_u32;
> + uint16_t *data_u16;
> + uint8_t *data_u8;
> int ret = 0;
>
> - ret = smu_cmn_get_metrics_table(smu,
> - &metrics_external,
> - true);
> - if (ret)
> + mutex_lock(&smu->metrics_lock);
> +
> + ret = smu_cmn_get_metrics_table_locked(smu,
> + NULL,
> + true);
> + if (ret) {
> + mutex_unlock(&smu->metrics_lock);
> return ret;
> + }
>
> smu_cmn_init_soft_gpu_metrics(gpu_metrics, 1, 3);
>
> - gpu_metrics->temperature_edge = metrics->TemperatureEdge;
> - gpu_metrics->temperature_hotspot = metrics->TemperatureHotspot;
> - gpu_metrics->temperature_mem = metrics->TemperatureMem;
> - gpu_metrics->temperature_vrgfx = metrics->TemperatureVrGfx;
> - gpu_metrics->temperature_vrsoc = metrics->TemperatureVrSoc;
> - gpu_metrics->temperature_vrmem = metrics->TemperatureVrMem0;
> + GET_METRICS_MEMBER(TemperatureEdge, &data_u16);
> + gpu_metrics->temperature_edge = *data_u16;
> +
> + GET_METRICS_MEMBER(TemperatureHotspot, &data_u16);
> + gpu_metrics->temperature_hotspot = *data_u16;
> +
> + GET_METRICS_MEMBER(TemperatureMem, &data_u16);
> + gpu_metrics->temperature_mem = *data_u16;
> +
> + GET_METRICS_MEMBER(TemperatureVrGfx, &data_u16);
> + gpu_metrics->temperature_vrgfx = *data_u16;
> +
> + GET_METRICS_MEMBER(TemperatureVrSoc, &data_u16);
> + gpu_metrics->temperature_vrsoc = *data_u16;
> +
> + GET_METRICS_MEMBER(TemperatureVrMem0, &data_u16);
> + gpu_metrics->temperature_vrmem = *data_u16;
>
> - gpu_metrics->average_gfx_activity = metrics->AverageGfxActivity;
> - gpu_metrics->average_umc_activity = metrics->AverageUclkActivity;
> - gpu_metrics->average_mm_activity = metrics->VcnActivityPercentage;
> + GET_METRICS_MEMBER(AverageGfxActivity, &data_u16);
> + gpu_metrics->average_gfx_activity = *data_u16;
>
> - gpu_metrics->average_socket_power = metrics->AverageSocketPower;
> - gpu_metrics->energy_accumulator = metrics->EnergyAccumulator;
> + GET_METRICS_MEMBER(AverageUclkActivity, &data_u16);
> + gpu_metrics->average_umc_activity = *data_u16;
>
> - if (metrics->AverageGfxActivity <= SMU_11_0_7_GFX_BUSY_THRESHOLD)
> - gpu_metrics->average_gfxclk_frequency = metrics->AverageGfxclkFrequencyPostDs;
> + GET_METRICS_MEMBER(VcnActivityPercentage, &data_u16);
> + gpu_metrics->average_mm_activity = *data_u16;
> +
> + GET_METRICS_MEMBER(AverageSocketPower, &data_u16);
> + gpu_metrics->average_socket_power = *data_u16;
> +
> + GET_METRICS_MEMBER(EnergyAccumulator, &data_u32);
> + gpu_metrics->energy_accumulator = *data_u32;
> +
> + GET_METRICS_MEMBER(AverageGfxActivity, &data_u16);
> + if (*data_u16 <= SMU_11_0_7_GFX_BUSY_THRESHOLD)
> + GET_METRICS_MEMBER(AverageGfxclkFrequencyPostDs, &data_u16);
> else
> - gpu_metrics->average_gfxclk_frequency = metrics->AverageGfxclkFrequencyPreDs;
> - gpu_metrics->average_uclk_frequency = metrics->AverageUclkFrequencyPostDs;
> - gpu_metrics->average_vclk0_frequency = metrics->AverageVclk0Frequency;
> - gpu_metrics->average_dclk0_frequency = metrics->AverageDclk0Frequency;
> - gpu_metrics->average_vclk1_frequency = metrics->AverageVclk1Frequency;
> - gpu_metrics->average_dclk1_frequency = metrics->AverageDclk1Frequency;
> -
> - gpu_metrics->current_gfxclk = metrics->CurrClock[PPCLK_GFXCLK];
> - gpu_metrics->current_socclk = metrics->CurrClock[PPCLK_SOCCLK];
> - gpu_metrics->current_uclk = metrics->CurrClock[PPCLK_UCLK];
> - gpu_metrics->current_vclk0 = metrics->CurrClock[PPCLK_VCLK_0];
> - gpu_metrics->current_dclk0 = metrics->CurrClock[PPCLK_DCLK_0];
> - gpu_metrics->current_vclk1 = metrics->CurrClock[PPCLK_VCLK_1];
> - gpu_metrics->current_dclk1 = metrics->CurrClock[PPCLK_DCLK_1];
> -
> - gpu_metrics->throttle_status = metrics->ThrottlerStatus;
> + GET_METRICS_MEMBER(AverageGfxclkFrequencyPreDs, &data_u16);
> + gpu_metrics->average_gfxclk_frequency = *data_u16;
> +
> + GET_METRICS_MEMBER(AverageUclkFrequencyPostDs, &data_u16);
> + gpu_metrics->average_uclk_frequency = *data_u16;
> +
> + GET_METRICS_MEMBER(AverageVclk0Frequency, &data_u16);
> + gpu_metrics->average_vclk0_frequency = *data_u16;
> +
> + GET_METRICS_MEMBER(AverageDclk0Frequency, &data_u16);
> + gpu_metrics->average_dclk0_frequency = *data_u16;
> +
> + GET_METRICS_MEMBER(AverageVclk1Frequency, &data_u16);
> + gpu_metrics->average_vclk1_frequency = *data_u16;
> +
> + GET_METRICS_MEMBER(AverageDclk1Frequency, &data_u16);
> + gpu_metrics->average_dclk1_frequency = *data_u16;
> +
> + GET_METRICS_MEMBER(CurrClock, &data_u32);
> + gpu_metrics->current_gfxclk = data_u32[PPCLK_GFXCLK];
> + gpu_metrics->current_socclk = data_u32[PPCLK_SOCCLK];
> + gpu_metrics->current_uclk = data_u32[PPCLK_UCLK];
> + gpu_metrics->current_vclk0 = data_u32[PPCLK_VCLK_0];
> + gpu_metrics->current_dclk0 = data_u32[PPCLK_DCLK_0];
> + gpu_metrics->current_vclk1 = data_u32[PPCLK_VCLK_1];
> + gpu_metrics->current_dclk1 = data_u32[PPCLK_DCLK_1];
> +
> + gpu_metrics->throttle_status =
> + sienna_cichlid_get_throttler_status_locked(smu);
> gpu_metrics->indep_throttle_status =
> - smu_cmn_get_indep_throttler_status(metrics->ThrottlerStatus,
> + smu_cmn_get_indep_throttler_status(gpu_metrics->throttle_status,
> sienna_cichlid_throttler_map);
>
> - gpu_metrics->current_fan_speed = metrics->CurrFanSpeed;
> + GET_METRICS_MEMBER(CurrFanSpeed, &data_u16);
> + gpu_metrics->current_fan_speed = *data_u16;
>
> - ret = smu_cmn_get_smc_version(smu, NULL, &smu_version);
> - if (ret)
> - return ret;
> + if (((smu->adev->asic_type == CHIP_SIENNA_CICHLID) && smu->smc_fw_version > 0x003A1E00) ||
> + ((smu->adev->asic_type == CHIP_NAVY_FLOUNDER) && smu->smc_fw_version > 0x00410400)) {
> + GET_METRICS_MEMBER(PcieWidth, &data_u8);
> + gpu_metrics->pcie_link_width = *data_u8;
>
> - if (((adev->asic_type == CHIP_SIENNA_CICHLID) && smu_version > 0x003A1E00) ||
> - ((adev->asic_type == CHIP_NAVY_FLOUNDER) && smu_version > 0x00410400)) {
> - gpu_metrics->pcie_link_width = metrics->PcieWidth;
> - gpu_metrics->pcie_link_speed = link_speed[metrics->PcieRate];
> + GET_METRICS_MEMBER(PcieRate, &data_u8);
> + gpu_metrics->pcie_link_speed = link_speed[*data_u8];
> } else {
> gpu_metrics->pcie_link_width =
> smu_v11_0_get_current_pcie_link_width(smu);
> @@ -3633,6 +3717,8 @@ static ssize_t sienna_cichlid_get_gpu_metrics(struct smu_context *smu,
> smu_v11_0_get_current_pcie_link_speed(smu);
> }
>
> + mutex_unlock(&smu->metrics_lock);
> +
> gpu_metrics->system_clock_counter = ktime_get_boottime_ns();
>
> *table = (void *)gpu_metrics;
>
More information about the amd-gfx
mailing list