diff options
Diffstat (limited to 'drivers/gpu/drm/amd/pm')
| -rw-r--r-- | drivers/gpu/drm/amd/pm/amdgpu_pm.c | 163 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/pm/inc/amdgpu_smu.h | 535 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/pm/inc/smu11_driver_if_vangogh.h | 1 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/pm/inc/smu_types.h | 2 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu10_hwmgr.c | 166 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu10_hwmgr.h | 3 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c | 10 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/pm/swsmu/smu11/navi10_ppt.c | 2 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/pm/swsmu/smu11/sienna_cichlid_ppt.c | 332 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/pm/swsmu/smu11/vangogh_ppt.c | 881 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/pm/swsmu/smu11/vangogh_ppt.h | 26 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/pm/swsmu/smu12/renoir_ppt.c | 224 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/pm/swsmu/smu12/smu_v12_0.c | 1 |
13 files changed, 2298 insertions, 48 deletions
diff --git a/drivers/gpu/drm/amd/pm/amdgpu_pm.c b/drivers/gpu/drm/amd/pm/amdgpu_pm.c index 7b6ef05a1d35..f5d97b97353a 100644 --- a/drivers/gpu/drm/amd/pm/amdgpu_pm.c +++ b/drivers/gpu/drm/amd/pm/amdgpu_pm.c @@ -730,11 +730,18 @@ static ssize_t amdgpu_set_pp_table(struct device *dev, * * - minimum and maximum engine clock labeled OD_SCLK * - * - maximum memory clock labeled OD_MCLK + * - minimum(not available for Vega20 and Navi1x) and maximum memory + * clock labeled OD_MCLK * * - three <frequency, voltage> points labeled OD_VDDC_CURVE. * They can be used to calibrate the sclk voltage curve. * + * - voltage offset(in mV) applied on target voltage calculation. + * This is available for Sienna Cichlid, Navy Flounder and Dimgrey + * Cavefish. For these ASICs, the target voltage calculation can be + * illustrated by "voltage = voltage calculated from v/f curve + + * overdrive vddgfx offset" + * * - a list of valid ranges for sclk, mclk, and voltage curve points * labeled OD_RANGE * @@ -755,6 +762,11 @@ static ssize_t amdgpu_set_pp_table(struct device *dev, * 600mV. "vc 2 1000 1000" will update point3 with clock set * as 1000Mhz and voltage 1000mV. * + * To update the voltage offset applied for gfxclk/voltage calculation, + * enter the new value by writing a string that contains "vo offset". + * This is supported by Sienna Cichlid, Navy Flounder and Dimgrey Cavefish. + * And the offset can be a positive or negative value. + * * - When you have edited all of the states as needed, write "c" (commit) * to the file to commit your changes * @@ -795,6 +807,8 @@ static ssize_t amdgpu_set_pp_od_clk_voltage(struct device *dev, type = PP_OD_COMMIT_DPM_TABLE; else if (!strncmp(buf, "vc", 2)) type = PP_OD_EDIT_VDDC_CURVE; + else if (!strncmp(buf, "vo", 2)) + type = PP_OD_EDIT_VDDGFX_OFFSET; else return -EINVAL; @@ -802,7 +816,8 @@ static ssize_t amdgpu_set_pp_od_clk_voltage(struct device *dev, tmp_str = buf_cpy; - if (type == PP_OD_EDIT_VDDC_CURVE) + if ((type == PP_OD_EDIT_VDDC_CURVE) || + (type == PP_OD_EDIT_VDDGFX_OFFSET)) tmp_str++; while (isspace(*++tmp_str)); @@ -898,6 +913,7 @@ static ssize_t amdgpu_get_pp_od_clk_voltage(struct device *dev, size = smu_print_clk_levels(&adev->smu, SMU_OD_SCLK, buf); size += smu_print_clk_levels(&adev->smu, SMU_OD_MCLK, buf+size); size += smu_print_clk_levels(&adev->smu, SMU_OD_VDDC_CURVE, buf+size); + size += smu_print_clk_levels(&adev->smu, SMU_OD_VDDGFX_OFFSET, buf+size); size += smu_print_clk_levels(&adev->smu, SMU_OD_RANGE, buf+size); } else if (adev->powerplay.pp_funcs->print_clock_levels) { size = amdgpu_dpm_print_clock_levels(adev, OD_SCLK, buf); @@ -1346,6 +1362,138 @@ static ssize_t amdgpu_set_pp_dpm_fclk(struct device *dev, return count; } +static ssize_t amdgpu_get_pp_dpm_vclk(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct drm_device *ddev = dev_get_drvdata(dev); + struct amdgpu_device *adev = drm_to_adev(ddev); + ssize_t size; + int ret; + + if (amdgpu_in_reset(adev)) + return -EPERM; + + ret = pm_runtime_get_sync(ddev->dev); + if (ret < 0) { + pm_runtime_put_autosuspend(ddev->dev); + return ret; + } + + if (is_support_sw_smu(adev)) + size = smu_print_clk_levels(&adev->smu, SMU_VCLK, buf); + else + size = snprintf(buf, PAGE_SIZE, "\n"); + + pm_runtime_mark_last_busy(ddev->dev); + pm_runtime_put_autosuspend(ddev->dev); + + return size; +} + +static ssize_t amdgpu_set_pp_dpm_vclk(struct device *dev, + struct device_attribute *attr, + const char *buf, + size_t count) +{ + struct drm_device *ddev = dev_get_drvdata(dev); + struct amdgpu_device *adev = drm_to_adev(ddev); + int ret; + uint32_t mask = 0; + + if (amdgpu_in_reset(adev)) + return -EPERM; + + ret = amdgpu_read_mask(buf, count, &mask); + if (ret) + return ret; + + ret = pm_runtime_get_sync(ddev->dev); + if (ret < 0) { + pm_runtime_put_autosuspend(ddev->dev); + return ret; + } + + if (is_support_sw_smu(adev)) + ret = smu_force_clk_levels(&adev->smu, SMU_VCLK, mask); + else + ret = 0; + + pm_runtime_mark_last_busy(ddev->dev); + pm_runtime_put_autosuspend(ddev->dev); + + if (ret) + return -EINVAL; + + return count; +} + +static ssize_t amdgpu_get_pp_dpm_dclk(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct drm_device *ddev = dev_get_drvdata(dev); + struct amdgpu_device *adev = drm_to_adev(ddev); + ssize_t size; + int ret; + + if (amdgpu_in_reset(adev)) + return -EPERM; + + ret = pm_runtime_get_sync(ddev->dev); + if (ret < 0) { + pm_runtime_put_autosuspend(ddev->dev); + return ret; + } + + if (is_support_sw_smu(adev)) + size = smu_print_clk_levels(&adev->smu, SMU_DCLK, buf); + else + size = snprintf(buf, PAGE_SIZE, "\n"); + + pm_runtime_mark_last_busy(ddev->dev); + pm_runtime_put_autosuspend(ddev->dev); + + return size; +} + +static ssize_t amdgpu_set_pp_dpm_dclk(struct device *dev, + struct device_attribute *attr, + const char *buf, + size_t count) +{ + struct drm_device *ddev = dev_get_drvdata(dev); + struct amdgpu_device *adev = drm_to_adev(ddev); + int ret; + uint32_t mask = 0; + + if (amdgpu_in_reset(adev)) + return -EPERM; + + ret = amdgpu_read_mask(buf, count, &mask); + if (ret) + return ret; + + ret = pm_runtime_get_sync(ddev->dev); + if (ret < 0) { + pm_runtime_put_autosuspend(ddev->dev); + return ret; + } + + if (is_support_sw_smu(adev)) + ret = smu_force_clk_levels(&adev->smu, SMU_DCLK, mask); + else + ret = 0; + + pm_runtime_mark_last_busy(ddev->dev); + pm_runtime_put_autosuspend(ddev->dev); + + if (ret) + return -EINVAL; + + return count; +} + static ssize_t amdgpu_get_pp_dpm_dcefclk(struct device *dev, struct device_attribute *attr, char *buf) @@ -2025,6 +2173,8 @@ static struct amdgpu_device_attr amdgpu_device_attrs[] = { AMDGPU_DEVICE_ATTR_RW(pp_dpm_mclk, ATTR_FLAG_BASIC|ATTR_FLAG_ONEVF), AMDGPU_DEVICE_ATTR_RW(pp_dpm_socclk, ATTR_FLAG_BASIC|ATTR_FLAG_ONEVF), AMDGPU_DEVICE_ATTR_RW(pp_dpm_fclk, ATTR_FLAG_BASIC|ATTR_FLAG_ONEVF), + AMDGPU_DEVICE_ATTR_RW(pp_dpm_vclk, ATTR_FLAG_BASIC|ATTR_FLAG_ONEVF), + AMDGPU_DEVICE_ATTR_RW(pp_dpm_dclk, ATTR_FLAG_BASIC|ATTR_FLAG_ONEVF), AMDGPU_DEVICE_ATTR_RW(pp_dpm_dcefclk, ATTR_FLAG_BASIC), AMDGPU_DEVICE_ATTR_RW(pp_dpm_pcie, ATTR_FLAG_BASIC), AMDGPU_DEVICE_ATTR_RW(pp_sclk_od, ATTR_FLAG_BASIC), @@ -2067,7 +2217,8 @@ static int default_attr_update(struct amdgpu_device *adev, struct amdgpu_device_ } else if (DEVICE_ATTR_IS(pp_od_clk_voltage)) { *states = ATTR_STATE_UNSUPPORTED; if ((is_support_sw_smu(adev) && adev->smu.od_enabled) || - (!is_support_sw_smu(adev) && hwmgr->od_enabled)) + (is_support_sw_smu(adev) && adev->smu.fine_grain_enabled) || + (!is_support_sw_smu(adev) && hwmgr->od_enabled)) *states = ATTR_STATE_SUPPORTED; } else if (DEVICE_ATTR_IS(mem_busy_percent)) { if (adev->flags & AMD_IS_APU || asic_type == CHIP_VEGA10) @@ -2087,6 +2238,12 @@ static int default_attr_update(struct amdgpu_device *adev, struct amdgpu_device_ } else if (DEVICE_ATTR_IS(gpu_metrics)) { if (asic_type < CHIP_VEGA12) *states = ATTR_STATE_UNSUPPORTED; + } else if (DEVICE_ATTR_IS(pp_dpm_vclk)) { + if (!(asic_type == CHIP_VANGOGH)) + *states = ATTR_STATE_UNSUPPORTED; + } else if (DEVICE_ATTR_IS(pp_dpm_dclk)) { + if (!(asic_type == CHIP_VANGOGH)) + *states = ATTR_STATE_UNSUPPORTED; } if (asic_type == CHIP_ARCTURUS) { diff --git a/drivers/gpu/drm/amd/pm/inc/amdgpu_smu.h b/drivers/gpu/drm/amd/pm/inc/amdgpu_smu.h index 4bdbcce7092d..e2e59fb0f754 100644 --- a/drivers/gpu/drm/amd/pm/inc/amdgpu_smu.h +++ b/drivers/gpu/drm/amd/pm/inc/amdgpu_smu.h @@ -465,124 +465,653 @@ struct smu_context uint32_t gfx_default_soft_max_freq; uint32_t gfx_actual_hard_min_freq; uint32_t gfx_actual_soft_max_freq; + + bool fine_grain_enabled; + bool fine_grain_started; }; struct i2c_adapter; +/** + * struct pptable_funcs - Callbacks used to interact with the SMU. + */ struct pptable_funcs { + /** + * @run_btc: Calibrate voltage/frequency curve to fit the system's + * power delivery and voltage margins. Required for adaptive + * voltage frequency scaling (AVFS). + */ int (*run_btc)(struct smu_context *smu); + + /** + * @get_allowed_feature_mask: Get allowed feature mask. + * &feature_mask: Array to store feature mask. + * &num: Elements in &feature_mask. + */ int (*get_allowed_feature_mask)(struct smu_context *smu, uint32_t *feature_mask, uint32_t num); + + /** + * @get_current_power_state: Get the current power state. + * + * Return: Current power state on success, negative errno on failure. + */ enum amd_pm_state_type (*get_current_power_state)(struct smu_context *smu); + + /** + * @set_default_dpm_table: Retrieve the default overdrive settings from + * the SMU. + */ int (*set_default_dpm_table)(struct smu_context *smu); + int (*set_power_state)(struct smu_context *smu); + + /** + * @populate_umd_state_clk: Populate the UMD power state table with + * defaults. + */ int (*populate_umd_state_clk)(struct smu_context *smu); + + /** + * @print_clk_levels: Print DPM clock levels for a clock domain + * to buffer. Star current level. + * + * Used for sysfs interfaces. + */ int (*print_clk_levels)(struct smu_context *smu, enum smu_clk_type clk_type, char *buf); + + /** + * @force_clk_levels: Set a range of allowed DPM levels for a clock + * domain. + * &clk_type: Clock domain. + * &mask: Range of allowed DPM levels. + */ int (*force_clk_levels)(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t mask); + + /** + * @od_edit_dpm_table: Edit the custom overdrive DPM table. + * &type: Type of edit. + * &input: Edit parameters. + * &size: Size of &input. + */ int (*od_edit_dpm_table)(struct smu_context *smu, enum PP_OD_DPM_TABLE_COMMAND type, long *input, uint32_t size); + + /** + * @get_clock_by_type_with_latency: Get the speed and latency of a clock + * domain. + */ int (*get_clock_by_type_with_latency)(struct smu_context *smu, enum smu_clk_type clk_type, struct pp_clock_levels_with_latency *clocks); + /** + * @get_clock_by_type_with_voltage: Get the speed and voltage of a clock + * domain. + */ + int (*get_clock_by_type_with_voltage)(struct smu_context *smu, + enum amd_pp_clock_type type, + struct + pp_clock_levels_with_voltage + *clocks); + + /** + * @get_power_profile_mode: Print all power profile modes to + * buffer. Star current mode. + */ int (*get_power_profile_mode)(struct smu_context *smu, char *buf); + + /** + * @set_power_profile_mode: Set a power profile mode. Also used to + * create/set custom power profile modes. + * &input: Power profile mode parameters. + * &size: Size of &input. + */ int (*set_power_profile_mode)(struct smu_context *smu, long *input, uint32_t size); + + /** + * @dpm_set_vcn_enable: Enable/disable VCN engine dynamic power + * management. + */ int (*dpm_set_vcn_enable)(struct smu_context *smu, bool enable); + + /** + * @dpm_set_jpeg_enable: Enable/disable JPEG engine dynamic power + * management. + */ int (*dpm_set_jpeg_enable)(struct smu_context *smu, bool enable); + + /** + * @read_sensor: Read data from a sensor. + * &sensor: Sensor to read data from. + * &data: Sensor reading. + * &size: Size of &data. + */ int (*read_sensor)(struct smu_context *smu, enum amd_pp_sensors sensor, void *data, uint32_t *size); + + /** + * @pre_display_config_changed: Prepare GPU for a display configuration + * change. + * + * Disable display tracking and pin memory clock speed to maximum. Used + * in display component synchronization. + */ int (*pre_display_config_changed)(struct smu_context *smu); + + /** + * @display_config_changed: Notify the SMU of the current display + * configuration. + * + * Allows SMU to properly track blanking periods for memory clock + * adjustment. Used in display component synchronization. + */ int (*display_config_changed)(struct smu_context *smu); + int (*apply_clocks_adjust_rules)(struct smu_context *smu); + + /** + * @notify_smc_display_config: Applies display requirements to the + * current power state. + * + * Optimize deep sleep DCEFclk and mclk for the current display + * configuration. Used in display component synchronization. + */ int (*notify_smc_display_config)(struct smu_context *smu); + + /** + * @is_dpm_running: Check if DPM is running. + * + * Return: True if DPM is running, false otherwise. + */ bool (*is_dpm_running)(struct smu_context *smu); + + /** + * @get_fan_speed_rpm: Get the current fan speed in RPM. + */ int (*get_fan_speed_rpm)(struct smu_context *smu, uint32_t *speed); + + /** + * @set_watermarks_table: Configure and upload the watermarks tables to + * the SMU. + */ int (*set_watermarks_table)(struct smu_context *smu, struct pp_smu_wm_range_sets *clock_ranges); + + /** + * @get_thermal_temperature_range: Get safe thermal limits in Celcius. + */ int (*get_thermal_temperature_range)(struct smu_context *smu, struct smu_temperature_range *range); + + /** + * @get_uclk_dpm_states: Get memory clock DPM levels in kHz. + * &clocks_in_khz: Array of DPM levels. + * &num_states: Elements in &clocks_in_khz. + */ int (*get_uclk_dpm_states)(struct smu_context *smu, uint32_t *clocks_in_khz, uint32_t *num_states); + + /** + * @set_default_od_settings: Set the overdrive tables to defaults. + */ int (*set_default_od_settings)(struct smu_context *smu); + + /** + * @set_performance_level: Set a performance level. + */ int (*set_performance_level)(struct smu_context *smu, enum amd_dpm_forced_level level); + + /** + * @display_disable_memory_clock_switch: Enable/disable dynamic memory + * clock switching. + * + * Disabling this feature forces memory clock speed to maximum. + * Enabling sets the minimum memory clock capable of driving the + * current display configuration. + */ int (*display_disable_memory_clock_switch)(struct smu_context *smu, bool disable_memory_clock_switch); + + /** + * @dump_pptable: Print the power play table to the system log. + */ void (*dump_pptable)(struct smu_context *smu); + + /** + * @get_power_limit: Get the device's power limits. + */ int (*get_power_limit)(struct smu_context *smu); + + /** + * @set_df_cstate: Set data fabric cstate. + */ int (*set_df_cstate)(struct smu_context *smu, enum pp_df_cstate state); + + /** + * @allow_xgmi_power_down: Enable/disable external global memory + * interconnect power down. + */ int (*allow_xgmi_power_down)(struct smu_context *smu, bool en); + + /** + * @update_pcie_parameters: Update and upload the system's PCIe + * capabilites to the SMU. + * &pcie_gen_cap: Maximum allowed PCIe generation. + * &pcie_width_cap: Maximum allowed PCIe width. + */ int (*update_pcie_parameters)(struct smu_context *smu, uint32_t pcie_gen_cap, uint32_t pcie_width_cap); + + /** + * @i2c_init: Initialize i2c. + * + * The i2c bus is used internally by the SMU voltage regulators and + * other devices. The i2c's EEPROM also stores bad page tables on boards + * with ECC. + */ int (*i2c_init)(struct smu_context *smu, struct i2c_adapter *control); + + /** + * @i2c_fini: Tear down i2c. + */ void (*i2c_fini)(struct smu_context *smu, struct i2c_adapter *control); + + /** + * @get_unique_id: Get the GPU's unique id. Used for asset tracking. + */ void (*get_unique_id)(struct smu_context *smu); + + /** + * @get_dpm_clock_table: Get a copy of the DPM clock table. + * + * Used by display component in bandwidth and watermark calculations. + */ int (*get_dpm_clock_table)(struct smu_context *smu, struct dpm_clocks *clock_table); + + /** + * @init_microcode: Request the SMU's firmware from the kernel. + */ int (*init_microcode)(struct smu_context *smu); + + /** + * @load_microcode: Load firmware onto the SMU. + */ int (*load_microcode)(struct smu_context *smu); + + /** + * @fini_microcode: Release the SMU's firmware. + */ void (*fini_microcode)(struct smu_context *smu); + + /** + * @init_smc_tables: Initialize the SMU tables. + */ int (*init_smc_tables)(struct smu_context *smu); + + /** + * @fini_smc_tables: Release the SMU tables. + */ int (*fini_smc_tables)(struct smu_context *smu); + + /** + * @init_power: Initialize the power gate table context. + */ int (*init_power)(struct smu_context *smu); + + /** + * @fini_power: Release the power gate table context. + */ int (*fini_power)(struct smu_context *smu); + + /** + * @check_fw_status: Check the SMU's firmware status. + * + * Return: Zero if check passes, negative errno on failure. + */ int (*check_fw_status)(struct smu_context *smu); + + /** + * @setup_pptable: Initialize the power play table and populate it with + * default values. + */ int (*setup_pptable)(struct smu_context *smu); + + /** + * @get_vbios_bootup_values: Get default boot values from the VBIOS. + */ int (*get_vbios_bootup_values)(struct smu_context *smu); + + /** + * @check_fw_version: Print driver and SMU interface versions to the + * system log. + * + * Interface mismatch is not a critical failure. + */ int (*check_fw_version)(struct smu_context *smu); + + /** + * @powergate_sdma: Power up/down system direct memory access. + */ int (*powergate_sdma)(struct smu_context *smu, bool gate); + + /** + * @set_gfx_cgpg: Enable/disable graphics engine course grain power + * gating. + */ int (*set_gfx_cgpg)(struct smu_context *smu, bool enable); + + /** + * @write_pptable: Write the power play table to the SMU. + */ int (*write_pptable)(struct smu_context *smu); + + /** + * @set_driver_table_location: Send the location of the driver table to + * the SMU. + */ int (*set_driver_table_location)(struct smu_context *smu); + + /** + * @set_tool_table_location: Send the location of the tool table to the + * SMU. + */ int (*set_tool_table_location)(struct smu_context *smu); + + /** + * @notify_memory_pool_location: Send the location of the memory pool to + * the SMU. + */ int (*notify_memory_pool_location)(struct smu_context *smu); + + /** + * @system_features_control: Enable/disable all SMU features. + */ int (*system_features_control)(struct smu_context *smu, bool en); + + /** + * @send_smc_msg_with_param: Send a message with a parameter to the SMU. + * &msg: Type of message. + * ¶m: Message parameter. + * &read_arg: SMU response (optional). + */ int (*send_smc_msg_with_param)(struct smu_context *smu, enum smu_message_type msg, uint32_t param, uint32_t *read_arg); + + /** + * @send_smc_msg: Send a message to the SMU. + * &msg: Type of message. + * &read_arg: SMU response (optional). + */ int (*send_smc_msg)(struct smu_context *smu, enum smu_message_type msg, uint32_t *read_arg); + + /** + * @init_display_count: Notify the SMU of the number of display + * components in current display configuration. + */ int (*init_display_count)(struct smu_context *smu, uint32_t count); + + /** + * @set_allowed_mask: Notify the SMU of the features currently allowed + * by the driver. + */ int (*set_allowed_mask)(struct smu_context *smu); + + /** + * @get_enabled_mask: Get a mask of features that are currently enabled + * on the SMU. + * &feature_mask: Array representing enabled feature mask. + * &num: Elements in &feature_mask. + */ int (*get_enabled_mask)(struct smu_context *smu, uint32_t *feature_mask, uint32_t num); + + /** + * @feature_is_enabled: Test if a feature is enabled. + * + * Return: One if enabled, zero if disabled. + */ int (*feature_is_enabled)(struct smu_context *smu, enum smu_feature_mask mask); + + /** + * @disable_all_features_with_exception: Disable all features with + * exception to those in &mask. + */ int (*disable_all_features_with_exception)(struct smu_context *smu, enum smu_feature_mask mask); + + /** + * @notify_display_change: Enable fast memory clock switching. + * + * Allows for fine grained memory clock switching but has more stringent + * timing requirements. + */ int (*notify_display_change)(struct smu_context *smu); + + /** + * @set_power_limit: Set power limit in watts. + */ int (*set_power_limit)(struct smu_context *smu, uint32_t n); + + /** + * @init_max_sustainable_clocks: Populate max sustainable clock speed + * table with values from the SMU. + */ int (*init_max_sustainable_clocks)(struct smu_context *smu); + + /** + * @enable_thermal_alert: Enable thermal alert interrupts. + */ int (*enable_thermal_alert)(struct smu_context *smu); + + /** + * @disable_thermal_alert: Disable thermal alert interrupts. + */ int (*disable_thermal_alert)(struct smu_context *smu); + + /** + * @set_min_dcef_deep_sleep: Set a minimum display fabric deep sleep + * clock speed in MHz. + */ int (*set_min_dcef_deep_sleep)(struct smu_context *smu, uint32_t clk); + + /** + * @display_clock_voltage_request: Set a hard minimum frequency + * for a clock domain. + */ int (*display_clock_voltage_request)(struct smu_context *smu, struct pp_display_clock_request *clock_req); + + /** + * @get_fan_control_mode: Get the current fan control mode. + */ uint32_t (*get_fan_control_mode)(struct smu_context *smu); + + /** + * @set_fan_control_mode: Set the fan control mode. + */ int (*set_fan_control_mode)(struct smu_context *smu, uint32_t mode); + + /** + * @set_fan_speed_rpm: Set a static fan speed in RPM. + */ int (*set_fan_speed_rpm)(struct smu_context *smu, uint32_t speed); + + /** + * @set_xgmi_pstate: Set inter-chip global memory interconnect pstate. + * &pstate: Pstate to set. D0 if Nonzero, D3 otherwise. + */ int (*set_xgmi_pstate)(struct smu_context *smu, uint32_t pstate); + + /** + * @gfx_off_control: Enable/disable graphics engine poweroff. + */ int (*gfx_off_control)(struct smu_context *smu, bool enable); + + + /** + * @get_gfx_off_status: Get graphics engine poweroff status. + * + * Return: + * 0 - GFXOFF(default). + * 1 - Transition out of GFX State. + * 2 - Not in GFXOFF. + * 3 - Transition into GFXOFF. + */ uint32_t (*get_gfx_off_status)(struct smu_context *smu); + + /** + * @register_irq_handler: Register interupt request handlers. + */ int (*register_irq_handler)(struct smu_context *smu); + + /** + * @set_azalia_d3_pme: Wake the audio decode engine from d3 sleep. + */ int (*set_azalia_d3_pme)(struct smu_context *smu); + + /** + * @get_max_sustainable_clocks_by_dc: Get a copy of the max sustainable + * clock speeds table. + * + * Provides a way for the display component (DC) to get the max + * sustainable clocks from the SMU. + */ int (*get_max_sustainable_clocks_by_dc)(struct smu_context *smu, struct pp_smu_nv_clock_table *max_clocks); + + /** + * @baco_is_support: Check if GPU supports BACO (Bus Active, Chip Off). + */ bool (*baco_is_support)(struct smu_context *smu); + + /** + * @baco_get_state: Get the current BACO state. + * + * Return: Current BACO state. + */ enum smu_baco_state (*baco_get_state)(struct smu_context *smu); + + /** + * @baco_set_state: Enter/exit BACO. + */ int (*baco_set_state)(struct smu_context *smu, enum smu_baco_state state); + + /** + * @baco_enter: Enter BACO. + */ int (*baco_enter)(struct smu_context *smu); + + /** + * @baco_exit: Exit Baco. + */ int (*baco_exit)(struct smu_context *smu); + + /** + * @mode1_reset_is_support: Check if GPU supports mode1 reset. + */ bool (*mode1_reset_is_support)(struct smu_context *smu); + + /** + * @mode1_reset: Perform mode1 reset. + * + * Complete GPU reset. + */ int (*mode1_reset)(struct smu_context *smu); + + /** + * @mode2_reset: Perform mode2 reset. + * + * Mode2 reset generally does not reset as many IPs as mode1 reset. The + * IPs reset varies by asic. + */ int (*mode2_reset)(struct smu_context *smu); + + /** + * @get_dpm_ultimate_freq: Get the hard frequency range of a clock + * domain in MHz. + */ int (*get_dpm_ultimate_freq)(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t *min, uint32_t *max); + + /** + * @set_soft_freq_limited_range: Set the soft frequency range of a clock + * domain in MHz. + */ int (*set_soft_freq_limited_range)(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t min, uint32_t max); + + /** + * @set_power_source: Notify the SMU of the current power source. + */ int (*set_power_source)(struct smu_context *smu, enum smu_power_src_type power_src); + + /** + * @log_thermal_throttling_event: Print a thermal throttling warning to + * the system's log. + */ void (*log_thermal_throttling_event)(struct smu_context *smu); + + /** + * @get_pp_feature_mask: Print a human readable table of enabled + * features to buffer. + */ size_t (*get_pp_feature_mask)(struct smu_context *smu, char *buf); + + /** + * @set_pp_feature_mask: Request the SMU enable/disable features to + * match those enabled in &new_mask. + */ int (*set_pp_feature_mask)(struct smu_context *smu, uint64_t new_mask); + + /** + * @get_gpu_metrics: Get a copy of the GPU metrics table from the SMU. + * + * Return: Size of &table + */ ssize_t (*get_gpu_metrics)(struct smu_context *smu, void **table); + + /** + * @enable_mgpu_fan_boost: Enable multi-GPU fan boost. + */ int (*enable_mgpu_fan_boost)(struct smu_context *smu); + + /** + * @gfx_ulv_control: Enable/disable ultra low voltage. + */ int (*gfx_ulv_control)(struct smu_context *smu, bool enablement); + + /** + * @deep_sleep_control: Enable/disable deep sleep. + */ int (*deep_sleep_control)(struct smu_context *smu, bool enablement); + + /** + * @get_fan_parameters: Get fan parameters. + * + * Get maximum fan speed from the power play table. + */ int (*get_fan_parameters)(struct smu_context *smu); + + /** + * @post_init: Helper function for asic specific workarounds. + */ int (*post_init)(struct smu_context *smu); + + /** + * @interrupt_work: Work task scheduled from SMU interrupt handler. + */ void (*interrupt_work)(struct smu_context *smu); + + /** + * @gpo_control: Enable/disable graphics power optimization if supported. + */ int (*gpo_control)(struct smu_context *smu, bool enablement); + + /** + * @gfx_state_change_set: Send the current graphics state to the SMU. + */ int (*gfx_state_change_set)(struct smu_context *smu, uint32_t state); + + /** + * @set_fine_grain_gfx_freq_parameters: Set fine grain graphics clock + * parameters to defaults. + */ int (*set_fine_grain_gfx_freq_parameters)(struct smu_context *smu); }; @@ -636,6 +1165,12 @@ enum smu_cmn2asic_mapping_type { #define FEA_MAP(fea) \ [SMU_FEATURE_##fea##_BIT] = {1, FEATURE_##fea##_BIT} +#define FEA_MAP_REVERSE(fea) \ + [SMU_FEATURE_DPM_##fea##_BIT] = {1, FEATURE_##fea##_DPM_BIT} + +#define FEA_MAP_HALF_REVERSE(fea) \ + [SMU_FEATURE_DPM_##fea##CLK_BIT] = {1, FEATURE_##fea##_DPM_BIT} + #define TAB_MAP(tab) \ [SMU_TABLE_##tab] = {1, TABLE_##tab} diff --git a/drivers/gpu/drm/amd/pm/inc/smu11_driver_if_vangogh.h b/drivers/gpu/drm/amd/pm/inc/smu11_driver_if_vangogh.h index 1c19eae93ff1..6e23a3f803a7 100644 --- a/drivers/gpu/drm/amd/pm/inc/smu11_driver_if_vangogh.h +++ b/drivers/gpu/drm/amd/pm/inc/smu11_driver_if_vangogh.h @@ -141,7 +141,6 @@ typedef struct { uint32_t MaxGfxClk; uint8_t NumDfPstatesEnabled; - uint8_t NumDpmLevelsEnabled; uint8_t NumDcfclkLevelsEnabled; uint8_t NumDispClkLevelsEnabled; //applies to both dispclk and dppclk uint8_t NumSocClkLevelsEnabled; diff --git a/drivers/gpu/drm/amd/pm/inc/smu_types.h b/drivers/gpu/drm/amd/pm/inc/smu_types.h index 720d15612fe1..8e428c728e0e 100644 --- a/drivers/gpu/drm/amd/pm/inc/smu_types.h +++ b/drivers/gpu/drm/amd/pm/inc/smu_types.h @@ -211,6 +211,7 @@ __SMU_DUMMY_MAP(SetGpoFeaturePMask), \ __SMU_DUMMY_MAP(DisallowGpo), \ __SMU_DUMMY_MAP(Enable2ndUSB20Port), \ + __SMU_DUMMY_MAP(RequestActiveWgp), \ #undef __SMU_DUMMY_MAP #define __SMU_DUMMY_MAP(type) SMU_MSG_##type @@ -240,6 +241,7 @@ enum smu_clk_type { SMU_OD_MCLK, SMU_OD_VDDC_CURVE, SMU_OD_RANGE, + SMU_OD_VDDGFX_OFFSET, SMU_CLK_COUNT, }; diff --git a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu10_hwmgr.c b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu10_hwmgr.c index e57e64bbacdc..88322781e447 100644 --- a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu10_hwmgr.c +++ b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu10_hwmgr.c @@ -251,7 +251,7 @@ static int smu10_set_hard_min_gfxclk_by_freq(struct pp_hwmgr *hwmgr, uint32_t cl smu10_data->gfx_actual_soft_min_freq = clock; smum_send_msg_to_smc_with_parameter(hwmgr, PPSMC_MSG_SetHardMinGfxClk, - smu10_data->gfx_actual_soft_min_freq, + clock, NULL); } return 0; @@ -558,7 +558,8 @@ static int smu10_hwmgr_backend_init(struct pp_hwmgr *hwmgr) /* enable the pp_od_clk_voltage sysfs file */ hwmgr->od_enabled = 1; - + /* disabled fine grain tuning function by default */ + data->fine_grain_enabled = 0; return result; } @@ -597,6 +598,7 @@ static int smu10_dpm_force_dpm_level(struct pp_hwmgr *hwmgr, uint32_t min_mclk = hwmgr->display_config->min_mem_set_clock/100; uint32_t index_fclk = data->clock_vol_info.vdd_dep_on_fclk->count - 1; uint32_t index_socclk = data->clock_vol_info.vdd_dep_on_socclk->count - 1; + uint32_t fine_grain_min_freq = 0, fine_grain_max_freq = 0; if (hwmgr->smu_version < 0x1E3700) { pr_info("smu firmware version too old, can not set dpm level\n"); @@ -613,6 +615,14 @@ static int smu10_dpm_force_dpm_level(struct pp_hwmgr *hwmgr, switch (level) { case AMD_DPM_FORCED_LEVEL_HIGH: case AMD_DPM_FORCED_LEVEL_PROFILE_PEAK: + data->fine_grain_enabled = 0; + + smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetMinGfxclkFrequency, &fine_grain_min_freq); + smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetMaxGfxclkFrequency, &fine_grain_max_freq); + + data->gfx_actual_soft_min_freq = fine_grain_min_freq; + data->gfx_actual_soft_max_freq = fine_grain_max_freq; + smum_send_msg_to_smc_with_parameter(hwmgr, PPSMC_MSG_SetHardMinGfxClk, data->gfx_max_freq_limit/100, @@ -648,6 +658,14 @@ static int smu10_dpm_force_dpm_level(struct pp_hwmgr *hwmgr, NULL); break; case AMD_DPM_FORCED_LEVEL_PROFILE_MIN_SCLK: + data->fine_grain_enabled = 0; + + smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetMinGfxclkFrequency, &fine_grain_min_freq); + smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetMaxGfxclkFrequency, &fine_grain_max_freq); + + data->gfx_actual_soft_min_freq = fine_grain_min_freq; + data->gfx_actual_soft_max_freq = fine_grain_max_freq; + smum_send_msg_to_smc_with_parameter(hwmgr, PPSMC_MSG_SetHardMinGfxClk, min_sclk, @@ -658,6 +676,14 @@ static int smu10_dpm_force_dpm_level(struct pp_hwmgr *hwmgr, NULL); break; case AMD_DPM_FORCED_LEVEL_PROFILE_MIN_MCLK: + data->fine_grain_enabled = 0; + + smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetMinGfxclkFrequency, &fine_grain_min_freq); + smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetMaxGfxclkFrequency, &fine_grain_max_freq); + + data->gfx_actual_soft_min_freq = fine_grain_min_freq; + data->gfx_actual_soft_max_freq = fine_grain_max_freq; + smum_send_msg_to_smc_with_parameter(hwmgr, PPSMC_MSG_SetHardMinFclkByFreq, min_mclk, @@ -668,6 +694,14 @@ static int smu10_dpm_force_dpm_level(struct pp_hwmgr *hwmgr, NULL); break; case AMD_DPM_FORCED_LEVEL_PROFILE_STANDARD: + data->fine_grain_enabled = 0; + + smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetMinGfxclkFrequency, &fine_grain_min_freq); + smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetMaxGfxclkFrequency, &fine_grain_max_freq); + + data->gfx_actual_soft_min_freq = fine_grain_min_freq; + data->gfx_actual_soft_max_freq = fine_grain_max_freq; + smum_send_msg_to_smc_with_parameter(hwmgr, PPSMC_MSG_SetHardMinGfxClk, SMU10_UMD_PSTATE_GFXCLK, @@ -703,6 +737,14 @@ static int smu10_dpm_force_dpm_level(struct pp_hwmgr *hwmgr, NULL); break; case AMD_DPM_FORCED_LEVEL_AUTO: + data->fine_grain_enabled = 0; + + smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetMinGfxclkFrequency, &fine_grain_min_freq); + smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetMaxGfxclkFrequency, &fine_grain_max_freq); + + data->gfx_actual_soft_min_freq = fine_grain_min_freq; + data->gfx_actual_soft_max_freq = fine_grain_max_freq; + smum_send_msg_to_smc_with_parameter(hwmgr, PPSMC_MSG_SetHardMinGfxClk, min_sclk, @@ -741,6 +783,14 @@ static int smu10_dpm_force_dpm_level(struct pp_hwmgr *hwmgr, NULL); break; case AMD_DPM_FORCED_LEVEL_LOW: + data->fine_grain_enabled = 0; + + smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetMinGfxclkFrequency, &fine_grain_min_freq); + smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetMaxGfxclkFrequency, &fine_grain_max_freq); + + data->gfx_actual_soft_min_freq = fine_grain_min_freq; + data->gfx_actual_soft_max_freq = fine_grain_max_freq; + smum_send_msg_to_smc_with_parameter(hwmgr, PPSMC_MSG_SetHardMinGfxClk, data->gfx_min_freq_limit/100, @@ -759,6 +809,7 @@ static int smu10_dpm_force_dpm_level(struct pp_hwmgr *hwmgr, NULL); break; case AMD_DPM_FORCED_LEVEL_MANUAL: + data->fine_grain_enabled = 1; case AMD_DPM_FORCED_LEVEL_PROFILE_EXIT: default: break; @@ -948,6 +999,8 @@ static int smu10_print_clock_levels(struct pp_hwmgr *hwmgr, struct smu10_voltage_dependency_table *mclk_table = data->clock_vol_info.vdd_dep_on_fclk; uint32_t i, now, size = 0; + uint32_t min_freq, max_freq = 0; + uint32_t ret = 0; switch (type) { case PP_SCLK: @@ -983,18 +1036,28 @@ static int smu10_print_clock_levels(struct pp_hwmgr *hwmgr, break; case OD_SCLK: if (hwmgr->od_enabled) { - size = sprintf(buf, "%s:\n", "OD_SCLK"); + ret = smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetMinGfxclkFrequency, &min_freq); + if (ret) + return ret; + ret = smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetMaxGfxclkFrequency, &max_freq); + if (ret) + return ret; + size = sprintf(buf, "%s:\n", "OD_SCLK"); size += sprintf(buf + size, "0: %10uMhz\n", - (data->gfx_actual_soft_min_freq > 0) ? data->gfx_actual_soft_min_freq : data->gfx_min_freq_limit/100); - size += sprintf(buf + size, "1: %10uMhz\n", data->gfx_max_freq_limit/100); + (data->gfx_actual_soft_min_freq > 0) ? data->gfx_actual_soft_min_freq : min_freq); + size += sprintf(buf + size, "1: %10uMhz\n", + (data->gfx_actual_soft_max_freq > 0) ? data->gfx_actual_soft_max_freq : max_freq); } break; case OD_RANGE: if (hwmgr->od_enabled) { - uint32_t min_freq, max_freq = 0; - smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetMinGfxclkFrequency, &min_freq); - smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetMaxGfxclkFrequency, &max_freq); + ret = smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetMinGfxclkFrequency, &min_freq); + if (ret) + return ret; + ret = smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetMaxGfxclkFrequency, &max_freq); + if (ret) + return ret; size = sprintf(buf, "%s:\n", "OD_RANGE"); size += sprintf(buf + size, "SCLK: %7uMHz %10uMHz\n", @@ -1414,23 +1477,96 @@ static int smu10_set_fine_grain_clk_vol(struct pp_hwmgr *hwmgr, enum PP_OD_DPM_TABLE_COMMAND type, long *input, uint32_t size) { + uint32_t min_freq, max_freq = 0; + struct smu10_hwmgr *smu10_data = (struct smu10_hwmgr *)(hwmgr->backend); + int ret = 0; + if (!hwmgr->od_enabled) { pr_err("Fine grain not support\n"); return -EINVAL; } - if (size != 2) { - pr_err("Input parameter number not correct\n"); + if (!smu10_data->fine_grain_enabled) { + pr_err("Fine grain not started\n"); return -EINVAL; } if (type == PP_OD_EDIT_SCLK_VDDC_TABLE) { - if (input[0] == 0) - smu10_set_hard_min_gfxclk_by_freq(hwmgr, input[1]); - else if (input[0] == 1) - smu10_set_soft_max_gfxclk_by_freq(hwmgr, input[1]); - else + if (size != 2) { + pr_err("Input parameter number not correct\n"); return -EINVAL; + } + + if (input[0] == 0) { + smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetMinGfxclkFrequency, &min_freq); + if (input[1] < min_freq) { + pr_err("Fine grain setting minimum sclk (%ld) MHz is less than the minimum allowed (%d) MHz\n", + input[1], min_freq); + return -EINVAL; + } + smu10_data->gfx_actual_soft_min_freq = input[1]; + } else if (input[0] == 1) { + smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetMaxGfxclkFrequency, &max_freq); + if (input[1] > max_freq) { + pr_err("Fine grain setting maximum sclk (%ld) MHz is greater than the maximum allowed (%d) MHz\n", + input[1], max_freq); + return -EINVAL; + } + smu10_data->gfx_actual_soft_max_freq = input[1]; + } else { + return -EINVAL; + } + } else if (type == PP_OD_RESTORE_DEFAULT_TABLE) { + if (size != 0) { + pr_err("Input parameter number not correct\n"); + return -EINVAL; + } + smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetMinGfxclkFrequency, &min_freq); + smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetMaxGfxclkFrequency, &max_freq); + + smu10_data->gfx_actual_soft_min_freq = min_freq; + smu10_data->gfx_actual_soft_max_freq = max_freq; + + ret = smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_SetHardMinGfxClk, + min_freq, + NULL); + if (ret) + return ret; + + ret = smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_SetSoftMaxGfxClk, + max_freq, + NULL); + if (ret) + return ret; + } else if (type == PP_OD_COMMIT_DPM_TABLE) { + if (size != 0) { + pr_err("Input parameter number not correct\n"); + return -EINVAL; + } + + if (smu10_data->gfx_actual_soft_min_freq > smu10_data->gfx_actual_soft_max_freq) { + pr_err("The setting minimun sclk (%d) MHz is greater than the setting maximum sclk (%d) MHz\n", + smu10_data->gfx_actual_soft_min_freq, smu10_data->gfx_actual_soft_max_freq); + return -EINVAL; + } + + ret = smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_SetHardMinGfxClk, + smu10_data->gfx_actual_soft_min_freq, + NULL); + if (ret) + return ret; + + ret = smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_SetSoftMaxGfxClk, + smu10_data->gfx_actual_soft_max_freq, + NULL); + if (ret) + return ret; + } else { + return -EINVAL; } return 0; diff --git a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu10_hwmgr.h b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu10_hwmgr.h index 6c9b5f060902..808e0ecbe1f0 100644 --- a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu10_hwmgr.h +++ b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu10_hwmgr.h @@ -283,6 +283,7 @@ struct smu10_hwmgr { uint32_t vclk_soft_min; uint32_t dclk_soft_min; uint32_t gfx_actual_soft_min_freq; + uint32_t gfx_actual_soft_max_freq; uint32_t gfx_min_freq_limit; uint32_t gfx_max_freq_limit; /* in 10Khz*/ @@ -299,6 +300,8 @@ struct smu10_hwmgr { bool need_min_deep_sleep_dcefclk; uint32_t deep_sleep_dcefclk; uint32_t num_active_display; + + bool fine_grain_enabled; }; struct pp_hwmgr; diff --git a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c index 8b867a6d52b5..8e1e97e31411 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c +++ b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c @@ -402,6 +402,10 @@ static int smu_set_funcs(struct amdgpu_device *adev) break; case CHIP_RENOIR: renoir_set_ppt_funcs(smu); + /* enable the fine grain tuning function by default */ + smu->fine_grain_enabled = true; + /* close the fine grain tuning function by default */ + smu->fine_grain_started = false; break; case CHIP_VANGOGH: vangogh_set_ppt_funcs(smu); @@ -478,9 +482,6 @@ static int smu_late_init(void *handle) smu_set_fine_grain_gfx_freq_parameters(smu); - if (adev->asic_type == CHIP_VANGOGH) - return 0; - if (!smu->pm_enabled) return 0; @@ -490,6 +491,9 @@ static int smu_late_init(void *handle) return ret; } + if (adev->asic_type == CHIP_VANGOGH) + return 0; + ret = smu_set_default_od_settings(smu); if (ret) { dev_err(adev->dev, "Failed to setup default OD settings!\n"); diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu11/navi10_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu11/navi10_ppt.c index 51e83123f72a..7ebf9588983f 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu11/navi10_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu11/navi10_ppt.c @@ -1673,7 +1673,7 @@ static int navi10_read_sensor(struct smu_context *smu, *size = 4; break; case AMDGPU_PP_SENSOR_GFX_SCLK: - ret = navi10_get_current_clk_freq_by_table(smu, SMU_GFXCLK, (uint32_t *)data); + ret = navi10_get_smu_metrics_data(smu, METRICS_AVERAGE_GFXCLK, (uint32_t *)data); *(uint32_t *)data *= 100; *size = 4; break; 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 9608745d732f..24f3c96a5e5e 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 @@ -314,6 +314,12 @@ static int sienna_cichlid_check_powerplay_table(struct smu_context *smu) table_context->thermal_controller_type = powerplay_table->thermal_controller_type; + /* + * Instead of having its own buffer space and get overdrive_table copied, + * smu->od_settings just points to the actual overdrive_table + */ + smu->od_settings = &powerplay_table->overdrive_table; + return 0; } @@ -907,6 +913,22 @@ static bool sienna_cichlid_is_support_fine_grained_dpm(struct smu_context *smu, return dpm_desc->SnapToDiscrete == 0 ? true : false; } +static bool sienna_cichlid_is_od_feature_supported(struct smu_11_0_7_overdrive_table *od_table, + enum SMU_11_0_7_ODFEATURE_CAP cap) +{ + return od_table->cap[cap]; +} + +static void sienna_cichlid_get_od_setting_range(struct smu_11_0_7_overdrive_table *od_table, + enum SMU_11_0_7_ODSETTING_ID setting, + uint32_t *min, uint32_t *max) +{ + if (min) + *min = od_table->min[setting]; + if (max) + *max = od_table->max[setting]; +} + static int sienna_cichlid_print_clk_levels(struct smu_context *smu, enum smu_clk_type clk_type, char *buf) { @@ -915,11 +937,16 @@ static int sienna_cichlid_print_clk_levels(struct smu_context *smu, struct smu_dpm_context *smu_dpm = &smu->smu_dpm; struct smu_11_0_dpm_context *dpm_context = smu_dpm->dpm_context; PPTable_t *pptable = (PPTable_t *)table_context->driver_pptable; + struct smu_11_0_7_overdrive_table *od_settings = smu->od_settings; + OverDriveTable_t *od_table = + (OverDriveTable_t *)table_context->overdrive_table; int i, size = 0, ret = 0; uint32_t cur_value = 0, value = 0, count = 0; uint32_t freq_values[3] = {0}; uint32_t mark_index = 0; uint32_t gen_speed, lane_width; + uint32_t min_value, max_value; + uint32_t smu_version; switch (clk_type) { case SMU_GFXCLK: @@ -995,6 +1022,70 @@ static int sienna_cichlid_print_clk_levels(struct smu_context *smu, (lane_width == dpm_context->dpm_tables.pcie_table.pcie_lane[i]) ? "*" : ""); break; + case SMU_OD_SCLK: + if (!smu->od_enabled || !od_table || !od_settings) + break; + + if (!sienna_cichlid_is_od_feature_supported(od_settings, SMU_11_0_7_ODCAP_GFXCLK_LIMITS)) + break; + + size += sprintf(buf + size, "OD_SCLK:\n"); + size += sprintf(buf + size, "0: %uMhz\n1: %uMhz\n", od_table->GfxclkFmin, od_table->GfxclkFmax); + break; + + case SMU_OD_MCLK: + if (!smu->od_enabled || !od_table || !od_settings) + break; + + if (!sienna_cichlid_is_od_feature_supported(od_settings, SMU_11_0_7_ODCAP_UCLK_LIMITS)) + break; + + size += sprintf(buf + size, "OD_MCLK:\n"); + size += sprintf(buf + size, "0: %uMhz\n1: %uMHz\n", od_table->UclkFmin, od_table->UclkFmax); + break; + + case SMU_OD_VDDGFX_OFFSET: + if (!smu->od_enabled || !od_table || !od_settings) + break; + + /* + * OD GFX Voltage Offset functionality is supported only by 58.41.0 + * and onwards SMU firmwares. + */ + smu_cmn_get_smc_version(smu, NULL, &smu_version); + if ((adev->asic_type == CHIP_SIENNA_CICHLID) && + (smu_version < 0x003a2900)) + break; + + size += sprintf(buf + size, "OD_VDDGFX_OFFSET:\n"); + size += sprintf(buf + size, "%dmV\n", od_table->VddGfxOffset); + break; + + case SMU_OD_RANGE: + if (!smu->od_enabled || !od_table || !od_settings) + break; + + size = sprintf(buf, "%s:\n", "OD_RANGE"); + + if (sienna_cichlid_is_od_feature_supported(od_settings, SMU_11_0_7_ODCAP_GFXCLK_LIMITS)) { + sienna_cichlid_get_od_setting_range(od_settings, SMU_11_0_7_ODSETTING_GFXCLKFMIN, + &min_value, NULL); + sienna_cichlid_get_od_setting_range(od_settings, SMU_11_0_7_ODSETTING_GFXCLKFMAX, + NULL, &max_value); + size += sprintf(buf + size, "SCLK: %7uMhz %10uMhz\n", + min_value, max_value); + } + + if (sienna_cichlid_is_od_feature_supported(od_settings, SMU_11_0_7_ODCAP_UCLK_LIMITS)) { + sienna_cichlid_get_od_setting_range(od_settings, SMU_11_0_7_ODSETTING_UCLKFMIN, + &min_value, NULL); + sienna_cichlid_get_od_setting_range(od_settings, SMU_11_0_7_ODSETTING_UCLKFMAX, + NULL, &max_value); + size += sprintf(buf + size, "MCLK: %7uMhz %10uMhz\n", + min_value, max_value); + } + break; + default: break; } @@ -1694,6 +1785,243 @@ static int sienna_cichlid_get_dpm_ultimate_freq(struct smu_context *smu, return ret; } +static void sienna_cichlid_dump_od_table(struct smu_context *smu, + OverDriveTable_t *od_table) +{ + struct amdgpu_device *adev = smu->adev; + uint32_t smu_version; + + dev_dbg(smu->adev->dev, "OD: Gfxclk: (%d, %d)\n", od_table->GfxclkFmin, + od_table->GfxclkFmax); + dev_dbg(smu->adev->dev, "OD: Uclk: (%d, %d)\n", od_table->UclkFmin, + od_table->UclkFmax); + + smu_cmn_get_smc_version(smu, NULL, &smu_version); + if (!((adev->asic_type == CHIP_SIENNA_CICHLID) && + (smu_version < 0x003a2900))) + dev_dbg(smu->adev->dev, "OD: VddGfxOffset: %d\n", od_table->VddGfxOffset); +} + +static int sienna_cichlid_set_default_od_settings(struct smu_context *smu) +{ + OverDriveTable_t *od_table = + (OverDriveTable_t *)smu->smu_table.overdrive_table; + OverDriveTable_t *boot_od_table = + (OverDriveTable_t *)smu->smu_table.boot_overdrive_table; + int ret = 0; + + ret = smu_cmn_update_table(smu, SMU_TABLE_OVERDRIVE, + 0, (void *)od_table, false); + if (ret) { + dev_err(smu->adev->dev, "Failed to get overdrive table!\n"); + return ret; + } + + memcpy(boot_od_table, od_table, sizeof(OverDriveTable_t)); + + sienna_cichlid_dump_od_table(smu, od_table); + + return 0; +} + +static int sienna_cichlid_od_setting_check_range(struct smu_context *smu, + struct smu_11_0_7_overdrive_table *od_table, + enum SMU_11_0_7_ODSETTING_ID setting, + uint32_t value) +{ + if (value < od_table->min[setting]) { + dev_warn(smu->adev->dev, "OD setting (%d, %d) is less than the minimum allowed (%d)\n", + setting, value, od_table->min[setting]); + return -EINVAL; + } + if (value > od_table->max[setting]) { + dev_warn(smu->adev->dev, "OD setting (%d, %d) is greater than the maximum allowed (%d)\n", + setting, value, od_table->max[setting]); + return -EINVAL; + } + + return 0; +} + +static int sienna_cichlid_od_edit_dpm_table(struct smu_context *smu, + enum PP_OD_DPM_TABLE_COMMAND type, + long input[], uint32_t size) +{ + struct smu_table_context *table_context = &smu->smu_table; + OverDriveTable_t *od_table = + (OverDriveTable_t *)table_context->overdrive_table; + struct smu_11_0_7_overdrive_table *od_settings = + (struct smu_11_0_7_overdrive_table *)smu->od_settings; + struct amdgpu_device *adev = smu->adev; + enum SMU_11_0_7_ODSETTING_ID freq_setting; + uint16_t *freq_ptr; + int i, ret = 0; + uint32_t smu_version; + + if (!smu->od_enabled) { + dev_warn(smu->adev->dev, "OverDrive is not enabled!\n"); + return -EINVAL; + } + + if (!smu->od_settings) { + dev_err(smu->adev->dev, "OD board limits are not set!\n"); + return -ENOENT; + } + + if (!(table_context->overdrive_table && table_context->boot_overdrive_table)) { + dev_err(smu->adev->dev, "Overdrive table was not initialized!\n"); + return -EINVAL; + } + + switch (type) { + case PP_OD_EDIT_SCLK_VDDC_TABLE: + if (!sienna_cichlid_is_od_feature_supported(od_settings, + SMU_11_0_7_ODCAP_GFXCLK_LIMITS)) { + dev_warn(smu->adev->dev, "GFXCLK_LIMITS not supported!\n"); + return -ENOTSUPP; + } + + for (i = 0; i < size; i += 2) { + if (i + 2 > size) { + dev_info(smu->adev->dev, "invalid number of input parameters %d\n", size); + return -EINVAL; + } + + switch (input[i]) { + case 0: + if (input[i + 1] > od_table->GfxclkFmax) { + dev_info(smu->adev->dev, "GfxclkFmin (%ld) must be <= GfxclkFmax (%u)!\n", + input[i + 1], od_table->GfxclkFmax); + return -EINVAL; + } + + freq_setting = SMU_11_0_7_ODSETTING_GFXCLKFMIN; + freq_ptr = &od_table->GfxclkFmin; + break; + + case 1: + if (input[i + 1] < od_table->GfxclkFmin) { + dev_info(smu->adev->dev, "GfxclkFmax (%ld) must be >= GfxclkFmin (%u)!\n", + input[i + 1], od_table->GfxclkFmin); + return -EINVAL; + } + + freq_setting = SMU_11_0_7_ODSETTING_GFXCLKFMAX; + freq_ptr = &od_table->GfxclkFmax; + break; + + default: + dev_info(smu->adev->dev, "Invalid SCLK_VDDC_TABLE index: %ld\n", input[i]); + dev_info(smu->adev->dev, "Supported indices: [0:min,1:max]\n"); + return -EINVAL; + } + + ret = sienna_cichlid_od_setting_check_range(smu, od_settings, + freq_setting, input[i + 1]); + if (ret) + return ret; + + *freq_ptr = (uint16_t)input[i + 1]; + } + break; + + case PP_OD_EDIT_MCLK_VDDC_TABLE: + if (!sienna_cichlid_is_od_feature_supported(od_settings, SMU_11_0_7_ODCAP_UCLK_LIMITS)) { + dev_warn(smu->adev->dev, "UCLK_LIMITS not supported!\n"); + return -ENOTSUPP; + } + + for (i = 0; i < size; i += 2) { + if (i + 2 > size) { + dev_info(smu->adev->dev, "invalid number of input parameters %d\n", size); + return -EINVAL; + } + + switch (input[i]) { + case 0: + if (input[i + 1] > od_table->UclkFmax) { + dev_info(smu->adev->dev, "UclkFmin (%ld) must be <= UclkFmax (%u)!\n", + input[i + 1], od_table->UclkFmax); + return -EINVAL; + } + + freq_setting = SMU_11_0_7_ODSETTING_UCLKFMIN; + freq_ptr = &od_table->UclkFmin; + break; + + case 1: + if (input[i + 1] < od_table->UclkFmin) { + dev_info(smu->adev->dev, "UclkFmax (%ld) must be >= UclkFmin (%u)!\n", + input[i + 1], od_table->UclkFmin); + return -EINVAL; + } + + freq_setting = SMU_11_0_7_ODSETTING_UCLKFMAX; + freq_ptr = &od_table->UclkFmax; + break; + + default: + dev_info(smu->adev->dev, "Invalid MCLK_VDDC_TABLE index: %ld\n", input[i]); + dev_info(smu->adev->dev, "Supported indices: [0:min,1:max]\n"); + return -EINVAL; + } + + ret = sienna_cichlid_od_setting_check_range(smu, od_settings, + freq_setting, input[i + 1]); + if (ret) + return ret; + + *freq_ptr = (uint16_t)input[i + 1]; + } + break; + + case PP_OD_RESTORE_DEFAULT_TABLE: + memcpy(table_context->overdrive_table, + table_context->boot_overdrive_table, + sizeof(OverDriveTable_t)); + fallthrough; + + case PP_OD_COMMIT_DPM_TABLE: + sienna_cichlid_dump_od_table(smu, od_table); + + ret = smu_cmn_update_table(smu, SMU_TABLE_OVERDRIVE, + 0, (void *)od_table, true); + if (ret) { + dev_err(smu->adev->dev, "Failed to import overdrive table!\n"); + return ret; + } + break; + + case PP_OD_EDIT_VDDGFX_OFFSET: + if (size != 1) { + dev_info(smu->adev->dev, "invalid number of parameters: %d\n", size); + return -EINVAL; + } + + /* + * OD GFX Voltage Offset functionality is supported only by 58.41.0 + * and onwards SMU firmwares. + */ + smu_cmn_get_smc_version(smu, NULL, &smu_version); + if ((adev->asic_type == CHIP_SIENNA_CICHLID) && + (smu_version < 0x003a2900)) { + dev_err(smu->adev->dev, "OD GFX Voltage offset functionality is supported " + "only by 58.41.0 and onwards SMU firmwares!\n"); + return -EOPNOTSUPP; + } + + od_table->VddGfxOffset = (int16_t)input[0]; + + sienna_cichlid_dump_od_table(smu, od_table); + break; + + default: + return -ENOSYS; + } + + return ret; +} + static int sienna_cichlid_run_btc(struct smu_context *smu) { return smu_cmn_send_smc_msg(smu, SMU_MSG_RunDcBtc, NULL); @@ -2372,7 +2700,7 @@ static void sienna_cichlid_fill_i2c_req(SwI2cRequest_t *req, bool write, { int i; - req->I2CcontrollerPort = 0; + req->I2CcontrollerPort = 1; req->I2CSpeed = 2; req->SlaveAddress = address; req->NumCmds = numbytes; @@ -2817,6 +3145,8 @@ static const struct pptable_funcs sienna_cichlid_ppt_funcs = { .mode1_reset = smu_v11_0_mode1_reset, .get_dpm_ultimate_freq = sienna_cichlid_get_dpm_ultimate_freq, .set_soft_freq_limited_range = smu_v11_0_set_soft_freq_limited_range, + .set_default_od_settings = sienna_cichlid_set_default_od_settings, + .od_edit_dpm_table = sienna_cichlid_od_edit_dpm_table, .run_btc = sienna_cichlid_run_btc, .set_power_source = smu_v11_0_set_power_source, .get_pp_feature_mask = smu_cmn_get_pp_feature_mask, diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu11/vangogh_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu11/vangogh_ppt.c index 8cb4fcee9a2c..a79dd04f81a2 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu11/vangogh_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu11/vangogh_ppt.c @@ -31,6 +31,9 @@ #include "smu_v11_5_ppsmc.h" #include "smu_v11_5_pmfw.h" #include "smu_cmn.h" +#include "soc15_common.h" +#include "asic_reg/gc/gc_10_3_0_offset.h" +#include "asic_reg/gc/gc_10_3_0_sh_mask.h" /* * DO NOT use these for err/warn/info/debug messages. @@ -118,6 +121,7 @@ static struct cmn2asic_msg_mapping vangogh_message_map[SMU_MSG_MAX_COUNT] = { MSG_MAP(StopDramLogging, PPSMC_MSG_StopDramLogging, 0), MSG_MAP(SetSoftMinCclk, PPSMC_MSG_SetSoftMinCclk, 0), MSG_MAP(SetSoftMaxCclk, PPSMC_MSG_SetSoftMaxCclk, 0), + MSG_MAP(RequestActiveWgp, PPSMC_MSG_RequestActiveWgp, 0), }; static struct cmn2asic_mapping vangogh_feature_mask_map[SMU_FEATURE_COUNT] = { @@ -162,6 +166,9 @@ static struct cmn2asic_mapping vangogh_feature_mask_map[SMU_FEATURE_COUNT] = { FEA_MAP(A55_DPM), FEA_MAP(CVIP_DSP_DPM), FEA_MAP(MSMU_LOW_POWER), + FEA_MAP_REVERSE(SOCCLK), + FEA_MAP_REVERSE(FCLK), + FEA_MAP_HALF_REVERSE(GFX), }; static struct cmn2asic_mapping vangogh_table_map[SMU_TABLE_COUNT] = { @@ -242,6 +249,12 @@ static int vangogh_get_smu_metrics_data(struct smu_context *smu, case METRICS_AVERAGE_SOCCLK: *value = metrics->SocclkFrequency; break; + case METRICS_AVERAGE_VCLK: + *value = metrics->VclkFrequency; + break; + case METRICS_AVERAGE_DCLK: + *value = metrics->DclkFrequency; + break; case METRICS_AVERAGE_UCLK: *value = metrics->MemclkFrequency; break; @@ -252,7 +265,8 @@ static int vangogh_get_smu_metrics_data(struct smu_context *smu, *value = metrics->UvdActivity; break; case METRICS_AVERAGE_SOCKETPOWER: - *value = metrics->CurrentSocketPower; + *value = (metrics->CurrentSocketPower << 8) / + 1000 ; break; case METRICS_TEMPERATURE_EDGE: *value = metrics->GfxTemperature / 100 * @@ -366,6 +380,10 @@ static int vangogh_get_allowed_feature_mask(struct smu_context *smu, *(uint64_t *)feature_mask |= FEATURE_MASK(FEATURE_GFX_DPM_BIT) | FEATURE_MASK(FEATURE_MP0CLK_DPM_BIT) + | FEATURE_MASK(FEATURE_SOCCLK_DPM_BIT) + | FEATURE_MASK(FEATURE_VCN_DPM_BIT) + | FEATURE_MASK(FEATURE_FCLK_DPM_BIT) + | FEATURE_MASK(FEATURE_DCFCLK_DPM_BIT) | FEATURE_MASK(FEATURE_DS_SOCCLK_BIT) | FEATURE_MASK(FEATURE_PPT_BIT) | FEATURE_MASK(FEATURE_TDC_BIT) @@ -379,6 +397,12 @@ static int vangogh_get_allowed_feature_mask(struct smu_context *smu, if (adev->pm.pp_feature & PP_DCEFCLK_DPM_MASK) *(uint64_t *)feature_mask |= FEATURE_MASK(FEATURE_DCFCLK_DPM_BIT); + if (adev->pm.pp_feature & PP_MCLK_DPM_MASK) + *(uint64_t *)feature_mask |= FEATURE_MASK(FEATURE_FCLK_DPM_BIT); + + if (adev->pm.pp_feature & PP_SCLK_DPM_MASK) + *(uint64_t *)feature_mask |= FEATURE_MASK(FEATURE_GFX_DPM_BIT); + if (smu->adev->pg_flags & AMD_PG_SUPPORT_ATHUB) *(uint64_t *)feature_mask |= FEATURE_MASK(FEATURE_ATHUB_PG_BIT); @@ -402,10 +426,63 @@ static bool vangogh_is_dpm_running(struct smu_context *smu) return !!(feature_enabled & SMC_DPM_FEATURE); } +static int vangogh_get_dpm_clk_limited(struct smu_context *smu, enum smu_clk_type clk_type, + uint32_t dpm_level, uint32_t *freq) +{ + DpmClocks_t *clk_table = smu->smu_table.clocks_table; + + if (!clk_table || clk_type >= SMU_CLK_COUNT) + return -EINVAL; + + switch (clk_type) { + case SMU_SOCCLK: + if (dpm_level >= clk_table->NumSocClkLevelsEnabled) + return -EINVAL; + *freq = clk_table->SocClocks[dpm_level]; + break; + case SMU_VCLK: + if (dpm_level >= clk_table->VcnClkLevelsEnabled) + return -EINVAL; + *freq = clk_table->VcnClocks[dpm_level].vclk; + break; + case SMU_DCLK: + if (dpm_level >= clk_table->VcnClkLevelsEnabled) + return -EINVAL; + *freq = clk_table->VcnClocks[dpm_level].dclk; + break; + case SMU_UCLK: + case SMU_MCLK: + if (dpm_level >= clk_table->NumDfPstatesEnabled) + return -EINVAL; + *freq = clk_table->DfPstateTable[dpm_level].memclk; + + break; + case SMU_FCLK: + if (dpm_level >= clk_table->NumDfPstatesEnabled) + return -EINVAL; + *freq = clk_table->DfPstateTable[dpm_level].fclk; + break; + default: + return -EINVAL; + } + + return 0; +} + static int vangogh_print_fine_grain_clk(struct smu_context *smu, enum smu_clk_type clk_type, char *buf) { - int size = 0; + DpmClocks_t *clk_table = smu->smu_table.clocks_table; + SmuMetrics_t metrics; + int i, size = 0, ret = 0; + uint32_t cur_value = 0, value = 0, count = 0; + bool cur_value_match_level = false; + + memset(&metrics, 0, sizeof(metrics)); + + ret = smu_cmn_get_metrics_table(smu, &metrics, false); + if (ret) + return ret; switch (clk_type) { case SMU_OD_SCLK: @@ -424,13 +501,733 @@ static int vangogh_print_fine_grain_clk(struct smu_context *smu, smu->gfx_default_hard_min_freq, smu->gfx_default_soft_max_freq); } break; + case SMU_SOCCLK: + /* the level 3 ~ 6 of socclk use the same frequency for vangogh */ + count = clk_table->NumSocClkLevelsEnabled; + cur_value = metrics.SocclkFrequency; + break; + case SMU_VCLK: + count = clk_table->VcnClkLevelsEnabled; + cur_value = metrics.VclkFrequency; + break; + case SMU_DCLK: + count = clk_table->VcnClkLevelsEnabled; + cur_value = metrics.DclkFrequency; + break; + case SMU_MCLK: + count = clk_table->NumDfPstatesEnabled; + cur_value = metrics.MemclkFrequency; + break; + case SMU_FCLK: + count = clk_table->NumDfPstatesEnabled; + ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_GetFclkFrequency, 0, &cur_value); + if (ret) + return ret; + break; default: break; } + switch (clk_type) { + case SMU_SOCCLK: + case SMU_VCLK: + case SMU_DCLK: + case SMU_MCLK: + case SMU_FCLK: + for (i = 0; i < count; i++) { + ret = vangogh_get_dpm_clk_limited(smu, clk_type, i, &value); + if (ret) + return ret; + if (!value) + continue; + size += sprintf(buf + size, "%d: %uMhz %s\n", i, value, + cur_value == value ? "*" : ""); + if (cur_value == value) + cur_value_match_level = true; + } + + if (!cur_value_match_level) + size += sprintf(buf + size, " %uMhz *\n", cur_value); + break; + default: + break; + } + + return size; +} + +static int vangogh_get_profiling_clk_mask(struct smu_context *smu, + enum amd_dpm_forced_level level, + uint32_t *vclk_mask, + uint32_t *dclk_mask, + uint32_t *mclk_mask, + uint32_t *fclk_mask, + uint32_t *soc_mask) +{ + DpmClocks_t *clk_table = smu->smu_table.clocks_table; + + if (level == AMD_DPM_FORCED_LEVEL_PROFILE_MIN_MCLK) { + if (mclk_mask) + *mclk_mask = clk_table->NumDfPstatesEnabled - 1; + + if (fclk_mask) + *fclk_mask = clk_table->NumDfPstatesEnabled - 1; + + if (soc_mask) + *soc_mask = 0; + } else if (level == AMD_DPM_FORCED_LEVEL_PROFILE_PEAK) { + if (mclk_mask) + *mclk_mask = 0; + + if (fclk_mask) + *fclk_mask = 0; + + if (soc_mask) + *soc_mask = 1; + + if (vclk_mask) + *vclk_mask = 1; + + if (dclk_mask) + *dclk_mask = 1; + } else if (level == AMD_DPM_FORCED_LEVEL_PROFILE_STANDARD) { + if (mclk_mask) + *mclk_mask = 0; + + if (fclk_mask) + *fclk_mask = 0; + + if (soc_mask) + *soc_mask = 1; + + if (vclk_mask) + *vclk_mask = 1; + + if (dclk_mask) + *dclk_mask = 1; + } + + return 0; +} + +bool vangogh_clk_dpm_is_enabled(struct smu_context *smu, + enum smu_clk_type clk_type) +{ + enum smu_feature_mask feature_id = 0; + + switch (clk_type) { + case SMU_MCLK: + case SMU_UCLK: + case SMU_FCLK: + feature_id = SMU_FEATURE_DPM_FCLK_BIT; + break; + case SMU_GFXCLK: + case SMU_SCLK: + feature_id = SMU_FEATURE_DPM_GFXCLK_BIT; + break; + case SMU_SOCCLK: + feature_id = SMU_FEATURE_DPM_SOCCLK_BIT; + break; + case SMU_VCLK: + case SMU_DCLK: + feature_id = SMU_FEATURE_VCN_DPM_BIT; + break; + default: + return true; + } + + if (!smu_cmn_feature_is_enabled(smu, feature_id)) + return false; + + return true; +} + +static int vangogh_get_dpm_ultimate_freq(struct smu_context *smu, + enum smu_clk_type clk_type, + uint32_t *min, + uint32_t *max) +{ + int ret = 0; + uint32_t soc_mask; + uint32_t vclk_mask; + uint32_t dclk_mask; + uint32_t mclk_mask; + uint32_t fclk_mask; + uint32_t clock_limit; + + if (!vangogh_clk_dpm_is_enabled(smu, clk_type)) { + switch (clk_type) { + case SMU_MCLK: + case SMU_UCLK: + clock_limit = smu->smu_table.boot_values.uclk; + break; + case SMU_FCLK: + clock_limit = smu->smu_table.boot_values.fclk; + break; + case SMU_GFXCLK: + case SMU_SCLK: + clock_limit = smu->smu_table.boot_values.gfxclk; + break; + case SMU_SOCCLK: + clock_limit = smu->smu_table.boot_values.socclk; + break; + case SMU_VCLK: + clock_limit = smu->smu_table.boot_values.vclk; + break; + case SMU_DCLK: + clock_limit = smu->smu_table.boot_values.dclk; + break; + default: + clock_limit = 0; + break; + } + + /* clock in Mhz unit */ + if (min) + *min = clock_limit / 100; + if (max) + *max = clock_limit / 100; + + return 0; + } + if (max) { + ret = vangogh_get_profiling_clk_mask(smu, + AMD_DPM_FORCED_LEVEL_PROFILE_PEAK, + &vclk_mask, + &dclk_mask, + &mclk_mask, + &fclk_mask, + &soc_mask); + if (ret) + goto failed; + + switch (clk_type) { + case SMU_UCLK: + case SMU_MCLK: + ret = vangogh_get_dpm_clk_limited(smu, clk_type, mclk_mask, max); + if (ret) + goto failed; + break; + case SMU_SOCCLK: + ret = vangogh_get_dpm_clk_limited(smu, clk_type, soc_mask, max); + if (ret) + goto failed; + break; + case SMU_FCLK: + ret = vangogh_get_dpm_clk_limited(smu, clk_type, fclk_mask, max); + if (ret) + goto failed; + break; + case SMU_VCLK: + ret = vangogh_get_dpm_clk_limited(smu, clk_type, vclk_mask, max); + if (ret) + goto failed; + break; + case SMU_DCLK: + ret = vangogh_get_dpm_clk_limited(smu, clk_type, dclk_mask, max); + if (ret) + goto failed; + break; + default: + ret = -EINVAL; + goto failed; + } + } + if (min) { + switch (clk_type) { + case SMU_UCLK: + case SMU_MCLK: + ret = vangogh_get_dpm_clk_limited(smu, clk_type, mclk_mask, min); + if (ret) + goto failed; + break; + case SMU_SOCCLK: + ret = vangogh_get_dpm_clk_limited(smu, clk_type, soc_mask, min); + if (ret) + goto failed; + break; + case SMU_FCLK: + ret = vangogh_get_dpm_clk_limited(smu, clk_type, fclk_mask, min); + if (ret) + goto failed; + break; + case SMU_VCLK: + ret = vangogh_get_dpm_clk_limited(smu, clk_type, vclk_mask, min); + if (ret) + goto failed; + break; + case SMU_DCLK: + ret = vangogh_get_dpm_clk_limited(smu, clk_type, dclk_mask, min); + if (ret) + goto failed; + break; + default: + ret = -EINVAL; + goto failed; + } + } +failed: + return ret; +} + +static int vangogh_get_power_profile_mode(struct smu_context *smu, + char *buf) +{ + static const char *profile_name[] = { + "FULL_SCREEN_3D", + "VIDEO", + "VR", + "COMPUTE", + "CUSTOM"}; + uint32_t i, size = 0; + int16_t workload_type = 0; + + if (!buf) + return -EINVAL; + + for (i = 0; i <= PP_SMC_POWER_PROFILE_CUSTOM; i++) { + /* + * Conv PP_SMC_POWER_PROFILE* to WORKLOAD_PPLIB_*_BIT + * Not all profile modes are supported on vangogh. + */ + workload_type = smu_cmn_to_asic_specific_index(smu, + CMN2ASIC_MAPPING_WORKLOAD, + i); + + if (workload_type < 0) + continue; + + size += sprintf(buf + size, "%2d %14s%s\n", + i, profile_name[i], (i == smu->power_profile_mode) ? "*" : " "); + } + return size; } +static int vangogh_set_power_profile_mode(struct smu_context *smu, long *input, uint32_t size) +{ + int workload_type, ret; + uint32_t profile_mode = input[size]; + + if (profile_mode > PP_SMC_POWER_PROFILE_CUSTOM) { + dev_err(smu->adev->dev, "Invalid power profile mode %d\n", profile_mode); + return -EINVAL; + } + + /* conv PP_SMC_POWER_PROFILE* to WORKLOAD_PPLIB_*_BIT */ + workload_type = smu_cmn_to_asic_specific_index(smu, + CMN2ASIC_MAPPING_WORKLOAD, + profile_mode); + if (workload_type < 0) { + dev_err_once(smu->adev->dev, "Unsupported power profile mode %d on VANGOGH\n", + profile_mode); + return -EINVAL; + } + + ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_ActiveProcessNotify, + 1 << workload_type, + NULL); + if (ret) { + dev_err_once(smu->adev->dev, "Fail to set workload type %d\n", + workload_type); + return ret; + } + + smu->power_profile_mode = profile_mode; + + return 0; +} + +static int vangogh_set_soft_freq_limited_range(struct smu_context *smu, + enum smu_clk_type clk_type, + uint32_t min, + uint32_t max) +{ + int ret = 0; + + if (!vangogh_clk_dpm_is_enabled(smu, clk_type)) + return 0; + + switch (clk_type) { + case SMU_GFXCLK: + case SMU_SCLK: + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_SetHardMinGfxClk, + min, NULL); + if (ret) + return ret; + + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_SetSoftMaxGfxClk, + max, NULL); + if (ret) + return ret; + break; + case SMU_FCLK: + case SMU_MCLK: + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_SetHardMinFclkByFreq, + min, NULL); + if (ret) + return ret; + + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_SetSoftMaxFclkByFreq, + max, NULL); + if (ret) + return ret; + break; + case SMU_SOCCLK: + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_SetHardMinSocclkByFreq, + min, NULL); + if (ret) + return ret; + + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_SetSoftMaxSocclkByFreq, + max, NULL); + if (ret) + return ret; + break; + case SMU_VCLK: + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_SetHardMinVcn, + min << 16, NULL); + if (ret) + return ret; + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_SetSoftMaxVcn, + max << 16, NULL); + if (ret) + return ret; + break; + case SMU_DCLK: + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_SetHardMinVcn, + min, NULL); + if (ret) + return ret; + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_SetSoftMaxVcn, + max, NULL); + if (ret) + return ret; + break; + default: + return -EINVAL; + } + + return ret; +} + +static int vangogh_force_clk_levels(struct smu_context *smu, + enum smu_clk_type clk_type, uint32_t mask) +{ + uint32_t soft_min_level = 0, soft_max_level = 0; + uint32_t min_freq = 0, max_freq = 0; + int ret = 0 ; + + soft_min_level = mask ? (ffs(mask) - 1) : 0; + soft_max_level = mask ? (fls(mask) - 1) : 0; + + switch (clk_type) { + case SMU_SOCCLK: + ret = vangogh_get_dpm_clk_limited(smu, clk_type, + soft_min_level, &min_freq); + if (ret) + return ret; + ret = vangogh_get_dpm_clk_limited(smu, clk_type, + soft_max_level, &max_freq); + if (ret) + return ret; + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_SetSoftMaxSocclkByFreq, + max_freq, NULL); + if (ret) + return ret; + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_SetHardMinSocclkByFreq, + min_freq, NULL); + if (ret) + return ret; + break; + case SMU_MCLK: + case SMU_FCLK: + ret = vangogh_get_dpm_clk_limited(smu, + clk_type, soft_min_level, &min_freq); + if (ret) + return ret; + ret = vangogh_get_dpm_clk_limited(smu, + clk_type, soft_max_level, &max_freq); + if (ret) + return ret; + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_SetSoftMaxFclkByFreq, + max_freq, NULL); + if (ret) + return ret; + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_SetHardMinFclkByFreq, + min_freq, NULL); + if (ret) + return ret; + break; + case SMU_VCLK: + ret = vangogh_get_dpm_clk_limited(smu, + clk_type, soft_min_level, &min_freq); + if (ret) + return ret; + + ret = vangogh_get_dpm_clk_limited(smu, + clk_type, soft_max_level, &max_freq); + if (ret) + return ret; + + + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_SetHardMinVcn, + min_freq << 16, NULL); + if (ret) + return ret; + + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_SetSoftMaxVcn, + max_freq << 16, NULL); + if (ret) + return ret; + + break; + case SMU_DCLK: + ret = vangogh_get_dpm_clk_limited(smu, + clk_type, soft_min_level, &min_freq); + if (ret) + return ret; + + ret = vangogh_get_dpm_clk_limited(smu, + clk_type, soft_max_level, &max_freq); + if (ret) + return ret; + + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_SetHardMinVcn, + min_freq, NULL); + if (ret) + return ret; + + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_SetSoftMaxVcn, + max_freq, NULL); + if (ret) + return ret; + + break; + default: + break; + } + + return ret; +} + +static int vangogh_force_dpm_limit_value(struct smu_context *smu, bool highest) +{ + int ret = 0, i = 0; + uint32_t min_freq, max_freq, force_freq; + enum smu_clk_type clk_type; + + enum smu_clk_type clks[] = { + SMU_SOCCLK, + SMU_VCLK, + SMU_DCLK, + SMU_MCLK, + SMU_FCLK, + }; + + for (i = 0; i < ARRAY_SIZE(clks); i++) { + clk_type = clks[i]; + ret = vangogh_get_dpm_ultimate_freq(smu, clk_type, &min_freq, &max_freq); + if (ret) + return ret; + + force_freq = highest ? max_freq : min_freq; + ret = vangogh_set_soft_freq_limited_range(smu, clk_type, force_freq, force_freq); + if (ret) + return ret; + } + + return ret; +} + +static int vangogh_unforce_dpm_levels(struct smu_context *smu) +{ + int ret = 0, i = 0; + uint32_t min_freq, max_freq; + enum smu_clk_type clk_type; + + struct clk_feature_map { + enum smu_clk_type clk_type; + uint32_t feature; + } clk_feature_map[] = { + {SMU_MCLK, SMU_FEATURE_DPM_FCLK_BIT}, + {SMU_FCLK, SMU_FEATURE_DPM_FCLK_BIT}, + {SMU_SOCCLK, SMU_FEATURE_DPM_SOCCLK_BIT}, + {SMU_VCLK, SMU_FEATURE_VCN_DPM_BIT}, + {SMU_DCLK, SMU_FEATURE_VCN_DPM_BIT}, + }; + + for (i = 0; i < ARRAY_SIZE(clk_feature_map); i++) { + + if (!smu_cmn_feature_is_enabled(smu, clk_feature_map[i].feature)) + continue; + + clk_type = clk_feature_map[i].clk_type; + + ret = vangogh_get_dpm_ultimate_freq(smu, clk_type, &min_freq, &max_freq); + + if (ret) + return ret; + + ret = vangogh_set_soft_freq_limited_range(smu, clk_type, min_freq, max_freq); + + if (ret) + return ret; + } + + return ret; +} + +static int vangogh_set_peak_clock_by_device(struct smu_context *smu) +{ + int ret = 0; + uint32_t socclk_freq = 0, fclk_freq = 0; + uint32_t vclk_freq = 0, dclk_freq = 0; + + ret = vangogh_get_dpm_ultimate_freq(smu, SMU_FCLK, NULL, &fclk_freq); + if (ret) + return ret; + + ret = vangogh_set_soft_freq_limited_range(smu, SMU_FCLK, fclk_freq, fclk_freq); + if (ret) + return ret; + + ret = vangogh_get_dpm_ultimate_freq(smu, SMU_SOCCLK, NULL, &socclk_freq); + if (ret) + return ret; + + ret = vangogh_set_soft_freq_limited_range(smu, SMU_SOCCLK, socclk_freq, socclk_freq); + if (ret) + return ret; + + ret = vangogh_get_dpm_ultimate_freq(smu, SMU_VCLK, NULL, &vclk_freq); + if (ret) + return ret; + + ret = vangogh_set_soft_freq_limited_range(smu, SMU_VCLK, vclk_freq, vclk_freq); + if (ret) + return ret; + + ret = vangogh_get_dpm_ultimate_freq(smu, SMU_DCLK, NULL, &dclk_freq); + if (ret) + return ret; + + ret = vangogh_set_soft_freq_limited_range(smu, SMU_DCLK, dclk_freq, dclk_freq); + if (ret) + return ret; + + return ret; +} + +static int vangogh_set_performance_level(struct smu_context *smu, + enum amd_dpm_forced_level level) +{ + int ret = 0; + uint32_t soc_mask, mclk_mask, fclk_mask; + uint32_t vclk_mask = 0, dclk_mask = 0; + + switch (level) { + case AMD_DPM_FORCED_LEVEL_HIGH: + ret = vangogh_force_dpm_limit_value(smu, true); + break; + case AMD_DPM_FORCED_LEVEL_LOW: + ret = vangogh_force_dpm_limit_value(smu, false); + break; + case AMD_DPM_FORCED_LEVEL_AUTO: + ret = vangogh_unforce_dpm_levels(smu); + break; + case AMD_DPM_FORCED_LEVEL_PROFILE_STANDARD: + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_SetHardMinGfxClk, + VANGOGH_UMD_PSTATE_STANDARD_GFXCLK, NULL); + if (ret) + return ret; + + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_SetSoftMaxGfxClk, + VANGOGH_UMD_PSTATE_STANDARD_GFXCLK, NULL); + if (ret) + return ret; + + ret = vangogh_get_profiling_clk_mask(smu, level, + &vclk_mask, + &dclk_mask, + &mclk_mask, + &fclk_mask, + &soc_mask); + if (ret) + return ret; + + vangogh_force_clk_levels(smu, SMU_MCLK, 1 << mclk_mask); + vangogh_force_clk_levels(smu, SMU_FCLK, 1 << fclk_mask); + vangogh_force_clk_levels(smu, SMU_SOCCLK, 1 << soc_mask); + vangogh_force_clk_levels(smu, SMU_VCLK, 1 << vclk_mask); + vangogh_force_clk_levels(smu, SMU_DCLK, 1 << dclk_mask); + + break; + case AMD_DPM_FORCED_LEVEL_PROFILE_MIN_SCLK: + ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_SetHardMinVcn, + VANGOGH_UMD_PSTATE_PEAK_DCLK, NULL); + if (ret) + return ret; + + ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_SetSoftMaxVcn, + VANGOGH_UMD_PSTATE_PEAK_DCLK, NULL); + if (ret) + return ret; + break; + case AMD_DPM_FORCED_LEVEL_PROFILE_MIN_MCLK: + ret = vangogh_get_profiling_clk_mask(smu, level, + NULL, + NULL, + &mclk_mask, + &fclk_mask, + NULL); + if (ret) + return ret; + + vangogh_force_clk_levels(smu, SMU_MCLK, 1 << mclk_mask); + vangogh_force_clk_levels(smu, SMU_FCLK, 1 << fclk_mask); + break; + case AMD_DPM_FORCED_LEVEL_PROFILE_PEAK: + ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_SetHardMinGfxClk, + VANGOGH_UMD_PSTATE_PEAK_GFXCLK, NULL); + if (ret) + return ret; + + ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_SetSoftMaxGfxClk, + VANGOGH_UMD_PSTATE_PEAK_GFXCLK, NULL); + if (ret) + return ret; + + ret = vangogh_set_peak_clock_by_device(smu); + break; + case AMD_DPM_FORCED_LEVEL_MANUAL: + case AMD_DPM_FORCED_LEVEL_PROFILE_EXIT: + default: + break; + } + return ret; +} + static int vangogh_read_sensor(struct smu_context *smu, enum amd_pp_sensors sensor, void *data, uint32_t *size) @@ -513,7 +1310,7 @@ static int vangogh_set_watermarks_table(struct smu_context *smu, if (clock_ranges) { if (clock_ranges->num_reader_wm_sets > NUM_WM_RANGES || - clock_ranges->num_writer_wm_sets > NUM_WM_RANGES) + clock_ranges->num_writer_wm_sets > NUM_WM_RANGES) return -EINVAL; for (i = 0; i < clock_ranges->num_reader_wm_sets; i++) { @@ -631,14 +1428,16 @@ static int vangogh_od_edit_dpm_table(struct smu_context *smu, enum PP_OD_DPM_TAB if (input[0] == 0) { if (input[1] < smu->gfx_default_hard_min_freq) { - dev_warn(smu->adev->dev, "Fine grain setting minimum sclk (%ld) MHz is less than the minimum allowed (%d) MHz\n", + dev_warn(smu->adev->dev, + "Fine grain setting minimum sclk (%ld) MHz is less than the minimum allowed (%d) MHz\n", input[1], smu->gfx_default_hard_min_freq); return -EINVAL; } smu->gfx_actual_hard_min_freq = input[1]; } else if (input[0] == 1) { if (input[1] > smu->gfx_default_soft_max_freq) { - dev_warn(smu->adev->dev, "Fine grain setting maximum sclk (%ld) MHz is greater than the maximum allowed (%d) MHz\n", + dev_warn(smu->adev->dev, + "Fine grain setting maximum sclk (%ld) MHz is greater than the maximum allowed (%d) MHz\n", input[1], smu->gfx_default_soft_max_freq); return -EINVAL; } @@ -676,8 +1475,10 @@ static int vangogh_od_edit_dpm_table(struct smu_context *smu, enum PP_OD_DPM_TAB return -EINVAL; } else { if (smu->gfx_actual_hard_min_freq > smu->gfx_actual_soft_max_freq) { - dev_err(smu->adev->dev, "The setting minimun sclk (%d) MHz is greater than the setting maximum sclk (%d) MHz\n", - smu->gfx_actual_hard_min_freq, smu->gfx_actual_soft_max_freq); + dev_err(smu->adev->dev, + "The setting minimun sclk (%d) MHz is greater than the setting maximum sclk (%d) MHz\n", + smu->gfx_actual_hard_min_freq, + smu->gfx_actual_soft_max_freq); return -EINVAL; } @@ -722,6 +1523,33 @@ static int vangogh_set_fine_grain_gfx_freq_parameters(struct smu_context *smu) return 0; } +static int vangogh_get_dpm_clock_table(struct smu_context *smu, struct dpm_clocks *clock_table) +{ + DpmClocks_t *table = smu->smu_table.clocks_table; + int i; + + if (!clock_table || !table) + return -EINVAL; + + for (i = 0; i < NUM_SOCCLK_DPM_LEVELS; i++) { + clock_table->SocClocks[i].Freq = table->SocClocks[i]; + clock_table->SocClocks[i].Vol = table->SocVoltage[i]; + } + + for (i = 0; i < NUM_FCLK_DPM_LEVELS; i++) { + clock_table->FClocks[i].Freq = table->DfPstateTable[i].fclk; + clock_table->FClocks[i].Vol = table->DfPstateTable[i].voltage; + } + + for (i = 0; i < NUM_FCLK_DPM_LEVELS; i++) { + clock_table->MemClocks[i].Freq = table->DfPstateTable[i].memclk; + clock_table->MemClocks[i].Vol = table->DfPstateTable[i].voltage; + } + + return 0; +} + + static int vangogh_system_features_control(struct smu_context *smu, bool en) { struct amdgpu_device *adev = smu->adev; @@ -733,6 +1561,38 @@ static int vangogh_system_features_control(struct smu_context *smu, bool en) return 0; } +static int vangogh_post_smu_init(struct smu_context *smu) +{ + struct amdgpu_device *adev = smu->adev; + uint32_t tmp; + uint8_t aon_bits = 0; + /* Two CUs in one WGP */ + uint32_t req_active_wgps = adev->gfx.cu_info.number/2; + uint32_t total_cu = adev->gfx.config.max_cu_per_sh * + adev->gfx.config.max_sh_per_se * adev->gfx.config.max_shader_engines; + + /* if all CUs are active, no need to power off any WGPs */ + if (total_cu == adev->gfx.cu_info.number) + return 0; + + /* + * Calculate the total bits number of always on WGPs for all SA/SEs in + * RLC_PG_ALWAYS_ON_WGP_MASK. + */ + tmp = RREG32_KIQ(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_ALWAYS_ON_WGP_MASK)); + tmp &= RLC_PG_ALWAYS_ON_WGP_MASK__AON_WGP_MASK_MASK; + + aon_bits = hweight32(tmp) * adev->gfx.config.max_sh_per_se * adev->gfx.config.max_shader_engines; + + /* Do not request any WGPs less than set in the AON_WGP_MASK */ + if (aon_bits > req_active_wgps) { + dev_info(adev->dev, "Number of always on WGPs greater than active WGPs: WGP power save not requested.\n"); + return 0; + } else { + return smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_RequestActiveWgp, req_active_wgps, NULL); + } +} + static const struct pptable_funcs vangogh_ppt_funcs = { .check_fw_status = smu_v11_0_check_fw_status, @@ -761,6 +1621,13 @@ static const struct pptable_funcs vangogh_ppt_funcs = { .set_default_dpm_table = vangogh_set_default_dpm_tables, .set_fine_grain_gfx_freq_parameters = vangogh_set_fine_grain_gfx_freq_parameters, .system_features_control = vangogh_system_features_control, + .feature_is_enabled = smu_cmn_feature_is_enabled, + .set_power_profile_mode = vangogh_set_power_profile_mode, + .get_power_profile_mode = vangogh_get_power_profile_mode, + .get_dpm_clock_table = vangogh_get_dpm_clock_table, + .force_clk_levels = vangogh_force_clk_levels, + .set_performance_level = vangogh_set_performance_level, + .post_init = vangogh_post_smu_init, }; void vangogh_set_ppt_funcs(struct smu_context *smu) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu11/vangogh_ppt.h b/drivers/gpu/drm/amd/pm/swsmu/smu11/vangogh_ppt.h index eab455493076..c56d4583dc72 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu11/vangogh_ppt.h +++ b/drivers/gpu/drm/amd/pm/swsmu/smu11/vangogh_ppt.h @@ -28,9 +28,29 @@ extern void vangogh_set_ppt_funcs(struct smu_context *smu); /* UMD PState Vangogh Msg Parameters in MHz */ -#define VANGOGH_UMD_PSTATE_GFXCLK 700 -#define VANGOGH_UMD_PSTATE_SOCCLK 678 -#define VANGOGH_UMD_PSTATE_FCLK 800 +#define VANGOGH_UMD_PSTATE_STANDARD_GFXCLK 1100 +#define VANGOGH_UMD_PSTATE_STANDARD_SOCCLK 600 +#define VANGOGH_UMD_PSTATE_STANDARD_FCLK 800 +#define VANGOGH_UMD_PSTATE_STANDARD_VCLK 705 +#define VANGOGH_UMD_PSTATE_STANDARD_DCLK 600 + +#define VANGOGH_UMD_PSTATE_PEAK_GFXCLK 1300 +#define VANGOGH_UMD_PSTATE_PEAK_SOCCLK 600 +#define VANGOGH_UMD_PSTATE_PEAK_FCLK 800 +#define VANGOGH_UMD_PSTATE_PEAK_VCLK 705 +#define VANGOGH_UMD_PSTATE_PEAK_DCLK 600 + +#define VANGOGH_UMD_PSTATE_MIN_SCLK_GFXCLK 400 +#define VANGOGH_UMD_PSTATE_MIN_SCLK_SOCCLK 1000 +#define VANGOGH_UMD_PSTATE_MIN_SCLK_FCLK 800 +#define VANGOGH_UMD_PSTATE_MIN_SCLK_VCLK 1000 +#define VANGOGH_UMD_PSTATE_MIN_SCLK_DCLK 800 + +#define VANGOGH_UMD_PSTATE_MIN_MCLK_GFXCLK 1100 +#define VANGOGH_UMD_PSTATE_MIN_MCLK_SOCCLK 1000 +#define VANGOGH_UMD_PSTATE_MIN_MCLK_FCLK 400 +#define VANGOGH_UMD_PSTATE_MIN_MCLK_VCLK 1000 +#define VANGOGH_UMD_PSTATE_MIN_MCLK_DCLK 800 /* RLC Power Status */ #define RLC_STATUS_OFF 0 diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu12/renoir_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu12/renoir_ppt.c index dc75db8af371..1f6a774278b1 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu12/renoir_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu12/renoir_ppt.c @@ -188,6 +188,7 @@ static int renoir_get_dpm_clk_limited(struct smu_context *smu, enum smu_clk_type return -EINVAL; *freq = clk_table->SocClocks[dpm_level].Freq; break; + case SMU_UCLK: case SMU_MCLK: if (dpm_level >= NUM_FCLK_DPM_LEVELS) return -EINVAL; @@ -343,6 +344,138 @@ failed: return ret; } +static int renoir_od_edit_dpm_table(struct smu_context *smu, + enum PP_OD_DPM_TABLE_COMMAND type, + long input[], uint32_t size) +{ + int ret = 0; + + if (!smu->fine_grain_enabled) { + dev_warn(smu->adev->dev, "Fine grain is not enabled!\n"); + return -EINVAL; + } + + if (!smu->fine_grain_started) { + dev_warn(smu->adev->dev, "Fine grain is enabled but not started!\n"); + return -EINVAL; + } + + switch (type) { + case PP_OD_EDIT_SCLK_VDDC_TABLE: + if (size != 2) { + dev_err(smu->adev->dev, "Input parameter number not correct\n"); + return -EINVAL; + } + + if (input[0] == 0) { + if (input[1] < smu->gfx_default_hard_min_freq) { + dev_warn(smu->adev->dev, + "Fine grain setting minimum sclk (%ld) MHz is less than the minimum allowed (%d) MHz\n", + input[1], smu->gfx_default_hard_min_freq); + return -EINVAL; + } + smu->gfx_actual_hard_min_freq = input[1]; + } else if (input[0] == 1) { + if (input[1] > smu->gfx_default_soft_max_freq) { + dev_warn(smu->adev->dev, + "Fine grain setting maximum sclk (%ld) MHz is greater than the maximum allowed (%d) MHz\n", + input[1], smu->gfx_default_soft_max_freq); + return -EINVAL; + } + smu->gfx_actual_soft_max_freq = input[1]; + } else { + return -EINVAL; + } + break; + case PP_OD_RESTORE_DEFAULT_TABLE: + if (size != 0) { + dev_err(smu->adev->dev, "Input parameter number not correct\n"); + return -EINVAL; + } + smu->gfx_actual_hard_min_freq = smu->gfx_default_hard_min_freq; + smu->gfx_actual_soft_max_freq = smu->gfx_default_soft_max_freq; + + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_SetHardMinGfxClk, + smu->gfx_actual_hard_min_freq, + NULL); + if (ret) { + dev_err(smu->adev->dev, "Restore the default hard min sclk failed!"); + return ret; + } + + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_SetSoftMaxGfxClk, + smu->gfx_actual_soft_max_freq, + NULL); + if (ret) { + dev_err(smu->adev->dev, "Restore the default soft max sclk failed!"); + return ret; + } + break; + case PP_OD_COMMIT_DPM_TABLE: + if (size != 0) { + dev_err(smu->adev->dev, "Input parameter number not correct\n"); + return -EINVAL; + } else { + if (smu->gfx_actual_hard_min_freq > smu->gfx_actual_soft_max_freq) { + dev_err(smu->adev->dev, + "The setting minimun sclk (%d) MHz is greater than the setting maximum sclk (%d) MHz\n", + smu->gfx_actual_hard_min_freq, + smu->gfx_actual_soft_max_freq); + return -EINVAL; + } + + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_SetHardMinGfxClk, + smu->gfx_actual_hard_min_freq, + NULL); + if (ret) { + dev_err(smu->adev->dev, "Set hard min sclk failed!"); + return ret; + } + + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_SetSoftMaxGfxClk, + smu->gfx_actual_soft_max_freq, + NULL); + if (ret) { + dev_err(smu->adev->dev, "Set soft max sclk failed!"); + return ret; + } + } + break; + default: + return -ENOSYS; + } + + return ret; +} + +static int renoir_set_fine_grain_gfx_freq_parameters(struct smu_context *smu) +{ + uint32_t min = 0, max = 0; + uint32_t ret = 0; + + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_GetMinGfxclkFrequency, + 0, &min); + if (ret) + return ret; + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_GetMaxGfxclkFrequency, + 0, &max); + if (ret) + return ret; + + smu->gfx_default_hard_min_freq = min; + smu->gfx_default_soft_max_freq = max; + smu->gfx_actual_hard_min_freq = 0; + smu->gfx_actual_soft_max_freq = 0; + + return 0; +} + static int renoir_print_clk_levels(struct smu_context *smu, enum smu_clk_type clk_type, char *buf) { @@ -358,6 +491,30 @@ static int renoir_print_clk_levels(struct smu_context *smu, return ret; switch (clk_type) { + case SMU_OD_RANGE: + if (smu->fine_grain_enabled) { + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_GetMinGfxclkFrequency, + 0, &min); + if (ret) + return ret; + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_GetMaxGfxclkFrequency, + 0, &max); + if (ret) + return ret; + size += sprintf(buf + size, "OD_RANGE\nSCLK: %10uMhz %10uMhz\n", min, max); + } + break; + case SMU_OD_SCLK: + if (smu->fine_grain_enabled) { + min = (smu->gfx_actual_hard_min_freq > 0) ? smu->gfx_actual_hard_min_freq : smu->gfx_default_hard_min_freq; + max = (smu->gfx_actual_soft_max_freq > 0) ? smu->gfx_actual_soft_max_freq : smu->gfx_default_soft_max_freq; + size += sprintf(buf + size, "OD_SCLK\n"); + size += sprintf(buf + size, "0:%10uMhz\n", min); + size += sprintf(buf + size, "1:%10uMhz\n", max); + } + break; case SMU_GFXCLK: case SMU_SCLK: /* retirve table returned paramters unit is MHz */ @@ -398,23 +555,35 @@ static int renoir_print_clk_levels(struct smu_context *smu, cur_value = metrics.ClockFrequency[CLOCK_FCLK]; break; default: - return -EINVAL; + break; } - for (i = 0; i < count; i++) { - ret = renoir_get_dpm_clk_limited(smu, clk_type, i, &value); - if (ret) - return ret; - if (!value) - continue; - size += sprintf(buf + size, "%d: %uMhz %s\n", i, value, - cur_value == value ? "*" : ""); - if (cur_value == value) - cur_value_match_level = true; - } + switch (clk_type) { + case SMU_GFXCLK: + case SMU_SCLK: + case SMU_SOCCLK: + case SMU_MCLK: + case SMU_DCEFCLK: + case SMU_FCLK: + for (i = 0; i < count; i++) { + ret = renoir_get_dpm_clk_limited(smu, clk_type, i, &value); + if (ret) + return ret; + if (!value) + continue; + size += sprintf(buf + size, "%d: %uMhz %s\n", i, value, + cur_value == value ? "*" : ""); + if (cur_value == value) + cur_value_match_level = true; + } - if (!cur_value_match_level) - size += sprintf(buf + size, " %uMhz *\n", cur_value); + if (!cur_value_match_level) + size += sprintf(buf + size, " %uMhz *\n", cur_value); + + break; + default: + break; + } return size; } @@ -724,15 +893,31 @@ static int renoir_set_performance_level(struct smu_context *smu, switch (level) { case AMD_DPM_FORCED_LEVEL_HIGH: + smu->fine_grain_started = 0; + smu->gfx_actual_hard_min_freq = smu->gfx_default_hard_min_freq; + smu->gfx_actual_soft_max_freq = smu->gfx_default_soft_max_freq; + ret = renoir_force_dpm_limit_value(smu, true); break; case AMD_DPM_FORCED_LEVEL_LOW: + smu->fine_grain_started = 0; + smu->gfx_actual_hard_min_freq = smu->gfx_default_hard_min_freq; + smu->gfx_actual_soft_max_freq = smu->gfx_default_soft_max_freq; + ret = renoir_force_dpm_limit_value(smu, false); break; case AMD_DPM_FORCED_LEVEL_AUTO: + smu->fine_grain_started = 0; + smu->gfx_actual_hard_min_freq = smu->gfx_default_hard_min_freq; + smu->gfx_actual_soft_max_freq = smu->gfx_default_soft_max_freq; + ret = renoir_unforce_dpm_levels(smu); break; case AMD_DPM_FORCED_LEVEL_PROFILE_STANDARD: + smu->fine_grain_started = 0; + smu->gfx_actual_hard_min_freq = smu->gfx_default_hard_min_freq; + smu->gfx_actual_soft_max_freq = smu->gfx_default_soft_max_freq; + ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_SetHardMinGfxClk, RENOIR_UMD_PSTATE_GFXCLK, @@ -785,6 +970,10 @@ static int renoir_set_performance_level(struct smu_context *smu, break; case AMD_DPM_FORCED_LEVEL_PROFILE_MIN_SCLK: case AMD_DPM_FORCED_LEVEL_PROFILE_MIN_MCLK: + smu->fine_grain_started = 0; + smu->gfx_actual_hard_min_freq = smu->gfx_default_hard_min_freq; + smu->gfx_actual_soft_max_freq = smu->gfx_default_soft_max_freq; + ret = renoir_get_profiling_clk_mask(smu, level, &sclk_mask, &mclk_mask, @@ -796,9 +985,14 @@ static int renoir_set_performance_level(struct smu_context *smu, renoir_force_clk_levels(smu, SMU_SOCCLK, 1 << soc_mask); break; case AMD_DPM_FORCED_LEVEL_PROFILE_PEAK: + smu->fine_grain_started = 0; + smu->gfx_actual_hard_min_freq = smu->gfx_default_hard_min_freq; + smu->gfx_actual_soft_max_freq = smu->gfx_default_soft_max_freq; + ret = renoir_set_peak_clock_by_device(smu); break; case AMD_DPM_FORCED_LEVEL_MANUAL: + smu->fine_grain_started = 1; case AMD_DPM_FORCED_LEVEL_PROFILE_EXIT: default: break; @@ -1159,6 +1353,8 @@ static const struct pptable_funcs renoir_ppt_funcs = { .set_pp_feature_mask = smu_cmn_set_pp_feature_mask, .get_gpu_metrics = renoir_get_gpu_metrics, .gfx_state_change_set = renoir_gfx_state_change_set, + .set_fine_grain_gfx_freq_parameters = renoir_set_fine_grain_gfx_freq_parameters, + .od_edit_dpm_table = renoir_od_edit_dpm_table, }; void renoir_set_ppt_funcs(struct smu_context *smu) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu12/smu_v12_0.c b/drivers/gpu/drm/amd/pm/swsmu/smu12/smu_v12_0.c index 522d55004655..06abf2a7ce9e 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu12/smu_v12_0.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu12/smu_v12_0.c @@ -225,6 +225,7 @@ int smu_v12_0_set_soft_freq_limited_range(struct smu_context *smu, enum smu_clk_ break; case SMU_FCLK: case SMU_MCLK: + case SMU_UCLK: ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_SetHardMinFclkByFreq, min, NULL); if (ret) return ret; |