提交 f1c37859 编写于 作者: E Evan Quan 提交者: Alex Deucher

drm/amd/powerplay: add Arcturus support for gpu metrics export

Add Arcturus gpu metrics export interface.
Signed-off-by: NEvan Quan <evan.quan@amd.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>
上级 e4c9200d
......@@ -79,6 +79,8 @@
/* possible frequency drift (1Mhz) */
#define EPSILON 1
#define smnPCIE_ESM_CTRL 0x111003D0
static const struct cmn2asic_msg_mapping arcturus_message_map[SMU_MSG_MAX_COUNT] = {
MSG_MAP(TestMessage, PPSMC_MSG_TestMessage, 0),
MSG_MAP(GetSmuVersion, PPSMC_MSG_GetSmuVersion, 1),
......@@ -234,6 +236,13 @@ static int arcturus_tables_init(struct smu_context *smu)
return -ENOMEM;
smu_table->metrics_time = 0;
smu_table->gpu_metrics_table_size = sizeof(struct gpu_metrics_v1_0);
smu_table->gpu_metrics_table = kzalloc(smu_table->gpu_metrics_table_size, GFP_KERNEL);
if (!smu_table->gpu_metrics_table) {
kfree(smu_table->metrics_table);
return -ENOMEM;
}
return 0;
}
......@@ -2242,6 +2251,88 @@ static void arcturus_log_thermal_throttling_event(struct smu_context *smu)
kgd2kfd_smi_event_throttle(smu->adev->kfd.dev, throttler_status);
}
static int arcturus_get_current_pcie_link_speed(struct smu_context *smu)
{
struct amdgpu_device *adev = smu->adev;
uint32_t esm_ctrl;
/* TODO: confirm this on real target */
esm_ctrl = RREG32_PCIE(smnPCIE_ESM_CTRL);
if ((esm_ctrl >> 15) & 0x1FFFF)
return (((esm_ctrl >> 8) & 0x3F) + 128);
return smu_v11_0_get_current_pcie_link_speed(smu);
}
static ssize_t arcturus_get_gpu_metrics(struct smu_context *smu,
void **table)
{
struct smu_table_context *smu_table = &smu->smu_table;
struct gpu_metrics_v1_0 *gpu_metrics =
(struct gpu_metrics_v1_0 *)smu_table->gpu_metrics_table;
SmuMetrics_t metrics;
int ret = 0;
mutex_lock(&smu->metrics_lock);
ret = smu_cmn_update_table(smu,
SMU_TABLE_SMU_METRICS,
0,
smu_table->metrics_table,
false);
if (ret) {
dev_info(smu->adev->dev, "Failed to export SMU metrics table!\n");
mutex_unlock(&smu->metrics_lock);
return ret;
}
smu_table->metrics_time = jiffies;
memcpy(&metrics, smu_table->metrics_table, sizeof(SmuMetrics_t));
mutex_unlock(&smu->metrics_lock);
smu_v11_0_init_gpu_metrics_v1_0(gpu_metrics);
gpu_metrics->temperature_edge = metrics.TemperatureEdge;
gpu_metrics->temperature_hotspot = metrics.TemperatureHotspot;
gpu_metrics->temperature_mem = metrics.TemperatureHBM;
gpu_metrics->temperature_vrgfx = metrics.TemperatureVrGfx;
gpu_metrics->temperature_vrsoc = metrics.TemperatureVrSoc;
gpu_metrics->temperature_vrmem = metrics.TemperatureVrMem;
gpu_metrics->average_gfx_activity = metrics.AverageGfxActivity;
gpu_metrics->average_umc_activity = metrics.AverageUclkActivity;
gpu_metrics->average_mm_activity = metrics.VcnActivityPercentage;
gpu_metrics->average_socket_power = metrics.AverageSocketPower;
gpu_metrics->energy_accumulator = metrics.EnergyAccumulator;
gpu_metrics->average_gfxclk_frequency = metrics.AverageGfxclkFrequency;
gpu_metrics->average_socclk_frequency = metrics.AverageSocclkFrequency;
gpu_metrics->average_uclk_frequency = metrics.AverageUclkFrequency;
gpu_metrics->average_vclk0_frequency = metrics.AverageVclkFrequency;
gpu_metrics->average_dclk0_frequency = metrics.AverageDclkFrequency;
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];
gpu_metrics->current_dclk0 = metrics.CurrClock[PPCLK_DCLK];
gpu_metrics->throttle_status = metrics.ThrottlerStatus;
gpu_metrics->current_fan_speed = metrics.CurrFanSpeed;
gpu_metrics->pcie_link_width =
smu_v11_0_get_current_pcie_link_width(smu);
gpu_metrics->pcie_link_speed =
arcturus_get_current_pcie_link_speed(smu);
*table = (void *)gpu_metrics;
return sizeof(struct gpu_metrics_v1_0);
}
static const struct pptable_funcs arcturus_ppt_funcs = {
/* init dpm */
.get_allowed_feature_mask = arcturus_get_allowed_feature_mask,
......@@ -2319,6 +2410,7 @@ static const struct pptable_funcs arcturus_ppt_funcs = {
.log_thermal_throttling_event = arcturus_log_thermal_throttling_event,
.get_pp_feature_mask = smu_cmn_get_pp_feature_mask,
.set_pp_feature_mask = smu_cmn_set_pp_feature_mask,
.get_gpu_metrics = arcturus_get_gpu_metrics,
};
void arcturus_set_ppt_funcs(struct smu_context *smu)
......
......@@ -274,6 +274,9 @@ struct smu_table_context
void *overdrive_table;
void *boot_overdrive_table;
uint32_t gpu_metrics_table_size;
void *gpu_metrics_table;
};
struct smu_dpm_context {
......
......@@ -272,5 +272,7 @@ int smu_v11_0_get_current_pcie_link_speed_level(struct smu_context *smu);
int smu_v11_0_get_current_pcie_link_speed(struct smu_context *smu);
void smu_v11_0_init_gpu_metrics_v1_0(struct gpu_metrics_v1_0 *gpu_metrics);
#endif
#endif
......@@ -417,10 +417,12 @@ int smu_v11_0_fini_smc_tables(struct smu_context *smu)
struct smu_table_context *smu_table = &smu->smu_table;
struct smu_dpm_context *smu_dpm = &smu->smu_dpm;
kfree(smu_table->gpu_metrics_table);
kfree(smu_table->boot_overdrive_table);
kfree(smu_table->overdrive_table);
kfree(smu_table->max_sustainable_clocks);
kfree(smu_table->driver_pptable);
smu_table->gpu_metrics_table = NULL;
smu_table->boot_overdrive_table = NULL;
smu_table->overdrive_table = NULL;
smu_table->max_sustainable_clocks = NULL;
......@@ -1971,3 +1973,15 @@ int smu_v11_0_get_current_pcie_link_speed(struct smu_context *smu)
return link_speed[speed_level];
}
void smu_v11_0_init_gpu_metrics_v1_0(struct gpu_metrics_v1_0 *gpu_metrics)
{
memset(gpu_metrics, 0xFF, sizeof(struct gpu_metrics_v1_0));
gpu_metrics->common_header.structure_size =
sizeof(struct gpu_metrics_v1_0);
gpu_metrics->common_header.format_revision = 1;
gpu_metrics->common_header.content_revision = 0;
gpu_metrics->system_clock_counter = ktime_get_boottime_ns();
}
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册